AN EVALUATION OF BLUE WATER PREDICTION IN SOUTHERN PART OF IRAN USING SWAT

Environmental Engineering and Management Journal January 2016, Vol.15, No. 1, 175-188

http://omicron.ch.tuiasi.ro/EEMJ/

“Gheorghe Asachi” Technical University of Iasi, Romania

AN EVALUATION OF BLUE WATER PREDICTION

IN SOUTHERN PART OF IRAN

USING THE SOIL AND WATER ASSESSMENT TOOL (SWAT)

Milad Jajarmizadeh1, Sobri Harun1, Rozi Abdullah2, Mohsen Salarpour1

1Department of Hydraulic and Hydrology, Universiti Teknologi Malaysia, 81310 Johor, Malaysia 2School of Civil Engineering, Universiti Sains Malaysia

Abstract Soil and water are the two major resources in the Earth’s hydro biological and geological systems. The hydrology of arid areas has become a topic of interest recently for hydrologists as water shortage at these areas can affect the agriculture, irrigation, and industry as a whole. This has also prompted water resource planners to more thoroughly investigate water resource crisis at arid areas. In this respect, the Soil and Water Assessment Tool (SWAT), a semi-distributed hydrological model, can be a subsidiary tool to be used in the prediction of surface runoff (blue water). This paper presents the application of SWAT on the Roodan watershed, which is located in the southern part of Iran and has 215 mm of annual precipitation. SWAT was engaged to know more about the daily flow and to evaluate the runoff volume. Three continuous scenarios were defined over the 21 years (1988-1992, 1993-2001, 2002-2008) for the land use map as it was found that continuous update of this layer were basically done during these periods. Results of sensitivity analysis showed that parameters related to transmission losses are most sensitive for this watershed. Furthermore, the SWAT had also visualized from the input data that the sub basins which have been designated for agricultural activities from 1988 to 2008 were at the southwest, center and northeast parts of Roodan watershed. Strength of modeling was evaluated by percentage of observations covered by the 95 Percentage prediction uncertainty (P-factor) and relative width of 95 % probability band (R-factor). The P and R factors in this study were recorded at, for calibration and validation periods, 50 % and 0.18 (calibration), and 50 % and 0.17 (validation) respectively. Nash-Sutcliffe and PBIAS obtained for calibration period were 0.75 and 1.5 %, and those for validation period were 0.64 and 21 %. However, results showed an underestimation trend for most peak flows during the modeling of daily stream flow. Nevertheless, the annual runoff volume for calibration and validation periods depicted a promising performance and thus validated the usage of SWAT as a subsidiary hydrological tool for water management projects attributed with stream flow and runoff volume. Key words: arid area, runoff volume, stream flow, SWAT Received: April, 2012; Revised final: December, 2012; Accepted: January, 2013

Author to whom all correspondence should be addressed: e-mail: [email protected]; Phone: +60173135142

1. Introduction Sustainable development of watersheds is an

important task especially for low rainfall areas (arid regions) of the world. Clarke (1993) stated that watershed problems are caused by degradation of the natural environments, anthropogenic activities, and climate change. These have resulted in more frequent occurrence of flood, drought, and water pollution. Balon and Dehnad (2006) reported that in Iran, a

large proportion of fresh water is used for the agriculture field. Therefore, there is a need to employ better techniques to assist in water resource management, for example delineation of watersheds.

A model is an expression of a portion of the natural or human constructed world and it can primarily be classified as physical, analog, or mathematical (Brooks et al., 1991; Dingman, 2002). Mathematical models had been employed for hydrological applications since the second half of the

Jajarmizadeh et al./Environmental Engineering and Management Journal 15 (2016), 1, 175-188

176

nineteenth century to estimate hydrological processes such as maximum flow and surface runoff. Today, hydrologists use computer technologies to provide automated application of existing analytical methods and to explore new theories (Shaw, 1994). In fact, hydrological models have now become a popular tool in the soil and water management field. Thompson et al. (2004) and Singh and Frevert (2006) pointed out such models are powerful for their fast processing speed, low evaluation cost for different management strategies, and level of protection from undesirable outcomes.

The development of semi-distributed models as a tradeoff between lumped and fully distributed models is still relatively in the hydrology field (Arnolds et al., 1993). The model algorithms are simple but physically based and in these models, the spatial heterogeneity is considered by means of observable physical characteristics of the catchment such as land use, soils and topography. Semi-distributed models are capable of predicting the average behavior of a catchment using small homogeneous units, and then aggregating them into a few predefined positions (Wilby, 1997). A semi-distributed model is advantageous in the sense that it presents a basin using average features, and this is one of its distinctive features from a lumped model. In addition, it is less data demanding than distributed models.

For a decade, the Soil and Water Assessment Tool (SWAT) has been applied to study and simulate various subjects such as climate variance and land-cover changes (Glavan et al., 2013; Junfeng et al., 2005; Wang et al., 2008) runoff and sediment (Tripathi et al., 2004; Xu et al., 2009); and nutrient loading (Huang et al., 2009). Some studies have also employed SWAT to run hydrological analysis (Kannan et al., 2007) and for applications in large scale basins (Schuol et al., 2008).

Influence of land use/ land management scenario on water quality and prediction of blue and green water has become an important subject recently, particularly for agricultural area (Glavan et al., 2013; Volk et al., 2009) a comprehensive review on related published literature can be found in Gassman et al. (2007). In Iran, the most recent studies that utilized SWAT include those done by Ghaffari et al. (2010, 2011), Faramarzi et al. (2009), Rostamian et al. (2008), and Masih et al. (2011). On the hand, in this study, the SWAT model was applied to generate daily stream flow and extract runoff volume for the Roodan watershed. Located in the southern part of Iran, this watershed basically has low precipitation, but it is also one of the main catchments in the Hormozgan province. This watershed is potentially useful for agricultural activities and farming. From an economy perspective, the watershed may prosper from animal husbandry, orchard farm, and even tourism. However, the fact that this watershed often suffers from shortage of water resource management during flood and drought season has hindered the full

development of this watershed (Ab Rah Saz Shargh, 2009; Morid et al., 2001). Therefore, this simulation is timely to gain more cognitive insights into the behavior of surface water resources at the Roodan watershed. Other studies have also been done on other parts of Iran, but these studies are conducted to either evaluate the impact of land use changes (Ghaffari et al., 2010), effect of accuracy in the input maps on SWAT modeling (Ghaffari et al., 2011), or to simulate blue water (water yield plus deep aquifer recharge), green water flow (actual and potential evapotranspiration) (Faramarzi et al., 2009) and sediment yield (Rostamian et al., 2008).

