Rainfall Amount, Intensity, Duration, and Frequency Relationships in the Mae Chaem Watershed in...

458 VOLUME 5J O U R N A L O F H Y D R O M E T E O R O L O G Y

q 2004 American Meteorological Society

Rainfall Amount, Intensity, Duration, and Frequency Relationships in the Mae ChaemWatershed in Southeast Asia

KOJI DAIRAKU

Atmospheric Environment Division, National Institute for Environmental Studies, Tsukuba, Ibaraki, Japan

SEITA EMORI*

Frontier Research System for Global Change, Yokohama, Japan

TAIKAN OKI

Institute of Industrial Science, the University of Tokyo, Tokyo, Japan

(Manuscript received 7 July 2003, in final form 8 November 2003)

ABSTRACT

A dense tipping-bucket rain gauge network was established in the Mae Chaem watershed in the mountainsof northwestern Thailand as part of the Global Energy and Water Cycle Experiment (GEWEX) Asian MonsoonExperiment-Tropics (GAME-T). Investigations of rainfall amounts, intensities, durations, and frequencies in therainy season revealed strong orographic rainfall enhancement in the region. The larger amount of high-altituderainfall was attributed to duration and frequency rather than intensity. Despite large rainfall variations, similarpatterns were found in the two study years, 1998 and 1999.

1. Introduction

Hydrological changes and the impact of those chang-es constitute a fundamental global-warming-related con-cern (Vorosmarty et al. 2000). Faced with threats tohuman life and natural ecosystems, such as droughts,floods, and soil erosion, water resource planners must,increasingly, make future risk assessments (Schnur2002). Historically, stable climatic conditions have beenassumed for water resource management, planning, andcivil engineering designs. However, global climatechange may lead to changes in rainfall events (Palmerand Ralsanen 2002). We must learn, therefore, to adaptto climate changes and cope with climate variability(Zwiers 2002). Hydrological predictions that account forglobal climate change mainly use general circulationmodels (GCMs) (e.g., Coe 2000; Koster et al. 2000;Vorosmarty et al. 2000; Palmer and Ralsanen 2002;Milly et al. 2002). However, coarse spatial resolutions

*Current affiliation: Atmospheric Environment Division, NationalInstitute for Environmental Studies, Tsukuba, Ibaraki, Japan.

Corresponding author address: Dr. Koji Dairaku, Atmospheric En-vironment Division, National Institute for Environmental Studies, 16-2Onogawa Tsukuba Ibaraki 305-0053, Japan.E-mail: [email protected]

and uncertain physical processes limit the amount ofregional-scale estimates available to planners.

Many researchers have examined temporal and spatialprecipitation distributions, especially the distribution ofextreme precipitation values. These studies are impor-tant for water resource and flood control managementand for designing and planning various engineering pro-jects (Smith 1992; Palmer and Ralsanen 2002). Precip-itation generally increases with elevation, and mountainranges can create leeward-side ‘‘rain shadows.’’ Thisimplies some underlying mechanisms that play an im-portant role in organizing precipitation systems in amountainous region. However, worldwide, only an ir-regular, coarse grid of precipitation measurement dataexists, and it is not sufficient to say that mechanismsare fully understood. Researchers have attempted to de-velop a number of geostatistical interpolation methodsto obtain reliable rainfall and rainfall-risk maps inmountainous regions with sparse precipitation measure-ment coverage (Phillips et al. 1992; Yamada et al. 1995;Weisse and Bois 2001; Kyriakidis et al. 2001). Otherresearchers have investigated orographic-induced me-teorological mechanisms and statistical orographic rainstructures (Smith 1986; Stow et al. 1991; Wratt et al.1996; Barros and Kuligowski 1998; Purdy et al. 2001;Miniscloux et al. 2001; Neiman et al. 2002). Only about6600 stations worldwide archive rain gauge data through

JUNE 2004 459D A I R A K U E T A L .

the Global Telecommunications Network or other re-gional or national data collection centers (Rudolf et al.1994). Both precipitation data and data archives are alsomostly located in mid- or high-latitude developed coun-tries. Consequently, most studies, have focused on mid-latitude developed countries, which are more likely tohave the necessary high-temporal-resolution rainfall re-cords. In recent years, hydrometeorological networkshave been installed in some ungauged high-elevationbasins. However, only a few studies (Ramage 1964;Thauvin and Lebel 1991; Oki and Musiake 1994; Uenoet al. 1994, 2001; Shuin et al. 1996; Barros et al. 2000,2003; Lang and Barros 2002, Barros and Lang 2003)have examined rainfall characteristics in a tropical re-gion. Tropical countries typically have sparse opera-tional rain gauge networks and extremely limited high-temporal-resolution (1 h or less) precipitation records.

Rain gauges provide relatively accurate point mea-surements of precipitation, although rain gauge obser-vations do suffer from systematic errors and biases ascaused by wind-induced undercatch, wetting, and evap-oration loss (Neff 1977; Golubev et al. 1992; Yang etal. 1998). Point-to-area and gauge gaps are the biggestproblems for gauge analysis. Researchers have used in-frared and microwave radiance satellite observations toretrieve precipitation information from many parts ofthe globe. However, estimates made from satellite ob-servations contain nonnegligible random errors and bi-ases due to the indirect relationship between observa-tions and precipitation, inadequate sampling, and al-gorithmic imperfections (Xie and Arkin 1997). Huffmanet al. (1995) noted that rain gauge analysis at the GlobalPrecipitation Climatology Centre (GPCC; Rudolf et al.1994) revealed important quantitative differences fromGPCP satellite estimates, particularly in Southeast Asiaand Central America. Xie and Arkin (1997) also notedlarge biases and random errors in a GPCC gauge, po-sitioned in an area of Thailand affected by significantmountain-induced precipitation variations.

