Skip to content
BY 4.0 license Open Access Published by De Gruyter Open Access October 3, 2020

Hydrological process simulation in Manas River Basin using CMADS

Xinchen Gu, Guang Yang, Xinlin He, Li Zhao, Xiaolong Li, Pengfei Li, Bing Liu, Yongli Gao, Lianqing Xue and Aihua Long
From the journal Open Geosciences


The inability to conduct hydrological simulations in areas that lack historical meteorological data is an important factor limiting the development of watershed models, understanding of watershed water resources, and ultimate development of effective sustainability policies. This study focuses on the Manas River Basin (MRB), which is a high-altitude area with no meteorological stations and is located on the northern slope of the Tianshan Mountains, northern China. The hydrological processes were simulated using the China Meteorological Assimilation Driving Datasets for the SWAT model (CMADS) using the Soil and Water Assessment Tool (SWAT) model. Simulated runoff was corrected using calibration/uncertainty and sensitivity program for the SWAT. Through parameter sensitivity analysis, parameter calibration, and verification, the Nash–Sutcliffe efficiency (NSE), adjusted R-square (Radj2), and percentage bias (PBIAS) were selected for evaluation. The results were compared with statistics obtained from Kenswat Hydrological Station, where the monthly runoff simulation efficiency was NSE=0.64, Radj2=0.69, and PBIAS=0.9, and the daily runoff simulation efficiency was NSE=0.75, Radj2 = 0.75, PBIAS = −1.5. These results indicate that by employing CMADS data, hydrological processes within the MRB can be adequately simulated. This finding is significant, as CMADS provide continuous temporal, detailed, and high-spatial-resolution meteorological data that can be used to build a hydrological model with adequate accuracy in areas that lack historical meteorological data.

1 Introduction

It is difficult to make accurate and reasonable simulations of water resources and hydrological processes in areas that lack meteorological stations, such as cold dry areas in alpine regions where there are large fluctuations in runoff and where the complex topography and large fluctuations make it difficult to construct or maintain stations [1]. However, existing studies have shown that when meteorological data are lacking, the hydrological processes in models are greatly affected [2,3]. The MRB is located in the arid area with a unique mountain–oasis–desert ecosystem. As the largest oasis farming area in Xinjiang and the fourth largest irrigated agricultural area in China, its ecological problems have always been the focus of scientists [4,5,6,7,8,9,10,11].

The modeling of the watershed hydrological model without historical meteorological data is the focus of scholars [12,13,14,15]. The SWAT model is a semi-distributed hydrological model based on physical mechanisms. It has been successfully applied to watershed research in different countries and regions and is one of the most widely used hydrological models for evaluating water resources in different types of basins globally [16,17,18,19,20,21]. For example, Pulighe et al. [22] used the SWAT model to predict the runoff and nutrient load of a Mediterranean semiarid small watershed, and the evaluation results from the model were found to be consistent with observed values. Pradhan et al. [23] applied the SWAT model and an artificial neural network to basins in tropical, subtropical, arid, and semiarid regions in Asia and showed that the SWAT model and the artificial neural network model performed better when used in low- and high-flow simulations instead of average traffic, respectively. Mosbahi and Benabdallah [24] evaluated soil erosion land management in a semiarid watershed in Tunis using the SWAT model, and the SWAT model successfully reproduced the correlation between water flow, simulated runoff, and sediment yield. In addition, Mengistu et al. [12] used a physical similarity regionalization method to allocate, calibrate, and verify, using the SWAT model in a catchment area where observed data were limited, and accordingly proposed a new method for estimating the water balance composition in a data-deficient area in South Africa.

While many studies have shown the benefits of using the SWAT model, research has also been conducted to improve its simulation accuracy, especially in the MRB. For example, Luo et al. [25] extended the single reservoir basic flow method in the SWAT model to the MRB in Tianshan, China, and thought that the delayed response of a basic flow was more reasonable. However, owing to a lack of basic flow measurement data, the study was unable to determine which of the filter- or model-based methods provided the most representative evaluation results. Yang et al. [26] applied the SWAT model to assess the hydrological heterogeneity in arid and alpine basins in Northwest China; the results showed the SWAT model to be a robust and highly accurate tool for characterizing flow processes in alpine regions. The hydrologic heterogeneity in high-altitude areas was also determined to be sensitive to energy, whereas that in low-altitude areas was sensitive to drought stress. In addition, Meng et al. [13] drove a SWAT model using CMADS and found that the SWAT model provided satisfactory results through parameter calibration in areas that had a high glacier supply rate. Furthermore, Zhang [14] used a SWAT model driven by model resolution imaging spectrometer and tropical rainfall measuring mission remote sensing meteorological data and predicted that the runoff from the Manas River would continue to increase under future climate scenarios.

The main water source of the Manas River is perennial glacial meltwater and seasonal snowmelt water, and there are no traditional meteorological stations within the basin. Although the studies mentioned earlier have attempted to extend different methods and data to the MRB and establish a watershed model, the research content has been relatively single and no substantive solution to the lack of hydrometeorological data in the MRB has been found yet. For example, the physical similarity regionalization method [10,12] is a common method in the areas that are lacking in data. This method cannot meet the needs of more refined hydrological models to enable the development and associated management of this river basin [8,9]. Therefore, a more reasonable hydrological process simulation is required to provide scientific and theoretical support for policymakers. Due to the unique situation within the MRB and the lack of existing hydrological and meteorological data, it is currently not possible to analyze the water resources and implement relevant water policies. To fully understand the hydrological processes with the MRB, it is necessary to use a distributed hydrological model with existing meteorological and reanalysis data as the driving forces to simulate the hydrological processes within the MRB. In this respect, the SWAT model was employed and driven by CMADS meteorological data in this study, and land use data, soil data, and digital elevation were also employed. The applicability of using CMADS to compensate for the scarcity of meteorological data was then discussed from the perspective of the significance of different hydrological parameters.

