Ecological Restoration and its Effects on a Regional Climate: The Source Region of the Yellow River,...

8
Ecological Restoration and Its Eects on a Regional Climate: The Source Region of the Yellow River, China Zhouyuan Li, Xuehua Liu,* ,Tianlin Niu, De Kejia, Qingping Zhou, §,Tianxiao Ma, and Yunyang Gao State Key Joint Laboratory of Environment Simulation and Pollution Control, and School of Environment, Tsinghua University, Beijing 100084, Peoples Republic of China Qinghai Academy of Animal Science and Veterinary Medicine, Qinghai University, Xining, Qinghai 810016, Peoples Republic of China § Southwest University for Nationalities, Chengdu, Sichuan 610041, Peoples Republic of China College of Forestry, Beijing Forestry University, Beijing 100083, Peoples Republic of China * S Supporting Information ABSTRACT: The source region of the Yellow River, China, experienced degradation during the 1980s and 1990s, but eective ecological restoration projects have restored the alpine grassland ecosystem. The local government has taken action to restore the grassland area since 1996. Remote sensing monitoring results show an initial restoration of this alpine grassland ecosystem with the structural transformation of land cover from 2000 to 2009 as low- and high-coverage grassland recovered. From 2000 to 2009, the low-coverage grassland area expanded by over 25% and the bare soil area decreased by approximately 15%. To examine the relationship between ecological structure and function, surface temperature (T s ) and evapotranspi- ration (ET) levels were estimated to study the dynamics of the hydro-heat pattern. The results show a turning point in approximately the year 2000 from a declining ET to a rising ET, eventually reaching the 1990 level of approximately 1.5 cm/day. We conclude that grassland coverage expansion has improved the regional hydrologic cycle as a consequence of ecological restoration. Thus, we suggest that long-term restoration and monitoring eorts would help maintain the climatic adjustment functions of this alpine grassland ecosystem. INTRODUCTION Because of climate change and human interference, many ecosystems in the world are experiencing an accelerated degradation, 13 especially for sensitive alpine and mountain grassland ecosystems. 46 China has a large population and rapid economic growth and is now realizing the urgent obligation to save degrading land, specically in the western area. The alpine grasslandwetland ecosystem in the source region of the Yellow River is located in the northeastern boundary of the Tibetan Plateau, which has vast and diverse grasslands and many lakes. It provides irreplaceable ecological services for both the local and downstream areas. 7 Natural and human factors played dierent roles during the stages when the source region experienced ecological degradation and restoration. Between the 1980s and 2000s, the grassland in the source region of the Yellow River experienced two stages, degradation and restoration, which could be attributed to both natural and anthropogenic factors (Table 1). Before and during the 1980s, water and grassland resources in the source region were abundant enough to support the local husbandry. 8 On the other hand, overgrazing and extensive pasturing gradually caused the decline in the quality of the grassland. In addition, a continuous drought made the grasslands unlikely to recover by natural means, with some zones beginning the process of desertication. 9,10 Meanwhile, glacier thawing and landslides, together with the eects of wind and river ows, caused surface soil erosion, leading to a severe decrease in grass coverage and landscape fragmentation. 11 Since the 1990s, human interference from transportation and mining in the region has increased under the national plan of the Western Development. 8 Human activities have disturbed the living conditions of wildlife and exiled some large carnivores. 12 The local original food chain almost collapsed. 13 Rodents were the dominant consumers of the vegetation and exponentially grew in the 1990s, severely threatening the grassland resources. 1416 In 1996, the central and local government of China decided to implement a series of ecological conservation and protection projects to restore the grassland ecosystem in the source region. 17 These projects Received: April 11, 2014 Revised: April 19, 2015 Accepted: April 20, 2015 Article pubs.acs.org/est © XXXX American Chemical Society A DOI: 10.1021/es505985q Environ. Sci. Technol. XXXX, XXX, XXXXXX

Transcript of Ecological Restoration and its Effects on a Regional Climate: The Source Region of the Yellow River,...

Ecological Restoration and Its Effects on a Regional Climate TheSource Region of the Yellow River ChinaZhouyuan Lidagger Xuehua Liudagger Tianlin Niudagger De KejiaDagger Qingping ZhousectDagger Tianxiao Ma∥

and Yunyang Gaodagger

daggerState Key Joint Laboratory of Environment Simulation and Pollution Control and School of Environment Tsinghua UniversityBeijing 100084 Peoplersquos Republic of ChinaDaggerQinghai Academy of Animal Science and Veterinary Medicine Qinghai University Xining Qinghai 810016 Peoplersquos Republic ofChinasectSouthwest University for Nationalities Chengdu Sichuan 610041 Peoplersquos Republic of China∥College of Forestry Beijing Forestry University Beijing 100083 Peoplersquos Republic of China

S Supporting Information

ABSTRACT The source region of the Yellow River China experienceddegradation during the 1980s and 1990s but effective ecological restorationprojects have restored the alpine grassland ecosystem The local governmenthas taken action to restore the grassland area since 1996 Remote sensingmonitoring results show an initial restoration of this alpine grasslandecosystem with the structural transformation of land cover from 2000 to2009 as low- and high-coverage grassland recovered From 2000 to 2009 thelow-coverage grassland area expanded by over 25 and the bare soil areadecreased by approximately 15 To examine the relationship betweenecological structure and function surface temperature (Ts) and evapotranspi-ration (ET) levels were estimated to study the dynamics of the hydro-heatpattern The results show a turning point in approximately the year 2000 froma declining ET to a rising ET eventually reaching the 1990 level ofapproximately 15 cmday We conclude that grassland coverage expansion has improved the regional hydrologic cycle as aconsequence of ecological restoration Thus we suggest that long-term restoration and monitoring efforts would help maintainthe climatic adjustment functions of this alpine grassland ecosystem

INTRODUCTION

Because of climate change and human interference manyecosystems in the world are experiencing an accelerateddegradation1minus3 especially for sensitive alpine and mountaingrassland ecosystems4minus6 China has a large population andrapid economic growth and is now realizing the urgentobligation to save degrading land specifically in the westernarea The alpine grasslandminuswetland ecosystem in the sourceregion of the Yellow River is located in the northeasternboundary of the Tibetan Plateau which has vast and diversegrasslands and many lakes It provides irreplaceable ecologicalservices for both the local and downstream areas7

Natural and human factors played different roles during thestages when the source region experienced ecologicaldegradation and restoration Between the 1980s and 2000sthe grassland in the source region of the Yellow Riverexperienced two stages degradation and restoration whichcould be attributed to both natural and anthropogenic factors(Table 1) Before and during the 1980s water and grasslandresources in the source region were abundant enough tosupport the local husbandry8 On the other hand overgrazingand extensive pasturing gradually caused the decline in the

quality of the grassland In addition a continuous droughtmade the grasslands unlikely to recover by natural means withsome zones beginning the process of desertification910

Meanwhile glacier thawing and landslides together with theeffects of wind and river flows caused surface soil erosionleading to a severe decrease in grass coverage and landscapefragmentation11 Since the 1990s human interference fromtransportation and mining in the region has increased under thenational plan of the Western Development8 Human activitieshave disturbed the living conditions of wildlife and exiled somelarge carnivores12 The local original food chain almostcollapsed13 Rodents were the dominant consumers of thevegetation and exponentially grew in the 1990s severelythreatening the grassland resources14minus16 In 1996 the centraland local government of China decided to implement a seriesof ecological conservation and protection projects to restore thegrassland ecosystem in the source region17 These projects

Received April 11 2014Revised April 19 2015Accepted April 20 2015

Article

pubsacsorgest

copy XXXX American Chemical Society A DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

included converting grazing land to grassland enclosing desertto nurture grass ecological migration and building habitats forthe natural predators of rodentsThe objective of continuous monitoring is essential to learn

how the ecosystem degraded and assess how it recoveredRemote sensing provides an effective and high-efficiencysolution to investigate land usecover changes (LUCC) inthe ecosystem18 In fact knowledge of LUCC of the sourceregion has been accumulated1011 Most studies using remotesensing showed only the direct results of the grassland changesbut a few identified the indirect effects of land surfacesuccession with respect to the ecosystem vegetation and theregional climate as a wholeAs a result of the impact from climate change the hydrologic

cycle requires particular attention as the fundamental supportof local primary productivity19 Evapotranspiration (ET) linksthe groundwater to atmospheric water and has been assumed tobe one of the most important factors in the hydrologic cycle20

Many previous studies about ET in the source region of theYellow River showed climate change either through meteoro-logical observation or soil humidity measurements21minus23 butfew studies have estimated ET at a large scale in a continuousspace to assess the behavior of ET and its relationship to theland surface structure Remote sensing retrieval helps estimateET at a large scale and has a much lower cost and higherefficiency than conventional ground observation methodsFrom calculation of the energy balance of the land surfaceand obtaining heat flux patterns using satellite and meteoro-logical data ET can be then estimated at a level of each pixel ofthe regional map24

Numerous studies on land ecosystem and regional climatefeedback mechanisms have been conducted worldwide Fromthe early 1980s to the early 1990s with remote sensingapproaches not yet widely applied for LUCC and meteoro-logical observation researchers employed experimental obser-vation data and numerical simulations to visualize regionalhydrologic cycles and their dynamic patterns to conduct andverify feedback theories25minus27 These studies which weretypically conducted in the field generated fairly limited resultswith respect to spatial resolution and mainly forest ecosystemswere examined From the mid-1990s to the present remotesensing methods have become more common as a reliable andefficient way to generate LUCC and hydro-heat data in aspatially explicit manner and feedback theory verification andrefinement methods have become more efficient andcomprehensive through the use of more specified statisticalanalysis tools28minus30 Human interference with LUCC andconsequent effects on regional climates have become a keyconcern31 The biological and physical processes affecting landsurfaces have been more widely discussed to elucidate the

underlying mechanisms of regional climatic systems based onfield observations and spatial retrieval methods for variousecosystems32minus34

Because we now understand how land cover and ET patternshave changed at the ground level in recent years interactionsbetween LUCC and hydrologic cycles in the source region ofthe Yellow River can be more thoroughly examined Weexamined LUCC trends in the source region of the YellowRiver for 1990minus2009 Surface temperature (Ts) and ET levelswere retrieved using remote sensing data and land surfaceenergy balance models which were used to denote surfaceenergy and hydrological conditions With the LUCC Ts andET results we describe the effects of ecological restoration onthe regional climate of the study area Our study contributesfindings on alpine grasslandminuswetland ecosystems that furthersubstantiate theories on the interactions between terrestrialecosystems and regional climatic factors in response toecological restoration these findings will be of use forcomparisons to projects with similar spatiotemporal dimensionsand methodologies

MATERIALS AND METHODS

Study Area The study area was determined to be the corecomponent in the source region of the Yellow River shown inthe Landsat-5 Thematic Mapper (TM) image The area liesbetween 33deg 45prime and 35deg 26prime N and between 96deg 54prime and 99deg12prime E it covers a total area of approximately 30 625 km2

(Figure 1) The altitude of the region is higher than 3800 m onaverage The area slopes downward from southwest tonortheast ranging from a combined landform of low mountainsto smooth plateaus

Identification of Land Cover Changes The remotesensing data consisted of five scenes of Landsat-5 TM images(30 times 30 m resolution) of the study area acquired on thefollowing dates August 30 1990 July 24 1994 August 9 2000July 25 2006 and July 17 2009 We obtained the data from theGlobal Land Cover Facility (GLCF) (httpglcfumdedu)The land covers of the images from 1990 to 2009 wereclassified into five types by applying the method of maximumlikelihood in ERDAS IMAGINE 2011 To set the interpretationcriterion for classification we defined land with no vegetationconstruction areas and sands as bare soil grass coverage lessthan 30 as low-coverage grassland grass coverage between 30and 70 as middle-coverage grassland grass coverage morethan 70 as high-coverage grassland and lakes and rivers aswater bodies The 230 Global Positioning System (GPS)-sampled points were taken during the fieldwork in 2011 and2012 in total including 83 for training and 147 for validatingrespectively in the classification criterions that were markedindicatively on the map of the study area (Figure 1) (see Tables

Table 1 Natural and Human Factors Influencing the Source Region of the Yellow River during the Period from the 1980s to2010s

natural factors anthropogenic factors

periods stages positive negative positive negative

1980s andbefore

origin abundant resources ofgrassland and wetland

less interference

1990s degrading draft global warming wind and watererosion effects rodents increasing

overgrazingtransportationdevelopment

2000s tocurrent

restoring grazing prohibition projects ofecological conservation andrestoration

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

B

S-1 and S-2 of the Supporting Information) A confusion matrixand Kappa statistics were used to make an accuracy assessmentwhich showed that the final result of the classification wasacceptable (see Table S-3 of the Supporting Information)Retrieval of Vegetation Conditions Ts and ET

Patterns Satellite-derived vegetation indices such as thenormalized difference vegetation index (NDVI) have beenclosely associated with primary vegetation production34 NDVIis defined as follows

= minus +r r r rNDVI ( )( )NIR red NIR red

where rNIR and rred represent surface reflectance levels averagedover wavelength ranges of infrared and visible infrared regionsof the spectrum respectively NDVI values and trends wereused to denote vegetation degradation or restoration trendsbased on LUCC dataRegional climate indicators regarding surface energy and

water budget levels were examined using the same set of

Landsat-5 TM images Ts was calculated as an indicator ofsurface energy conditions based on the following formula3536

ε= +T K K Lln( 1)s 2 1 6 (1)

where K1 = 60776 times 106 W cmminus2 srminus1 μmminus1 K2 = 126056 times106 W cmminus2 srminus1 μmminus1 with K1 and K2 being radiationconstants for Landsat-5 images L6 is the spectral radiance ofband 6 in Landsat-5 images and ε is the atmospheric emissivitylevel determined on the basis of the NDVIWe measured the value adding product (VAP) of plain areas

in the atmospheric correction (ATCOR2) module platform ofthe ERDAS 2011 remote sensing processing software toretrieve ET (cmday) data for the region Principles of theproduct are based on the following surface energy balanceequations37

= + +R H G LEn (2)

=ET LE286 (3)

where the terms denote composites of net radiation (RnW mminus2) sensible heat flux (H W mminus2) ground heat flux (GW mminus2) and latent heat flux (LE W mminus2) The modelingmethod involved two main tasks (1) calculating the radiationbalance based on remote sensing pixel reflectance levels and (2)calculating the heat balance using field knowledge that includessurface vegetation and meteorological conditions (see Table S-4 and Figure S-1 of the Supporting Information)

Comparisons to Meteorological and HydrologicalRecords To compare the estimated results with meteoro-logical and hydrological observations air temperature (degC)water balance (mm) and runoff (mm) variations during thesummers (JuneminusSeptember) of 1990minus2009 for the study areawere also recorded The water balance was defined as thedifference between precipitation (mm) and evaporation (mm)levels Because the moving average is typically used with timeseries data to smooth out short-term fluctuations and reflectlonger term trends we calculated the moving average tovisualize observation trends by creating a series of water balanceand runoff averages for the full data set The originalmeteorological data were provided by the China MeteorologicalData Sharing Service System (httpcdccmagovcn) asobserved at the meteorological station of Madoi County andrunoff data were provided by the hydrological station ofHuangheyan in Qinghai shown in Figure 1 To validate Ts andET values estimated via remote sensing and values estimatedbased on observations we compared interpolated data of

Figure 1 Study area source region of the Yellow River China (yellowdots denote ground sample points for satellite image interpretation)

Figure 2 Land cover changes of the study area from 1990 to 2009

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

C

observed spatial air temperature and ET with the estimatedvalues The process and results are described in Figure S-2 ofthe Supporting InformationStatistical Analysis Among the key variables we focused

on albedo (α) and the NDVI α reflects the roughness of theland surface and is related to vegetation conditions33 TheNDVI is used to quantify ground vegetation α NDVI Ts andET values retrieved from the raster-based results were extractedfrom pixels of different land cover types and then used to

perform the correlation analysis in SPSS 130 (SPSS Inc

Chicago IL) Pairs of variables (eg NDVIminusα αminusTs TsminusETand ETminusNDVI) were selected on the basis of causeminuseffectvegetation and regional climate processes3839 which were

closely associated with feedback loop pathways In addition α

NDVI Ts and ET trends from 1990 to 2009 were observed

according to different land cover patterns The correlations

among α NDVI Ts and ET and corresponding trends are

Figure 3 (a) Ts changes and annual air temperature changes shown by the curve and linear regression (b) ET changes and annual water balance andrunoff variations in the summers of 1990minus2009 for the study area shown as three-period moving average (Mavg) curves

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

D

anticipated to reveal interactions between surface structures andregional hydro-heat processes

RESULTS

Grassland Degradation and Recovery Regional landcover changes showed patterns of grassland degradation andrecovery in the summers of 1990minus2009 (Figure 2) Theproportion of high-coverage grassland expanded gradually fromapproximately 27 in 1990 to greater than 30 in 2009 Low-coverage grassland areas increased rapidly after 2000accounting for the largest proportion of land (39) Inaddition proportions of the other two land cover types middle-coverage grassland and bare soils decreased from 21 to 12and from 26 to 11 respectively Water body coverage levelsremained constant at a level of 8 for the entire period Thesetrends denote land cover changes after approximately 2000Hydro-heat Patterns and Regional Climate Change Ts

presents a spatial pattern with higher values of approximately2975 K in the north and lower values of approximately 2932 Kin the south (Figure 3a) For different land covers Ts values aresorted in ascending order (high- middle- and low-coveragegrasslands and bare soils) In the study area average summer airtemperatures increased by almost 2 degC over the past 20 years(Figure 3a) In general Ts values of the study area for thesummers of 1990minus2009 follow observed regional warmingtrends We also found that the Ts values of bare soil and low-and middle-coverage grassland areas appeared to decrease from2006 to 2009 (Figure 4)ET levels were lower at approximately 085 cmday in the

north and higher at approximately 175 cmday in the south(Figure 3b) ET values were highest in the water bodies andareas surrounding wetlands also exhibited higher ET values ofapproximately 155 cmday The ET values for the other typesof land cover were ranked in the following ascending order

bare soils and low- middle- and high-coverage grasslandsSandy riverbank areas generated very low ET values ofapproximately 037 cmday The ET value for all of the landcover types initially declined from 1990 to 2000 and thenincreased to 1990 levels In grasslands and bare soils ET levelsreached approximately 15 cmday in 1990 and decreased toapproximately 10 cmday in approximately 2000 denoting anET turning point Thereafter grassland and bare soil ET levelsreturned to approximately 15 cmday in 2006 and 2009(Figures 3b and 4) Meteorological records for the summers of1990minus2009 show that the water balance level reached its lowestvalue in 1996 and the region has since returned to highermoisture levels

Correlation between Key Variables and α and NDVITrends The correlation analysis results show how associationsbetween the key variables form the regional climate feedbackloop (Figure 4) Pearson linear coefficients are underlined andadded to each pair In 1990 1994 and 2000 all pairs weresignificantly correlated with the exception of the αminusTs pairThe scatter graphs show that NDVI was negatively related to αTs was negatively related to ET and ET was positively relatedto NDVI although the relationship between α and Ts remainsunclear In 1990 the αminusTs pair exhibited an insignificantlynegative relation whereas the 2006 and 2009 results generallyshowed a positive numerical relationship between the twovariables Apart from the hydro-heat pattern changes of Ts andET noted above α and NDVI trends were extracted on thebasis of different land cover types revealing that α reached amaximum value in 1994 and that NDVI initially declined butincreased after 1994 denoting the start of the restoration phase

DISCUSSION

Consequences of Ecosystem Degradation As an inlandplateau grasslandminuswetland ecosystem changes in the land

Figure 4 Correlation analysis pairs of NDVIminusα αminusTs TsminusET and ETminusNDVI and change trends of the key variables for 1990minus2009 Numbersmarked in each box denote the Pearson correlation coefficients for each pair of variables lowast and lowastlowast denote the significance levels of 005 and 001respectively

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

E

cover structure in this study area are related to the temperatureand hydrologic patterns of this region From 1990 to 2000 asgrassland areas degraded while bare soils expanded albedolevels of the land surface reached peak level in 1994 At thesame time surface temperatures increased continuously partlybecause of global warming but also because of a reduction inevaporative cooling40 On the basis of Ts spatial patterns thesurface heating effects on bare soils of the northern area becamemore pronounced reflecting patterns similar to those of theurban heat island effect40 This regional surface temperaturechange is attributable to several factors such as elevationradiation surface reflectivity precipitation and evaporation4041

However quantitative attributes of these factors were notexamined in this study The correlation analysis of the keyvariables showed an association between vegetation and hydro-heat environments of the study area based on the retrievedspatial data although these data are too limited to fully reflectcausality in the ecological degradation and restoration feedbackloop Temporal dimension variable change trends show that theα and NDVI turning point occurred in 1994 and that ET hasincreased since 2000 and this could be interpreted as a delayedresponse to improved vegetation conditionsEcosystem Restoration Effects According to references

and local policy documents humans began to play a positiverole in the regional feedback process and attempted to controldegradation trends in the study area beginning in 19967842

