Land use land cover change as a casual factor for climate variability and trends in the Bilate Watershed, Ethiopia Authors: Samuel Shibeshi Bikeko, Venkatesham E. Published: 2024 ABSTRACT Climate change and land use dynamics are critical issues facing many regions worldwide, particularly in developing countries. This study examines the spatiotemporal changes in land use and land cover (LULC) and their impact on climate variability in the Bilate Watershed, Ethiopia, from 1994 to 2024. Utilizing multispectral satellite imagery from Landsat 5, 7, and 8, along with meteorological data from five weather stations, LULC classification was performed using the Random Forest algorithm on the Google Earth Engine platform. To analyze climatic variability and trends, the Mann–Kendall trend test, the Standardized Precipitation Index (SPI), and the Standardized Temperature Index (STI) were employed. The findings indicate a significant decline in forest cover, with an accelerated annual loss of approximately 4681.2 hectares between 2014 and 2024. Concurrently, agricultural land expanded by about 1141 hectares annually, and urban areas grew by 24.3 hectares per year in recent years. Seasonal mean rainfall variation showed significant declines in the upper catchment, with Bega (p = 0.004, Sen’s slope = − 3.819 mm), Belg (p = 0.006, Sen’s slope = − 7.972 mm), and Kiremt (p = 0.005, Sen’s slope = − 7.117 mm), while the lower catchment experienced a notable increase during the Belg season (p = 0.025, Sen’s slope = 6.424 mm), highlighting uneven water availability across the watershed. Furthermore, pronounced warming trends were observed in the upper catchment (Bega: p = 0.002, Sen’s slope = 0.029; Belg: p = 0.001, Sen’s slope = 0.030; Kiremt: p = 0.004, Sen’s slope = 0.018), with moderate warming noted in the middle catchment during the Kiremt season (p = 0.020, Sen’s slope = 0.016). These LULC changes have significantly impacted climate variability, emphasizing the critical influence of human activities on regional climate dynamics. This study underscores the urgent need for sustainable land management and conservation strategies to mitigate the challenges posed by deforestation, urbanization, and agricultural expansion. Introduction Land use and land cover change (LULCC) significantly contributes to climate change through various mechanisms, including deforestation (Liu et al., 2016), agricultural expansion (Prestele et al., 2017), urbanization (Hassan et al., 2016), and the degradation of wetlands and grasslands (Zhang et al., 2014). Deforestation releases stored carbon dioxide (CO2) into the atmosphere, with studies indicating that it accounts for about 10–15% of global anthropogenic CO2 emissions (Baccini et al., 2017; Harris et al., 2021). The conversion of natural landscapes into agricultural lands leads to substantial greenhouse gas (GHG) emissions, particularly methane (CH4) and nitrous oxide (N2O) from livestock and rice paddies (Carlson et al., 2017; Tubiello et al., 2007). Urbanization increases local temperatures through the heat island effect and boosts GHG emissions due to higher energy (Seto et al., 2012; Zhao & Wan, 2021). Draining wetlands and degrading peatlands result in the oxidation of stored organic matter, releasing CO2 and CH4 (Dargie et al., 2017; Hooijer et al., 2012). Additionally, the conversion of grasslands and shrublands disrupts carbon and water cycles, reducing biodiversity and carbon storage capacity (Liu et al., 2023; Wilcox et al., 2017). Climate change is causing significant disruptions in many developing nations, particularly those whose economies heavily rely on climate-sensitive industries with limited adaptability (Mesfin et al., 2020). The variability and trends in climate have profound effects on the environment and social development, which are crucial for supporting a growing human population (Mishraa et al., 2010). Understanding these climatic trends is essential because numerous global challenges, including food insecurity, water scarcity, biodiversity loss, and health issues, are intrinsically linked to climate change (Luo et al., 2013). Extensive research has been conducted to analyze climate trends across various regions and periods, to explore the causes of these trends, and to examine their effects on the environment and human societies (Yang et al., 2013). Ethiopia is known for its unpredictable climate fluctuations and irregular rainfall and has been experiencing extreme temperatures, often leading to significant disasters (Gebrehiwot, 2013). Most of these disasters are climate-related, exacerbated by ecosystem degradation due to unregulated human activities. Poverty further intensifies the issue. Other factors contributing to Ethiopia’s vulnerability to climate variability and change include its heavy dependence on rain-fed agriculture, underdeveloped water resources, rapid population growth, weak institutional frameworks, and a general lack of awareness (Alemayehu & Tilahun, 2019). Recent climate variability has posed substantial challenges to Ethiopia, affecting natural disasters, resource degradation, food security, water and energy supply, poverty alleviation, and sustainable development initiatives (Stavi et al., 2021). In recent decades, Ethiopia has experienced a temperature increase of approximately 0.37 °C, indicating a shift in its climate (Gebrechorkos et al., 2019). Since the 1990s, there has been a notable decline in rainfall, significantly impacting the country’s water supply and agricultural productivity (Mekonen & Berlie, 2020). Both increased human activities and natural climate variability have played substantial roles in the observed changes in precipitation across Ethiopia (Rowell et al., 2015). Studies in Ethiopia (Arndt et al., 2011; Matewos, 2019; Mersha & van Laerhoven, 2018; Mideksa, 2010) have highlighted the significant impact of climate changes on crop productivity. According to Eshetu et al. (2020), agricultural productivity in Ethiopia is projected to decline due to climate variability between 2030 and 2050, potentially increasing the nation’s dependence on food aid. This underscores the need for further research to identify observed climate changes and develop effective