Water resources are influenced by changes in land use and land cover (LULC), such as industrialization, urbanization, forestry, and agriculture. This study has aimed to analyze past and predicted LULC dynamics and their impacts on the components of the water balance in the Central Rift Valley (CRV) sub-basins in Ethiopia. The Soil and Water Assessment Tool (SWAT) and the Land Change Modeler (LCM) were employed to evaluate the impacts of past and future LULC dynamics in the Ketar, Meki and Shalla sub-basins. The SWAT models were calibrated with flow data from 1990 to 2001 and were validated with flows from 2004 to 2010, using SWAT-CUP in the SUFI-2 algorithm. LCM with Multi-Layer Perceptron (MLP) neural network method for land transition scenario analysis and a Markov Chain method for predictions, as well as SWAT models with fixing-changing methods for simulations, were used to evaluate the condition of hydrological processes under the influence of changes in LULC. The analyses resulted in an annual runoff variation from − 20.2 to 32.3%, water yield from − 10.9 to 13.3%, and evapotranspiration from − 4.4 to 14.4% in the sub-basins, due to changes in LULC. Integrated land use planning is recommended for the management of water resources.
Hinweise
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Introduction
Changes in land use and land cover can modify the carbon cycle, surface and sub-surface water conditions and biodiversity, among others, on local, regional, and global scales (Chemura et al. 2020). These changes must be managed appropriately for sustainable development of the water resources in a region (Gashaw et al. 2018). Appropriate management is based on understanding the key factors that affect the water resources of a given region. LULC is one of the top factors affecting the conditions of water resources in a given region (Wagner et al. 2023). To manage change in LULC and water resources effectively in a basin, it is important to assess the historical LULC dynamics, and also the potential future LULC dynamics (Leta et al. 2021a). Changes in LULC can be caused by increasing population and economic growth, which put pressure on the ecosystems that provide water and water-related services (Deche et al. 2023; Nigusie and Dananto 2021). Settlement and agricultural land expansion, deforestation, pollution, etc. are the most common drivers directly affecting these ecosystems (Elias et al. 2019).
Changes in land use take place very rapidly these days, due to the rapid increase in demand for resources. In countries with limited water resources, rapid changes in land use aggravate the problems of water scarcity (Kundu et al. 2017). Population growth is always followed by an increase in demand for land and other resources. In the process, loads that are beyond the carrying capacity of the land, in combination with unsuitable management, will lead to land degradation. For example, in developing regions, bigger families and smaller land holdings will lead to deforestation in rural areas. Elias et al. (2019) have stated that the reasons for possible deforestation are multi-faceted. However, the main reason is that there is no proper land use policy in place. In addition, changes in LULC due to human intervention will affect the integrity of natural resources. Water, energy, land, and food are naturally linked through complex networks of direct and indirect effects in the ecosystems. The changes will have a significant influence on the quantity and/or quality of stream flows (Nigusie and Dananto 2021). Land uses can change naturally and can be changed by human interventions due to population growth (Hasan et al. 2020; Shrestha et al. 2018; Tang 2020; Tekleab and Kassew 2019; Zhang et al. 2020).
Anzeige
Land cover dynamics has become a concern of the twenty-first century, with significant implications for human survival. It is, therefore, always necessary to have a suitable policy ready for implementation. In order to be ready with a policy, it is necessary to understand the trends in land use change. Changes in land use have potentially huge impacts on water resources (Ayalew et al. 2022; Shumet and Mengistu 2016; Yang et al. 2023), but quantifying these impacts is still a challenging problem in hydrology. Even when there is little or no human intervention, hydrological systems incorporate variations in the flow of water, solutes, sediments, and energy (Truneh et al. 2023). Understanding the impacts of land use, therefore, necessitates integrated scientific approaches. Nowadays, direct measurements, remote sensing and hydrological modeling studies are tools that shed the light by which the impacts of changes in land use on water resources can be assessed and quantified (Baker and Miller 2013; Stonestrom et al. 2009; Sulamo et al. 2021). In this study, apart from other methods applied so far, we employed integrated use of hydrological model, SWAT and land use change modeler, LCM, to evaluate the impacts of LULC changes on hydrological components. Future potential impacts of LULC changes were also estimated based on the trends in the past with LCM.
Several studies conducted in many parts of the Ethiopian Rift Valley region (CRV) and beyond have investigated the expansion of agricultural land at the expense of natural vegetation, forest, shrubs and grass land (Belihu et al. 2020; Legesse et al. 2003; Sulamo et al. 2021; Tekleab and Kassew 2019; Wolde et al. 2021; Yifru et al. 2021). These studies have revealed that the water resources in the lakes are highly sensitive to changes in LULC. According to the studies, significant changes have been observed in the hydrology of the CRV lakes in Ethiopia over the past 4 decades. For example, Lake Abiyata declined in size in past years (Ayenew 2007). The volume of Lake Ziway has also decreased due to overexploitation and reduced recharging. The stream flows to the lake have been reduced because of changes in land use in the upper catchments (Desta et al. 2015). In addition, the development of large-scale irrigation, industrial abstraction from the CRV lakes, the introduction of intensive agricultural practices, together with poor water management practices, have modified the hydrology of most lakes in the region (Seyoum et al. 2015). However, little consideration has been given to the interaction between the water balance components in the catchment and the changes in LULC. Thus, this study aimed to address this gap.