2 Materials and methods

2.1 Study area

The MRB is located in the foothills of the northern Tianshan Mountains in China (Figure 1) and covers an area spanning 5.05 × 103 km2. The MRB has a continental arid climate: the annual average temperature is approximately 5.9°C (maximum and minimum temperature extremes are 40.0°C and −32.0°C, respectively), and the amount of yearly precipitation and evaporation are 110–300 mm and 1,500–2,000 mm, respectively [14]. The highest and lowest elevations of this basin are 5,138 m above mean sea level (AMSL) and 840 m, respectively. Perennial glacier cover exists at 3,600 m AMSL and higher, and perennial glacial meltwater and seasonal snowmelt water provide most of the water for the Manas River and the MRB. A large amount of perennial glacial meltwater and seasonal snowmelt water is produced in the summer, which is a relatively stable water resource within the basin, and low amounts of summer precipitation over a short duration occur in the Piedmont area, thus supplementing the water resource. Kenswat Hydrometric Station (43°58′E, 85°57′N, 840 m) is located at the pour point where the Manas River flows out of the mountain, and this is also the pour point of the entire basin. As shown in Figure 1, the Manas River originates from the Tianshan Mountains; it has an average annual flow of 12.9 × 108 m3 and total water resources of 25.73 × 108 m3. With respect to hydrogeological conditions, the fault and fold geological structural belts, which have the same east-west axis as the Tianshan Mountains, are distributed in the MRB, and they form a large anticline structure composed of Tertiary strata that block the groundwater connection between the basin and the downstream plain area. When river runoff passes through the anticline structure and enters the downstream alluvial plain, the surface water and the underground water have a mutual transport relationship.

Figure 1 Manas River Basin is located in the middle of the Eurasian continent and on the north slope of Tianshan Mountain. The elevation difference of the basin is large.

Figure 1

Manas River Basin is located in the middle of the Eurasian continent and on the north slope of Tianshan Mountain. The elevation difference of the basin is large.

2.2 CMADS data

Owing to the lack of meteorological stations and available historical meteorological data in the study area, this study used the CMADS dataset version 1.1 as the source of detailed and continuous high- spatial-resolution meteorological data. CMADS is a public dataset that was developed by Prof. Xianyong Meng from the China Agricultural University (CAU). CMADS integrates local analysis and prediction system/space–time multiscale analysis system technology and adopts many technologies and scientific methods including data nesting, resampling model projection, and bilinear interpolation [11]. CMADS series datasets can be used to drive various hydrological models such as SWAT, variable infiltration capacity, and the storm water management model. The datasets also allow the users to easily extract various meteorological elements and conduct a detailed climate analysis. The data sources of the CMADS series include nearly 40,000 regional automatic stations within 2,421 national automatic and commercial assessment centers in China [15]. This ensures the wide applicability of CMADS data in China and greatly improves the accuracy of data. The spatial resolution of the CMADS data grid is 0.25° × 0.25°, and the data relate to the years spanning from 2008 to 2016. CMADS data were obtained over a nine-year period from 1 January 2008 to 31 December 2016 and they include many watersheds throughout East Asia [15,27,28,29]. Studies have also used CMADS data and the Penman–Monteith method to calculate potential evapotranspiration (PET) across China, and a reliable performance has been noted [29,30].

Precipitation data were stitched using the climate prediction center morphing technique (CMORPH) global precipitation products and the National Meteorological Information Center’s data of China (which is based on CMORPH integrated precipitation products): the latter contains daily precipitation records observed at 2,400 national meteorological stations and CMORPH satellite’s inversion precipitation products. The inversion algorithm for incoming solar radiation at the ground surface employs the discrete longitudinal method by Stamens et al. [31] to calculate radiation transmission [32]. The SWAT was then employed to automatically read data relating to five meteorological elements (rainfall, temperature, relative humidity, solar radiation, and wind speed data) obtained from 12 grid nodes (Table 1).

Table 1

CMADS meteorological grid node information

Grid nodesLatitudeLongitudeAltitude (m)

2.3 SWAT model

The SWAT watershed area is discretized from a given digital elevation model (DEM) to several sub-basins. The hydrological response unit (HRU) of each sub-basin has similar land use, soil type, and slope classification, which is the smallest spatial unit of the model. The watershed HRU is a region that has relatively single and uniform underlying surface characteristics and has similar hydrological characteristics. To further reveal this combination difference within a basin and reflect the spatial heterogeneity of the underlying surface of the basin, it is necessary to divide each sub-basin into several HRUs, each of which has only a single land use type, soil type, and class of terrain slope. The model is then used to simulate hydrological processes such as evapotranspiration, filtration, surface runoff, groundwater runoff, and sediment erosion in each HRU. The runoff from each HRU first flows into the main channel of each sub-basin and then flows from one sub-basin to another until it finally reaches the pour point.

The water balance equation can be expressed as follows:


where SWt is the final soil moisture content (mm); SW0 is the initial soil moisture content (mm); t is the simulation time (days); Rday is the daily precipitation (mm water); Qsurf is the daily surface runoff (mm water); Ea is the daily evapotranspiration (mm); wseep is the amount of water entering the aeration zone from the soil profile on a given day (mm); and Qgw is the return flow on a given date (mm). SW0 is the initial soil moisture content obtained from soil moisture data provided by harmonized world soil database (HWSD).

