Abstract

This study investigates the impact of different digital elevation model (DEM) resolutions on the topological attributes and simulated runoff, as well as the sensitivity of runoff parameters in the Mahabad Dam watershed in Iran. The watershed and streamlines were delineated in ArcGIS, and the hydrologic analyses were performed using the Soil and Water Assessment Tool (SWAT). The sensitivity analysis on runoff parameters was performed, using the Sequential Uncertainties FItting Ver. 2 algorithm, in the SWAT Calibration and Uncertainty Procedures (SWAT-CUP) program. The results indicated that the sensitivity of runoff parameters, watershed surface area, and elevations changed under different DEM resolutions. As the distribution of slopes changed using different DEMs, surface parameters were most affected. Furthermore, higher amounts of runoff were generated when DEMs with finer resolutions were implemented. In comparison with the observed value of 8 m3/s at the watershed outlet, the 12.5 m DEM showed more realistic results (6.77 m3/s). Comparatively, the 12.5 m DEM generated 0.74% and 2.73% more runoff compared with the 30 and 90 m DEMs, respectively. The findings of this study indicate that in order to reduce computation time, researchers may use DEMs with coarser resolutions at the expense of minor decreases in accuracy.

INTRODUCTION

The application of practical models and appropriate data in hydrological studies are vital to understand the processes occurring in the watershed scale (Yaseen et al. 2019). The ability of these models to represent the hydrological processes and estimate variables such as runoff, sediment, and nutrient yields within a watershed greatly depends on the quality of the input data, which are the primary sources of errors in estimated hydrological variables (Earls & Dixon 2005).

The application of hydrological models, in particular for rainfall-runoff simulation and flood prediction, has received extensive attention in recent years, and numerous research studies have been conducted in these fields (e.g. Wang et al. 2015; Chau 2017; Fotovatikhah et al. 2018; Moazenzadeh et al. 2018). The digital elevation model (DEM) is one of the essential input files for the hydrological models used in estimating a variety of variables in a watershed (Cotter et al. 2003; Chaubey et al. 2005). It is a digital (raster) dataset of elevations in x, y, and z coordinates, which represents the physical parameters of the watershed regarding the flow direction, drainage network, and drainage slopes (Freeman 1991). The DEM resolution can affect important watershed characteristics such as surface area, shape, length, and slope. The surface area of the watershed reflects the generated volume of water from rainfall; the shape of a watershed influences the shape of its characteristic hydrograph; the length affects the travel time of water through a watershed; and the slope affects the momentum of runoff (Babaei et al. 2019; Nazari-Sharabian et al. 2019a).

Significant developments in remote sensing technology have led to the development of high-quality DEMs with different resolutions for both commercial and research purposes. In ArcGIS, the spatial analysis computation speed is closely associated with the resolution of the input data. Therefore, in order to speed up the computation procedure, users may use DEMs with coarser resolutions.

The Soil and Water Assessment Tool (SWAT) is among the software packages widely used for watershed-scale studies. This model uses DEMs for watershed delineation. Cotter et al. (2003) used the SWAT model to evaluate the impact of different DEM resolutions on the uncertainties of predicted values of runoff, sediment, nitrate-nitrogen (NO3-N), and total phosphorus (TP) transport in the Moores Creek watershed in Washington County, Arkansas, USA. The authors concluded that the model's output uncertainty depends upon the input parameters' uncertainty. Chaubey et al. (2005) found that the watershed delineation, stream network, and sub-basin classification in the SWAT are affected by the DEM resolution. The authors showed that a coarser DEM resolution results in decreased runoff, sediment, NO3-N, and TP load predictions. Moreover, Dixon & Earls (2009) used three DEMs with 30, 90, and 300 m resolutions to compare the predicted streamflow in the Charlie Creek drainage basin, located in the Peace River drainage basin of central Florida, USA, using the SWAT model. The authors reported that the model is sensitive to the resolution of DEMs, and resampling may not be adequate in modeling stream flows using a distributed watershed model.

Lin et al. (2010) studied the effect of DEMs with different resolutions (varying from 5 to 140 m) on hydrological parameters in the Xiekengxi River watershed in the Zhejiang Province of China. The authors showed that runoff values were more sensitive to coarser DEM resolutions, but not so sensitive to finer resolutions. Peter et al. (2013) studied the sediment delivery estimates in a coastal watershed in South Carolina, USA, using four DEMs with 3, 10, 30, and 90 m resolutions. The researchers noted that finer resolution DEMs resulted in more accurate slope results and sediment output. Furthermore, Zhang et al. (2014) studied the impact of different DEM resolutions on the SWAT model outputs of the sediment and nutrient yield in the agricultural watershed of the Xiangxi River, Gorges Reservoir in China. The authors used 17 DEMs with resolutions varying from 30 to 1,000 m and analyzed the results of the model outputs of sediments and nutrients for each resolution. The researchers noticed that the sediment yield was significantly affected by the DEM resolution, and the prediction of dissolved oxygen load was significantly affected by DEM resolutions coarser than 500 m. Moreover, the authors noticed that total nitrogen, NO3-N, and TP loads were slightly affected by the DEM resolution, while the ammonia nitrogen (NH4-N) load was unaffected.

More recently, Liffner et al. (2018) studied the sensitivity of hypsometric properties to the DEM resolution, DEM type, and polynomial order through assessing differences in hypsometric properties derived from 417 catchments and subcatchments in southern Australia. The researchers found significant sensitivity (p-value <0.05) of hypsometric properties across the DEM types and polynomial orders. Using four different DEMs and the Universal Soil Loss Equation, Chen et al. (2018) conducted a new analysis to compute the amounts of sheet and rill erosion of the Shihmen Reservoir watershed in northern Taiwan. The authors concluded that the DEM created from airborne Light Detection and Ranging (LiDAR), with the highest vertical resolution among the four DEMs, yielded the highest amount of soil erosion. The lowest amounts of soil erosion were observed when using the two DEMs created from satellite images with the lowest vertical resolution, and a medium soil erosion amount was observed when using the DEM created from aerial photographs.

According to the literature, studies have noted that the DEM resolution has a direct impact on hydrologic model simulations. However, the use of finer resolution spatial data does not necessarily improve the performance of hydrological model simulations (Ndomba & Birhanu 2008). It was noted that no studies have investigated the relationship between the DEM resolution and the sensitivity of parameters contributing to the variable of interest, i.e. runoff. Therefore, using DEMs with different resolutions, this study seeks to find the answer to the following questions:

  • 1.

    How are the sensitivities of parameters contributing to runoff affected?

  • 2.

    What topological factors in the delineated watershed are affected?

  • 3.

    What is the impact of the DEM resolution on runoff yield in the watershed?

MATERIALS AND METHODS

For the case study of the Mahabad Dam watershed in Iran, three DEMs with 12.5, 30, and 90 m resolutions were implemented. These DEMs were chosen for this study, as they are publicly available from online databases (i.e. the Phased Array Type L-band Synthetic Aperture Radar (ALOS PALSAR), Space Shuttle Radar Topography Mission (SRTM 2018), and Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER)), as well as for comparison purposes. To delineate the watershed and perform the hydrologic analyses from 1991 to 2012 (the time period that data records were available from hydrometric stations in the region), the SWAT model Ver. 2012.10.21, which runs in ArcGIS using the ArcSWAT extension, was employed. According to the Food and Agriculture Organization of the United States (FAO) (2018a; 2018b), the SWAT is the most comprehensive environmental modeling software available (http://www.fao.org/land-water/land/land-governance/land-resources-planning-toolbox/category/details/en/c/1111246/). In comparison with other hydrological models, the SWAT is physically based, computationally efficient, and capable of continuous simulation over long time periods (Gassman et al. 2007).

