Abstract
Land use/cover change is one of the factors responsible for changing the water balance of the watershed by altering the magnitude of surface runoff, interflow, base flow, and evapotranspiration. This study was aimed at evaluating the impacts of land use/cover change on the water balance of Bilate watershed between 1989, 2002, and 2015. The water balance simulation model (WaSiM) was used to access the impacts of land use/cover change on water balance. The model was calibrated (1989–2003) and validated (2007–2015) using the streamflow of at Bilate Tena gauging station. The result of land-use dynamics showed land use/cover change has a significant impact on the water balance of the watershed: on runoff production, base flow, interflow, evapotranspiration, and total simulation flow. In the study watershed, the change in total simulated flow increased by 77.83%; surface runoff, interflow, and base flow increased by 80.23%, 75.69%, and 87.79% respectively; and evapotranspiration decreased by 6% throughout the study period (1989–2015). The results obtained from this study revealed that the watershed is under land/cover change that shows its impacts on hydrological processes and water balance. Thus, effective information regarding the environmental response of land use/cover change is important to hydrologists, land-use planners, watershed management, and decision-makers for sustainable water resource projects and ecosystem services. Therefore, the policy-makers, planners, and stakeholders should design strategies to ensure the sustainability of the watershed hydrology for the sake of protecting agricultural activities, and urban planning and management systems within the watershed area.
HIGHLIGHTS
The research aimed to understand the LULC change impact on water balance.
The impacts on hydrological cycle processes were determined.
It is possible to use the WaSiM model for any watersheds.
The model can be used for watershed managers.
The model can be used as input for data construction of hydraulic structures.
Graphical Abstract
INTRODUCTION
Long-term spatial and temporal variation of water balance components such as surface runoff, soil moisture, evapotranspiration (ET), groundwater, and streamflow can be influenced by many factors within a watershed, including land use and climate change (Deng et al. 2015). Land use and land cover are highly dynamic especially in developing countries that have agricultural-based economy and rapidly increasing populations.
Land cover is defined as the topography and biophysical characteristics of the earth's surface such as vegetation, water, organisms, soil, and structures created by human activities (Lambin et al. 2003). Land cover change has a significant influence on the quantity or quality of streamflow (Lambin et al. 2003). The need to manage our physical environment sustainably is caused by a growing population and the enhanced capabilities of humans to utilize the earth's resources (Farmar-Bowers et al. 2006).
The important part of sustainable resource use is to manage the land cover where it is, has been, or is likely to become under large stress. The human activities in utilizing and managing these land resources mainly affect the biophysical characteristics, whereas land-use change is any physical, biological, or chemical change in the conditions or the resources due to management to satisfy human interests (Farmar-Bowers et al. 2006).
Land-use change is one of the most visible changes in the landscapes of the world. Along with climate change, land-use change has a strong impact on the water budget and hydrology of river catchments (Defries & Eshleman 2004). One of the main challenges in recent hydrological research is assessing the effect of diverse environmental changes. Climate and land use/cover are the main factors affecting the hydrological behavior of catchments (Hörmann et al. 2005; Brath et al. 2006; Huisman et al. 2009).
Different studies applying different modeling approaches have identified possible land-use change impacts on catchment hydrology. Understanding the hydrological processes is crucial towards better water and land resource management; for example, hydrology, which is largely influenced by land cover and is highly important to agricultural productivity (Easton et al. 2010). Large changes in land use have often been associated with changes in the local hydrology, as hydrologic responses of a catchment are influenced by the land cover (Nejad Hashemi et al. 2011).
The major effect of land use/cover change is likely to alter the hydrologic response of sub-basins and change water availability (Mengistu 2009). The land cover under little vegetation is subjected to high surface runoff and low water retention (Tufa & Srinivasarao 2014). The land use/cover plays a fundamental role in driving hydrological processes within a sub-basin (Gwate et al. 2015). These include changes in water demands such as irrigation, changes in groundwater recharge, and runoff, and changes in water quality from agricultural runoff (Guo et al. 2008).
Therefore, a far better understanding of land use/cover change, its effect, and interaction with the hydrology of a basin are highly essential in water supply from altered hydrological processes of infiltration, groundwater recharge, and runoff, and changes in water quality from agricultural runoff (Guo et al. 2008).
Ethiopia is one of the most populous countries in Africa with a population of over 94 million people (CSA 2013). Eighty-five percent of the population lives in rural areas and directly depends on the land for its livelihood (Asmamaw et al. 2012). This means the demand for land is increasing as the population increases. Agriculture, which depends on the availability of seasonal rainfall, is the main economy of the country. People need land for food production and housing and it is common practice to clear the forest for farming and housing activities. Therefore, the result of these activities is land use/cover changes due to daily human intervention. Hence, understanding how the land cover changes influence the hydrology of the watershed enables planners to formulate policies to minimize the undesirable effects of future land cover changes. Small-scale sub-basin-based hydrological information considering land use/cover change is crucial for hydrological processes assessment for irrigated agriculture or any use of water. Water availability is becoming a critical factor in so many sectors that the need to assess the anticipated impacts of land use/cover change on hydrology is unquestionable (Tubiello & van der Velde 2007).
Bilate watershed, which is one of the sub-watersheds of the Rift Valley basin, is facing the above types of challenges. Deforestation is a day-to-day activity for people living in and around the watershed due to increasing demand for charcoal, construction, domestic needs, expansion of arable land, and grazing areas (Degelo 2007). This continuous change in land use/cover is expected to impact the water balance of the watershed by changing the magnitude and pattern of the components of hydrological processes, which are surface runoff, baseflow, interflow, and evapotranspiration, which results in increasing the extent of the water management problem.
So, studying the impacts of land use/cover change on the water balance for Bilate watershed was crucial to solving a wide variety of water resources problems, including design of hydraulic structures; urban and highway drainage; planning of flood-control works; source pollution; disposal of waste material; evaluation of environmental impacts of land use and management practices; planning of soil conservation works and agricultural products. This enables the local governments and policymakers to formulate and implement effective and appropriate response strategies to minimize the undesirable effect of future land use/cover change (PHE; Ethiopia Consortium 2011). Similar research was done by Schulla & Jasper (2012), Kebede et al. (2013), Hagos (2014), and Kaiser (2014) using the WaSiM model. Therefore, this research aims to determine the land use/cover change on the water balance of the Bilate watershed using the WaSiM model.
