Understanding the effects of soil data quality on SWAT model performance and hydrological processes in Tamedroust watershed (Morocco)

Journal of African Earth Sciences 160 (2019) 103616

Contents lists available at ScienceDirect

Journal of African Earth Sciences journal homepage: www.elsevier.com/locate/jafrearsci

Understanding the effects of soil data quality on SWAT model performance and hydrological processes in Tamedroust watershed (Morocco)

T

Yassine Bouslihima,∗, Aicha Rochdia, Namira El Amrani Paazab, Lorena Liuzzoc a

Renewable Energy, Environment and Development Laboratory, Faculty of Science and Technology, Hassan First University, Settat, Morocco Department of Applied Geology, Faculty of Science and Technology, Hassan First University, Settat, Morocco c Università degli Studi di Enna Kore, Cittadella Universitaria, 94100 Enna, Italy b

ARTICLE INFO

ABSTRACT

Keywords: Hydrological modeling SWAT model Soil data effect Hydrological processes Tamedroust watershed (Morocco)

Hydrological modeling remains an indispensable tool for water resources research in order to understand the watershed functioning and to effectively manage water supplies. The study of the hydrologic functioning of a semi-arid watershed is a real challenge in the lack of reliable and regular data. In this work, we propose (1) to evaluate the performance of Soil and Water Assessment Tool (SWAT) on the TAMEDROUST watershed (Morocco) under extremely contrasted climatic conditions during the period 1998–2002 and (2) to test the effects of the soil data resolution on the watershed response. For this second purpose, two soil databases are compared: (i) HWSD-2L, a low-resolution soil database with three soil types obtained from the Harmonized World Soil Database produced by FAO and (ii) TAMED-SOL, a refined database with eleven soil types, performed from field measurements and laboratory analyses. Results before calibration showed a considerable variability and a significant effect of the soil characteristics on the different components of the hydrological cycle such as SW (soil water content) and GWQ groundwater discharge into reach). Flows obtained by HWSD-2L soil database were consistently lower than those simulated by TAMED-SOIL but both databases overpredicted streamflow if compared to observed data. To harmonize the data of the two databases, we proceeded to a harmonization of the soil depth by considering in HWSD-2L only the surface layer, thus creating a third database HWSD-1L which was also considered in the evaluation of the model response. After the calibration period, both soils databases (TAMED-SOIL and HWSD-2L) improved the model performance in terms of streamflow, with values of R2 and NSE between 0.64 and 0.65. Model validation was acceptable and similar for both databases with R2 and NSE values of 0.76 and 0.57. We can conclude that the quality and resolution of soil data affects all hydrological cycle components, mainly SW and water yield (WYLD), but do not lead to significant discrepancies in streamflow simulation, especially after the model calibration.

1. Introduction Hydrological modeling is based on several input data such as meteorological, topography, land use data and soil properties. SWAT (Soil and Water Assessment Tool) is a watershed scale model developed to predict the impact of land management practices on water, sediment, and agricultural chemical yields in large, complex watersheds with varying soils, land use, and management conditions (Di Luzio et al., 2002). SWAT is being used extensively in the United States and Europe to improve the assessment of climate change impacts on water supply and quality (Easton et al., 2008; Santhi et al., 2001; Spruill et al., 2000). Soil plays a key role in the hydrological cycle; it captures and stores water, making it available for absorption by crops, and thus minimizing ∗

surface evaporation and maximizing water use efficiency and productivity (Gibbon, 2012). The lack of reliable and harmonized soil data has considerably hampered land degradation assessments and water resources sustainable management and studies. To overcome this problem, international organizations have been working on the implementation of global databases on soils since the end of the 1980s (Nachtergaele et al., 2010; Oldeman and Van Engelen, 1993; Ribeiro et al., 2018). Different studies have analyzed the effects of different soil characteristics input data on hydrological processes using SWAT or other hydrological models. Levick et al. (2004) found that the runoff using State Soil Geographic (STATSGO) soils was generally higher than those simulated with Soil Survey Geographic (SSURGO). Geza and McCray

Corresponding author. E-mail addresses: [email protected], [email protected] (Y. Bouslihim).

https://doi.org/10.1016/j.jafrearsci.2019.103616 Received 15 November 2018; Received in revised form 3 July 2019; Accepted 28 August 2019 Available online 29 August 2019 1464-343X/ © 2019 Elsevier Ltd. All rights reserved.

Journal of African Earth Sciences 160 (2019) 103616

Y. Bouslihim, et al.

(2008) compared the effect of SSURGO (highest spatial resolution) and STATSGO (lowest spatial resolution) soil databases in streamflow and water quality predictions. Results showed that, before calibration, SSURGO overpredicted total streamflow compared to STATSGO, but less stream loading in terms of sediment and sediment-attached nutrients components, after calibration results suggest a better estimation by SSURGO than STATSGO, since Nash–Sutcliffe Efficiency (NSE) value was 0.70 and 0.61 for SSURGO and STATSGO, respectively. Mukundan et al. (2010) evaluated the effect of STATSGO and SSURGO soil databases of streamflow and sediment in North Fork Broad River (182 km2). Results show that the model predictions of flow and sediment by the two models were similar, and the differences were statistically insignificant. Ye et al. (2011) analyzed the impact of soil data with two different spatial resolutions, in a large humid watershed of Xinjiang River (15 535 km2), and found that soil data resolution affects soil water storage. The finer resolution produced higher monthly soil water storage simulation than lower resolution, but do not lead to significant discrepancies in streamflow simulation. In a small watershed of Turkey Creek (126 km2), Zhao (2016) tested the effect of two soil datasets, FAO (world reference base) and GSCC (the Genetic Soil Classification of China). Results show that after model calibration, both soils datasets improved the model performance in terms of discharge, but soil water content (SW) is more sensitive to soil properties. Worqlul et al. (2018) evaluated the effects of soil characteristics on SWAT runoff and water balance in paired-watersheds in the Upper Blue Nile Basin. The study indicated that the SWAT model captured the observed flow very well for both calibration and validation period. Most of these previous studies focused on comparing open source databases available online (STATSGO and SSURGO) or at different national organizations such as GSCC, that cover a limited part of the world. Due to the non-availability of these data in many regions with very serious soil data availability limitations, modelers have been forced to create their own databases. Adding that spatial information of soil physical properties is costly, requires numerous soil observations and laboratory analysis not easy to obtain (Ye et al., 2011), which makes the preparation of SWAT input database a very long and tedious task. In Morocco, input data effect on hydrological modeling quality has been little studied and most of studies focused on the application of SWAT model (Bouslihim et al., 2016; Briak et al., 2016; Brouziyne et al., 2018; Fadil et al., 2013; Fadil et al., 2011), without verifying the effect of input data on different hydrological components of the watershed. This study was conducted to understand the effect of soil data on the hydrological behavior in an understudied watershed in which most studies were aimed at shallow aquifer located in the Chaouia basin (El Mansouri, 1993; Smaoui et al., 2012). In this project, we created our soil database called “TAMED-SOIL” for which many soil samples were done, and several soil analyses were made throughout the watershed area to determine some soil characteristics needed to create SWAT model database. This detailed soil database and the Harmonized World Soil Database (HWSD) (Ribeiro et al., 2018) produced by FAO were used as inputs for SWAT model to study the effects of soil data quality on the hydrological behavior of Tamedroust watershed. We hope that this study can help users to highlight the effects of soil data quality on the hydrological modeling behavior of the watershed, and to verify if a high-resolution soil database will yield better results.