adaptation measures. Previous studies on the Bilate Watershed in Ethiopia have revealed significant impacts of LULC changes on hydrology, soil erosion, and crop production. Kuma et al. (2021) indicate that agricultural expansion, deforestation, and population pressure have led to substantial deforestation and land degradation, thereby altering the region’s hydrology and increasing soil erosion. The conversion of forests and grasslands into agricultural lands has resulted in reduced soil fertility and productivity, exacerbating food insecurity (Tamire et al., 2022). Additionally, deforestation has decreased evapotranspiration, leading to higher temperatures and reduced rainfall, which exacerbates water scarcity and adversely affects agricultural productivity (Sulamo et al., 2021). While these studies have extensively examined the impacts of LULC changes on hydrology, soil erosion, and agricultural productivity, there is a notable gap in research regarding the long-term impacts of LULC changes on climate change and variability. This research aims to address this gap by investigating the profound impacts of LULC changes on climate change and variability over time, thus filling the void left by previous research efforts. Therefore, this research work is intended to (i) assess the spatiotemporal changes of LULC over the past 30 years (1994–2024) in the Bilate Watershed and (ii) investigate climate variability and trends over the Bilate Watershed, thereby addressing the gaps left by previous research efforts. Method Study area The Bilate Watershed is located in Ethiopia, originating from the Gurage Mountains and draining into the Lake Abaya-Chamo Basin. It covers 5625 km2 and spans from the Ethiopian Highlands to the Rift Valley lowlands. The altitude ranges from 1146 to 3393 m above sea level. Geographically, its location extends from 6°36′N 38°00′E to 8°05′N 38°12′E (Fig. 1).Fig. 1Location map of the study area Location map of the study area Data source Satellite data To study and analyze LULC changes in the Bilate Watershed, multispectral satellite images of the area were acquired for four distinct years: 1994, 2004, 2014, and 2024. The temporal span of these images enables a comprehensive analysis of LULC dynamics over 30 years. Table 1 summarizes the acquisition details for each set of satellite images. Table 1Satellite data acquisitionYearSensor typeSpatial resolutionAcquisition dateSource1994Landsat MSS30 m12/1/1994www.glovis.gov2004Landsat TM30 m11/14/2004www.glovis.gov2014Landsat 8 OLI/TIRS30 m11/28/2014www.glovis.gov2024Landsat 8 OLI/TIRS30 m02/16/2024www.glovis.gov Table 1Satellite data acquisitionYearSensor typeSpatial resolutionAcquisition dateSource1994Landsat MSS30 m12/1/1994www.glovis.gov2004Landsat TM30 m11/14/2004www.glovis.gov2014Landsat 8 OLI/TIRS30 m11/28/2014www.glovis.gov2024Landsat 8 OLI/TIRS30 m02/16/2024www.glovis.gov Satellite data acquisition Metrological data For climate variability and trend analysis in the Bilate Watershed, meteorological data was obtained from the Ethiopian Meteorological Institute (EMI). Data collection spanned five stations strategically distributed across the watershed’s three catchments. In the upper catchment, data was gathered from Fonko and Hossaina stations. Alaba Kulito and Durame stations were selected for the middle catchment, while Bodity station was used for the lower catchment. This distribution of meteorological stations ensures comprehensive coverage of the watershed, facilitating detailed analysis of climatic trends and variability across different altitudinal zones and land use areas. Reference data acquisition The research conducted an accuracy assessment for the classified satellite images using a total of 300 Ground Control Points (GCPs). These GCPs were systematically collected from September 20, 2023, to October 28, 2024, ensuring comprehensive spatial coverage across five distinct LULC categories. The GCPs for the Settlement category were collected from impervious surfaces and roads in the towns of Durame, Hossaina, and Alaba Kulito. Forest GCPs were obtained through collaboration with the Bilate Watershed Authority. Bare land GCPs were sourced from open spaces predominantly covered with sand. Agricultural GCPs were gathered from various individual farmlands within the watershed. For the water body category, GCPs were referenced against Budameda, Tido Kubi, and Mekbib Petros lakes located in the Bilate catchment area. Data preprocessing Data preprocessing for both satellite and meteorological datasets involves several critical steps to ensure the data is accurate and suitable for analysis. For satellite data, the process includes radiometric and geometric corrections to adjust pixel values and align images with real-world coordinates, respectively. This is followed by an image sub-setting to focus on the Bilate Watershed, and cloud masking to exclude cloud-covered areas. For meteorological data, preprocessing begins with collecting data from five stations (Fonko, Hossaina, Alaba Kulito, Durame, and Bodity) and includes cleaning to handle missing values and outliers, homogenization for consistency, and temporal aggregation for trend analysis. Data analysis The overall methodology of the study is outlined in the flowchart (Fig. 2). Following the preprocessing of satellite and meteorological data, the data analysis phase commenced. LULC classification was executed using the Google Earth Engine (GEE) platform, employing the Random Forest (RF) classification method to classify images for the years 1994, 2004, 2014, and 2024 within the river basin. Post-classification, an accuracy assessment was carried out to evaluate how well the classification results matched the ground reality, utilizing the collected GCP. Subsequently, change detection analysis was performed to identify land use transitions and to determine the dynamics over the 30 years. This analysis helped in understanding which land use categories were converted into others and the extent of these changes. In parallel, climate variability and trend analysis were conducted using the meteorological data for the river basin. The Mann–Kendall test was applied to assess trends in the climatic data. Additionally, for