The government planned ecological conservation and restora-tion projects and has implemented these projects from 1996 tothe present These projects have aimed to improve primaryproductivity levels and ecosystem self-control capacities tomaintain healthy grasslands in the source region Theseimproved vegetation conditions may stabilize surface temper-atures and increase soil moisture and ET levels as shown inrelevant observations Runoff levels in the area have increasedsince 2000 and this trend has been significantly correlated withprecipitation patterns43minus45 In the present study comparisonsbetween annual water balance and runoff variations show adecline in ET until 2000 (Figure 3b) suggesting that runoff andET patterns interacted closely in the hydrologic cycleRegional Climate Feedback Mechanisms of the

Terrestrial Ecosystem Previous studies on interactionsbetween the land surface and hydroclimatological conditionshave mainly employed methods and tools such as satellitecloud images cloud frequency analyses canopy microclimatol-ogy sensors and meteorological observations to examine anddefine cloud formation feedback effects46minus48 Most of thesestudies examined forest ecosystems with less of a focus onalpine grassland ecosystems Still micro-to-macro processanalyses presented in previous studies suggest the presence ofregional climate feedback mechanisms ET improvements viavegetation recovery helped increase humidity levels in turnincreasing boundary convection and local precipitationlevels4748

In the neighboring QinghaiminusTibet Plateau source regionwhich is similar to the study area recent observations andstudies have shown a positive correlation between ET andhumidity as well as a general increase in precipitation andrunoff levels over the last 5minus10 years4445 Although causalinferences between restored vegetation levels and improvedhydrological conditions in the study area remain unclear basedon existing results we expect to reveal such mechanisms ingreater depth with the use of more observations andexperiments on alpine grassland ecosystems We conclude

that lower ET trends led to lower air moisture levels inhibitingcloud formation and local precipitation It is assumed thatdrought conditions inhibit vegetation growth potentiallyaccelerating grassland degradation further as shown in Figure5

Coupling System of the Atmospheric and TerrestrialHydrocycle According to principles of modern hydro-climatology the atmosphere serves as a climatic frame anddrives energy flows and water vapor transportation patternsSubsystems of the atmosphere and terrestrial ecosystems werecoupled to form a complete hydrocycle system Dividing thehydrocycle system into two branches [the atmospherichydrocycle (AHC) and terrestrial hydrocycle (THC)] helpedreveal global climate patterns50

The dominant factors of AHC mainly include atmosphericcirculation and energy flux trends in patterns of circulationAtmospheric circulation models primarily determine biomes atthe continental scale and strongly affect the construction andsuccession of ecosystems influencing global land coverstructures THC involves terrestrialminusatmosphere interfaceprocesses wherein the atmosphere interacts with geographicalhydrological and biological processes driven by key factors tosupport the development of ecosystems as shown in our studyKey factors include surface energy flux water availability andgroundwater balance ET plays a dominant role in THCbecause ET is driven by radiation and relies on thetransformation of groundwater into atmospheric water vaporwhich is widely related to soil and vegetation processes As partof the land cover structure vegetation determines the speedand means through which groundwater and precipitation movefrom the land surface to the atmosphere In turn studies haveshown that THC plays a key role in climate systems through itsinvolvement in the hydrologic cycleTo situate this analysis of long-term AHC impacts within

processes of ecological restoration it is also essential tounderstand this phenomenon and to make predictionsObservations show that the Asian monsoon has been lessactive over the last 20 years resulting in the drying andwarming of the QinghaiminusTibet Plateau51 Drought conditionsmay continue over a long period of time Therefore long-termand advanced ecological restoration and monitoring efforts are

Figure 5 Regional climate feedback loops of the degraded andrestored scenarios of typical land use landscapes for the study area(adapted with permission from ref 49 Copyright 1984 Elsevier)

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

F

essential Although previous climate change studies havepresented global climate models and have discussed energyuse strategies that mitigate global warming few works haveexamined regional THC processes to assess or manage regionalclimatic and hydrological changes Few climate policy studieshave determined the net impact of biophysical changesresulting from land use pattern changes33 Our work in thesource region of the Yellow River serves as an initial attempt toexplore the feasibility of applying innovative methodologies toaddress gaps in knowledge and to evaluate ecologicalrestoration strategies in future studies

ASSOCIATED CONTENTS Supporting InformationClassification criterions of the study area and the accuracyassessment for the land cover classification detailed methodsfor the calculation models of ATCOR2 and validation of theestimated results with meteorological data The SupportingInformation is available free of charge on the ACS Publicationswebsite at DOI 101021es505985q

AUTHOR INFORMATIONCorresponding AuthorTelephone +86-10-6279-4119 E-mail xuehua-hjxmailtsinghuaeducnNotesThe authors declare no competing financial interest

ACKNOWLEDGMENTSThis project was financially supported by the fund from theState Key Joint Laboratory of Environment Simulation andPollution Control China (11Y02ESPCT) and the projectnamed The Relationship of Birds Migration Patterns andHabitat Factors in Poyang Lake (2010CB530300-04) Theauthors thank the anonymous reviewers for their critiqueswhich improved earlier versions of this manuscript

REFERENCES(1) Armesto J J Bautista S Del Val E Ferguson B Garciacutea XGaxiola A Godinez-Alvarez H Gann G Lopez-Barrera FManson R Nu nez-Avila M Ortiz-Arrona C Tognetti PWilliams-Linera G Towards an ecological restoration networkReversing land degradation in Latin America Front Ecol Environ2007 5 (4) w1minusw4 DOI 1018901540-9295(2007)5[w1TAERNR]20CO2(2) Baustian M M Georgia M Dreelin E A Esselman PSchultze S R Qian L Awb T G Luo L Rose J BA Onehundred year review of the socioeconomic and ecological systems ofLake St Clair North America J Great Lakes Res 2014 40 15minus26DOI 101016jjglr201311006(3) Grumbine R E Assessing environmental security in ChinaFront Ecol Environ 2014 12 (7) 403minus411 DOI 101890130147(4) Bullon T Environmental assessment and land change analysis insemi-natural land covers applicable to land management Int J ApplEarth Obs Geoinf 2015 34 147minus156 DOI 101016jjag201408006(5) Liu J Linderman M Ouyang Z An L Yang J Zhang HEcological degradation in protected areas The case of Wolong NatureReserve for giant pandas Science 2001 292 (5514) 98minus101DOI 101126science1058104(6) Poyatos R Latron J Llorens P Land use and land coverchange after agricultural abandonment Mt Res Dev 2003 23 (4)362minus368 DOI 1016590276-4741(2003)023[0362LUALCC]20CO2(7) Feng J Wang T Xie C Eco-environmental degradation in thesource region of the Yellow River northeast QinghaiminusXizang Plateau

Environ Monit Assess 2006 122 (1minus3) 125minus143 DOI 101007s10661-005-9169-2(8) Wang G Cheng G Eco-environmental changes and causativeanalysis in the source regions of the Yangtze and Yellow Rivers ChinaEnviron Syst Decis 2000 20 (3) 221minus232 DOI 101023A1006703831018(9) Pan J Liu J Yellow River source area of land use and landscapepattern change and its ecological effects J Arid Land Res Environ2005 19 (4) 69minus74 (in Chinese) DOI 103969jissn1003-7578200504014(10) Xu J Song L Zhao Z Hu Y Liu C Monitoring grasslanddegradation dynamically at Maduo County in source region of YellowRiver in past 15 years based on remote sensing Arid Land Geogr 201235 (4) 615minus622 (in Chinese) DOI 1013826jcnkicn65-1103x201204018(11) Yang J Ding Y Chen R Spatial and temporal of variations ofalpine vegetation cover in the source regions of the Yangtze andYellow Rivers of the Tibetan Plateau from 1982 to 2001 Environ Geol2006 50 (3) 313minus322 DOI 101007s00254-006-0210-8(12) Davidson A D Detling J K Brown J H Ecological roles andconservation challenges of social burrowing herbivorous mammals inthe worldrsquos grasslands Front Ecol Environ 2012 10 (9) 477minus486DOI 101890110054(13) Qu J Li W Yang M Ji W Zhang Y Life history of theplateau pika (Ochotona curzoniae) in alpine meadows of the TibetanPlateau Mamm Biol 2013 78 (1) 68minus72 DOI 101016jmambio201209005(14) Liu J Nie H Studies on the population productivity ecology ofplateau pika III Trend of population dynamics in plateau pika withdensity-independent and density-dependent vital rates Acta TheriolSin 1992 12 (2) 139minus146 (in Chinese)(15) Liu W Zhou L Wang X The study of different grazingintensity on the role of plants and rodent studies Acta Ecol Sin 199919 (3) 88minus94 (in Chinese)(16) Bai W Zhang Y Xie G Shen Z Causes analysis of grasslanddegradation Maduo County in Yellow River source region Acta EcolSin 2002 13 (7) 823minus826 (in Chinese) DOI 1013287j1001-933220020194(17) Fang Y Qin D Ding Y Frozen soil change and adaptation ofanimal husbandry A case of the source regions of Yangtze and YellowRivers Environ Sci Policy 2011 14 (5) 555minus568 DOI 101016jenvsci201103012(18) Verburg P H van de Steeg J Veldkamp A Willemen LFrom land cover change to land function dynamics A major challengeto improve land characterization J Environ Manage 2009 90 (3)1327minus1335 DOI 101016jjenvman200808005(19) Barnett T P Adam J C Lettenmaier A D P Potentialimpacts of a warming climate on water availability in snow-dominatedregions Nature 2005 438 303minus309 DOI 101038nature04141(20) Chattopadhyay N Hulme M Evaporation and potentialevapotranspiration in India under conditions of recent and futureclimate change Agric For Meteorol 1997 87 55minus73 DOI 101016S0168-1923(97)00006-3(21) Liu M Xia Z Han S Wang X Tang Z Relationshipbetween variation of evapotranspiration and ecological deterioration insource region of Yellow River J Hohai Uni 2009 37 (6) 631minus634 (inChinese)(22) Yi X Yin Y Li G Peng J Temperature variation in recent50 years in the three-river headwaters region of Qinghai Province ActaGeol Sin 2011 66 (11) 1451minus1465 (in Chinese)(23) Hou W Li Y Soil surface humidity index and sensitivityanalysis of the climate factors that affect it in the Yellow River sourceregions J Glaciol Geocryol 2010 32 (6) 1226minus1233 (in Chinese)(24) Mauser W Schadlich S Modelling the spatial distribution ofevapotranspiration on different scales using remote sensing data JHydrol 1998 212 250minus267 DOI 101016S0022-1694(98)00228-5(25) Salati E Vose P B Amazon basin A system in equilibriumScience 1984 225 (4658) 129minus138 DOI 1023071693078

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

G

(26) Schlesinger W H Reynolds J F Cunningham G LHuenneke L F Jarrell W M Virginia R A Whitford W GBiological feedbacks in global desertification Science 1990 247 (4946)1043minus1048 DOI 101126science24749461043(27) Shukla J Nobre C Sellers P Amazon deforestation andclimate change Science 1990 247 (4948) 1322minus1325 DOI 101126science24749481322(28) Baron J S Hartman M D Kittel T G F Band L E OjimaD S Lammers R B Effects of land cover water redistribution andtemperature on ecosystem processes in the South Platte basin EcolAppl 1998 8 (4) 1037minus1051 DOI 1018901051-0761(1998)008[1037eolcwr]20co2(29) Boegh E Soegaard H Thomsen A Evaluating evapotranspi-ration rates and surface conditions using Landsat TM to estimateatmospheric resistance and surface resistance Remote Sens Environ2002 79 (2minus3) 329minus343 DOI 101016S0034-4257(01)00283-8(30) Sterling S M Ducharne A Polcher J The impact of globalland-cover change on the terrestrial water cycle Nat Clim Change2013 3 (4) 385minus390 DOI 101038nclimate1690(31) Loarie S R Lobell D B Asner G P Mu Q Field C BDirect impacts on local climate of sugar-cane expansion in Brazil NatClim Change 2011 1 (2) 105minus109 DOI 101038nclimate1067(32) Teuling A J Seneviratne S I Stockli R et al Contrastingresponse of European forest and grassland energy exchange toheatwaves Nat Geosci 2010 3 (10) 722minus727 DOI 101038ngeo950(33) Anderson R G Canadell J G Randerson J T et alBiophysical considerations in forestry for climate protection FrontEcol Environ 2011 9 (3) 174minus182 DOI 101890090179(34) Campo-Bescos M A Munoz-Carpena R Kaplan D ASouthworth J Zhu L Waylen P R Beyond precipitationPhysiographic gradients dictate the relative importance of environ-mental drivers on savanna vegetation PLoS One 2013 8 (8)No e72348 DOI 101371journalpone0072348(35) Schneider K Mauser W Processing and accuracy of Landsatthematic mapper data for lake surface temperature measurement Int JRemote Sens 1996 17 (11) 2027minus2041 DOI 10108001431169608948757(36) Li F Jackson T J Kustas W P Schmugge T J French AN Cosh M H Bindlish R Deriving land surface temperature fromLandsat 5 and 7 during SMEX02SMACEX Remote Sens Environ2004 92 (4) 521minus534 DOI 101016jrse200402018(37) Bastiaanssen W G M Menenti M Feddes R A Holtslag AA M A remote sensing surface energy balance algorithm for land(SEBAL) J Hydrol 1998 212 198minus212 DOI 101016S0022-1694(98)00253-4(38) Bonan G B Forests and climate change Forcings feedbacksand the climate benefits of forests Science 2008 320 (5882) 1444minus1449 DOI 101126science1155121(39) Li Z Liu X Ma T Kejia D Zhou Q Yao B Niu TRetrieval of the surface evapotranspiration patterns in the alpinegrasslandminuswetland ecosystem applying SEBAL model in the sourceregion of the Yellow River China Ecol Modell 2013 270 64minus75DOI 101016jecolmodel201309004(40) Zhao L Lee X Smith R B Oleson K Strong contributionsof local background climate to urban heat islands Nature 2014 511(7508) 216minus219 DOI 101038nature13462(41) van Heerwaarden C C de Arellano J V Gounou AGuichard F Couvreux F Understanding the daily cycle ofevapotranspiration A method to quantify the influence of forcingsand feedbacks J Hydrometeorol 2010 11 (6) 1405minus1422DOI 1011752010jhm12721(42) Chang G G Li F X Li L Changes and Restoration in SourceRegion of Three Rivers of China Metrology Press Beijing China 2010(in Chinese)(43) Tong L Xu X Fu Y Li S Wetland changes and theirresponses to climate change in the three-river headwaters region ofChina since the 1990s Energies 2014 7 (4) 2515minus2534DOI 103390en7042515

(44) Zhang Y Zhao X Zhao S Feng C Correlation betweenevapotranspiration and climate factors in warm steppe in source regionof Yangtze Yellow and Yalu Tsangpo Rivers J Desert Res 2010 30(2) 363minus368 (in Chinese)(45) Jiang Y Li D Variations of Tangnaihai runoff andprecipitation and temperature in the upper reach of the YellowRiverMeteorol Disaster Reduct Res 2011 34 (2) 51minus57 (in Chinese)(46) Molders N Kramm G Influence of wildfire induced land-cover changes on clouds and precipitation in interior AlaskaA casestudy Atmos Res 2007 84 (2) 142minus168 DOI 101016jatmos-res200606004(47) Obregon A Gehrig-Downie C Gradstein S R RollenbeckR Bendix J Canopy level fog occurrence in a tropical lowland forestof French Guiana as a prerequisite for high epiphyte diversity AgricFor Meteorol 2011 151 (3) 290minus300 DOI 101016jagrfor-met201011003(48) Harding R J Blyth E M Tuinenburg O A Wiltshire ALand atmosphere feedbacks and their role in the water resources of theGanges basin Sci Total Environ 2013 468 (S1) S85minusS92DOI 101016jscitotenv201303016(49) Bradshaw A D Ecological principles and land reclamationpractice Landscape Plann 1984 11 (1) 35minus48(50) Shelton M L Hydroclimatology Perspectives and ApplicationsCambridge Press Cambridge UK 2009(51) Liu Y Lu S Li S Gao Y Numerical simulation of impact ofland surface changes on regional climatic environment in sourceregions of Yellow River Plateau Meteorol 2009 28 (2) 327minus334 (inChinese)

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

H

included converting grazing land to grassland enclosing desertto nurture grass ecological migration and building habitats forthe natural predators of rodentsThe objective of continuous monitoring is essential to learn

how the ecosystem degraded and assess how it recoveredRemote sensing provides an effective and high-efficiencysolution to investigate land usecover changes (LUCC) inthe ecosystem18 In fact knowledge of LUCC of the sourceregion has been accumulated1011 Most studies using remotesensing showed only the direct results of the grassland changesbut a few identified the indirect effects of land surfacesuccession with respect to the ecosystem vegetation and theregional climate as a wholeAs a result of the impact from climate change the hydrologic

cycle requires particular attention as the fundamental supportof local primary productivity19 Evapotranspiration (ET) linksthe groundwater to atmospheric water and has been assumed tobe one of the most important factors in the hydrologic cycle20

Many previous studies about ET in the source region of theYellow River showed climate change either through meteoro-logical observation or soil humidity measurements21minus23 butfew studies have estimated ET at a large scale in a continuousspace to assess the behavior of ET and its relationship to theland surface structure Remote sensing retrieval helps estimateET at a large scale and has a much lower cost and higherefficiency than conventional ground observation methodsFrom calculation of the energy balance of the land surfaceand obtaining heat flux patterns using satellite and meteoro-logical data ET can be then estimated at a level of each pixel ofthe regional map24

Numerous studies on land ecosystem and regional climatefeedback mechanisms have been conducted worldwide Fromthe early 1980s to the early 1990s with remote sensingapproaches not yet widely applied for LUCC and meteoro-logical observation researchers employed experimental obser-vation data and numerical simulations to visualize regionalhydrologic cycles and their dynamic patterns to conduct andverify feedback theories25minus27 These studies which weretypically conducted in the field generated fairly limited resultswith respect to spatial resolution and mainly forest ecosystemswere examined From the mid-1990s to the present remotesensing methods have become more common as a reliable andefficient way to generate LUCC and hydro-heat data in aspatially explicit manner and feedback theory verification andrefinement methods have become more efficient andcomprehensive through the use of more specified statisticalanalysis tools28minus30 Human interference with LUCC andconsequent effects on regional climates have become a keyconcern31 The biological and physical processes affecting landsurfaces have been more widely discussed to elucidate the

underlying mechanisms of regional climatic systems based onfield observations and spatial retrieval methods for variousecosystems32minus34

Because we now understand how land cover and ET patternshave changed at the ground level in recent years interactionsbetween LUCC and hydrologic cycles in the source region ofthe Yellow River can be more thoroughly examined Weexamined LUCC trends in the source region of the YellowRiver for 1990minus2009 Surface temperature (Ts) and ET levelswere retrieved using remote sensing data and land surfaceenergy balance models which were used to denote surfaceenergy and hydrological conditions With the LUCC Ts andET results we describe the effects of ecological restoration onthe regional climate of the study area Our study contributesfindings on alpine grasslandminuswetland ecosystems that furthersubstantiate theories on the interactions between terrestrialecosystems and regional climatic factors in response toecological restoration these findings will be of use forcomparisons to projects with similar spatiotemporal dimensionsand methodologies

MATERIALS AND METHODS

Study Area The study area was determined to be the corecomponent in the source region of the Yellow River shown inthe Landsat-5 Thematic Mapper (TM) image The area liesbetween 33deg 45prime and 35deg 26prime N and between 96deg 54prime and 99deg12prime E it covers a total area of approximately 30 625 km2

(Figure 1) The altitude of the region is higher than 3800 m onaverage The area slopes downward from southwest tonortheast ranging from a combined landform of low mountainsto smooth plateaus

Identification of Land Cover Changes The remotesensing data consisted of five scenes of Landsat-5 TM images(30 times 30 m resolution) of the study area acquired on thefollowing dates August 30 1990 July 24 1994 August 9 2000July 25 2006 and July 17 2009 We obtained the data from theGlobal Land Cover Facility (GLCF) (httpglcfumdedu)The land covers of the images from 1990 to 2009 wereclassified into five types by applying the method of maximumlikelihood in ERDAS IMAGINE 2011 To set the interpretationcriterion for classification we defined land with no vegetationconstruction areas and sands as bare soil grass coverage lessthan 30 as low-coverage grassland grass coverage between 30and 70 as middle-coverage grassland grass coverage morethan 70 as high-coverage grassland and lakes and rivers aswater bodies The 230 Global Positioning System (GPS)-sampled points were taken during the fieldwork in 2011 and2012 in total including 83 for training and 147 for validatingrespectively in the classification criterions that were markedindicatively on the map of the study area (Figure 1) (see Tables

Table 1 Natural and Human Factors Influencing the Source Region of the Yellow River during the Period from the 1980s to2010s

natural factors anthropogenic factors

periods stages positive negative positive negative

1980s andbefore

origin abundant resources ofgrassland and wetland

less interference

1990s degrading draft global warming wind and watererosion effects rodents increasing

overgrazingtransportationdevelopment

2000s tocurrent

restoring grazing prohibition projects ofecological conservation andrestoration

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

B

S-1 and S-2 of the Supporting Information) A confusion matrixand Kappa statistics were used to make an accuracy assessmentwhich showed that the final result of the classification wasacceptable (see Table S-3 of the Supporting Information)Retrieval of Vegetation Conditions Ts and ET