Based on dynamic storage models and the gradient of the terrain, the soil hydraulic conductivity (SOL_K), spatiotemporal changes in soil moisture, and characteristics of soil internal flow can be calculated using the SWAT. However, it is also important to consider lateral flow with respect to collecting soil water in the areas with high surface water conductivity. In this respect, Sloan [33] combined the SWAT with a groundwater flow movement storage model to simultaneously calculate seepage. Shallow aquifers collect groundwater into the main channel of the sub-basin, and surface runoff is thus from rainfall that remains after plant closure and soil infiltration. The Green–Ampt infiltration method and the soil conservation service (SCS) curve method can be used to estimate surface runoff [34]. In addition, the peak runoff rate reflects the erosion force of rainstorms and can be used to predict the amount of sediment loss.

Evapotranspiration includes canopy evaporation and soil evaporation, and many methods can be used to simulate PET, including the Priestley–Taylor method [35], the Penman–Montes method, and the Hargreaves method [36], all of which are included in SWAT.

Based on the study of evapotranspiration in arid and semiarid areas, Hargreaves and Samani’s formula [36] can be used to calculate PET based on temperature and solar radiation. As this method provides good calculation results, the United Nations Food and Agriculture Organization (FAO) recommends using it when meteorological data are lacking. In this study, therefore, the Hargreaves method [36] is used to simulate PET as follows:


where ET0Har is PET calculated using Hargreaves’ method (mm/day); Tx is the daily maximum temperature (°C); Tn is the minimum temperature (°C); and Ra is the solar radiation at the top of the atmosphere (MJ/(m2 day)) and can be calculated using the latitude and atmospheric top radiation table provided by the FAO.


The DEM data for the basin were provided by the geospatial data cloud platform of the computer network information center of the Chinese Academy of Sciences ( DEM is ASTER GDEM v2 released by the Ministry of Economy, the Trade and Industry of Japan, and the National Aeronautics and Space Administration of the United States in 2015, with a resolution of 30 m. ArcMap was used to fill the depressions relating to DEM data to reduce errors, and to calculate the DEM and generate the boundaries, river network, and slope data for the basin and sub-basins. Google’s satellite image data provided by the ArcGIS Online basemap were employed (Figure 1).

The soil data with a resolution of 1 km employed were based on the HWSD version 1.1 provided by the scientific data center for cold and dry areas ( Land use data with a resolution of 30 m for the summer of 2015 were obtained from the National Geoscience Data Sharing Centre (, which uses Landsat-8 remote sensing images with a small amount of cloud and rich surface feature information that was obtained through remote sensing interpretations.

In this study, in accordance with the DEM, the SWAT program automatically divided the MRB into 21 sub-basins. In each of these sub-basins, the runoff flows into the pour point and then into the next sub-basin. The runoff ultimately flows into sub-basin 1, which is the river pour point, and the Kenswat Hydrological Station is also located in sub-basin 1. The 12 CMADS meteorological grid nodes located around the MRB are shown in Figure 2(b). The SWAT model used the data from these 12 meteorological grid nodes (Table 1).

Figure 2 (a) The SWAT program divided the MRB into 21 sub-basins and (b) employed data from 12 CMADS meteorological grid nodes within the area. The CMADS meteorological grid in the figure almost perfectly covers the whole basin.

Figure 2

(a) The SWAT program divided the MRB into 21 sub-basins and (b) employed data from 12 CMADS meteorological grid nodes within the area. The CMADS meteorological grid in the figure almost perfectly covers the whole basin.

The land use, soil type, and slope-gradient distribution within the Manas watershed are shown in Figure 3. The soil type distribution is shown in Figure 3(a) and its percentage distribution is as follows: gelic leptosol soil (47.61%), glacier (21.54%), mollic leptosol soil (18.17%), haplic greyzem soil (4.83%), luvic kastanozem soil (4.32%), haplic chernozem soil (2.60%), haplic kastanozem soil (0.64%), and calcaric fluvisol soil (0.30%). The percentage distributions of the land use types shown in Figure 3(b) are as follows: grassland (48.49%), impervious surface (26.20%), ice on glacier (20.29%), forest (4.97%), cropland (0.04%), and bareland (0.01%). Furthermore, the percentage slope-gradient distribution relating to Figure 3(c) is as follows: 0–15° (7.65%), 15–25° (9.12%), 25–75° (63.19%), and ≥75° (9.12%).

Figure 3 Map of the HRUs showing the different (a) soil types, (b) land use types, and (c) slope gradients within the MRB. Different soil types, land use, and slope gradient make up different HRUs.

Figure 3

Map of the HRUs showing the different (a) soil types, (b) land use types, and (c) slope gradients within the MRB. Different soil types, land use, and slope gradient make up different HRUs.

The elevation differs considerably within the area, and the soil types and vegetation coverage thus also show significant vertical zone differentiation. Water absorption capacity thus differs in accordance with the soil properties. For example, the particle sizes of the various soil types differ, and permeability thus differs, which affects the amount of water absorbed and subsequent runoff. Furthermore, the different vegetation types and their litter layers have varying degrees of precipitation interception and can thus enhance surface runoff and prolong infiltration time. Vegetation with higher coverage reduces the depth of surface runoff and increases soil and underground runoff. The interaction between these factors means that the confluence of precipitation and ice/snowmelt water is complex and is thus an important hydrological process.

In summary, the SWAT model of the MRB was ultimately constructed using sub-basin divisions, imported data from CMADS meteorological grid nodes, land use data, soil type data, slope division data, HRU generation, a time scale, and a preheating period setting.

2.5 Model calibration and validation

In this study, the sequential uncertainty fitting algorithm in SWAT-CUP was used to estimate the unknown parameters iteratively through a sequential fitting process. This program also considers the uncertainty of the model input, model structure, input parameters, and observation data. During the calibration process, this program uses the global sensitivity analysis method, considers the balance phenomenon in the calibration process, and uses the t-stat (Student’s t-test) and p-value method to evaluate sensitivity. These different parameter settings generate acceptable runoff curves. In this process, the higher the sensitivity of the parameters, the greater the absolute value of the t-stat and the closer the p-value is to zero [37].