The main objective of this study was to simulate the daily runoff (blue water) at the Roodan watershed using SWAT. The annual runoff volume generated from the model will also be discussed. 2. Methodology 2.1. Case study

The Roodan watershed lies between the

Hormozgan and Kerman provinces in Iran (in the range of 26o57’ - 28o31’N and 56o47’ – 57o54’E) and covers an area of 10570 km2. The mean elevation is 781 m above sea level (Fig. 1). Generally, the watershed is mountainous in the north and east direction and plain in the center and southern part. For the period of 1978 to 2008, the average annual precipitation was 215 mm (Ab Rah Saz Shargh, 2009). The heaviest precipitation occurred from October to March and warmer months had no substantial precipitation. For the same period, the mean daily temperature was 25 0C. The predominant soil type in the northern part and center is a heterogeneous mix of clay, silt and sand. Lands in Roodan can be divided into shrub land (range brush), mix grassland with shrub land, rocky terrain, lands used for irrigated agriculture and orchard farms, and urban area (cities and villages). Generally, the climate of Roodan is arid with short and high intensity rainfall. In addition, there is a dam - Esteghlal Dam located at the outlet of the watershed, this dam holds an important role in collecting surface water for development of downstream area. 2.2. Model description

The development of SWAT version 2009 is a continuation of the United States Department of Agriculture (USDA) Agricultural Research Service (ARS) modeling experience that has spanned over a period of roughly 30 years (Krysanova and Arnold, 2009). SWAT provides continuous time simulations with a high level of spatial detail by allowing further division of a watershed (Arnold et al., 1998). The model functions based on daily time step (Neitsch et al., 2005) and its main objective is to predict the impact of land management practices on water, sediment, and agricultural chemical yields in large and complex watersheds with varying soils, land uses

An evaluation of blue water prediction in Southern part of Iran using SWAT

177

and management conditions over long periods of time. SWAT simulates the hydrologic cycle of a sub-basin using water balance as presented in Eq. (1):

)(1

0 gwseepasurf

t

idayt QWEQRSWSW

(1)

where SWt is the final water content (mm), SW0 is the initial water content on day i (mm), t is time (day), Rday is amount of precipitation on day i (mm), Qsurf is the amount of runoff on day i (mm), Ea is the amount of evapotransporation on day i (mm), Wseep is the amount of water entering the vadose zone from the soil profile on day i (mm), and Qgw is the amount of return flow on day i (mm).

The most important feature of SWAT is its hydrological response units (HRUs), which are used to explain spatial heterogeneity in terms of land cover, soil type and slope within the catchment (Freeze and Cherry, 1979). As a continuous time model, SWAT has been used for the study of long term results on large and complex basins in single gauge watershed and multiple gauge watershed (Ghaffari et al., 2010; Schuol et al., 2008). The major advances and reviews for SWAT model can be found in Van Griensven and Meixner (2003); Abbaspour and Yang (2006); and Kim et al. (2007). A fully detailed description on SWAT model can be found in Neitsch et al. (2005). The Environmental Protection Agency (EPA) has suggested SWAT model for assessment of watershed management in various fields such as for rainfall-runoff modeling, land use change, sedimentation, and erosion (Lenhart et al., 2002).

2.3. Land use update module

The SWAT model launches the land use update file (lup.dat) to consider the relevant updates during the period of simulation (Winchell et al., 2010).

Lup.dat is an optional file that contains the information on development and land uses in a sub-basin, these may include specified change or both original and targeted (new) land uses. Lup.dat allows HRU fraction to be updated during a simulation run by definition of percentage of changes. This file operates on HRUs for desired dates and remains effective for the remaining simulation until the next update.

Thus, this lup.dat file is obviously useful to initialize conservative measures in mid-simulation. With this optional procedure, the reliability of the model of any given dates can be further enhanced (Neitsch et al., 2010). A comprehensive explanation on this procedure can be found in Pai and Saraswat (2011). 2.4. Required data 2.4.1. Digital elevation modeling (DEM)

DEM is prepared for Roodan watershed with a resolution of 90 m from a number of topography maps scaled at 1:25000 available from the Iran topography organization. A mesh size map which is between 50-90 m resolutions is sufficient for SWAT model (Chaplot, 2005; FitzHugh and Mackay, 2000). A more accurate hydrographic segmentation and sub-watershed boundary were obtained in this study by using a digital river layer of the Roodan watershed and specifying the optimal minimum sub-watershed area (Arabi et al., 2006; Biesbrouck et al., 2002).

2.4.2. Soil map

The Food and Agriculture Organization (FAO) soil map, which provides data for 5000 types of soil (Reynolds et al., 1999; Schuol et al., 2008; Faramarzi et al., 2009; Setegn et al., 2010) was used to generate the soil input file for the Roodan watershed (FAO, 1995).

Fig. 1. Location of Roodan watershed in Iran according to Ab Rah Saz Shargh (2009)


178

In this study, in order to obtain a better SWAT

model, the soil map was prepared using a geology map with an accuracy of 1:25000. Soil samples obtained from case studies and in collaboration with the agricultural organization of the Hormozgan province in Iran were also used for the development of this soil map.