Patterns Satellite-derived vegetation indices such as thenormalized difference vegetation index (NDVI) have beenclosely associated with primary vegetation production34 NDVIis defined as follows

= minus +r r r rNDVI ( )( )NIR red NIR red

where rNIR and rred represent surface reflectance levels averagedover wavelength ranges of infrared and visible infrared regionsof the spectrum respectively NDVI values and trends wereused to denote vegetation degradation or restoration trendsbased on LUCC dataRegional climate indicators regarding surface energy and

water budget levels were examined using the same set of

Landsat-5 TM images Ts was calculated as an indicator ofsurface energy conditions based on the following formula3536

ε= +T K K Lln( 1)s 2 1 6 (1)

where K1 = 60776 times 106 W cmminus2 srminus1 μmminus1 K2 = 126056 times106 W cmminus2 srminus1 μmminus1 with K1 and K2 being radiationconstants for Landsat-5 images L6 is the spectral radiance ofband 6 in Landsat-5 images and ε is the atmospheric emissivitylevel determined on the basis of the NDVIWe measured the value adding product (VAP) of plain areas

in the atmospheric correction (ATCOR2) module platform ofthe ERDAS 2011 remote sensing processing software toretrieve ET (cmday) data for the region Principles of theproduct are based on the following surface energy balanceequations37

= + +R H G LEn (2)

=ET LE286 (3)

where the terms denote composites of net radiation (RnW mminus2) sensible heat flux (H W mminus2) ground heat flux (GW mminus2) and latent heat flux (LE W mminus2) The modelingmethod involved two main tasks (1) calculating the radiationbalance based on remote sensing pixel reflectance levels and (2)calculating the heat balance using field knowledge that includessurface vegetation and meteorological conditions (see Table S-4 and Figure S-1 of the Supporting Information)

Comparisons to Meteorological and HydrologicalRecords To compare the estimated results with meteoro-logical and hydrological observations air temperature (degC)water balance (mm) and runoff (mm) variations during thesummers (JuneminusSeptember) of 1990minus2009 for the study areawere also recorded The water balance was defined as thedifference between precipitation (mm) and evaporation (mm)levels Because the moving average is typically used with timeseries data to smooth out short-term fluctuations and reflectlonger term trends we calculated the moving average tovisualize observation trends by creating a series of water balanceand runoff averages for the full data set The originalmeteorological data were provided by the China MeteorologicalData Sharing Service System (httpcdccmagovcn) asobserved at the meteorological station of Madoi County andrunoff data were provided by the hydrological station ofHuangheyan in Qinghai shown in Figure 1 To validate Ts andET values estimated via remote sensing and values estimatedbased on observations we compared interpolated data of

Figure 1 Study area source region of the Yellow River China (yellowdots denote ground sample points for satellite image interpretation)

Figure 2 Land cover changes of the study area from 1990 to 2009

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

C

observed spatial air temperature and ET with the estimatedvalues The process and results are described in Figure S-2 ofthe Supporting InformationStatistical Analysis Among the key variables we focused

on albedo (α) and the NDVI α reflects the roughness of theland surface and is related to vegetation conditions33 TheNDVI is used to quantify ground vegetation α NDVI Ts andET values retrieved from the raster-based results were extractedfrom pixels of different land cover types and then used to

perform the correlation analysis in SPSS 130 (SPSS Inc

Chicago IL) Pairs of variables (eg NDVIminusα αminusTs TsminusETand ETminusNDVI) were selected on the basis of causeminuseffectvegetation and regional climate processes3839 which were

closely associated with feedback loop pathways In addition α

NDVI Ts and ET trends from 1990 to 2009 were observed

according to different land cover patterns The correlations

among α NDVI Ts and ET and corresponding trends are

Figure 3 (a) Ts changes and annual air temperature changes shown by the curve and linear regression (b) ET changes and annual water balance andrunoff variations in the summers of 1990minus2009 for the study area shown as three-period moving average (Mavg) curves

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

D

anticipated to reveal interactions between surface structures andregional hydro-heat processes

RESULTS

Grassland Degradation and Recovery Regional landcover changes showed patterns of grassland degradation andrecovery in the summers of 1990minus2009 (Figure 2) Theproportion of high-coverage grassland expanded gradually fromapproximately 27 in 1990 to greater than 30 in 2009 Low-coverage grassland areas increased rapidly after 2000accounting for the largest proportion of land (39) Inaddition proportions of the other two land cover types middle-coverage grassland and bare soils decreased from 21 to 12and from 26 to 11 respectively Water body coverage levelsremained constant at a level of 8 for the entire period Thesetrends denote land cover changes after approximately 2000Hydro-heat Patterns and Regional Climate Change Ts

presents a spatial pattern with higher values of approximately2975 K in the north and lower values of approximately 2932 Kin the south (Figure 3a) For different land covers Ts values aresorted in ascending order (high- middle- and low-coveragegrasslands and bare soils) In the study area average summer airtemperatures increased by almost 2 degC over the past 20 years(Figure 3a) In general Ts values of the study area for thesummers of 1990minus2009 follow observed regional warmingtrends We also found that the Ts values of bare soil and low-and middle-coverage grassland areas appeared to decrease from2006 to 2009 (Figure 4)ET levels were lower at approximately 085 cmday in the

north and higher at approximately 175 cmday in the south(Figure 3b) ET values were highest in the water bodies andareas surrounding wetlands also exhibited higher ET values ofapproximately 155 cmday The ET values for the other typesof land cover were ranked in the following ascending order

bare soils and low- middle- and high-coverage grasslandsSandy riverbank areas generated very low ET values ofapproximately 037 cmday The ET value for all of the landcover types initially declined from 1990 to 2000 and thenincreased to 1990 levels In grasslands and bare soils ET levelsreached approximately 15 cmday in 1990 and decreased toapproximately 10 cmday in approximately 2000 denoting anET turning point Thereafter grassland and bare soil ET levelsreturned to approximately 15 cmday in 2006 and 2009(Figures 3b and 4) Meteorological records for the summers of1990minus2009 show that the water balance level reached its lowestvalue in 1996 and the region has since returned to highermoisture levels

Correlation between Key Variables and α and NDVITrends The correlation analysis results show how associationsbetween the key variables form the regional climate feedbackloop (Figure 4) Pearson linear coefficients are underlined andadded to each pair In 1990 1994 and 2000 all pairs weresignificantly correlated with the exception of the αminusTs pairThe scatter graphs show that NDVI was negatively related to αTs was negatively related to ET and ET was positively relatedto NDVI although the relationship between α and Ts remainsunclear In 1990 the αminusTs pair exhibited an insignificantlynegative relation whereas the 2006 and 2009 results generallyshowed a positive numerical relationship between the twovariables Apart from the hydro-heat pattern changes of Ts andET noted above α and NDVI trends were extracted on thebasis of different land cover types revealing that α reached amaximum value in 1994 and that NDVI initially declined butincreased after 1994 denoting the start of the restoration phase

DISCUSSION

Consequences of Ecosystem Degradation As an inlandplateau grasslandminuswetland ecosystem changes in the land

Figure 4 Correlation analysis pairs of NDVIminusα αminusTs TsminusET and ETminusNDVI and change trends of the key variables for 1990minus2009 Numbersmarked in each box denote the Pearson correlation coefficients for each pair of variables lowast and lowastlowast denote the significance levels of 005 and 001respectively

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

E

cover structure in this study area are related to the temperatureand hydrologic patterns of this region From 1990 to 2000 asgrassland areas degraded while bare soils expanded albedolevels of the land surface reached peak level in 1994 At thesame time surface temperatures increased continuously partlybecause of global warming but also because of a reduction inevaporative cooling40 On the basis of Ts spatial patterns thesurface heating effects on bare soils of the northern area becamemore pronounced reflecting patterns similar to those of theurban heat island effect40 This regional surface temperaturechange is attributable to several factors such as elevationradiation surface reflectivity precipitation and evaporation4041

However quantitative attributes of these factors were notexamined in this study The correlation analysis of the keyvariables showed an association between vegetation and hydro-heat environments of the study area based on the retrievedspatial data although these data are too limited to fully reflectcausality in the ecological degradation and restoration feedbackloop Temporal dimension variable change trends show that theα and NDVI turning point occurred in 1994 and that ET hasincreased since 2000 and this could be interpreted as a delayedresponse to improved vegetation conditionsEcosystem Restoration Effects According to references

and local policy documents humans began to play a positiverole in the regional feedback process and attempted to controldegradation trends in the study area beginning in 19967842

The government planned ecological conservation and restora-tion projects and has implemented these projects from 1996 tothe present These projects have aimed to improve primaryproductivity levels and ecosystem self-control capacities tomaintain healthy grasslands in the source region Theseimproved vegetation conditions may stabilize surface temper-atures and increase soil moisture and ET levels as shown inrelevant observations Runoff levels in the area have increasedsince 2000 and this trend has been significantly correlated withprecipitation patterns43minus45 In the present study comparisonsbetween annual water balance and runoff variations show adecline in ET until 2000 (Figure 3b) suggesting that runoff andET patterns interacted closely in the hydrologic cycleRegional Climate Feedback Mechanisms of the

Terrestrial Ecosystem Previous studies on interactionsbetween the land surface and hydroclimatological conditionshave mainly employed methods and tools such as satellitecloud images cloud frequency analyses canopy microclimatol-ogy sensors and meteorological observations to examine anddefine cloud formation feedback effects46minus48 Most of thesestudies examined forest ecosystems with less of a focus onalpine grassland ecosystems Still micro-to-macro processanalyses presented in previous studies suggest the presence ofregional climate feedback mechanisms ET improvements viavegetation recovery helped increase humidity levels in turnincreasing boundary convection and local precipitationlevels4748

In the neighboring QinghaiminusTibet Plateau source regionwhich is similar to the study area recent observations andstudies have shown a positive correlation between ET andhumidity as well as a general increase in precipitation andrunoff levels over the last 5minus10 years4445 Although causalinferences between restored vegetation levels and improvedhydrological conditions in the study area remain unclear basedon existing results we expect to reveal such mechanisms ingreater depth with the use of more observations andexperiments on alpine grassland ecosystems We conclude

that lower ET trends led to lower air moisture levels inhibitingcloud formation and local precipitation It is assumed thatdrought conditions inhibit vegetation growth potentiallyaccelerating grassland degradation further as shown in Figure5

Coupling System of the Atmospheric and TerrestrialHydrocycle According to principles of modern hydro-climatology the atmosphere serves as a climatic frame anddrives energy flows and water vapor transportation patternsSubsystems of the atmosphere and terrestrial ecosystems werecoupled to form a complete hydrocycle system Dividing thehydrocycle system into two branches [the atmospherichydrocycle (AHC) and terrestrial hydrocycle (THC)] helpedreveal global climate patterns50

The dominant factors of AHC mainly include atmosphericcirculation and energy flux trends in patterns of circulationAtmospheric circulation models primarily determine biomes atthe continental scale and strongly affect the construction andsuccession of ecosystems influencing global land coverstructures THC involves terrestrialminusatmosphere interfaceprocesses wherein the atmosphere interacts with geographicalhydrological and biological processes driven by key factors tosupport the development of ecosystems as shown in our studyKey factors include surface energy flux water availability andgroundwater balance ET plays a dominant role in THCbecause ET is driven by radiation and relies on thetransformation of groundwater into atmospheric water vaporwhich is widely related to soil and vegetation processes As partof the land cover structure vegetation determines the speedand means through which groundwater and precipitation movefrom the land surface to the atmosphere In turn studies haveshown that THC plays a key role in climate systems through itsinvolvement in the hydrologic cycleTo situate this analysis of long-term AHC impacts within

processes of ecological restoration it is also essential tounderstand this phenomenon and to make predictionsObservations show that the Asian monsoon has been lessactive over the last 20 years resulting in the drying andwarming of the QinghaiminusTibet Plateau51 Drought conditionsmay continue over a long period of time Therefore long-termand advanced ecological restoration and monitoring efforts are

Figure 5 Regional climate feedback loops of the degraded andrestored scenarios of typical land use landscapes for the study area(adapted with permission from ref 49 Copyright 1984 Elsevier)

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

F

essential Although previous climate change studies havepresented global climate models and have discussed energyuse strategies that mitigate global warming few works haveexamined regional THC processes to assess or manage regionalclimatic and hydrological changes Few climate policy studieshave determined the net impact of biophysical changesresulting from land use pattern changes33 Our work in thesource region of the Yellow River serves as an initial attempt toexplore the feasibility of applying innovative methodologies toaddress gaps in knowledge and to evaluate ecologicalrestoration strategies in future studies

ASSOCIATED CONTENTS Supporting InformationClassification criterions of the study area and the accuracyassessment for the land cover classification detailed methodsfor the calculation models of ATCOR2 and validation of theestimated results with meteorological data The SupportingInformation is available free of charge on the ACS Publicationswebsite at DOI 101021es505985q

AUTHOR INFORMATIONCorresponding AuthorTelephone +86-10-6279-4119 E-mail xuehua-hjxmailtsinghuaeducnNotesThe authors declare no competing financial interest

ACKNOWLEDGMENTSThis project was financially supported by the fund from theState Key Joint Laboratory of Environment Simulation andPollution Control China (11Y02ESPCT) and the projectnamed The Relationship of Birds Migration Patterns andHabitat Factors in Poyang Lake (2010CB530300-04) Theauthors thank the anonymous reviewers for their critiqueswhich improved earlier versions of this manuscript

REFERENCES(1) Armesto J J Bautista S Del Val E Ferguson B Garciacutea XGaxiola A Godinez-Alvarez H Gann G Lopez-Barrera FManson R Nu nez-Avila M Ortiz-Arrona C Tognetti PWilliams-Linera G Towards an ecological restoration networkReversing land degradation in Latin America Front Ecol Environ2007 5 (4) w1minusw4 DOI 1018901540-9295(2007)5[w1TAERNR]20CO2(2) Baustian M M Georgia M Dreelin E A Esselman PSchultze S R Qian L Awb T G Luo L Rose J BA Onehundred year review of the socioeconomic and ecological systems ofLake St Clair North America J Great Lakes Res 2014 40 15minus26DOI 101016jjglr201311006(3) Grumbine R E Assessing environmental security in ChinaFront Ecol Environ 2014 12 (7) 403minus411 DOI 101890130147(4) Bullon T Environmental assessment and land change analysis insemi-natural land covers applicable to land management Int J ApplEarth Obs Geoinf 2015 34 147minus156 DOI 101016jjag201408006(5) Liu J Linderman M Ouyang Z An L Yang J Zhang HEcological degradation in protected areas The case of Wolong NatureReserve for giant pandas Science 2001 292 (5514) 98minus101DOI 101126science1058104(6) Poyatos R Latron J Llorens P Land use and land coverchange after agricultural abandonment Mt Res Dev 2003 23 (4)362minus368 DOI 1016590276-4741(2003)023[0362LUALCC]20CO2(7) Feng J Wang T Xie C Eco-environmental degradation in thesource region of the Yellow River northeast QinghaiminusXizang Plateau

Environ Monit Assess 2006 122 (1minus3) 125minus143 DOI 101007s10661-005-9169-2(8) Wang G Cheng G Eco-environmental changes and causativeanalysis in the source regions of the Yangtze and Yellow Rivers ChinaEnviron Syst Decis 2000 20 (3) 221minus232 DOI 101023A1006703831018(9) Pan J Liu J Yellow River source area of land use and landscapepattern change and its ecological effects J Arid Land Res Environ2005 19 (4) 69minus74 (in Chinese) DOI 103969jissn1003-7578200504014(10) Xu J Song L Zhao Z Hu Y Liu C Monitoring grasslanddegradation dynamically at Maduo County in source region of YellowRiver in past 15 years based on remote sensing Arid Land Geogr 201235 (4) 615minus622 (in Chinese) DOI 1013826jcnkicn65-1103x201204018(11) Yang J Ding Y Chen R Spatial and temporal of variations ofalpine vegetation cover in the source regions of the Yangtze andYellow Rivers of the Tibetan Plateau from 1982 to 2001 Environ Geol2006 50 (3) 313minus322 DOI 101007s00254-006-0210-8(12) Davidson A D Detling J K Brown J H Ecological roles andconservation challenges of social burrowing herbivorous mammals inthe worldrsquos grasslands Front Ecol Environ 2012 10 (9) 477minus486DOI 101890110054(13) Qu J Li W Yang M Ji W Zhang Y Life history of theplateau pika (Ochotona curzoniae) in alpine meadows of the TibetanPlateau Mamm Biol 2013 78 (1) 68minus72 DOI 101016jmambio201209005(14) Liu J Nie H Studies on the population productivity ecology ofplateau pika III Trend of population dynamics in plateau pika withdensity-independent and density-dependent vital rates Acta TheriolSin 1992 12 (2) 139minus146 (in Chinese)(15) Liu W Zhou L Wang X The study of different grazingintensity on the role of plants and rodent studies Acta Ecol Sin 199919 (3) 88minus94 (in Chinese)(16) Bai W Zhang Y Xie G Shen Z Causes analysis of grasslanddegradation Maduo County in Yellow River source region Acta EcolSin 2002 13 (7) 823minus826 (in Chinese) DOI 1013287j1001-933220020194(17) Fang Y Qin D Ding Y Frozen soil change and adaptation ofanimal husbandry A case of the source regions of Yangtze and YellowRivers Environ Sci Policy 2011 14 (5) 555minus568 DOI 101016jenvsci201103012(18) Verburg P H van de Steeg J Veldkamp A Willemen LFrom land cover change to land function dynamics A major challengeto improve land characterization J Environ Manage 2009 90 (3)1327minus1335 DOI 101016jjenvman200808005(19) Barnett T P Adam J C Lettenmaier A D P Potentialimpacts of a warming climate on water availability in snow-dominatedregions Nature 2005 438 303minus309 DOI 101038nature04141(20) Chattopadhyay N Hulme M Evaporation and potentialevapotranspiration in India under conditions of recent and futureclimate change Agric For Meteorol 1997 87 55minus73 DOI 101016S0168-1923(97)00006-3(21) Liu M Xia Z Han S Wang X Tang Z Relationshipbetween variation of evapotranspiration and ecological deterioration insource region of Yellow River J Hohai Uni 2009 37 (6) 631minus634 (inChinese)(22) Yi X Yin Y Li G Peng J Temperature variation in recent50 years in the three-river headwaters region of Qinghai Province ActaGeol Sin 2011 66 (11) 1451minus1465 (in Chinese)(23) Hou W Li Y Soil surface humidity index and sensitivityanalysis of the climate factors that affect it in the Yellow River sourceregions J Glaciol Geocryol 2010 32 (6) 1226minus1233 (in Chinese)(24) Mauser W Schadlich S Modelling the spatial distribution ofevapotranspiration on different scales using remote sensing data JHydrol 1998 212 250minus267 DOI 101016S0022-1694(98)00228-5(25) Salati E Vose P B Amazon basin A system in equilibriumScience 1984 225 (4658) 129minus138 DOI 1023071693078

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

G

(26) Schlesinger W H Reynolds J F Cunningham G LHuenneke L F Jarrell W M Virginia R A Whitford W GBiological feedbacks in global desertification Science 1990 247 (4946)1043minus1048 DOI 101126science24749461043(27) Shukla J Nobre C Sellers P Amazon deforestation andclimate change Science 1990 247 (4948) 1322minus1325 DOI 101126science24749481322(28) Baron J S Hartman M D Kittel T G F Band L E OjimaD S Lammers R B Effects of land cover water redistribution andtemperature on ecosystem processes in the South Platte basin EcolAppl 1998 8 (4) 1037minus1051 DOI 1018901051-0761(1998)008[1037eolcwr]20co2(29) Boegh E Soegaard H Thomsen A Evaluating evapotranspi-ration rates and surface conditions using Landsat TM to estimateatmospheric resistance and surface resistance Remote Sens Environ2002 79 (2minus3) 329minus343 DOI 101016S0034-4257(01)00283-8(30) Sterling S M Ducharne A Polcher J The impact of globalland-cover change on the terrestrial water cycle Nat Clim Change2013 3 (4) 385minus390 DOI 101038nclimate1690(31) Loarie S R Lobell D B Asner G P Mu Q Field C BDirect impacts on local climate of sugar-cane expansion in Brazil NatClim Change 2011 1 (2) 105minus109 DOI 101038nclimate1067(32) Teuling A J Seneviratne S I Stockli R et al Contrastingresponse of European forest and grassland energy exchange toheatwaves Nat Geosci 2010 3 (10) 722minus727 DOI 101038ngeo950(33) Anderson R G Canadell J G Randerson J T et alBiophysical considerations in forestry for climate protection FrontEcol Environ 2011 9 (3) 174minus182 DOI 101890090179(34) Campo-Bescos M A Munoz-Carpena R Kaplan D ASouthworth J Zhu L Waylen P R Beyond precipitationPhysiographic gradients dictate the relative importance of environ-mental drivers on savanna vegetation PLoS One 2013 8 (8)No e72348 DOI 101371journalpone0072348(35) Schneider K Mauser W Processing and accuracy of Landsatthematic mapper data for lake surface temperature measurement Int JRemote Sens 1996 17 (11) 2027minus2041 DOI 10108001431169608948757(36) Li F Jackson T J Kustas W P Schmugge T J French AN Cosh M H Bindlish R Deriving land surface temperature fromLandsat 5 and 7 during SMEX02SMACEX Remote Sens Environ2004 92 (4) 521minus534 DOI 101016jrse200402018(37) Bastiaanssen W G M Menenti M Feddes R A Holtslag AA M A remote sensing surface energy balance algorithm for land(SEBAL) J Hydrol 1998 212 198minus212 DOI 101016S0022-1694(98)00253-4(38) Bonan G B Forests and climate change Forcings feedbacksand the climate benefits of forests Science 2008 320 (5882) 1444minus1449 DOI 101126science1155121(39) Li Z Liu X Ma T Kejia D Zhou Q Yao B Niu TRetrieval of the surface evapotranspiration patterns in the alpinegrasslandminuswetland ecosystem applying SEBAL model in the sourceregion of the Yellow River China Ecol Modell 2013 270 64minus75DOI 101016jecolmodel201309004(40) Zhao L Lee X Smith R B Oleson K Strong contributionsof local background climate to urban heat islands Nature 2014 511(7508) 216minus219 DOI 101038nature13462(41) van Heerwaarden C C de Arellano J V Gounou AGuichard F Couvreux F Understanding the daily cycle ofevapotranspiration A method to quantify the influence of forcingsand feedbacks J Hydrometeorol 2010 11 (6) 1405minus1422DOI 1011752010jhm12721(42) Chang G G Li F X Li L Changes and Restoration in SourceRegion of Three Rivers of China Metrology Press Beijing China 2010(in Chinese)(43) Tong L Xu X Fu Y Li S Wetland changes and theirresponses to climate change in the three-river headwaters region ofChina since the 1990s Energies 2014 7 (4) 2515minus2534DOI 103390en7042515