Based on the comprehensive evaluation method, Moriasi proposed four quantitative statistical indicators [38]: the NSE, PBIAS, relative root-mean-square error (RRMSE), and the Radj2. This study used three common indicators, the NSE, PBIAS, and Radj2, to quantify the model performance, as these three performance indexes match the simulation value of the quantitative numerical model with the measured value. The NSE is a statistical indicator that quantifies the relative size of the residual variance [39], and the PBIAS shows whether the average trend of the simulation data is greater or less than that of the corresponding observation data [40], which is different from the variance used for observation data.

The NSE, Radj2, and PBIAS are calculated as follows:


where R2 is the R-square; Radj2 is the adjusted R-square; NSE is the Nash–Sutcliffe efficiency coefficient; PBIAS is the deviation of the data being evaluated (expressed as a percentage); oi is the measured runoff (m3/s); si is the simulated runoff (m3/s); o¯ is the mean measured runoff during the simulation period (m3/s); s¯ is the mean simulated runoff during the simulation period (m3/s); n is the number of flow measurements in the analysis; and k is the number of independent variables.

The closer the NSE and Radj2 values are to 1, the better the SWAT performance. Furthermore, when PBIAS is close to zero, the simulation is more accurate. Moriasi proposed that when NSE ≥ 0.5 and PBIAS is within ± 25%, the model is considered to provide a satisfactory simulation. This study used the criteria developed by Moriasi (Table 2) to conduct the evaluation.

Table 2

Classification of performance grades for statistical indicators: NSE, PBIAS, and Radj2

Performance gradesNSEPBIAS
Very good1.00 ≥ NSE ≥ 0.75|PBIAS| < 10
Good0.75 > NSE ≥ 0.6515 ≥ |PBIAS| > 10
Satisfactory0.65 > NSE ≥ 0.525 ≥ |PBIAS| > 15
UnsatisfactoryNSE < 0.50|PBIAS| > 25

3 Analysis of results

In accordance with the topography and river network, ArcSWAT was used to divide the study area into 21 sub-basins. To obtain high-resolution land use, soil properties, and slope, these sub-basins were further divided into 136 HRUs. The SWAT model divided each sub-basin into several HRUs and then selected the HRUs as the hydrological unit to be simulated. On this basis, monthly and daily-scale simulations were conducted using the SWAT model to determine the model’s capabilities and the model was calibrated according to the runoff observed at the Kenswat Hydrometric Station. The results of the sensitivity analysis determined 15 parameters that could be used to correct the model, and accuracy requirements were achieved in the validation period. The final runoff simulation results were NSE = 0.64, Radj2 = 0.69, and PBIAS = −0.9 (monthly); and NSE = 0.75, Radj2 = 0.75, and PBIAS = −1.5 (daily).

3.1 Sensitivity analysis

A sensitivity analysis was conducted to determine the parameters that had a strong influence on the snowmelt water runoff simulation. The sensitivity analysis was conducted by incorporating the Latin-Hypercube–One factor-At-a-Time (LH-OAT) method into SWAT-CUP. A parameter sensitivity analysis was conducted for common parameters; the parameters with high model sensitivity were selected for parameter adjustment, and 15 parameters were finally selected (Table 3).

Table 3

Sensitivity and optimal parameter values

ParametersDescriptionSensitivity priorityt-statp-valueOptimal value
R_Precipitation{2010001–2016365}.pcpDaily rainfall from 2010 to 2016 (mm)1−27.591.5 × 10−1241.90
V_SMTMP.bsnSnowmelt base temperature (°C)215.131.2 × 10−466.71
V_SMFMX.bsnMaximum melt rate for snow during year (occurs on summer solstice) (mm/°C day)3−14.712.0 × 10−441.09
R_SOL_K.solSaturated hydraulic conductivity (mm/h)4−12.571.0 × 10−33−0.46
V_SFTMP.bsnSnowfall temperature (°C)57.171.4 × 10−126.34
R_SLSUBBSN.hruAverage slope length (m)64.997.2 × 10−70.39
R_CN2.mgtSCS runoff curve number for moisture condition II7−4.624.4 × 10−6−0.45
R_HRU_SLP.hruAverage slope steepness (m/m)8−3.534.3 × 10−40.29
R_SOL_BD.solMoist bulk density (g/cm3)9−3.141.7 × 10−30.10
V_RCHRG_DP.gwDeep aquifer percolation fraction102.133.3 × 10−20.29
V_ALPHA_BF.gwBaseflow alpha factor (days)11−1.530.130.58
R_SOL_AWC.solAvailable water capacity of the soil layer (mm/mm)120.950.340.11
V_ESCO.bsnSoil evaporation compensation factor13−0.950.340.29
V_GWQMN.gwThreshold depth of water required in shallow aquifer for return flow to occur (mm)140.860.39412.83
R_OV_N.hruManning’s “n” value for overland flow15−0.740.46−0.25

R multiplies the existing value with (1 + the given value); V replaces the exiting value with the given value

Table 3 lists the sensitivity of the parameters considered in the final calibration results. The precipitation {2010001–2016365} parameter had the highest sensitivity and was adjusted to 1.90 times the rainfall in 2010–2016. The optimal value of the snowmelt base temperature (SMTMP) was 6.71, which indicated that snowmelt began at 6.71°C. Furthermore, the optimal value of the snowfall base temperature (SFTMP) was 6.34, which indicated that the rainfall began to change to snowfall at 6.34°C. This table also lists other parameters, such as hydraulic conductivity (SOL_K), the SCS runoff curve coefficient (CN2), slope (HRU_SLP), and the soil effective water content (SOL_AWC).

