Separate and Combined Impacts of Land Cover and Climate Changes on Hydrological Responses of Dhidhessa River Basin, Ethiopia

Land cover and climate changes greatly influence hydrologic responses of a basin. However, the response vary from basin to basin depending on the nature and severity of the changes and basin characteristics. Moreover, the combined impacts of the changes affect hydrologic responses of a basin in an offsetting or synergistic manner. This study quantified the separate and combined impacts, and the relative contributions of land cover and climate changes on multiple hydrological regimes (i.e., surface runoff, streamflow, groundwater recharge evapotranspiration) for the Dhidhessa Subbasin. Land cover and climate change data were obtained from a recent study completed for the basin. Calibrated Soil and Water Analysis Tool (SWAT) was used to quantify the impacts. The result showed that SWAT model performed well for the Dhidhessa Subbasin in predicting the water balance components. Substantial land cover change as well as an increasing temperature and rainfall trends were reported in the river basin during the past three decades. In response to these changes, surface runoff, streamflow and actual evapotranspiration (AET) increased while groundwater recharge declined. Surface runoff was more sensitive to land cover than to climate changes whereas streamflow and AET were more sensitive to climate change than to land cover change. The combined impacts played offsetting effect on groundwater recharge and AET while inconsistent effects within study periods for other hydrologic responses. Overall, the predicted hydrologic responses will have negative impacts on agricultural production and water resources availability. Therefore, the implementation of integrated watershed management strategies such as soil and water conservation and afforestation could reverse the negative impacts.


Introduction
Land cover and climate changes are key components of global environmental changes [1,2,3,4]. Human-induced land cover change affects the biophysics, biogeochemistry and biogeography of the terrestrial surface and the atmosphere [5,6]. These changes in material and energy cycle of the Earth's systems affect energy balance between land surface and atmosphere subsequently resulting in climate and ecosystem changes [6,7,8,9]. Overall, land cover change together with climate forcing influences the atmosphere and land surface processes [9,10].
Land cover change affects hydrological processes by altering evapotranspiration (ET), soil water holding capacity of a soil, infiltration, precipitation interception capacity [10,11,12] and runoff travel times [12]. Consequently, land cover change modifies the quantity and quality of surface and subsurface water flows [3,13], the magnitude and frequency of floods and droughts [14], and regional and global climate systems. At basin scale, the effect of land cover changes on hydrological processes are reflected in the supply-and-demand balance of water resources, which in turn affect ecosystems, environment and economy [1]. Generally, the separate and the combined impacts of land cover and climate changes influence the environment particularly hydrological processes of a basin. Therefore, investigating the impacts of land cover and climate changes on hydrological processes became a major research topic during the last few decades. Studies demonstrated that land cover and climate changes greatly affect hydrologic responses of a basin [15,16,17]. The magnitude of the change, however, varies from basin to basin depending on the basin characteristics such as geography, geology, topography, climate conditions, and intensity of land cover changes. Hydrological responses to land cover and climate changes could exhibit spatiotemporal variability even within a basin [16,18]. Moreover, the relative contribution of land cover and climate changes varies from region to region.
As a result, research findings on the topic remain inconclusive. For example, some studies reported that climate change influences hydrologic response more than land cover change [4,15,19]; others claimed that hydrologic responses are more sensitive to land cover change than to climate change [20,21,22]. On the other hand, some studies argued that it is vital to consider 4 impacts of both land cover and climate changes on hydrologic responses than examining the separate impacts [19,23]. The combined impacts could have either synergistic or offsetting effects compared to the impact resulting from neither factor [15,19]. For example, Shrestha and Htut [19] reported that the combined impacts of land cover and climate changes on streamflow is more significant than their separate impacts in Mynamar. Zhang et al. [4], however, reported that hydrological response to the combined effect is similar to those induced by climate change alone in China. These studies showed that impacts of climate and land cover changes on hydrological responses can't be generalized as they could be basin specific. Therefore, considering the separate and combined impacts of land cover and climate changes on hydrologic responses is essential to understand the hydrological processes and design effective water resource management strategies [24,25].
Despite this acknowledgement, only limited studies [15,26,27] have investigated the impacts of both land cover and climate changes on hydrological responses. According to Krysanova and White [28], for example, only 30 peer-reviewed papers published between 2004 and 2015 addressed the combined effects of land cover change and climate change at global scale using SWAT model. In comparison, 210 and 109 peer-reviewed papers were published considering the separate impacts of climate change and land cover change, respectively between 2004 and 2015. However, papers published on the combined impacts of land cover and climate changes have been increasing recently but still more papers are dealing with the separate than the combined impacts. In addition, studies showed the need to investigate the impacts of the changes on hydrologic responses at local scale [29,30]. This is because the effect can be evident at subbasin scale than at basin scale [4,31]. In addition, most of the existing studies examined one or limited number of hydrologic responses (e.g., surface runoff, streamflow, evapotranspiration or groundwater recharge) to analyze the impacts [8,12,17,29,32,33,34].
The separate and combined impacts of land cover and climate changes on hydrological processes can be investigated using process based hydrological model [32,35]. Hydrological model (e.g., Soil and Water Assessment Tool; SWAT) is best options for integrated modeling approach because of their capability to predict hydrologic responses to spatiotemporally variable inputs (e.g., land cover and precipitation). These inputs can be derived from satellite remote sensing, or field (ground) data. Remote sensing is more viable for data-scarce basins due to its cost-and time- effectiveness, large area coverage, consistent data quality, and provision of long-term spatiotemporal global information [36,37,38]. However, remote sensing data have unavoidable errors emanated from different sources of errors, therefore, should be calibrated and validated before use before using it for any applications.
The Dhidhessa River basin has experienced substantial land cover and climate changes during the last three decades [39,40]. However, the hydrologic responses to these changes as well as the relative contribution of land cover and climate changes on hydrologic responses are not investigated in most basins of Ethiopia including for the Dhidhessa Subbasin. In this study, we examined both the separate and the combined effects of land cover and climate changes on hydrological responses at regional (local) scale for Dhidhessa River basin. We used an integrated modeling approach to quantify the impacts on multiple hydrological processes including surface runoff, streamflow (water yield), groundwater recharge and evapotranspiration using satellite rainfall product. The Dhidhessa River basin has experienced substantial land cover and climate changes during the last three decades [39,40]. This study analyzed the effects of these changes on multiple hydrological responses of the basin that contributes more than 25% of Blue Nile River's streamflow during dry season. Results of this study can be valuable to manage existing and planned water resources projects in Didhessa River basin, Blue Nile basin, and other hydrologically similar basins in Ethiopia and elsewhere.