It is, therefore, essential to evaluate the impacts of changes in LULC due to natural and human intervention on the components of the water balance of the area (Kalogiannidis et al. 2023; Schilling et al. 2008). Quantifying the impacts of land use changes in the past on the water cycle component will help to identify and rank the critical LULC elements that have a significant input into altering the water balance environment (Chauhan et al. 2020). Forecasting possible future land use changes based on past changes will also help to indicate how the possible future impact of LULC changes can be managed (Chauhan et al. 2020; Schilling et al. 2008; Tayebzadeh Moghadam et al. 2021). In past studies, greater emphasis has been laid on the hydrology of a lake and on the impacts of changes in climate in the CRV sub-basins (Gadissa et al. 2019; Musie et al. 2020; Truneh et al. 2023; Ulsido et al. 2013). The work presented here is, therefore, aimed at analyzing the impacts of past and future potential land use changes on the components of the water balance in the CRV sub-basins in Ethiopia. We analyze the extent to which changes in land use can modify the water balance components in the region. This study uses the SWAT model with different static land use maps for a given timeframe, while the climatic and other parameters are fixed. The target is to assess only the impacts of LULC together with LCM embedded in TerrSet 2020 software (Chauhan et al. 2020; Leta et al. 2021a; Saddique et al. 2020). The approach followed integrated modeling approaches to address the land use dynamics and its impacts on the components of the water balance (Bucha et al. 2024; Yifru et al. 2021).
Materials and methods
Description of the study area
CRV is located in the East Africa region, in the upper head of the rift valley basin in Ethiopia. Geographically, the basin area extends from 38°15′00″ E and 39°27′0″ E to 7°00′0″ N and 8°30′00″ N and covers an area of 15,301.96 km2 (Fig. 1). It is part of the main Ethiopian rift and comprises a significant part of the great African rift valley system that stretches from the Red Sea to Mozambique, passing through Ethiopia, Kenya and Tanzania. In Ethiopia, the rift is divided into three subsystems: Chew Bahir (Lake Stephanie), CRV, and the Afar triangle (Elias et al. 2019). CRV comprises the major Ziway, Langano, Abiyata, and Shalla lakes.
Fig. 1
Major river basins in Ethiopia (bottom left), the location of CRV in Ethiopia (upper left), and CRV sub-basins (right) adopted from Truneh et al. 2023
×
Anzeige
Data sources
To build model input files, SWAT-2012 requires a digital elevation model (DEM), land cover and land use information, soils, and basic climate data. SWAT subdivides a watershed into Hydrological Response Units (HRU) and treats an HRU as a homogeneous block of land use, management techniques and soil properties, and then quantifies the relative impact of vegetation, management, soil, land use and climate changes within each HRU (Arnold et al. 2011). Subdividing the watershed allows users to analyze hydrological processes in different sub-basins within a larger watershed and to understand the impacts of regional land use management. Accordingly, CRV was subdivided into sub-basins based on their outlet points, as indicated in Fig. 2, with their watershed boundaries indicated by a black line. The outlet points at each of the sub-basin were taken at the hydrological gauging stations of the respective sub-basin.
Fig. 2
Sub-basins of the CRV basin, stream networks and their outlets (hydrological stations) where the impacts of changes in LULC for each sub-basin were analyzed (Truneh et al. 2023)
×
Basic climate data such as the daily observed precipitation, the maximum and minimum temperature, wind speed, hours of sunshine, and relative humidity data from six stations in the CRV region were collected from the National Meteorological Agency (NMA) of Ethiopia. To calibrate and validate the SWAT model, hydrological data of the river discharge at the outlet points of the sub-basins were obtained from the Ministry of Water and Energy (MW&E) of Ethiopia.
Land cover and land use maps were obtained from the Ethiopian Geospatial Information Institute (GSII). These maps presented the land cover and land use based on Landsat TM and ETM + and Sentinel satellite imagery from 2003, 2008, and 2013 and 2020. The maps were checked for accuracy with ground truth and converted from the images to LULC map with kappa coefficient values above 83% as indicated in the source document from the institute, GSII. Major LULC such as rangeland and shrubs, forest, agricultural land, urban areas and settlements, and water were mapped at each time step. Land use data adjusted to ground truth points for years 2003, 2008, 2013 and 2020 were thus used to analyze the impacts of changes in land use on the components of the water balance in the region. The objective behind selecting these land use years were to evaluate the impacts of past 20 years land use changes and to predict the potential changes in the next 30 years. Besides the data were gathered and processed for LULC changes by the institute every 5 years and in the years indicated for the past periods.
Soil data maps for determining soil parameters such as texture, hydrological soil group (HSG) and available water content for soils, as needed to run SWAT, were obtained from the Ministry of Agriculture and Natural Resources (MANR). DEM data were obtained from the Oromia Bureau of Agriculture and Natural Resources (OBANR). The soil hydro-physical properties determine and define the existence and the quantity of each component of the water balance (Báťková et al. 2020). Soil hydraulic characteristics, especially the soil water retention curve and hydraulic conductivity, are essential for many agricultural, environmental, and engineering applications (Matula et al. 2007). The soil physical properties and the area coverage of each soil type were classified according to the SWAT classification standards. The soil type and its distribution in the sub-basins are indicated in Fig. 3. The soil type classification is based on the SWAT classification codes.