The SWAT model inputs for this study include the DEMs, land-use map, soil map, and meteorological data (precipitation, temperature, solar radiation, wind, and relative humidity). Moreover, for the sensitivity analysis of the parameters contributing to the runoff, the Sequential Uncertainties FItting Ver. 2 (SUFI-2) module of the SWAT Calibration and Uncertainty Procedures (SWAT-CUP) program Ver. 5.1.6.2 was used. Several other tools are available for the sensitivity analysis and calibration/validation of the SWAT model, but as SWAT-CUP is recommended by the developers of the SWAT model and is freely available, this model was chosen for this study. The following flowchart, given in Figure 1, summarizes the procedure of this study.

Figure 1

Flowchart of this study.

Figure 1

Flowchart of this study.

Soil and Water Assessment Tool

Among the most commonly used, continuous-time, semi-distributed, and physically based models is the SWAT model. To simulate processes within a watershed, the SWAT model integrates weather, surface and groundwater hydrology, soil properties, plant growth, nutrient cycles, and land management practices (Arnold et al. 1998). Based on interior outlet points along the stream network, the model divides the watershed area into several sub-basins. Watersheds, also known as basins or catchments, are physically delineated by the area upstream from a specific outlet point. Watershed delineation based on DEMs is the prerequisite to setting up a SWAT model. There are two methods for watershed delineation in a SWAT model: the DEM-based method and the pre-defined method, in which users can define the reaches and sub-basins manually. Moreover, based on the soil type, land use, and slope, the SWAT model divides the basin into hydrological response units (HRUs) and runs the simulations at the HRU level (Shang et al. 2012).

Surface runoff

In the SWAT model, surface runoff is predicted for the daily rainfall using the Soil Conservation Service (SCS) curve number (CN) method. In the SCS method, surface runoff occurs when the rainfall (in mm) for the day (Rday) is greater than the initial abstraction (i.e. losses such as evapotranspiration, depression storage, and infiltration). The SCS CN method is an empirical conceptual method developed for the computation of surface runoff under varying soil types and land uses. The SCS CN is: 
formula
(1)
where Qsurf is the collected runoff (mm of H2O), Rday is the precipitation depth for the day (mm of H2O), S is the retention parameter (mm of H2O), and Ia is the initial abstraction loss (mm of H2O). The retention parameter S is calculated using the following equation: 
formula
(2)

The SCS CN is a function of the soil's permeability, land use, and antecedent soil water conditions. Typical CN values for various land covers and soil types can be found in hydrology reference books.

SWAT Calibration and Uncertainty Procedures

The sensitivity analysis is the process of determining the significance of the effect of a one or multi-parameter combination on the output of a model or target function. The SWAT-CUP program has been developed for the calibration, validation, and sensitivity analysis of the SWAT model parameters and uses five different calibration procedures: SUFI-2, particle swarm optimization (PSO), generalized likelihood uncertainty estimation (GLUE), parameter solution (ParaSol), and Markov Chain Monte Carlo (MCMC). For large-scale models, in which the calibration process can be very time-consuming, the semi-automated SUFI-2 is quite efficient. To quantify the sensitivity of each parameter, the SWAT-CUP program uses a multiple regression approach, as presented in the following equation: 
formula
(3)
where g is the objective function value, α is the regression constant, and β is the coefficient of parameters. The relative significance of each parameter b is then identified by a t-test. The t-stat is the coefficient of a parameter divided by its standard error. It is a measure of the precision with which the regression coefficient is measured. If a coefficient is large compared to its standard error, then it is probably different from zero, and the parameter is sensitive. The Student's t-distribution describes how the mean of a sample with a certain number of observations is expected to behave. The p-value for each term tests the null hypothesis that the coefficient is equal to zero (no effect). A low p-value (<0.05) indicates that the null hypothesis can be rejected. In other words, a predictor that has a low p-value is likely to be a meaningful addition to the model because changes in the predictor's value are related to changes in the response variable. Conversely, a larger p-value suggests that changes in the predictor are not associated with changes in the response, meaning that the parameter is not very sensitive. A p-value of <0.05 is the generally accepted point at which to reject the null hypothesis (i.e. the coefficient of the parameter is different from zero). With a p-value of 0.05, there is only a 5% chance that the results would have occurred in a random distribution. Therefore, with a 95% probability, the variable has some effect. In this analysis, a higher value of t-stat (lower p-value) shows a higher sensitivity of the parameter (Abbaspour et al. 2007).
To calibrate a SWAT model, the SWAT-CUP program adjusts model parameters within a user-defined range to achieve the desired value of an objective function. These functions include: the multiplicative form of the squared error (mult); the summation form of the squared error (sum); the coefficient of determination (R2); the chi-squared test (Chi2); the Nash–Sutcliffe model efficiency coefficient (NS); and modified R2, where the coefficient of determination is multiplied by the coefficient of the regression line between measured and simulated data, b (bR2); the SSQR method (SSQR); percent bias (PBIAS); the Kling–Gupta efficiency; the RMSE (root-mean-square error)-observations standard deviation ratio (RSR); and the modified Nash–Sutcliffe efficiency factor (MNS). In the SWAT-CUP model, the user should only choose one objective function, and the model works to obtain the best results based on the selected objective function. However, the values of all 11 objective functions are reported at the end. In this study, R2 (Equation (4)), NS (Equation (5)), and PBIAS (Equation (6)) are used to show how well the model was calibrated and validated. The objective function selected to perform the calibration/validation of the model in this study was the NS, considering its popularity among hydrologists in water resource studies, as well as the ease of the interpretation of its value. 
formula
(4)
 
formula
(5)
 
formula
(6)
where ‘Q’ is a variable such as streamflow, ‘m’ and ‘s’ stand for measured and simulated, the bar indicates the average, and ‘i’ is the ith measured or simulated value. Higher R2 values show that the model fits the observed values better. The NS function has a range of −∞ to 1. NS = 1 corresponds to a perfect match of simulated values to the observed data. The values between 0 and 1 indicate that the simulated and observed values are close to each other, whereas values less than 0 show that the model has no predictive power (Moriasi et al. 2007). Moreover, PBIAS measures the average tendency of the simulated data to be larger or smaller than the observations. The optimum value is zero, where low magnitude values indicate better simulations. Positive values indicate model underestimation, and negative values indicate model overestimation (Gupta et al. 1999).

Study area and data

In this study, the Mahabad Dam watershed in Iran, which is topologically located in a mountainous area, was selected as the case study. Therefore, changes in the DEM resolution can have a higher impact on watershed delineation and topologic features, and consequently, the sensitivity of parameters and runoff yield, compared with a flatter watershed area (Nazari-Sharabian et al. 2019b). This watershed is located in the West-Azerbaijan province in the northwest of Iran (36°44′N, 45°39′E) and is one of the Urmia Lake basins. The area of the watershed is approximately 808 km2 and is mostly covered by agricultural fields and grasslands. The average precipitation in this region is 350 mm/yr, and it is mostly in the form of snow during cold months of the year. The Kauter and Beytas Rivers originate from the southern heights of the plain and run to the north in parallel. They join and create the Mahabad Dam reservoir and continue running as the Mahabad River (Figure 2). Based on the hydrometric station records, the average flow rates of the Kauter and Beytas Rivers during 1988–2012 were 6.18 and 1.82 m3/s, respectively (I.R. of Iran Meteorological Organization).