The Chaouia climate regime is described as semi-arid with an average annual rainfall of 298 mm/year (for the period between 1993 and 2009), characterized by irregular rainfall (El Assaoui, 2017). Generally, the most humid months are December, January and February. Moreover, the months of June, July and August are often completely dry (Fig. 2). The average annual temperature is about 18 °C. The flow regime of these basins is formed by long periods of low or no-existent flows, and some short-term severe floods between November and February (El Houssine et al., 2014). As shown in Fig. 3, the average annual precipitation data shows significant variability between the years in a period of 17 years (1993–2009), in which many dry periods occurred (1993–1994, 1998–2001, 2005, 2007 and 2009). These periods are interspersed with wet periods of shorter cycles. After 2009, no hydrological data were available following the construction of a dam on the Tamedroust River. 2.2. SWAT model set up SWAT is a semi-distributed, process-based river basin model, developed by Arnold et al. (1998) on behalf of the US Department for Agriculture (USDA) to predict the impact of land management practices on water, sediment, and agricultural chemical yields in large complex watersheds with a variety of soils, land use, and management conditions (Neitsch et al., 2011). In SWAT, the watershed is divided into sub-basins based on topography, and then each sub-basin is further conceptually divided into Hydrologic Response Units (HRUs), which have a unique combination of soil, land use and slope (Worqlul et al., 2018). The hydrological cycle in the model is estimated by the following equation of water balance (Neitsch et al., 2011): t

SWt = SW0 +

(Rday

Qsurf

Ea

Wseep

Qgw )

i=1

where SW is the soil water content, “t” is time in days, and Rday, Qsurf, Ea, Wseep and Qgw are respectively, daily amounts of precipitation, surface runoff, evapotranspiration, water entering the vadose zone from the soil profile, and return flow (all units are in mm). In the current study, the Soil Conservation Service Curve Number (SCS–CN) (USDA, 1972) and the Penman-Monteith method (Monteith, 1965) were used to estimate surface runoff and evapotranspiration, respectively. The Muskingum method developed by Mccarthy (1938) was used for flow routing. 2.3. Model input The use of a semi-distributed river basin model such as SWAT (Arnold et al., 2012), taking into account the different characteristics of the watershed and requires the preparation of a diversified database includes topography, land cover, soil, meteorological and hydrological data. Slope map (Fig. 4) was extracted from Digital Elevation Model (DEM) ASTER-GDEM2, developed jointly by the U.S. National Aeronautics and Space Administration (NASA) and Japan's Ministry of Economy, Trade, and Industry (METI) with a resolution of 30 m. The hydrological station Tamedroust (x = 299450 m, y = 277540 m) was used as the outlet to delineate the watershed and extract its various features. The land use map was extracted from a Landsat-8 satellite image, with a resolution of 30 m, which is identical to the spatial resolution of the DEM; this image was processed by the supervised classification method with ArcGIS software and is categorized into five types (Fig. 5). Water represents only a small part of the surface (0.01%). In general, the bare soil occupies the most prominent part with a percentage of 80%; the other types occupy the rest with 11.45%, 4.92% and 3.39% for pasture, urban and agriculture, respectively.

2. Material and methods 2.1. Description of the study area The Tamedroust watershed is one of Settat-Ben Ahmed plateau watersheds (Chaouia, Morocco) located between 35° 50′N, 7° 00′W and 33° 05′N, 7° 34′W, as shown in Fig. 1, extended over an area of 642 km2 with altitude varying between 304 m and 809 m. 2

Journal of African Earth Sciences 160 (2019) 103616

Y. Bouslihim, et al.

Fig. 1. Watershed boundary and stream network of Tamedroust River.

of Meteorology) ICARDA-SYRIE and IDRC-CANADA (El Oumri et al., 1995). All soils type, are reclassified by their texture and spatially represented, as shown in Fig. 6b using ArcGIS. For example, the Calcimagnesian soil is classified into three groups: silty in the upstream portion of the watershed, silty-clay in the middle and clayey in the downstream part of the watershed. After this classification, several fieldwork, sampling and soil analyses were carried out. Other texture results from Bagri and Rochdi (2008) were used to complete the TAMED-SOIL database. It should be noted that soil analysis encompasses a series of steps from field sampling to laboratory testing and, eventually, to interpretation. None of the steps is independent of the others; the care taken at one step affects the result obtained at another one (Jones, 2001). Therefore all sampling and sample preparation were fully respected. All soil analysis carried in the laboratory were done on 2 mm sieved soil (Gregorich and Carter, 2007). Indeed, SWAT model requires mainly, soil texture (clay, silt, sand content, % of soil weight), rock fragment content (% of total soil weight), soil organic carbon (C, % of soil weight) and electrical conductivity (dS/m) which were carried out following the standards in force in the soil analysis. The Walkley and Black method (Walkley and Black, 1934) was used to measure organic carbon content. Soil texture was determined by Robinson Pipette method (Robinson, 1933, 1922). The pipette method of particle-size analysis is a sedimentation procedure which utilizes pipette sampling at controlled depths and times (Day, 1965). Other SWAT soil parameters such as SOL_BD (moist bulk density), SOL_AWC (available water capacity) and SOL_K (saturated hydraulic conductivity) were estimated using equations (Saxton and Rawls, 2006). USLE_K (USLE equation soil erodibility K factor) was calculated using the equations described by Neitsch et al. (2011) and Singh (1995). The moist soil albedo (SOL_ALB) was extracted for each soil by using Landsat-8 satellite image and compared with some results from Hinse et al. (1989). ( The soil depth was verified in situ and supplemented by previous geological studies of the region (El Bouqdaoui, 1995).