The 15 hydrological parameters used are all within a reasonable range, and some are further discussed later to explore their practical significance.

3.2 Model calibration and validation

The runoff data from Kenswat Hydrological Station from 2008 to 2016 were used to preheat, calibrate, and verify the use of the SWAT model within the MRB. In this study, the warm-up time was set to two years (2008–2009) to initialize the SWAT model. The above-mentioned 15 parameters and parameter values (Table 3) were determined by the calibration of surface runoff data from 2010 to 2013, and the evaluation standard met the requirements. The 2014–2016 surface runoff data were then used for the verification, and the 15 parameter values determined during calibration were entered into the SWAT model to evaluate similarities and differences between simulated runoff values and measured runoff values. In general, the runoff simulated by the SWAT model was similar to that of measured runoff, and the flood occurrence times of simulated and measured runoff data were in good agreement.

In the monthly scale simulation (Figure 4), the peak summer flows in 2010, 2012, and 2014 were higher than the measured runoff; the peak summer flows in 2011 and 2013 were lower than the measured runoff; and the peak summer flows in 2015 and 2016 were in better agreement with observed values. The simulation results were generally good throughout 2010–2016 (with the exception of 2015), although the winter flow in other years was lower than measured values.

Figure 4 Observed (at Kenswat hydrometric station) and calibrated monthly average runoff between 2010 and 2016 showing high consistency between the results. The runoff was almost perfectly reproduced in 2015–2016.

Figure 4

Observed (at Kenswat hydrometric station) and calibrated monthly average runoff between 2010 and 2016 showing high consistency between the results. The runoff was almost perfectly reproduced in 2015–2016.

The results of the daily-scale simulation (Figure 5) show that the peak summer flows in 2010, 2012, 2013, and 2016 were similar to the actual measurements, but peak summer flows in 2011, 2014, and 2015 were quite different from the actual measurements (and that of 2014 was relatively large). The winter flows in 2010 and 2014 were relatively low, while those of other years were relatively high.

Figure 5 Measured runoff values (at Kenswat Hydrometric Station) and simulated runoff values on a daily scale showing high consistency between the results, especially the runoff in summer.

Figure 5

Measured runoff values (at Kenswat Hydrometric Station) and simulated runoff values on a daily scale showing high consistency between the results, especially the runoff in summer.

The calibration results for the monthly scale simulation (2010–2013) (Table 4) were NSE = 0.64, Radj2 = 0.69, and PBIAS = 0.90, and the validation results (2014–2016) were NSE = 0.82, Radj2 = 0.83, and PBIAS = −3.80. The calibration results for the daily-scale simulation (2010–2013) were NSE = 0.75, Radj2 = 0.75, and PBIAS = −1.50, and the validation results (2014–2016) were NSE = 0.66, Radj2 = 0.67, and PBIAS = −12.60.

Table 4

Values obtained during calibration and verification of evaluation indexes

Statistical indicatorsCalibration (2010–2013)Validation (2014–2016)

The performance criteria for the monthly and daily-scale simulations shown in Table 2 reveal very good calibration and verification simulation results for three performance parameters, NSE, Radj2, and PBIAS. The model’s simulation values are consistent with the measured values; therefore, the accuracy of the model satisfies the requirements after parameter calibration.

4 Discussion

The MRB is a small watershed located on the northern slope of the Tianshan Mountains. In the past 40 years, owing to vigorous developments in water-saving irrigation, the artificial oasis area in the MRB has expanded by 129.56% [8], and the social and economic effects relating to these oases have actively promoted development in this area [9]. The demand for water resources in the MRB has thus simultaneously increased [41]. Although human activities mainly affect the plain area below the Kenswat Hydrometric Station in the MRB, the generation and concentration of runoff in the mountain area are mainly affected by climate change.

It is necessary to simulate and analyze water resources within the basin to enable their control and associated sustainable regional development. However, owing to the characteristics of the MRB, existing hydrological and meteorological data are scarce and thus cannot be used in model simulations of water resources. Therefore, in this study, CMADS data were applied to the SWAT model to explore the hydrological processes occurring within the basin and to compensate for the missing hydrological and meteorological data.

The MRB was divided into 21 sub-basins and 136 HRUs using the SWAT model. The basin was modeled using ArcMap-ArcSWAT 2012 and adjusted using SWAT-CUP. Most of the parameters used in this process were found to fall within a reasonable range. In addition, the optimal value of the SMTMP was found to be 6.71, and the optimal value of the SFTMP was 6.34; this indicates that the temperature of snowmelt water in the MRB was about 6°C, which is consistent with previous studies that have concluded that snowfall events in the Tianshan Mountain area and its surrounding areas basically occur between −35 and 5°C [42].

Simulation data can be based on data obtained from a wide range of data sources, such as meteorological observation stations, to compensate for lacking historical meteorological data within an area [11,15,43]. When establishing the hydrological model, the spatial resolution of the simulation dataset was high, and available CMADS data were thus employed. Compared with the Climatic Research Unit (CRU) data and other widely used datasets, CMADS data have a higher spatial resolution and are more suitably applied in a small watershed [43,44,45,46,47]. For a watershed lacking meteorological data, the CMADS dataset can ensure accuracy on large spatial and temporal scales. It is common to employ available meteorological data from stations to calibrate a dataset, as these can smooth out the typicality and particularity of the watershed with respect to certain types of meteorological elements. Applying such data is necessary for smoothing any distortions within the dataset. However, there was no available historical meteorological data in the MRB due to the lack of stations. In this study, there was a certain amount of distortion in the rainfall data from CMADS when applied to the watershed. The CMADS rainfall data applied to the MRB were thus adjusted (1.9 times the original), and the resulting adjusted rainfall data able to drive the SWAT model adequately and model the MRB were found. This result shows that for areas that lack meteorological data, datasets such as CMADS and CRU can be used for hydrological analysis, simulation, and modeling. The SWAT model adapted well to the MRB and a more reasonable hydrological process was obtained when CMADS data were combined with the model. With respect to the simulation effect and scale, compared with the hydrological model [10,12] established by the physical similarity regionalization method, CMADS data were capable of meeting the needs of increased refinement and scale diversification.