Fig. 3
Soil type distributions in the sub-basins (Truneh et al. 2023)
×
Sub-basin delineation and land use reclassification
The CRV region was divided into sub-basins based on their discharge outlet (monitoring) stations. Three of the major sub-basins—Ketar, Meki and Shalla—(Fig. 2), were delineated with SWAT HRU tools and Arc-GIS to analyze and quantify the hydrological impacts of past LULC changes as well as future potential changes with the use of a calibrated SWAT model. The major LULCs of the sub-basins were re-classified into Agriculture, Forest, Rangeland, Water, and Settlements, based on the similarities of their hydrological response (Sawicz et al. 2011; Wagener et al. 2007). The classified LULC categories are presented in Table 1. Figure 4 also indicates the re-classified LULC maps of the CRV basin for each time step in the past.
Table 1
Description of the reclassified LULC classes of the CRV basin
Urban, residential areas, roads, industrial zones, barren land, dry sand, and dry stream channels, bed rocks, dry land mass and lava outflows
URHD
5
Water
Rivers, lakes, ponds, and wetlands
WATR
The detail area coverages of each land use categories are indicated in Tables 6 and 7 of “Results” section
Fig. 4
Map of past re-classified LULC maps of the CRV sub-basins for each time step
×
SWAT model setup, calibration, validation, and performance evaluation
SWAT is a versatile model of a watershed, which is capable of simulating a range of processes from rainfall-runoff processes to many other important parameters (Stonestrom et al. 2009). The operations of the model involve soil characteristics, hydrology, weather, land management, plant growth, pesticides, and nutrients in its subcomponents (Abbaspour et al. 2015). SWAT can also be applied on a wide-scale watershed with high efficiency of computation, and is very appropriate for analyzing the impacts of land use changes (Arnold et al. 2011; Kundu et al. 2017; Stonestrom et al. 2009).
The model analyzes the water balance of a basin based on the basic water balance equations, as stated in Arnold et al. 2011, and is defined as
where SWt is the soil water content (mm) at time t, SW0 is the initial soil water content (mm), t is the simulation period (days), Rdayi is the amount of precipitation on the i-th day (mm), Qsurfi is the amount of surface runoff on the i-th day (mm), Eai is the amount of evapotranspiration on the i-th day (mm), Wseepi is the amount of water entering the vadose zone from the soil profile on the i-th day (mm), and Qgwi is the amount of base flow on the i-th day (mm).
One of the critical parameters that are evaluated for sustainable water resource management of the study area is the water yield. The water yield is the aggregate sum of the water leaving the HRU and entering the principal channel during a time step (Arnold et al. 2011). In this study, the SWAT models were, therefore, set, calibrated and validated to analyze the impacts of LULC changes on the components of the water balance separately for the sub-basins, based on their outlet points:
where Wyld is the water yield (mm), Qsur is the surface runoff (mm), Qlat is the contribution of the lateral flow to the stream (mm), Qgw is the contribution of groundwater to the streamflow (mm), and Tloss is the transmission losses (mm) from the tributary in the HRU by means of transmission through the bed.
Calibration and validation
The calibration was performed with the flow data from 1990 to 2001 and was validated with the data from 2004 to 2010. The data from the years 2002 and 2003 were jumped to offer time span for the data used in the calibrations and validations. The basic LULC used during calibration was the land use year in 2003. It is calibrated and validated using monthly monitored stream flows from the outlets of the Ketar, Meki and Jidu (Shalla) rivers. Calibration and validation of the SWAT models were performed with the use of SWAT-CUP, a calibration uncertainty program for SWAT, with the SUFI-2 algorithm. The models were set to run for the baseline periods from 1984 to 2010 for each of the sub-basins.
Model accuracy and performance evaluation for SWAT
The accuracy and the performance of the model were evaluated and checked before it was used for simulation in the sites. This sets the model to better resemble the sites. In this work, the SWAT model was calibrated, validated and its performances were evaluated against the coefficient of determination (R2), Nash–Sutcliffe efficiency (NSE) and Percentage of bias (PBIAS), using the monitored stream flows from the outlets of the Ketar, Meki and Jidu river gauging stations. The models were evaluated by a performance scale set according to Moriasi et al. 2007 (Table 2).
Table 2
The model performance scale.
Source: Das et al. (2019), Kundu et al. (2017) and Moriasi et al. (2007)
Performance rating
PBIAS (%)
NSE
R2
Very good
PBIAS < ± 10
0.75 ≤ NSE ≤ 1.00
0.75 ≤ R2 ≤ 1.00
Good
± 10 < PBIAS < ± 15
0.65 < NSE < 0.75
0.60 < R2 < 0.75
Satisfactory
± 15 < PBIAS < ± 25
0.50 < NSE < 0.65
0.50 < R2 < 0.60
Unsatisfactory
PBIAS > ± 25
NSE < 0.50
R2 < 0.50
The outputs of the performance for our models for each of the sub-basin were presented in “Results” section.