drought vulnerability analysis, the Standardized Precipitation Index (SPI) and Standardized Temperature Index (STI) were utilized. These indices helped in identifying periods of drought and assessing their severity.Fig. 2Methodological flowchart Methodological flowchart Land use land cover classification The Random Forest (RF) classification method was used for image classification. The choice of employing the Random Forest (RF) machine learning approach for land cover classification is driven by its recognized superiority in achieving high classification accuracy. Particularly within Google Earth Engine (GEE), RF is widely adopted for both classification and regression tasks, as emphasized by the work of (Shelestov et al., 2017). This algorithm is selected for its capacity to provide robust and accurate predictions through an ensemble learning paradigm. The RF process involves the generation of multiple decision trees, each constructed based on random subsets of features, thereby enhancing diversity and robustness. Bootstrapping initiates the classification procedure by creating varied sets of decision trees through random subsampling of the dataset with replacement. During tree construction, feature randomization is introduced to consider only subsets of features at each node, mitigating the risk of overfitting and enhancing model stability. In classification tasks, the ultimate prediction is determined through a majority voting mechanism among the individual trees, constituting an ensemble decision-making process expressed mathematically in Eq. 1. The percentage of training and test data in the RF model is typically allocated around 70–80% for training and 20–30% for testing, ensuring a balance between model training and evaluation (Ramo & Chuvieco, 2017). Hyperparameters, such as the number of trees, maximum tree depth, and features considered at each split, play a crucial role in RF model performance. The optimum values for these parameters are determined through hyperparameter tuning, with techniques like random search exploring the parameter space. The specific values depend on the characteristics of the dataset and the complexity of the land cover classification task.1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$D\mathit=\mathit{\left\{{\left(x1,y1\right),\;\left(x2,y2\right),\dots,\left(xn,yn\right)}\right\}}$$\end{document}D=x1,y1,x2,y2,⋯,xn,ynwhere D is the bootstrapped dataset, xi stands as a vector of M feature values for the ith sample, and yi is the corresponding class label for classification. Accuracy assessment The accuracy assessment of classified land cover maps is a crucial step in ensuring the reliability of satellite-based classifications. For the accuracy assessment, GPS data, existing land cover maps, and Google Earth imagery were employed. Once validation points from various sources are acquired, accuracy assessment statistics, including User Accuracy, Producer Accuracy, and Overall Accuracy, were computed (Eq. 2–4). Accuracy metrics equations were adapted from (Dagne et al., 2023). These statistics provide quantitative measures of the classification's performance, offering insights into the reliability and precision of the classified land cover maps over the specified time frame. This comprehensive validation strategy helps overcome challenges related to ground data accessibility and ensures the robustness of the land cover classification results.2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$OA=\frac{Number\;of\;correctly\;classified\;pixel}{Total\;number\;of\;pixels}\ast100$$\end{document}OA=NumberofcorrectlyclassifiedpixelTotalnumberofpixels*1003\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$UA_i=\frac{Number\;of\;correctly\;classified\;pixelf\;or\;class\;i}{Total\;number\;of\;pixels\;classified\;as\;class\;i}\ast100$$\end{document}UAi=NumberofcorrectlyclassifiedpixelforclassiTotalnumberofpixelsclassifiedasclassi*1004\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$PA_i=\frac{Number\;of\;correctly\;classified\;pixel\;for\;class\;i}{Total\;number\;of\;pixels\;of\;class\;i}\ast100$$\end{document}PAi=NumberofcorrectlyclassifiedpixelforclassiTotalnumberofpixelsofclassi*100 Climate variability and trends To investigate the evolution of climatic characteristics over time, various analytical approaches are employed. The Standardized Precipitation Index (SPI) assesses changes in precipitation patterns by examining the frequency and intensity of precipitation events, providing insights into trends and variability. The Mann–Kendall (MK) trend test is utilized to detect significant climate trends in time series data, identifying monotonic increases or decreases that indicate long-term changes. Concurrently, the Standardized Temperature Index (STI) measures temperature variability across a study area’s observed time scale, offering a standardized metric for comparing temperature fluctuations. Standardize Precipitation Index (SPI) The Standard Precipitation Index (SPI) is a widely utilized tool for assessing and monitoring precipitation anomalies over different time scales, crucial for identifying periods of drought or excess rainfall. The SPI standardizes precipitation data, allowing for meaningful comparisons across diverse climates and geographical locations. This index can be determined using the following equation (Eq. 5) adapted from (Karavitis et al., 2011):5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Z=\frac{Xi-\overline{Xi}}{S }$$\end{document}Z=Xi-Xi¯S Z is the Standardized Precipitation Index (SPI), representing the deviation of annual precipitation Xi from the mean (Xi) of the historical record, scaled by the standard deviation S of the annual precipitation. Standardize Temperature Index (STI) The Standardized Temperature Index (STI) is a statistical measure used to assess and quantify temperature anomalies over a specific period. It is particularly useful in identifying and evaluating periods of unusually high or low temperatures, thereby aiding in the analysis of climate variability and trends (Eq. 6), and its mathematical equation is adapted from (Li et al., 2021). The STI is analogous to the Standardized Precipitation Index (SPI), which is used for drought assessment