Understanding rainfall variability within regions isindispensable for applying global climate research re-sults to regional climate and water cycles. As Dettingerand Diaz (2000) noted, relatively high-altitude catch-ments feed river basins, but weather stations are locatedat lower altitudes, leading to a low-altitude bias in pre-cipitation estimates for most river basins. Oki (2001)also noted that if a climate model produces weaker andmore continuous rainfall intensity than occurs in reality,evapotranspiration from intercepted water values shouldalso increase, resulting in exaggeratedly low soil mois-ture and runoff values. Direct runoff calculated by landsurface models (LSMs) could also appear lower becauseof the weaker rain rates.

Shuin et al. (1996) studied temporal and spatial rain-fall at five ground rainfall stations in order to geosta-tistically interpolate rainfall and link spatial rainfall datato a radar rain gauge on Indonesia’s Mount Merapi. Theycompared the rainfall characteristics with characteristics

from Mount Fuji in Japan. The study revealed cleardiurnal and seasonal precipitation variations on MountMerapi. The small spatial and temporal rainfall scaleswere mainly associated with convective rains. Ueno etal. (1994) investigated diurnal variation of rainfallamounts, intensity, and frequency using tipping-bucketrain gauges at four stations located on the ThanggulaMountain range in the central Tibetan Plateau. The re-sults showed insignificant diurnal variations in rainfallamount, intensity, and frequency. Using newly installedrain gauges for a 4-month study, Barros et al. (2000)investigated rainfall characteristics of the 1999 monsoonin the complex orography of central Nepal. Their re-search revealed higher intensity rains over shorter du-rations at the lower-elevation stations.

Since 1998, Global Energy and Water Cycle Exper-iment (GEWEX) Asian Monsoon Experiment-Tropics(GAME-T) researchers have conducted high-temporal-resolution ground-based rainfall observations in moun-tainous areas of Thailand (Fig. 1). Fifteen tipping-buck-et-type rain gauges with subhourly (1 min) time reso-lutions were installed at sites 380–2565 m above sealevel in the Mae Chaem watershed (Fig. 2). The wa-tershed covers an area of 3853 km2 (Dairaku et al. 2000;Kuraji et al. 2001). Research revealed that altitude-re-lated rainfall enhancement in the Mae Chaem watershedcorrelated to the total duration of rainfall and was in-dependent of the mean rainfall intensity (Dairaku et al.2000). Kuraji et al. (2001) further investigated rainfallspatial scales, mean rainfall intensity, and total rainfallduration in the wet and dry seasons.

Previous studies, however, have not examined thecharacteristics of individual rainfall events. That is,while studies have investigated average or total rainfallcharacteristics such as mean rainfall intensities and totalrainfall durations, they have not investigated rainfallfrequency, amount, intensity, and duration on an eventbasis. This paper investigates these relationships in theMae Chaem watershed, a tropical mountainous regionusing high-temporal-resolution tipping-bucket raingauge data to understand the statistical precipitationstructures in the rainy season that induce significant biasand spatial sampling error. The sections that follow de-scribe the study area, dataset, methodology, and results.

2. Methods

a. Study area and dataset

The Mae Chaem watershed study area is one of theGAME-T research sites (Dairaku et al. 2000; Kuraji etal. 2001). Located in the northwest Chao Phraya Riverbasin, the Mae Chaem watershed has an area of 3853km2 (Fig. 1). The highest peak in the watershed, DoiInthanon, rises 2535 m above sea level. Tipping-bucket-type rain gauges (20-cm orifice diameter; 0.5 mm pre-cipitation per tip) with 1-s resolutions have collectedrainfall data in the mountainous area since November


FIG. 1. Solid rectangle indicates the location of the Mae Chaem watershed in Thailand.

TABLE 1. Location and elevation of rain gauges in Fig. 2.

Station name Lat (N) Lon (E) Elev (m)

Wat ChanBo KaeoMae YodMae SaMae Jon Luang*

198049090188529120188499530188499200188409090

988179370988309420988069070988209060988289380

95513301091

6271356

Doi InthanonMae KlangMae ChaemResearch StationSirikit Plantation

188359170188309580188309070188319200188219570

988299110988289130988229130988179400988289050

24961339

47110941225

Mae NingMae LongOb LuangHuay Bong**Mae Tho**

188369410188259580188139210188099010188159040

988139030988139340988289580988259400988129340

15851369

326792

1220

* Unused station in 1998.** Unused stations in 1998 and 1999.

1997 (Fig. 2). The observation network consisted ofonly seven stations before June 1998 but was increasedto its present status after that time. Table 1 shows thelocations of Mae Chaem watershed rain gauges at ele-vations ranging from 326 to 2496 m. The watershed’saverage altitude is 948 m.