(44) Zhang Y Zhao X Zhao S Feng C Correlation betweenevapotranspiration and climate factors in warm steppe in source regionof Yangtze Yellow and Yalu Tsangpo Rivers J Desert Res 2010 30(2) 363minus368 (in Chinese)(45) Jiang Y Li D Variations of Tangnaihai runoff andprecipitation and temperature in the upper reach of the YellowRiverMeteorol Disaster Reduct Res 2011 34 (2) 51minus57 (in Chinese)(46) Molders N Kramm G Influence of wildfire induced land-cover changes on clouds and precipitation in interior AlaskaA casestudy Atmos Res 2007 84 (2) 142minus168 DOI 101016jatmos-res200606004(47) Obregon A Gehrig-Downie C Gradstein S R RollenbeckR Bendix J Canopy level fog occurrence in a tropical lowland forestof French Guiana as a prerequisite for high epiphyte diversity AgricFor Meteorol 2011 151 (3) 290minus300 DOI 101016jagrfor-met201011003(48) Harding R J Blyth E M Tuinenburg O A Wiltshire ALand atmosphere feedbacks and their role in the water resources of theGanges basin Sci Total Environ 2013 468 (S1) S85minusS92DOI 101016jscitotenv201303016(49) Bradshaw A D Ecological principles and land reclamationpractice Landscape Plann 1984 11 (1) 35minus48(50) Shelton M L Hydroclimatology Perspectives and ApplicationsCambridge Press Cambridge UK 2009(51) Liu Y Lu S Li S Gao Y Numerical simulation of impact ofland surface changes on regional climatic environment in sourceregions of Yellow River Plateau Meteorol 2009 28 (2) 327minus334 (inChinese)

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

H

S-1 and S-2 of the Supporting Information) A confusion matrixand Kappa statistics were used to make an accuracy assessmentwhich showed that the final result of the classification wasacceptable (see Table S-3 of the Supporting Information)Retrieval of Vegetation Conditions Ts and ET

Patterns Satellite-derived vegetation indices such as thenormalized difference vegetation index (NDVI) have beenclosely associated with primary vegetation production34 NDVIis defined as follows

= minus +r r r rNDVI ( )( )NIR red NIR red

where rNIR and rred represent surface reflectance levels averagedover wavelength ranges of infrared and visible infrared regionsof the spectrum respectively NDVI values and trends wereused to denote vegetation degradation or restoration trendsbased on LUCC dataRegional climate indicators regarding surface energy and

water budget levels were examined using the same set of

Landsat-5 TM images Ts was calculated as an indicator ofsurface energy conditions based on the following formula3536

ε= +T K K Lln( 1)s 2 1 6 (1)

where K1 = 60776 times 106 W cmminus2 srminus1 μmminus1 K2 = 126056 times106 W cmminus2 srminus1 μmminus1 with K1 and K2 being radiationconstants for Landsat-5 images L6 is the spectral radiance ofband 6 in Landsat-5 images and ε is the atmospheric emissivitylevel determined on the basis of the NDVIWe measured the value adding product (VAP) of plain areas

in the atmospheric correction (ATCOR2) module platform ofthe ERDAS 2011 remote sensing processing software toretrieve ET (cmday) data for the region Principles of theproduct are based on the following surface energy balanceequations37

= + +R H G LEn (2)

=ET LE286 (3)

where the terms denote composites of net radiation (RnW mminus2) sensible heat flux (H W mminus2) ground heat flux (GW mminus2) and latent heat flux (LE W mminus2) The modelingmethod involved two main tasks (1) calculating the radiationbalance based on remote sensing pixel reflectance levels and (2)calculating the heat balance using field knowledge that includessurface vegetation and meteorological conditions (see Table S-4 and Figure S-1 of the Supporting Information)

Comparisons to Meteorological and HydrologicalRecords To compare the estimated results with meteoro-logical and hydrological observations air temperature (degC)water balance (mm) and runoff (mm) variations during thesummers (JuneminusSeptember) of 1990minus2009 for the study areawere also recorded The water balance was defined as thedifference between precipitation (mm) and evaporation (mm)levels Because the moving average is typically used with timeseries data to smooth out short-term fluctuations and reflectlonger term trends we calculated the moving average tovisualize observation trends by creating a series of water balanceand runoff averages for the full data set The originalmeteorological data were provided by the China MeteorologicalData Sharing Service System (httpcdccmagovcn) asobserved at the meteorological station of Madoi County andrunoff data were provided by the hydrological station ofHuangheyan in Qinghai shown in Figure 1 To validate Ts andET values estimated via remote sensing and values estimatedbased on observations we compared interpolated data of

Figure 1 Study area source region of the Yellow River China (yellowdots denote ground sample points for satellite image interpretation)

Figure 2 Land cover changes of the study area from 1990 to 2009

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

C

observed spatial air temperature and ET with the estimatedvalues The process and results are described in Figure S-2 ofthe Supporting InformationStatistical Analysis Among the key variables we focused

on albedo (α) and the NDVI α reflects the roughness of theland surface and is related to vegetation conditions33 TheNDVI is used to quantify ground vegetation α NDVI Ts andET values retrieved from the raster-based results were extractedfrom pixels of different land cover types and then used to

perform the correlation analysis in SPSS 130 (SPSS Inc

Chicago IL) Pairs of variables (eg NDVIminusα αminusTs TsminusETand ETminusNDVI) were selected on the basis of causeminuseffectvegetation and regional climate processes3839 which were

closely associated with feedback loop pathways In addition α

NDVI Ts and ET trends from 1990 to 2009 were observed

according to different land cover patterns The correlations

among α NDVI Ts and ET and corresponding trends are

Figure 3 (a) Ts changes and annual air temperature changes shown by the curve and linear regression (b) ET changes and annual water balance andrunoff variations in the summers of 1990minus2009 for the study area shown as three-period moving average (Mavg) curves

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

D

anticipated to reveal interactions between surface structures andregional hydro-heat processes

RESULTS

Grassland Degradation and Recovery Regional landcover changes showed patterns of grassland degradation andrecovery in the summers of 1990minus2009 (Figure 2) Theproportion of high-coverage grassland expanded gradually fromapproximately 27 in 1990 to greater than 30 in 2009 Low-coverage grassland areas increased rapidly after 2000accounting for the largest proportion of land (39) Inaddition proportions of the other two land cover types middle-coverage grassland and bare soils decreased from 21 to 12and from 26 to 11 respectively Water body coverage levelsremained constant at a level of 8 for the entire period Thesetrends denote land cover changes after approximately 2000Hydro-heat Patterns and Regional Climate Change Ts

presents a spatial pattern with higher values of approximately2975 K in the north and lower values of approximately 2932 Kin the south (Figure 3a) For different land covers Ts values aresorted in ascending order (high- middle- and low-coveragegrasslands and bare soils) In the study area average summer airtemperatures increased by almost 2 degC over the past 20 years(Figure 3a) In general Ts values of the study area for thesummers of 1990minus2009 follow observed regional warmingtrends We also found that the Ts values of bare soil and low-and middle-coverage grassland areas appeared to decrease from2006 to 2009 (Figure 4)ET levels were lower at approximately 085 cmday in the

north and higher at approximately 175 cmday in the south(Figure 3b) ET values were highest in the water bodies andareas surrounding wetlands also exhibited higher ET values ofapproximately 155 cmday The ET values for the other typesof land cover were ranked in the following ascending order

bare soils and low- middle- and high-coverage grasslandsSandy riverbank areas generated very low ET values ofapproximately 037 cmday The ET value for all of the landcover types initially declined from 1990 to 2000 and thenincreased to 1990 levels In grasslands and bare soils ET levelsreached approximately 15 cmday in 1990 and decreased toapproximately 10 cmday in approximately 2000 denoting anET turning point Thereafter grassland and bare soil ET levelsreturned to approximately 15 cmday in 2006 and 2009(Figures 3b and 4) Meteorological records for the summers of1990minus2009 show that the water balance level reached its lowestvalue in 1996 and the region has since returned to highermoisture levels

Correlation between Key Variables and α and NDVITrends The correlation analysis results show how associationsbetween the key variables form the regional climate feedbackloop (Figure 4) Pearson linear coefficients are underlined andadded to each pair In 1990 1994 and 2000 all pairs weresignificantly correlated with the exception of the αminusTs pairThe scatter graphs show that NDVI was negatively related to αTs was negatively related to ET and ET was positively relatedto NDVI although the relationship between α and Ts remainsunclear In 1990 the αminusTs pair exhibited an insignificantlynegative relation whereas the 2006 and 2009 results generallyshowed a positive numerical relationship between the twovariables Apart from the hydro-heat pattern changes of Ts andET noted above α and NDVI trends were extracted on thebasis of different land cover types revealing that α reached amaximum value in 1994 and that NDVI initially declined butincreased after 1994 denoting the start of the restoration phase

DISCUSSION

Consequences of Ecosystem Degradation As an inlandplateau grasslandminuswetland ecosystem changes in the land

Figure 4 Correlation analysis pairs of NDVIminusα αminusTs TsminusET and ETminusNDVI and change trends of the key variables for 1990minus2009 Numbersmarked in each box denote the Pearson correlation coefficients for each pair of variables lowast and lowastlowast denote the significance levels of 005 and 001respectively

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

E

cover structure in this study area are related to the temperatureand hydrologic patterns of this region From 1990 to 2000 asgrassland areas degraded while bare soils expanded albedolevels of the land surface reached peak level in 1994 At thesame time surface temperatures increased continuously partlybecause of global warming but also because of a reduction inevaporative cooling40 On the basis of Ts spatial patterns thesurface heating effects on bare soils of the northern area becamemore pronounced reflecting patterns similar to those of theurban heat island effect40 This regional surface temperaturechange is attributable to several factors such as elevationradiation surface reflectivity precipitation and evaporation4041

However quantitative attributes of these factors were notexamined in this study The correlation analysis of the keyvariables showed an association between vegetation and hydro-heat environments of the study area based on the retrievedspatial data although these data are too limited to fully reflectcausality in the ecological degradation and restoration feedbackloop Temporal dimension variable change trends show that theα and NDVI turning point occurred in 1994 and that ET hasincreased since 2000 and this could be interpreted as a delayedresponse to improved vegetation conditionsEcosystem Restoration Effects According to references

and local policy documents humans began to play a positiverole in the regional feedback process and attempted to controldegradation trends in the study area beginning in 19967842

The government planned ecological conservation and restora-tion projects and has implemented these projects from 1996 tothe present These projects have aimed to improve primaryproductivity levels and ecosystem self-control capacities tomaintain healthy grasslands in the source region Theseimproved vegetation conditions may stabilize surface temper-atures and increase soil moisture and ET levels as shown inrelevant observations Runoff levels in the area have increasedsince 2000 and this trend has been significantly correlated withprecipitation patterns43minus45 In the present study comparisonsbetween annual water balance and runoff variations show adecline in ET until 2000 (Figure 3b) suggesting that runoff andET patterns interacted closely in the hydrologic cycleRegional Climate Feedback Mechanisms of the

Terrestrial Ecosystem Previous studies on interactionsbetween the land surface and hydroclimatological conditionshave mainly employed methods and tools such as satellitecloud images cloud frequency analyses canopy microclimatol-ogy sensors and meteorological observations to examine anddefine cloud formation feedback effects46minus48 Most of thesestudies examined forest ecosystems with less of a focus onalpine grassland ecosystems Still micro-to-macro processanalyses presented in previous studies suggest the presence ofregional climate feedback mechanisms ET improvements viavegetation recovery helped increase humidity levels in turnincreasing boundary convection and local precipitationlevels4748

In the neighboring QinghaiminusTibet Plateau source regionwhich is similar to the study area recent observations andstudies have shown a positive correlation between ET andhumidity as well as a general increase in precipitation andrunoff levels over the last 5minus10 years4445 Although causalinferences between restored vegetation levels and improvedhydrological conditions in the study area remain unclear basedon existing results we expect to reveal such mechanisms ingreater depth with the use of more observations andexperiments on alpine grassland ecosystems We conclude

that lower ET trends led to lower air moisture levels inhibitingcloud formation and local precipitation It is assumed thatdrought conditions inhibit vegetation growth potentiallyaccelerating grassland degradation further as shown in Figure5

Coupling System of the Atmospheric and TerrestrialHydrocycle According to principles of modern hydro-climatology the atmosphere serves as a climatic frame anddrives energy flows and water vapor transportation patternsSubsystems of the atmosphere and terrestrial ecosystems werecoupled to form a complete hydrocycle system Dividing thehydrocycle system into two branches [the atmospherichydrocycle (AHC) and terrestrial hydrocycle (THC)] helpedreveal global climate patterns50

The dominant factors of AHC mainly include atmosphericcirculation and energy flux trends in patterns of circulationAtmospheric circulation models primarily determine biomes atthe continental scale and strongly affect the construction andsuccession of ecosystems influencing global land coverstructures THC involves terrestrialminusatmosphere interfaceprocesses wherein the atmosphere interacts with geographicalhydrological and biological processes driven by key factors tosupport the development of ecosystems as shown in our studyKey factors include surface energy flux water availability andgroundwater balance ET plays a dominant role in THCbecause ET is driven by radiation and relies on thetransformation of groundwater into atmospheric water vaporwhich is widely related to soil and vegetation processes As partof the land cover structure vegetation determines the speedand means through which groundwater and precipitation movefrom the land surface to the atmosphere In turn studies haveshown that THC plays a key role in climate systems through itsinvolvement in the hydrologic cycleTo situate this analysis of long-term AHC impacts within

processes of ecological restoration it is also essential tounderstand this phenomenon and to make predictionsObservations show that the Asian monsoon has been lessactive over the last 20 years resulting in the drying andwarming of the QinghaiminusTibet Plateau51 Drought conditionsmay continue over a long period of time Therefore long-termand advanced ecological restoration and monitoring efforts are

Figure 5 Regional climate feedback loops of the degraded andrestored scenarios of typical land use landscapes for the study area(adapted with permission from ref 49 Copyright 1984 Elsevier)

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

F

essential Although previous climate change studies havepresented global climate models and have discussed energyuse strategies that mitigate global warming few works haveexamined regional THC processes to assess or manage regionalclimatic and hydrological changes Few climate policy studieshave determined the net impact of biophysical changesresulting from land use pattern changes33 Our work in thesource region of the Yellow River serves as an initial attempt toexplore the feasibility of applying innovative methodologies toaddress gaps in knowledge and to evaluate ecologicalrestoration strategies in future studies

ASSOCIATED CONTENTS Supporting InformationClassification criterions of the study area and the accuracyassessment for the land cover classification detailed methodsfor the calculation models of ATCOR2 and validation of theestimated results with meteorological data The SupportingInformation is available free of charge on the ACS Publicationswebsite at DOI 101021es505985q

AUTHOR INFORMATIONCorresponding AuthorTelephone +86-10-6279-4119 E-mail xuehua-hjxmailtsinghuaeducnNotesThe authors declare no competing financial interest

ACKNOWLEDGMENTSThis project was financially supported by the fund from theState Key Joint Laboratory of Environment Simulation andPollution Control China (11Y02ESPCT) and the projectnamed The Relationship of Birds Migration Patterns andHabitat Factors in Poyang Lake (2010CB530300-04) Theauthors thank the anonymous reviewers for their critiqueswhich improved earlier versions of this manuscript

REFERENCES(1) Armesto J J Bautista S Del Val E Ferguson B Garciacutea XGaxiola A Godinez-Alvarez H Gann G Lopez-Barrera FManson R Nu nez-Avila M Ortiz-Arrona C Tognetti PWilliams-Linera G Towards an ecological restoration networkReversing land degradation in Latin America Front Ecol Environ2007 5 (4) w1minusw4 DOI 1018901540-9295(2007)5[w1TAERNR]20CO2(2) Baustian M M Georgia M Dreelin E A Esselman PSchultze S R Qian L Awb T G Luo L Rose J BA Onehundred year review of the socioeconomic and ecological systems ofLake St Clair North America J Great Lakes Res 2014 40 15minus26DOI 101016jjglr201311006(3) Grumbine R E Assessing environmental security in ChinaFront Ecol Environ 2014 12 (7) 403minus411 DOI 101890130147(4) Bullon T Environmental assessment and land change analysis insemi-natural land covers applicable to land management Int J ApplEarth Obs Geoinf 2015 34 147minus156 DOI 101016jjag201408006(5) Liu J Linderman M Ouyang Z An L Yang J Zhang HEcological degradation in protected areas The case of Wolong NatureReserve for giant pandas Science 2001 292 (5514) 98minus101DOI 101126science1058104(6) Poyatos R Latron J Llorens P Land use and land coverchange after agricultural abandonment Mt Res Dev 2003 23 (4)362minus368 DOI 1016590276-4741(2003)023[0362LUALCC]20CO2(7) Feng J Wang T Xie C Eco-environmental degradation in thesource region of the Yellow River northeast QinghaiminusXizang Plateau

Environ Monit Assess 2006 122 (1minus3) 125minus143 DOI 101007s10661-005-9169-2(8) Wang G Cheng G Eco-environmental changes and causativeanalysis in the source regions of the Yangtze and Yellow Rivers ChinaEnviron Syst Decis 2000 20 (3) 221minus232 DOI 101023A1006703831018(9) Pan J Liu J Yellow River source area of land use and landscapepattern change and its ecological effects J Arid Land Res Environ2005 19 (4) 69minus74 (in Chinese) DOI 103969jissn1003-7578200504014(10) Xu J Song L Zhao Z Hu Y Liu C Monitoring grasslanddegradation dynamically at Maduo County in source region of YellowRiver in past 15 years based on remote sensing Arid Land Geogr 201235 (4) 615minus622 (in Chinese) DOI 1013826jcnkicn65-1103x201204018(11) Yang J Ding Y Chen R Spatial and temporal of variations ofalpine vegetation cover in the source regions of the Yangtze andYellow Rivers of the Tibetan Plateau from 1982 to 2001 Environ Geol2006 50 (3) 313minus322 DOI 101007s00254-006-0210-8(12) Davidson A D Detling J K Brown J H Ecological roles andconservation challenges of social burrowing herbivorous mammals inthe worldrsquos grasslands Front Ecol Environ 2012 10 (9) 477minus486DOI 101890110054(13) Qu J Li W Yang M Ji W Zhang Y Life history of theplateau pika (Ochotona curzoniae) in alpine meadows of the TibetanPlateau Mamm Biol 2013 78 (1) 68minus72 DOI 101016jmambio201209005(14) Liu J Nie H Studies on the population productivity ecology ofplateau pika III Trend of population dynamics in plateau pika withdensity-independent and density-dependent vital rates Acta TheriolSin 1992 12 (2) 139minus146 (in Chinese)(15) Liu W Zhou L Wang X The study of different grazingintensity on the role of plants and rodent studies Acta Ecol Sin 199919 (3) 88minus94 (in Chinese)(16) Bai W Zhang Y Xie G Shen Z Causes analysis of grasslanddegradation Maduo County in Yellow River source region Acta EcolSin 2002 13 (7) 823minus826 (in Chinese) DOI 1013287j1001-933220020194(17) Fang Y Qin D Ding Y Frozen soil change and adaptation ofanimal husbandry A case of the source regions of Yangtze and YellowRivers Environ Sci Policy 2011 14 (5) 555minus568 DOI 101016jenvsci201103012(18) Verburg P H van de Steeg J Veldkamp A Willemen LFrom land cover change to land function dynamics A major challengeto improve land characterization J Environ Manage 2009 90 (3)1327minus1335 DOI 101016jjenvman200808005(19) Barnett T P Adam J C Lettenmaier A D P Potentialimpacts of a warming climate on water availability in snow-dominatedregions Nature 2005 438 303minus309 DOI 101038nature04141(20) Chattopadhyay N Hulme M Evaporation and potentialevapotranspiration in India under conditions of recent and futureclimate change Agric For Meteorol 1997 87 55minus73 DOI 101016S0168-1923(97)00006-3(21) Liu M Xia Z Han S Wang X Tang Z Relationshipbetween variation of evapotranspiration and ecological deterioration insource region of Yellow River J Hohai Uni 2009 37 (6) 631minus634 (inChinese)(22) Yi X Yin Y Li G Peng J Temperature variation in recent50 years in the three-river headwaters region of Qinghai Province ActaGeol Sin 2011 66 (11) 1451minus1465 (in Chinese)(23) Hou W Li Y Soil surface humidity index and sensitivityanalysis of the climate factors that affect it in the Yellow River sourceregions J Glaciol Geocryol 2010 32 (6) 1226minus1233 (in Chinese)(24) Mauser W Schadlich S Modelling the spatial distribution ofevapotranspiration on different scales using remote sensing data JHydrol 1998 212 250minus267 DOI 101016S0022-1694(98)00228-5(25) Salati E Vose P B Amazon basin A system in equilibriumScience 1984 225 (4658) 129minus138 DOI 1023071693078