based on precipitation data.6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\text{STI}=\frac{Ti-\mu }{\sigma }$$\end{document}STI=Ti-μσwhere Ti is the temperature for a specific period (seasonal), µ is the mean temperature over a reference period, σ is the standard deviation of the temperature over the same reference period. Mann Kendall Test The Mann–Kendall Trend test assesses monotonic trends in time-series data, such as temperature and precipitation from 1994 to 2024, using metrological records. It calculates the Mann–Kendall test statistic S, which evaluates the sum of signs of pairwise differences in the data series. The variance Var (S) considers tied groups within the data, adjusting for any non-independence. The resulting standard normal test statistic Z is derived from S and Va (S), indicating the significance of observed trends: positive Z values denote increasing trends, negative values indicate decreasing trends, and values exceeding critical thresholds suggest statistical significance. This non-parametric approach accommodates data of all distributions and is crucial for detecting climate and environmental changes over time reliably (Tong et al., 2014):7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Z=\;\begin{bmatrix}\frac{S-1}{\sqrt{V\left(S\right)}},&for&S>0\\0&for&s=0\\\frac{S+1}{\sqrt{V\left(S\right)}},&for&S<0\end{bmatrix}$$\end{document}Z=S-1VS,forS>00fors=0S+1VS,forS<08\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$S=\sum_{i=1}^{n-1}\sum_{j=i+1}^{n}sign({X}_{j}-{X}_{i})$$\end{document}S=∑i=1n-1∑j=i+1nsign(Xj-Xi)9\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$sign=\;\begin{bmatrix}1&for&(X_j-X_i)>0\\0&for&\left(X_j-X_i\right)=0\\-1&for&\left(X_j-X_i\right)<0\end{bmatrix}$$\end{document}sign=1for(Xj-Xi)>00forXj-Xi=0-1forXj-Xi<010\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$V\left(S\right)=\frac{1}{18}\left[n\left(n-1\right)\left(2n+5\right)-\sum_{p=1}^{q}{t}_{p}({t}_{p}-1)(2{t}_{p}+5)\right]$$\end{document}VS=118nn-12n+5-∑p=1qtp(tp-1)(2tp+5) Result Spatiotemporal dynamics of LULC over the past 30 years (1994–2024) Between 1994 and 2004 (Fig. 3), the Bilate Watershed saw a significant reduction in forest cover, dropping from 284,048.74 hectares (50.50% of the basin) to 215,361.21 hectares (38.29%). This represents a substantial loss of 68,687.53 hectares, highlighting extensive deforestation. Conversely, agricultural land expanded considerably, from 265,283.26 hectares (47.16%) to 331,394.54 hectares (58.91%), indicating a growth of 66,111.28 hectares. The built-up areas also increased markedly, from a mere 4.14 hectares to 3,013.64 hectares (0.54%), driven by urbanization and infrastructure development. The area of water bodies saw a modest increase from 769.95 hectares (0.14%) to 1040.32 hectares (0.18%), while bare land decreased slightly from 12,375.51 hectares (2.20%) to 11,706.47 hectares (2.08%) (Table 2). From 2014 to 2024 (Fig. 4), the trends observed earlier continued, with forest cover further reducing from 210,786.52 hectares (37.47%) to 163,974.43 hectares (29.15%), a loss of 46,812.09 hectares. Agricultural land, which had seen a slight decline by 2014, increased again from 329,668.21 hectares (58.61%) to 341,078.85 hectares (60.64%), marking an addition of 11,410.64 hectares. Bare land dramatically increased from 15,134.67 hectares (2.69%) to 48,799.57 hectares (8.68%), highlighting severe land degradation issues. The built-up area grew significantly from 3504.45 hectares (0.62%) to 7,251.68 hectares (1.29%), reflecting ongoing urbanization. Water bodies, however, experienced a reduction from 3382.87 hectares (0.60%) to 1331.32 hectares (0.24%) (Table 2), indicating possible issues with water management or conservation.Fig. 3Land cover maps of Bilate Watershed between 1994 and 2004Table 2Land use land cover dynamics over the past 30 years (1994–2024)1994200420142024Class nameArea (Ha)%Area (Ha)%Area (Ha)%Area (Ha)%Forest284048.7450.50215361.2138.29210786.5237.47163974.4329.15Water body769.950.141040.320.183382.870.601331.320.24Bare land12375.512.2011706.472.0815134.672.6948799.578.68Built up4.140.003013.640.543504.450.627251.681.29Agriculture265283.2647.16331394.5458.91329668.2158.61341078.8560.64Total562481.60100.00562516.18100.00562476.72100.00562435.85100.00Fig. 4Land cover maps of Bilate Watershed between 2014 and 2024 Land cover maps of Bilate Watershed between 1994 and 2004 Table 2Land use land cover dynamics over the past 30 years (1994–2024)1994200420142024Class nameArea (Ha)%Area (Ha)%Area (Ha)%Area (Ha)%Forest284048.7450.50215361.2138.29210786.5237.47163974.4329.15Water body769.950.141040.320.183382.870.601331.320.24Bare land12375.512.2011706.472.0815134.672.6948799.578.68Built up4.140.003013.640.543504.450.627251.681.29Agriculture265283.2647.16331394.5458.91329668.2158.61341078.8560.64Total562481.60100.00562516.18100.00562476.72100.00562435.85100.00 Land use land cover dynamics over the past 30 years (1994–2024) Land cover maps of Bilate Watershed between 2014 and 2024 Rate of change From 1994 to 2004, the Bilate Watershed underwent notable changes in land use and land cover (Fig. 3), reflecting significant environmental and socio-economic shifts. Forest areas experienced a substantial decline, with an average annual loss of 686.9 hectares, largely driven by logging, agricultural expansion, and urbanization (Kuma et al., 2021). This period also saw a modest increase in water bodies by 27.01 hectares per year, likely due to improved watershed management and natural hydrological processes. Concurrently, bare land decreased by 66.95 hectares annually, suggesting successful soil conservation and land reclamation efforts. The built-up area expanded rapidly at a rate of 300.47 hectares per year, indicating robust population growth and infrastructural development. Agricultural land grew dramatically by 6611 hectares annually, underscoring the escalating demand for food production and land conversion (Belayneh et al., 2013; Wubie et al., 2016). These changes highlight the complex interplay between natural processes and human activities, necessitating sustainable land management practices to mitigate adverse environmental impacts. The decade from 2004 to 2014 