2.4.3. Land use map

The land use map of this Roodan watershed was prepared in accordance to a satellite image, available current land use map (1:25000), statistics on the development of agricultural areas from 2002 to 2008 from the agriculture organization of Hormozgan, Iran, and site visit to various parts of the case study area in year 2007 and 2008. Previous land uses for the periods of 1988-1992 and 1993-2001 were collected through available land use maps (1:25000) and statistics of agricultural area from the same organization.

The land use map portrays major changes in the agricultural area for the past 21 years (total changes are approximately 1.1 % of total watershed area). Accordingly, the land use update input table (LupInput.dat) was generated by converting the original land use (agricultural area) to targeted land use, namely shrub-land area (SHRB) and mix grass-land/shrub-land area (MIGS) by definition of alternation percentage on given dates in year 1988, 1993 and 2002. SWAT allows the HRUs fraction to convert from original to targeted land use by considering the percentage of changes made. In this study, the HRUs fraction changed from SHRB and MIGS to agricultural area from 1988 to 2008. This means that the SWAT model had updated and changed the land use map from 1988 to 2008 in the middle of a continuous model simulation.

2.4.4. Weather stations and river discharge gauge

The data were obtained from the Meteorological Organization of Iran (IRIMO). The meteorological data collected included daily precipitation and maximum and minimum air temperature from 1988 to 2008. Similarly, the daily discharge from 1988-2008 was recorded from the discharge gauge located at the outlet of the Roodan watershed, which is named as “Barantin-Minab”.

2.4.5. Model set-up

SWAT version 2009 was used for the simulation of Roodan watershed. Generally, the watershed simulation involved five major steps - DEM set-up; stream definition; outlet and inlet definition; watershed outlets selection and definition; and calculation of sub-basin parameters. The land use dataset together with the soil and slope dataset were then imported, overlaid and linked to the SWAT database. Slope classes are particularly important if sub-basins have a wide range of slopes within HRUs (Winchell et al., 2010). For this watershed, three slope classifications were defined, namely 0 to 5 %, 5

to 20 % and more than 20%. Subsequently, the low value threshold was set at 5 % for the land use, soil and slope distribution in the HRU definition stage (Raneesh et al., 2010). After that, the watershed was divided into 513 HRUs for the whole catchment and 45 sub-basins. Hargreaves method was chosen for calculation of potential evapotranspiration and the SCS Curve Number was used for calculation of runoff volume in accordance with antecedent soil moisture condition type II.

Due to the aridity of the case study area, the reach evaporation coefficient (EVRCH.bsn) was adjusted based on measured value of daily stream flow and by considering the contribution of base flow to main channels. Furthermore, it seems that the reach evaporation adjustment factor in the original estimation using SWAT tends to overestimate evaporation from arid lands, and therefore Neitsch et al. (2010) has recommended that this factor has to be significantly adjusted for arid areas. Finally, the prepared model was run using a year of warm-up period. Table 1 shows the required data for the setting up of the Roodan watershed.

2.4.6. Land-use updates (lup.dat)

The land uses in this watershed has been aforementioned, so as the importance of incorporating the lup.dat file in the simulation using SWAT. 21 sub-basins were identified to include the original land use (agricultural area) and targeted land uses (shrub-land or shrub land/grassland). Changes are identified when there are significant alterations in the original and targeted land uses in a catchment (Winchell et al., 2010). For example from agricultural lands (original land use) to shrub-land/grassland (targeted land used), and this is the way the HRUs in each sub-basin, especially those that had the same soil features and slope classes, are merged (or combined) for the year 1988, 1993, and 2002. Land use change has been shown for Roodan watershed (Fig.2).

The SWAT model used in this study also visualized that most of the sub-basins that had been developed primarily for agricultural activities from 1988 to 2008 were in the southwest, center and northeast areas of the Roodan watershed, as shown in Fig. 3. This is in accordance to the percentages of changing agricultural area, which feed in database of SWAT Roodan watershed (Fig. 3).

2.5. SUFI-2 algorithm

Calibration and validation of semi-distributed

models such as SWAT can be a tough task for modelers. The reason is that a model has many parameters for adjustment. Usually, there are two approaches for calibration of SWAT (i) manually calibration and (ii) SWAT-CUP (calibration and uncertainty procedures) program, which is a calibration and uncertainty analysis tool (Abbaspour et al., 2007). In this research, the SWAT-CUP


179

program was used together with the Sequential Uncertainty Fitting (SUFI-2) algorithm.

SUFI-2 is a type of inverse modeling (IM) algorithm that is used for calibration and validation of hydrological models, including SWAT. In SUFI-2, all sources of uncertainties are identified and these include uncertainty in driving variables, conceptual model, parameters, and measured data. There are two main factors that are considered in the index of evaluation for the results; firstly, the degree to which all uncertainties are accounted for (quantified by a measure referred to as the P-factor) and is expressed

in percentage of measured data bracketed by the 95% prediction uncertainty (95PPU). The 95PPU in this research was calculated at the 2.5 % and 97.5% levels of cumulative distribution for an output variable obtained through Latin Hypercube sampling.

Another factor that shows the strength of a calibration and uncertainty analysis is the R-factor, which is the mean thickness of the 95PPU band divided by the standard deviation of the observed data. In theory, the value for P-factor ranges between 0 to 100 %, while that of R factor should be between nil to infinity.