Fig. 2. Mean monthly rainfall and temperature variation at Tamedroust watershed (1993–2009).

For soil characteristics, two different databases were used. The first is a global database HWSD developed by the Food and Agriculture Organization of the United Nations (FAO), the International Institute for Applied Systems Analysis (IIASA), and the Institute of Soil Science, Chinese Academy of Science. This database contains 16000 map units with two different soil layers (0–300 mm and 300–1000 mm deep) (Nachtergaele et al., 2012). Most of the soil data requested by the SWAT model were directly obtained from this database (sequence number, drainage class, organic carbon, and soil texture data), using other sources to complete it. Table 1 lists all parameters with the sources and/ or methods used to calculate them; we call this data HWSD-2L, where 2L means two layers (Fig. 6a). The second database TAMED-SOIL is the result of an extensive research conducted by our research group over a period of eighteen months (from April/2016 to September/2017), based on the schematic soil map with a scale of 1:50,0000 realized by INRA-Morocco (National Institute for Agronomic Research), DMN-Morocco (National Direction 3

Journal of African Earth Sciences 160 (2019) 103616

Y. Bouslihim, et al.

Fig. 3. Average annual precipitation variation for the period between 1993 and 2009.

To calibrate the watershed response, hydrological and meteorological data including daily rainfall and streamflow, were collected from the Hydraulic Basin Agency of Bouregreg and Chaouia (ABHBC) corresponding to Tamedroust station (the only one located at the outlet of the watershed). While all the other climatic data (temperature, wind, relative humidity, and solar radiation) were obtained from the NCEP Climate Forecast System Reanalysis (CFSR) (Saha et al., 2010).

percentage of land use and slope. Comparing the two soil databases used in this project (Fig. 6), the resolution of the soil map affects the number of Hydrological Response Units (HRUs). The high number of soil types in TAMED-SOIL increases the combinations between soil, land use and slope. The number of HRUS created by TAMED-SOIL is very high compared to HWSD-2L soil data (421 against 164).

2.4. Defining SWAT hydrologic response units

2.5. Model sensitivity analysis, calibration and validation

Hydrologic response units (HRUs) were created using land use, slope and soil. Several choices are available for creating these HRUs through SWAT model, focusing either on dominant HRUs or on given minimum limits for each input. In our case, we tried to show the effect of soil on hydrological modeling, for that 0% was chosen as a minimum percentage of soil class over the watershed area to integrate the maximum information related to the soil, and 5% as the minimum

The model was calibrated and evaluated on a monthly time step and the simulation was divided into 3 periods: 3 years (January/1995 to December/1997) for model initialization (warm-up), 3 years (from January/1998 to December/2000) for calibration and 2 years (from January/2001 to December/2002) for validation. Sensitivity analysis aims to identify the key parameters that affect model performance and it plays important roles in model

Fig. 4. Slope map of Tamedroust watershed. 4

Journal of African Earth Sciences 160 (2019) 103616

Y. Bouslihim, et al.

Fig. 5. Land use of Tamedroust watershed. Table 1 SWAT (Soil and Water Assessment Tool) input parameters for each soil type. Parameter

Description

Unit

Source HWSD-2L

SOL-Z SOL_BD SOL_AWC SOL_K SOL_CBN SOL_CLAY SOL_SILT SOL_SAND ROCK SAL_ALB USLE_K SOL_EC

Depth from soil surface to bottom of layer Moist bulk density Available water capacity of the soil layer Saturated hydraulic conductivity Organic carbon content Clay content Silt content Sand content Rock fragment content Moist soil albedo USLE equation soil erodibility (K) factor Electrical conductivity

mm g·cm−3 mm H2O·mm−1 sol mm·Hr−1 % of soil weight % of soil weight % of soil weight % of soil weight % of total weight _ 0.013 metric ton m2 hr/(m3 -metric ton cm) dS·m−1

TAMED-SOIL

HWSD database Soil analysis equations (Saxton and Rawls, 2006) equations (Saxton and Rawls, 2006) equations (Saxton and Rawls, 2006) HWSD database Soil analysis HWSD database Soil analysis HWSD database Soil analysis HWSD database Soil analysis HWSD database Soil analysis Landsat-8 image (Neitsch et al., 2011; Williams, 1995) equations HWSD database Soil analysis

Table 2 Physical properties of different soil type. Soil database

Soil

Percentage

SOL-ZMX (mm)

Texture

HYDGRP

SOL-AWC (mm H2O·mm−1 sol)

HWSD-2L

CL-Calcisols (dominant soil)

73.42

1000

L

C

PH-Phaezems (dominant soil)

26.03

1000

L

C

VR-Vertisols (dominant soil)

0.55

1000

C

D

Calcisols -Clay Calcisols -Clay Loam Calcisols -Loam Rankers -Sandy Loam Rankers -Clay Loam Xerosols -Sandy Clay Xerosols -Sandy Clay Loam (am) Xerosols -Clay Loam Xerosols -Sandy Clay Loam (av) Vertisols -Clay Calcisols x Rankers -Loam