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

G

(26) Schlesinger W H Reynolds J F Cunningham G LHuenneke L F Jarrell W M Virginia R A Whitford W GBiological feedbacks in global desertification Science 1990 247 (4946)1043minus1048 DOI 101126science24749461043(27) Shukla J Nobre C Sellers P Amazon deforestation andclimate change Science 1990 247 (4948) 1322minus1325 DOI 101126science24749481322(28) Baron J S Hartman M D Kittel T G F Band L E OjimaD S Lammers R B Effects of land cover water redistribution andtemperature on ecosystem processes in the South Platte basin EcolAppl 1998 8 (4) 1037minus1051 DOI 1018901051-0761(1998)008[1037eolcwr]20co2(29) Boegh E Soegaard H Thomsen A Evaluating evapotranspi-ration rates and surface conditions using Landsat TM to estimateatmospheric resistance and surface resistance Remote Sens Environ2002 79 (2minus3) 329minus343 DOI 101016S0034-4257(01)00283-8(30) Sterling S M Ducharne A Polcher J The impact of globalland-cover change on the terrestrial water cycle Nat Clim Change2013 3 (4) 385minus390 DOI 101038nclimate1690(31) Loarie S R Lobell D B Asner G P Mu Q Field C BDirect impacts on local climate of sugar-cane expansion in Brazil NatClim Change 2011 1 (2) 105minus109 DOI 101038nclimate1067(32) Teuling A J Seneviratne S I Stockli R et al Contrastingresponse of European forest and grassland energy exchange toheatwaves Nat Geosci 2010 3 (10) 722minus727 DOI 101038ngeo950(33) Anderson R G Canadell J G Randerson J T et alBiophysical considerations in forestry for climate protection FrontEcol Environ 2011 9 (3) 174minus182 DOI 101890090179(34) Campo-Bescos M A Munoz-Carpena R Kaplan D ASouthworth J Zhu L Waylen P R Beyond precipitationPhysiographic gradients dictate the relative importance of environ-mental drivers on savanna vegetation PLoS One 2013 8 (8)No e72348 DOI 101371journalpone0072348(35) Schneider K Mauser W Processing and accuracy of Landsatthematic mapper data for lake surface temperature measurement Int JRemote Sens 1996 17 (11) 2027minus2041 DOI 10108001431169608948757(36) Li F Jackson T J Kustas W P Schmugge T J French AN Cosh M H Bindlish R Deriving land surface temperature fromLandsat 5 and 7 during SMEX02SMACEX Remote Sens Environ2004 92 (4) 521minus534 DOI 101016jrse200402018(37) Bastiaanssen W G M Menenti M Feddes R A Holtslag AA M A remote sensing surface energy balance algorithm for land(SEBAL) J Hydrol 1998 212 198minus212 DOI 101016S0022-1694(98)00253-4(38) Bonan G B Forests and climate change Forcings feedbacksand the climate benefits of forests Science 2008 320 (5882) 1444minus1449 DOI 101126science1155121(39) Li Z Liu X Ma T Kejia D Zhou Q Yao B Niu TRetrieval of the surface evapotranspiration patterns in the alpinegrasslandminuswetland ecosystem applying SEBAL model in the sourceregion of the Yellow River China Ecol Modell 2013 270 64minus75DOI 101016jecolmodel201309004(40) Zhao L Lee X Smith R B Oleson K Strong contributionsof local background climate to urban heat islands Nature 2014 511(7508) 216minus219 DOI 101038nature13462(41) van Heerwaarden C C de Arellano J V Gounou AGuichard F Couvreux F Understanding the daily cycle ofevapotranspiration A method to quantify the influence of forcingsand feedbacks J Hydrometeorol 2010 11 (6) 1405minus1422DOI 1011752010jhm12721(42) Chang G G Li F X Li L Changes and Restoration in SourceRegion of Three Rivers of China Metrology Press Beijing China 2010(in Chinese)(43) Tong L Xu X Fu Y Li S Wetland changes and theirresponses to climate change in the three-river headwaters region ofChina since the 1990s Energies 2014 7 (4) 2515minus2534DOI 103390en7042515

(44) Zhang Y Zhao X Zhao S Feng C Correlation betweenevapotranspiration and climate factors in warm steppe in source regionof Yangtze Yellow and Yalu Tsangpo Rivers J Desert Res 2010 30(2) 363minus368 (in Chinese)(45) Jiang Y Li D Variations of Tangnaihai runoff andprecipitation and temperature in the upper reach of the YellowRiverMeteorol Disaster Reduct Res 2011 34 (2) 51minus57 (in Chinese)(46) Molders N Kramm G Influence of wildfire induced land-cover changes on clouds and precipitation in interior AlaskaA casestudy Atmos Res 2007 84 (2) 142minus168 DOI 101016jatmos-res200606004(47) Obregon A Gehrig-Downie C Gradstein S R RollenbeckR Bendix J Canopy level fog occurrence in a tropical lowland forestof French Guiana as a prerequisite for high epiphyte diversity AgricFor Meteorol 2011 151 (3) 290minus300 DOI 101016jagrfor-met201011003(48) Harding R J Blyth E M Tuinenburg O A Wiltshire ALand atmosphere feedbacks and their role in the water resources of theGanges basin Sci Total Environ 2013 468 (S1) S85minusS92DOI 101016jscitotenv201303016(49) Bradshaw A D Ecological principles and land reclamationpractice Landscape Plann 1984 11 (1) 35minus48(50) Shelton M L Hydroclimatology Perspectives and ApplicationsCambridge Press Cambridge UK 2009(51) Liu Y Lu S Li S Gao Y Numerical simulation of impact ofland surface changes on regional climatic environment in sourceregions of Yellow River Plateau Meteorol 2009 28 (2) 327minus334 (inChinese)

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

H

observed spatial air temperature and ET with the estimatedvalues The process and results are described in Figure S-2 ofthe Supporting InformationStatistical Analysis Among the key variables we focused

on albedo (α) and the NDVI α reflects the roughness of theland surface and is related to vegetation conditions33 TheNDVI is used to quantify ground vegetation α NDVI Ts andET values retrieved from the raster-based results were extractedfrom pixels of different land cover types and then used to

perform the correlation analysis in SPSS 130 (SPSS Inc

Chicago IL) Pairs of variables (eg NDVIminusα αminusTs TsminusETand ETminusNDVI) were selected on the basis of causeminuseffectvegetation and regional climate processes3839 which were

closely associated with feedback loop pathways In addition α

NDVI Ts and ET trends from 1990 to 2009 were observed

according to different land cover patterns The correlations

among α NDVI Ts and ET and corresponding trends are

Figure 3 (a) Ts changes and annual air temperature changes shown by the curve and linear regression (b) ET changes and annual water balance andrunoff variations in the summers of 1990minus2009 for the study area shown as three-period moving average (Mavg) curves

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

D

anticipated to reveal interactions between surface structures andregional hydro-heat processes

RESULTS

Grassland Degradation and Recovery Regional landcover changes showed patterns of grassland degradation andrecovery in the summers of 1990minus2009 (Figure 2) Theproportion of high-coverage grassland expanded gradually fromapproximately 27 in 1990 to greater than 30 in 2009 Low-coverage grassland areas increased rapidly after 2000accounting for the largest proportion of land (39) Inaddition proportions of the other two land cover types middle-coverage grassland and bare soils decreased from 21 to 12and from 26 to 11 respectively Water body coverage levelsremained constant at a level of 8 for the entire period Thesetrends denote land cover changes after approximately 2000Hydro-heat Patterns and Regional Climate Change Ts

presents a spatial pattern with higher values of approximately2975 K in the north and lower values of approximately 2932 Kin the south (Figure 3a) For different land covers Ts values aresorted in ascending order (high- middle- and low-coveragegrasslands and bare soils) In the study area average summer airtemperatures increased by almost 2 degC over the past 20 years(Figure 3a) In general Ts values of the study area for thesummers of 1990minus2009 follow observed regional warmingtrends We also found that the Ts values of bare soil and low-and middle-coverage grassland areas appeared to decrease from2006 to 2009 (Figure 4)ET levels were lower at approximately 085 cmday in the

north and higher at approximately 175 cmday in the south(Figure 3b) ET values were highest in the water bodies andareas surrounding wetlands also exhibited higher ET values ofapproximately 155 cmday The ET values for the other typesof land cover were ranked in the following ascending order

bare soils and low- middle- and high-coverage grasslandsSandy riverbank areas generated very low ET values ofapproximately 037 cmday The ET value for all of the landcover types initially declined from 1990 to 2000 and thenincreased to 1990 levels In grasslands and bare soils ET levelsreached approximately 15 cmday in 1990 and decreased toapproximately 10 cmday in approximately 2000 denoting anET turning point Thereafter grassland and bare soil ET levelsreturned to approximately 15 cmday in 2006 and 2009(Figures 3b and 4) Meteorological records for the summers of1990minus2009 show that the water balance level reached its lowestvalue in 1996 and the region has since returned to highermoisture levels

Correlation between Key Variables and α and NDVITrends The correlation analysis results show how associationsbetween the key variables form the regional climate feedbackloop (Figure 4) Pearson linear coefficients are underlined andadded to each pair In 1990 1994 and 2000 all pairs weresignificantly correlated with the exception of the αminusTs pairThe scatter graphs show that NDVI was negatively related to αTs was negatively related to ET and ET was positively relatedto NDVI although the relationship between α and Ts remainsunclear In 1990 the αminusTs pair exhibited an insignificantlynegative relation whereas the 2006 and 2009 results generallyshowed a positive numerical relationship between the twovariables Apart from the hydro-heat pattern changes of Ts andET noted above α and NDVI trends were extracted on thebasis of different land cover types revealing that α reached amaximum value in 1994 and that NDVI initially declined butincreased after 1994 denoting the start of the restoration phase

DISCUSSION

Consequences of Ecosystem Degradation As an inlandplateau grasslandminuswetland ecosystem changes in the land

Figure 4 Correlation analysis pairs of NDVIminusα αminusTs TsminusET and ETminusNDVI and change trends of the key variables for 1990minus2009 Numbersmarked in each box denote the Pearson correlation coefficients for each pair of variables lowast and lowastlowast denote the significance levels of 005 and 001respectively

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

E

cover structure in this study area are related to the temperatureand hydrologic patterns of this region From 1990 to 2000 asgrassland areas degraded while bare soils expanded albedolevels of the land surface reached peak level in 1994 At thesame time surface temperatures increased continuously partlybecause of global warming but also because of a reduction inevaporative cooling40 On the basis of Ts spatial patterns thesurface heating effects on bare soils of the northern area becamemore pronounced reflecting patterns similar to those of theurban heat island effect40 This regional surface temperaturechange is attributable to several factors such as elevationradiation surface reflectivity precipitation and evaporation4041

However quantitative attributes of these factors were notexamined in this study The correlation analysis of the keyvariables showed an association between vegetation and hydro-heat environments of the study area based on the retrievedspatial data although these data are too limited to fully reflectcausality in the ecological degradation and restoration feedbackloop Temporal dimension variable change trends show that theα and NDVI turning point occurred in 1994 and that ET hasincreased since 2000 and this could be interpreted as a delayedresponse to improved vegetation conditionsEcosystem Restoration Effects According to references

and local policy documents humans began to play a positiverole in the regional feedback process and attempted to controldegradation trends in the study area beginning in 19967842

The government planned ecological conservation and restora-tion projects and has implemented these projects from 1996 tothe present These projects have aimed to improve primaryproductivity levels and ecosystem self-control capacities tomaintain healthy grasslands in the source region Theseimproved vegetation conditions may stabilize surface temper-atures and increase soil moisture and ET levels as shown inrelevant observations Runoff levels in the area have increasedsince 2000 and this trend has been significantly correlated withprecipitation patterns43minus45 In the present study comparisonsbetween annual water balance and runoff variations show adecline in ET until 2000 (Figure 3b) suggesting that runoff andET patterns interacted closely in the hydrologic cycleRegional Climate Feedback Mechanisms of the

Terrestrial Ecosystem Previous studies on interactionsbetween the land surface and hydroclimatological conditionshave mainly employed methods and tools such as satellitecloud images cloud frequency analyses canopy microclimatol-ogy sensors and meteorological observations to examine anddefine cloud formation feedback effects46minus48 Most of thesestudies examined forest ecosystems with less of a focus onalpine grassland ecosystems Still micro-to-macro processanalyses presented in previous studies suggest the presence ofregional climate feedback mechanisms ET improvements viavegetation recovery helped increase humidity levels in turnincreasing boundary convection and local precipitationlevels4748

In the neighboring QinghaiminusTibet Plateau source regionwhich is similar to the study area recent observations andstudies have shown a positive correlation between ET andhumidity as well as a general increase in precipitation andrunoff levels over the last 5minus10 years4445 Although causalinferences between restored vegetation levels and improvedhydrological conditions in the study area remain unclear basedon existing results we expect to reveal such mechanisms ingreater depth with the use of more observations andexperiments on alpine grassland ecosystems We conclude

that lower ET trends led to lower air moisture levels inhibitingcloud formation and local precipitation It is assumed thatdrought conditions inhibit vegetation growth potentiallyaccelerating grassland degradation further as shown in Figure5

Coupling System of the Atmospheric and TerrestrialHydrocycle According to principles of modern hydro-climatology the atmosphere serves as a climatic frame anddrives energy flows and water vapor transportation patternsSubsystems of the atmosphere and terrestrial ecosystems werecoupled to form a complete hydrocycle system Dividing thehydrocycle system into two branches [the atmospherichydrocycle (AHC) and terrestrial hydrocycle (THC)] helpedreveal global climate patterns50

The dominant factors of AHC mainly include atmosphericcirculation and energy flux trends in patterns of circulationAtmospheric circulation models primarily determine biomes atthe continental scale and strongly affect the construction andsuccession of ecosystems influencing global land coverstructures THC involves terrestrialminusatmosphere interfaceprocesses wherein the atmosphere interacts with geographicalhydrological and biological processes driven by key factors tosupport the development of ecosystems as shown in our studyKey factors include surface energy flux water availability andgroundwater balance ET plays a dominant role in THCbecause ET is driven by radiation and relies on thetransformation of groundwater into atmospheric water vaporwhich is widely related to soil and vegetation processes As partof the land cover structure vegetation determines the speedand means through which groundwater and precipitation movefrom the land surface to the atmosphere In turn studies haveshown that THC plays a key role in climate systems through itsinvolvement in the hydrologic cycleTo situate this analysis of long-term AHC impacts within

processes of ecological restoration it is also essential tounderstand this phenomenon and to make predictionsObservations show that the Asian monsoon has been lessactive over the last 20 years resulting in the drying andwarming of the QinghaiminusTibet Plateau51 Drought conditionsmay continue over a long period of time Therefore long-termand advanced ecological restoration and monitoring efforts are

Figure 5 Regional climate feedback loops of the degraded andrestored scenarios of typical land use landscapes for the study area(adapted with permission from ref 49 Copyright 1984 Elsevier)

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

F

essential Although previous climate change studies havepresented global climate models and have discussed energyuse strategies that mitigate global warming few works haveexamined regional THC processes to assess or manage regionalclimatic and hydrological changes Few climate policy studieshave determined the net impact of biophysical changesresulting from land use pattern changes33 Our work in thesource region of the Yellow River serves as an initial attempt toexplore the feasibility of applying innovative methodologies toaddress gaps in knowledge and to evaluate ecologicalrestoration strategies in future studies

ASSOCIATED CONTENTS Supporting InformationClassification criterions of the study area and the accuracyassessment for the land cover classification detailed methodsfor the calculation models of ATCOR2 and validation of theestimated results with meteorological data The SupportingInformation is available free of charge on the ACS Publicationswebsite at DOI 101021es505985q

AUTHOR INFORMATIONCorresponding AuthorTelephone +86-10-6279-4119 E-mail xuehua-hjxmailtsinghuaeducnNotesThe authors declare no competing financial interest

ACKNOWLEDGMENTSThis project was financially supported by the fund from theState Key Joint Laboratory of Environment Simulation andPollution Control China (11Y02ESPCT) and the projectnamed The Relationship of Birds Migration Patterns andHabitat Factors in Poyang Lake (2010CB530300-04) Theauthors thank the anonymous reviewers for their critiqueswhich improved earlier versions of this manuscript

REFERENCES(1) Armesto J J Bautista S Del Val E Ferguson B Garciacutea XGaxiola A Godinez-Alvarez H Gann G Lopez-Barrera FManson R Nu nez-Avila M Ortiz-Arrona C Tognetti PWilliams-Linera G Towards an ecological restoration networkReversing land degradation in Latin America Front Ecol Environ2007 5 (4) w1minusw4 DOI 1018901540-9295(2007)5[w1TAERNR]20CO2(2) Baustian M M Georgia M Dreelin E A Esselman PSchultze S R Qian L Awb T G Luo L Rose J BA Onehundred year review of the socioeconomic and ecological systems ofLake St Clair North America J Great Lakes Res 2014 40 15minus26DOI 101016jjglr201311006(3) Grumbine R E Assessing environmental security in ChinaFront Ecol Environ 2014 12 (7) 403minus411 DOI 101890130147(4) Bullon T Environmental assessment and land change analysis insemi-natural land covers applicable to land management Int J ApplEarth Obs Geoinf 2015 34 147minus156 DOI 101016jjag201408006(5) Liu J Linderman M Ouyang Z An L Yang J Zhang HEcological degradation in protected areas The case of Wolong NatureReserve for giant pandas Science 2001 292 (5514) 98minus101DOI 101126science1058104(6) Poyatos R Latron J Llorens P Land use and land coverchange after agricultural abandonment Mt Res Dev 2003 23 (4)362minus368 DOI 1016590276-4741(2003)023[0362LUALCC]20CO2(7) Feng J Wang T Xie C Eco-environmental degradation in thesource region of the Yellow River northeast QinghaiminusXizang Plateau

Environ Monit Assess 2006 122 (1minus3) 125minus143 DOI 101007s10661-005-9169-2(8) Wang G Cheng G Eco-environmental changes and causativeanalysis in the source regions of the Yangtze and Yellow Rivers ChinaEnviron Syst Decis 2000 20 (3) 221minus232 DOI 101023A1006703831018(9) Pan J Liu J Yellow River source area of land use and landscapepattern change and its ecological effects J Arid Land Res Environ2005 19 (4) 69minus74 (in Chinese) DOI 103969jissn1003-7578200504014(10) Xu J Song L Zhao Z Hu Y Liu C Monitoring grasslanddegradation dynamically at Maduo County in source region of YellowRiver in past 15 years based on remote sensing Arid Land Geogr 201235 (4) 615minus622 (in Chinese) DOI 1013826jcnkicn65-1103x201204018(11) Yang J Ding Y Chen R Spatial and temporal of variations ofalpine vegetation cover in the source regions of the Yangtze andYellow Rivers of the Tibetan Plateau from 1982 to 2001 Environ Geol2006 50 (3) 313minus322 DOI 101007s00254-006-0210-8(12) Davidson A D Detling J K Brown J H Ecological roles andconservation challenges of social burrowing herbivorous mammals inthe worldrsquos grasslands Front Ecol Environ 2012 10 (9) 477minus486DOI 101890110054(13) Qu J Li W Yang M Ji W Zhang Y Life history of theplateau pika (Ochotona curzoniae) in alpine meadows of the TibetanPlateau Mamm Biol 2013 78 (1) 68minus72 DOI 101016jmambio201209005(14) Liu J Nie H Studies on the population productivity ecology ofplateau pika III Trend of population dynamics in plateau pika withdensity-independent and density-dependent vital rates Acta TheriolSin 1992 12 (2) 139minus146 (in Chinese)(15) Liu W Zhou L Wang X The study of different grazingintensity on the role of plants and rodent studies Acta Ecol Sin 199919 (3) 88minus94 (in Chinese)(16) Bai W Zhang Y Xie G Shen Z Causes analysis of grasslanddegradation Maduo County in Yellow River source region Acta EcolSin 2002 13 (7) 823minus826 (in Chinese) DOI 1013287j1001-933220020194(17) Fang Y Qin D Ding Y Frozen soil change and adaptation ofanimal husbandry A case of the source regions of Yangtze and YellowRivers Environ Sci Policy 2011 14 (5) 555minus568 DOI 101016jenvsci201103012(18) Verburg P H van de Steeg J Veldkamp A Willemen LFrom land cover change to land function dynamics A major challengeto improve land characterization J Environ Manage 2009 90 (3)1327minus1335 DOI 101016jjenvman200808005(19) Barnett T P Adam J C Lettenmaier A D P Potentialimpacts of a warming climate on water availability in snow-dominatedregions Nature 2005 438 303minus309 DOI 101038nature04141(20) Chattopadhyay N Hulme M Evaporation and potentialevapotranspiration in India under conditions of recent and futureclimate change Agric For Meteorol 1997 87 55minus73 DOI 101016S0168-1923(97)00006-3(21) Liu M Xia Z Han S Wang X Tang Z Relationshipbetween variation of evapotranspiration and ecological deterioration insource region of Yellow River J Hohai Uni 2009 37 (6) 631minus634 (inChinese)(22) Yi X Yin Y Li G Peng J Temperature variation in recent50 years in the three-river headwaters region of Qinghai Province ActaGeol Sin 2011 66 (11) 1451minus1465 (in Chinese)(23) Hou W Li Y Soil surface humidity index and sensitivityanalysis of the climate factors that affect it in the Yellow River sourceregions J Glaciol Geocryol 2010 32 (6) 1226minus1233 (in Chinese)(24) Mauser W Schadlich S Modelling the spatial distribution ofevapotranspiration on different scales using remote sensing data JHydrol 1998 212 250minus267 DOI 101016S0022-1694(98)00228-5(25) Salati E Vose P B Amazon basin A system in equilibriumScience 1984 225 (4658) 129minus138 DOI 1023071693078