Figure 2

(a) Land-use map and (b) soil map of the Mahabad Dam watershed.

Figure 2

(a) Land-use map and (b) soil map of the Mahabad Dam watershed.

Tables 1 and 2 represent the land use and soil classification of the watershed, respectively.

Table 1

Land-use classification of the Mahabad Dam watershed

Land-use typeWatershed area (%)
Dryland farming 66.37 
Dense pasture 13.82 
Non-dense pasture 11.09 
Irrigated farming 4.49 
Forest 2.95 
Water 1.03 
Urban (medium density) 0.13 
Land-use typeWatershed area (%)
Dryland farming 66.37 
Dense pasture 13.82 
Non-dense pasture 11.09 
Irrigated farming 4.49 
Forest 2.95 
Water 1.03 
Urban (medium density) 0.13 
Table 2

Soil types in the Mahabad Dam watershed

Soil typeSand (%)Silt (%)Clay (%)Watershed area (%)
Taconic 43 35 23 72 
Benson 35 37 30 28 
Soil typeSand (%)Silt (%)Clay (%)Watershed area (%)
Taconic 43 35 23 72 
Benson 35 37 30 28 

To delineate the watershed, DEMs with 12.5, 30, and 90 m resolutions were acquired from the Advanced Land Observation Satellite: the ALOS PALSAR 2018 (http://www.eorc.jaxa.jp/ALOS/en/index.htm), SRTM 2018 (https://www2.jpl.nasa.gov/srtm/), and ASTER global DEM data source 2018 (https://asterweb.jpl.nasa.gov/gdem.asp). The soil map was retrieved from the FAO soils portal using the Harmonized World Soil Database Ver. 1.2 (http://www.fao.org/soils-portal/soil-survey/soil-maps-and-databases/harmonized-world-soil-database-v12/en/2018). The land-use map and the data from the hydrometric stations were provided by the Mahab Ghodss Consulting Engineering Company (2014) (http://www.mahabghodss.com/). Moreover, meteorological data were provided by the Islamic Republic of Iran Meteorological Organization (IRIMO) (2014) for the weather station of Mahabad from 1988 to 2012 (http://www.irimo.ir/eng/).

RESULTS AND DISCUSSION

Sensitivity analysis

To build a realistic watershed model, it has to be calibrated. Since parameters represent processes, and several parameters contribute to watershed yields, a sensitivity analysis has to be performed prior to model calibration in order to determine the parameters that most affect the variable of interest in the watershed, e.g. runoff. Therefore, a sensitivity analysis helps to decrease the number of parameters in the calibration procedure by eliminating the parameters identified as not sensitive. The parameters analyzed using the SUFI-2 module of the SWAT-CUP program, along with a short description of each parameter, are presented in Table 3. The SWAT model uses numerous parameters to simulate the water quality and surface runoff in a watershed. Among those, the parameters presented in Table 3 are the most important affecting runoff generation and movement in a watershed.

Table 3

Parameters contributing to the runoff in the SWAT model

ParametersDescription
SHALLST.gw Initial depth of water in the shallow aquifer (mm) 
DEEPST.gw Initial depth of water in the deep aquifer (mm) 
GWHT.gw Initial groundwater height (m) 
ALPHA_BF.gw Baseflow alpha factor (1/days) 
GW_DELAY.gw Groundwater delay time (days) 
GWQMN.gw Threshold depth of water in the shallow aquifer required for return flow to occur (mm) 
GW_REVAP.gw Groundwater ‘revap’ coefficient 
REVAPMN.gw Threshold depth of water in the shallow aquifer for ‘revap’ or percolation to the deep aquifer to occur (mm) 
RCHRG_DP.gw Deep aquifer percolation fraction 
GW_SPYLD.gw Specific yield of the shallow aquifer (m3/m3
SNO_SUB.sub Initial snow water content (mm) 
PLAPS.sub Precipitation lapse rate (mm/km) 
TLAPS.sub Temperature lapse rate (°C/km) 
CH_K1.sub Effective hydraulic conductivity in tributary channel alluvium (mm/h) 
CH_N1.sub Manning's ‘n’ value for the tributary channels 
CN2.mgt Initial SCS runoff CN for moisture condition II 
SFTMP.bsn Snowfall temperature (°C) 
SMTMP.bsn Snow melt base temperature (°C) 
SURLAG.bsn Surface runoff lag coefficient 
SMFMX.bsn Melt factor for snow on June 21 (mm/°C-day) 
SMFMN.bsn Melt factor for snow on December 21 (mm/°C-day) 
TIMP.bsn Snow pack temperature lag factor 
CH_N2.rte Manning's ‘n’ value for the main channel 
CH_K2.rte Effective hydraulic conductivity in main channel alluvium (mm/h) 
ALPHA_BNK.rte Baseflow alpha factor for bank storage (days) 
ESCO.hru Soil evaporation compensation factor 
EPCO.hru Plant uptake compensation factor 
CANMX.hru Maximum canopy storage (mm) 
OV_N.hru Manning's ‘n’ value for the overland flow 
SOL_ZMX.sol Maximum rooting depth of the soil profile (mm) 
SOL_AWC.sol Available water capacity of the soil layer (mm H2O/mm soil) 
SOL_K.sol Saturated hydraulic conductivity (mm/h) 
SOL_BD.sol Moist bulk density (g/cm3
ParametersDescription
SHALLST.gw Initial depth of water in the shallow aquifer (mm) 
DEEPST.gw Initial depth of water in the deep aquifer (mm) 
GWHT.gw Initial groundwater height (m) 
ALPHA_BF.gw Baseflow alpha factor (1/days) 
GW_DELAY.gw Groundwater delay time (days) 
GWQMN.gw Threshold depth of water in the shallow aquifer required for return flow to occur (mm) 
GW_REVAP.gw Groundwater ‘revap’ coefficient 
REVAPMN.gw Threshold depth of water in the shallow aquifer for ‘revap’ or percolation to the deep aquifer to occur (mm) 
RCHRG_DP.gw Deep aquifer percolation fraction 
GW_SPYLD.gw Specific yield of the shallow aquifer (m3/m3
SNO_SUB.sub Initial snow water content (mm) 
PLAPS.sub Precipitation lapse rate (mm/km) 
TLAPS.sub Temperature lapse rate (°C/km) 
CH_K1.sub Effective hydraulic conductivity in tributary channel alluvium (mm/h) 
CH_N1.sub Manning's ‘n’ value for the tributary channels 
CN2.mgt Initial SCS runoff CN for moisture condition II 
SFTMP.bsn Snowfall temperature (°C) 
SMTMP.bsn Snow melt base temperature (°C) 
SURLAG.bsn Surface runoff lag coefficient 
SMFMX.bsn Melt factor for snow on June 21 (mm/°C-day) 
SMFMN.bsn Melt factor for snow on December 21 (mm/°C-day) 
TIMP.bsn Snow pack temperature lag factor 
CH_N2.rte Manning's ‘n’ value for the main channel 
CH_K2.rte Effective hydraulic conductivity in main channel alluvium (mm/h) 
ALPHA_BNK.rte Baseflow alpha factor for bank storage (days) 
ESCO.hru Soil evaporation compensation factor 
EPCO.hru Plant uptake compensation factor 
CANMX.hru Maximum canopy storage (mm) 
OV_N.hru Manning's ‘n’ value for the overland flow 
SOL_ZMX.sol Maximum rooting depth of the soil profile (mm) 
SOL_AWC.sol Available water capacity of the soil layer (mm H2O/mm soil) 
SOL_K.sol Saturated hydraulic conductivity (mm/h) 
SOL_BD.sol Moist bulk density (g/cm3

For sensitivity analysis, the model was run for 70 iterations (at least twice the number of parameters) for each DEM (Abbaspour et al. 2007). Figure 3 shows the results of the sensitivity analyses of the parameters contributing to the runoff in the SWAT model of the Mahabad Dam watershed for the 12.5, 30, and 90 m DEMs. Parameters with higher t-stat values (lower p-values) are more sensitive, and changing their values has a higher impact on the runoff yield in the watershed.