Table 1. Applied data for SWAT setup in Roodan watershed

Data type Scale Source Data description Topography 1:25000 Iran Survey Organization Elevation, Counter line ،Latitude and Altitude,

Channel slopes and Lengths Digital river network 1:25000 Iran Survey Organization Main and tributary river in Roodan basin Soil map 1:25000 AbRah Saz Shargh, (2009),

and Hormozgan Agricultural Organization

Attributed data with soil physical & chemical properties

Land use 1:25000 Ab Rah Saz Shargh (2009), and Hormozgan Agricultural Organization

Land cover classification 7 land use classes: agriculture area, orchard, range grass, range shrub, rock, city and village

Statistics of developing agriculture area

--- Hormozgan Agricultural Organization

Increasing agricultural area during the periods of 1988-1992, 1993-2001, and 2002-2008

Weather 12 stations (daily rainfall) 5 station (temperature)

Ministry of power, and IRIMO Daily precipitation and temperature

Fig. 2. Land use change for Roodan watershed (Ab Rah Saz Shargh, 2009)


180

Fig. 3. Visualized developing agricultural area in the Roodan sub-basins using SWAT (1988-2008)

A P factor of one and R factor of zero shows a

simulation that corresponds exactly to observed data and this is the ideal case. However, a modeler should seek to have the best solution for the modeling, not an idealistic simulation (Abbaspour et al., 1997, 2004, 2007). Usually, besides the R and P factors, accuracy of a model can be judged by three coefficients, namely the coefficient of determination (R2), Nash-Sutcliffe (NS) and PBIAS (Moriasi et al., 2007) between observed and simulated data. Ideally, R2 equals to one and this shows a complete harmony between measured and simulated data. Meanwhile, NS is acceptable to be ideally one (Krause et al., 2005). Moriasi et al. ( 2007) stated that absolute value of PBIAS that ranges from 15 to 25 shows that the SWAT model is satisfactory, the model is rated as good when the value is from 10 to 15, and very good when it is smaller than 10. The R2, NS and PBIAS values in this study were estimated using Eqs. (2-4) respectively:

2

5.0

1 1

22

12

n

i

n

isimavesimobsaveobs

n

isimavesimobsaveobs

QQQQ

QQQQ

R (2)

where n is the number of data, Qsim and Qobs are the simulated and observed runoff at event i ,Qsimave and

Qobsave are the average simulated and observed runoff over the simulation period. The NS coefficient was calculated as in Eq. (3):

n

i

2aveobs

n

i

2obssim

QQ

QQ1NS (3)

where n is the number data, Qsim and Qobs are the simulated and observed stream flow at time step i, and Qavg is the average observed stream flow over the simulation period. The PBIAS coefficient was calculated as in Eq. (4):

n

iobs

n

isimobs

Q

QQPBIAS

100

(4)

where Qobs and Qsim are measured and predicted values at time step i. PBIAS is an absolute value representing a model’s capability to evaluate the simulated data.

2.5.1. Sensitivity analysis

In this study, 26 hydrological parameters were used for sensitivity analysis, as given in the user’s manual of SWAT 2009 (Kannan et al., 2007; Winchell et al., 2010). The identification of sensitive parameters in Roodan watershed was done using a combination of Latin-hypercube and One factor-At-a-Time sampling schemata. This method is explained in detail by Van Griensven et al. (2006) and Schuol et al. (2008) has stated that it has fast processing speed and is able to rank the order of the parameters. The sensitive parameters that would be used were chosen through the LH-OAT analysis and our cognition on case studies together with available literatures in arid and semi-arid regions (Faramarzi et al., 2009; Ghaffari et al., 2010; Masih et al., 2011). A more in depth sensitivity analysis was performed using SUFI-2 algorithm during the calibration process.

2.5.2. Calibration and validation

The calibration and validation processes of a model are important to develop a reliable model. In this study, the calibration and uncertainty procedures were conducted using the SUFI-2 algorithm. The time period that was used in the calibration process was from January 1989 to December 2002 and the validation period started from January 2003 to December 2008. The year 1988 was defined as the warm period.

3. Results and discussion

3.1. Sensitivity analysis

The result of sensitivity analysis is listed in Table 2. The 12 most sensitive parameters based on the last iteration and the ranking given by SUFI-2


181

algorithm were identified to generate the stream flow model.

Generally, results indicated that the most sensitive parameters were those related to the process of channel routing, ground water, surface runoff, soil properties, and evapotranspiration. The seven most sensitive parameters had their p-value equal to zero and are indicated using bold letters in Table 2. Furthermore, it was found that channel routing parameters played an important role in modeling surface runoff due to their significant impact on transmission losses. Among all these sensitive parameters, both CH_K2 (Effective hydraulic conductivity of main channel) and CH_L2 (length of main channel) were the most sensitive for the Roodan watershed.

However, only calibration of CH_K2 led to better results in the uncertainty analysis and objective functions. On the other hand, adjustment of EVRCH.bsn parameter managed to produce slightly better stream flow in the watershed, but its relative sensitivity in SUFI-2 algorithm was not substantial. The sensitivity of different parameters in Iranian watersheds has also been discussed in other published works before. For example, Faramarzi et al. (2009) and Masih et al. (2011) reported that the curve number, base flow alpha factor, maximum canopy index, surface runoff lag coefficient, available water capacity of the soil layer, soil evaporation compensation factor, and effective

hydraulic conductivity of the main channel are sensitive in the Iran climate. Masih et al. (2011) also discovered that parameters attributed with snowfall are slightly sensitive in the Karkheh basin (a mountainous semi-arid area). On the contrary, we found that in the south of Iran (Roodan watershed), parameters attributed with snowfall are not substantially significant.

3.2. Calibration and validation

It is essential to calibrate a rainfall-runoff model at a basin scale to reduce any possible uncertainties. Uncertainties are outcome of processes that are not considered or have been simplified by the model or processes that the modeler is unfamiliar with (Abbaspour et al., 2007). As mentioned earlier, the strength of a calibration is expressed in terms of the P and R factors as well as R2, NS and PBIAS. Table 3 shows the calibration and validation results for the Roodan watershed.

To obtain desirable result, it is better to seek for a larger P factor, but a smaller R factor using SUFI-2. For calibration, the assumed ideal P-factor is 100 % and R-factor is nil (Abbaspour et al, 2007). In this study, both calibration and validation periods gave satisfactory NS and R2 values, which, at the same time, corresponded to those reported in Oeurng et al. (2011).