MATERIALS AND METHODS
Description of the study area
Bilate River is one of the inland rivers of Ethiopia, whose source is located in the Gurage Mountains in central Ethiopia. The Bilate watershed is a part of the Main Ethiopian Rift valley basin which is part of the Great Rift Valley. The Bilate River watershed (BRW) covers an area of about 5,625 km2 and is located in the Southern Ethiopian rift valley and partly in the western Ethiopian Highlands (Figure 1), and its elevation is about 1,175 meters.
Input data and methods
Input data for WaSiM model
WaSiM allows various model configurations depending on the aim of the application and on the amount and quality of input data. It is possible to combine various sub-model components and to run the model in various spatial and temporal discretizations.
- (A)
Digital Elevation Model (DEM) data
Initially, 30 m by 30 m resolution DEM data was downloaded from Global Earth Explorer (USGS) https://earthexplorer.usgs.gov or https://www.usgs.gov/core science systems/science analytics and synthesis/gap/science/land-cover-data).
- (B)
Soil data
The FAO/UNESCO-Soil Map of East Africa (2012), available in Arc/Info format with a scale of 1: 1,000,000, was obtained from the GIS and Remote Sensing Department, Ministry of Water, Irrigation, and Electricity of Ethiopia (MOWIE). These data were used as input for the WaSiM model.
- (C)
Land use/cover data
The land use/cover image with three years of spatial resolutions of 30 meters (1989, 2002, and 2015) (https://www.usgs.gov/core science systems/science analytics and synthesis/gap/science/land-cover-data) or https://earthexplorer.usgs.gov were downloaded from USGS Earth Explorer and prepared using ERDAS software and ArcGIS. The WaSiM model was run for three different years' intervals of Landsat data (1989, 2002, and 2015) of land use such as Grassland/Pasture, Barren lands, Rangeland Scrublands, Cultivation/Agriculture, and Mixed Forest.
The following parameter values used as input in the WaSiM model were generated for those land uses within study period intervals: albedo, leaf area index (LAI), leaf surface resistance (Rsc), Intercept Cap, rs_evaporation, aerodynamic roughness length (Z0), vegetation cover fraction (VCF); root depth within those different land use/cover types were obtained from different works of literature and website; http://www.unigiessen.de/~gh1461/plapada/php/list/contentorhttp://www.unigiessen.de/~gh1461/plapada/plapada.html.
- (D)
Meteorological data
The meteorological data required for this study were collected from the National Meteorological Agency of Ethiopia. The daily meteorological data collected for this study include precipitation, maximum and minimum temperature, relative humidity and wind speed, and sunshine hours from the years 1987–2015 in and around the watershed area for 4 stations, as shown in Table 1.
Station name . | Longitude (E) . | Latitude (N) . | Altitude (m) . | Mean annual RF (mm) . | Percentage missed . |
---|---|---|---|---|---|
A.Kulito | 38° 05′ 38.00″ | 7°18′38″ | 1,772 | 1,025 | 0.74 |
Boditi | 37° 57′ 18.00″ | 6° 57′ 13.00″ | 2,043 | 1,154 | 1.97 |
Fonko | 37° 58′ 4.99″ | 7° 38′ 31.99″ | 2,246 | 1,093 | 9.17 |
Hosana | 37° 51′ 14.00″ | 7° 34′ 1.99″ | 2,307 | 1,100 | 3.74 |
Station name . | Longitude (E) . | Latitude (N) . | Altitude (m) . | Mean annual RF (mm) . | Percentage missed . |
---|---|---|---|---|---|
A.Kulito | 38° 05′ 38.00″ | 7°18′38″ | 1,772 | 1,025 | 0.74 |
Boditi | 37° 57′ 18.00″ | 6° 57′ 13.00″ | 2,043 | 1,154 | 1.97 |
Fonko | 37° 58′ 4.99″ | 7° 38′ 31.99″ | 2,246 | 1,093 | 9.17 |
Hosana | 37° 51′ 14.00″ | 7° 34′ 1.99″ | 2,307 | 1,100 | 3.74 |
The missing meteorological daily data were filled by using the Arithmetic mean values method; hence, the total missed values were counted and compared with the data for each year, the percentage of missed data of all stations was less than 10%.
- (E)
Streamflow data
Discharges of two gauging stations, Alaba Kulito and Tena (on Bilate River) are found in the watershed and daily flow data were collected from the Ministry of Water, Irrigation, and Electricity of Ethiopia for both gauging stations.
From the two gauging stations, the Bilate Tena gauging station was selected, because Bilate Tena gauging station was found in the same location as the outlet of the Bilate watershed. Flow data was required for performing sensitivity analysis, calibration, and validation of the model from 1989 to 2015 for the period of 27 years (Table 2).