Figure 3

Sensitivity of parameters contributing to the runoff in the SWAT model using: (a) the 12.5 m DEM, (b) the 30 m DEM, and (c) the 90 m DEM.

Figure 3

Sensitivity of parameters contributing to the runoff in the SWAT model using: (a) the 12.5 m DEM, (b) the 30 m DEM, and (c) the 90 m DEM.

According to Figure 3, in the first quartile of the sensitivity analysis, results for the 12.5 m DEM, in order, were the Manning's ‘n’ value for the tributary channels (CH_N1.sub), maximum rooting depth of the soil profile (SOL_ZMX.sol), initial depth of water in the shallow aquifer (SHALLST.gw), baseflow alpha factor for bank storage (ALPHA_BNK.rte), initial snow water content (SNO_SUB.sub), snowfall temperature (SFTMP.bsn), melt factor for snow on June 21 (SMFMX.bsn), initial depth of water in the deep aquifer (DEEPST.gw), and melt factor for snow on December 21 (SMFMN.bsn); these were also the most sensitive parameters. On the other hand, the last quartile shows that, in order, the precipitation lapse rate (TLAPS.sub), groundwater delay time (GW_DELAY.gw), maximum canopy storage (CANMX.hru), surface runoff lag coefficient (SURLAG.bsn), threshold depth of water in the shallow aquifer required for return flow to occur (GWQMN.gw), saturated hydraulic conductivity (SOL_K.sol), effective hydraulic conductivity in tributary channel alluvium (CH_K1.sub), and snowmelt base temperature (SMTMP.bsn) were the least sensitive parameters.

Moreover, for the 30 m DEM, the snowmelt base temperature (SMTMP.bsn), snowfall temperature (SFTMP.bsn), moist bulk density (SOL_BD.sol), deep aquifer percolation fraction (RCHRG_DP.gw), melt factor for snow on December 21 (SMFMN.bsn), soil evaporation compensation factor (ESCO.hru), saturated hydraulic conductivity (SOL_K.sol), initial SCS runoff CN for moisture condition II (CN2.mgt), and threshold depth of water in the shallow aquifer required for return flow to occur (GWQMN.gw) were the most sensitive parameters. On the other hand, the last quartile shows that, in order, the initial snow water content (SNO_SUB.sub), available water capacity of the soil layer (SOL_AWC.sol), Manning's ‘n’ value for the tributary channels (CH_N1.sub), surface runoff lag coefficient (SURLAG.bsn), effective hydraulic conductivity in main channel alluvium (CH_K2.rte), maximum rooting depth of the soil profile (SOL_ZMX.sol), Manning's ‘n’ value for the main channel (CH_N2.rte), and the initial depth of water in the deep aquifer (DEEPST.gw) were the least sensitive parameters.

Finally, for the 90 m DEM, the snowmelt base temperature (SMTMP.bsn), snowfall temperature (SFTMP.bsn), moist bulk density (SOL_BD.sol), deep aquifer percolation fraction (RCHRG_DP.gw), soil evaporation compensation factor (ESCO.hru), saturated hydraulic conductivity (SOL_K.sol), melt factor for snow on December 21 (SMFMN.bsn), initial SCS runoff CN for moisture condition II (CN2.mgt), and threshold depth of water in the shallow aquifer required for return flow to occur (GWQMN.gw) were the most sensitive parameters. On the other hand, in order, the effective hydraulic conductivity in tributary channel alluvium (CH_K1.sub), available water capacity of the soil layer (SOL_AWC.sol), Manning's ‘n’ value for the tributary channels (CH_N1.sub), surface runoff lag coefficient (SURLAG.bsn), maximum rooting depth of the soil profile (SOL_ZMX.sol), effective hydraulic conductivity in main channel alluvium (CH_K2.rte), and Manning's ‘n’ value for the main channel (CH_N2.rte) were the least sensitive parameters in the last quartile.

Furthermore, to compare the parameters with the highest sensitivity in all three DEMs, the first quartile of Figures 3(a)–3(c) is presented in one graph, as shown in Figure 4.

Figure 4

Sensitivity of highly sensitive parameters in all three DEMs.

Figure 4

Sensitivity of highly sensitive parameters in all three DEMs.

The comparison between the sensitivity of parameters in different DEM resolutions shows that the 30 and 90 m DEMs share the same highly sensitive parameters, while the 12.5 m DEM only shares the snowfall temperature (SFTMP.bsn) and melt factor for snow on December 21 (SMFMN.bsn) with the other two DEMs. These findings indicate that for this mountainous watershed, located in a region where snowfall is significant, the mean air temperature at which precipitation is equally likely to be rain as snow/freezing rain and the melt factor for snow on December 21 are among the most sensitive parameters, regardless of the DEM resolution. Moreover, since the 30 and 90 m DEMs share the same highly sensitive parameters, while the 12.5 m DEM only shares two parameters with them, it can be concluded that there exists a resolution threshold in which above or below it, certain parameters become more or less sensitive.

Moreover, the snowmelt base temperature (SMTMP.bsn) shows the highest sensitivity in both the 30 and 90 m DEMs. In the 12.5 m DEM, the Manning's ‘n’ value for the tributary channels (CH_N1.sub) is the most sensitive parameter. This can be explained by the fact that, by using finer resolution DEMs, more land features appear, and therefore, more tributary streams form in the watershed, which ultimately join the mainstream. As a result, surface parameters, such as the Manning's ‘n’ value, become more important. Among soil parameters, the 12.5 m DEM shows that the maximum rooting depth of the soil profile (SOL_ZMX.sol) is the most sensitive parameter, while in the 30 and 90 m DEMs, the soil bulk density (SOL_BD.sol), which expresses the ratio of the mass of solid particles to the total volume of the soil, and the saturated hydraulic conductivity of soil (SOL_K.sol), which is a measure of the ease of water movement through the soil, are the most sensitive parameters. Moreover, the groundwater parameters show that the initial depth of water in the shallow aquifer (SHALLST.gw) and the initial depth of water in the deep aquifer (DEEPST.gw) are the most sensitive parameters in the 12.5 m DEM; however, the deep aquifer percolation fraction (RCHRG_DP.gw), which is the fraction of percolation from the root zone recharging the deep aquifer, and the threshold depth of water in the shallow aquifer required for return flow to occur (GWQMN.gw) are the most sensitive parameters in the 30 and 90 m DEMs.

