Time series land use/land cover mapping and change detection to support policies on sustainable environmental and economic management

Abstract

The upper Tekeze River Basin is facing challenges of widespread deforestation and natural vegetation cover degradation that could exacerbate the water scarcity, food insecurity and extreme poverty in the region. Using remote sensing and GIS, this study quantified the land use land cover change trend in the last three decades and analyzed the current land use / cover statues in the basin. A hybrid classification technique is applied to obtain better classification accuracy. Moreover, for automated cloud and cloud shadow detection the newly developed Mountainous Fmask is used. Using post classification change detection technique, seven major land use/cover classes were identified. These classes remained the dominant classes during the study period, showing marked changes in the area coverage within them. Based on the error matrix statistical indices, the classification accuracies of each class are found to be strong. The overall accuracy and the kappa coefficient for the 2021 map are 91% and 89%, respectively. The techniques used have contributed to improving the accuracy of the classification process and helped the classified images to practically match the ground truths. The analysis revealed settlement expansion by 570.31% in parallel with the expansion of farmland by 52.32% during the period 1991–2021. In contrast, the forestland decreased significantly, by 75.55%. The environmental degradation and unplanned use of land resources could have contributed to why the upper Tekeze basin is experiencing worsening poverty, water scarcity and food insecurity. Thus, land use/cover time series modeling is essential for various purposes, including land use planning and, managing natural resources. In this regards this study provides basic information for implementing sustainable environmental conservation strategies in the area. Furthermore, the applied methodologies may have practical applications in other similar areas.

Introduction

The alteration of land use and land cover is a worldwide environmental concern. There are few unchanged landscapes on Earth, aside from a few isolated and inaccessible regions (Lambin et al. 2003; Riebsame et al. 1994; Ayele et al. 2018). As cited in Hassen et al. (2021), the extent, intensity, and rate of land use/land cover change (LULCC) are more significant and complicated than they were in the past. The change is faster and more noticeable in developing countries (Meyfroidt and Lambin 2011).

Although LULCC can occur either naturally or due to human factors, human activities are the primary forces of this change (Kabba and Li 2011). Human activities such as urbanization, agricultural area expansion, deforestation, and other similar activities have resulted in numerous changes in LULC (Dhanasekarapandian et al. 2015; Chen et al. 2021).

The increasing human population and its impact on land have grown at an exponential rate over the past century (Anantha et al. 2021). The increased human demand places more stress on the planet's finite resources, which in turn influences land use patterns (Islam et al. 2018). As a result of population growth, deforestation has become a common practice in many developing nations with the goal of increasing agricultural output (Berihun et al. 2019; Girma et al. 2022).

Therefore, more research is being published on the subject as a result of the extensive effects of land use changes on climate, ecosystem biophysical assets, and ecosystem services (Zhai et al. 2020).

LULC is driven by many environmental and social factors (Gómez et al. 2016). Apart from the absence of appropriate land use policies, there is a tendency to oversimplify the reasons for the changes in land use and land cover (Tesfaye et al. 2019). Although LULC change is a worldwide problem, the dynamics of this process have varied and been inconsistent due to local conditions (Berihun et al. 2019). To improve our understanding of and response to these dynamics, future studies should consider the relationships among LULC change, the major drivers, and the actors involved (Zhai et al. 2020; Hersperger et al. 2010).

Growing populations and land degradation are making sub-Saharan Africans (SSAs) more vulnerable to economic and environmental change (Christmann 2020). Similarly, Ethiopia has experienced pronounced LULC changes during recent decades, with a considerable expansion of cultivated land at the expense of other LULC types (Kindu et al. 2016; Bewket and Abebe 2013). Ethiopia, among the SSA nations, whose highlands were once 90% forested, is currently less than 4% (Eweg et al. 1998). The Ethiopia highlands are Africa's most populated agricultural regions (Eweg et al. 1998; Samal and Gedam 2021; Temtime and Eshete 2023). Population growth in densely populated highlands is a major driver of land cover dynamics, and the fact that most rural Ethiopians depend on agriculture worsens the LULC issue.