No. . | Description of data type . | Source . | Resolution . | Years of data . |
---|---|---|---|---|
1 | DEM | Global Earth Explorer (USGS) | 30 m by 30 m | 1989 |
2 | Land use/cover image | http://www.unigiessen.de/~gh1461/plapada/php/list/contentorhttp://www.unigiessen.de/~gh1461/plapada/plapada.html | 30 m | 1989, 2002 & 2015 |
3 | Soil data | Ministry of Water, Irrigation and Electricity of Ethiopia | 30 m | 1992 |
4 | Meteorological data | National Meteorological Agency of Ethiopia | – | 1987–2015 |
5 | Stream flow | Ministry of Water, Irrigation and Electricity of Ethiopia | – | 1987–2015 |
No. . | Description of data type . | Source . | Resolution . | Years of data . |
---|---|---|---|---|
1 | DEM | Global Earth Explorer (USGS) | 30 m by 30 m | 1989 |
2 | Land use/cover image | http://www.unigiessen.de/~gh1461/plapada/php/list/contentorhttp://www.unigiessen.de/~gh1461/plapada/plapada.html | 30 m | 1989, 2002 & 2015 |
3 | Soil data | Ministry of Water, Irrigation and Electricity of Ethiopia | 30 m | 1992 |
4 | Meteorological data | National Meteorological Agency of Ethiopia | – | 1987–2015 |
5 | Stream flow | Ministry of Water, Irrigation and Electricity of Ethiopia | – | 1987–2015 |
Landsat image ID . | Sensor type . | Date acquired . | Path/row . | LULC year . |
---|---|---|---|---|
055-1224 | ETM + | Dec 24, 1989 | 168/55 | 1989 |
054-1130 | ETM + | Dec.12, 1989 | 168/54 | 1989 |
054-1109 | ETM + | Nov.09, 2002 | 168/54 | 2002 |
054-1116 | ETM + | Nov.16, 2002 | 169/54 | 2002 |
055-1224 | ETM + | Dec.24, 2015 | 169/55 | 2015 |
055-1217 | ETM + | Dec.17, 2015 | 168/55 | 2015 |
Landsat image ID . | Sensor type . | Date acquired . | Path/row . | LULC year . |
---|---|---|---|---|
055-1224 | ETM + | Dec 24, 1989 | 168/55 | 1989 |
054-1130 | ETM + | Dec.12, 1989 | 168/54 | 1989 |
054-1109 | ETM + | Nov.09, 2002 | 168/54 | 2002 |
054-1116 | ETM + | Nov.16, 2002 | 169/54 | 2002 |
055-1224 | ETM + | Dec.24, 2015 | 169/55 | 2015 |
055-1217 | ETM + | Dec.17, 2015 | 168/55 | 2015 |
Preprocessing of data
Missing meteorological data estimation
The missing data of meteorological daily data were filled by using the Arithmetic mean values method; hence, the total missed values were counted and compared with the data for each year, the percentage of missed data of all stations was less than 10%.
Consistency test for meteorological data
The data of the given meteorological stations was checked with the help of a double mass-curve method with reference to their neighborhood stations. It was tested using the XLSTAT 2017 Software SNHT test (Amiri 2011).
In the case of the R statistic (R stands for Range), the null and alternative hypotheses are given by H0: the T variables are not homogeneous for what concerns their mean. Two-sided test: Ha the T variables follow one or more distributions that have the same mean. The double mass curve was used to check the consistency of the rainfall stations in the study area, and the analysis shows that the stations were consistent over the considered period.
Filling missed hydrological data
The daily flow data are archived based on m3/s and transformed into mm/time step before implementation into WaSiM, since the available meteorological and hydrological data cover the same period from 1989 to 2015 used.
Y = dependent variables
X = explanatory/independent variables
Areal rainfall
Areal rainfall is the average rainfall over an area, referred to as the areal rainfall distribution and is restricted to a long-term average value. It is expressed as a mean depth (mm) over the catchment area and used to know the distribution of rainfall for the calibration and validation period (Rutebuka et al. 2020). Figure 2 shows the Thiessen polygon and areal proportion of each of the four selected stations in the sub-basins.
Land use/cover data preparation and processing
- (A)
Landsat image processing
After delineating the watershed of the study area, land use/cover data preparation and processing is very crucial to have land cover data for the watershed. Landsat ETM+ was selected for the period of 1989, 2002, and 2015 respectively. To avoid a seasonal variation in vegetation pattern and distribution throughout a year, the selection of the acquired data was made as much as possible in the same annual season of the acquired years. The images used in this study area were orthorectified to a Universal Transverse Mercator projection using datum WGS (World Geodetic System) 84 zone 37N. The acquisition dates, sensor, path/row, resolution, and the producers of the satellite images used in this study are summarized in Table 3 below.
- (B)
Land use/cover classifications
The Land use/cover change studies were differentiated using the available data source such as remote sensing, any other relevant information, and previous local knowledge. Hence, based on the prior knowledge of the study area, ERDAS Imagine, and additional information from previous research in the study area (Degelo 2007; Wagesho 2014; Getahun 2017), the types of land use and land cover were identified for the Bilate watershed. The descriptions of these land use and land covers are given as follows in Table 4.
Major land use land cover . | Their definitions . |
---|---|
Agricultural/cultivation-lands | These include a diverse class of cultivated land, plots covered by food and commercial crops (croplands) and land units covered by residuals after immediate harvest. |
Mixed-forest/Forest-lands | Forestlands usually have tree crown areal density capable of modulating the micro climate and water holding capacity of the watershed. They range from densely populated tall trees of tropical rain forest used for timber to moderately grown green forest. Forestlands could be evergreen, deciduous or mixed forestland. |
Range and Shrub-lands | Range lands are typical to arid and semiarid lands characterized by xerophytic vegetation and transition zones from forest land to sparse woodlands whereas Shrub lands are a plant community characterized by vegetation dominated by shrubs, often also including grasses, herbs, and geophytes. |
Grass-lands/Pasture | Grasslands are land units where the potential natural vegetation is predominantly grasses and grass-like plants. It is dominated by naturally occurring grasses as well as those areas of actual range land that have been modified to include grasses, whereas pasture land is an area covered with grass or other plants suitable for the grazing of livestock. |
Water and Marshy land | Area that remains water logged and swampy throughout the year, and rivers. But water or marshy land (Boyo Lake) was not considered for this study because there is no full data to construct a lake model module in the WaSiM control file. |
Barren land | Land of limited ability to support life and in which less than one-third of the area has vegetation or another cover. |
Major land use land cover . | Their definitions . |
---|---|
Agricultural/cultivation-lands | These include a diverse class of cultivated land, plots covered by food and commercial crops (croplands) and land units covered by residuals after immediate harvest. |
Mixed-forest/Forest-lands | Forestlands usually have tree crown areal density capable of modulating the micro climate and water holding capacity of the watershed. They range from densely populated tall trees of tropical rain forest used for timber to moderately grown green forest. Forestlands could be evergreen, deciduous or mixed forestland. |
Range and Shrub-lands | Range lands are typical to arid and semiarid lands characterized by xerophytic vegetation and transition zones from forest land to sparse woodlands whereas Shrub lands are a plant community characterized by vegetation dominated by shrubs, often also including grasses, herbs, and geophytes. |
Grass-lands/Pasture | Grasslands are land units where the potential natural vegetation is predominantly grasses and grass-like plants. It is dominated by naturally occurring grasses as well as those areas of actual range land that have been modified to include grasses, whereas pasture land is an area covered with grass or other plants suitable for the grazing of livestock. |
Water and Marshy land | Area that remains water logged and swampy throughout the year, and rivers. But water or marshy land (Boyo Lake) was not considered for this study because there is no full data to construct a lake model module in the WaSiM control file. |
Barren land | Land of limited ability to support life and in which less than one-third of the area has vegetation or another cover. |
All the three raster land uses of the watershed were classified into six major types (Grassland/Pasture, Range and Shrub lands, cultivation/agriculture, Mixed Forest, Settlements, and Barren land). To differentiate the cultivated land from barren lands, the season of the land use/land cover downloading was selected as December and November. To parameterize the land use in a distributed way, a land-use grid was required. The land use grid was parameterized with a land-use table that describes each grid cell with a parameter data set according to the grid classification. A specific value, which refers to a land-use type in the control-file, is assigned to each cell of this grid. The characteristics (e.g., root depth, resistance, LAI, VCF) of these types are declared in the land use table in the control-file. Most of the parameters describe a seasonal cycle with maximum (e.g., leaf area index – LAI) or minimum (e.g., stomata resistance – rsc) values during the vegetation period. It is easily shown that the increase of cultivation/agriculture, barren lands, and settlement area causes the decrease of mixed forest area, grasslands/pasture, and range and shrub land over the last 27 years for three LULC maps.