Impacts on streamflow

In the SWAT model, the stream networks were delineated based on the elevation and slope distribution characteristics of the land. The streams were laid out based on the model's recommended minimum drainage areas of 1,601.23 ha for the 12.5 m DEM, 1,589.78 ha for the 30 m DEM, and 1,571.78 ha for the 90 m DEM. Moreover, five slope classes (0–10, 10–20, 20–40, 40–60, and 60–9,999) were defined in the model. As a result, a total number of 23 sub-basins and 82 HRUs were formed for all DEM resolutions. HRUs were generated based on land uses, soil types, and slope classes that made up at least 20% of a given sub-basin's area. Figure 5 shows the area delineation of the Mahabad Dam watershed in the GIS along with the sub-basins and the slope map of the watershed.

Figure 5

(a) Sub-basins in the model and (b) slope map of the watershed.

Figure 5

(a) Sub-basins in the model and (b) slope map of the watershed.

Using monthly streamflow data and implementing the DEM with the finest resolution (12.5 m), the SWAT model was calibrated from 1988 to 2004. For model auto-calibration, SWAT-CUP was run for 500 iterations in a single simulation, as recommended by the model developers. Moreover, iterations higher than 500 were also tested to be compared with the results from 500 iterations, but the changes were negligible. Out of the 33 parameters analyzed for their sensitivity on the runoff yield in the watershed, the first 17 (the half of the parameters that were more sensitive) were used in the model calibration. It should be noted, that all, or even less than half, of the parameters could be used for the model calibration, depending on the time available, cost permitted, and precision required. A 3-year model warmup from 1988 to 1991 was set in order to stabilize baseflow conditions in the model. Following calibration, streamflow was validated for the period of 2005–2012. Table 4 shows the calibrated values for the parameters, and Figures 6(a) and 6(b) show the observed and simulated streamflow values for the Kauter and Beytas Rivers, respectively.

Table 4

Calibrated values for parameters affecting streamflow in the Mahabad Dam watershed

ParametersDescriptionSub-basinsCalibrated value
SHALLST.gw Initial depth of water in the shallow aquifer (mmH2O) All 318.85 
DEEPST.gw Initial depth of water in the deep aquifer (mm) All 7,566.89 
GW_REVAP.gw Groundwater ‘revap’ coefficient All 0.02 
ALPHA_BF.gw Baseflow alpha factor (1/days) All 0.87 
RCHRG_DP.gw Deep aquifer percolation fraction All 0.127 
SNO_SUB.sub Initial snow water content (mm) All 28.308 
CH_N1.sub Manning's ‘n’ value for the tributary channels 11, 13, 14, 21, 24–26, 28, 36 26.46 
15–20, 22, 23, 27, 29–35, 37–45 17.279 
CN2.mgt Initial SCS runoff CN for moisture condition II 11, 13, 14, 21, 24–26, 28, 36 91.3 
15–20, 22, 23, 27, 29–35, 37–45 86.14 
ALPHA_BNK.rte Baseflow alpha factor for bank storage (days) All 0.787 
SFTMP.bsn Snowfall temperature (°C) All 4.83 
SMFMX.bsn Melt factor for snow on June 21 (mm/°C-day) All 4.6 
SMFMN.bsn Melt factor for snow on December 21 (mm/°C-day) All 0.55 
TIMP.bsn Snow pack temperature lag factor All 0.28 
OV_N.hru Manning's ‘n’ value for the overland flow 11, 13, 14, 21, 24–26, 28, 36 10.78 
15–20, 22, 23, 27, 29–35, 37–45 22.1 
SOL_ZMX.sol_TACONIC Maximum rooting depth of the soil profile (mm) All 1,124.15 
SOL_ZMX.sol_BENSON All 2,036.35 
SOL_AWC.sol_TACONIC Available water capacity of the soil layer (mmH2O/mm soil) All 0.12 
SOL_AWC.sol_BENSON All 0.21 
SOL_BD.sol_TACONIC Moist bulk density (g/cm3All 0.9 
SOL_BD.sol_BENSON All 1.26 
ParametersDescriptionSub-basinsCalibrated value
SHALLST.gw Initial depth of water in the shallow aquifer (mmH2O) All 318.85 
DEEPST.gw Initial depth of water in the deep aquifer (mm) All 7,566.89 
GW_REVAP.gw Groundwater ‘revap’ coefficient All 0.02 
ALPHA_BF.gw Baseflow alpha factor (1/days) All 0.87 
RCHRG_DP.gw Deep aquifer percolation fraction All 0.127 
SNO_SUB.sub Initial snow water content (mm) All 28.308 
CH_N1.sub Manning's ‘n’ value for the tributary channels 11, 13, 14, 21, 24–26, 28, 36 26.46 
15–20, 22, 23, 27, 29–35, 37–45 17.279 
CN2.mgt Initial SCS runoff CN for moisture condition II 11, 13, 14, 21, 24–26, 28, 36 91.3 
15–20, 22, 23, 27, 29–35, 37–45 86.14 
ALPHA_BNK.rte Baseflow alpha factor for bank storage (days) All 0.787 
SFTMP.bsn Snowfall temperature (°C) All 4.83 
SMFMX.bsn Melt factor for snow on June 21 (mm/°C-day) All 4.6 
SMFMN.bsn Melt factor for snow on December 21 (mm/°C-day) All 0.55 
TIMP.bsn Snow pack temperature lag factor All 0.28 
OV_N.hru Manning's ‘n’ value for the overland flow 11, 13, 14, 21, 24–26, 28, 36 10.78 
15–20, 22, 23, 27, 29–35, 37–45 22.1 
SOL_ZMX.sol_TACONIC Maximum rooting depth of the soil profile (mm) All 1,124.15 
SOL_ZMX.sol_BENSON All 2,036.35 
SOL_AWC.sol_TACONIC Available water capacity of the soil layer (mmH2O/mm soil) All 0.12 
SOL_AWC.sol_BENSON All 0.21 
SOL_BD.sol_TACONIC Moist bulk density (g/cm3All 0.9 
SOL_BD.sol_BENSON All 1.26 
Figure 6

Observed vs. simulated streamflow in (a) the Kauter River and (b) the Beytas River.

Figure 6

Observed vs. simulated streamflow in (a) the Kauter River and (b) the Beytas River.

The PBIAS value for streamflow calibration in the Kauter River is positive, which indicates the model's underestimation. On the other hand, the PBIAS value for streamflow validation in the Kauter River and streamflow calibration and validation in the Beytas River are negative, indicating the model's overestimation. The difference between simulated and observed values can be attributed to several factors, including measurement errors, instrument errors, and model errors. However, based on the NS and R2 values, the calibration and validation results are satisfactory.

Physical characteristics that affect runoff are land use, vegetation, soil type, drainage area, basin shape, elevation, slope, topography, orientation direction, drainage network patterns, and ponds, lakes, reservoirs, sinks, etc., in the basin, as these may prevent or alter runoff from continuing downstream.

In this study, land use, vegetation, and soil types were not altered during the simulations, and no ponds, lakes, reservoirs, sinks, etc., existed upstream of the hydrometric stations where the stream flows were measured. Therefore, the changes in DEM resolutions were expected to influence land features, such as the surface (drainage) areas, elevations, and slope distributions. These features, all together, influence drainage network patterns and ultimately surface runoff generation. After the watershed delineation and model calibration, these features were analyzed to determine the effects of implementing DEMs with different resolutions on them. The results are presented in Tables 5 and 6.