Wind-induced undercatch, wetting, and evaporationlosses can cause systematic errors and biases in raingauge observations (Neff 1977; Golubev et al. 1992;Yang et al. 1998). Wetting loss is evaporation from theinner-wall surface of the rain gauge. Evaporation lossis defined as the undermeasurement of water lost byevaporation before observation. Wind-induced under-catch errors are extremely significant, especially for sol-id precipitation like snowfall, but in the GAME-T ex-perimental watershed, solid precipitation is rarely ob-served. It has been reported that for liquid rain, under-catch error using an unshielded rain gauge is only about4%, and wetting error and evaporation are negligible(Golubev et al. 1992). Thus, error correction was notapplied to the rainfall records in this study.

To investigate the precipitation characteristics in therainy season, the rainfall records used in this study wereobtained from 12 sites from 1 June to 31 October 1998and from 13 sites from 1 June to 31 October 1999 thathad complete data. The data collection dates allowed

for investigation of rainy-season rainfall amount, inten-sity, duration, and frequency relationships. Because ofincomplete measurements during the study period, 1998data from the Mae Jon Luang, Mae Tho, and Huay Bongsites and 1999 data from the Mae Tho and Huay Bongsites were not used in this study.


FIG. 2. GAME-T rainfall observation stations in the Mae Chaem watershed.

FIG. 3. Concept of the mean-time method for defining rainfallevents and durations.

b. Climate factors that describe rainfallcharacteristics

In previous studies (Dairaku et al. 2000; Kuraji et al.2001), only two factors, average rainfall intensity andtotal rainfall duration, were considered in the exami-nation of orographic rainfall characteristics in the MaeChaem watershed. In order to investigate rainfallamount, intensity, duration, and frequency relationships,the duration of each rainfall event should first be definedusing the interval-time method or the mean-time methoddescribed by Dairaku et al. (2000).

This study used the mean-time method to define arainfall event and its rainfall duration and to judgewhether a recorded rainfall tipping time met the meantime. As shown in Fig. 3, tn (t1, t2, t3, . . .) representsthe tipping time recorded by a tipping-bucket rain gauge,Tm represents the mean time, and T1 represents the start-ing time of the calculation:

T 1 T (k 2 1) % t , T 1 T k (k ^ 1).1 m n 1 m

(1)

By this definition, if tn satisfies the condition (1), whenk is a certain value, rain will fall continually during theTm. This study uses 1 h as the mean time (Tm).

For each rainfall event, the rainfall amount, intensity,and duration relationship is defined as

R 5 I T ,i i i (2)

where Ri (millimeters) represents the rainfall amount, Ii

represents rainfall intensity (millimeters per hour), andTi represents the rainfall duration (hour) of each rainfall


FIG. 4. Relationships between elevation and (a) rainfall amount, (b) average rainfall intensity, (c) average rainfallduration, and (d) frequency in 1998 and 1999. In this legend, 6–10 denotes the months Jun–Oct.

event i. The total rainfall amount from 1 June through31 October 1998 and 1999 is represented by

R 5 I T , (3)O Oi i i

and mean rainfall intensity is represented by

R I TO Oi i iI 5 5 . (4)

T TO Oi i

Here, S Ti is the total rainfall duration. Average rainfallduration is defined by

1T 5 T , (5)O iN

where N represents the number (or frequency) of rainfallevents.

c. Rainfall amount, intensity, duration, and frequencyrelationships

Using the above definitions, the total rainfall amountwas divided into three factors: average rainfall intensity,average rainfall duration, and frequency. The resultantequation is as follows:

I TO i iR 5 I T 5 TO O Oi i i iTO i

I T TO Oi i i5 N 5 ITN. (6)

NTO i

This equation introduces average rainfall duration andfrequency by defining each rainfall event. This allowsus to analyze rainfall amount, intensity, duration, andfrequency relationships.

3. Results and discussion

a. Relating rainfall amount, intensity, duration, andfrequency to elevation in the Mae Chaemwatershed

Figure 4 shows altitude-related orographic rainfallcharacteristics in the watershed from June to October1998 and 1999. Equation (6) defines the four variables.Figure 4a clearly reveals varying rainfall amounts byelevation throughout the rainy season and large inter-annual rainfall amount variations. As illustrated in Fig.4b, average rainfall intensity did not correlate to ele-vation. In other words, the high rainfall at high elevationstations shown in Fig. 4a was not attributed to high


FIG. 5. Variation of precipitation amounts in the Jun–Oct period, 1979–99. Data from CMAP,GPCP, GPCC, and TRMM4B43 were averaged over a 28 by 1.58 area (17.58–19.58N by 97.58–998E). Mae Chaem watershed precipitation was estimated by substituting the average elevationvalue of the watershed (948 m) for the linear regression shown in Fig. 4a.

rainfall intensities. Figure 4c shows that average rainfalldurations increased with elevation, while Fig. 4d showsthat rainfall frequencies also depended on elevation. Inother words, orographic factors contributed to longerrainfall durations and higher rainfall frequencies.