After modifying the CMADS rainfall data used in this study, an ideal hydrological simulation effect was obtained. This indicates that CMADS rainfall data may require modification prior to use in watershed hydrological simulation studies, as described in a previous study [13]. However, due to the different scale model simulations, certain ecological fallacies [48] may be introduced; for example, the model quality index on a monthly scale was not synchronous with that on a daily scale. In the future work, the aim will be to quantitatively evaluate the accuracy of using CMADS meteorological data (not only rainfall data) in the MRB, to provide enhanced results for use in water resource analysis and associated management. In general, the SWAT model based on CMADS data provided continuous and high-spatial-resolution meteorological data and enabled good simulation of the annual hydrological conditions of the MRB, which can be used to explain the impact of climate change and human activities on water resources in the basin. These data have good applicability within the MRB and can serve as a reference when conducting a hydrological analysis of this basin in Northwest China, which has few meteorological stations [49].

5 Conclusions

This study used CMADS data and the SWAT model to simulate and verify the runoff within the MRB. With respect to the uncertainty of the model simulation, monthly and daily runoff data were obtained from Kenswat Hydrological Station and calibrated and verified by conducting a parameter sensitivity analysis and parameter optimization calibration. A hydrological model of the MRB was thus established, and the following conclusions were drawn:

  1. Through the parameter sensitivity analysis and parameter calibration verification of the temperature index method, the SWAT model was shown to have good adaptability for use in the MRB. However, with respect to the impact of climate change and human factors within the MRB, and the potentially large errors in the summers of some years that occurred in the model, it will be necessary to conduct further studies to continuously verify whether the CMADS data and SWAT model are suitable for use in this region.

  2. Runoff at Kenswat Hydrological Station in the MRB was located and simulated using CMADS data and the SWAT model. These results show that the SWAT model driven by CMADS can well reproduce the hydrological process of the MRB on monthly and daily scales.

  3. CMADS can provide continuous temporal and high-spatial-resolution meteorological data for simulating the water resources in the MRB, without the need for meteorological data. However, compared with traditional weather stations, the weather data provided by CMADS require adjustment prior to modeling.

In summary, CMADS data and the SWAT model were used to simulate the hydrology of the MRB, which has a high glacier recharge rate. The results show that CMADS data can serve as a reference for studying the use of the SWAT model in evaluating water resources in cold and dry areas that lack meteorological data.


This research was funded by the National Natural Science Foundation of China (Grant No. U1803244); Key Technologies Research and Development Program (Grant No. 2017YFC0404303); Xinjiang Production and Construction Corps (Grant Nos. 2018CB023, CZ027204, 2018AB027, and 2018BC007); and Shihezi University (Grant Nos. CXRC201801 and RCZK2018C22). This work was also supported by the Talent Program of Xinjiang Production and Construction Corps and Key Laboratory of Modern Water-Saving Irrigation.

  1. Author contributions: XG, GY, and LZ conceptualized the research; methodology was outlined by GY, LZ, and XG; PL and XL worked with the software; the validation was done by XG, GY, and XH; formal analysis was carried out by BL and XL; the study was investigated by BL; the necessary resources were provided by LX; the data curation was done by GY; the original draft was prepared by XG, GY, and XH; the visualization was created by LA and YG; XH, YG, and XG supervised the process; the whole project was administrated by XH; and funding was managed by XH. All authors have read and agreed to the publication of this manuscript.

  2. Conflicts of interest: The authors declare no conflict of interest.


[1] Wang XL, Luo Y, Sun L, Zhang YQ. Assessing the effects of precipitation and temperature changes on hydrological processes in a glacier-dominated catchment. Hydrol Proc. 2015;29:4830–45, 10.1002/hyp.10538.Search in Google Scholar

[2] Kan GY, He XY, Li JR, Ding LQ, Hong Y, Zhang HB, et al. Computer aided numerical methods for hydrological model calibration: an overview and recent development. Arch Comput Methods Eng. 2019;26:35–59, 10.1007/s11831-017-9224-5.Search in Google Scholar

[3] Olichwer T, Tarka R. Impact of climate change on the groundwater run-off in south-west Poland. Open Geosci. 2015;7:1–14, 10.1515/geo-2015-0001.Search in Google Scholar

[4] Liu Q, Yang Z, Shi H, Wang Z. Ecological risk assessment of geohazards in Natural World Heritage Sites: an empirical analysis of Bogda. Tianshan Open Geosci. 2019;11:327–40, 10.1515/geo-2019-0026.Search in Google Scholar

[5] Yang M, Kang L, Chen H, Zhou M, Zhang J. Lithological mapping of East Tianshan area using integrated data fused by Chinese GF-1 PAN and ASTER multi-spectral data. Open Geosci. 2018;10:532–43, 10.1515/geo-2018-0042.Search in Google Scholar

[6] Zhou KF, Wang SS. Spectral properties of weathered and fresh rock surfaces in the Xiemisitai metallogenic belt, NW Xinjiang, China. Open Geosci. 2017;9:322–39. 10.1515/geo-2017-0027.Search in Google Scholar

[7] Turk KGB, Aljughaiman AS. Land use/land cover assessment as related to soil and irrigation water salinity over an oasis in arid environment. Open Geosci. 2020;12:220–31. 10.1515/geo-2020-0103.Search in Google Scholar