Sensitive parameters in the models
The sensitive parameters from the land use, soil, land slope, basin management and groundwater categories were identified during calibration, and their values were adjusted accordingly. Sensitivity analysis decides which variables should be adjusted to obtain better results (Abbaspour et al. 2015; Khalilian and Shahvari 2019; Kundu et al. 2017). A set of parameters is selected for the sensitivity analysis, as given by various researchers and from the documents of SWAT and during calibration while setting up the models for the sites (Du et al. 2013; Kundu et al. 2017; Moreira et al. 2018). The sensitive parameters identified during calibration were tabulated (Table 3) with their lower and upper limit values. The calibration and validation help to adjust the values of these sensitive parameters.
Table 3
The most sensitive parameters that were identified in the sub-basins
No.
Name
Description
Lower bound
Upper bound
Process
1
CN2
Soil conservation service runoff curve number for moisture condition II
35
98
Management/runoff
2
ESCO
Soil evaporation compensation factor
0
1
Evaporation
3
SOL_Z
Soil depth
0
3500
Soil
4
SOL_AWC
Available water capacity of the soil layer (mm/mm soil)
0
1
Soil
5
GWQ_MN
Threshold depth of water in the shallow aquifer required for return flow to occur (mm)
0
5000
Groundwater
6
REVAPMN
Threshold depth of water in the shallow aquifer for revap to occur (mm)
0
500
Groundwater
7
EPCO
Plant evaporation compensation factor
0
1
Evaporation
8
GW REVAP
Groundwater-revap coefficient
0.02
0.2
Groundwater
9
SURLAG
Surface runoff lag coefficient
0.05
24
Run-off
10
ALPHA_BF
Baseflow alpha factor (days)
0
1
Groundwater
11
SOL_K
Soil conductivity (mm/h)
0
2000
Soil
12
GW_DELAY
Groundwater delay (days)
0
500
Groundwater
13
BIOMIX
Biological mixing efficiency
0
1
Management
14
CH_K2
Hydraulic conductivity in main channel (mm/h)
− 0.01
500
Channel
15
RCHRG_DP
Deep aquifer percolation fraction
0
1
Ground water
16
HRU_SLP
Average slope steepness (m m−1)
0
1
Geomorphology
Past land use and land cover change analyses, and future prediction methods
Past land use and land cover change analysis
The quantification and evaluation of the changes in each LULC category were analyzed with the SWAT model and with LCM embedded in TerrSet 2020. In the SWAT model, the area coverage of each LULC category in each sub-basin was calculated with the HRU analysis tool. The HRU tool calculates the static area coverage of the supplied land use map as categorized for a defined timeframe, e.g., year 2013 or year 2020.
LCM will calculate the percentage of gains and/or losses in the area of each land use category between two-time steps. The change evaluation will be performed for the whole areas of the two supplied maps, which have been georeferenced exactly to indicate the same place but with different time periods. Thus, the dynamics of each major LULC category of the sub-basin between two defined time periods was analyzed.
Land use and land cover change prediction
The predicted land uses were performed by LCM in TerrSet2020, a geospatial monitoring and modeling system. TerrSet 2020 is an integrated model developed in Clark University Lab, USA, for geospatial monitoring, evaluation, and modeling. In the analysis, LCM determines the dynamics of LULC change, how much change in land cover took place between earlier and later LULC images, and then calculates the relative amounts of transitions of the variable. LCM is used to predict and project changes using multiple land cover categories. It uses the Cellular Automata-Markov Chain (CA_MC), which is a stochastic modeling method used to simulate the future LULC change over time from past changes (“Land Change Modeler in TerrSet” 2022; Leta et al. 2021b). It predicts the spatial structure of various LULC categories and scenarios based on the Transition Potential Matric (TPM).
LCM has three functional units: the change analysis, transition potentials and change prediction categories, together with many other sub-functions. The change analysis sub-function analyzes the trend of the spatial changes and creates maps. The transition potential subsection helps to simulate the future potential transition scenarios. The change prediction subsection predicts the land cover and validates the predicted land cover, based on the transition scenarios developed under the transition potential. LCM uses various approaches to produce maps of the transition potential. In this analysis, multi-layer perceptron (MLP) approaches were selected. This approach is more flexible and more dynamic than the others when multiple transition types are modeled. In the prediction section, the Markov Chain was used to generate transition probability matrices between LULC classes. LCM predicts the possible land use that would occur in the future based on past land use changes according to the transition potential scenarios and the probability matrices.
LCM validation
Validation is a process for assessing the quality of the predicted LULC map against a reference map (Leta et al. 2021a). The Landsat images for 2008 and 2013 were utilized after these categories were harmonized to simulate the 2020 LULC image. A comparison of the simulated LULC image with the actual map was developed. The LULCs of years 2008 and 2013 were provided to validate LCM, and the model was validated by simulating the recent LULC map of 2020. The validation process in LCM involves cross-tabulation in a three-way comparison between the earlier land cover map (2008), the predicted land cover map (2020), and the current map (2020). The validation of the LCM model was used to make a statistical assessment of the quality of the predicted 2020 LULC image against the 2020 reference image.