Table 2. List of sensitive parameters and their ranking for Roodan watershed in SUFI-2 algorithm

Sensitivity Rank

Parameter Description t-Value* p-Value* Optimal

parameter value 1 **v_CH__K2.rte Effective hydraulic conductivity of

main channel-20.5 0 55.6

2 v_ALPHA__BF.gw Base flow alpha factor 12.66 0 0.9

3 v_CN2.mgt__SHRB SCS runoff curve number for antecedent moisture condition type II for Shrub land

11.4 0 89

4 v_CN2.mgt__MIGS SCS runoff curve number for antecedent moisture condition type II for Mixed Grassland and Shrub land

5.1 0 51

5 ***r_SOL__AWC(1).sol Available water capacity of the soil layer

-5 0 0.11

6 v_ESCO.hru Soil evaporation compensation factor 2.46 0 0.56 7 v_SURLAG.bsn Surface runoff lag coefficient -1.7 0 11 8 v_GWQMN.gw Threshold depth of water in the shallow

aquifer required for return flow to occur 1.2 0.21 3094

9 r_SOL_K(1)..sol Soil conductivity -1 0.28 0.01 10 v_CANMX.hru Maximum canopy index 0.96 0.33 16.6 11 v_CH_N2.rte Manning coefficient for channel -0.95 0.34 0.13 12 v_RCHRG_DP.gw Groundwater recharge to deep aquifer -0.53 0.59 0.42

* t-value and p-value show measure (large absolute value) and significance of sensitivity (close to zero) respectively for each parameter; **v: parameter value is replaced by given value or absolute change; ***r: parameter value is multiplied by (1 + a given value) or relative change (Abbaspour et al., 2007)

Table 3. Criteria for examining the accuracy of calibration (1989-2002) and validation (2003-2008) for daily flow

Index Calibration Validation P-factor % 50 50 R-factor 0.18 0.17

Nash and Sutcliffe coefficient % (NS) 75 64 Correlation coefficient % (R2) 75 64

PBIAS% 1.5 21


182

PBIAS, on the other hand, gave good

calibration results, but the results for the validation period are only satisfactory. Nevertheless, this phenomenon corresponded to that reported in Moriasi et al. (2007) and this may be caused by light development in the watershed (e.g. hydraulic structures, tunnel, and etc.) and moderate usage of water resources. This is because, generally, accuracy of hydrological models decreases in regions that have higher human activities and uncertainty in the hydrological models also always increases in such regions.

The NS and R2 values in daily time scale obtained for stream flow simulation for Pole Chehre (area 10860 km2) as reported in Masih et al. (2011) were 0.75 and 0.82 respectively for a calibration period of around 8 years. On the other hand, in this research, the value for both NS and R2 was 0.75 for a calibration period of around 14 years. In term of the validation period, Masih et al. (2011) reported that the NS and R2 values for the Pole Chehre basin were

0.71 and 0.76 respectively for a time span of seven years. For this research, the NS and R2 values were both0.64 for a validation period of around 7 years. In conclusion, the results depicted an acceptable agreement with those of the Pole Chehre basin.

3.3. Daily Stream flow evaluation

Fig. 4 shows a comparison between observed and simulated stream flow in cubic meter per second (CMS) for calibration (1989-2002) and validation (2003-2008) periods. For both periods, the result showed under-estimation in most peak flows. The largest flow was recorded in the year 1993 according to IRIMO, but its simulation was still under-estimated. In addition, some stream flows were over-estimated such as at the end of the validation period. Figure 4 shows that SWAT is reasonably capable in replicating the observed flow trend. The statistical performance was satisfactory for both periods as shown in Table 3. Meanwhile, Fig. 5 shows the residual error for both calibration and validation periods. Additionally, Fig. 7 and 8 show the result of 95PPU generated as output of the SUFI-2 calculation for calibration and validation periods.

Fig. 6 shows the scatter plot of observed flows versus simulated flows for the Roodan watershed; the SWAT model tends to underestimate between reference line (1:1) and fitted line of total data (between observed and simulated flows) for both calibration and validation periods. Table 4 depicts the percentile absolute error between observed and simulated flows. From the table, it can be seen that 95 % of the data have differences of 20.66 (CMS) in validation period and 43.5 (CMS) for calibration period. Generally, percentiles of absolute error for validation are better than those of calibration.

In comparison with recent studies, Oeurng et al. (2011) found that the fluctuation in peak flows did not fully match the oceanic climate for agricultural

basin. Also, in any case, SWAT cannot simulate flood discharge when the river is overflowing. In another recent study by Nie et al. (2011), the pattern of rainfall had considerable effect on the simulated peak flows.

It seems that using an average rainfall as input data for SWAT model increases uncertainty for simulated stream flow and this may be caused by the limiting function of SWAT where it only allows the model to simulate small precipitation event. In regard to this, Hjelmfelt (1980) reported that most arid regions like the Roodan watershed have short but high intensity precipitation. Therefore, this condition is expected to affect the simulation of peak flows. On the other hand, Hantush and Kalin (2005) reported that over simplification in simulating rainfall-runoff model is one of the sources for underestimation in hydrological models and this effect is more profound in low precipitation period modeling. When a rainfall event is directly transformed to runoff and lateral flows, SWAT gives better stream flow simulation, as supported by Abbaspour et al. (2007). Masih et al. (2011) has also stated that semi-distributed models such as SWAT are sensitive to density of rain gauges for prediction of stream flow in semi-arid areas in Iran (Karkheh Basin). Such under-estimation and over-estimation in stream flow by the SWAT model is governed by the intensity of rainfall as recorded by rain gauges. Presumably, transmission loss is also one of the key reasons that have led to compromised simulation of peak flows, especially for arid and semi-arid regions. This is because such areas have two highly sensitive parameters related to the transmission process, namely CH_K(2) and CH_L(2).

Thus, it can be clarified that transmission losses, which prevent lateral flow and interaction of less-understood processes such as behavior of groundwater recharge/river, can affect simulation of moderate and peak flows.