marked further dynamic shifts in the Bilate Watershed’s land cover (Fig. 4). The rate of forest loss, although still significant, slowed to 457.5 hectares per year, possibly due to initial conservation efforts and changes in land use policies (Mariye et al., 2022). Water bodies saw a more pronounced increase, growing by 130.2 hectares annually (Table 3), which can be attributed to enhanced water management and conservation measures. However, bare land began to increase significantly at a rate of 342.8 hectares per year, indicating rising deforestation and land degradation issues. The expansion of built-up areas slowed to 49.1 hectares per year, reflecting more stringent urban planning and land use regulations. Interestingly, agricultural land decreased 172.6 hectares annually, potentially due to soil degradation, reduced fertility, and conversion of agricultural land to other uses. These trends underscore the necessity for ongoing and improved land management strategies to balance development with environmental sustainability. Table 3Rate change of land use land cover changes from 1994 and 20241994–20042004–20142014–20241994–2024Forest − 686.9 − 457.5 − 4,681.20 − 4,002.49Waterbody27.01130.2 − 20.5118.7Bare land − 66.95342.8336.6540.47Built up300.4749.124.3241.42Agricultures6,611 − 172.611412526.2 Table 3Rate change of land use land cover changes from 1994 and 20241994–20042004–20142014–20241994–2024Forest − 686.9 − 457.5 − 4,681.20 − 4,002.49Waterbody27.01130.2 − 20.5118.7Bare land − 66.95342.8336.6540.47Built up300.4749.124.3241.42Agricultures6,611 − 172.611412526.2 Rate change of land use land cover changes from 1994 and 2024 Between 2014 and 2024, the trends in land use and land cover change became even more pronounced, with significant implications for the region's environmental and socio-economic landscape. Forest areas continued to decline sharply, with an alarming annual loss rate of 4681.2 hectares, exacerbating issues such as habitat loss, biodiversity decline, and increased carbon emissions. The area covered by water bodies decreased slightly by 20.51 hectares per year, raising concerns about water resource management and climate change impacts (Seyam et al., 2023). Bare land expanded at a rate of 336.65 hectares annually, reflecting ongoing land degradation and deforestation pressures. Meanwhile, the growth rate of built-up areas slowed further to 24.3 hectares annually (Table 3), suggesting more controlled and sustainable urban development. Agricultural land began to increase again at a rate of 1141 hectares per year, indicating efforts in land recovery and sustainable farming practices. These changes highlight the urgent need for comprehensive land management strategies, environmental conservation initiatives, and sustainable development practices to address the complex challenges facing the Bilate Watershed (Dagne et al., 2023). Accuracy assessment The accuracy assessment results for land cover classification in the Bilate Watershed from 1994 to 2024 indicate varying levels of producer and user accuracy across different land cover types, with overall accuracy and kappa values reflecting improvements over time. In 1994, the overall accuracy was 0.63 with a kappa value of 0.52, indicating moderate agreement. By 2004, overall accuracy increased to 0.67 and the kappa value to 0.59, showing improved classification reliability. The trend continued in 2014, with overall accuracy reaching 0.74 and a kappa of 0.68, reflecting substantial agreement. In 2024, the overall accuracy slightly declined to 0.71, with a kappa value of 0.65, still indicating substantial agreement (Table 4). Notably, the accuracy of forest classification fluctuated, with the highest user accuracy of 1.00 observed in both 1994 and 2024, while water bodies and built-up areas showed significant variability in accuracy over the years (Table 5). Table 4Accuracy assessment results (1994, 2004, 2014, and 2024)1994200420142024Land coverProducerUserProducerUserProducerUserProducerUserForest0.671.000.680.400.640.500.681.00Water Body0.760.860.680.950.960.791.000.85Bare land0.600.501.000.830.920.720.960.44Built up0.360.260.950.490.650.620.750.60Agricultures0.800.920.801.000.701.000.901.00Overall0.630.670.740.71Kappa0.520.590.680.65Table 5Seasonal rainfall variation across each catchment in the Bilate WatershedSeasonKendall’s taup-valueSen’s slopeUpper catchmentBega − 0.3150.004* − 3.819Belg − 0.3000.006* − 7.972Kiremt − 0.3050.005* − 7.117Middle catchmentBega − 0.0840.445 − 0.740Belg0.0540.6291.540Kiremt − 0.1930.0078* − 3.515Lower catchmentBega0.1430.1921.630Belg0.2440.0256.424Kiremt0.0220.8490.615 Table 4Accuracy assessment results (1994, 2004, 2014, and 2024)1994200420142024Land coverProducerUserProducerUserProducerUserProducerUserForest0.671.000.680.400.640.500.681.00Water Body0.760.860.680.950.960.791.000.85Bare land0.600.501.000.830.920.720.960.44Built up0.360.260.950.490.650.620.750.60Agricultures0.800.920.801.000.701.000.901.00Overall0.630.670.740.71Kappa0.520.590.680.65 Accuracy assessment results (1994, 2004, 2014, and 2024) Table 5Seasonal rainfall variation across each catchment in the Bilate WatershedSeasonKendall’s taup-valueSen’s slopeUpper catchmentBega − 0.3150.004* − 3.819Belg − 0.3000.006* − 7.972Kiremt − 0.3050.005* − 7.117Middle catchmentBega − 0.0840.445 − 0.740Belg0.0540.6291.540Kiremt − 0.1930.0078* − 3.515Lower catchmentBega0.1430.1921.630Belg0.2440.0256.424Kiremt0.0220.8490.615 Seasonal rainfall variation across each catchment in the Bilate Watershed Climate variability and trends The analysis of seasonal mean rainfall variation across the Bilate Watershed reveals significant spatial disparities. In the upper catchment, all three seasons Bega,1Belg,2 and Kiremt3 show a significant decline in rainfall, with Kendall’s tau values of − 0.315, − 0.300, and − 0.305, respectively, and corresponding p-values all below 0.01. The Sen’s slope indicates sharp reductions, particularly during Belg (− 7.972 mm) and Kiremt (− 7.117 mm), highlighting critical water shortages for agriculture. In the middle catchment, the Kiremt season also exhibits a significant decline (Kendall’s tau = − 0.193, p = 0.0078), with a Sen’s slope of − 3.515 mm, while Bega shows a non-significant decrease and Belg a minor, non-significant increase. Conversely, the lower catchment displays a significant rise in Belg rainfall (Kendall’s tau = 0.244, p = 0.025), with a Sen’s slope of 6.424 mm, suggesting improved water availability during this season, although Bega and Kiremt remain relatively stable without significant trends. These patterns underscore the uneven distribution of rainfall changes across the watershed, with marked reductions in the upper and middle catchments (Figs. 5 and 6), while the lower catchment (Fig. 7) benefits from an increase in Belg rainfall, potentially impacting agriculture and water resources differently across the regions (Figs. 8 and 9).Fig. 5Mean seasonal rainfall (mm) variation in the lower catchment of Bilate WatershedFig. 