- (C)
Accuracy assessment
The overall accuracy is used to indicate the accuracy of the whole classification (i.e. number of correctly classified pixels divided by the total number of pixels in the error matrix), whereas the other two measures indicate the accuracy of individual classes. User's accuracy is regarded as the probability that a pixel classified on the map represents that class on the ground or reference data, whereas the product's accuracy represents the probability that a pixel on reference data has been correctly classified.
In this study, the assessment was carried out using the original image for 1989 maps and the Google Earth Images for 2002 and 2015, together with previous knowledge of the area, as reference data to generate the testing data set. A total of 83, 85, and 85 testing sample points were selected randomly for the years 1989, 2002, and 2015, respectively. After completing the accuracy process as indicated in Table 6, the overall accuracy estimated as 87% is acceptable. The land cover vector data were converted into an appropriate raster format, grid size, and different land covers. The raster format of the land use map is converted to vector, to ASCII, and then to grid format, which is required as input for the WaSiM hydrological model.
Soil data preparation
The parameterization of the soil's physical properties is crucial for any hydrological model application. The soil hydraulic properties, which are saturated and unsaturated hydraulic conductivity and water retention, control the main hydrological processes (Fox et al. 2005).
The watershed was discretized into five different soil types, presented in Table 5. For the soil parameterization, the method of multiple soil horizons was used, where each soil type may have a different number of horizons. Each soil horizon has different hydraulic properties and may consist of a different number of layers of various thicknesses (Figure 3).
WaSiM codes for major soil map . | Definition for major soil map . |
---|---|
1 | Eutric Vertisols |
2 | Vitric Andosols |
3 | Chromic Luvisols |
4 | Humic Nitisols |
5 | Lithic Leptosols |
WaSiM codes for major soil map . | Definition for major soil map . |
---|---|
1 | Eutric Vertisols |
2 | Vitric Andosols |
3 | Chromic Luvisols |
4 | Humic Nitisols |
5 | Lithic Leptosols |
List of parameters . | Description . |
---|---|
Rsc | For land use |
krec (recession constant) | For the saturated hydraulic conductivities |
Qomax | For base flow(when the soil is saturated) |
dr (drainage density) | For interflow |
List of parameters . | Description . |
---|---|
Rsc | For land use |
krec (recession constant) | For the saturated hydraulic conductivities |
Qomax | For base flow(when the soil is saturated) |
dr (drainage density) | For interflow |
The parameterization of the soil physical properties for each horizon was based on the Van Genuchten parameters after Wösten et al. (1999) and HWSD (Harmonized World Soil Data) Viewer was used to determine percentages of silt, clay, and sand in each layer of the soil.
WaSiM model setup
The control file of the model was adjusted as per the watershed characteristics and available input data. Meteorological input data of the model were interpolated for each grid cell in the watershed and are followed by simulation of the main hydrological processes such as evapotranspiration, interception, infiltration, and the separation of discharge into the direct flow, interflow, base flow, and total simulated flow. These calculations are built modularly and can be adapted to the physical characteristics of the watershed. All spatial data were prepared in a raster data set (grid) with a resolution of 30 m by 30 m.
Sensitivity analysis
The sensitivity analysis includes test runs in which the value of only one coefficient or parameter is changed, while the values of the others remain constant. The WaSiM model sensitive parameters for this study were selected from different findings, which were made using the WaSiM model in different watersheds in Ethiopia and other basins out of our country. The sensitivity analysis was checked using the manual method by setting the values of the sensitive parameters on the WaSiM control file one after the other. From the model runs, sensitivity analysis mainly focused on the unsaturated zone model parameters, land use model (rsc or leaf surface resistance), and soil model (K recession) in the WaSiM-control file but the most sensitive parameters were found in the unsaturated zone model, as shown in Table 6.
Determining the water balance of the watershed
The water budget simulation section of WaSiM comprises a chain of modules that combine both the physical and empirical descriptions of water flow. In this study, constant climatic conditions under changing land use or land cover were considered. The following components or model modules were used to calculate the water balance of watersheds.
Potential and real evapotranspiration
Interception
Infiltration and the unsaturated zone module
Soil model
RESULTS AND DISCUSSION
Land use/cover analysis
Overall accuracy, producer's accuracy, and user's accuracy
The accuracy assessment is used to determine the correctness of the classified image. It was performed using a confusion matrix. The overall accuracy gives the overall results of the confusion matrix. It is calculated by dividing the total number of correct pixels (diagonals) by the total number of pixels in the confusion matrix. The overall accuracy for the maps of 1989, 2002, and 2015 were 87, 80, and 91% respectively hence, they fulfill the minimum requirements.
The producer's accuracy tells us how well a certain area can be classified. It is obtained by dividing the number of correctly classified pixels in the category by the total number of pixels of the category in the reference data. The producer's accuracy is also known as an Omission Error, which is the probability of reference pixels being classified correctly. It gives only the proportion of correctly classified pixels. The overall result of the producer's accuracy ranges from 69% to 93% as indicated in Table 7.