Climate Prediction Center (CPC) Merged Analysis ofPrecipitation (CMAP; Xie and Arkin 1997), the GlobalPrecipitation Climatology Project (GPCP; Huffman etal. 1997), GPCC (Rudolf et al. 1994), and the TropicalRainfall Measuring Mission (TRMM) product 3B43(Adler et al. 2000) datasets were averaged over the 28by 1.58 area located at 17.58–19.58N by 97.58–998E,which corresponds to the Mae Chaem watershed.Though the area (28 by 1.58) is larger than the gaugearea, the differences between the values averaged over28 by 1.58 area and the values averaged just over theMae Chaem watershed are very small because of thevery coarse grid of global datasets. Average precipita-tion was estimated for the GAME-T gauge array bytaking the average elevation value (948 m) within thewatershed boundary (Fig. 2) on the linear regression inFig. 4a.

Figure 5 shows interannual variations in precipitationamounts from June to October of 1979–99 for all avail-able years from each dataset. All datasets agreed that1998 was relatively dry, as compared to previous years,while in 1999 all the global precipitation datasets in-dicated close-to-average rainfall amounts. However, theGAME-T rain gauges placed 1999 rainfall far above theaverage of global datasets. The accurate estimation ofan areal representative precipitation and quality evalu-ations among datasets are questions beyond the scopeof this paper.

A relatively strong dry anomaly in 1998 annual pre-cipitation was found to prevail in most of Southeast Asiaby using the CMAP dataset. A wet anomaly was seen

in 1999 in the area, though the precipitation around thewatershed was close to the normal (not shown). Duringthe period 1997–2000, the global climate system ex-perienced a transition from a strong warm event (ElNino) of the ENSO cycle in sea surface temperature in1997/98 to a strong cold event (La Nina) in 1999/2000(Anyamba et al. 2002). However, the influence of theglobal circulation associated with ENSO on the meso-scale rainfall characteristics in our study region is notidentified clearly.

In summary, all datasets showed 1998 as a relativelydry year and found large rainfall amount variations be-tween 1998 and 1999. Orographic rainfall characteris-tics attributed to higher rainfall duration and frequencyrather than to rainfall intensity were also common inboth 1998 and 1999.

b. Rainfall events that contribute to orographicrainfall

To investigate the rainfall characteristics indicated inFig. 4 in more detail, the rainfall events were dividedinto four categories according to rainfall duration andintensity.

1) RAINFALL DURATION, FREQUENCY, AND

AMOUNT RELATIONSHIPS

Table 2 shows the average number of rainfall eventsand average rainfall amount for each rainfall durationcategory from June to October 1998 and 1999. Eachcategory was arbitrarily determined by rainfall durationto have a statistically sufficient number of samples.Though some different categorizations were investigat-ed (e.g., changing the 2–5-h category to a 3–4- or 3–


TABLE 2. Average number of events, ratio of average number of events to the total, average rainfall amount, ratio of average rainfallamount to the total, increment and coefficient of determination of the linear regression between elevation and frequency, and increment andcoefficient of determination of the linear regression between elevation and rainfall amount for each rainfall duration (T ) category from Junto Oct 1998 and 1999.

Rainfall duration (h) T 5 1 2 % T % 5 6 % T % 10 11 % T

Average No. of events

Ratio of No. of events to total (%)

1998199919981999

100.9158.9

54.856.6

74.5106.8

40.538.1

7.012.3

3.84.4

1.82.61.00.9

Average rainfall amount (mm)

Ratio of rainfall amount to total (%)

1998199919981999

130.8178.4

15.014.0

504.1706.6

58.055.6

138.1236.5

15.918.6

96.8149.6

11.111.8

Increment of linear regression(No. of events per 100 m)

Coefficient of determination (R2)

1998199919981999

3.75.50.74*0.58*

4.05.10.70*0.71*

0.51.20.44*0.64*

0.20.40.65*0.69*

Increment of linear regression[Rainfall amount (mm) per 100 m]


1998199919981999

3.44.50.43**0.28

23.431.10.45**0.70*

9.120.20.47**0.51*

11.934.30.56*0.68*

* 1% significance level.** 5% significance level.

6-h category), no significant differences resulted. Thus,categorization did not affect this paper’s conclusions.

As shown in Table 2, most rainfall events (approxi-mately 95%) lasted less than 5 h. A large number ofevents lasted approximately 1 h (54.8% in 1998; 56.6%in 1999), though the rainfall amounts from these eventsaccounted for a small percentage of the total observedrainfall (15.0% in 1998; 14.0% in 1999). The 2–5-hcategory accounted for the greatest rainfall amounts(58.0% in 1998; 55.6% in 1999). Rainfall events thatlasted over 6 h contributed a relatively large rainfallamount (around 30%) to the total, though they onlyaccounted for a small number of events (approximately5%).

Figures 6 and 7 examine the results further and focus,in particular, on spatial characteristics. Figure 6 showsthe relationships between elevation and the four rainfallduration categories. Figures 6a–d show the 1-, 2–5-, 6–10-, and over-11-h categories, respectively.

As indicated in Fig. 6a, the number of 1-h rainfallevents increased with elevation. The year 1999 had near-ly twice as many 1-h rainfall events as 1998, regardlessof station elevation. Figure 6b indicates a similar trend.The number of relatively long rainfall events was muchsmaller than those for the previous two categories (seeFigs. 6c and 6d). Nonetheless, a clear correlation existsbetween the number of rainfall events and elevation. Ascompared to 1998, the number of rainfall events in 1999increased only at the high-altitude stations. For longerrainfall duration categories (Figs. 6c and 6d), however,1998 and 1999 showed no significant differences in thenumber of low-elevation events.