Methods for analyzing the impacts of the change in LULC on the components of the water balance (the fixing-changing method)
For water resources management, it is important to understanding the response of the watershed to changes in LULC. Classified LULC maps (2003, 2008, 2013 and 2020) and predicted LULC maps (2030, 2040 and 2050) were, therefore, used to reveal the hydrological impacts of LULC changes. The LULC maps were used separately, while all other SWAT inputs were kept similar. The “fixing-changing method” is a method for changing LULC maps while keeping other inputs, e.g., weather data files, fixed in the calibrated SWAT model, in order to quantify the sole impacts of the LULC. This method has been employed by many researchers in the past for land use change analyses (Chauhan et al. 2020; Gashaw et al. 2018; Woldesenbet et al. 2017). However, in this study, the impacts on water cycle components were identified apart from other studies. The changes in the components of the water balance were analyzed in relation to the water balance outputs of the land use data that was used during the SWAT calibration and validations, i.e., LULC data for year 2003. A general flow chart of the methods is presented in Fig. 5.
Fig. 5
Graphical illustration of the methodologies
×
Results and discussion
Results for SWAT model calibration, validation, and performance evaluation
The calibration results indicate good agreement between the simulated and observed discharges in the sub-basins. The results for simulated and observed discharges in the sub-basins were evaluated against R2, NSE and PBIAS during calibration and validation. The values in the Ketar sub-basin are in good agreement with R2 > 0.6, NSE > 0.5 and PBIAS < = ± 25, Fig. 6. Similarly, the results showed that the simulated and observed monthly discharges were in a good agreement during calibration and validation for the Meki and Shalla sub-basins and presented in Truneh et al. (2023).
Fig. 6
Calibration (a) and validation (b) results of the models for CRV sub-basins
×
Sensitive parameters
Various hydrological parameters built into SWAT that have been found to be important for hydrological modeling were selected and adjusted accordingly. The most sensitive parameters that were identified represent the land use and land cover, soil characteristics and groundwater categories. The following parameters were identified as highly sensitive in the Ketar sub-basin: EPCO, RCHRG_DP, SOL_K, GW_DELAY, CN2, REVAPMIN, and SURLAG. Similarly, ESCO, REVAPMIN, GWQMN, HRU_SLP and GW-DEALY were very highly sensitive parameters in the Meki sub-basin, and ESCO, CH_K2, SOL_K, GWQMN were very highly sensitive in the Shalla sub-basin based on t-stat values (Truneh et al. 2023). The description, values and ranges of values of the parameters are presented in Tables 3, 4 and 5.
Table 4
Sensitivity or Mean of Index; I, values of the selected parameters for the sub-basins, according to their “t-stat” results as per the scale indicated in Table 6 (parameters with very high sensitivities are in bold)
Parameter**
Ketar
Meki
Shalla
t-stat value
Sensitivity
t-stat value
Sensitivity
t-stat value
Sensitivity
R__CN2.mgt
1.408
Very high
− 0.394
Negligible
− 0.111
Negligible
V__ALPHA_BF.gw
0.046
Low
− 0.997
Negligible
− 1.643
Negligible
A__GW_DELAY.gw
1.206
Very high
1.951
Very high
− 1.032
Negligible
A__GWQMN.gw
0.783
High
1.564
Very high
2.685
Very high
A__REVAPMN.gw
1.970
Very high
1.441
Very high
− 1.116
Negligible
A__GW_REVAP.gw
0.710
High
0.844
High
NI*
NI*
V__ESCO.bsn
0.905
High
1.181
Very high
1.739
Very high
V__EPCO.bsn
1.013
Very High
− 1.210
Negligible
− 1.513
Negligible
A__SURLAG.bsn
2.329
Very high
− 1.242
Negligible
0.744
High
R__SOL_AWC(..).sol
− 1.034
Negligible
− 3.957
Negligible
NI*
NI*
R__SOL_K(..).sol
1.202
Very high
− 1.417
Negligible
1.197
Very high
V__CH_K2.rte
− 0.551
Negligible
NI*
NI*
1.926
Very high
R__SOL_Z(..).sol
NI*
NI*
NI*
NI*
NI*
NI*
V__RCHRG_DP.gw
1.137
Very high
NI*
NI*
− 1.986
Negligible
R__HRU_SLP.hru
NI*
NI*
1.799
Very high
0.084
Low
R__BIOMIX.mgt
NI*
NI*
1.669
Very high
0.798
High
Table 5
Parameter sensitivity scale classes assigned in SWAT
(Source: Lenhart 2002)
Class
Mean of index (I)
Category of sensitivity
1
0 ≤ I ≤ 0.05
Small to negligible
2
0.05 ≤ I ≤ 0.2
Medium
3
0.2 ≤ I < 1
High
4
I ≥ 1
Very high
The results of past changes in LULC
Changes in LULC in past years were delineated in SWAT models, and they are presented in Table 6. In the past, most of the land in the CRV sub-basins was covered by agricultural land masses, followed by range lands, forest lands, settlements, and water bodies, in the order of their area coverages. Range lands were greatly reduced, and agricultural lands have increased in all the sub-basins at the expense of range land, and to some extent at the expense of forests and other land covers over time periods in the past (Table 6).