. | LULC classification data . | . | . | ||||||
---|---|---|---|---|---|---|---|---|---|
. | . | CL . | RL . | GL . | MF . | S . | BL . | Total . | Users accuracy . |
LULC classification data | CL | 18 | 1 | 1 | 20 | 90 | |||
RL | 1 | 12 | 1 | 1 | 15 | 80 | |||
GL | 1 | 16 | 1 | 18 | 89 | ||||
MF | 1 | 1 | 10 | 12 | 83 | ||||
S | 1 | 1 | 15 | 17 | 88 | ||||
BL | 1 | 14 | 15 | 93 | |||||
Total | 22 | 13 | 18 | 11 | 17 | 15 | 97 | ||
Producers accuracy(%) | 82 | 92 | 89 | 90 | 88 | 93 | Overall accuracy = 87 |
. | LULC classification data . | . | . | ||||||
---|---|---|---|---|---|---|---|---|---|
. | . | CL . | RL . | GL . | MF . | S . | BL . | Total . | Users accuracy . |
LULC classification data | CL | 18 | 1 | 1 | 20 | 90 | |||
RL | 1 | 12 | 1 | 1 | 15 | 80 | |||
GL | 1 | 16 | 1 | 18 | 89 | ||||
MF | 1 | 1 | 10 | 12 | 83 | ||||
S | 1 | 1 | 15 | 17 | 88 | ||||
BL | 1 | 14 | 15 | 93 | |||||
Total | 22 | 13 | 18 | 11 | 17 | 15 | 97 | ||
Producers accuracy(%) | 82 | 92 | 89 | 90 | 88 | 93 | Overall accuracy = 87 |
Note: CL, cultivation/agricultural land, RL, Range land, GL, grass land, MF, mixed forest, S, settlement, and BL, bare land.
User's accuracy is the ratio between the total number of pixels correctly belonging to a class (diagonal elements) and the total number of pixels assigned to the same class by the classification procedure (row total). This quantity explains the probability that a pixel of the classified image truly corresponds to the class to which it has been assigned. In this study, the user's accuracy ranges from 80% to 93%.
Land use/cover maps
The land use/land cover map of 1989 in Tables 8 and 9 shows that the total cultivated land/agriculture coverage class was about 28.92% of the total area of the watershed. It increased rapidly and became 34.12% and 43.3 of the watershed in 2002 and 2015 land use land cover maps respectively. This is mainly because of the population growth that caused the increase in demand for new cultivation land and settlement, which in turn resulted in shrinking of other types of land use land cover of the area. The forest coverage in 1989 was about 21.79% of the total area of the watershed. However, in the year the 2002 and 2015, this was reduced to almost 12.79% and 6.4% of the total area respectively. This is most probably because of the deforestation activities that have taken place for agriculture and the expansion of settlements.
WaSiM codes for three LULC data . | Definition for major LULC . | ||
---|---|---|---|
LU1989 . | LU2002 . | LU2015 . | |
1 | 1 | 2 | Grasslands/Pasture |
2 | 2 | 5 | Range and Shrub lands |
3 | 3 | 4 | Cultivation/agriculture |
4 | 4 | 1 | Mixed Forest |
5 | 5 | 3 | Settlements |
6 | 6 | 6 | Barren lands |
WaSiM codes for three LULC data . | Definition for major LULC . | ||
---|---|---|---|
LU1989 . | LU2002 . | LU2015 . | |
1 | 1 | 2 | Grasslands/Pasture |
2 | 2 | 5 | Range and Shrub lands |
3 | 3 | 4 | Cultivation/agriculture |
4 | 4 | 1 | Mixed Forest |
5 | 5 | 3 | Settlements |
6 | 6 | 6 | Barren lands |
. | Percentage land use/land cover . | Percentage change . | ||||
---|---|---|---|---|---|---|
Land use/land cover class . | 1989 . | 2002 . | 2015 . | 1989–2002 . | 2002–2015 . | 1989–2015 . |
Grasslands/Pasture | 10.48% | 11.11% | 8.08% | 6.00% | -20.79% | -22.00% |
Range and Shrub lands | 26.35% | 23.24% | 15.06% | -12.00% | -35.00% | -43.00% |
Cultivation/agriculture | 28.92% | 34.12% | 43.30% | 18.00% | 27.00% | 50.00% |
Mixed Forest | 21.79% | 12.30% | 6.40% | -43.00% | -48.00% | -71.00% |
Settlements | 10.25% | 16.23% | 21.98% | 58.00% | 35.00% | 114.00% |
Barren lands | 2.20% | 3.00% | 5.18% | 30.00% | 73.00% | 135.00% |
. | Percentage land use/land cover . | Percentage change . | ||||
---|---|---|---|---|---|---|
Land use/land cover class . | 1989 . | 2002 . | 2015 . | 1989–2002 . | 2002–2015 . | 1989–2015 . |
Grasslands/Pasture | 10.48% | 11.11% | 8.08% | 6.00% | -20.79% | -22.00% |
Range and Shrub lands | 26.35% | 23.24% | 15.06% | -12.00% | -35.00% | -43.00% |
Cultivation/agriculture | 28.92% | 34.12% | 43.30% | 18.00% | 27.00% | 50.00% |
Mixed Forest | 21.79% | 12.30% | 6.40% | -43.00% | -48.00% | -71.00% |
Settlements | 10.25% | 16.23% | 21.98% | 58.00% | 35.00% | 114.00% |
Barren lands | 2.20% | 3.00% | 5.18% | 30.00% | 73.00% | 135.00% |
The individual class areas and change statistics for the three periods are summarized in Table 9. The results of previous studies showed the same fact in the Bilate watershed. For example, Wagesho (2014) and Wakjira (2016) reported that cultivation and settlements of Bilate watershed were increased by 61.6% and the mixed forest decreased to 67.7% from 1976 to 2000. Hence, the impacts of land use/cover change of the Bilate watershed are indicated in Figures 4 and 5.
Sensitivity analysis
The executed sensitivity analysis mainly focused on the unsaturated zone model parameters. Table 10 shows the results of the sensitivity analysis. Both Q0 and kB are quite sensitive to the base flow, as expected. Furthermore, kD is considerably sensitive to peak flow. Finally, the drainage parameter dr seems to be quite sensitive to the base flow. An increase in the kB value always results in an increased base flow value for low flow conditions at the beginning of the calibration period. Finally, the higher the kD value, the lower the direct flow and when the value of Ki increased, the value of the interflow becomes lower. A similar analysis was done by Kaiser (2014).