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

G

(26) Schlesinger W H Reynolds J F Cunningham G LHuenneke L F Jarrell W M Virginia R A Whitford W GBiological feedbacks in global desertification Science 1990 247 (4946)1043minus1048 DOI 101126science24749461043(27) Shukla J Nobre C Sellers P Amazon deforestation andclimate change Science 1990 247 (4948) 1322minus1325 DOI 101126science24749481322(28) Baron J S Hartman M D Kittel T G F Band L E OjimaD S Lammers R B Effects of land cover water redistribution andtemperature on ecosystem processes in the South Platte basin EcolAppl 1998 8 (4) 1037minus1051 DOI 1018901051-0761(1998)008[1037eolcwr]20co2(29) Boegh E Soegaard H Thomsen A Evaluating evapotranspi-ration rates and surface conditions using Landsat TM to estimateatmospheric resistance and surface resistance Remote Sens Environ2002 79 (2minus3) 329minus343 DOI 101016S0034-4257(01)00283-8(30) Sterling S M Ducharne A Polcher J The impact of globalland-cover change on the terrestrial water cycle Nat Clim Change2013 3 (4) 385minus390 DOI 101038nclimate1690(31) Loarie S R Lobell D B Asner G P Mu Q Field C BDirect impacts on local climate of sugar-cane expansion in Brazil NatClim Change 2011 1 (2) 105minus109 DOI 101038nclimate1067(32) Teuling A J Seneviratne S I Stockli R et al Contrastingresponse of European forest and grassland energy exchange toheatwaves Nat Geosci 2010 3 (10) 722minus727 DOI 101038ngeo950(33) Anderson R G Canadell J G Randerson J T et alBiophysical considerations in forestry for climate protection FrontEcol Environ 2011 9 (3) 174minus182 DOI 101890090179(34) Campo-Bescos M A Munoz-Carpena R Kaplan D ASouthworth J Zhu L Waylen P R Beyond precipitationPhysiographic gradients dictate the relative importance of environ-mental drivers on savanna vegetation PLoS One 2013 8 (8)No e72348 DOI 101371journalpone0072348(35) Schneider K Mauser W Processing and accuracy of Landsatthematic mapper data for lake surface temperature measurement Int JRemote Sens 1996 17 (11) 2027minus2041 DOI 10108001431169608948757(36) Li F Jackson T J Kustas W P Schmugge T J French AN Cosh M H Bindlish R Deriving land surface temperature fromLandsat 5 and 7 during SMEX02SMACEX Remote Sens Environ2004 92 (4) 521minus534 DOI 101016jrse200402018(37) Bastiaanssen W G M Menenti M Feddes R A Holtslag AA M A remote sensing surface energy balance algorithm for land(SEBAL) J Hydrol 1998 212 198minus212 DOI 101016S0022-1694(98)00253-4(38) Bonan G B Forests and climate change Forcings feedbacksand the climate benefits of forests Science 2008 320 (5882) 1444minus1449 DOI 101126science1155121(39) Li Z Liu X Ma T Kejia D Zhou Q Yao B Niu TRetrieval of the surface evapotranspiration patterns in the alpinegrasslandminuswetland ecosystem applying SEBAL model in the sourceregion of the Yellow River China Ecol Modell 2013 270 64minus75DOI 101016jecolmodel201309004(40) Zhao L Lee X Smith R B Oleson K Strong contributionsof local background climate to urban heat islands Nature 2014 511(7508) 216minus219 DOI 101038nature13462(41) van Heerwaarden C C de Arellano J V Gounou AGuichard F Couvreux F Understanding the daily cycle ofevapotranspiration A method to quantify the influence of forcingsand feedbacks J Hydrometeorol 2010 11 (6) 1405minus1422DOI 1011752010jhm12721(42) Chang G G Li F X Li L Changes and Restoration in SourceRegion of Three Rivers of China Metrology Press Beijing China 2010(in Chinese)(43) Tong L Xu X Fu Y Li S Wetland changes and theirresponses to climate change in the three-river headwaters region ofChina since the 1990s Energies 2014 7 (4) 2515minus2534DOI 103390en7042515

(44) Zhang Y Zhao X Zhao S Feng C Correlation betweenevapotranspiration and climate factors in warm steppe in source regionof Yangtze Yellow and Yalu Tsangpo Rivers J Desert Res 2010 30(2) 363minus368 (in Chinese)(45) Jiang Y Li D Variations of Tangnaihai runoff andprecipitation and temperature in the upper reach of the YellowRiverMeteorol Disaster Reduct Res 2011 34 (2) 51minus57 (in Chinese)(46) Molders N Kramm G Influence of wildfire induced land-cover changes on clouds and precipitation in interior AlaskaA casestudy Atmos Res 2007 84 (2) 142minus168 DOI 101016jatmos-res200606004(47) Obregon A Gehrig-Downie C Gradstein S R RollenbeckR Bendix J Canopy level fog occurrence in a tropical lowland forestof French Guiana as a prerequisite for high epiphyte diversity AgricFor Meteorol 2011 151 (3) 290minus300 DOI 101016jagrfor-met201011003(48) Harding R J Blyth E M Tuinenburg O A Wiltshire ALand atmosphere feedbacks and their role in the water resources of theGanges basin Sci Total Environ 2013 468 (S1) S85minusS92DOI 101016jscitotenv201303016(49) Bradshaw A D Ecological principles and land reclamationpractice Landscape Plann 1984 11 (1) 35minus48(50) Shelton M L Hydroclimatology Perspectives and ApplicationsCambridge Press Cambridge UK 2009(51) Liu Y Lu S Li S Gao Y Numerical simulation of impact ofland surface changes on regional climatic environment in sourceregions of Yellow River Plateau Meteorol 2009 28 (2) 327minus334 (inChinese)

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

H

anticipated to reveal interactions between surface structures andregional hydro-heat processes

RESULTS

Grassland Degradation and Recovery Regional landcover changes showed patterns of grassland degradation andrecovery in the summers of 1990minus2009 (Figure 2) Theproportion of high-coverage grassland expanded gradually fromapproximately 27 in 1990 to greater than 30 in 2009 Low-coverage grassland areas increased rapidly after 2000accounting for the largest proportion of land (39) Inaddition proportions of the other two land cover types middle-coverage grassland and bare soils decreased from 21 to 12and from 26 to 11 respectively Water body coverage levelsremained constant at a level of 8 for the entire period Thesetrends denote land cover changes after approximately 2000Hydro-heat Patterns and Regional Climate Change Ts

presents a spatial pattern with higher values of approximately2975 K in the north and lower values of approximately 2932 Kin the south (Figure 3a) For different land covers Ts values aresorted in ascending order (high- middle- and low-coveragegrasslands and bare soils) In the study area average summer airtemperatures increased by almost 2 degC over the past 20 years(Figure 3a) In general Ts values of the study area for thesummers of 1990minus2009 follow observed regional warmingtrends We also found that the Ts values of bare soil and low-and middle-coverage grassland areas appeared to decrease from2006 to 2009 (Figure 4)ET levels were lower at approximately 085 cmday in the

north and higher at approximately 175 cmday in the south(Figure 3b) ET values were highest in the water bodies andareas surrounding wetlands also exhibited higher ET values ofapproximately 155 cmday The ET values for the other typesof land cover were ranked in the following ascending order

bare soils and low- middle- and high-coverage grasslandsSandy riverbank areas generated very low ET values ofapproximately 037 cmday The ET value for all of the landcover types initially declined from 1990 to 2000 and thenincreased to 1990 levels In grasslands and bare soils ET levelsreached approximately 15 cmday in 1990 and decreased toapproximately 10 cmday in approximately 2000 denoting anET turning point Thereafter grassland and bare soil ET levelsreturned to approximately 15 cmday in 2006 and 2009(Figures 3b and 4) Meteorological records for the summers of1990minus2009 show that the water balance level reached its lowestvalue in 1996 and the region has since returned to highermoisture levels

Correlation between Key Variables and α and NDVITrends The correlation analysis results show how associationsbetween the key variables form the regional climate feedbackloop (Figure 4) Pearson linear coefficients are underlined andadded to each pair In 1990 1994 and 2000 all pairs weresignificantly correlated with the exception of the αminusTs pairThe scatter graphs show that NDVI was negatively related to αTs was negatively related to ET and ET was positively relatedto NDVI although the relationship between α and Ts remainsunclear In 1990 the αminusTs pair exhibited an insignificantlynegative relation whereas the 2006 and 2009 results generallyshowed a positive numerical relationship between the twovariables Apart from the hydro-heat pattern changes of Ts andET noted above α and NDVI trends were extracted on thebasis of different land cover types revealing that α reached amaximum value in 1994 and that NDVI initially declined butincreased after 1994 denoting the start of the restoration phase

DISCUSSION

Consequences of Ecosystem Degradation As an inlandplateau grasslandminuswetland ecosystem changes in the land

Figure 4 Correlation analysis pairs of NDVIminusα αminusTs TsminusET and ETminusNDVI and change trends of the key variables for 1990minus2009 Numbersmarked in each box denote the Pearson correlation coefficients for each pair of variables lowast and lowastlowast denote the significance levels of 005 and 001respectively

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

E

cover structure in this study area are related to the temperatureand hydrologic patterns of this region From 1990 to 2000 asgrassland areas degraded while bare soils expanded albedolevels of the land surface reached peak level in 1994 At thesame time surface temperatures increased continuously partlybecause of global warming but also because of a reduction inevaporative cooling40 On the basis of Ts spatial patterns thesurface heating effects on bare soils of the northern area becamemore pronounced reflecting patterns similar to those of theurban heat island effect40 This regional surface temperaturechange is attributable to several factors such as elevationradiation surface reflectivity precipitation and evaporation4041

However quantitative attributes of these factors were notexamined in this study The correlation analysis of the keyvariables showed an association between vegetation and hydro-heat environments of the study area based on the retrievedspatial data although these data are too limited to fully reflectcausality in the ecological degradation and restoration feedbackloop Temporal dimension variable change trends show that theα and NDVI turning point occurred in 1994 and that ET hasincreased since 2000 and this could be interpreted as a delayedresponse to improved vegetation conditionsEcosystem Restoration Effects According to references

and local policy documents humans began to play a positiverole in the regional feedback process and attempted to controldegradation trends in the study area beginning in 19967842

The government planned ecological conservation and restora-tion projects and has implemented these projects from 1996 tothe present These projects have aimed to improve primaryproductivity levels and ecosystem self-control capacities tomaintain healthy grasslands in the source region Theseimproved vegetation conditions may stabilize surface temper-atures and increase soil moisture and ET levels as shown inrelevant observations Runoff levels in the area have increasedsince 2000 and this trend has been significantly correlated withprecipitation patterns43minus45 In the present study comparisonsbetween annual water balance and runoff variations show adecline in ET until 2000 (Figure 3b) suggesting that runoff andET patterns interacted closely in the hydrologic cycleRegional Climate Feedback Mechanisms of the

Terrestrial Ecosystem Previous studies on interactionsbetween the land surface and hydroclimatological conditionshave mainly employed methods and tools such as satellitecloud images cloud frequency analyses canopy microclimatol-ogy sensors and meteorological observations to examine anddefine cloud formation feedback effects46minus48 Most of thesestudies examined forest ecosystems with less of a focus onalpine grassland ecosystems Still micro-to-macro processanalyses presented in previous studies suggest the presence ofregional climate feedback mechanisms ET improvements viavegetation recovery helped increase humidity levels in turnincreasing boundary convection and local precipitationlevels4748

In the neighboring QinghaiminusTibet Plateau source regionwhich is similar to the study area recent observations andstudies have shown a positive correlation between ET andhumidity as well as a general increase in precipitation andrunoff levels over the last 5minus10 years4445 Although causalinferences between restored vegetation levels and improvedhydrological conditions in the study area remain unclear basedon existing results we expect to reveal such mechanisms ingreater depth with the use of more observations andexperiments on alpine grassland ecosystems We conclude

that lower ET trends led to lower air moisture levels inhibitingcloud formation and local precipitation It is assumed thatdrought conditions inhibit vegetation growth potentiallyaccelerating grassland degradation further as shown in Figure5

Coupling System of the Atmospheric and TerrestrialHydrocycle According to principles of modern hydro-climatology the atmosphere serves as a climatic frame anddrives energy flows and water vapor transportation patternsSubsystems of the atmosphere and terrestrial ecosystems werecoupled to form a complete hydrocycle system Dividing thehydrocycle system into two branches [the atmospherichydrocycle (AHC) and terrestrial hydrocycle (THC)] helpedreveal global climate patterns50

The dominant factors of AHC mainly include atmosphericcirculation and energy flux trends in patterns of circulationAtmospheric circulation models primarily determine biomes atthe continental scale and strongly affect the construction andsuccession of ecosystems influencing global land coverstructures THC involves terrestrialminusatmosphere interfaceprocesses wherein the atmosphere interacts with geographicalhydrological and biological processes driven by key factors tosupport the development of ecosystems as shown in our studyKey factors include surface energy flux water availability andgroundwater balance ET plays a dominant role in THCbecause ET is driven by radiation and relies on thetransformation of groundwater into atmospheric water vaporwhich is widely related to soil and vegetation processes As partof the land cover structure vegetation determines the speedand means through which groundwater and precipitation movefrom the land surface to the atmosphere In turn studies haveshown that THC plays a key role in climate systems through itsinvolvement in the hydrologic cycleTo situate this analysis of long-term AHC impacts within

processes of ecological restoration it is also essential tounderstand this phenomenon and to make predictionsObservations show that the Asian monsoon has been lessactive over the last 20 years resulting in the drying andwarming of the QinghaiminusTibet Plateau51 Drought conditionsmay continue over a long period of time Therefore long-termand advanced ecological restoration and monitoring efforts are

Figure 5 Regional climate feedback loops of the degraded andrestored scenarios of typical land use landscapes for the study area(adapted with permission from ref 49 Copyright 1984 Elsevier)

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

F

essential Although previous climate change studies havepresented global climate models and have discussed energyuse strategies that mitigate global warming few works haveexamined regional THC processes to assess or manage regionalclimatic and hydrological changes Few climate policy studieshave determined the net impact of biophysical changesresulting from land use pattern changes33 Our work in thesource region of the Yellow River serves as an initial attempt toexplore the feasibility of applying innovative methodologies toaddress gaps in knowledge and to evaluate ecologicalrestoration strategies in future studies

ASSOCIATED CONTENTS Supporting InformationClassification criterions of the study area and the accuracyassessment for the land cover classification detailed methodsfor the calculation models of ATCOR2 and validation of theestimated results with meteorological data The SupportingInformation is available free of charge on the ACS Publicationswebsite at DOI 101021es505985q

AUTHOR INFORMATIONCorresponding AuthorTelephone +86-10-6279-4119 E-mail xuehua-hjxmailtsinghuaeducnNotesThe authors declare no competing financial interest

ACKNOWLEDGMENTSThis project was financially supported by the fund from theState Key Joint Laboratory of Environment Simulation andPollution Control China (11Y02ESPCT) and the projectnamed The Relationship of Birds Migration Patterns andHabitat Factors in Poyang Lake (2010CB530300-04) Theauthors thank the anonymous reviewers for their critiqueswhich improved earlier versions of this manuscript

REFERENCES(1) Armesto J J Bautista S Del Val E Ferguson B Garciacutea XGaxiola A Godinez-Alvarez H Gann G Lopez-Barrera FManson R Nu nez-Avila M Ortiz-Arrona C Tognetti PWilliams-Linera G Towards an ecological restoration networkReversing land degradation in Latin America Front Ecol Environ2007 5 (4) w1minusw4 DOI 1018901540-9295(2007)5[w1TAERNR]20CO2(2) Baustian M M Georgia M Dreelin E A Esselman PSchultze S R Qian L Awb T G Luo L Rose J BA Onehundred year review of the socioeconomic and ecological systems ofLake St Clair North America J Great Lakes Res 2014 40 15minus26DOI 101016jjglr201311006(3) Grumbine R E Assessing environmental security in ChinaFront Ecol Environ 2014 12 (7) 403minus411 DOI 101890130147(4) Bullon T Environmental assessment and land change analysis insemi-natural land covers applicable to land management Int J ApplEarth Obs Geoinf 2015 34 147minus156 DOI 101016jjag201408006(5) Liu J Linderman M Ouyang Z An L Yang J Zhang HEcological degradation in protected areas The case of Wolong NatureReserve for giant pandas Science 2001 292 (5514) 98minus101DOI 101126science1058104(6) Poyatos R Latron J Llorens P Land use and land coverchange after agricultural abandonment Mt Res Dev 2003 23 (4)362minus368 DOI 1016590276-4741(2003)023[0362LUALCC]20CO2(7) Feng J Wang T Xie C Eco-environmental degradation in thesource region of the Yellow River northeast QinghaiminusXizang Plateau

Environ Monit Assess 2006 122 (1minus3) 125minus143 DOI 101007s10661-005-9169-2(8) Wang G Cheng G Eco-environmental changes and causativeanalysis in the source regions of the Yangtze and Yellow Rivers ChinaEnviron Syst Decis 2000 20 (3) 221minus232 DOI 101023A1006703831018(9) Pan J Liu J Yellow River source area of land use and landscapepattern change and its ecological effects J Arid Land Res Environ2005 19 (4) 69minus74 (in Chinese) DOI 103969jissn1003-7578200504014(10) Xu J Song L Zhao Z Hu Y Liu C Monitoring grasslanddegradation dynamically at Maduo County in source region of YellowRiver in past 15 years based on remote sensing Arid Land Geogr 201235 (4) 615minus622 (in Chinese) DOI 1013826jcnkicn65-1103x201204018(11) Yang J Ding Y Chen R Spatial and temporal of variations ofalpine vegetation cover in the source regions of the Yangtze andYellow Rivers of the Tibetan Plateau from 1982 to 2001 Environ Geol2006 50 (3) 313minus322 DOI 101007s00254-006-0210-8(12) Davidson A D Detling J K Brown J H Ecological roles andconservation challenges of social burrowing herbivorous mammals inthe worldrsquos grasslands Front Ecol Environ 2012 10 (9) 477minus486DOI 101890110054(13) Qu J Li W Yang M Ji W Zhang Y Life history of theplateau pika (Ochotona curzoniae) in alpine meadows of the TibetanPlateau Mamm Biol 2013 78 (1) 68minus72 DOI 101016jmambio201209005(14) Liu J Nie H Studies on the population productivity ecology ofplateau pika III Trend of population dynamics in plateau pika withdensity-independent and density-dependent vital rates Acta TheriolSin 1992 12 (2) 139minus146 (in Chinese)(15) Liu W Zhou L Wang X The study of different grazingintensity on the role of plants and rodent studies Acta Ecol Sin 199919 (3) 88minus94 (in Chinese)(16) Bai W Zhang Y Xie G Shen Z Causes analysis of grasslanddegradation Maduo County in Yellow River source region Acta EcolSin 2002 13 (7) 823minus826 (in Chinese) DOI 1013287j1001-933220020194(17) Fang Y Qin D Ding Y Frozen soil change and adaptation ofanimal husbandry A case of the source regions of Yangtze and YellowRivers Environ Sci Policy 2011 14 (5) 555minus568 DOI 101016jenvsci201103012(18) Verburg P H van de Steeg J Veldkamp A Willemen LFrom land cover change to land function dynamics A major challengeto improve land characterization J Environ Manage 2009 90 (3)1327minus1335 DOI 101016jjenvman200808005(19) Barnett T P Adam J C Lettenmaier A D P Potentialimpacts of a warming climate on water availability in snow-dominatedregions Nature 2005 438 303minus309 DOI 101038nature04141(20) Chattopadhyay N Hulme M Evaporation and potentialevapotranspiration in India under conditions of recent and futureclimate change Agric For Meteorol 1997 87 55minus73 DOI 101016S0168-1923(97)00006-3(21) Liu M Xia Z Han S Wang X Tang Z Relationshipbetween variation of evapotranspiration and ecological deterioration insource region of Yellow River J Hohai Uni 2009 37 (6) 631minus634 (inChinese)(22) Yi X Yin Y Li G Peng J Temperature variation in recent50 years in the three-river headwaters region of Qinghai Province ActaGeol Sin 2011 66 (11) 1451minus1465 (in Chinese)(23) Hou W Li Y Soil surface humidity index and sensitivityanalysis of the climate factors that affect it in the Yellow River sourceregions J Glaciol Geocryol 2010 32 (6) 1226minus1233 (in Chinese)(24) Mauser W Schadlich S Modelling the spatial distribution ofevapotranspiration on different scales using remote sensing data JHydrol 1998 212 250minus267 DOI 101016S0022-1694(98)00228-5(25) Salati E Vose P B Amazon basin A system in equilibriumScience 1984 225 (4658) 129minus138 DOI 1023071693078