Particularly in the Ethiopian highlands, research has shown that agricultural land (the backbone of the economy) has been expanding at the expense of forested areas (Chen et al. 2020; Akin and Berberoglu 2023). In the Ethiopian highlands, LULC changes resulted in surface runoff, which reduced the water retention capacity, water infiltration, and stream flow, causing wetland loss and lake drying (Samal and Gedam 2021). The land degradation and unplanned use of land resources could have contributed to why northern Ethiopia is experiencing worsening poverty and food insecurity.

Forecasting changes in an area and putting into practice evidence-based, pertinent policies require an understanding of both the historical pattern and the current condition of LULC (Girma et al. 2022). Through an investigation of the empirical and statistical associations between LULC fluctuations in the basin over the last three decades (1991–2001, 2001–2011, and 2011–2021), this study aimed to produce reliable LULC time series data for the UTRB that can be used as an input in the basin’s natural resources management. This study attempted to assess the LULC changes in the UTRB using GIS through the following specific tasks: (i) using GIS to track and analyze the evolution of LULC categories from 1991 to 2021 and (ii) creating a LULC change map with associated transition/change trends and data for LULC classes in the study area.

In contrast to the majority of previous studies in the area, which used either supervised or unsupervised classification method, this study, used a hybrid image classification technique—which combines unsupervised and supervised classification techniques— to improve the classification accuracy (Congalton 2015). Additionally, this study applies the recently developed approach known as Mountainous Fmask (MFmask) for Landsat4-8 images for automated cloud and cloud shadow detection (Qiu et al. 2017) because the UTRB is located in the northern highlands of Ethiopia. While there is a wealth of literature on LULC change for the study area, LULC studies that apply hybrid image classification technique (unsupervised followed by super vised) and Fmask (MFmask) for automated cloud and cloud shadow detection are rare. We think that these techniques helped the classified images become more accurate and to closely reflect the actual circumstances on the ground.

Therefore, the methods used in this study could be used as a guide for future research of a similar nature.

Materials and methods

Description of the study area

Northern Ethiopia is home to the Upper Tekeze River Basin (UTRB), a sub-basin of the Blue Nile basin located in the northwestern area of the country, specifically between geographical grid systems of 12° 0′ 12.20ʺ and 14° 45′ 42.29ʺ N latitude and 36° 32′ 07.70ʺ to 39° 46′ 23.8 ʺ E longitude. The topography of the basin varies greatly. The basin can be classified into three physiographic regions in general: the western lowlands, intermountain valleys and grabens, and the highlands; 70% of the basin being high lands, meaning areas above 1500 m above sea level (a.s.l.) (Hagos et al. 2015). Within the basin, slope can vary greatly, from 0 to 376.5%. The four main soil types that make up the majority of the catchment area in the basin are the lithic leptosols, haplic livisols, euric leptosols, and euric vertisols (Basin 2015).

The elevation ranges from 881 m a.s.l. at the basin outlet to 4620 m a.s.l. at the highest point (Fig. 1). The Tekeze River originates in the southern part of the watershed, adjacent to the Ras Dashen Mountains. It proceeds in a northerly direction, subsequently changing course to the west until it converges with the Atbara River in the northeastern portion of Sudan. The study site encompasses a surface area of 45,694 square kilometers. The onset of the dry season commences in October and extends for four months. Approximately 66% of the annual rainfall occurs in July and August. The basin experiences mean annual rainfall ranging from approximately 400 mm in the northeast to over 1200 mm in the highlands of the southwest (Gebremicael et al. 2017b). The mean annual temperature in the basin ranges from approximately 10 °C in the elevated areas to over 28 °C in the lower regions.