11.40 17.36 20.63 4.32 9.85 9.39 13.55 3.99 1.64 2.35 5.52

300 300 300 450 450 500 500 500 500 1000 300

C CL L SL CL SC SCL CL SCL C L

D C C B C C C C C D B

6.93a 5.6b 5.07a 2.87b 0.6a 0.53b 0.12 0.10 0.12 0.1 0.11 0.1 0.07 0.12 0.1 0.11 0.1

TAMED-SOIL

Where SOL-ZMX is the maximum rooting depth of soil profile (mm), HYDGRP is the soil hydrologic group (A, B, C or D), SOL-AWC is the available water capacity of the soil layer (mm H2O/mm). a First layer. b Second layer. 5

Journal of African Earth Sciences 160 (2019) 103616

Y. Bouslihim, et al.

Fig. 6. Tamedroust watershed soil maps on (a) HWSD-2L database (b) TAMED-SOIL database.

parameterization (Song et al., 2015). It is thus possible to reduce the number of parameters to be included in the calibration, reducing the efforts required in calibration and increasing probability of converging towards a powerful combination (Arnold et al., 2012). The p-value and t-stat were used to evaluate the significance of the relative sensitivity. Larger absolute t-stat means greater sensitivity and, p-value close to zero represents higher significance (Abbaspour et al., 2007). The coefficient of determination R2 (Krause et al., 2005) and Nash–Sutcliffe Efficiency NSE (Nash and Sutcliffe, 1970) were used to evaluate the accuracy of calibration and validation. The R2 values vary from zero to one, a closer value to one represents a perfect correlation, while zero indicates no correlation. NSE values can range between negative infinity and one (Moriasi et al., 2007). The closer the NSE value to one the better is the estimation of the streamflow by the model (Geza and McCray, 2008).

3. Results Simulation results were discussed in two phases, before and after calibration. A comparison before calibration is highly recommended because it allowed us to evaluate the direct effect of inputs on hydrological behavior because the calibration hides the differences between soils databases used. Also, the uncalibrated model results can show how well each dataset predicts streamflow before calibration, which would indicate the effort required for calibration when using each data set (Geza and McCray, 2008). 3.1. Modeling results before calibration 3.1.1. Streamflow Fig. 7 shows the simulated monthly streamflow using both databases before calibration. Generally, the simulated streamflow using the different databases were higher than the observed values during most of the time. Moreover, flows obtained using the HWSD-2L soil database were consistently lower than those simulated by TAMED-SOIL 6

Journal of African Earth Sciences 160 (2019) 103616

Y. Bouslihim, et al.

Fig. 7. Comparison before calibration of observed and simulated monthly streamflow using HWSD-2L and TAMED-SOIL databases.

database. We should not forget that we used the same input data (slope, land-use and meteorological data), so the only difference in output data is the soil characteristics.

yield (WYLD) such as SURFQ and GWQ using two different soil databases shows the significant contribution of GWQ in the results of water yield at TAMED-SOIL database. When using the HWSD-2L water gets stuck in the soil layer or evaporates according to the high values of SW and ET. On the other hand when using the TAMED-SOIL water is approximately equally distributed between the different components of the water cycle. The ET estimated using the TAMED-SOIL database was generally lower than simulated using HWSD-2L database, but water percolating out of the root zone was higher when TAMED-SOIL was used. Soil water content (SW) shows considerable variability; the obtained values using the HWSD-2L database were higher than those obtained using TAMEDSOIL (581.9 mm versus 112.59 mm). To explain this significant difference in results and especially for SW, we proceeded to homogenize HWSD-2L soils and limit SOL-ZMX (the maximum rooting depth of soil profile (mm)) in 30 cm, keeping just the topsoil part, to compare it with TAMED-SOIL, assuming that these databases have close values of depth. We call this new database HWSD-1L where 1L means one layer. The

3.1.2. Hydrological components Water yield (WYLD) was taken as the sum of surface runoff, lateral flow in the soil profile and groundwater return flow (Heatwole, 1995). Fig. 8 summarizes the annual average values obtained for each parameter during 5 years (1998–2002), these results show and confirm the big difference between water yield values, 114.28 and 28.79 mm for TAMED-SOIL and HWSD-2L, respectively. Water yield results give us just a general idea about the final reaction of both models. Although, for a detailed analysis of these results, it is necessary to analyze the other components of the hydrological cycle like surface runoff (SURQ), groundwater discharge into reach (GWQ), actual evapotranspiration (ET), soil water content (SW) and the amount of water percolation out of root zone (PERC). Comparing the hydrological parameters contributing in the water

Fig. 8. Comparison of hydrological components simulated by using the two different soil databases TAMED-SOIL and HWSD-2L (before calibration). 7

Journal of African Earth Sciences 160 (2019) 103616

Y. Bouslihim, et al.

Table 4 R2 and NSE values in calibration period 1998 to 2000 and validation period 2001 to 2002 for the three soil databases. Calibration

TAMED-SOIL HWSD-1L HWSD-2L

Validation

R2

NSE

R2

NSE

0.65 0.65 0.65

0.64 0.64 0.64

0.76 0.74 0.76

0.57 0.54 0.57

Likewise, according to Moriasi et al. (2007), the statistical comparison shows satisfactory results of calibration for all three databases, NSE and R2 values were 0.64 and 0.65, respectively (Fig. 10). The Validation involves running the model using the best parameters values obtained during the calibration process and comparing the predictions to observed data for another period not used in the calibration. In our case, the chosen period is 2 years (2001–2002), an extended period is recommended for the calibration and validation periods, but unfortunately, data quality was not sufficient to select other periods. Despite the short period, model validation for daily streamflow simulation showed a performance of NSE and R2 values greater than 0.74 and 0.54, receptively for all three databases presented in Fig. 10 and Table 4.