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

G

(26) Schlesinger W H Reynolds J F Cunningham G LHuenneke L F Jarrell W M Virginia R A Whitford W GBiological feedbacks in global desertification Science 1990 247 (4946)1043minus1048 DOI 101126science24749461043(27) Shukla J Nobre C Sellers P Amazon deforestation andclimate change Science 1990 247 (4948) 1322minus1325 DOI 101126science24749481322(28) Baron J S Hartman M D Kittel T G F Band L E OjimaD S Lammers R B Effects of land cover water redistribution andtemperature on ecosystem processes in the South Platte basin EcolAppl 1998 8 (4) 1037minus1051 DOI 1018901051-0761(1998)008[1037eolcwr]20co2(29) Boegh E Soegaard H Thomsen A Evaluating evapotranspi-ration rates and surface conditions using Landsat TM to estimateatmospheric resistance and surface resistance Remote Sens Environ2002 79 (2minus3) 329minus343 DOI 101016S0034-4257(01)00283-8(30) Sterling S M Ducharne A Polcher J The impact of globalland-cover change on the terrestrial water cycle Nat Clim Change2013 3 (4) 385minus390 DOI 101038nclimate1690(31) Loarie S R Lobell D B Asner G P Mu Q Field C BDirect impacts on local climate of sugar-cane expansion in Brazil NatClim Change 2011 1 (2) 105minus109 DOI 101038nclimate1067(32) Teuling A J Seneviratne S I Stockli R et al Contrastingresponse of European forest and grassland energy exchange toheatwaves Nat Geosci 2010 3 (10) 722minus727 DOI 101038ngeo950(33) Anderson R G Canadell J G Randerson J T et alBiophysical considerations in forestry for climate protection FrontEcol Environ 2011 9 (3) 174minus182 DOI 101890090179(34) Campo-Bescos M A Munoz-Carpena R Kaplan D ASouthworth J Zhu L Waylen P R Beyond precipitationPhysiographic gradients dictate the relative importance of environ-mental drivers on savanna vegetation PLoS One 2013 8 (8)No e72348 DOI 101371journalpone0072348(35) Schneider K Mauser W Processing and accuracy of Landsatthematic mapper data for lake surface temperature measurement Int JRemote Sens 1996 17 (11) 2027minus2041 DOI 10108001431169608948757(36) Li F Jackson T J Kustas W P Schmugge T J French AN Cosh M H Bindlish R Deriving land surface temperature fromLandsat 5 and 7 during SMEX02SMACEX Remote Sens Environ2004 92 (4) 521minus534 DOI 101016jrse200402018(37) Bastiaanssen W G M Menenti M Feddes R A Holtslag AA M A remote sensing surface energy balance algorithm for land(SEBAL) J Hydrol 1998 212 198minus212 DOI 101016S0022-1694(98)00253-4(38) Bonan G B Forests and climate change Forcings feedbacksand the climate benefits of forests Science 2008 320 (5882) 1444minus1449 DOI 101126science1155121(39) Li Z Liu X Ma T Kejia D Zhou Q Yao B Niu TRetrieval of the surface evapotranspiration patterns in the alpinegrasslandminuswetland ecosystem applying SEBAL model in the sourceregion of the Yellow River China Ecol Modell 2013 270 64minus75DOI 101016jecolmodel201309004(40) Zhao L Lee X Smith R B Oleson K Strong contributionsof local background climate to urban heat islands Nature 2014 511(7508) 216minus219 DOI 101038nature13462(41) van Heerwaarden C C de Arellano J V Gounou AGuichard F Couvreux F Understanding the daily cycle ofevapotranspiration A method to quantify the influence of forcingsand feedbacks J Hydrometeorol 2010 11 (6) 1405minus1422DOI 1011752010jhm12721(42) Chang G G Li F X Li L Changes and Restoration in SourceRegion of Three Rivers of China Metrology Press Beijing China 2010(in Chinese)(43) Tong L Xu X Fu Y Li S Wetland changes and theirresponses to climate change in the three-river headwaters region ofChina since the 1990s Energies 2014 7 (4) 2515minus2534DOI 103390en7042515

(44) Zhang Y Zhao X Zhao S Feng C Correlation betweenevapotranspiration and climate factors in warm steppe in source regionof Yangtze Yellow and Yalu Tsangpo Rivers J Desert Res 2010 30(2) 363minus368 (in Chinese)(45) Jiang Y Li D Variations of Tangnaihai runoff andprecipitation and temperature in the upper reach of the YellowRiverMeteorol Disaster Reduct Res 2011 34 (2) 51minus57 (in Chinese)(46) Molders N Kramm G Influence of wildfire induced land-cover changes on clouds and precipitation in interior AlaskaA casestudy Atmos Res 2007 84 (2) 142minus168 DOI 101016jatmos-res200606004(47) Obregon A Gehrig-Downie C Gradstein S R RollenbeckR Bendix J Canopy level fog occurrence in a tropical lowland forestof French Guiana as a prerequisite for high epiphyte diversity AgricFor Meteorol 2011 151 (3) 290minus300 DOI 101016jagrfor-met201011003(48) Harding R J Blyth E M Tuinenburg O A Wiltshire ALand atmosphere feedbacks and their role in the water resources of theGanges basin Sci Total Environ 2013 468 (S1) S85minusS92DOI 101016jscitotenv201303016(49) Bradshaw A D Ecological principles and land reclamationpractice Landscape Plann 1984 11 (1) 35minus48(50) Shelton M L Hydroclimatology Perspectives and ApplicationsCambridge Press Cambridge UK 2009(51) Liu Y Lu S Li S Gao Y Numerical simulation of impact ofland surface changes on regional climatic environment in sourceregions of Yellow River Plateau Meteorol 2009 28 (2) 327minus334 (inChinese)

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

H

cover structure in this study area are related to the temperatureand hydrologic patterns of this region From 1990 to 2000 asgrassland areas degraded while bare soils expanded albedolevels of the land surface reached peak level in 1994 At thesame time surface temperatures increased continuously partlybecause of global warming but also because of a reduction inevaporative cooling40 On the basis of Ts spatial patterns thesurface heating effects on bare soils of the northern area becamemore pronounced reflecting patterns similar to those of theurban heat island effect40 This regional surface temperaturechange is attributable to several factors such as elevationradiation surface reflectivity precipitation and evaporation4041

However quantitative attributes of these factors were notexamined in this study The correlation analysis of the keyvariables showed an association between vegetation and hydro-heat environments of the study area based on the retrievedspatial data although these data are too limited to fully reflectcausality in the ecological degradation and restoration feedbackloop Temporal dimension variable change trends show that theα and NDVI turning point occurred in 1994 and that ET hasincreased since 2000 and this could be interpreted as a delayedresponse to improved vegetation conditionsEcosystem Restoration Effects According to references

and local policy documents humans began to play a positiverole in the regional feedback process and attempted to controldegradation trends in the study area beginning in 19967842

The government planned ecological conservation and restora-tion projects and has implemented these projects from 1996 tothe present These projects have aimed to improve primaryproductivity levels and ecosystem self-control capacities tomaintain healthy grasslands in the source region Theseimproved vegetation conditions may stabilize surface temper-atures and increase soil moisture and ET levels as shown inrelevant observations Runoff levels in the area have increasedsince 2000 and this trend has been significantly correlated withprecipitation patterns43minus45 In the present study comparisonsbetween annual water balance and runoff variations show adecline in ET until 2000 (Figure 3b) suggesting that runoff andET patterns interacted closely in the hydrologic cycleRegional Climate Feedback Mechanisms of the

Terrestrial Ecosystem Previous studies on interactionsbetween the land surface and hydroclimatological conditionshave mainly employed methods and tools such as satellitecloud images cloud frequency analyses canopy microclimatol-ogy sensors and meteorological observations to examine anddefine cloud formation feedback effects46minus48 Most of thesestudies examined forest ecosystems with less of a focus onalpine grassland ecosystems Still micro-to-macro processanalyses presented in previous studies suggest the presence ofregional climate feedback mechanisms ET improvements viavegetation recovery helped increase humidity levels in turnincreasing boundary convection and local precipitationlevels4748

In the neighboring QinghaiminusTibet Plateau source regionwhich is similar to the study area recent observations andstudies have shown a positive correlation between ET andhumidity as well as a general increase in precipitation andrunoff levels over the last 5minus10 years4445 Although causalinferences between restored vegetation levels and improvedhydrological conditions in the study area remain unclear basedon existing results we expect to reveal such mechanisms ingreater depth with the use of more observations andexperiments on alpine grassland ecosystems We conclude

that lower ET trends led to lower air moisture levels inhibitingcloud formation and local precipitation It is assumed thatdrought conditions inhibit vegetation growth potentiallyaccelerating grassland degradation further as shown in Figure5

Coupling System of the Atmospheric and TerrestrialHydrocycle According to principles of modern hydro-climatology the atmosphere serves as a climatic frame anddrives energy flows and water vapor transportation patternsSubsystems of the atmosphere and terrestrial ecosystems werecoupled to form a complete hydrocycle system Dividing thehydrocycle system into two branches [the atmospherichydrocycle (AHC) and terrestrial hydrocycle (THC)] helpedreveal global climate patterns50

The dominant factors of AHC mainly include atmosphericcirculation and energy flux trends in patterns of circulationAtmospheric circulation models primarily determine biomes atthe continental scale and strongly affect the construction andsuccession of ecosystems influencing global land coverstructures THC involves terrestrialminusatmosphere interfaceprocesses wherein the atmosphere interacts with geographicalhydrological and biological processes driven by key factors tosupport the development of ecosystems as shown in our studyKey factors include surface energy flux water availability andgroundwater balance ET plays a dominant role in THCbecause ET is driven by radiation and relies on thetransformation of groundwater into atmospheric water vaporwhich is widely related to soil and vegetation processes As partof the land cover structure vegetation determines the speedand means through which groundwater and precipitation movefrom the land surface to the atmosphere In turn studies haveshown that THC plays a key role in climate systems through itsinvolvement in the hydrologic cycleTo situate this analysis of long-term AHC impacts within

processes of ecological restoration it is also essential tounderstand this phenomenon and to make predictionsObservations show that the Asian monsoon has been lessactive over the last 20 years resulting in the drying andwarming of the QinghaiminusTibet Plateau51 Drought conditionsmay continue over a long period of time Therefore long-termand advanced ecological restoration and monitoring efforts are

Figure 5 Regional climate feedback loops of the degraded andrestored scenarios of typical land use landscapes for the study area(adapted with permission from ref 49 Copyright 1984 Elsevier)

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

F

essential Although previous climate change studies havepresented global climate models and have discussed energyuse strategies that mitigate global warming few works haveexamined regional THC processes to assess or manage regionalclimatic and hydrological changes Few climate policy studieshave determined the net impact of biophysical changesresulting from land use pattern changes33 Our work in thesource region of the Yellow River serves as an initial attempt toexplore the feasibility of applying innovative methodologies toaddress gaps in knowledge and to evaluate ecologicalrestoration strategies in future studies

ASSOCIATED CONTENTS Supporting InformationClassification criterions of the study area and the accuracyassessment for the land cover classification detailed methodsfor the calculation models of ATCOR2 and validation of theestimated results with meteorological data The SupportingInformation is available free of charge on the ACS Publicationswebsite at DOI 101021es505985q

AUTHOR INFORMATIONCorresponding AuthorTelephone +86-10-6279-4119 E-mail xuehua-hjxmailtsinghuaeducnNotesThe authors declare no competing financial interest

ACKNOWLEDGMENTSThis project was financially supported by the fund from theState Key Joint Laboratory of Environment Simulation andPollution Control China (11Y02ESPCT) and the projectnamed The Relationship of Birds Migration Patterns andHabitat Factors in Poyang Lake (2010CB530300-04) Theauthors thank the anonymous reviewers for their critiqueswhich improved earlier versions of this manuscript

REFERENCES(1) Armesto J J Bautista S Del Val E Ferguson B Garciacutea XGaxiola A Godinez-Alvarez H Gann G Lopez-Barrera FManson R Nu nez-Avila M Ortiz-Arrona C Tognetti PWilliams-Linera G Towards an ecological restoration networkReversing land degradation in Latin America Front Ecol Environ2007 5 (4) w1minusw4 DOI 1018901540-9295(2007)5[w1TAERNR]20CO2(2) Baustian M M Georgia M Dreelin E A Esselman PSchultze S R Qian L Awb T G Luo L Rose J BA Onehundred year review of the socioeconomic and ecological systems ofLake St Clair North America J Great Lakes Res 2014 40 15minus26DOI 101016jjglr201311006(3) Grumbine R E Assessing environmental security in ChinaFront Ecol Environ 2014 12 (7) 403minus411 DOI 101890130147(4) Bullon T Environmental assessment and land change analysis insemi-natural land covers applicable to land management Int J ApplEarth Obs Geoinf 2015 34 147minus156 DOI 101016jjag201408006(5) Liu J Linderman M Ouyang Z An L Yang J Zhang HEcological degradation in protected areas The case of Wolong NatureReserve for giant pandas Science 2001 292 (5514) 98minus101DOI 101126science1058104(6) Poyatos R Latron J Llorens P Land use and land coverchange after agricultural abandonment Mt Res Dev 2003 23 (4)362minus368 DOI 1016590276-4741(2003)023[0362LUALCC]20CO2(7) Feng J Wang T Xie C Eco-environmental degradation in thesource region of the Yellow River northeast QinghaiminusXizang Plateau

Environ Monit Assess 2006 122 (1minus3) 125minus143 DOI 101007s10661-005-9169-2(8) Wang G Cheng G Eco-environmental changes and causativeanalysis in the source regions of the Yangtze and Yellow Rivers ChinaEnviron Syst Decis 2000 20 (3) 221minus232 DOI 101023A1006703831018(9) Pan J Liu J Yellow River source area of land use and landscapepattern change and its ecological effects J Arid Land Res Environ2005 19 (4) 69minus74 (in Chinese) DOI 103969jissn1003-7578200504014(10) Xu J Song L Zhao Z Hu Y Liu C Monitoring grasslanddegradation dynamically at Maduo County in source region of YellowRiver in past 15 years based on remote sensing Arid Land Geogr 201235 (4) 615minus622 (in Chinese) DOI 1013826jcnkicn65-1103x201204018(11) Yang J Ding Y Chen R Spatial and temporal of variations ofalpine vegetation cover in the source regions of the Yangtze andYellow Rivers of the Tibetan Plateau from 1982 to 2001 Environ Geol2006 50 (3) 313minus322 DOI 101007s00254-006-0210-8(12) Davidson A D Detling J K Brown J H Ecological roles andconservation challenges of social burrowing herbivorous mammals inthe worldrsquos grasslands Front Ecol Environ 2012 10 (9) 477minus486DOI 101890110054(13) Qu J Li W Yang M Ji W Zhang Y Life history of theplateau pika (Ochotona curzoniae) in alpine meadows of the TibetanPlateau Mamm Biol 2013 78 (1) 68minus72 DOI 101016jmambio201209005(14) Liu J Nie H Studies on the population productivity ecology ofplateau pika III Trend of population dynamics in plateau pika withdensity-independent and density-dependent vital rates Acta TheriolSin 1992 12 (2) 139minus146 (in Chinese)(15) Liu W Zhou L Wang X The study of different grazingintensity on the role of plants and rodent studies Acta Ecol Sin 199919 (3) 88minus94 (in Chinese)(16) Bai W Zhang Y Xie G Shen Z Causes analysis of grasslanddegradation Maduo County in Yellow River source region Acta EcolSin 2002 13 (7) 823minus826 (in Chinese) DOI 1013287j1001-933220020194(17) Fang Y Qin D Ding Y Frozen soil change and adaptation ofanimal husbandry A case of the source regions of Yangtze and YellowRivers Environ Sci Policy 2011 14 (5) 555minus568 DOI 101016jenvsci201103012(18) Verburg P H van de Steeg J Veldkamp A Willemen LFrom land cover change to land function dynamics A major challengeto improve land characterization J Environ Manage 2009 90 (3)1327minus1335 DOI 101016jjenvman200808005(19) Barnett T P Adam J C Lettenmaier A D P Potentialimpacts of a warming climate on water availability in snow-dominatedregions Nature 2005 438 303minus309 DOI 101038nature04141(20) Chattopadhyay N Hulme M Evaporation and potentialevapotranspiration in India under conditions of recent and futureclimate change Agric For Meteorol 1997 87 55minus73 DOI 101016S0168-1923(97)00006-3(21) Liu M Xia Z Han S Wang X Tang Z Relationshipbetween variation of evapotranspiration and ecological deterioration insource region of Yellow River J Hohai Uni 2009 37 (6) 631minus634 (inChinese)(22) Yi X Yin Y Li G Peng J Temperature variation in recent50 years in the three-river headwaters region of Qinghai Province ActaGeol Sin 2011 66 (11) 1451minus1465 (in Chinese)(23) Hou W Li Y Soil surface humidity index and sensitivityanalysis of the climate factors that affect it in the Yellow River sourceregions J Glaciol Geocryol 2010 32 (6) 1226minus1233 (in Chinese)(24) Mauser W Schadlich S Modelling the spatial distribution ofevapotranspiration on different scales using remote sensing data JHydrol 1998 212 250minus267 DOI 101016S0022-1694(98)00228-5(25) Salati E Vose P B Amazon basin A system in equilibriumScience 1984 225 (4658) 129minus138 DOI 1023071693078

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

G

(26) Schlesinger W H Reynolds J F Cunningham G LHuenneke L F Jarrell W M Virginia R A Whitford W GBiological feedbacks in global desertification Science 1990 247 (4946)1043minus1048 DOI 101126science24749461043(27) Shukla J Nobre C Sellers P Amazon deforestation andclimate change Science 1990 247 (4948) 1322minus1325 DOI 101126science24749481322(28) Baron J S Hartman M D Kittel T G F Band L E OjimaD S Lammers R B Effects of land cover water redistribution andtemperature on ecosystem processes in the South Platte basin EcolAppl 1998 8 (4) 1037minus1051 DOI 1018901051-0761(1998)008[1037eolcwr]20co2(29) Boegh E Soegaard H Thomsen A Evaluating evapotranspi-ration rates and surface conditions using Landsat TM to estimateatmospheric resistance and surface resistance Remote Sens Environ2002 79 (2minus3) 329minus343 DOI 101016S0034-4257(01)00283-8(30) Sterling S M Ducharne A Polcher J The impact of globalland-cover change on the terrestrial water cycle Nat Clim Change2013 3 (4) 385minus390 DOI 101038nclimate1690(31) Loarie S R Lobell D B Asner G P Mu Q Field C BDirect impacts on local climate of sugar-cane expansion in Brazil NatClim Change 2011 1 (2) 105minus109 DOI 101038nclimate1067(32) Teuling A J Seneviratne S I Stockli R et al Contrastingresponse of European forest and grassland energy exchange toheatwaves Nat Geosci 2010 3 (10) 722minus727 DOI 101038ngeo950(33) Anderson R G Canadell J G Randerson J T et alBiophysical considerations in forestry for climate protection FrontEcol Environ 2011 9 (3) 174minus182 DOI 101890090179(34) Campo-Bescos M A Munoz-Carpena R Kaplan D ASouthworth J Zhu L Waylen P R Beyond precipitationPhysiographic gradients dictate the relative importance of environ-mental drivers on savanna vegetation PLoS One 2013 8 (8)No e72348 DOI 101371journalpone0072348(35) Schneider K Mauser W Processing and accuracy of Landsatthematic mapper data for lake surface temperature measurement Int JRemote Sens 1996 17 (11) 2027minus2041 DOI 10108001431169608948757(36) Li F Jackson T J Kustas W P Schmugge T J French AN Cosh M H Bindlish R Deriving land surface temperature fromLandsat 5 and 7 during SMEX02SMACEX Remote Sens Environ2004 92 (4) 521minus534 DOI 101016jrse200402018(37) Bastiaanssen W G M Menenti M Feddes R A Holtslag AA M A remote sensing surface energy balance algorithm for land(SEBAL) J Hydrol 1998 212 198minus212 DOI 101016S0022-1694(98)00253-4(38) Bonan G B Forests and climate change Forcings feedbacksand the climate benefits of forests Science 2008 320 (5882) 1444minus1449 DOI 101126science1155121(39) Li Z Liu X Ma T Kejia D Zhou Q Yao B Niu TRetrieval of the surface evapotranspiration patterns in the alpinegrasslandminuswetland ecosystem applying SEBAL model in the sourceregion of the Yellow River China Ecol Modell 2013 270 64minus75DOI 101016jecolmodel201309004(40) Zhao L Lee X Smith R B Oleson K Strong contributionsof local background climate to urban heat islands Nature 2014 511(7508) 216minus219 DOI 101038nature13462(41) van Heerwaarden C C de Arellano J V Gounou AGuichard F Couvreux F Understanding the daily cycle ofevapotranspiration A method to quantify the influence of forcingsand feedbacks J Hydrometeorol 2010 11 (6) 1405minus1422DOI 1011752010jhm12721(42) Chang G G Li F X Li L Changes and Restoration in SourceRegion of Three Rivers of China Metrology Press Beijing China 2010(in Chinese)(43) Tong L Xu X Fu Y Li S Wetland changes and theirresponses to climate change in the three-river headwaters region ofChina since the 1990s Energies 2014 7 (4) 2515minus2534DOI 103390en7042515