Descriptions of the study area
The Dhidhessa River basin drains to the Blue Nile River (Figure 1). It is one of the largest and most important Blue Nile River basins in terms of its physiography and hydrology [41]. Temperature and rainfall in the Dhidhessa River basin exhibit substantial spatial and seasonal variability. The mean daily maximum and minimum air temperatures in the river basin range from 20-33 °C and 6-19 °C , respectively. The mean annual rainfall ranges from 1200 mm to 2200 mm in the river basin. Soils in the river basin are generally deep and have high organic content implying they have high infiltration potential. The dominant soil type is Acrisols while Cambisols and Nitisols are common [42]. Igneous, sedimentary and metamorphic rocks are common in the basin with igneous rock, particularly basalt being dominant [43]. Forest, shrubland, grassland, and agriculture are the primary land cover types in the basin [40]. Major crops include perennial and cash crops like coffee, Mango and Avocado [42]. The river basin is notable for its plant and animal biodiversity.    [39] reported the suitability of these products for the Dhidhessa River basin. Climate change information done using these data was obtained from Kabite et al. [39]. Streamflow data for Dhidhessa River from 1982 to 2014 was obtained from Ethiopian Ministry of Water, Irrigation and Energy (MoWIE) for a station near the town of Arjo ( Figure 1). Only one gauge station was used to calibrate and validate the SWAT model because most of the subbasin are generally ungauged. This is one of the limitation of this study. The streamflow data was used to calibrate SWAT as described in the section above.
Generally, land cover and climate changes information for the Dhidhessa River basin were obtained from Kabite et al. [40] and Kabite et al. [39], respectively.

Methodology
Distributed process-based hydrological models are effective for assessing the separate and combined impacts of land cover and climate changes on hydrological processes [44]. Soil and Water Assessment Tool (SWAT) is a semi-distributed process-based hydrological model that was developed to predict the impacts of land management practices on water, sediment, and 8 agricultural chemical yields [45]. It is now widely used for various applications throughout the world including in the Blue Nile basin for water balance studies, soil erosion, land cover or climate change impact assessment [46,47,48]. The capability of SWAT to characterize hydrologic responses of a basin for various land-use and climate change scenarios was well documented [1,3,49,50]. The deterministic nature of SWAT makes the model suitable to separate the contributions of climate and land cover changes on hydrologic response [33]. As such, SWAT was used in this study. As previously described, spatial datasets such as land cover, DEM and soil and the temporal datasets like weather data (i.e., rainfall and temperature), obtained from various sources were used to setup and calibrate SWAT.