Figure 7 illustrates the relationships between eleva-tion and rainfall amounts for the categorized events. Asshown in Fig. 7a, the rainfall amounts of the shortest

duration category did not clearly correlate with eleva-tion. The total rainfall amount for the 1-h category wasnot large (about 15%, as shown in Table 2), as men-tioned above. Figure 7b indicates the major contributorto orographic rainfall. As shown in Figs. 7c and 7d,despite the smaller number of events, the 2–5-h cate-gory, as well as the even rarer longer duration category,contributed relatively large amounts to orographic rain-fall.

2) RAINFALL INTENSITY, FREQUENCY, AND

AMOUNT RELATIONSHIPS

Table 3 shows the average number of rainfall eventsand the average rainfall amount for the four intensity-related categories, less than 1, 1–5, 5–11, and over 11mm h21 from June to October 1998 and 1999. Over90% of the total rainfall events had rainfall intensitiesweaker than 5 mm h21. Intensities weaker than 1 mmh21 contributed a large number of events (50.4% in1998; 52.7% in 1999), but only a small accumulation(8.5% in 1998; 9.6% in 1999).

One can suspect that some part of this frequent verylight rain (1 mm h21), is due to dew formation or fogdrips. However, it is generally very difficult to catchdewdrops and fog drips by 0.5 mm per tip rain gauge.Fog moves horizontally, and forest canopy can catch itand make drips. But, a rain gauge, which has a simpleshape, could not be expected to form them. Convectiverains in the evening that quickly pass by a rain gaugewould explain some parts of the light rainfall events.

The 1–5 mm h21 intensity category was responsiblefor the greatest rainfall amounts (54.1% in 1998; 58.0%in 1999). While the number of rainfall events with great-er than 5 mm h21 intensities was very small (about 7%–


FIG. 6. Relationship between elevation and the frequency of each rainfall duration category from Jun to Oct1998 and 1999.

8%), those rainfall events contributed a relatively largeamount of rainfall (over 30%).

Figure 8 shows the relationships between elevationand rainfall events divided into each rainfall intensitycategory. As shown in Figs. 8a and 8b, elevation andthe number of relatively weak rainfall events correlatedwell. On the other hand, Figs. 8c and 8d show thatrelatively strong intensity rainfall events showed almostno association with elevation. Only 1999 had weaklycorrelated events (Fig. 8c). The year 1999 had approx-imately twice as many relatively weak intensity rainfallevents as 1998 (see Figs. 8a and 8b). As Fig. 8c indi-cates, 1999 had 1.5 times as many comparatively strongrainfall events, though both 1999 and 1998 data showeda little correlation between stronger rain intensities andelevation. There were no significant differences in thenumber of strong intensity rainfall events between 1998and 1999, as shown in Fig. 8d.

Figure 9 illustrates elevation and rainfall amount re-lationships by rainfall intensity category. As shown inFig. 9a and Table 3, events with less than 1 mm h21

rainfall intensity correlated comparatively well with el-evation. However, 1 mm h21 rainfall intensity eventscontributed little to high elevation rainfall (less than10%), despite the large number of these events (about

50% of the total). The 1–5 mm h21 rainfall intensitycategory also correlated well with elevation and was amajor contributor to orographic rainfall (see Fig. 9b).The 1–5 mm h21 category contributed over half therainfall and had a large increment in the linear regression[33.3 mm (100 m)21 in 1998 and 53.9 mm (100 m)21

in 1999]. As shown in Fig. 9c, 1999 strongly contributedto orographic rainfall; however, 1998 did not [9.4 mm(100 m)21 in 1998 and 28.7 mm (100 m)21 in 1999].Figure 9d indicates that the stronger intensity rainfallevents did not contribute to orographic rainfall and hada small increment in the linear regression [less than 2.5mm (100 m)21]. Although only about 7%–8% of therainfall events had an intensity greater than 5 mm h21,these rainfall events accounted for a relatively large por-tion of each station’s rainfall (over 30%). The 1–5 and5–11 mm h21 rain intensity category showed clear dif-ferences between 1998 and 1999, as shown in Fig. 9.

Rainfall events with 1–5 mm h21 intensities occurredmost frequently and correlated best with elevation.Strong intensity rainfall events (e.g., over 5 mm h21)were very limited in the number of events. Except forthe 5–11 mm h21 rain intensity category in 1999, thelarge contribution by these events did not correlate withelevation.


FIG. 7. Relationship between elevation and the rainfall amount of each rainfall duration category from Jun to Oct1998 and 1999.

TABLE 3. Average number of events, ratio of average number of events to the total, average rainfall amount, ratio of average rainfallamount to the total, increment and coefficient of determination of the linear regression between elevation and frequency, and increment andcoefficient of determination of the linear regression between elevation and rainfall amount for each rainfall intensity (I) category from Junto Oct 1998 and 1999.