Fig. 1

Location map of the Upper Tekeze River Basin

Despite tremendous efforts on soil and water conservation (SWC) practices, the basin is characterized by severe land degradation due to deforestation, overgrazing and cultivation on the rugged topography (Gebremicael et al. 2017a; Tesfaye et al. 2019).

The Tekeze hydropower dam, highest arch dam in Africa with a height of 180 m and capacity of 300 megawatts electricity, was constructed in the upper basin in 2009 (Welde 2016).

Data sources and analysis method

Three Landsat images at 30 m resolution from Landsat 5_TM (1991, 2001, and 2011) and one at 30 m resolution from Landsat 8_OLI (2021) are used in this study. The images were downloaded from the USGS Centre for Earth Resources Observation and Science (EROS) at (http://earthexplorer.usgs.gov). Although the clearest, most detailed view of the ground from space is provided by satellite imaging with a resolution of 30 cm, to minimize inaccuracies resulting from unique seasonal differences in vegetation distribution, images of dry and the same yearly season are utilized (Dibaba et al. 2020; Gidey et al. 2017). All images are spatially georeferenced in the Universal Transverse Mercator (UTM) projection with the datum World Geodetic System (WGS) 1984 UTM zone 37 N. This approach was helpful for comparison of changes and patterns that occurred in the time period under discussion. Relative cloud-free dates and years (January, December, and November) are used for the acquisition of the Landsat imagery over the study area. The resolution, acquisition date, swath/row, duration, and employed sensors of the satellite images are listed in Table 1. Similarly, the 30 m*30 m digital elevation model (DEM) of the study area was downloaded from http://earthexplorer.usgs.gov. In addition, the image classification and the cleaning, refining, and verifying of classified images are aided by secondary sources such as Google Earth ground control points (GCPs) and ancillary data to infer key topographic metrics. The images are processed in Tiff format, and the details of the image properties are summarized in Table 1. To ensure that no misalignments existed in the region under study, it was necessary to geo-rectify the collected satellite data against the base map using ground control points. The downloaded satellite images were processed using ENVI software version 5.3 and ArcGIS 10.3 (ESRI GIS software) to categorize the different land features of the study area. A classification sieve is applied to refine the classified images, recode the isolated pixels, and remove the salt-and-pepper effect.

Table 1 Landsat images used for LULC change analysis

Image preprocessing and classification

Stacking/composite, mosaicking and sub-setting

After determining the band combinations—bands 4, 3, and 2 for the Landsat 5_TM images and bands 5, 4, and 3 for the Landsat 8_OLI image—to be useful for image display, enhancement and classification, multiple Landsat spectra are combined into a single raster dataset with the help of a layer stacking tool. Then, since the study area falls across eight scenes, image mosaicking is carried out. The images extracted by the mask tool are then applied to subset the mosaicked images to include only the study area, the area of interest (AOI).

Although the function-of-mask (Fmask) technique has performed well in operational cloud identification for Landsat images, according to Qiu et al. (2017), it has certain problems for mountainous locations. Therefore, considering the erratic mountainous topography of the study area located in the Northern Highlands of Ethiopia, a newly developed method known as mountainous Fmask (MFmask) for Landsat4-8 images is utilized for automated cloud and cloud shadow detection. The MFmask method is intended for cloud and cloud shadow identification in hilly regions, where the Fmask algorithm is not working effectively. It was developed based on the success of the Fmask algorithm (Qiu et al. 2017). Brightness Temperature (BT), Digital Elevation Models (DEMs), and Landsat Top of Atmosphere (TOA) reflectance are the inputs used by the MFmask algorithm. With the use of DEMs, MFmask is more effective than Fmask at separating land and water pixels in mountainous regions.