Description of SWAT
SWAT is designed to operate at various spatiotemporal scales, environmental conditions, and different spatial and temporal details [47]. SWAT is integrated with Geographic Information System (GIS) and handles the spatiotemporal heterogeneity of basin characteristics such as topography, land cover, soil and climate condition [32,51]. For this study, SWAT model was carried out at monthly timescale. In addition, a tool known as SWAT-CUP simplifies calibration, sensitivity analysis and uncertainty analysis of SWAT model. Unlike many distributed models, SWAT is not data intensive [52]. As such, SWAT is suitable for data scarce basins like Dhidhessa River basin. SWAT simulates hydrological responses using water the balance equation as, where is the final soil water content (mm), is the initial soil water content on day i 9 SWAT uses a two-level discretization scheme: i) sub-basin creation based on topographic information, and ii) Hydrological Response Unit (HRU) creation by further discretizing of the subbasins based on land use, soil type slope classes. HRU is a basic computational unit assumed to have homogeneous hydrologic response. Hydrological processes are first simulated at the HRU level and then routed through the reaches using the Muskingum routing method [45]. SWAT estimates surface runoff using the modified United State Department of Agriculture (USDA) SCS curve number method (Eq. 2 and 3), and potential evapotranspiration using the Hargreaves method. In this study, a minimum threshold area of 400 km 2 was used for determining the number of subbasins, and 5% threshold for soil, slope, and land use were used for the HRU definition.
Consequently, the Dhidhessa River basin was discretized in to 41 subbasins and 800 HRUs.

SWAT model calibration
SWAT model should be calibrated before using the model output for any application.
SWAT-CUP is one of the widely used software for SWAT model calibration [54]. The Sequential Uncertainty Fitting: SUFI-2 algorithm (SUFI-2) available from SWAT-CUP was used for the calibration effort. SUFI-2 provides more reasonable and balanced predictions than the generalized likelihood uncertainty estimation (GLUES) and the parameter solution (ParaSol) methods also supported by SWAT-CUP [53,55].
Determining the most sensitive parameters for a given watershed is the first requirement for SWAT model calibration. For this study, 19 parameters were considered for calibration based on recommendations by previous studies [56,57]. Global sensitivity analysis was performed using SWAT-CUP on the parameters, and 11 were deemed sensitive for the Dhidhessa River basin. or smaller than their observed ones. In addition, the total uncertainty in the model prediction is commonly measured by P-and R-factor. P-factor represents the percentage of observed data enveloped by our modelling result, the 95 percent prediction uncertainty (95PPU). And the Rfactor represents the ratio of the average width of the 95PPU band to the standard deviation of observed data. For the realistic model prediction, P-factor ≥0.7 and R-factor ≤1.5 is required [54,58].

Simulation Scenarios
The following procedure was employed to assess the separate and combined impacts of land cover and climate changes on hydrological response. First, the climate data was divided in to three (i.e., 1982 to 1995, 1995 to 2008 and 2008 to 2018) based on a trend analysis we did on the climate data [39]. Next, SWAT model was calibrated for each land cover reference map, and was used to a baseline model and models for the eight land cover-climate change scenarios shown in Table 1. Finally, results for each scenarios were compared to the baseline results, and then relative impacts of land cover change and climate change were quantified.
To assess the separate impacts of land cover and climate changes on hydrological processes, scenario models were developed using the "one-factor-at-a-time" approach in which one factor is changed at a time while fixing the other condition [4,16]. Such technique helps to separate the relative contributions of land cover and climate change impacts. On the other hand, the combined impacts of the land cover and climate changes was assessed from scenario models developed by changing both land cover and climate. To develop the scenarios, CHIRPS2 rainfall and ENACTS temperature data were split into three periods as shown in Table 1.  [3]. For this study, a statistical method recommended by Yin et al. [34], and described in equations 5 and 7 was adopted.
where ∆ : the combined impacts of land cover and climate changes (in %); , : simulation with changing both land cover and climate from the baseline reference years; , : baseline simulation; ∆ , : the separate impacts of land cover change; , , : simulation with the same climate with the baseline simulation while changing the land cover; ∆ , : the separate impacts of climate changes; , , : simulation with the same land cover with the baseline simulation while changing the climate reference year.
In summary, the separate and combined effects of the land cover and climate changes were quantified as follows.
i. Scenario 1 was considered baseline for the three periods while scenario 4 was used as baseline for the 2008 to 2018 period. ii.
The separate impacts of land cover changes were determined by comparing results of scenario 2 and scenario 1, scenario 8 and scenario 4, and scenario 7 and scenario 1. iii.
The separate impacts of climate change was determined by comparing scenarios 3 and 1, scenarios 6 and 4 and scenarios 5 and 1. iv.
The combined impacts of land cover and climate changes was determined using scenario 4 and 1, scenario 9 and 4, and scenario 9 and 1.  the basin has deep soil, permeable lithology, is vegetated and receives high rainfall. All these characteristics favor high infiltration and percolation thus resulting in substantial baseflow.     model performance classification. Moreover, the P-factor and R-factor values for all models are acceptable [62]. Overall, both the statistical measurements shown in Table 3 and the graphical results shown in Figure 2 indicate that streamflow predictions of the calibrated models are in good agreement with the observed streamflow. This indicates that the models can represent hydrological response of the Dhidhessa River basin and are suitable for further application.