Fig. 9. Comparison before calibration between simulated monthly streamflow using the three soil databases.

comparison between the three databases TAMED-SOIL, HWSD-2L and HWSD-1L showed that the results of simulated streamflow using HWSD1L database have a great similarity with those simulated using TAMEDSOIL database (Fig. 9) and strongly correlated with a R2 equal to 98%, HWSD-2L produced lower values than HWSD-1L and TAMED-SOIL databases. 3.2. Modeling results after calibration

3.2.2. Hydrological components Simulation results of some hydrological components (ET, SW and WYLD) are shown in Table 5, using the three soil databases during 1998–2002 period. The ET and SW simulated using HWSD-2L database were higher than those simulated by using TAMED-SOIL and HWSD-1L databases. Furthermore, WYLD estimated with TAMED-SOIL databases was lower than those simulated by using HWSD-2L and HWSD-1L databases. WYLD values obtained are 16.43 mm, 15.51 mm and 14.60 mm for HWSD-1L, HWSD-2L and TAMED-SOIL respectively. Fig. 11 shows the spatial distribution of annual average soil water content at the sub-basin level using a single scale for the three databases; the values of each sub-basin are also presented next to the maps. Simulated values by HWSD-1L (Fig. 11b) were very close with a standard deviation (STDEV) equal to 1.2. The presence of two dominant soil types with the same texture (loamy) and a homogeneous depth for all sequences (30 cm), which makes the watershed a homogeneous unit in terms of soil, can explain these results. Differently, the simulated values using HWSD-2L (Fig. 11c) are very high as compared to the two other databases with a STDEV equal to 14.5, and as already explained this is related to the depth that reaches 1 m in most sequences and just 30 cm in others including the varied values of SOL-AWC from each sequence to another. The third TAMED-SOIL (Fig. 11a) database as previously mentioned contains 11 soil types with widely varying texture classes,

Sequential Uncertainty Fitting (SUFI-2) technique with SWATCalibration and Uncertainty Programs (SWAT-CUP) (Abbaspour, 2013) was used for automatic model calibration and sensitivity analysis. In order to select the most sensitive parameters, twenty parameters related to soil, tile drainage, surface water, and groundwater were tested. Based on the results of the sensitivity analysis p-value and t-stat were used to eliminate no sensitive parameters from the calibration process (Arnold et al., 2012). Eight parameters were found to be the most sensitive: runoff curve number (CN2), Manning's “n” value for overland flow (OV_N), average slope length (SLSUBBSN), depth of the sub-surface drain available (DDRAIN_BSN), water capacity of the soil layer (SOL_AWC), moist bulk density (SOL_BD), base flow alpha-factor (ALPHA_BF) and depth to impervious layer in soil profile (DEP_IMP). All parameters are listed in Table 3 with their optimal values. 3.2.1. Streamflow Streamflow calibration was performed at Tamedroust watershed outlet, and the model was carefully calibrated over a three years. Simulations results with different soil databases were compared with the observed values using graphical and statistical methods (R2 and NSE). All R2 and NSE values after calibration are shown in Table 4. Table 3 Sensitive parameters and their fitted values for the three databases using SUFI-2. Parameters

CN2 SLSUBBSN OV_N SOL_AWC ALPHA_BF DEP_IMP DDRAIN_BSN SOL_BD

Method_(Initial range)

r_(-0.3, 0.3) v_(10, 150) v_(0.01, 30) r_(-0.5, 0.5) v_(0, 1) v_(0, 6000) v_(0, 2000) r_(-0.5, 0.5)

TAMED-SOIL

HWSD-2L

HWSD-1L

rank

fitted value

rank

fitted value

rank

fitted value

4 6 3 5 8 2 1 7

−0.29 47.6 8.72 0.42 0.85 53.25 1074 0.48

8 7 4 5 6 1 2 3

−0.29 71.14 3.95 −0.31 0.84 904.6 1424.5 0.48

3 8 5 4 7 2 1 6

−0.28 51 7.23 0.38 0.54 460 1044 0.4

Where CN2: curve number condition II. SLSUBBSN: average slope length. OV_N: Manning's “n” value for overland flow. SOL_AWC: available water capacity of the soil layer. ALPHA_BF: baseflow alpla factor. DEP_IMP: Depth to impervious layer in soil profile. DDRAIN_BSN: depth of the sub-surface drain. SOL_BD: moist bulk density. r: means the existing parameter value is multiplied by 1 + a given value. v: means the existing parameter value is to be replaced by a given value. 8

Journal of African Earth Sciences 160 (2019) 103616

Y. Bouslihim, et al.

Fig. 10. Scatter plot of simulated vs. observed streamflow for the calibration (1998–2000) and the validation (2001–2002) periods for the three databases.

depths, and SOL_AWC values, which gave us a proper distribution of SW values in different sub-basins with a STDEV equal to 20.46. In the upper part of the watershed the soils have a sandy clay loam texture, sandy loam or loam which explains the low values of SW (subbasins 7, 21, 22 and 23) as shown in Fig. 11a, in the downstream part, the soils are clayey with a good quantity of organic matter. SW values are higher than the upstream part, and as we know, SW is controlled mainly by soil texture and organic matter. Soil with a high proportion of silt and clay particles hold more water (subbasins 2 and 6).

4. Discussion The statistical comparison shows satisfactory calibration and validation results for all the databases involved in the project and that the resolution of soil data did not contribute significantly to improve the results achieved after the hydrological model calibration. As confirmed by studies conducted by Mukundan et al. (2010) and Di Luzio et al. (2005), which suggested that a less detailed soil data could be used to save time and effort needed to create a detailed soil dataset, especially in a larger watershed. The results of these studies indicated that after calibration, the variations in model streamflow predictions were statistically insignificant. This is understandable since the calibration can

Table 5 Comparison after calibration of some hydrological components simulated by using three soil databases from 1998 to 2002. Month

1 2 3 4 5 6 7 8 9 10 11 12 Average Sum

TAMED-SOIL database (mm)

HWSD-1L soil database (mm)

HWSD-2L soil database (mm)