Parameter . | Description . | Optimum value . | Ranges . | |
---|---|---|---|---|
Min . | Max . | |||
kD [h] | Recession constant for direct runoff | 220 | 0 | 220 |
KI [h] | Recession constant for interflow | 100 | 0 | 100 |
dr [m-1] | Drainage density | 60 | 10 | 80 |
KB [h] | Recession constant for base flow | 0.155 | 0 | 1 |
Krec [-] | Recession constant for saturated hydraulic conductivity with depth | 0.9 | 0.1 | 1 |
rs-evaporation [s/m] | Soil surface resistance (for evaporation only) | 80 | 20 | 100 |
rsc [s/m] | Leaf surface resistance | 250 | 50 | 300 |
Parameter . | Description . | Optimum value . | Ranges . | |
---|---|---|---|---|
Min . | Max . | |||
kD [h] | Recession constant for direct runoff | 220 | 0 | 220 |
KI [h] | Recession constant for interflow | 100 | 0 | 100 |
dr [m-1] | Drainage density | 60 | 10 | 80 |
KB [h] | Recession constant for base flow | 0.155 | 0 | 1 |
Krec [-] | Recession constant for saturated hydraulic conductivity with depth | 0.9 | 0.1 | 1 |
rs-evaporation [s/m] | Soil surface resistance (for evaporation only) | 80 | 20 | 100 |
rsc [s/m] | Leaf surface resistance | 250 | 50 | 300 |
Model calibration and validation
The WaSiM sensitive parameters were identified/selected from other authors' findings: Schulla & Jasper (2012), Kebede et al. (2013), Hagos (2014), and Kaiser (2014). Finally, the sensitive parameters for this study are listed in Table 10 and the maximum, minimum, and optimum values of the sensitive parameters of this study are identified, as indicated in the table.
As reported in Kebede et al. (2013), the parameters such as rs_evaporation (soil surface resistance for evaporation only) and rsc (leaf surface resistance) were also calibrated manually. The other parameters in the soil water dynamics of the WaSiM-ETH were KD, and KI (Table 10), which control the surface runoff and interflow storage effects in the Richards equation, which was used for this study. Finally, the parameters Krec, dr, KD, and KI were found to be very sensitive. From all of these sensitive parameters, dr was the most sensitive parameter.
As the model analysis of this research indicated, the hydrographs were good at simulating the daily, weekly, and monthly scales. The monthly simulation indicates better than daily and weekly simulation for this study. From the results indicated in Table 10, the index of agreement d lies in the range of 0.0–1.0 with higher values indicating better agreement. Similarly, NSE and EVC range from minus infinity to 1.0, with higher values indicating better agreement, and a value of 1.0 being the optimal value.
The coefficient of efficiency (R2) and Nash Sutcliffe model efficiency (NSE) values were used to examine model performance and the result indicates 0.85 and 0.89 to the coefficient of efficiency (R2) and 0.85 and 0.89 to Nash Sutcliffe model efficiency (NSE) during calibration and validation respectively, but the model shows underestimation in some years because some flow data problems show the outliers (Figures 6–9). From the other findings, Wagesho (2014) reported 0.78 (R2) and 0.611 (NSE) for the calibration and 0.78 (R2) and 0.623 (NSE) for validation using the HEC HMS model and Getahun (2017) reported, 0.79 (R2) and 0.78 (NSE) for calibration and 0.64 (R2) and 0.60 (NSE) for validation using the SWAT model in Bilate watershed.
To evaluate the general model performance of the distributed hydrological model, Brincker et al. (2001) suggest the following graduation of the achieved model efficiency. Based on general performance criteria, the model indicated good performance since the values of DV and R2 for the calibration and validation period were 17.42, 0.85, and 10.5, 0.89 respectively (Table 11).
Performance of WaSiM model for Bilate watershed
Impacts of LULCC on water balance of Bilate watershed
The analysis of the LULCC contribution was made on surface runoff; interflow, base flow, total simulated discharge, and evapotranspiration as characteristics of the hydrological process of the watershed. The contribution of surface runoff, total simulated flow, and interflow have increased from 1989 to 2015, as indicated in Figure 11. This was related to the surface cover of the watershed since changing the forest land of the watershed to agricultural land accelerated the runoff rate and reduced the infiltration of soil moisture content (Table 8). From the result of the land use land cover map, areas of forest have decreased from 1989 to 2015, which has contributed to the increasing surface runoff contribution. In the same manner, Wagesho (2014) reported that the simulated surface runoff component increased progressively since the 1970s in Bilate watershed. As in Bahati et al. (2021), the historical/current minimum, maximum, and mean annual flow of Muziz river, future minimum, maximum, and mean annual flow will increase respectively.
On the other hand, the rate of evapotranspiration has decreased from 1989 to 2015, indicating losses are mainly through evapotranspiration. These result revealed that the land use/land cover change has significant impacts on infiltration rates, on runoff production, total simulation flow, interflow, base flow, evapotranspiration, and water retention capacity of the soil or change in storage of the soil; hence, it affects the water balance of the watershed. This is because the land cover under little vegetation is subjected to high surface runoff, low water retention, and low evapotranspiration (Tufa & Srinivasarao 2014). The changes in water balance (hydrological process) under the land use/land cover changes are summarized in Figure 10.
As shown in Tables 12 and 13, the simulated water balance for the Bilate watershed using the WaSiM-ETH reveals the interflow component of the water balance takes a higher fraction of simulated discharge. The change in soil water storage ΔS is the result of the balance, being positive when the profile has a net gain of water, and negative for a net loss (Reichardt et al. 1995). From the results, the mean annual stream flows were evaluated due to land use/land cover change in the Bilate watershed, as shown in Figure 11.