Preliminary image processing, which includes geographic and temporal normalization and geometric and atmospheric corrections, was conducted on all LULC maps. This was necessary to establish a direct connection between the data that were acquired and the actual occurrences that took place in the research area. The ENVI 5.3 FLAASH (Fast Line of Sight Atmospheric Analysis of Spectral Hypercubes) module (ENVI 2009) is used for atmospheric corrections. With FLAASH, it is possible to correct wavelengths up to 3 µm in the visible, near-infrared, and shortwave infrared domains using first-principles atmospheric correction. Moreover, FLAASH includes additional features: (1) adjacency effect correction (pixel mixing from surface-reflected radiance scattering), (2) an option to calculate the amount of aerosol/haze and scene-average visibility, when dealing with especially stressful atmospheric conditions such as clouds, FLAASH employs the most cutting-edge methods, (3) cirrus and opaque cloud classification map, and (4) adjustable spectral polishing for cloud suppression(ENVI 2009).

LULC classification

LULC change detection is analyzed in GIS using the raw data from a preprocessed Landsat image as an input. Separate image classification is carried out for each of the four reference study years (1991, 2001, 2011, and 2021) by using a hybrid classification method, unsupervised followed by supervised techniques, to achieve better classification accuracy (Congalton 2015). The layout of the main applied research methods is presented in Fig. 2.

Fig. 2

Flow of the methodology applied

Under unsupervised classification, the iterative self-organizing data analysis (ISODATA) clustering algorithm is used to determine spectral classes. Then, supervised classification using the support vector machine (SVM) algorithm followed.

Unsupervised classification is sometimes called as clustering. This technique identifies similarities in the data. It groups far from one another and close to one another data types into separate clusters. There are no assigned class values in unsupervised learning. In this study as unsupervised learning the ISODATA algorithm is used. The ISODATA algorithm is a modification of the k-means clustering algorithm. In order to produce a best collection of output classes, the ISODATA classification technique combines additional steps for splitting, combining, and discarding trial classes with elements of the k-means approach. The unsupervised training helps in getting initial information on the data and aids in preparing the dataset ready for use in the supervised classification process (Siddiqui and Punia 2008; Rudrapal 2015).To improve the accuracy of the classification, following the unsupervised technique the study applied supervised classification. In the supervised classification, this study used Support Vector Machines (SVM) algorithm. In this process sample pixels of known identity are used to classify pixels of unknown character. Even with a very small amount of training data, SVM produces good results. The training algorithm bases on assigning fresh example into one of the two categories, making the decision simple and error free. To map samples of several classes with a distinct gap as large as feasible, SVM builds a hyper-plane. Every sample for a particular training set is labeled to fall into one of the two groups. SVM makes use of the kernel approach concept, which maps data via a non-linear transformation to a high dimensional space (Rudrapal 2015). In this study the Radial Basis Function (RBF) Kernel (Gaussian Kernel) is used. Cross-validation technique is applied in tuning the Kernel parameters.

As the SVM method needs only a few sample signatures (Awad and Khanna 2015), from the total of 592 ground truth reference data points, a total of 144 data points are used as training samples in the supervised classification for developing training sites to produce a signature of each land cover class. The identified training samples from each LULC were assigned to corresponding individual LULC classes before conducting supervised image classification. Then, the Landsat images were classified into seven major LULC classes (Table 2). To refine the classified images, patches of isolated pixels are recorded, and salt and pepper effects are removed using a classification sieve.

Table 2 UTRB LULC classes and their descriptions

Classification accuracy assessments

A total of 592 ground-truth data points (global positioning system (GPS) and Google Earth images) were collected in October 2020, assuming that there was no significant LULC change between 2020 and 2021. Out of the total ground-truth information, 448 are used to verify the accuracy of the classified images following the method of Rwanga and Ndambuki (2017). One popular way to show the results of how well the classification performed is to use a confusion matrix. Using the confusion matrix approach, which lists the reference data in columns and the classified data in rows, the classification accuracy results are estimated by computing the user accuracy, producer accuracy, overall accuracy (OA), and kappa coefficient (K). The overall accuracy and the kappa coefficient are calculated using Eqs. (1) and (2) (Foody 2002). According to Congalton (2015) Lu and Weng (2007), the difference between the real agreement of the classified map and the chance agreement of the random classifier compared to the reference data is measured using the kappa coefficient, which represents the total classification accuracy. The overall accuracy, on the other hand, represents the overall accuracy of the classification without going to how well the individual classes are classified. Individual class accuracies are measured using producer and user accuracies. Producer accuracy is based on commission errors (classifying an area to a class category where it does not belong), and user accuracy is based on omission errors (excluding a class category from the category to which it belongs) (Table 4).