ET

SW

WYLD

ET

SW

WYLD

ET

SW

WYLD

19.47 15.70 23.98 26.93 13.74 1.53 0.00 0.00 1.34 10.11 11.88 16.80 11.79 141.46

34.29 23.52 19.37 9.84 0.79 0.00 0.00 0.00 1.18 14.81 22.50 38.00 13.69 164.30

1.24 0.76 0.66 0.58 0.20 0.12 0.10 0.09 0.12 0.62 5.85 4.28 1.22 14.60

22.23 18.65 27.93 31.49 17.91 3.08 0.00 0.00 1.52 10.24 12.14 17.85 13.59 163.04

43.72 31.59 25.81 14.42 2.24 0.00 0.00 0.00 1.10 15.30 26.43 47.67 17.36 208.27

1.54 1.08 0.88 0.74 0.32 0.18 0.12 0.08 0.10 0.58 6.20 4.60 1.37 16.43

19.83 16.19 25.82 34.99 39.32 10.45 2.29 1.21 1.99 12.30 12.73 18.12 16.27 195.26

77.49 67.95 63.72 47.88 14.28 4.67 2.38 1.17 1.80 13.96 32.56 71.98 33.32 399.85

1.42 1.09 0.84 0.73 0.33 0.21 0.16 0.12 0.13 0.60 5.47 4.40 1.29 15.51

Where ET is the actual evapotranspiration, SW is the soil water content and WYLD is the water yield. 9

Journal of African Earth Sciences 160 (2019) 103616

Y. Bouslihim, et al.

Fig. 11. The spatial distribution of Soil Water content by using a) TAMED-SOIL, b) HWSD-1L and c) HWSD-2L databases during the time period of 1998–2002.

masks the direct effect of inputs on the model results, which requires analysis and evaluation of different results before the calibration, as recommended by Geza and McCray (2008). The main difficulty encountered is that the model has underestimated the streamflow in several days during floods and systematically after the main peak for all databases, this demonstrates the major limitation for the runoff modeling in arid and semi-arid regions (Beven, 2001; Pilgrim et al., 1988; Wheater et al., 2007). In such meteorological conditions, with limited data availability and poor spatial distribution of measurement stations, this was an additional challenge, that was overcome with good reliability (Näschen et al., 2018; Tuo et al., 2016). Several studies have shown that soil characteristics such as soil depth, hydrological group, and hydraulic conductivity can influence hydrological cycle components in the watershed (Geroy et al., 2011; Mohanty and Mousli, 2000; Wang and Melesse, 2006). However, in our case, it has been noted that only the soil water content (SW) has the most significant variation between all water cycle components. These results are consistent with other studies such as Ye et al. (2011) and Zhao (2016). Specifically, in our case, HWSD-2L soil database contains two soil

layers for each type soil (0–300 mm and 300–1000 mm for topsoil and subsoil, respectively), with a maximum rooting depth of soil profile (SOL_ZMX) being equal to 1 m. On the other hand, the SOL_ZMX values of the dominant soils of the TAMED-SOIL database are between 300 and 450 mm. This significant difference in soil depth may be the most plausible explanation for higher SW values observed in the HWSD-L2 soil database, which affect all the other components of the hydrological cycle. In addition, the clear difference between SOL_AWC values in the two databases (Table 2) can be considered as one of the main factors responsible. An increase in SOL_ AWC allows the soil to retain more water and then decreases the streamflow (Opere and Okello, 2011). 5. Conclusion This article presents the performance of the SWAT model using two different soil databases to evaluate the effect of soils data quality on the hydrological behavior and water balance analysis in a semi-arid watershed. The comparison between the two soil databases was made before and after calibration. Results indicated that the quality and the resolution of the soil map affect the number of HRUs because the high number of soil types increases the combinations between different soil, 10

Journal of African Earth Sciences 160 (2019) 103616

Y. Bouslihim, et al.

land use and slope. The soil depth affects the various components of the hydrological cycle such as SW and GWQ and has a direct effect on WYLD. The fine soil resolution produces higher WYLD values than the lower resolution (114.28 and 28.79 mm for TAMED-SOIL and HWSD2L, respectively). After calibration, the statistical comparison shows satisfactory results of calibration and validation for all soil databases (TAMED-SOIL, HWSD-1L and HWSD-2L). A significant variation has been observed in the other parameters such as SW, WYLD and ET, as mentioned in results section. We can, therefore, conclude that the model can give good results after streamflow calibration, but it does not mean that the other components are well simulated. The use of a detailed soil map or the modification of some parameters, depth, for example, can influence all the results. Consequently, before each project, researchers need to select the appropriate resolution of each input data taking into account the results and the expected objectives.