Hydrologic responses to land cover and climate changes
The hydrological responses to land cover and climate changes for the 1982-1995 to 1995-2008, 1995-2008 to 2008-2018 and 1982-1995

Relative contributions of the separate and combined impacts of land cover and climate changes on hydrologic responses
Relative contributions of the separate and combined effects of land cover and climate changes on hydrological responses were calculated based on equations 5 to 7 (Table 4). Results show that the relative contributions vary among the hydrological variables and the study periods.
For example, climate change increased AET while land cover change decreased the variable. Land cover change increased surface runoff but decreased groundwater recharge. On the other hand, the effect of land cover and climate changes on other hydrologic responses vary among study periods.
Surface runoff was more sensitive to land cover change than to climate change. AET and streamflow were more sensitive to climate change than to land cover change. The relative importance of land cover and climate change on groundwater recharge vary among study periods.
The combined effects was more important than the separate effect on surface runoff and AET  Another interesting finding of this study is that the combined effects of land cover and

Discussion
This study showed that SWAT model is suitable for studying the impacts of land cover and climate changes on hydrologic responses for the Dhidhessa River basin. The performance of the model was very good for the basin based on Moriasi et al. [62] model performance classification.
The high baseflow ratio for the three simulated scenarios and observed streamflow showed that basflow is the major contributors for the Dhidhessa River streamflow. This has also been reported by several previous studies [41,[59][60][61]. As previously described, the basin has deep soil, permeable lithology, is vegetated and receives high rainfall. All these characteristics favor high infiltration and percolation thus resulting in substantial baseflow.
The changes in hydrologic responses observed in this study is attributed to the land cover and climate changes of the Dhidhessa River basin. Substantial spatiotemporal land cover and climate changes were reported during the last three decades in the Dhidhessa River basin [39,40].
For example, agriculture, settlement, and water bodies increased while bush and shrubland, and impacts on the environment and agricultural production [66]. Likewise, the wetting and warming climatic condition observed in the Dhidhessa River basin will have a positive implications for agriculture and water resources availability. However, the combined effects of climate and land cover changes may diffuse the positive impacts, particularly on water resources. Likewise, climate changes was reported in the Dhidhessa River basin during the last three decades [39]. Rainfall generally increased but statistically significant increasing trends was reported in some parts of the Dhidhessa River basin. On the other hand, statistically significant increasing trends of both minimum and maximum temperature were reported during the last three decades in the river basin.
Generally, wetting and warming of Ddhidhessa River basin was reported for the Dhidhessa River basin during the last three decades [40].
We think surface runoff increased in the Dhidhessa River basin primarily because of increase in agricultural land (by 265%) at the expense of declines in bush and shrub by 47% and forestland by 20%. In addition, increasing rainfall in the basin during the analysis period has also contributed for the increased surface runoff. Losses of natural vegetation increase surface runoff by reducing infiltration and time of concentration. Agricultural activities could disturb soil structure and reduce soil infiltration capacity resulting in high volume of runoff production [67,68].
Thus, we believe land cover changes coupled with increased rainfall has led to increased surface runoff in the Dhidhessa River basin. Decreased in infiltration also explains the decline in ground water recharge during the study period. Vegetation cover influences groundwater recharge as it The findings of this study is consistent with previous studies [22,70,71]. For example, Chimdessa et al. [70] reported increasing streamflow in the Dhidhessa River basin. Berihun et al. [71] reported increased surface runoff in response to land cover change in the upper Blue Nile basin.
The findings of this study on the relative contribution of land cover and climate changes are in agreement with previous studies reported in the upper Awash basin [48] and in Heihe River basin [22]. Several studies have reported that surface runoff is more sensitive to land cover change than to climate change [4,34,71]. In contrary, there are studies that have argued that surface runoff is more sensitive to climate change than to the land cover change [22,72]. Similarly, a study of small watershed located in the upper Blue Nile basin showed that AET is more sensitive to land cover change than to climate change [71]. This study also showed that streamflow in the Dhidhessa River basin is more sensitive to climate change than to land cover change. This finding is consistent with several previous studies [22,70,73], and disagrees with findings of others [4,34]. This implies that relative sensitivity of streamflow to climate change and land cover could vary from basin to basin. This underscores the challenges to generalize relative impacts of the changes and the need to analyze impacts on basins and various hydrologic variables on a case-by-case basis.
The combined impacts of land cover and climate changes on hydrologic responses reported in this study are consistent with findings of Tamm et al. [74] and Berihun et al. [71]. Studies showed that the offsetting and synergetic effects of land cover and climate changes on hydrological variables depends on the severity of the changes and on basin characteristics [22,75,76]. The result also showed that surface runoff was more sensitive to land cover and climate changes during the last three decades in the Dhidhessa River basin.
Overall, the hydrological responses of the Dhidhessa River basin has undergone significant changes during the last three decades. The changes in hydrologic regimes are indicators of changes in watershed conditions [77]. Increase in surface runoff and decrease in groundwater recharge in the Dhidhessa River basin indicates deterioration of land resources in the basin. Increased surface runoff aggravates soil erosion, reservoir siltation, and deteriorates water quality. As such, we think changes in land cover and climate have adverse impacts on agricultural production and water resource availability in the Dhidhessa River basin. The problem could be more severe as the trend continuous [39,40].
If carefully planned and implemented, integrated watershed management and climate adaptation strategies reverse the negative impacts. Soil and water conservation practices and afforestation could play key roles in reversing the impacts. Conservation measures reduce surface runoff by slowing down runoff and by enhancing infiltration and percolation. They also maintain or improve soil fertility and create favorable condition for the soil to support vegetation growth and hold water. Finally, afforestation could modify the climatic condition of the area and the surroundings. Generally, afforestation and soil and water conservation practices could combat climate change and watershed degradation, and improve agricultural production and water resource availability.