6Mean seasonal rainfall (mm) variation in the middle catchment of Bilate WatershedFig. 7Mean seasonal rainfall (mm) variation in the upper catchment of Bilate WatershedFig. 8Seasonal mean temperature variation in the lower catchment of Bilate WatershedFig. 9Seasonal mean temperature variation in the middle catchment of Bilate Watershed Mean seasonal rainfall (mm) variation in the lower catchment of Bilate Watershed Mean seasonal rainfall (mm) variation in the middle catchment of Bilate Watershed Mean seasonal rainfall (mm) variation in the upper catchment of Bilate Watershed Seasonal mean temperature variation in the lower catchment of Bilate Watershed Seasonal mean temperature variation in the middle catchment of Bilate Watershed The analysis of seasonal mean temperature variation across the Bilate Watershed reveals significant spatial and temporal disparities (Table 6). In the upper catchment, all seasons Bega, Belg, and Kiremt show strong warming trends, with Kendall’s tau values of 0.341, 0.429, and 0.414, respectively, and corresponding p-values below 0.01. The Sen’s slope highlights notable temperature increases, particularly during Belg (0.030 °C) and Bega (0.029 °C), signaling heightened vulnerability to warming. In the middle catchment, temperature trends are more mixed, with Bega showing no significant change, while Kiremt displays a significant warming trend (Kendall’s tau = 0.254, p = 0.020) and Belg showing a slight, non-significant increase. In the lower catchment, no statistically significant temperature changes are observed across any season, with minimal Sen’s slope values and p-values exceeding 0.1. These findings highlight the pronounced warming in the upper catchment (Fig. 10), moderate warming in the middle catchment (Fig. 9) during Kiremt, and relative stability in the lower catchment (Fig. 8), indicating uneven temperature impacts across the Bilate Watershed. Table 6Seasonal mean temperature variation across each catchment in the Bilate WatershedSeasonKendall’s taup-valueSen’s slopeUpper catchmentBega0.3410.002**0.029Belg0.4290.001**0.030Kiremt0.4140.004**0.018Middle catchmentBega − 0.0110.928 − 0.001Belg0.1160.2910.013Kiremt0.2540.020*0.016Lower catchmentBega − 0.1660.129 − 0.014Belg0.0780.4790.006Kiremt0.1650.1320.010Fig. 10Seasonal mean temperature variation in the upper catchment of Bilate Watershed Table 6Seasonal mean temperature variation across each catchment in the Bilate WatershedSeasonKendall’s taup-valueSen’s slopeUpper catchmentBega0.3410.002**0.029Belg0.4290.001**0.030Kiremt0.4140.004**0.018Middle catchmentBega − 0.0110.928 − 0.001Belg0.1160.2910.013Kiremt0.2540.020*0.016Lower catchmentBega − 0.1660.129 − 0.014Belg0.0780.4790.006Kiremt0.1650.1320.010 Seasonal mean temperature variation across each catchment in the Bilate Watershed Seasonal mean temperature variation in the upper catchment of Bilate Watershed The observed differences in temperature trends across the three catchments can be explained by several key factors. Variations in land use, such as deforestation and agriculture, can impact local temperatures. Reduced vegetation in the middle and upper catchments can lead to higher surface temperatures, while the lower catchment might retain more moisture and have cooler temperatures due to better vegetation cover. This relationship between land use and temperature is well documented in studies like those by (Lambin et al., 2003; Meyfroidt et al., 2010). Anthropogenic climate change affects temperature patterns differently across elevations. Lower catchments may experience more significant warming due to urbanization and increased greenhouse gas emissions, while upper catchments might show mixed trends due to the varying impacts of global warming on high-altitude climates. The (Alley et al., 2007; Sharma et al., 2019; Sheffield & Wood, 2008; Wang et al., 2022) reports provide comprehensive evidence on how global warming is altering temperature patterns worldwide. Drought Severity Index (DSI) Standardized Precipitation Index (SPI) Figure 11 depicts the SPI for the upper, middle, and lower catchments from 1994 to 2024, illustrating the variability in drought conditions. The upper catchment exhibits significant fluctuations, with notable wet periods around 1995, 2010, and 2020, where SPI values exceed 2.0, and severe droughts around 2000 and 2015, where values drop below − 1.0. This variability suggests a high sensitivity to climatic changes, likely influenced by its altitude and orographic effects. The middle catchment shows similar trends but with moderated extremes, indicating less sensitivity compared to the upper catchment. Noteworthy wet periods are observed around 1995 and 2010, with significant droughts around 2000 and 2015. The lower catchment displays the least variability, with less extreme SPI values throughout the observed period. Significant wet conditions are seen around 1995 and 2020, while droughts are less severe than in the other catchments. These patterns reflect the lower catchment's more stable climatic influences and lower altitude. The overall trends across all catchments highlight an increasing frequency of extreme precipitation events in recent years, potentially due to anthropogenic climate change (Sharma et al., 2019). Additionally, land use changes such as deforestation and urbanization in the middle and lower catchments may contribute to the moderated SPI values observed (Lambin & Meyfroidt, 