Inf. Syst. 03, 279–289. https://doi.org/10.4236/jgis.2011.34024. Fadil, A., Rhinane, H., Kaoukaya, A., Kharchf, Y., 2013. Comparaison de deux modèles hydrologiques sur une zone pilote du bassin versant de Bouregreg. Travaux de l’Institut Scientifique, Rabat, série Géologie & Géographie physique 23, 33–38. Geroy, I.J., Gribb, M.M., Marshall, H.-P., Chandler, D.G., Benner, S.G., McNamara, J.P., 2011. Aspect influences on soil water retention and storage. Hydrol. Process. 25, 3836–3842. Geza, M., McCray, J.E., 2008. Effects of soil data resolution on SWAT model stream flow and water quality predictions. J. Environ. Manag. 88, 393–406. Gibbon, D., 2012. Save and Grow: A Policymaker's Guide to the Sustainable Intensification of Smallholder Crop Production. Rome, Italy: Food and Agriculture Organization of the United Nations (2011), pp. 112, US 5.00. ISBN 978-92-5-1068717. Experimental Agriculture, 48(1), 154-154.Chicago. Gregorich, E.G., Carter, M.R., 2007. Soil Sampling and Methods of Analysis. CRC press. Heatwole, C.D., 1995. Water Quality Modeling: Proceedings of the International Symposium, April 2-5, 1995. American Society of Agricultural Engineers, Hyatt Hotel Orlando, Florida. Hinse, M., Gwyn, Q.H., Bonn, F., Merzouk, A., Badraoui, M., 1989. Réflectance spectrale des sols de Settat (Chaouia, Maroc). Jones Jr., J.B., 2001. Laboratory Guide for Conducting Soil Tests and Plant Analysis. CRC press. Krause, P., Boyle, D.P., Bäse, F., 2005. Comparison of different efficiency criteria for hydrological model assessment. Adv. Geosci. 5, 89–97. Levick, L.R., Semmens, D.J., Guertin, D.P., Burns, I.S., Scott, S.N., Unkrich, C.L., Goodrich, D.C., 2004. Adding global soils data to the automated geospatial watershed assessment tool (AGWA). Proc. 2nd SAHRA,(Sustainability of Semi-Arid Hydrol. and Riparian Areas), U. of Ariz. Internat’l. Sym. on Transboundary Water Manage Nov 16–19. Mccarthy, G., 1938. The unit hydrograph and flood routing. In: Proceedings of Conference of North Atlantic Division. US Army Corps of Engineers, pp. 608–609 1938. Mohanty, B.P., Mousli, Z., 2000. Saturated hydraulic conductivity and soil water retention properties across a soil-slope transition. Water Resour. Res. 36, 3311–3324. Monteith, J.L., 1965. Evaporation and environment. In: Symposia of the Society for Experimental Biology. Cambridge University Press (CUP), Cambridge, pp. 205–234. 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. Trans. ASABE (Am. Soc. Agric. Biol. Eng.) 50, 885–900. Mukundan, R., Radcliffe, D.E., Risse, L.M., 2010. Spatial resolution of soil data and channel erosion effects on SWAT model predictions of flow and sediment. J. Soil Water Conserv. 65, 92–104. Nachtergaele, F., van Velthuizen, H., Verelst, L., Batjes, N.H., Dijkshoorn, K., van Engelen, V.W.P., Fischer, G., Jones, A., Montanarela, L., 2010. The harmonized world soil database. In: Proceedings of the 19th World Congress of Soil Science, Soil Solutions for a Changing World, Brisbane, Australia, 1-6 August 2010, pp. 34–37. Nachtergaele, F., van Velthuizen, H., Verelst, L., Wiberg, D., 2012. Harmonized World Soil Database, Version 1.2, FAO, IIASA, ISRIC, ISSCAS. JRC. Näschen, K., Diekkrüger, B., Leemhuis, C., Steinbach, S., Seregina, L., Thonfeld, F., van der Linden, R., 2018. Hydrological modeling in data-scarce catchments: the kilombero floodplain in Tanzania. Water 10, 599. Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through conceptual models part I — a discussion of principles. J. Hydrol. 10, 282–290. https://doi.org/10.1016/00221694(70)90255-6. Neitsch, S.L., Arnold, J.G., Kiniry, J.R., Williams, J.R., 2011. Soil and Water Assessment Tool Theoretical Documentation Version 2009. Texas Water Resources Institute. Oldeman, L.R., Van Engelen, V.W.P., 1993. A world soils and terrain digital database (SOTER)—an improved assessment of land resources. Geoderma 60, 309–325. Opere, A.O., Okello, B.N., 2011. Hydrologic analysis for river Nyando using SWAT. Hydrol. Earth Syst. Sci. Discuss. 8, 1765–1797. Pilgrim, D.H., Chapman, T.G., Doran, D.G., 1988. Problems of rainfall-runoff modelling in arid and semiarid regions. Hydrol. Sci. J. 33, 379–400. Ribeiro, E., Batjes, N.H., van Oostrum, A.J.M., 2018. World Soil Information Service (WoSIS)-Towards the standardization and harmonization of world soil data. Procedures manual. Robinson, G.W., 1933. The dispersion of soils in mechanical analysis. Imp. Bur. Soil Sci. Tech. Commun 26. Robinson, G.W., 1922. A new method for the mechanical analysis of soils and other dispersions. J. Agric. Sci. 12, 306–321. Saha, S., Moorthi, S., Pan, H.-L., Wu, X., Wang, Jiande, Nadiga, S., Tripp, P., Kistler, R., Woollen, J., Behringer, D., Liu, H., Stokes, D., Grumbine, R., Gayno, G., Wang, Jun, Hou, Y.-T., Chuang, H., Juang, H.-M.H., Sela, J., Iredell, M., Treadon, R., Kleist, D., Van Delst, P., Keyser, D., Derber, J., Ek, M., Meng, J., Wei, H., Yang, R., Lord, S., van den Dool, H., Kumar, A., Wang, W., Long, C., Chelliah, M., Xue, Y., Huang, B., Schemm, J.-K., Ebisuzaki, W., Lin, R., Xie, P., Chen, M., Zhou, S., Higgins, W., Zou, C.Z., Liu, Q., Chen, Y., Han, Y., Cucurull, L., Reynolds, R.W., Rutledge, G., Goldberg, M., 2010. The NCEP climate Forecast System Reanalysis. Bull. Am. Meteorol. Soc. 91, 1015–1058. https://doi.org/10.1175/2010BAMS3001.1. Santhi, C., Arnold, J.G., Williams, J.R., Dugas, W.A., Srinivasan, R., Hauck, L.M., 2001. Validation of the swat model on a large rwer basin with point and nonpoint sources 1. JAWRA Journal of the American Water Resources Association 37, 1169–1188. Saxton, K.E., Rawls, W.J., 2006. Soil water characteristic estimates by texture and organic matter for hydrologic solutions. Soil Sci. Soc. Am. J. 70, 1569–1578. Singh, V.P., 1995. Computer models of watershed hydrology. Rev. Smaoui, H., Zouhri, L., Ouahsine, A., Carlier, E., 2012. Modelling of groundwater flow in heterogeneous porous media by finite element method. Hydrol. Process. 26, 558–569. Song, X., Zhang, J., Zhan, C., Xuan, Y., Ye, M., Xu, C., 2015. Global sensitivity analysis in