Conclusions
An integrated study that assesses the spatiotemporal land cover and climate changes, and their separate and combined impacts on various hydrological responses is important to plan land resources management strategies. This study examined the impacts of the substantial land cover and climate changes reported for the Dhidhessa River basin during the last three decades. The study result completed for the Dhidhessa River basin showed that the majority of the land cover changes involves conversion from natural vegetation to agriculture land. This type of land cover conversion has significant impacts on hydrologic responses of a basin. Similarly, previous study result showed that rainfall and temperature were generally increased in the basin during the past three decades. Overall, the significant land cover and climate changes reported for the Dhidhessa River basin during the last three decades, and their complex interactions have resulted in a significant changes in hydrological responses.
SWAT was calibrated for all land cover conditions for this study. The calibration results showed that the model is suitable to examine the separate and combined effects of the land cover and climate changes on hydrologic responses for the Dhidhessa River basin. Our study shows that in response to the observed land cover and climate changes in the basin, surface runoff, AET and streamflow increased by 40%, 5% and 4% whereas groundwater recharge declined by 3%, respectively during the last three decades. We think surface runoff increase was attributed to increase in rainfall and conversion of natural vegetation to agricultural land. Conversely, decline in groundwater recharge could be due to reduced infiltration and baseflow in response to conversion of vegetation cover to agriculture. Increased AET despite deforestation could be attributed to increased waterbody and crops in the basin. Streamflow increased during wet season but decreased during dry season dry season during the 1982-1995 to 2008-2018 period. These changes affect both agricultural activities and water resources availability negatively.
The relative contribution of land cover and climate changes on hydrological responses is not consistent among the hydrologic variables. AET and streamflow were more sensitive to climate change than to land cover change whereas runoff was more sensitive to land cover change than to Generally, in response to significant land cover and climate changes reported for the Dhidhessa River basin during the last three decades, significant hydrologic response changes were predicted. The simulated hydrologic responses have negative impacts on agricultural production and water resources availability. Future hydrologic regime could get worse if the land cover and climate changes trends continue [39]. Integrated watershed management strategies could reverse these problems. Funding: This research received no external funding.