Table 5

Surface area, elevations, and average runoff within the watershed

DEM resolution (m)Surface area (ha)Minimum elevation (m)Maximum elevation (m)Mean elevation (m)Runoff (m3/s)
12.5 80,062.78 1,337 2,824 1,797.68 6.77 
30 79,489.19 1,320 2,806 1,778.38 6.72 
90 78,591.09 1,328 2,799 1,774.96 6.59 
DEM resolution (m)Surface area (ha)Minimum elevation (m)Maximum elevation (m)Mean elevation (m)Runoff (m3/s)
12.5 80,062.78 1,337 2,824 1,797.68 6.77 
30 79,489.19 1,320 2,806 1,778.38 6.72 
90 78,591.09 1,328 2,799 1,774.96 6.59 
Table 6

Slope distribution within the watershed

Slope class (°)12.5 m DEM
30 m DEM
90 m DEM
Surface area (ha)% areaSurface area (ha)% areaSurface area (ha)% area
0–10 36.75 0.05 34.66 0.04 403.99 0.51 
10–20 3,291.78 4.11 5,241.78 6.59 11,748.03 14.95 
20–40 53,207.81 66.46 51,885.50 65.27 59,107.19 75.21 
40–60 23,526.44 29.38 22,327.25 28.09 7,331.88 9.33 
Slope class (°)12.5 m DEM
30 m DEM
90 m DEM
Surface area (ha)% areaSurface area (ha)% areaSurface area (ha)% area
0–10 36.75 0.05 34.66 0.04 403.99 0.51 
10–20 3,291.78 4.11 5,241.78 6.59 11,748.03 14.95 
20–40 53,207.81 66.46 51,885.50 65.27 59,107.19 75.21 
40–60 23,526.44 29.38 22,327.25 28.09 7,331.88 9.33 

According to Table 5, the DEM with the finest resolution shows the highest values of the surface area, as well as minimum, mean, and maximum elevations, along with the highest average runoff value in the watershed. Moreover, as the DEM resolution gets coarser, the surface areas and average runoff decrease. However, this pattern for minimum elevation is not regular, and the 30 m DEM shows the lowest elevation.

Results presented in Table 5 show consistent trends of runoff and surface area variations. The runoffs based on the 12.5, 30, and 90 m DEMs, all slightly decreased or increased with the resolution, correspondingly, as did the surface areas. The 12.5 m DEM generated 0.74% more runoff compared with the 30 m DEM and 2.73% more runoff compared with the 90 m DEM. Moreover, the 30 m DEM generated 1.97% more runoff compared with the 90 m DEM. In comparison with the observed value of 8 m3/s at the watershed outlet, the 12.5 m DEM showed the most realistic results.

As presented in Table 6, for all DEM resolutions, the 20–40° slope class is predominant. Moreover, among the three DEMs, and except for the 10–20° slope class, the 12.5 m DEM shows the highest percentage of surface areas in all slope classes; as the DEM resolution gets coarser, this percentage decreases. Since the 60–9,999 slope class covered 0% of the watershed area, it is not presented in Table 6. Furthermore, Table 7 presents the impacts of DEM resolutions on the surface area, runoff, and predominant slopes in the sub-basin scale.

Table 7

Sub-basin surface area, runoff, and predominant slope for each DEM

Sub-basinSurface area (ha)
Runoff (m3/s)
Predominant slope (°)
12.5 m30 m90 m12.5 m30 m90 m12.5 m30 m90 m
800.6 794.9 785.9 6.77 6.72 6.59 0–10 0–10 0–10 
523.4 522.2 516.4 5.01 4.99 4.90 20–40 20–40 20–40 
17.73 18.25 17.82 0.11 0.11 0.11 20–40 20–40 20–40 
57.02 56.96 56.03 0.36 0.36 0.35 20–40 20–40 20–40 
37.3 72.02 71.97 0.23 0.15 0.15 20–40 20–40 20–40 
277.2 272.7 269.4 1.76 1.73 1.69 20–40 20–40 20–40 
201.7 197 194.6 1.29 1.26 1.23 20–40 20–40 20–40 
36.64 36.73 36.39 0.23 0.23 0.23 20–40 20–40 20–40 
47.67 47.26 46.58 0.39 0.39 0.38 20–40 20–40 20–40 
10 75.41 75.32 73.9 0.47 0.47 0.46 20–40 20–40 20–40 
11 427 426.2 422 4.41 4.40 4.32 20–40 20–40 20–40 
12 282.8 281.9 279.2 3.53 3.51 3.45 20–40 20–40 20–40 
13 45.05 44.96 44.72 0.28 0.28 0.28 20–40 20–40 20–40 
14 19.64 17.19 19.39 0.12 0.11 0.12 20–40 20–40 20–40 
15 22.39 22.48 22.48 0.14 0.14 0.14 20–40 20–40 20–40 
16 136.1 132.1 130.8 0.79 0.76 0.74 20–40 20–40 20–40 
17 228.4 227.7 225.2 3.19 3.18 3.12 20–40 20–40 20–40 
18 28.91 28.64 28.02 0.18 0.18 0.17 40–60 40–60 40–60 
19 139.3 138.6 136.7 2.66 2.65 2.61 20–40 20–40 20–40 
20 45.78 45.73 44.94 0.95 0.95 0.94 20–40 20–40 20–40 
21 75.82 75.42 74.21 1.60 1.59 1.56 20–40 20–40 20–40 
22 16.95 16.72 16.29 0.36 0.35 0.34 20–40 20–40 20–40 
23 29.13 28.96 28.51 0.60 0.60 0.59 20–40 20–40 20–40 
Sub-basinSurface area (ha)
Runoff (m3/s)
Predominant slope (°)
12.5 m30 m90 m12.5 m30 m90 m12.5 m30 m90 m
800.6 794.9 785.9 6.77 6.72 6.59 0–10 0–10 0–10 
523.4 522.2 516.4 5.01 4.99 4.90 20–40 20–40 20–40 
17.73 18.25 17.82 0.11 0.11 0.11 20–40 20–40 20–40 
57.02 56.96 56.03 0.36 0.36 0.35 20–40 20–40 20–40 
37.3 72.02 71.97 0.23 0.15 0.15 20–40 20–40 20–40 
277.2 272.7 269.4 1.76 1.73 1.69 20–40 20–40 20–40 
201.7 197 194.6 1.29 1.26 1.23 20–40 20–40 20–40 
36.64 36.73 36.39 0.23 0.23 0.23 20–40 20–40 20–40 
47.67 47.26 46.58 0.39 0.39 0.38 20–40 20–40 20–40 
10 75.41 75.32 73.9 0.47 0.47 0.46 20–40 20–40 20–40 
11 427 426.2 422 4.41 4.40 4.32 20–40 20–40 20–40 
12 282.8 281.9 279.2 3.53 3.51 3.45 20–40 20–40 20–40 
13 45.05 44.96 44.72 0.28 0.28 0.28 20–40 20–40 20–40 
14 19.64 17.19 19.39 0.12 0.11 0.12 20–40 20–40 20–40 
15 22.39 22.48 22.48 0.14 0.14 0.14 20–40 20–40 20–40 
16 136.1 132.1 130.8 0.79 0.76 0.74 20–40 20–40 20–40 
17 228.4 227.7 225.2 3.19 3.18 3.12 20–40 20–40 20–40 
18 28.91 28.64 28.02 0.18 0.18 0.17 40–60 40–60 40–60 
19 139.3 138.6 136.7 2.66 2.65 2.61 20–40 20–40 20–40 
20 45.78 45.73 44.94 0.95 0.95 0.94 20–40 20–40 20–40 
21 75.82 75.42 74.21 1.60 1.59 1.56 20–40 20–40 20–40 
22 16.95 16.72 16.29 0.36 0.35 0.34 20–40 20–40 20–40 
23 29.13 28.96 28.51 0.60 0.60 0.59 20–40 20–40 20–40 

According to Table 7, the same pattern exists in the sub-basin level, as in the whole watershed, except for some relatively small sub-basins (3, 5, 8, 14, and 15) that show discrepancies. In the SWAT model, sub-basins are formed based on the confluences of streams. In this case study, the SWAT model formed small sub-basins at points where multiple streams met. These small sub-basins do not follow the same patterns as in the bigger ones, in which the finer resolution DEMs resulted in higher surface areas and runoff. This can be attributed to the formation of smaller streams as the DEM resolution increases, which causes the formation of smaller sub-basins at the confluences. However, the overall higher surface runoff in the watershed, when using finer resolution DEMs, can be attributed to larger surface areas. Therefore, as a result of higher surface areas, more runoff was generated on land.