$$ {\text{OA}} = \left( {{\text{X}}/{\text{Y}}} \right){*}100 $$

(1)

where OA is the overall accuracy, X is the number of correct points in the diagonals of the matrix, and Y is the total number of points taken as a reference/observation point.

$$ K = [N\mathop \sum \nolimits_{i = 1}^{r} x_{ii} - \mathop \sum \nolimits_{i = 1}^{r} \left( {x_{i + } *x_{i + } } \right)]/[N^{2} - \mathop \sum \nolimits_{i = 1}^{r} \left( {x_{i + } *x_{ + 1} } \right)] $$

(2)

where: K is the kappa coefficient, r is the number of rows and columns in the error matrix, N is the total number of observations, x_ii is the observation in row i and column i, x_i+ is the marginal total of row i, and x_+i is the marginal total of column i.

A kappa value greater than 0.80 indicates strong classification performance; a moderate classification performance is indicated by kappa values between 0.40 and 0.80; and a kappa value less than 0.40 indicates weak classification performance (Lillesand et al. 2004) (Table 4).

Post-classification and change detection

Using independently produced classified images and post-classification methods, change detection was conducted to assess the conversion of one LULC class to another class (Bogoliubova and Tymków 2014). In comparing the areas covered by different land cover types during the three different study periods, classification comparisons of land cover statistics are used to identify shift patterns, gains and losses (positive and negative, respectively) in each land cover class. Additionally, by applying Eqs. (3) and (4) to the data for each of the three time intervals (1991–2001, 2001–2011 and 2011–2021), as well as the overall study period, which spans from 1991 to 2021, the percentage change (%) and rate of change in hectares per year (ha/year) are calculated (Table 6). Positive values indicate an expansion in coverage, while negative values indicate a reduction.

$$ Percentage\;of\;change \left( {\Delta \% } \right) = [A_{initial} - A_{final} /A_{initial} ]* 100 $$

(3)

$$ Rate\; of change = {\raise0.7ex\hbox{${\left( {A_{final } - A_{initial } } \right)}$} \!\mathord{\left/ {\vphantom {{\left( {A_{final } - A_{initial } } \right)} N}}\right.\kern-0pt} \!\lower0.7ex\hbox{$N$}} $$

(4)

where: the rate of change is in ha/year; “A” is the extent of each LULC type in ha; and N is the time interval between the initial and final years.

Results

Land use/land cover maps

Following the methodology layout in Fig. 2, four LULC maps for 1991, 2001, 2011, and 2021 that show seven major LULC classes (barren land, farmland, forest, grassland, shrub land, settlement, and water body) are produced (Figs. 3, 4, 5 and 6 and Table 2).

Fig. 3

LULC map for 1991

Fig. 4

LULC map for 2001

Fig. 5

LULC map for 2011

Fig. 6

LULC map for 2021

These LULC classes remained the dominant LULC classes during the study period, showing marked changes in the area coverage within them. The area covered by barren land, farmland, forest, grassland, settlements, and water bodies in 1991 was 0.72%, 44.47%, 7.49%, 25.44%, 0.09%, 21.6%, and 0.19% of the total area, respectively. At the end of the study period, 2021, the areas covered by barren land, farmland, forest, grassland, settlements, and water bodies were 2.02%, 67.73%, 1.83%, 15.68%, 0.58%, 11.82%, and 0.34%, respectively (Table 3).