2011).Fig. 11Standardized Precipitation Index (SPI) over the three catchments Standardized Precipitation Index (SPI) over the three catchments Standardized Temperature Index (STI) The STI shows significant temperature anomalies and their variations across the upper, middle, and lower catchments from 1994 to 2024, revealing unique climatic dynamics in each region. The upper catchment shows pronounced variability from 1994 to 1999 with alternating warm and cool phases, attributed to its higher elevation, followed by more stable conditions from 1999 to 2004, except for a warm period around 2003–2004. Post-2004, increased temperature variability indicates rising anomalies likely due to broader climatic changes. The middle catchment experiences similar but less extreme variations during the same initial period, with stability between 1999 and 2004, and moderate variability thereafter, aligning with trends in the upper catchment. The lower catchment, influenced by its lower altitude, exhibits moderate variability from 1994 to 1999 and a significant warm anomaly around 2003–2004, with increasing variability from 2004 onwards, reflecting broader climatic influences observed in the upper and middle catchments (Fig. 12).Fig. 12Standardized Temperature Index (STI) over the three catchments Standardized Temperature Index (STI) over the three catchments Interplay between land use land cover change and climate variability The association between land cover change and climate variability in the Bilate Watershed over the past 30 years demonstrates a complex interplay where anthropogenic activities and natural climatic shifts exacerbate environmental changes. The significant reduction in forest cover, which dropped from 284,048.74 hectares in 1994 to 163,974.43 hectares in 2024, is linked to increased deforestation driven by agricultural expansion and urbanization. This loss of forest area correlates with rising temperatures and altered precipitation patterns. Studies have shown that deforestation leads to reduced evapotranspiration and changes in surface albedo, contributing to local and regional climate warming (Asante et al., 2021; Lawrence & Vandecar, 2015). The concurrent increase in agricultural land from 265,283.26 hectares to 341,078.85 hectares has further disrupted the local climate by increasing land surface temperatures due to the replacement of forests with cropland (Pielke et al., 2007). Climate variability, characterized by significant seasonal changes in rainfall and temperature across different catchments, further influences land cover dynamics. The observed decrease in seasonal rainfall in the upper catchment, coupled with an increase in temperature, suggests a feedback loop where reduced forest cover exacerbates drought conditions, which in turn inhibits forest regrowth and promotes land degradation. The rise in bare land from 12,375.51 hectares in 1994 to 48,799.57 hectares in 2024 underscores this degradation. The variability in rainfall patterns, with significant positive trends in the lower catchment rainfall, contrasts with the significant negative trends in the upper catchment rainfall, reflecting how land cover changes can differentially impact microclimates within a watershed (Eckstein et al., 2019; Sheffield & Wood, 2008). The shifts in land cover also impact drought severity indices such as the SPI and STI. The upper catchment, experiencing pronounced variability in these indices, highlights its sensitivity to both climatic changes and land cover modifications. Deforestation and agricultural expansion have likely increased the region's vulnerability to extreme climatic events, evidenced by the increased frequency of both severe droughts and wet periods in recent years. This pattern aligns with global studies indicating that land use changes, such as deforestation, can significantly amplify climate extremes by altering hydrological cycles and surface energy balances (Lambin & Meyfroidt, 2011; Rasul & Sharma, 2016; Sharma et al., 2019). Hence, the observed trends in the Bilate Watershed underscore the critical need for integrated land and climate management strategies to mitigate adverse environmental impacts. Discussion The significant decline in forest cover in the Bilate Watershed from 50.4% in 1994 to 29.1% in 2024 indicates widespread deforestation and forest degradation, aligning with trends observed across Ethiopia and other parts of Africa. Studies such as those by (Getahun et al., 2013) attribute this forest loss to agricultural expansion and population pressure. Similarly, (Debebe et al., 2023) noted that deforestation in Ethiopia is often driven by the need for agricultural land and fuelwood, with long-term implications for biodiversity and climate change. The increase in agricultural land from 47.1 to 60.6% over the study period in the Bilate Watershed reflects a common trend in many developing regions, driven by the need to meet the food demands of a growing population. This expansion, as reported by (Assefa & Bork, 2014; Gebrechorkos et al., 2019), poses a significant threat to sustainable land management by converting forested and marginal lands into agricultural use. The growth of built-up areas in the Bilate Watershed from 0.007 to 1.3% reflects significant urban development driven by economic growth and population increase, a trend mirrored in many rapidly developing regions (Seto et al., 2012; Seyam et al., 2023). Urbanization often leads to the conversion of agricultural and natural lands into urban infrastructure, with profound implications for local ecosystems and biodiversity. The variability in the extent of water bodies in the Bilate Watershed, with an overall reduction in recent years, highlights the impact of environmental changes and water management practices, as observed in other watersheds studied by (Ayele et al., 2017; Mekonnen & Hoekstra, 2016). Additionally, the increase in bare land from 2.2 to 8.6% suggests issues such as soil erosion and land degradation, consistent with findings from studies by (Mariye et al., 2022; Seyam et al., 2023). These trends underscore the importance of implementing sustainable land management practices and soil conservation measures to combat soil erosion and