Rainfall intensity (mm h21) I , 1 1 % I , 5 5 % I , 11 11 % I

Average No. of events

Ratio of No. of events to total (%)

1998199919981999

92.8147.8

50.452.7

76.8113.2

41.740.3

11.316.2

6.15.8

3.33.51.81.2

Average rainfall amount (mm)

Ratio of rainfall amount to total (%)

1998199919981999

73.6121.3

8.59.6

470.8736.1

54.158.0

218.7323.0

25.125.4

106.689.212.3

7.0Increment of linear regression

(No. of events per 100 m)Coefficient of determination (R2)

1998199919981999

4.56.30.77*0.63*

3.45.10.77*0.68*

0.40.80.280.40**

0.10.10.130.02

Increment of linear regression[Rainfall amount (mm) per 100 m]


1998199919981999

3.65.40.65*0.52*

33.353.90.86*0.79*

9.428.70.36**0.56*

1.62.20.010.06

* 1% significance level.** 5% significance level.


FIG. 8. Relationship between elevation and the frequency of each rainfall intensity category from Jun to Oct1998 and 1999.

3) RAINFALL DURATION, INTENSITY, FREQUENCY,AND AMOUNT RELATIONSHIPS

Clearly, the data show strong relationships betweenelevation and the number of events, regardless of rainfallduration (Fig. 6). Relatively long rainfall durationevents were responsible for orographic rainfall, as mea-sured in rainfall amounts (Fig. 7). Moreover, the numberof weak rainfall intensity events (less than 5 mm h21)significantly correlated with elevation (Figs. 8a and 8b).On the other hand, the number of strong rainfall inten-sity events (over 5 mm h21) was relatively independentof elevation (Figs. 8c and 8d). Weak rainfall events thuscontributed the most to orographic rainfall amounts(Fig. 9). Both 1998 and 1999 showed similar charac-teristics. Based on these facts, we speculate that rela-tively weak rainfall intensity events could cause a low-altitude bias in precipitation estimates (spatial samplingerror). On the other hand, strong rainfall intensity eventsthat might relate to incoming storms and large-scaleweather systems would not cause it.

These results expand on our previous research (Dair-aku et al. 2000) and prior knowledge of orographic rain-fall structures in the Mae Chaem watershed; rainfallamount enhancement was linked with altitude in the

Mae Chaem watershed, and correlations were foundwith total rainfall durations, but the mean rainfall in-tensity was found to be independent. In this paper, ourresearch revealed more detailed orographic rainfallstructures on an event basis; that is, rainfall events occurfrequently and tend to be of long duration (regardlessof rainfall intensity) at relatively high altitudes in thewatershed.

4. Conclusions

As one of the GAME-T research activities, a denseground-based rain gauge observation network was es-tablished in a northwest Thailand watershed. Gauge ob-servations revealed higher rainfall at higher elevationstations in 1998 and 1999. Observations have tradition-ally been made only at low elevations and in populatedregions, causing one of the errors in estimating arealprecipitation. Since the final quality of the merged anal-ysis is primarily determined by the quality of the inputdata sources, it is necessary to deepen our understandingof bias and error structures in the individual data sourcesand to better understand rainfall characteristics.

The present study investigated the temporal and spa-


FIG. 9. Relationship between elevation and the rainfall amount of each rainfall intensity category from Jun to Oct1998 and 1999.

tial distribution, which is one of the important indicesreflecting regional hydroclimate conditions. Relatingrainfall amount, intensity, duration, and frequency toelevation in the Mae Chaem watershed indicated thefollowing:

1) A strong relationship existed between elevation andthe number of events, regardless of an event’s rainfallduration; relatively long rainfall duration eventswere responsible for orographic rainfall, as measuredin rainfall amount.

2) The number of weak rainfall intensity events sig-nificantly correlated with elevation, while the num-ber of comparatively strong rainfall intensity eventswas relatively independent of elevation.

3) Relatively weak intensity rainfall events contributedthe most to orographic rainfall amounts.

In conclusion, orographic rainfall characteristics canbe attributed to the higher frequency of relatively weakrainfall intensity events at high elevations. These eventshave longer durations than the stronger rainfall intensityevents. These phenomena were common in 1998 and1999, despite large variation in rainfall amounts be-tween the two years.

During the rainy season, monsoonal moist westerly

wind is predominant in the lower troposphere. We pre-sume that the monsoonal background wind and land-form rather than synoptic disturbances induce a dynamicorographic upward motion. Therefore convective cloudsfrequently form in the mountain ranges. The cloud basein the season is around 1.5–2 km above sea level, whichis as high as mountains in this region. That is possiblyone of the causes of relatively weak and frequent pre-cipitation in the mountain. However, the relationshipsand associated mechanisms of orographic rainfall char-acteristics revealed in this study have not yet been thor-oughly explained. How do local circulations such asmountain-to-valley circulations interact with Asianmonsoon circulations? Questions such as this requirefurther investigation using satellite data for cloud ac-tivities, reanalysis atmospheric data, and numerical ex-periments that can model regional climates usingGAME-T gauge observations. Nonetheless, the newlyobtained rainfall data and proposed analysis method al-low for a better understanding of the tropical mountainclimate and its rainfall characteristics and can signifi-cantly impact future studies if long-term observationsare continued and then applied to other climate regions.