. | Daily . | Weekly . | Monthly . | |||
---|---|---|---|---|---|---|
Criteria . | Calibration 1989–2003 . | Validation 2007–2015 . | Calibration 1989–2003 . | Validation 2007–2015 . | Calibration 1989–2003 . | Validation 2007–2015 . |
EV | 0.77 | 0.82 | 0.87 | 0.70 | 0.89 | 0.93 |
R2 | 0.73 | 0.77 | 0.82 | 0.96 | 0.85 | 0.89 |
RMSE | 0.19 | 0.24 | 0.41 | 0.01 | 0.71 | 0.71 |
NS | 0.73 | 0.77 | 0.82 | 0.96 | 0.85 | 0.89 |
Coefficient of determination | 0.77 | 0.48 | 0.86 | 0.58 | 0.89 | 0.90 |
Index of agreement | 0.93 | 0.93 | 0.95 | 0.99 | 0.96 | 0.97 |
. | Daily . | Weekly . | Monthly . | |||
---|---|---|---|---|---|---|
Criteria . | Calibration 1989–2003 . | Validation 2007–2015 . | Calibration 1989–2003 . | Validation 2007–2015 . | Calibration 1989–2003 . | Validation 2007–2015 . |
EV | 0.77 | 0.82 | 0.87 | 0.70 | 0.89 | 0.93 |
R2 | 0.73 | 0.77 | 0.82 | 0.96 | 0.85 | 0.89 |
RMSE | 0.19 | 0.24 | 0.41 | 0.01 | 0.71 | 0.71 |
NS | 0.73 | 0.77 | 0.82 | 0.96 | 0.85 | 0.89 |
Coefficient of determination | 0.77 | 0.48 | 0.86 | 0.58 | 0.89 | 0.90 |
Index of agreement | 0.93 | 0.93 | 0.95 | 0.99 | 0.96 | 0.97 |
Year . | P . | ETR . | Q . | . | QDIR . | QINT . | QBAS . |
---|---|---|---|---|---|---|---|
1989 | 1041.13 | 940.62 | 64.55 | 35.96 | 5.65 | 44.01 | 14.89 |
1990 | 969.69 | 869.64 | 111.29 | −11.24 | 10.05 | 91.01 | 10.24 |
1991 | 917.52 | 806.86 | 52.62 | 58.04 | 4.51 | 40.29 | 7.82 |
1992 | 1168.82 | 976.98 | 122.14 | 69.70 | 14.86 | 86.45 | 20.84 |
1993 | 1148.50 | 1015.16 | 156.92 | − 23.58 | 17.78 | 126.04 | 13.10 |
1994 | 964.37 | 834.09 | 43.19 | 87.09 | 3.60 | 29.94 | 9.66 |
1995 | 882.95 | 752.16 | 55.02 | 75.77 | 4.86 | 38.55 | 11.60 |
1996 | 1112.69 | 969.60 | 118.03 | 25.06 | 14.69 | 87.48 | 15.86 |
1997 | 1003.41 | 834.65 | 48.33 | 120.43 | 6.99 | 34.48 | 6.86 |
1998 | 1015.15 | 884.91 | 70.29 | 59.95 | 7.66 | 53.67 | 8.96 |
1999 | 715.80 | 673.35 | 21.73 | 20.72 | 1.55 | 16.55 | 3.63 |
2000 | 929.08 | 765.87 | 65.70 | 97.51 | 7.61 | 47.96 | 10.12 |
2001 | 1166.17 | 988.80 | 102.91 | 74.46 | 11.99 | 79.88 | 11.04 |
2002 | 958.23 | 895.88 | 57.03 | 5.32 | 4.25 | 43.05 | 9.73 |
2003 | 1101.16 | 963.84 | 100.81 | 36.51 | 14.61 | 75.21 | 10.99 |
Average | 1006.31 | 878.16 | 79.37 | 48.78 | 8.71 | 59.64 | 11.02 |
Year . | P . | ETR . | Q . | . | QDIR . | QINT . | QBAS . |
---|---|---|---|---|---|---|---|
1989 | 1041.13 | 940.62 | 64.55 | 35.96 | 5.65 | 44.01 | 14.89 |
1990 | 969.69 | 869.64 | 111.29 | −11.24 | 10.05 | 91.01 | 10.24 |
1991 | 917.52 | 806.86 | 52.62 | 58.04 | 4.51 | 40.29 | 7.82 |
1992 | 1168.82 | 976.98 | 122.14 | 69.70 | 14.86 | 86.45 | 20.84 |
1993 | 1148.50 | 1015.16 | 156.92 | − 23.58 | 17.78 | 126.04 | 13.10 |
1994 | 964.37 | 834.09 | 43.19 | 87.09 | 3.60 | 29.94 | 9.66 |
1995 | 882.95 | 752.16 | 55.02 | 75.77 | 4.86 | 38.55 | 11.60 |
1996 | 1112.69 | 969.60 | 118.03 | 25.06 | 14.69 | 87.48 | 15.86 |
1997 | 1003.41 | 834.65 | 48.33 | 120.43 | 6.99 | 34.48 | 6.86 |
1998 | 1015.15 | 884.91 | 70.29 | 59.95 | 7.66 | 53.67 | 8.96 |
1999 | 715.80 | 673.35 | 21.73 | 20.72 | 1.55 | 16.55 | 3.63 |
2000 | 929.08 | 765.87 | 65.70 | 97.51 | 7.61 | 47.96 | 10.12 |
2001 | 1166.17 | 988.80 | 102.91 | 74.46 | 11.99 | 79.88 | 11.04 |
2002 | 958.23 | 895.88 | 57.03 | 5.32 | 4.25 | 43.05 | 9.73 |
2003 | 1101.16 | 963.84 | 100.81 | 36.51 | 14.61 | 75.21 | 10.99 |
Average | 1006.31 | 878.16 | 79.37 | 48.78 | 8.71 | 59.64 | 11.02 |
Where, P is Precipitation, Q is total runoff, is change in Storage, ETR is Evapotranspiration, QDIR is Direct flow, QINT Inter flow, QBASE is Base flow.