Table 6
Past LULC area percentage for each time step in the sub-basins
Years
Sub-basins
Total area in (ha)
Land use class cover in (%) out of the total area
Forest
Range land
Agriculture
Water
Settlement
2003
Ketar
346,885.95
3.77
28.39
65.56
0.24
2.04
Meki
193,582.58
7.01
26.1
63.98
0.88
2.03
Shalla
146,625.82
1.06
23
73.61
0.41
1.92
2008
Ketar
346,885.95
11.06
22.95
62.97
1.05
1.97
Meki
193,582.58
10.41
12.67
74.07
0.95
1.89
Shalla
146,625.82
0.43
20.56
78.22
0.32
0.46
2013
Ketar
346,885.95
18.22
6.43
74.39
0.18
0.77
Meki
193,582.58
17.71
8.51
72.68
0.61
0.49
Shalla
146,625.82
3.55
10.48
84.6
0
1.37
2020
Ketar
346,885.95
11.8
0.21
87.01
0.1
0.87
Meki
193,582.58
6.37
0.4
91.84
0.94
0.45
Shalla
146,625.82
0
1.55
97.68
0
0.77
In past years, the forest coverage in the Ketar sub-basin was 3.77% in 2003, 11.06% in 2008, 18.22% in 2013, but in 2020 the forest cover was only 11.8%. The change analyses indicate that in past years, the forest coverage had been increasing in this sub-basin, but in the more recent past, the coverage went down between 2013 and 2020, while the agricultural area coverage increased from 74.39 to 87.71% in the same period.
Almost all the range and bush lands in the sub-basins were changed to agriculture. The level of range and bush in the Ketar sub-basin was about 28.39% in 2003 and only 0.21% in 2020. In the Meki sub-basin, the level fell from about 26.1 to 0.4% between 2003 and 2020. Similarly, the coverage was reduced from 23 to 1.55% in the Shalla sub-basin. Details of the changes in past land use, and the changes in coverages in the sub-basins, are shown in Table 6.
The evaluations of the spatial and temporal changes between various LULC classes between 2003 and 2008, between 2008 and 20,013, between 2013 and 2020, and for the predicted time periods between 2030 and 2040 and between 2040 and 2050 were analyzed, and the results are shown in Fig. 7. The percentages of gains and losses were determined by LCM in the change analysis subsection for all the CRV maps. SWAT HRU analysis tools were employed to quantify the static coverage for each of the LULC categories in each sub-basin. The SWAT HRU analysis tools were employed. The results for past land uses are presented in Table 6 and the results for predicted land uses are shown in Table 7. Due to extreme agricultural practices, forest coverage had been devastated by 2020 in the Shalla sub-basin, where almost 98% of the sub-basin had been taken into intensive agricultural use. This severe change had also caused the water bodies to decline to almost zero in the sub-basin in 2020. Similar trends were also observed in the change analyses of the other sub-basins, see Table 6. An afforestation program and basin-wide land use planning and management interventions now need to be implemented in suitable places in the sub-basins in order to conserve water and other natural resources.
Fig. 7
Percentage of area gains and losses of past and predicted LULC categories within the time periods in the CRV sub-basins, as analyzed by LCM
Table 7
Predicted LULC area as a percentage for each time step in the sub-basins
Year
Sub-basins
Total area in (ha)
Land use class cover in (%) out of the total area
Forest
Range
Agriculture
Water
Settlement
2030
Ketar
346,885.95
18.24
6.40
74.40
0.18
0.78
Meki
193,582.58
8.12
1.80
82.93
4.19
2.91
Shalla
146,625.82
0.46
1.54
95.98
0.09
1.87
2040
Ketar
346,885.95
11.43
4.29
76.48
3.89
3.90
Meki
193,582.58
6.16
6.52
80.21
3.07
4.00
Shalla
146,625.82
7.74
0.41
88.97
1.49
1.33
2050
Ketar
346,885.95
16.78
2.69
72.95
3.49
4.08
Meki
193,582.58
10.81
4.37
76.99
3.77
4.02
Shalla
146,625.82
15.61
1.93
79.22
1.56
1.64
×
The decrease in settlement in the sub-basins does not indicate that the buildings and city expansion are reduced. However, during land reclassification based on hydrological response similarities, dry land masses, bare soil and sandy areas were categorized under settlement. The changes of dry land masses and bare soil in to forest, pasture land or agriculture significantly extrafolds the increase in urbanization and results in the reduction of the coverages of settlement in total in the subsequent years though there were increments in urbanizations.
The predicted LULC analysis results and a discussion
The future land uses were predicted on the basis of the probability matrix developed in the Markov Chain. The predicted LULC area coverages of the sub-basins are presented in Table 7. The LCM validation analysis indicates the results of its evaluation as hits, misses and false alarms. The hits are the exactly predicted values from the cross tabulations of the three images. The misses are cases where there were changes in the areas that the model was unable to predict, and false alarms refer to changes that are predicted but do not take place in reality. Although LCM does not incorporate the possible land use policy interventions during the prediction, the predicted land uses in the sub-basins have good validity. The maps of the predicted LULC for the future periods are presented in Fig. 8.