3.4. Evaluation of runoff volume

The runoff volume was evaluated using the Soil Conservation Service-Curve Number (SCS-CN) method (Neitsch et al., 2005). Fig. 9 shows the percentage of runoff volume for the calibration and validation periods as well as for the overall model. Pie chart (1) depicts equal contribution of both observed and simulated runoff volumes in the calibration process. On the contrary, pie chart (2) shows that there is a 10 % difference in contribution of runoff volume between simulated and observed data in the validation process. As for the entire model, the contribution towards runoff volume of both simulated and observed data has a difference of 2 %, as depicted in pie chart (3). Fig. 10 and 11 illustrate the cumulative daily runoff volume for the Roodan watershed in calibration and validation periods.


183

Fig. 4. Observed and simulated stream flow over calibration (1989-2002) and validation (2003-2008) periods

Fig. 5. Duration curves of residual errors (i.e., simulated flow subtracting observed flows)

Fig. 6. (A): Scatter plot of observed and simulated flows over calibration and (B): validation period

Fig. 7. Observed and simulated flow over calibration period by 95PPU


184

Fig. 8. Observed and simulated flow over validation period by 95PPU

Table 4. Percentile of absolute error between observed and simulated flow (CMS)

Percentile %

5 10 25 50 75 90 95 Validation 0.03 0.05 0.12 0.42 1.07 7.99 20.66 Calibration 0.38 0.69 1.2 1.69 3.72 14.65 43.54

Fig. 10 shows that the simulated daily runoff volumes have been underestimated over the period of 1990 to 1996.

From 1996 to the end of calibration period, the simulated daily flow was satisfactory. The underestimation can be caused by the aforementioned reasons, i.e., intensity and duration of precipitation, transmission losses, and interaction with ground water. Not only that, it is postulated that parameters related to the soils (e.g., SOL_AWC) and surface runoff processes (e.g., CN2.mgt) could have significantly affected as well.

For the validation period, the performance of this SWAT model was of satisfactory from 2003 to early 2006 (Fig. 11). Beyond that, the simulated daily runoff volume had been profoundly overestimated. This could be caused by recession of runoff volume in the SWAT model, as shown earlier in the studies of Yang et al. (2008). Generally, the trend of daily cumulative runoff volume is similar for both observed and simulated daily runoff volumes for both calibration and validation sample data.

A comparison between annual runoff volumes had been made and is as shown in Fig. 12. The result shows that the simulated annual runoff volume had been underestimated for 7 years whereas most of the annual runoff volume had been overestimated for every year. Furthermore, only two years of total annual runoff volumes had been simulated to be approximately the same (1997 and 2007). Thus, it can be concluded that SWAT model has shown a promising performance for both daily stream flow and attributed runoff volume modeling. In this study, however, SWAT did not satisfactorily reproduce the moderate and peak flows in the Roodan watershed; it only simulated a promising runoff volume. Therefore, SWAT can be used as a useful subsidiary tool for sustainable development of watersheds, water resource management in un-gauged

catchments, and decision making for plans regarding human activities planning.

Fig. 9. Pie chart (1), (2) and (3): contribution of observed

vs. simulated runoff volume (m3) for calibration and validation periods and the overall modeling


185

Fig. 10. Cumulative daily runoff volume for calibration period (Note: a billion is equal to a thousand millions)

Fig. 11. Cumulative daily runoff volume for validation period (Note: a billion is equal to a thousand millions)

Fig. 12. Annual runoff volume calculated by SCS-CN method using SWAT for each year (Note: a billion is equal to a thousand millions)

4. Conclusions

SWAT model had been used for modeling and simulation of stream flow and runoff volume in the Roodan watershed situated at the southern part of Iran. The area is within an arid region where the precipitation is low (around 215 mm per annum). Even though the available data is scarce, the results of the modeling are of satisfactory. Land use map update procedure was performed according to available statistics for the agricultural area for three periods starting at 1988, 1993 and 2002 to consider the impact of continuous land use changes.

In the sensitivity analysis, SWAT showed a high sensitivity towards parameters that are attributed to the process of channel routing, ground water, surface runoff, soil properties, and evapotranspiration.

In the sensitivity analysis, it was found that the adjustment of EVRCH.bsn parameter can be helpful as it can reduce over-estimation in evaporation for main channels. Results of uncertainty analysis showed that for both calibration and validation processes, SUFI-2 algorithm gave reasonably good results. The values of R2, NS, and PBIAS coefficients are 0.75, 0.75 and 1.5 %


186

respectively for the calibration period and 0.64, 0.64 and 21 % respectively for validation period.

SWAT also visualized 21 sub-basins that had been developed into agricultural fields from 1988 to 2008 where most of theses sub-basins are in the southwest, center and northeast areas of the Roodan watershed. The simulated stream flow by SWAT model shown a promising trend, although peak flows had not been obviously reproduced by the model; the SWAT model had underestimated most of the peak flows. It can be said that transmission losses and interaction of less-understood processes such as behavior of groundwater recharge/river can affect the simulation of moderate and peak flows. Nevertheless, the estimation of runoff volume using SWAT model showed good performance for both calibration and validation processes. Even though the final results showed that the simulated SWAT model did not fully match the peak stream flows, the estimation for annual runoff volume was fairly satisfactory. In other words, SWAT can still be used as a subsidiary hydrological tool in water resource management projects when data concerning runoff volumes has to be analyzed. Moreover, the SWAT model can also be used for planning of agriculture and irrigation activities.

To sum up, it is suggested that the development of this tool for research-based planning on agricultural activities in arid region and evaluation of attributed impacts on runoff, sedimentation, erosion and nutrient has to be carried out. The model shall, undeniably, remains beneficial for future scheme and development planning for the Roodan watershed and its sub-basins, but in depth assessments will still be needed to elevate the accuracy and reliability of this model. It is also suggested that application of SWAT model for stream flow modeling in arid climate over large and small scales should be further explored to make the model more flexible.

Acknowledgements We appreciate the cooperation and help given by the Department of Hydraulic and Hydrology and Centre of Information and Communication Technology (CICT) of Universiti Teknologi Malaysia; consultant engineers of Ab Rah Saz Shargh Corporation in Iran; and the regional water organization, agricultural organization, and natural resources organization of the Hormozgan province, Iran. References Abbaspour K.C., Genuchten M.T.van., Schulin R., Schlapp

I., (1997), A sequential uncertainty domain inverse procedure for estimating subsurface flow and transport parameters, Water Resource Research, 33, 1879-1892.