Table 3 LULC area coverage in hectares (ha) for 1991, 2001, 2011 and 2021

Accuracy assessment results

The accuracy of the image classification is evaluated using a confusion matrix, as discussed in Sect. "Classification accuracy assessments" above. Using the confusion matrix approach accuracy assessing parameters, the user’s accuracy, the producer’s accuracy for each LULC class, the overall accuracy (OA), and the kappa coefficient are calculated using Eqs. (3) and (4). The user’s accuracy and producer’s accuracy calculated for each of the LULC classes for the 2021 map are within the acceptable range (Table 4). In addition, the overall accuracy and the kappa coefficient of the 2021 map are 91% and 89%, respectively. These result is strong which indicates that the classified images practically match the ground truths. As discussed in Sect. "Classification accuracy assessments", a kappa coefficient exceeding 0.80 indicates a strong classification performance (Lillesand et al. 2004) (Table 4).

Table 4 Confusion matrix for 2021 classified LULC map

Post-classification and change detection

Throughout the three decades, there was a significant change in the area coverage of the identified LULC class types. The seven main LULC types found in the study base year (1991) continued to be major LULC features for the three-decade study period. However, during the study period, there were different degrees of gains and losses within the seven identified LULC classes in the basin (Table 5).

Table 5 LULC change (%) and annual rate of change during 1991–2001, 2001–2011, 2011–2021 and 1991–2021

Gain and loss analysis

This study closely examined each LULC loss and gain over the three-decade study period. According to the study results, forestland decreased significantly by 75.55% from 1991 to 2021 as a result of the drastic expansion of settlements and farmland by 570.31% and 52.32%, respectively, during the same period. The statistics regarding the gains and losses in Km² are presented in Table 6. The greatest increase (19.37%) in farmland area occurred during the first study phase (1991–2001); on the other hand, the greatest decrease (18.66%) in grassland area occurred during the second phase (2001–22011). The third phase exhibited the greatest loss (24.42%) in shrub-land. Based on the transition area matrix, Table 6 calculates the changes between the different LULC categories and the corresponding gains and losses during the study period (1991–2021).

Table 6 LULC transition area matrix (km²) during 1991–2021

The transition area matrix, which displays the causes of change as well as the direction of change over the course of the three distinct periods and the entire study period, is calculated. According to the transition matrix for the 1991–2021 research period (Table 6), the UTRB underwent considerable changes from one LULC category to another during the 1991–2021 period. The transition area matrix for each of the study periods (1991–2001, 2001–2011, and 2011–2021) are attached as additional materials.

Patterns of LULC Shift, 1991 to 2021

The LULC transition/shift trend in the last three decades in the UTRB has been continuous, with sharp declines in forests, shrublands, and grasslands. Forests declined at rates of 13,627.58 ha/year, 6640.66 ha/year, and 5590.73 ha/year during 1991–2001, 2001–2011 and 2011–2021, respectively, and the average annual decline rate over the 30-year period was 8619.66 ha/year. Similarly, shrub-land declined at rates of 12,288.58 ha/year, 14,940.24 ha/year, and 17,449.15 ha/year during 1991–2001, 2001–2011, and 2011–2021, respectively, with an average continuous annual rate of decrease of 14,892.66 ha/year over the 30 years. Grassland declined at rates of 13,315.28 ha/year, 19,207.69 ha/year, and 12,069.52 ha/year during 1991–2001, 2001–2011, and 2011–2021, respectively, at an average continuous annual rate of decrease of 14,864.17 ha/year. In contrast, farmland increased continuously at average annual rates of 39,351.26 ha/year, 37,359.48 ha/year, and 29,599.48 ha/year during 1991–2001, 2001–2011, and 2011–2021, respectively, and the average annual increase rate was 35,436.74 ha/year during 1991–2021, Table 4 and Fig. 7.