Fig. 8
LULC maps for each of the predicted time steps
×
In the predicted scenarios, the changes in land use vary from sub-basin to sub-basin. For example, the forest cover in the predicted years decreases from 18.24% in 2030 to 16.78% in 2050 in Ketar, but it increases from 0.46 to 15.61% in the Shalla sub-basin in the same period. The general trend is for agricultural and range land coverage to decrease in the time period between 2030 and 2050, but there will be increments in 2040. The settlement area will increase in the Ketar and Meki sub-basins, though it is predicted to decrease in Shalla. The decrease in settlement in Shalla is not necessarily related to a decrease in buildings. Nevertheless, it is assumed that factors such as dry land masses and sandy areas which were hydrologically categorized in the settlement group will be changed into agricultural areas and possibly into water bodies, as the coverage of water bodies is also predicted to increase.
Impacts of change in LULC on the components of the water balance in the sub-basins
The hydrological impacts of the LULC changes were evaluated for annual, seasonal and monthly distributions of the major water balance components: surface runoff, water yield and evapotranspiration in the region. The analyses were made separately for each sub-basin to better understand site-specific impacts. The climate factors were kept constant (fixed), while the land use maps were changed to evaluate the sole impacts of the LULC changes on the components of the water balance. Accordingly, the annual, seasonal, and monthly variations due to LULC changes for the Ketar, Meki and Shalla sub-basins will be discussed separately.
Ketar sub-basin
Surface runoff
The change in surface runoff in this sub-basin varies on an annual average from − 4.2 to 4.39% due to changes in land use in the course of the time periods between 2003 and 2050, in relation to the base year land use (LULC 2003) simulated values. The greatest reduction in runoff occurred in 2008, and the greatest increment was in 2020. One of the reasons for the reduction in surface runoff was the increase in forest cover from 3.77 to 11.06% over the years from 2003 to 2008. Forests usually increase retention and interception, and therefore, the runoff decreases significantly. Forest coverage was also on an increasing trend even in 2013, but due to the increase in agriculture and the reduction in range lands, the rate at which surface runoff occurred increased up to 2020. The monthly distribution of the runoff indicates that the changes in land use had an impact on the surface runoff. The surface runoff monthly distributions are presented in Fig. 9. The zigzag line along the land use years indicates the effects of changes in land use on the runoff. The runoff and the expansion in agricultural land and in urban areas are proved to be positively correlated. The results show that as agricultural land increases, the rate of runoff for the area also increases. Mainly, there has been an increase in agricultural land in the basin, at the expense of range land and forest cover, which counteract runoff enhancement.
Fig. 9
Monthly average distributions of surface runoff, water yield and ET across the land use change years, and the percentage of change in relation to the base year land use land cover (B. LULC 2003) in the Ketar sub-basin
×
Water yield
The annual change in water yield due to changes in land use has varied on an average from − 0.78 to 1.03% in the Ketar sub-basin over the past years and will vary in the predicted years. However, the amount of change in water yield in the sub-basin varies from season to season and even differs on a monthly basis, Fig. 9. Water yield is the main component of the water balance in this sub-basin in relation to the other major subcomponents, but the variation due to changes in land use is relatively small in percentage terms, Fig. 9. Nevertheless, a huge volume of water is affected, although it may seem small as a percentage of the total water balance. For example, 1% of the annual water yield in the Ketar sub-basin was 4.8 mm per unit area. When multiplied by the total area of the Ketar sub-basin, this amounts to more than 16 million m3 of water, which is a huge amount. Land use change management will, therefore, have a vital role in improving the available water yield in the sub-basin.
Evapotranspiration
Like other water balance components, ET also shows strong variations connected with changes in land use, and the monthly distributions of ET also vary, Fig. 9. For example, the ET varies on an annual average from − 2.08 to 5.36% due to changes in land use from the base simulation over the analyzed periods, 2003–2050. ET has a strong correlation with changes in land use and land cover in this sub-basin. Land management will, therefore, help to improve the availability of water resources and protection against losses through evaporation.
The variabilities in ET and runoff were stronger than the variability in water yield, as indicated in Fig. 9 as a percentage of the change. Surface runoff and ET are, therefore, greatly affected by changes in land use. This is highly related to changes in forest cover in the sub-basin and is, therefore, the critical element of the LULC in the sub-basin. Improvements in forest cover will, therefore, favor the availability of water resources.
Meki sub-basin
Surface runoff
The analyses of the model indicated that the annual average variation in surface runoff (Q) in the Meki sub-basin is from 5.15 to 25.37% for the years from 2003 to 2050, as indicated in Fig. 10. The annual sum of the surface runoff was 34 mm per square meter in 2003, while it was 36.97 mm per square meter in 2020. These changes were mainly due to the changes in LULC in the sub-basin, as other factors such as climate and management were kept constant in the model. In the predicted LULC scenarios, the annual surface runoff in the sub-basin will rise to 39.88 mm per unit area in 2050. This indicates that a huge amount of water will become additional runoff in the coming 30 years in the sub-basin, i.e., about 5.88 mm per unit square meter, because of the predicted LULC changes. This is similar to the findings of Musie et al. (2020), which also indicate that the land use change scenarios in the sub-basin will lead to an increase in surface runoff in the future (Musie et al. 2020). Water harvesting to store the future excess runoff in the sub-basin is, therefore, crucial for improving the water availability index of the sub-basin for use during peaks in demand.