Acknowledgments. This study was carried out as oneof the GEWEX/GAME activities supported by Grant


11201202 (representative: Prof. Katumi Musiake),‘‘Modeling Global Hydrological Cycles and World Wa-ter Resources Coupled with Human Activities’’ underthe ‘‘Research and Development of Hydrological Sys-tem Modeling and Water Resources System’’ of the CoreResearch for Evolutional Science and Technology(CREST) of the Japan Science and Technology Cor-poration (JST), as well as by ‘‘Water Resources andVariability in Asia in the 21st Century’’ of the ResearchFund to Promote Science and Technology of the Min-istry of Education, Culture, Sports, Science, and Tech-nology (MEXT) supervised by Dr. A. Kitoh at the Me-teorological Research Institute of Japan MeteorologicalAgency, and by the No. 5 project of the Research In-stitute for Humanity and Nature, ‘‘Integrated Manage-ment System for Water Issues for a Global Environ-mental Information Library and World Water Model,’’and a grant-in-aid for JSPS Fellows from the Ministryof Education, Culture, Sports, Science, and Technology,Japan. The research was directed by a GAME-Tropicssubcommittee.

Rainfall observation fieldwork in the Mae Chaem wa-tershed was conducted by Prof. Suzuki and Dr. Kurajiin cooperation with the Royal Forestry Department andKasetsart University, Thailand. The authors gratefullythank Dr. Yasushi Agata for helping to arrange theGAME-T observation data and Prof. Jun Matsumoto,Prof. Takehiko Satomura, Mr. Thada Sukhapunnaphan,and Prof. Manabu D. Yamanaka for their advice. Alsothanks to three anonymous reviewers whose commentsand suggestions helped to improve this paper. We usedparameter subsettings of TRMM level 3 standard prod-ucts, processed and provided by the National Aeronau-tics and Space Administration and the National SpaceDevelopment Agency of Japan for analyses.

REFERENCES

Adler, R. F., G. J. Huffman, D. T. Bolvin, S. Curtis, and E. J. Nelkin,2000: Tropical rainfall distributions determined using TRMMcombined with other satellite and rain gauge information. J.Appl. Meteor., 39, 2007–2023.

Anyamba, A., C. J. Tucker, and R. Mahoney, 2002: From El Ninoto La Nina: Vegetation response patterns over east and southernAfrica during the 1997–2000 period. J. Climate, 15, 3096–3103.

Barros, A. P., and R. J. Kuligowski, 1998: Orographic effects duringa severe wintertime rainstorm in the Appalachian Mountains.Mon. Wea. Rev., 126, 2648–2672.

——, and T. J. Lang, 2003: Monitoring the monsoon in the Himalayas:Observations in central Nepal, June 2001. Mon. Wea. Rev., 131,1408–1427.

——, M. Joshi, J. Putkonen, and D. W. Burbank, 2000: A study ofthe 1999 monsoon rainfall in a mountainous region in centralNepal using TRMM products and rain gauge observations. Geo-phys. Res. Lett., 27, 3683–3686.

——, G. Kim, E. Williams, and S. W. Nesbitt, 2003: Probing oro-graphic controls in the Himalayas during the monsoon usingsatellite imagery. Nat. Hazards Earth Syst. Sci., in press.

Coe, M. T., 2000: Modeling terrestrial hydrological systems at thecontinental scale: Testing the accuracy of an atmospheric GCM.J. Climate, 13, 686–704.

Dairaku, K., K. Kuraji, M. Suzuki, N. Tangtham, W. Jirasuktaveekul,

and K. Punyatrong, 2000: The effect of rainfall duration andintensity on orographic rainfall enhancement in a mountainousarea: A case study in the Mae Chaem watershed, Thailand. J.Japan Soc. Hydrol. Water Resour., 13, 57–68.

Dettinger, M. D., and H. F. Diaz, 2000: Global characteristics ofstream flow. seasonality and variability. J. Hydrometeor., 1, 289–310.

Golubev, V. S., P. Y. Groisman, and R. G. Quayle, 1992: An Eval-uation of the U.S. standard 8-in. nonrecording raingage at theValdai Polygon, USSR. J. Atmos. Oceanic Technol., 9, 624–629.

Huffman, G. J., R. F. Adler, B. Rudolf, U. Schneider, and P. R. Keehn,1995: Global precipitation estimates based on a technique forcombining satellite-based estimates, rain gauge analysis, andNWP model precipitation information. J. Climate, 8, 1284–1295.

——, and Coauthors, 1997: The Global Precipitation ClimatologyProject (GPCP) Combined Precipitation Dataset. Bull. Amer. Me-teor. Soc., 78, 5–20.

Koster, R. D., M. J. Suarez, and M. Heiser, 2000: Variance and pre-dictability of precipitation at seasonal-to-interannual timescales.J. Hydrometeor., 1, 26–46.

Kuraji, K., K. Punyatrong, and M. Suzuki, 2001: Altitudinal increasein rainfall in the Mae Chaem watershed, Thailand. J. Meteor.Soc. Japan, 79, 353–363.

Kyriakidis, P. C., J. Kim, and N. L. Miller, 2001: Geostatistical map-ping of precipitation from rain gauge data using atmospheric andterrain characteristics. J. Appl. Meteor., 40, 1855–1877.

Lang, T. J., and A. P. Barros, 2002: An investigation of the onset ofthe 1999 and 2000 monsoons in central Nepal. Mon. Wea. Rev.,130, 1299–1316.