[8] Yang G, Xue LQ, He XL, Wang C, Long AH. Change in land use and evapotranspiration in the Manas River Basin. China long-term water-saving measures. Sci Rep. 2017;7:17874. 10.1038/s41598-017-18030-5.Search in Google Scholar PubMed PubMed Central

[9] Yang G, Li F, Chen D, He X, Xue L, Long A. Assessment of changes in oasis scale and water management in the arid Manas River Basin, north western China. Sci Total Environ. 2019;691:506–15. 10.1016/j.scitotenv.2019.07.143.Search in Google Scholar PubMed

[10] He XL, Guo SL. Impacts of climate change on hydrology and water resources in the Manas River Basin. Adv Water Sci. 1998;78–84 (in Chinese with English abstract).Search in Google Scholar

[11] Meng XY, Wang H, Lei XH, Cai SY. Simulation, validation, and analysis of the hydrological components of Jing and Bo River Basin based on the SWAT model driven by CMADS. Acta Ecol Sin. 2017;37:7114–27.Search in Google Scholar

[12] Mengistu AG, Van RLD, Woyessa YE. Techniques for calibration and validation of SWAT model in data scarce arid and semi-arid catchments in South Africa. J Hydrol Region Stud. 2019;25:100621. 10.1016/j.ejrh.2019.100621.Search in Google Scholar

[13] Meng XY, Wang H, Lei XH, Cai SY, Wu HJ, Ji XN, et al. Hydrological modeling in the Manas River Basin using soil and water assessment tool driven by CMADS. Tehnicki Vjesnik-Technical Gaz. 2017;24:525–34. 10.17559/tv-20170108133334.Search in Google Scholar

[14] Zhang ZY. Modeling hydrological processes in main runoff generating area of Manasi River Basin, Xinjiang. PhD thesis. China: Shihezi University; 2018.Search in Google Scholar

[15] Meng X, Wang H. Significance of the China meteorological assimilation driving datasets for the SWAT model (CMADS) of East Asia. Water. 2017;9:765. 10.3390/w9100765.Search in Google Scholar

[16] Galvan L, Oiias M, Canovas CR, Sarmiento AM, Nieto JM. Hydrological modeling of a watershed affected by acid mine drainage (Odiel River, SW Spain). Assessment of the pollutant contributing areas. J Hydrol. 2016;540:196–206. 10.1016/j.jhydrol.2016.06.005.Search in Google Scholar

[17] Volk M, Bosch D, Nangia V, Narasimhan B. SWAT: agricultural water and nonpoint source pollution management at a watershed scale. Agric Water Manag. 2016;175:1–3. 10.1016/j.agwat.2016.06.013.Search in Google Scholar

[18] Ghaffari G, Keesstra S, Ghodousi J, Ahmadi H. SWAT-simulated hydrological impact of land-use change in the Zanjanrood Basin, Northwest Iran. Hydrol Proc. 2010;24:892–903. 10.1002/hyp.7530.Search in Google Scholar

[19] Zeynep O, Kentel E, Alp E. Determination of unit nutrient loads for different land uses in wet periods through modelling and optimization for a semi-arid region. J Hydrol. 2016;540:40–9. 10.1016/j.jhydrol.2016.05.074.Search in Google Scholar

[20] Cho KH, Pachepsky YA, Kim M, Pyo J, Park MH, Kim YM, et al. Modeling seasonal variability of fecal coliform in natural surface waters using the modified SWAT. J Hydrol. 2016;535:377–85. 10.1016/j.jhydrol.2016.01.084.Search in Google Scholar

[21] Pereira DR, Martinez MA, Pruski FF, Silva DD. Hydrological simulation in a basin of typical tropical climate and soil using the SWAT model. Part I: calibration and validation tests. J Hydrol Region Stud. 2016;7:14–37. 10.1016/j.ejrh.2016.05.002.Search in Google Scholar

[22] Pulighe G, Bonati G, Colangeli M, Traverso L, Lupia F, Altobelli F, et al. Predicting streamflow and nutrient loadings in a semi-arid Mediterranean watershed with ephemeral streams using the SWAT model. Agronomy-Basel. 2020;10:2. 10.3390/agronomy10010002.Search in Google Scholar

[23] Pradhan P, Tingsanchali T, Shrestha S. Evaluation of soil and water assessment tool and artificial neural network models for hydrologic simulation in different climatic regions of Asia. Sci Total Environ. 2020;701:134308. 10.1016/j.scitotenv.2019.134308.Search in Google Scholar PubMed

[24] Mosbahi M, Benabdallah S. Assessment of land management practices on soil erosion using SWAT model in a Tunisian semi-arid catchment. J Soils Sediment. 2020;20:1129–39. 10.1007/s11368-019-02443-y.Search in Google Scholar

[25] Luo Y, Arnold J, Allen P, Chen X. Baseflow simulation using SWAT model in an inland river basin in Tianshan Mountains, Northwest China. Hydrol Earth Syst Sci. 2012;16:1259–67. 10.5194/hess-16-1259-2012.Search in Google Scholar

[26] Yang L, Feng Q, Yin Z, Deo RC, Wen X, Si J, et al. Regional hydrology heterogeneity and the response to climate and land surface changes in arid alpine basin, northwest China. Catena. 2020;187:104345. 10.1016/j.catena.2019.104345.Search in Google Scholar

[27] Chun XS, Zheng HX, Hui Q, Miao LL, Xiao CY. China land soil moisture EnKF data assimilation based on satellite remote sensing data. Sci China Earth Sci. 2011;54:1430–40. 10.1007/s11430-010-4160-3.Search in Google Scholar

[28] Guo D, Wang H, Zhang X, Liu G. Evaluation and analysis of grid precipitation fusion products in Jinsha River Basin based on China meteorological assimilation datasets for the SWAT model. Water. 2019;11:253. 10.3390/w11020253.Search in Google Scholar