Fig. 7

LULC change trend (%) during 1991–2021

Discussion

This study of the UTRB covers an area of 4,569,458 ha of land. According the overall accuracy and kappa coefficient results the classification performance is strong, which indicates classified images to practically match the ground truths. This study identified seven major LULC classes in the basin and assessed the rate, trend, and magnitude of land cover changes over three decades (1991–2021). Due to human activities such as deforestation, increased cultivation/agriculture, and settlement projects, the UTRB experienced a substantial degree of LULC change. This is clearly seen visually comparing the maps of 1991 and 2021, Figs. 3 and 6. Changes in the LULC across the UTRB, a basin with an erratic highland terrain, caused by human activity have the potential to have detrimental effects on the basin's natural hazards, water resources, and socioeconomic dynamics. For this reason, LULC change time series modeling is essential for many applications, such as those pertaining to land use planning, protection and management of landslides, erosion, and basin potential water resource management.

Over the three decades between 1991 and 2021, the upper Tekeze River basin experienced a considerable transition in terms of the major LULC classes. According to the LULC change analysis in the study area, settlements expanded by 570.31% in parallel with the expansion of agricultural land by 52.32%. In contrast, forestland decreased significantly by 75.55% from 1991 to 2021 as a result of the drastic expansion of settlements, barren land, and agricultural/farmland during the same period. This could have resulted from the increased pressure that an increasing number of people and animals are exerting on the environment, which also likely contributed significantly to the rapid growth of agricultural lands. The increase in water body cover could mainly be due to the expansion of artificial water bodies, such as the construction of the 300 megawatt electricity Tekeze hydrodam in 2009, the highest arch dam in Africa, Figs. 5 and 6.

The wide-ranging agricultural land expansion and settlement of the UTRB caused by the loss of forest, grass, and shrub areas is evidence of the impact that people have had on the UTRB environment. Furthermore, the results of this study clearly showed that one of the events that led to the region's environmental degradation and water scarcity was the LULC shift caused by human activity. The LULC transitions we observed in the study area were in agreement with the findings of earlier studies carried out in the Upper Blue Nile basin and elsewhere in the nation (Girma et al. 2022; Berihun et al. 2019; Regasa et al. 2021; Gashaw et al. 2017; Hassen et al. 2021).

Conclusions

The general conclusion drawn from this study is the occurrence of widespread deforestation and natural vegetation cover degradation in the UTRB. This occurrence could be classified as human-induced LULC change. The frequency and severity of extreme weather events and natural disasters can be greatly influenced by changes in land use and land cover, such as settlement/urbanization and deforestation. Deforestation and land degradation can lead to increased frequency of extreme weather and natural disaster such as increased soil erosion, reduced water infiltration, and a higher risk of floods.

These detrimental outcomes could contribute to extreme poverty and food insecurity in the area. Additionally, the deterioration of natural plant cover and deforestation may have major worldwide effects on future increases in carbon stocks and negative effect in regional and global emissions reduction contribution.

This study's findings demonstrated that the UTRB is what it is now because of the decades of LULC management; it is now up to the LULC management of the future to improve the UTRB's contribution to the environment and the community lively hood that depends on it. Therefore, to better manage resources and implement conservation measures in the basin, LULC must be improved.

To limit LULC conversions and maintain a balance between food security, socioeconomic development, and sustainable natural resource management, we suggest the development and implementation of effective regional and national land use policies in Ethiopia. This necessitates active community involvement, stakeholder engagement, and appropriate laws and local regulations that cascade the policies.

In this regard, this study has assessed the change in the land and its attributes resulting from the impact of human and natural activity. The findings of this study will be useful input for government institutions, policymakers, and researchers in natural resource administration, land planning, targeting and allocating resources, and other land use interventions and studies.

Moreover, to better understand and respond to overall LULCC dynamics, we suggest that future studies focus on the details and interactions between LULC change, related drivers, and pertinent actors. This is important because, although LULC change is a global issue, local factors have led to inconsistent and diverse dynamics in the process.