Milly, P. C. D., R. T. Wetherald, K. A. Dunne, and T. L. Delworth,2002: Increasing risk of great floods in a changing climate. Na-ture, 415, 514–517.

Miniscloux, F., J. D. Creutin, and S. Anquetin, 2001: Geostatisticalanalysis of orographic rainbands. J. Appl. Meteor., 40, 1835–1854.

Neff, E. L., 1977: How much rain does a rain gage gage? J. Hydrol.,35, 213–220.

Neiman, P. J., F. M. Ralph, A. B. White, D. E. Kingsmill, and P. O.G. Persson, 2002: The statistical relationship between upslopeflow and rainfall in California’s coastal mountains: Observationsduring CALJET. Mon. Wea. Rev., 130, 1468–1492.

Oki, T., 2001: Modeling surface hydrology for global water cyclesimulations. Present and Future of Modeling Global Environ-mental Change: Toward Integrated Modeling, T. Matsuno, andH. Kida, Eds., Terra Scientific, 391–403.

——, and K. Musiake, 1994: Seasonal change of the diurnal cycleof precipitation over Japan and Malaysia. J. Appl. Meteor., 33,1445–1463.

Palmer, T. N., and J. Ralsanen, 2002: Quantifying the risk of extremeseasonal precipitation events in a changing climate. Nature, 415,512–514.

Phillips, D. L., J. Dolph, and D. Marks, 1992: A comparison ofgeostatistical procedures for spatial analysis of precipitation inmountainous terrain. Agric. For. Meteor., 58, 119–141.

Purdy, J. C., D. Harris, G. L. Austin, A. W. Seed, and W. Gray, 2001:A case study of orographic rainfall processes incorporating mul-tiscaling characterization techniques. J. Geophys. Res., 106,7837–7845.

Ramage, C. S., 1964: Diurnal variation of summer rainfall of Malaya.J. Trop. Geogr., 19, 62–68.

Rudolf, B., H. Hauschild, W. Rueth, and U. Schneider, 1994: Ter-restrial precipitation analyses: Operational method and requireddensity of point measurements. Global Precipitations and Cli-mate Change, M. Desbois and F. Desalmand, Eds., Springer-Verlag, 173–186.

Schnur, R., 2002: The investment forecast. Nature, 415, 483–484.Shuin, Y., H. Shibano, M. Suzuki, and T. Ohta, 1996: Temporal and

spatial characteristics of rainfall at a south-west slope of volcanicMt. Merapi in Indonesia. J. Japan. Soc. Erosion Control Eng.,48, 3–12.


Smith, J. A., 1992: Precipitation. Handbook of Hydrology, D. R.Maidment, Ed., McGraw-Hill, 83–130.

Smith, R. B., 1986: Current status of ALPEX research in the UnitedStates. Bull. Amer. Meteor. Soc., 67, 310–318.

Stow, D., S. Bradley, and W. Gray, 1991: A preliminary investigationof orographic rainfall enhancement over low hills near AucklandNew Zealand. J. Meteor. Soc. Japan, 69, 489–495.

Thauvin, V., and T. Lebel, 1991: IV: EPSAT-NIGER study of rainfallover the Sahel at small time steps using a dense network ofrecording raingauges. Hydrol. Processes, 5, 251–260.

Ueno, K., N. Endoh, T. Ohata, H. Yabuki, T. Koike, M. Koike, T.Ohta, and Z. Yinsheng, 1994: Characteristics of precipitationdistribution in Tanggula, Monsoon, 1993. Bull. Glacier Res., 12,39–47.

——, H. Fujii, H. Yamada, and L. Liu, 2001: Weak and frequentmonsoon precipitation over the Tibetan Plateau. J. Meteor. Soc.Japan, 79, 419–434.

Vorosmarty, C. J., P. Green, J. Salisbury, and R. B. Lammers, 2000:

Global water resources: Vulnerability from climate change andpopulation growth. Science, 289, 284–288.

Weisse, A. K., and P. Bois, 2001: Topographic effects on statisticalcharacteristics of heavy rainfall and mapping in the French Alps.J. Appl. Meteor., 40, 720–740.

Wratt, D. S., and Coauthors, 1996: The New Zealand Southern AlpsExperiment. Bull. Amer. Meteor. Soc., 77, 683–692.

Xie, P., and P. A. Arkin, 1997: Global precipitation: A 17-year monthlyanalysis based on gauge observations, satellite estimates, and nu-merical model outputs. Bull. Amer. Meteor. Soc., 78, 2539–2558.

Yamada, T., T. Hibino, T. Araki, and M. Nakatsugawa, 1995: Statis-tical characteristics of rainfall in mountainous basins (in Japa-nese with English summary). J. Hydraul. Coastal Environ. Eng.,527, 1–13.

Yang, D., B. E. Goodison, S. Ishida, and C. S. Benson, 1998: Ad-justment of daily precipitation data at ten climate stations inAlaska: Application of World Meteorological Organization in-tercomparison results. Water Resour. Res., 34, 241–256.

Zwiers, F. W., 2002: The 20-year forecast. Nature, 416, 690–691.

Rainfall Amount, Intensity, Duration, and Frequency Relationships in the Mae Chaem Watershed in...

Documents

Transcript of Rainfall Amount, Intensity, Duration, and Frequency Relationships in the Mae Chaem Watershed in...