Comparing the findings of this study with other studies, consistent results were found in the work of the following researchers: Wolock & Price (1994) reported that a decrease in the DEM resolution in simulating 71 areas in Pennsylvania, New York, and New Jersey, USA, resulted in decreased streamflow. Similarly, Chaubey et al. (2005) came to the same conclusion in the agricultural Moores Creek watershed in Virginia, USA. Moreover, Cho & Lee (2001) reported that the finer resolution DEM simulated increased average slope and, hence, higher estimated runoff from the Broadhead watershed in New Jersey, USA.

On the other hand, Di Luzio et al. (2005) concluded that the DEM choice was critical for the realistic definition of a watershed and sub-watershed boundaries, as well as topographic input, and consequently, simulated outputs. However, the authors reported that a decrease in the DEM resolution resulted in increased runoff in the Goodwin Creek watershed in Mississippi, USA. Moreover, Dixon & Earls (2009) and Bosch et al. (2004) also reported inconsistent results compared with the findings of this study for the case studies of Charlie Creek in Florida and Little River in Georgia, USA, respectively. The authors concluded that although SWAT outputs were sensitive to the resolutions of the DEMs, the changes in the DEM resolution did not show a certain pattern with runoff output. As mentioned earlier, in this study, only the impacts of DEMs with different resolutions were investigated related to runoff generation and the sensitivity of parameters attributed to it, without altering other inputs. In the later-mentioned studies, in addition to different DEM resolutions, researchers used different resolutions of other GIS inputs, such as land-use/land-cover maps and soil maps, which could have caused results that are inconsistent with the findings of the current study.

CONCLUSIONS AND RECOMMENDATIONS

This study evaluated the effects of three DEM resolutions (ALOS PALSAR 12.5 m, SRTM 30 m, and ASTER 90 m) on the runoff yield, as well as the sensitivity of the parameters contributing to runoff, for the case study of the Mahabad Dam watershed in Iran. The following are the main findings of this study:

  • Comparing the sensitivity of parameters in different DEM resolutions showed that the 30 and 90 m DEMs shared the same highly sensitive parameters: the snowmelt base temperature (SMTMP.bsn), snowfall temperature (SFTMP.bsn), moist bulk density (SOL_BD.sol), deep aquifer percolation fraction (RCHRG_DP.gw), soil evaporation compensation factor (ESCO.hru), saturated hydraulic conductivity (SOL_K.sol), melt factor for snow on December 21 (SMFMN.bsn), initial SCS runoff CN for moisture condition II (CN2.mgt), and threshold depth of water in the shallow aquifer required for return flow to occur (GWQMN.gw); however, the 12.5 m DEM only shared the snowfall temperature (SFTMP.bsn) and melt factor for snow on December 21 (SMFMN.bsn) with the other two DEMs.

  • In the mountainous watershed of the Mahabad Dam, located in a region where snowfall is significant, the mean air temperature at which precipitation is equally likely to be rain as snow/freezing rain and the melt factor for snow on December 21 were among the most sensitive parameters for all DEM resolutions.

  • In the 12.5 m DEM, the Manning's ‘n’ value for the tributary channels (CH_N1.sub) was the most sensitive parameter, which can be attributed to the fact that by using finer resolution DEMs, more tributary streams appeared in the watershed; therefore, surface parameters such as Manning's ‘n’ value became more important.

  • Topologically, the total watershed surface area and elevations in the watershed varied due to DEM resolutions.

  • The 12.5 m DEM (finest resolution) showed the highest values for the surface area, as well as the minimum, mean, and maximum elevations in the delineated watershed.

  • As the distribution of slope within the watershed changed in different DEM resolutions, the surface parameters were affected more than other parameters.

  • Under the 12.5 m DEM, higher amounts of runoff were generated in the watershed.

  • The results showed consistent trends of runoff and surface area variations.

  • Comparing the generated runoff for each case, a 0.74% increase was observed when using the 12.5 DEM instead of the 30 m DEM. Furthermore, the 12.5 DEM generated 2.73% more runoff compared with the 90 m DEM, and the 30 m DEM generated 1.97% more runoff compared with the 90 m DEM.

The results of this study indicate that the resolution of data is a sensitive issue in environmental modeling, and the choice of input DEM resolution depends on the watershed response of interest. As for this case study, the maximum increase in the runoff yield was 2.73% when using the 12.5 m DEM compared with the 90 m DEM. Depending on the expected model output accuracy, this difference can be either considered negligible or significant. If it is negligible, researchers can use coarser resolution DEMs, which are freely available and help decrease computation time.

This study had some limitations. For instance, data from the hydrometric stations were only available for the time period of 1991–2012. More recent data or data records covering a longer period could have improved the calibration/validation of the model and, ultimately, the analysis of runoff generation in the watershed. For a future extension of this study, it is recommended for researchers to use more DEMs of different resolutions from available sources. It would also be valuable to repeat the same procedure in different topologies, in order to come to a more general conclusion on the impact of the DEM resolution on watershed yields, as well as to identify resolution thresholds where new parameters become more sensitive. The same procedure can also be implemented to analyze the impacts of the DEM resolution on sediment and nutrient yields in a watershed and the sensitivity of their parameters. Furthermore, researchers can consider using different resolutions of other model inputs, such as land-use maps or soil maps, and investigate their impacts on the sensitivity of parameters and watershed yields.

CONFLICTS OF INTEREST

The authors declare no conflict of interest.