(44) Zhang Y Zhao X Zhao S Feng C Correlation betweenevapotranspiration and climate factors in warm steppe in source regionof Yangtze Yellow and Yalu Tsangpo Rivers J Desert Res 2010 30(2) 363minus368 (in Chinese)(45) Jiang Y Li D Variations of Tangnaihai runoff andprecipitation and temperature in the upper reach of the YellowRiverMeteorol Disaster Reduct Res 2011 34 (2) 51minus57 (in Chinese)(46) Molders N Kramm G Influence of wildfire induced land-cover changes on clouds and precipitation in interior AlaskaA casestudy Atmos Res 2007 84 (2) 142minus168 DOI 101016jatmos-res200606004(47) Obregon A Gehrig-Downie C Gradstein S R RollenbeckR Bendix J Canopy level fog occurrence in a tropical lowland forestof French Guiana as a prerequisite for high epiphyte diversity AgricFor Meteorol 2011 151 (3) 290minus300 DOI 101016jagrfor-met201011003(48) Harding R J Blyth E M Tuinenburg O A Wiltshire ALand atmosphere feedbacks and their role in the water resources of theGanges basin Sci Total Environ 2013 468 (S1) S85minusS92DOI 101016jscitotenv201303016(49) Bradshaw A D Ecological principles and land reclamationpractice Landscape Plann 1984 11 (1) 35minus48(50) Shelton M L Hydroclimatology Perspectives and ApplicationsCambridge Press Cambridge UK 2009(51) Liu Y Lu S Li S Gao Y Numerical simulation of impact ofland surface changes on regional climatic environment in sourceregions of Yellow River Plateau Meteorol 2009 28 (2) 327minus334 (inChinese)

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

H

essential Although previous climate change studies havepresented global climate models and have discussed energyuse strategies that mitigate global warming few works haveexamined regional THC processes to assess or manage regionalclimatic and hydrological changes Few climate policy studieshave determined the net impact of biophysical changesresulting from land use pattern changes33 Our work in thesource region of the Yellow River serves as an initial attempt toexplore the feasibility of applying innovative methodologies toaddress gaps in knowledge and to evaluate ecologicalrestoration strategies in future studies

ASSOCIATED CONTENTS Supporting InformationClassification criterions of the study area and the accuracyassessment for the land cover classification detailed methodsfor the calculation models of ATCOR2 and validation of theestimated results with meteorological data The SupportingInformation is available free of charge on the ACS Publicationswebsite at DOI 101021es505985q

AUTHOR INFORMATIONCorresponding AuthorTelephone +86-10-6279-4119 E-mail xuehua-hjxmailtsinghuaeducnNotesThe authors declare no competing financial interest

ACKNOWLEDGMENTSThis project was financially supported by the fund from theState Key Joint Laboratory of Environment Simulation andPollution Control China (11Y02ESPCT) and the projectnamed The Relationship of Birds Migration Patterns andHabitat Factors in Poyang Lake (2010CB530300-04) Theauthors thank the anonymous reviewers for their critiqueswhich improved earlier versions of this manuscript

REFERENCES(1) Armesto J J Bautista S Del Val E Ferguson B Garciacutea XGaxiola A Godinez-Alvarez H Gann G Lopez-Barrera FManson R Nu nez-Avila M Ortiz-Arrona C Tognetti PWilliams-Linera G Towards an ecological restoration networkReversing land degradation in Latin America Front Ecol Environ2007 5 (4) w1minusw4 DOI 1018901540-9295(2007)5[w1TAERNR]20CO2(2) Baustian M M Georgia M Dreelin E A Esselman PSchultze S R Qian L Awb T G Luo L Rose J BA Onehundred year review of the socioeconomic and ecological systems ofLake St Clair North America J Great Lakes Res 2014 40 15minus26DOI 101016jjglr201311006(3) Grumbine R E Assessing environmental security in ChinaFront Ecol Environ 2014 12 (7) 403minus411 DOI 101890130147(4) Bullon T Environmental assessment and land change analysis insemi-natural land covers applicable to land management Int J ApplEarth Obs Geoinf 2015 34 147minus156 DOI 101016jjag201408006(5) Liu J Linderman M Ouyang Z An L Yang J Zhang HEcological degradation in protected areas The case of Wolong NatureReserve for giant pandas Science 2001 292 (5514) 98minus101DOI 101126science1058104(6) Poyatos R Latron J Llorens P Land use and land coverchange after agricultural abandonment Mt Res Dev 2003 23 (4)362minus368 DOI 1016590276-4741(2003)023[0362LUALCC]20CO2(7) Feng J Wang T Xie C Eco-environmental degradation in thesource region of the Yellow River northeast QinghaiminusXizang Plateau

Environ Monit Assess 2006 122 (1minus3) 125minus143 DOI 101007s10661-005-9169-2(8) Wang G Cheng G Eco-environmental changes and causativeanalysis in the source regions of the Yangtze and Yellow Rivers ChinaEnviron Syst Decis 2000 20 (3) 221minus232 DOI 101023A1006703831018(9) Pan J Liu J Yellow River source area of land use and landscapepattern change and its ecological effects J Arid Land Res Environ2005 19 (4) 69minus74 (in Chinese) DOI 103969jissn1003-7578200504014(10) Xu J Song L Zhao Z Hu Y Liu C Monitoring grasslanddegradation dynamically at Maduo County in source region of YellowRiver in past 15 years based on remote sensing Arid Land Geogr 201235 (4) 615minus622 (in Chinese) DOI 1013826jcnkicn65-1103x201204018(11) Yang J Ding Y Chen R Spatial and temporal of variations ofalpine vegetation cover in the source regions of the Yangtze andYellow Rivers of the Tibetan Plateau from 1982 to 2001 Environ Geol2006 50 (3) 313minus322 DOI 101007s00254-006-0210-8(12) Davidson A D Detling J K Brown J H Ecological roles andconservation challenges of social burrowing herbivorous mammals inthe worldrsquos grasslands Front Ecol Environ 2012 10 (9) 477minus486DOI 101890110054(13) Qu J Li W Yang M Ji W Zhang Y Life history of theplateau pika (Ochotona curzoniae) in alpine meadows of the TibetanPlateau Mamm Biol 2013 78 (1) 68minus72 DOI 101016jmambio201209005(14) Liu J Nie H Studies on the population productivity ecology ofplateau pika III Trend of population dynamics in plateau pika withdensity-independent and density-dependent vital rates Acta TheriolSin 1992 12 (2) 139minus146 (in Chinese)(15) Liu W Zhou L Wang X The study of different grazingintensity on the role of plants and rodent studies Acta Ecol Sin 199919 (3) 88minus94 (in Chinese)(16) Bai W Zhang Y Xie G Shen Z Causes analysis of grasslanddegradation Maduo County in Yellow River source region Acta EcolSin 2002 13 (7) 823minus826 (in Chinese) DOI 1013287j1001-933220020194(17) Fang Y Qin D Ding Y Frozen soil change and adaptation ofanimal husbandry A case of the source regions of Yangtze and YellowRivers Environ Sci Policy 2011 14 (5) 555minus568 DOI 101016jenvsci201103012(18) Verburg P H van de Steeg J Veldkamp A Willemen LFrom land cover change to land function dynamics A major challengeto improve land characterization J Environ Manage 2009 90 (3)1327minus1335 DOI 101016jjenvman200808005(19) Barnett T P Adam J C Lettenmaier A D P Potentialimpacts of a warming climate on water availability in snow-dominatedregions Nature 2005 438 303minus309 DOI 101038nature04141(20) Chattopadhyay N Hulme M Evaporation and potentialevapotranspiration in India under conditions of recent and futureclimate change Agric For Meteorol 1997 87 55minus73 DOI 101016S0168-1923(97)00006-3(21) Liu M Xia Z Han S Wang X Tang Z Relationshipbetween variation of evapotranspiration and ecological deterioration insource region of Yellow River J Hohai Uni 2009 37 (6) 631minus634 (inChinese)(22) Yi X Yin Y Li G Peng J Temperature variation in recent50 years in the three-river headwaters region of Qinghai Province ActaGeol Sin 2011 66 (11) 1451minus1465 (in Chinese)(23) Hou W Li Y Soil surface humidity index and sensitivityanalysis of the climate factors that affect it in the Yellow River sourceregions J Glaciol Geocryol 2010 32 (6) 1226minus1233 (in Chinese)(24) Mauser W Schadlich S Modelling the spatial distribution ofevapotranspiration on different scales using remote sensing data JHydrol 1998 212 250minus267 DOI 101016S0022-1694(98)00228-5(25) Salati E Vose P B Amazon basin A system in equilibriumScience 1984 225 (4658) 129minus138 DOI 1023071693078

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

G

(26) Schlesinger W H Reynolds J F Cunningham G LHuenneke L F Jarrell W M Virginia R A Whitford W GBiological feedbacks in global desertification Science 1990 247 (4946)1043minus1048 DOI 101126science24749461043(27) Shukla J Nobre C Sellers P Amazon deforestation andclimate change Science 1990 247 (4948) 1322minus1325 DOI 101126science24749481322(28) Baron J S Hartman M D Kittel T G F Band L E OjimaD S Lammers R B Effects of land cover water redistribution andtemperature on ecosystem processes in the South Platte basin EcolAppl 1998 8 (4) 1037minus1051 DOI 1018901051-0761(1998)008[1037eolcwr]20co2(29) Boegh E Soegaard H Thomsen A Evaluating evapotranspi-ration rates and surface conditions using Landsat TM to estimateatmospheric resistance and surface resistance Remote Sens Environ2002 79 (2minus3) 329minus343 DOI 101016S0034-4257(01)00283-8(30) Sterling S M Ducharne A Polcher J The impact of globalland-cover change on the terrestrial water cycle Nat Clim Change2013 3 (4) 385minus390 DOI 101038nclimate1690(31) Loarie S R Lobell D B Asner G P Mu Q Field C BDirect impacts on local climate of sugar-cane expansion in Brazil NatClim Change 2011 1 (2) 105minus109 DOI 101038nclimate1067(32) Teuling A J Seneviratne S I Stockli R et al Contrastingresponse of European forest and grassland energy exchange toheatwaves Nat Geosci 2010 3 (10) 722minus727 DOI 101038ngeo950(33) Anderson R G Canadell J G Randerson J T et alBiophysical considerations in forestry for climate protection FrontEcol Environ 2011 9 (3) 174minus182 DOI 101890090179(34) Campo-Bescos M A Munoz-Carpena R Kaplan D ASouthworth J Zhu L Waylen P R Beyond precipitationPhysiographic gradients dictate the relative importance of environ-mental drivers on savanna vegetation PLoS One 2013 8 (8)No e72348 DOI 101371journalpone0072348(35) Schneider K Mauser W Processing and accuracy of Landsatthematic mapper data for lake surface temperature measurement Int JRemote Sens 1996 17 (11) 2027minus2041 DOI 10108001431169608948757(36) Li F Jackson T J Kustas W P Schmugge T J French AN Cosh M H Bindlish R Deriving land surface temperature fromLandsat 5 and 7 during SMEX02SMACEX Remote Sens Environ2004 92 (4) 521minus534 DOI 101016jrse200402018(37) Bastiaanssen W G M Menenti M Feddes R A Holtslag AA M A remote sensing surface energy balance algorithm for land(SEBAL) J Hydrol 1998 212 198minus212 DOI 101016S0022-1694(98)00253-4(38) Bonan G B Forests and climate change Forcings feedbacksand the climate benefits of forests Science 2008 320 (5882) 1444minus1449 DOI 101126science1155121(39) Li Z Liu X Ma T Kejia D Zhou Q Yao B Niu TRetrieval of the surface evapotranspiration patterns in the alpinegrasslandminuswetland ecosystem applying SEBAL model in the sourceregion of the Yellow River China Ecol Modell 2013 270 64minus75DOI 101016jecolmodel201309004(40) Zhao L Lee X Smith R B Oleson K Strong contributionsof local background climate to urban heat islands Nature 2014 511(7508) 216minus219 DOI 101038nature13462(41) van Heerwaarden C C de Arellano J V Gounou AGuichard F Couvreux F Understanding the daily cycle ofevapotranspiration A method to quantify the influence of forcingsand feedbacks J Hydrometeorol 2010 11 (6) 1405minus1422DOI 1011752010jhm12721(42) Chang G G Li F X Li L Changes and Restoration in SourceRegion of Three Rivers of China Metrology Press Beijing China 2010(in Chinese)(43) Tong L Xu X Fu Y Li S Wetland changes and theirresponses to climate change in the three-river headwaters region ofChina since the 1990s Energies 2014 7 (4) 2515minus2534DOI 103390en7042515

(44) Zhang Y Zhao X Zhao S Feng C Correlation betweenevapotranspiration and climate factors in warm steppe in source regionof Yangtze Yellow and Yalu Tsangpo Rivers J Desert Res 2010 30(2) 363minus368 (in Chinese)(45) Jiang Y Li D Variations of Tangnaihai runoff andprecipitation and temperature in the upper reach of the YellowRiverMeteorol Disaster Reduct Res 2011 34 (2) 51minus57 (in Chinese)(46) Molders N Kramm G Influence of wildfire induced land-cover changes on clouds and precipitation in interior AlaskaA casestudy Atmos Res 2007 84 (2) 142minus168 DOI 101016jatmos-res200606004(47) Obregon A Gehrig-Downie C Gradstein S R RollenbeckR Bendix J Canopy level fog occurrence in a tropical lowland forestof French Guiana as a prerequisite for high epiphyte diversity AgricFor Meteorol 2011 151 (3) 290minus300 DOI 101016jagrfor-met201011003(48) Harding R J Blyth E M Tuinenburg O A Wiltshire ALand atmosphere feedbacks and their role in the water resources of theGanges basin Sci Total Environ 2013 468 (S1) S85minusS92DOI 101016jscitotenv201303016(49) Bradshaw A D Ecological principles and land reclamationpractice Landscape Plann 1984 11 (1) 35minus48(50) Shelton M L Hydroclimatology Perspectives and ApplicationsCambridge Press Cambridge UK 2009(51) Liu Y Lu S Li S Gao Y Numerical simulation of impact ofland surface changes on regional climatic environment in sourceregions of Yellow River Plateau Meteorol 2009 28 (2) 327minus334 (inChinese)

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

H

(26) Schlesinger W H Reynolds J F Cunningham G LHuenneke L F Jarrell W M Virginia R A Whitford W GBiological feedbacks in global desertification Science 1990 247 (4946)1043minus1048 DOI 101126science24749461043(27) Shukla J Nobre C Sellers P Amazon deforestation andclimate change Science 1990 247 (4948) 1322minus1325 DOI 101126science24749481322(28) Baron J S Hartman M D Kittel T G F Band L E OjimaD S Lammers R B Effects of land cover water redistribution andtemperature on ecosystem processes in the South Platte basin EcolAppl 1998 8 (4) 1037minus1051 DOI 1018901051-0761(1998)008[1037eolcwr]20co2(29) Boegh E Soegaard H Thomsen A Evaluating evapotranspi-ration rates and surface conditions using Landsat TM to estimateatmospheric resistance and surface resistance Remote Sens Environ2002 79 (2minus3) 329minus343 DOI 101016S0034-4257(01)00283-8(30) Sterling S M Ducharne A Polcher J The impact of globalland-cover change on the terrestrial water cycle Nat Clim Change2013 3 (4) 385minus390 DOI 101038nclimate1690(31) Loarie S R Lobell D B Asner G P Mu Q Field C BDirect impacts on local climate of sugar-cane expansion in Brazil NatClim Change 2011 1 (2) 105minus109 DOI 101038nclimate1067(32) Teuling A J Seneviratne S I Stockli R et al Contrastingresponse of European forest and grassland energy exchange toheatwaves Nat Geosci 2010 3 (10) 722minus727 DOI 101038ngeo950(33) Anderson R G Canadell J G Randerson J T et alBiophysical considerations in forestry for climate protection FrontEcol Environ 2011 9 (3) 174minus182 DOI 101890090179(34) Campo-Bescos M A Munoz-Carpena R Kaplan D ASouthworth J Zhu L Waylen P R Beyond precipitationPhysiographic gradients dictate the relative importance of environ-mental drivers on savanna vegetation PLoS One 2013 8 (8)No e72348 DOI 101371journalpone0072348(35) Schneider K Mauser W Processing and accuracy of Landsatthematic mapper data for lake surface temperature measurement Int JRemote Sens 1996 17 (11) 2027minus2041 DOI 10108001431169608948757(36) Li F Jackson T J Kustas W P Schmugge T J French AN Cosh M H Bindlish R Deriving land surface temperature fromLandsat 5 and 7 during SMEX02SMACEX Remote Sens Environ2004 92 (4) 521minus534 DOI 101016jrse200402018(37) Bastiaanssen W G M Menenti M Feddes R A Holtslag AA M A remote sensing surface energy balance algorithm for land(SEBAL) J Hydrol 1998 212 198minus212 DOI 101016S0022-1694(98)00253-4(38) Bonan G B Forests and climate change Forcings feedbacksand the climate benefits of forests Science 2008 320 (5882) 1444minus1449 DOI 101126science1155121(39) Li Z Liu X Ma T Kejia D Zhou Q Yao B Niu TRetrieval of the surface evapotranspiration patterns in the alpinegrasslandminuswetland ecosystem applying SEBAL model in the sourceregion of the Yellow River China Ecol Modell 2013 270 64minus75DOI 101016jecolmodel201309004(40) Zhao L Lee X Smith R B Oleson K Strong contributionsof local background climate to urban heat islands Nature 2014 511(7508) 216minus219 DOI 101038nature13462(41) van Heerwaarden C C de Arellano J V Gounou AGuichard F Couvreux F Understanding the daily cycle ofevapotranspiration A method to quantify the influence of forcingsand feedbacks J Hydrometeorol 2010 11 (6) 1405minus1422DOI 1011752010jhm12721(42) Chang G G Li F X Li L Changes and Restoration in SourceRegion of Three Rivers of China Metrology Press Beijing China 2010(in Chinese)(43) Tong L Xu X Fu Y Li S Wetland changes and theirresponses to climate change in the three-river headwaters region ofChina since the 1990s Energies 2014 7 (4) 2515minus2534DOI 103390en7042515

(44) Zhang Y Zhao X Zhao S Feng C Correlation betweenevapotranspiration and climate factors in warm steppe in source regionof Yangtze Yellow and Yalu Tsangpo Rivers J Desert Res 2010 30(2) 363minus368 (in Chinese)(45) Jiang Y Li D Variations of Tangnaihai runoff andprecipitation and temperature in the upper reach of the YellowRiverMeteorol Disaster Reduct Res 2011 34 (2) 51minus57 (in Chinese)(46) Molders N Kramm G Influence of wildfire induced land-cover changes on clouds and precipitation in interior AlaskaA casestudy Atmos Res 2007 84 (2) 142minus168 DOI 101016jatmos-res200606004(47) Obregon A Gehrig-Downie C Gradstein S R RollenbeckR Bendix J Canopy level fog occurrence in a tropical lowland forestof French Guiana as a prerequisite for high epiphyte diversity AgricFor Meteorol 2011 151 (3) 290minus300 DOI 101016jagrfor-met201011003(48) Harding R J Blyth E M Tuinenburg O A Wiltshire ALand atmosphere feedbacks and their role in the water resources of theGanges basin Sci Total Environ 2013 468 (S1) S85minusS92DOI 101016jscitotenv201303016(49) Bradshaw A D Ecological principles and land reclamationpractice Landscape Plann 1984 11 (1) 35minus48(50) Shelton M L Hydroclimatology Perspectives and ApplicationsCambridge Press Cambridge UK 2009(51) Liu Y Lu S Li S Gao Y Numerical simulation of impact ofland surface changes on regional climatic environment in sourceregions of Yellow River Plateau Meteorol 2009 28 (2) 327minus334 (inChinese)

Environmental Science amp Technology Article

DOI 101021es505985qEnviron Sci Technol XXXX XXX XXXminusXXX

H