Year . | P . | ETR . | Q . | . | QDIR . | QINT . | QBAS . |
---|---|---|---|---|---|---|---|
2007 | 1133.59 | 1010.06 | 58.94 | 64.59 | 6.24 | 39.05 | 13.65 |
2008 | 1025.70 | 913.95 | 43.47 | 68.28 | 3.96 | 31.09 | 8.43 |
2009 | 910.39 | 813.42 | 36.00 | 60.97 | 2.37 | 25.22 | 8.41 |
2010 | 1197.51 | 1032.93 | 144.68 | 19.90 | 19.75 | 110.90 | 14.04 |
2011 | 945.56 | 877.70 | 42.08 | 25.78 | 3.36 | 29.56 | 9.17 |
2012 | 859.41 | 741.31 | 41.98 | 76.12 | 2.00 | 29.34 | 10.63 |
2013 | 1180.71 | 1022.47 | 112.40 | 45.84 | 11.02 | 80.76 | 20.62 |
2014 | 1105.60 | 904.38 | 153.19 | 48.03 | 15.86 | 115.63 | 21.69 |
2015 | 899.50 | 772.89 | 41.41 | 85.20 | 2.73 | 30.84 | 7.84 |
Average | 1028.66 | 898.79 | 74.91 | 54.97 | 7.48 | 54.71 | 12.72 |
Year . | P . | ETR . | Q . | . | QDIR . | QINT . | QBAS . |
---|---|---|---|---|---|---|---|
2007 | 1133.59 | 1010.06 | 58.94 | 64.59 | 6.24 | 39.05 | 13.65 |
2008 | 1025.70 | 913.95 | 43.47 | 68.28 | 3.96 | 31.09 | 8.43 |
2009 | 910.39 | 813.42 | 36.00 | 60.97 | 2.37 | 25.22 | 8.41 |
2010 | 1197.51 | 1032.93 | 144.68 | 19.90 | 19.75 | 110.90 | 14.04 |
2011 | 945.56 | 877.70 | 42.08 | 25.78 | 3.36 | 29.56 | 9.17 |
2012 | 859.41 | 741.31 | 41.98 | 76.12 | 2.00 | 29.34 | 10.63 |
2013 | 1180.71 | 1022.47 | 112.40 | 45.84 | 11.02 | 80.76 | 20.62 |
2014 | 1105.60 | 904.38 | 153.19 | 48.03 | 15.86 | 115.63 | 21.69 |
2015 | 899.50 | 772.89 | 41.41 | 85.20 | 2.73 | 30.84 | 7.84 |
Average | 1028.66 | 898.79 | 74.91 | 54.97 | 7.48 | 54.71 | 12.72 |
Where, P is Precipitation, Q is total runoff, is change in Storage, ETR is Evapotranspiration, QDIR is Direct flow, QINT Inter flow, QBASE is Base flow.
Hydrological processes . | LULC_1989 . | LULC_2002 . | LULC_2015 . |
---|---|---|---|
Evapotranspiration(ETR),mm | 890.84 | 887.84 | 837.58 |
Total simulated flow(Q), mm | 74.898 | 86.08 | 133.34 |
Surface runoff(QDIR), mm | 7.844 | 9.02 | 14.13 |
Inter flow(QINT), mm | 55.50 | 59.00 | 97.51 |
Base flow(QBAS), mm | 11.55 | 12.50 | 21.69 |
Hydrological processes . | LULC_1989 . | LULC_2002 . | LULC_2015 . |
---|---|---|---|
Evapotranspiration(ETR),mm | 890.84 | 887.84 | 837.58 |
Total simulated flow(Q), mm | 74.898 | 86.08 | 133.34 |
Surface runoff(QDIR), mm | 7.844 | 9.02 | 14.13 |
Inter flow(QINT), mm | 55.50 | 59.00 | 97.51 |
Base flow(QBAS), mm | 11.55 | 12.50 | 21.69 |
The annual simulation of hydrological processes was analyzed for LULC_1989, LULC_2002, and LULC_2015 data. The result indicated that there was a change in total simulation flow, evapotranspiration, surface runoff, base flow and inters flow in each land use/land cover data (Table 12). In the study year intervals (1989–2015), change in total simulated flow increased by 77.83%, and direct runoff, interflow, and base flow increased by 80.23%, 75.69%, and 87.79% respectively. Hence, evapotranspiration decreased by 6% throughout the study time.
Hydrological cycling in a watershed can be characterized and quantified by a water balance, which is the computation of all water fluxes at the boundaries of the system under consideration. It is an itemized statement of all gains, losses, and changes of water storage within a specified elementary volume of soil. From this study, rainfall was considered the gain, where evapotranspiration, total simulated flow (runoff, interflow, and base flow) were considered as the losses (Table 14).
CONCLUSION
In this study, we analyzed the impact of land use/land cover change on the water balance of the Bilate watershed. As part of our analysis, we considered six dominant land use/land covers including mixed forest, cultivation/agricultural land, barren land, grassland/pasture, range and shrub land, and settlement on the Bilate watershed for LULC_1989, LULC_2002, and LULC_ 2015. Like many before us, we found ArcGIS to be a very important tool for the preparation of input data for analyses and the WaSiM model to be important for considering land use land cover data, soil, and DEM data.
The advancement of computational power and the availability of spatial and temporal data have made hydrological models attractive tools to examine and analyze the characteristics of watersheds and how the hydrological process of the catchment functions under varying land-use dynamics. Particularly in this study, hydrological modeling is a useful tool for investigating interactions among the watershed components and hydrologic response analysis to LULCC at various spatial and temporal scales.
The simulated water balance for the Bilate watershed using the WaSiM-ETH showed the interflow component of the water balance takes a higher fraction of simulated discharge and also surface runoff and total simulation flow increased through the study period. Additionally, the ETR, ETP, and soil storage capacity decreased throughout the study periods. Overall, the hydrological model describes changes in the water balance from 1989 to 2015, which indicate that the change in total simulated flow increased by 77.83%, and direct runoff, interflow, and base flow increased by 80.23%, 75.69%, and 87.79% respectively. Additionally, evapotranspiration decreased by 6% throughout the study time.
The future sustainable land and water resources in the Bilate watershed depend on the spatial planning of land use to optimize environmental benefits. Factors that must be considered include managing surface runoff control, erosion protection, flood protection, and water availability. Finally, educating the community on the effect of the unplanned land-use practices on the environment, natural resources, and ecosystem are of paramount importance for the future sustainability of the watershed.
ACKNOWLEDGEMENTS
Special thanks are due to the Ethiopian National Meteorological Agency and Ethiopian Ministry of Water, Irrigation, and Electricity for providing us with the meteorological and hydrological data.
DISCLOSURE STATEMENT
The authors declare no conflict of interest.
DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.