Fig. 10
Monthly average distributions of surface runoff, water yield and ET during the land use change years, and the change percentage in relation to the base year land use land cover (B. LULC 2003) in the Meki sub-basin
×
Water yield
The variability in water yields due to the sole impacts of changes in land use on the annual average ranges from − 0.93 to 3.27% in this sub-basin. Water yield (WY) is the second most abundant water balance component in the Meki sub-basin. The annual average values of the simulated water yields range from 19.62 to 20.51 mm per unit area due to changes in land use. These variabilities in water yield due to the LULC dynamics will have their own effect on water-use planning and management in the sub-basin. Water yield enhancement strategies based on the sensitivities of the water balance to changes in LULC are absolutely essential.
Evapotranspiration
ET was a major component of the water balance and showed an average annual variation from − 4.43 to 7.39% from the simulation outputs for the base LULC years. The lowest annual ET recorded in this sub-basin was in 2013. In 2013, forest coverage was high. However, due to the significant reduction in open water bodies in the sub-basin from an area coverage of 0.95 to 0.61%, as indicated in Table 6, a reduction in ET was observed. This reduction in area coverage of water bodies has reduced ET significantly and surpasses the rate of ET increments from the forest area, as the evaporation from open water bodies is obviously high. The net balance indicates a reduction in ET in the periods between 2008 and 2013. All of these are sole impacts of changes in LULC in the sub-basin. Investigated land use planning and use according to water balance sensitivities to land use changes will help to improve water availabilities.
Shalla sub-basin
Surface runoff
As in the other sub-basins, changes in LULC have an impact on surface runoff in the Shalla sub-basin. The annual average simulated surface runoff varied from − 20 to 32.07% from the base land use year annual average simulated values. The surface runoff was smaller in relation to the water yield and in relation to ET, as can be observed from Fig. 11. In this sub-basin, the averaged sum of surface runoff ranges from 47.61 to 53.44 mm for different land use years. The total annual surface runoff is about 6.67% of the total annual rainfall in the sub-basin, which is a relatively small amount. The changes in LULC have been mainly from range land to agriculture. These changes have affected the runoff conditions in the sub-basin.
Fig. 11
Monthly average distributions of surface runoff, water yield and ET in the land use change years, and the percentage change in relation to the base year land use land cover (B. LULC 2003) in the Shalla sub-basin
×
Water yield
Water yield is a catchment water production capability which is naturally highly related to land use and land cover conditions. In the Shalla sub-basin, the monthly distribution of water yields has varied significantly, although the annual average variation in water yield from the base year annual average is from − 10.38 to 10.49%. As the impact of evapotranspiration is higher, afforestation alone may not be able to improve the generation capacity of the sub-basin. It is, therefore, of crucial importance to use investigated basin management to improve the water yield of the catchment. Integrated water resource management that incorporates all possible management factors will lead to improvements.
Evapotranspiration
ET is the main component of the water balance in the sub-basin. In all the land use years, the simulated ET annual average values increased by as much as 15.66%. This increment is due to changes in LULC in the sub-basin in past years, and due to the combined change effects of the LULC categories in the predicted future years. Overall catchment management and integrated land use planning will, therefore, improve the water resource availability of the sub-basin and will lead to a reduction in evaporation losses.
Overall trends of the water balance components in the sub-basins
The general trends in the major water balance components in the sub-basins are indicated in Fig. 12a–c. The increase in surface runoff and in evapotranspiration is more significant in the sub-basins. The change in annual water yield is not high, and it seems to be decreasing in the Meki sub-basin. Due to the changes in LULC in the sub-basins, it is of critical importance to address the reduction in water yield components, which are the crucial component for water availability. Water yield enhancement and land use management are, therefore, necessary to improve the water yield.
Fig. 12
Annual average values of major water balance components and their trends in the sub-basins due to changes in LULC in relation to the base year land uses; a surface runoff, b water yield, c evapotranspiration
×
Conclusion
Water resources are influenced by various land uses, such as industrialization, urbanization, forestry, and agriculture. Understanding the variations in the components of the water balance due to changes in LULC is important for effective water management. Our results indicate that changes in LULC mainly affect the evapotranspiration, surface runoff and water yield components of the water balance in the CRV sub-basins. An increase in forest cover in the sub-basins resulted in a reduction in runoff and an increase in evapotranspiration. Water yields were also affected by a change in forest cover and other land uses, such as agriculture and rangeland. Forest cover and changes in forest cover were, therefore, found to be the most decisive factor affecting the components of the water balance, followed by changes in agricultural land use and changes in rangeland.
Understanding the impacts of LULC change dynamics on water resources can also help engineers, planners and managers to develop management and development strategies to reduce the negative impacts of future LULC dynamics on water resources. It will also help policy-makers and government bodies to make better decisions on resource development and management. Land use and land cover conservation planning, based on site-specific LULC changes, is, therefore, crucial for proper surface and groundwater management.
Acknowledgements
We acknowledge financial support from the Czech National Agency for Agricultural Research, NAZV (Project No. QK 1910086). We would also like to thank the Ethiopian Meteorological Agency, the Oromia Bureau of Agriculture and Natural Resources, the Ethiopian Ministry of Water Resources, the Ethiopian Geospatial and Information Institute, and the Ethiopian Ministry of Agriculture and Natural Resources for providing the data and for their support during data collection. We would also like to thank Robin Healey for his language proof reading and for suggestions related to the text.
Declarations
Conflict of interest
We declare that we have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