REFERENCES

REFERENCES
Abbaspour
K. C.
Yang
J.
Maximov
I.
Siber
R.
Bogner
K.
Mieleitner
J.
Zobrist
J.
Srinivasan
R.
2007
Modelling hydrology and water quality in the pre-alpine/alpine Thur watershed using SWAT
.
Journal of Hydrology
333
,
413
430
.
doi:10.1016/j.jhydrol.2006.09.014
.
ALOS Research and Application Project of EORC, JAXA
2018
Available from: http://www.eorc.jaxa.jp/ALOS/en/index.htm (accessed 10 May 2018).
Arnold
J. G.
Srinivasan
R.
Muttiah
R. S.
Williams
J. R.
1998
Large area hydrologic modeling and assessment – part 1: model development
.
Journal of the American Water Resources Association
34
(
1
),
73
89
.
doi:10.1111/j.1752-1688.1998.tb05961.x
.
ASTER Global Digital Elevation Map
2018
Available from: https://asterweb.jpl.nasa.gov/gdem.asp (accessed 15 February 2018)
.
Bosch
D. D.
Sheridan
J. M.
Batten
H. L.
Arnold
J. G.
2004
Evaluation of the SWAT model on a coastal plain agricultural watershed
.
Transactions of the American Society of Agricultural Engineers
47
,
1493
1506
.
doi:10.13031/2013.17629
.
Chau
K. W.
2017
Use of meta-heuristic techniques in rainfall-runoff modelling
.
Water
9
(
3
),
article no. 186
.
doi:10.3390/w9030186
.
Chaubey
I.
Cotter
A. S.
Costello
T. A.
Soerens
T. S.
2005
Effect of DEM data resolution on SWAT output uncertainty
.
Hydrological Processes
19
,
621
628
.
doi:10.1002/hyp.5607
.
Chen
W.
Li
D. H.
Yang
K. J.
Tsai
F.
Seeboonruang
U.
2018
Identifying and comparing relatively high soil erosion sites with four DEMs
.
Ecological Engineering
120
,
449
463
.
doi:10.1016/j.ecoleng.2018.06.025
.
Cho
S. M.
Lee
M.
2001
Sensitivity considerations when modeling hydrologic processes with digital elevation model
.
Journal of the American Water Resources Association
37
,
931
934
.
doi:10.1111/j.1752-1688.2001.tb05523.x
.
Cotter
A. S.
Chaubey
I.
Costello
T. A.
Soerens
T. S.
Nelson
M. A.
2003
Water quality model output uncertainty as affected by spatial resolution of input data
.
Journal of the American Water Resources Association
39
,
977
986
.
doi:10.1111/j.1752-1688.2003.tb04420.x
.
Di Luzio
M.
Arnold
J. G.
Srinivasan
R.
2005
Effect of GIS data quality on small watershed streamflow and sediment simulations
.
Hydrological Processes
19
,
629
650
.
doi:10.1002/hyp.5612
.
Dixon
B.
Earls
J.
2009
Resample or not?! Effects of resolution of DEMs in watershed modelling
.
Hydrological Processes
23
,
1714
1724
.
doi:10.1002/hyp.7306
.
Earls
J.
Dixon
B.
2005
A comparative study of the effects of input resolution on the SWAT model
.
River Basin Management III
83
,
213
222
.
Food and Agriculture Organization of the United Nations (FAO)
2018a
Soil and Water Assessment Tool (SWAT)
. .
Food and Agriculture Organization of the United Nations (FAO)
2018b
Harmonized World Soil Database v 1.2
. .
Fotovatikhah
F.
Herrera
M.
Shamshirband
S.
Chau
K.
Ardabili
S. F.
Piran
J.
2018
Survey of computational intelligence as basis to big flood management: challenges, research directions and future work
.
Engineering Applications of Computational Fluid Mechanics
12
(
1
),
411
437
.
doi:10.1080/19942060.2018.1448896
.
Freeman
T. G.
1991
Calculating catchment area with divergent flow based on a regular grid
.
Computational Geosciences
17
(
3
),
413
422
.
doi:10.1016/0098-3004(91)90048-I
.
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
.
American Society of Agricultural and Biological Engineers
50
(
4
),
1211
1250
.
doi:10.13031/2013.23637
.
Gupta
H. V.
Sorooshian
S.
Yapo
P. O.
1999
Status of automatic calibration for hydrologic models: comparison with multilevel expert calibration
.
Journal of Hydrologic Engineering
4
(
2
),
135
143
.
doi:10.1061/(ASCE)1084-0699(1999)4:2(135)
.
I.R. OF IRAN Meteorological Organization (IRIMO)
2014
Available from: http://www.irimo.ir/eng/ (accessed 15 February 2018)
.
Liffner
J. W.
Hewa
G. A.
Peel
M. C.
2018
The sensitivity of catchment hypsometry and hypsometric properties to DEM resolution and polynomial order
.
Geomorphology
309
,
112
120
.
doi:10.1016/j.geomorph.2018.02.022
.
Lin
S.
Jing
C.
Chaplot
V.
Yu
X.
Zhang
Z.
Moore
N.
Wu
J.
2010
Effect of DEM resolution on SWAT outputs of runoff, sediment and nutrients
.
Hydrology and Earth System Sciences
7
,
4411
4435
.
doi:10.5194/hessd-7-4411-2010
.
Mahab Ghodss Consulting Engineering Company
2014
Available from: http://www.mahabghodss.com/ (accessed 15 February 2018)
.
Moazenzadeh
R.
Mohammadi
B.
Shamshirband
S.
Chau
K.
2018
Coupling a firefly algorithm with support vector regression to predict evaporation in northern Iran
.
Engineering Applications of Computational Fluid Mechanics
12
(
1
),
584
597
.
doi:10.1080/19942060.2018.1482476
.
Moriasi
D. N.
Arnold
J. G.
Van Liew
M. W.
Bingner
R. L.
Harmel
R. D.
Veith
T. L.
2007
Model evaluation guidelines for systematic quantification of accuracy in watershed simulations
.
Transactions of the American Society of Agricultural Engineers
50
,
885
900
.
doi:10.13031/2013.23153
.
Nazari-Sharabian
M.
Taheriyoun
M.
Karakouzian
M.
2019a
Surface runoff and pollutant load response to urbanization, climate variability, and low impact developments – a case study
.
Water Supply
19
(
8
).
doi:10.2166/ws.2019.123
.
Nazari-Sharabian
M.
Taheriyoun
M.
Ahmad
S.
Karakouzian
M.
Ahmadi
A.
2019b
Water quality modeling of Mahabad Dam watershed–reservoir system under climate change conditions, using SWAT and system dynamics
.
Water
11
,
article no. 394
.
doi:10.3390/w11020394
.
Ndomba
P.
Birhanu
B.
2008
Problems and prospects of SWAT model applications in Nilotic catchments: a review
.
Nile Water Science and Engineering Journal
1
,
41
52
.
Peter
C. B.
Ali
M. S.
Megan
W. L.
Mark
D. T.
Craig
S. T. D.
2013
Sediment delivery estimates in water quality models altered by resolution and source of topographic data
.
Journal of Environmental Quality
43
(
1
),
26
36
.
doi:10.2134/jeq2012.0148
.
Shang
X.
Wang
X.
Zhang
D.
Chen
W.
Chen
X.
Kong
H.
2012
An improved SWAT-based computational framework for identifying critical source areas for agricultural pollution at the lake basin scale
.
Ecological Modelling
226
,
1
10
.
doi:10.1016/j.ecolmodel.2011.11.030
.
Shuttle Radar Topography Mission (SRTM)
2018
Available from: https://www2.jpl.nasa.gov/srtm/ (accessed 10 May 2018)
.
Wang
W.
Chau
K.
Qiu
L.
Chen
Y.
2015
Improving forecasting accuracy of medium and long-term runoff using artificial neural network based on EEMD decomposition
.
Environmental Research
139
,
46
54
.
doi:10.1016/j.envres.2015.02.002
.
Wolock
D. M.
Price
C. V.
1994
Effects of digital elevation model map scale and data resolution on a topography-based watershed model
.
Water Resources Research
30
,
3041
3052
.
doi:10.1029/94WR01971
.
Zhang
P.
Liu
R.
Bao
Y.
Wang
J.
Yu
W.
Shen
Z.
2014
Uncertainty of SWAT model at different DEM resolutions in a large mountainous watershed
.
Water Research
53
(
1
),
132
144
.
doi:10.1016/j.watres.2014.01.018
.