[29] Liu X, Yang M, Meng X, Wen F, Sun G. Assessing the impact of reservoir parameters on runoff in the Yalong River Basin using the SWAT model. Water. 2019;11:643. 10.3390/w11040643.Search in Google Scholar

[30] Tian Y, Zhang K, Xu YP, Gao X, Wang J. Evaluation of potential evapotranspiration based on CMADS reanalysis dataset over China. Water. 2018;10:1126. 10.3390/w10091126.Search in Google Scholar

[31] Stamnes K, Tsay SC, Wiscombe W, Jayaweera K. Numerically stable algorithm for discrete-ordinate-method radiative-transfer in multiple-scattering and emitting layered media. Appl Opt. 1988;27:2502–9. 10.1364/ao.27.002502.Search in Google Scholar

[32] ChunXiang S, ZhengHui X, Hui Q, MiaoLing L, XiaoChun Y. China land soil moisture EnKF data assimilation based on satellite remote sensing data. Sci China Earth Sci. 2011;54:1430–40. 10.1007/s11430-010-4160-3.Search in Google Scholar

[33] Sloan PG, Moore ID. Modeling subsurface stormflow on steeply sloping forested watersheds. Water Resour Res. 1984;20:1815–22. 10.1029/WR020i012p01815.Search in Google Scholar

[34] Boughton WC. A review of the USDA SCS curve number method. Soil Res. 1989;27:511–23. 10.1071/sr9890511.Search in Google Scholar

[35] Priestley CHB. On the assessment of surface heat flux and evaporation using large-scale parameters. Monthly Weather Rev. 2009;100:81–92. 10.1175/1520-0493(1972)1002.3.CO;2.Search in Google Scholar

[36] Hargreaves GH, Samani ZA. Reference crop evapotranspiration from temperature. Appl Eng Agric. 1985;1:96–9. 10.13031/2013.26773.Search in Google Scholar

[37] Abbaspour KC, Yang J, Maximov I, Siber R, Bogner K, Mieleitner J, et al. Modelling hydrology and water quality in the pre-alpine/alpine Thur watershed using SWAT. J Hydrol. 2007;333:413–30. 10.1016/j.jhydrol.2006.09.014.Search in Google Scholar

[38] Moriasi DN, Arnold JG, van Liew MW, Bingner RL, Harmel RD, Veith TL. Model evaluation guidelines for systematic quantification of accuracy in watershed simulations. Trans ASAE. 2007;50:885–900. 10.13031/2013.23153.Search in Google Scholar

[39] Nash JE, Sutcliffe JV. River flow forecasting through conceptual models. Part I – A discussion of principles. J Hydrol. 1970;10:282–90. 10.1016/0022-1694(70)90255-6.Search in Google Scholar

[40] Gupta HV, Sorooshian S, Yapo PO. Status of automatic calibration for hydrologic models: comparison with multilevel expert calibration. J Hydrol Eng. 1999;4:135–43. 10.1061/(asce)1084-0699(1999)4:2(135).Search in Google Scholar

[41] Jacobs K, Lebel L, Buizer J, Addams L, Matson P, McCullough E, et al. Linking knowledge with action in the pursuit of sustainable water-resources management. Proc Natl Acad Sci U S A. 2016;113:4591–6. 10.1073/pnas.0813125107.Search in Google Scholar

[42] Qin Y, Ding YL. Change characteristics of different types of snowfall event in China’s Tianshan Mountains from 1961 to 2016. Adv Water Sci. 2019;30:457–66 (in Chinese with English abstract).Search in Google Scholar

[43] Meng X, Wang H, Wu Y, Long A, Wang J, Shi C, et al. Investigating spatiotemporal changes of the land-surface processes in Xinjiang using high-resolution CLM3.5 and CLDAS: Soil temperature. Sci Rep. 2017;7:1–14. 10.1038/s41598-017-10665-8.Search in Google Scholar

[44] Cao Y, Zhang J, Yang M, Lei X, Guo B, Yang L, et al. Application of SWAT model with CMADS data to estimate hydrological elements and parameter uncertainty based on SUFI-2 algorithm in the Lijiang River Basin, China. Water. 2018;10:742. 10.3390/w10060742.Search in Google Scholar

[45] Vu TT, Li L, Jun KS. Evaluation of multi-satellite precipitation products for streamflow simulations: a case study for the Han River Basin in the Korean Peninsula, East Asia. Water. 2018;10:642. 10.3390/w10050642.Search in Google Scholar

[46] Liu J, Shanguan D, Liu S, Ding Y. Evaluation and hydrological simulation of CMADS and CFSR reanalysis datasets in the Qinghai-Tibet Plateau. Water. 2018;10:513. 10.3390/w10040513.Search in Google Scholar

[47] Zhou Z, Gao X, Yang Z, Feng J, Meng C, Xu Z. Evaluation of hydrological application of CMADS in Jinhua River Basin, China. Water. 2019;11:138. 10.3390/w11010138.Search in Google Scholar

[48] Openshaw S. The modifiable areal unit problem. Concepts and techniques in modern geography No. 38, Vol. 38. Norwich, UK: GeoBooks; 1983.Search in Google Scholar

[49] Meng XY, Wang H, Shi CX, Wu YP, Ji XN. Establishment and evaluation of the China meteorological assimilation driving datasets for the SWAT model (CMADS). Water. 2018;10:18. 10.3390/w10111555.Search in Google Scholar

Received: 2020-04-25
Revised: 2020-08-12
Accepted: 2020-08-24
Published Online: 2020-10-03

© 2020 Xinchen Gu et al., published by De Gruyter

This work is licensed under the Creative Commons Attribution 4.0 International License.

Scroll Up Arrow