Abbaspour K.C., Johnson A., Van Genuchten M.T., (2004), Estimating uncertain flow and transport parameters using a sequential uncertainty fitting procedure, Vadose Zone Journal, 3, 1340- 1352.

Abbaspour K.C., Yang J., (2006), A Calibration and Uncertainty analysis Program for SWAT, Swiss Federal Institute Aquatic Science and Technology.

Abbaspour K.C., Yang,J., Maximov I., Siber R., Bogner K., Mieleitner J., Zobrist J., Srinivasan R., (2007), Modeling hydrology and water quality in the pre-Alpine/Alpine Thur watershed using SWAT, Journal of Hydrology, 333, 413–430.

Abbaspour K.C., (2008), User Manual for SWAT-CUP, SWAT Calibration and Uncertainty analysis Programs, Swiss Federal Institute of Aquatic Science and Technology, Eawag, Duebendorf, Switzerland, 93 pp.

Ab Rah Saz Shargh, (2009), Comprehensive studies of water resource management for Roodan watershed, (in Persian), Consulting Water Resource Engineering Corporation, register code 14800, Mashhad, Iran, On line at: www.Abrahsaz.com.

Arabi M., Govindaraju R.S., Hantush M.M., Engel B.A., (2006), Role of watershed subdivision on modeling the effectiveness of best management practices with SWAT, Journal of the American Water Resources Association, 42, 513-528.

Arnold J.G., Allen P.M., Bernhardt G., (1993), A com-prehensive surface-ground water flow model, Journal of Hydrology, 142, 47-69.

Arnold J.G., Srinivasan R., Muttiah R.S., Williams J.R.,(1998), Large area hydrologic modeling and assessment, Part I. Model development, Journal of American Water Resources Association, 34, 73-89.

Balon M., Dehnad F., (2006), Water crisis in arid and semi-arid regions- an international challenge, Symposium in Tehran Iran, 12-13, September 2006.

Biesbrouck B., Wyseure G., Van Orschoven J., Feyen J., (2002), AV SWAT 2000, Vital Decosterstraat 102, B-3000 Leuven Belgium, Katholieke Universiteit Leuven and Laboratory for Soil and Water Management.

Brooks K.N., Ffolliott P.F., Gregersen H.M., Thames J.L., (1991), Hydrology and the Management of Watersheds, Iowa State University Press, Ames, IA.

Chaplot V., (2005), Impact of DEM mesh size and soil map scale on SWAT runoff, sediment, and NO3-N loads predictions, Journal of Hydrology, 312, 207-222.

Clarke R., (1993), Water: The International Crisis. Cambridge, MA: MIT Press.

Dingman S.L., (2002), Physical Hydrology, 2nd (Ed.), Upper Saddle River, N.J. Prentice Hall.

Faramarzi M, Abbaspour K.C., Schulin R., Hong Y., (2009), Modeling blue and green water resources availability in Iran, Hydrological Processes, 23, 486-501.

FitzHugh T.W., Mackay D.S., (2000), Impacts of input parameter spatial aggregation on an agricultural nonpoint source pollution model, Journal of Hydrology, 236, 35–53.

Food and Agriculture Organization, (1995), The digital soil map of the world and derived soil properties, CD-ROM, Version 3.5, Rome.

Freeze R.A., Cherry J.A., (1979), Ground Water, Prentice Hall, Eaglewood Cliffs N.J.

Ghaffari G., Keesstra S., Ghodousi J., Ahmadi H., (2010), SWAT-simulated hydrological impact of land-use change in the Zanjanrood Basin, Northwest Iran, Hydrological Processes, 24, 892–903.

Ghaffari G., (2011), The impact of DEM resolution on runoff and sediment modeling results, Research Journal of Environmental Sciences, 5, 691-702.

Gassman P.W., Reyes M.R, Green C.H., Arnold J.G., (2007), The soil and water assessment tool, historical development, applications, and future research directions, Transactions of the ASABE, 50, 1211-1250.


187

Glavan M., Pintar M., Volk M., (2013), Land use change in a 200-year period and its effect on blue and green water flowin two Slovenian Mediterranean catchments-lessons for the future, Hydrological Processes, 27, 3964–3980.

Gosain A.K., Mani A., Dwivedi C., (2009), Hydrological Modeling Review, Climawater. Report No.1.

Hantush M.M., Kalin L., (2005), Uncertainty and sensitivity analysis of runoff and sediment yield in a small agricultural watershed with kineros, Hydrological Sciences Journal, 50, 1151-1172.

Hjelmfelt A.T., (1980), Empirical-investigation of curve number technique, Journal of the Hydraulics and Engineering Division-ASCE, 106, 1471-1476.

Huang Z., Xue B., Pang,Y., (2009), Simulation on stream flow and nutrient loading in Gucheng Lake, Low Yangtze river basin, based on swat model, Quaternary International, 208, 109-115.

Junfeng C., Xiubin L., Ming Z., (2005), simulating the impacts of climate variation and land-cover changes on basin hydrology, A case study of the Suomo basin, Science in China Series D Earth Sciences, 48, 1501-1509.

Kannan N., White S.M., Worrall F., Whelan M.J., (2007), Sensitivity Analysis identification of the best evapotranspiration and runoff options for hydrological modeling in SWAT 2000, Journal of Hydrology, 332, 456-466.

Kim N.W., Chung M., Seung Y., Arnold J.G., (2008), Development and application of the integrated SWAT-MODFLOW Model, Journal of Hydrology, 356, 1– 16.

Krause P., Boyle D.P., Base F., (2005), Comparison of different efficient criteria for hydrological model assessment, Advanced Geosciences, 5, 89-97.

Krysanova V., Arnold J.G., (2009), Advances in ecohydrological modeling with SWAT-A review, Hydrological Science Journal, 53, 939-947.