Appendix A. Supplementary data Supplementary data to this article can be found online at https:// doi.org/10.1016/j.jafrearsci.2019.103616. References Abbaspour, K.C., 2013. SWAT-CUP 2012. SWAT Calibration and Uncertainty Program—A User Manual. Abbaspour, K.C., Yang, J., Reichert, P., Vejdani, M., Haghighat, S., Srinivasan, R., 2007. SWAT-CUP: SWAT Calibration and Uncertainty Programs. A User Manual. EAWAG. Zurich, Switzerland. Arnold, J.G., Moriasi, D.N., Gassman, P.W., Abbaspour, K.C., White, M.J., Srinivasan, R., Santhi, C., Harmel, R.D., Van Griensven, A., Van Liew, M.W., others, 2012. SWAT: model use, calibration, and validation. Transactions of the ASABE 55, 1491–1508. Arnold, J.G., Srinivasan, R., Muttiah, R.S., Williams, J.R., 1998. Large area hydrologic modeling and assessment part I: model development. JAWRA Journal of the American Water Resources Association 34, 73–89. Bagri, M., Rochdi, A., 2008. Soil aggregate stability under two rainfall modes. In: Presented at the EUROSOIL, Vienna, Austria. Beven, K.J., 2001. Rainfall-runoff Modelling. Wiley Online Library. Bouslihim, Y., Kacimi, I., Brirhet, H., Khatati, M., Rochdi, A., Pazza, N.E.A., Miftah, A., Yaslo, Z., 2016. Hydrologic modeling using SWAT and GIS, application to subwatershed Bab-Merzouka (Sebou, Morocco). J. Geogr. Inf. Syst. 8, 20. Briak, H., Moussadek, R., Aboumaria, K., Mrabet, R., 2016. Assessing sediment yield in Kalaya gauged watershed (Northern Morocco) using GIS and SWAT model. International Soil and Water Conservation Research 4, 177–185. https://doi.org/10. 1016/j.iswcr.2016.08.002. Brouziyne, Y., Abouabdillah, A., Hirich, A., Bouabid, R., Zaaboul, R., Benaabidate, L., 2018. Modeling sustainable adaptation strategies toward a climate-smart agriculture in a Mediterranean watershed under projected climate change scenarios. Agric. Syst. 162, 154–163. Day, P.R., 1965. Particle fractionation and particle-size analysis. Methods of Soil Analysis. Part 1. Physical and Mineralogical Properties, Including Statistics of Measurement and Sampling Agronomymonograpp. 545–567. https://doi.org/10.2134/ agronmonogr9.1.c43. Di Luzio, M., Arnold, J.G., Srinivasan, R., 2005. Effect of GIS data quality on small watershed stream flow and sediment simulations. Hydrol. Process.: Int. J. 19, 629–650. Di Luzio, M., Srinivasan, R., Arnold, J.G., Neitsch, S.L., 2002. ArcView Interface for SWAT2000: User's Guide. TWRI Report TR-193, Texas Water Resources Institute, Texas. Easton, Z.M., Fuka, D.R., Walter, M.T., Cowan, D.M., Schneiderman, E.M., Steenhuis, T.S., 2008. Re-conceptualizing the soil and water assessment tool (SWAT) model to predict runoff from variable source areas. J. Hydrol. 348, 279–291. El Assaoui, 2017. Contribution à l’étude méthodologique de l’impact des changements climatiques sur les ressources en eaux souterraines: cas de la nappe de Berrechid. Univ. Hassan II, Maroc. El Bouqdaoui, K., 1995. Etude géologique et géotechnique de la ville de Settat. Cartographie géotechnique et substances utiles de la province de Settat. Doctorat de 3eme cycle. El Houssine, E.G., Bouabid, E.M., Mohamed, T., 2014. Surface flows in the plate of settatben ahmed and the plain of berrechid: endoreic hydrography. International Journal of Innovation and Scientific Research 9, 40–53. El Mansouri, B., 1993. Structure et modélisation quantitative de l’aquifère de Berrechid. Validation par l’approche géostatistique. Doctorat de l’université en géosciences. Hydrogéologie quantitative, option. El Oumri, El Mourid, Ambri, El Ouali, Gobel, Jmyi, 1995. Carte des sols de la zone test clé Projet “Caractérisation agro-écologique”’ Atelier sur la caractérisation agroécologique. Fadil, A., Rhinane, H., Kaoukaya, A., Kharchaf, Y., Bachir, O.A., 2011. Hydrologic modeling of the Bouregreg watershed (Morocco) using GIS and SWAT model. J. Geogr.

11

Journal of African Earth Sciences 160 (2019) 103616

Y. Bouslihim, et al. hydrological modeling: review of concepts, methods, theoretical framework, and applications. J. Hydrol. 523, 739–757. Spruill, C.A., Workman, S.R., Taraba, J.L., 2000. Simulation of daily and monthly stream discharge from small watersheds using the SWAT model. Transactions of the ASAE 43, 1431. Tuo, Y., Duan, Z., Disse, M., Chiogna, G., 2016. Evaluation of precipitation input for SWAT modeling in Alpine catchment: a case study in the Adige river basin (Italy). Sci. Total Environ. 573, 66–82. USDA, S., 1972. National Engineering Handbook, Hydrology, Section 4. United States Department of Agriculture, Soil Conservation Service Chapters 4–10. Walkley, A., Black, I.A., 1934. An examination of the Degtjareff method for determining soil organic matter, and a proposed modification of the chromic acid titration method. Soil Sci. 37, 29–38. Wang, X., Melesse, A.M., 2006. Effects of statsgo and ssurgo as inputs on swat model's

snowmelt simulation 1. JAWRA Journal of the American Water Resources Association 42, 1217–1236. Wheater, H., Sorooshian, S., Sharma, K.D., 2007. Hydrological Modelling in Arid and Semi-arid Areas. Cambridge University Press. Williams, J.R., 1995. The EPIC Model. In: Singh, V.P. (Ed.), Computer Models of Watershed Hydrology, Chapter 25, Water Resources, Publications, Highlands Ranch. Worqlul, A.W., Ayana, E.K., Yen, H., Jeong, J., MacAlister, C., Taylor, R., Gerik, T.J., Steenhuis, T.S., 2018. Evaluating hydrologic responses to soil characteristics using SWAT model in a paired-watersheds in the Upper Blue Nile Basin. Catena 163, 332–341. https://doi.org/10.1016/j.catena.2017.12.040. Ye, X., Zhang, Q., Viney, N.R., 2011. The effect of soil data resolution on hydrological processes modelling in a large humid watershed. Hydrol. Process. 25, 130–140. Zhao, A., 2016. Effect of different soil data on hydrological process modeling in Weihe River basin of Northwest China. Arabian Journal of Geosciences 9, 664.

12