maintain land productivity, as emphasized by (Hurni et al., 2005). The findings of this research indicate that the middle and lower catchments, more affected by human activities such as deforestation and urbanization, show moderated SPI values. This is consistent with (Mahmood et al., 2019; Pielke et al., 2011), which confirms that land use changes can significantly alter local climates by affecting surface albedo, evaporation rates, and atmospheric circulation. Studies by (Abera et al., 2020; Mengist et al., 2022) further emphasize the impact of agricultural expansion and urban development in Ethiopia, which aligns with the observed trends in the Bilate Watershed’s middle and lower catchments. Studies by (Sheffield & Wood, 2008; Wang et al., 2022) report increased climate variability, with more frequent and severe weather events like droughts and heavy rainfall. The research findings of more frequent extreme weather events in the Bilate Watershed corroborate this, showing increased SPI variability. Additionally, research by (Adger et al., 2006; Gebreegziabher et al., 2016; Kumssa & Jones, 2010; Mekonnen & Hoekstra, 2016) highlights similar trends of declining rainfall during critical periods in other parts of Ethiopia, reinforcing global patterns. The research findings mention significant seasonal rainfall variations, with the lower catchment showing an incremental trend, and the upper catchment experiencing a decreasing trend. This aligns with studies, such as those by (Trenberth et al., 2014), which emphasize the impact of changing seasonal weather patterns on regional hydrology. Other studies like those by (Antala et al., 2022; Matewos, 2019) confirm these trends, noting similar seasonal variations in rainfall across different Ethiopian regions. The interplay between land use/land cover changes and local climate variability is well-documented in scientific literature. Changes in forest cover, such as the decline observed in the Bilate Watershed, can significantly impact local microclimates. Forests play a crucial role in regulating temperature and precipitation patterns through evapotranspiration and canopy shading. As forests are converted to agricultural land or bare land, surface albedo increases, leading to higher surface temperatures and reduced evapotranspiration, which may exacerbate local warming and alter precipitation patterns (Mahmood et al., 2019; Pielke et al., 2007). Abera et al. (2020) further highlight that deforestation can lead to reduced rainfall in nearby areas by disrupting the water cycle. In the Bilate Watershed, the shift from forest to agricultural and bare land likely contributes to increased temperature variability and reduced rainfall, as evidenced by the increased SPI variability observed in the middle and lower catchments. The increase in bare land, linked to land degradation and erosion, exacerbates these changes, contributing to local warming and more frequent extreme weather events such as droughts and heat waves (Mariye et al., 2022). Urbanization and the expansion of built-up areas also have significant climatic implications. Studies by (Seto et al., 2012) and (Seyam et al., 2023) show that urbanization alters local climate by replacing vegetated surfaces with impervious materials like concrete and asphalt, which increases surface temperatures (urban heat island effect) and reduces natural infiltration, thereby affecting local hydrology. In the Bilate Watershed, the increase in built-up areas from 0.007 to 1.3% over the study period may contribute to increased local temperatures and reduced moisture retention, further influencing climatic conditions. Similar studies in other Ethiopian regions, such as those by (Mengist et al., 2022) and (Abera et al., 2020), confirm that urbanization can exacerbate local climate variability by influencing temperature and precipitation patterns. The relationship between built-up area expansion and increased SPI variability observed in this research is consistent with these findings. Additionally, the reduction in water bodies, as documented in studies (Mekonnen & Hoekstra, 2016), likely impacts local climate by reducing evaporative cooling and contributing to increased surface temperatures in adjacent areas. This reduction also affects the sustainability of water resources, leading to greater climate variability and increased SPI values during dry seasons (Wang et al., 2022). Conclusion The study of the Bilate Watershed over the past three decades reveals substantial land use and land cover (LULC) changes that have significantly influenced the local climate and environment. The sharp decline in forest cover has contributed to deforestation, localized warming, and altered rainfall patterns by reducing evapotranspiration and increasing surface albedo. Agricultural expansion has transformed natural ecosystems, leading to soil degradation, increased runoff, and shifts in the hydrological cycle, potentially driving more frequent extreme weather events. Results from the Mann–Kendall test highlight that these LULC changes have intensified climate variability, particularly in the upper and middle catchments. Urbanization has disproportionately contributed to local climate change, amplifying temperature anomalies through heat island effects. The reduction of water bodies and the expansion of bare land have further exacerbated temperature increases, destabilizing microclimates within the watershed. However, this study has limitations. First, while remote sensing data offers broad insights, it may lack the fine-grained detail needed for a more localized analysis of LULC impacts. Second, socio-economic drivers behind LULC changes were not deeply examined, which could provide a more holistic understanding of the changes. Future research should consider these factors alongside higher-resolution remote sensing data to better understand the complex relationships between LULC changes and local climate variability. Overall, these findings emphasize the urgent need for sustainable land management practices, reforestation efforts, and improved water management to mitigate the adverse impacts of LULC changes and strengthen the resilience of the Bilate Watershed to future climatic shifts.