Lenhart T., Eckhardt K., Fohrer N., Frede H.G., (2002), Comparison of two different approaches of sensitivity analysis, Physics and Chemistry of the Earth, 27, 645-654.

Masih I., Maskey S., Uhlenbrook S., Smakhtin V., (2011), Assessing the impact of areal precipitation input on stream flow simulations using the SWAT model, Journal of the American Water Resources Association, 47, 179-195.

Moriasi D.N., Arnold J.G., Van Liew M.W., Binger R.L., Harmel R.D., Veith T., (2007), Model evaluation guidelines for systematic quantification of accuracy in watershed simulations, Transactions of the ASABE, 50, 885-900.

Morid S., Ghaemi H., Mir Abolghasemi H., Abediny M., (2001), Evaluation of the HEC-1 model for flood forecasting and solution in the hormozgan province, Iran, Journal of Environment Hydrology, 9, 1-8.

Neitsch S.L., Arnold J.G., Kiniry J.R., William J.R., (2005), Soil and Water Assessment Tool Theoretical Documentation, Version 2005, Grassland, Soil and Water Research Laboratory, Agricultural Research Service,and Blackland Research Center, Texas Agricultural Experiment Station: Temple, Texas 76502, USA, 494.

Neitsch S.L., Arnold J.G., Kiniry J.R., Ssrinvasan R., Williams J.R., (2010), Soil and Water Assessment Tool, Input/Output File Documentation, Version 2009, Soil and Water Research Laboratory, Agricultural Research Service, and Blackland Research Center, Texas Agrilife Research. Texas water resources

institute technical report No.365: Texas A&M University, USA.

Nie W., YuanY., Kepner W., Nash M.S., Jackson M., Erickson C., (2011), Assessing impact of land use and land cover changes on hydrology for the upper San Pedro watershed, Journal of Hydrology, 407, 105-114.

Oeurng C., Sauvage S., Sanchez-Perez J., (2011), Assessment of hydrology, sediment and particulate organic carbon yield in large agricultural catchment using SWAT model, Journal of Hydrology, 401, 145-153.

Pai N., Saraswat D., (2011), SWAT 2009_LUC: A tool to activate the land use change module in SWAT 2009, Soil and Water division of ASABE, 54, 1649-1658.

Raneesh S.G., Thampi K.Y., Surya T.V., (2010), Influence of scale on SWAT model calibration for stream flow in a river basin in the humid tropics, Water Resources Management, 224, 4567-4578.

Reynolds C.A., Jackson T.J., Rawls W.J., (1999), Estimating available water content by linking the FAO soil map of the world with global soil profile database and pedo-transfer functions, Proceedings of the AGU 1999 Spring Conference, Boston, MA.

Rostamian R., Aazam J., Afyuni M., Mousavi F., Heidarpour M., Jalalian A., (2008), Application of a SWAT model for estimating runoff and sediment in two mountainous basins in central Iran, Hydrological Sciences Journal, 53, 977-988.

Schuol J., Abbaspour K.C., Srinivasan R., Yang H., (2008), Estimation of freshwater availability in the West African sub-continent using the SWAT hydrologic model, Journal of Hydrology, 352, 30-49.

Setegn S.G., Srinivasan R., Melesse A.M., Dargahi B., (2010), SWAT model application and prediction uncertainty analysis in the lake Tana basin, Ethiopia, Hydrological Processes, 24, 357-367.

Shaw Elizabeth M., (1994), Catchment Modeling, Hydrology in Practice, London: Chapman and Hall, 342-370.

Singh V.P., Frevert D.K., (2006), Watershed Models, U.S: Taylor &Francis Group.

Thompson J.R., Sorenson H.R., Gavin H., Refsgaard A., (2004), Application of the coupled MIKE SHE/MIKE 11 modeling system to a lowland wet grassland in southeast England, Journal of Hydrology, 293, 151-179.

Tripathi M.P, Panda R.K., Raghuwanshi N.S., Singh R., (2004), Hydrological modeling of small watershed using generated rainfall in the soil and water assessment tool model, Hydrology Processes, 18, 1811-1821.

Van Griensven A., Meixner T., (2003), Sensitivity, Optimization and Uncertainty Analysis for the Model Parameters of SWAT, 2nd International SWAT Conference, Bari, Italy, 1-4 July.

Van Griensven A., Meixner T., Grunwald S., Bishop T., Diluzio M., Srinivasan R., (2006), A global sensitivity analysis tool for the parameters of multi- variable catchment models, Journal of Hydrology, 324, 10-23.

Volk M., Liersch S., Schmidt G., (2009), Towards the implementation of the European Water FrameworkDirective? Lessons learned from water quality simulations in an agricultural watershed, Land Use Policy, 26, 580–588.

Wang S., Kang S., Zhang L., Li F., (2008), Modeling hydrological response to different land use and climate change scenarios in the Zamu river basin of northwest China, Hydrological Processes, 22, 202-2510.


188

Wilby R.L., (1997), Contemporary Hydrology, England: John and Sons.

Winchell M., Srinivasan R., Di Luzio M., Arnold J., (2010), Arc swat interface for SWAT 2009, Users’guide. Grassland, Soil and Water Research Laboratory, Agricultural Research Service, and Blackland Research Center, Texas Agricultural Experiment Station: Temple, Texas 76502, USA, 495.

Xu Z.X., Pang J.P., Liu C.M., Li J.Y., (2009), Assessment of runoff and sediment yield in the miyun reservoir catchment by using SWAT model, Hydrological Process, 23, 3619-3630.

Yang J., Reichert P., Abbaspour K.C., Xia J., Yang H., (2008), Comparing uncertainty analysis use for a SWAT application to Chaohe basin in China, Journal of Hydrology, 358, 1-23.

AN EVALUATION OF BLUE WATER PREDICTION IN SOUTHERN PART OF IRAN USING SWAT

Documents

Transcript of AN EVALUATION OF BLUE WATER PREDICTION IN SOUTHERN PART OF IRAN USING SWAT