Trend analysis, past dynamics and future prediction of land use and land cover change in upper Wabe-Shebele river basin

Abstract

A growing population has led to extensive farming at the expense of a natural environment. Changes in land use and cover have caused land degradation, and problematic groundwater recharge. The objective of this study was to evaluate the historical trend, simulations, and predictions of land use land cover change in the Upper Wabe-Shebele River Basin. The study accounted for 1992, 2007 and 2022 as well as it will predict the change for 2037 and 2052. Landsat TM for 1992, ETM + for 2007, and Landsat-8 OLI for 2022 were used. In QGIS 3.16, the maximum likelihood method was utilized for supervised image classification. Using CA-Markov and the Land Change Modeler land use and land cover change for 2037 and 2052 were predicted. Validity and accuracy of the model was evaluated using actual and predicted land use and land cover changes of 2022. Topography, proximity to a town, stream, roads, and population density were used as input for the model. The results showed that between 1992 and 2007, cultivated land increased by 17.07% on average at a rate of 1.05%, while settlement increased by 17.51% at a rate of 1.08% per year. Agricultural and settlement land increased by 22.97% and 30.12%, respectively. Between 1992 and 2022, the transition area matrix showed 2,330.25 and 1,145.77 km² of forest and grazing land were changed to settlement and cultivated land, respectively. Meanwhile, from 2022 to 2037, the quantity of land used for cultivated, grazing, and settlement is predicted to increase by 0.19, 3.66, and 23.8% in order. For 2037 and 2052, settlement and cultivated land were increased by 1.3 and 7.32% respectively. Finally, since natural ecosystem had been significantly disturbed by change in the study area, comprehensive rehabilitation and management is demanded.

1. Introduction

Land is an important, limited natural resource that could be modified by natural and anthropogenic activities [1]. These are intensifying large-scale modifications of land use and land cover. Land cover is defined as the observed bio-physical cover on the earth's surface, whereas land use is defined as the arrangements, activities, and inputs that people undertake on a certain land cover type [2]. Both of them have a significant impact on water availability, land productivity, and tourist attractions, which can generate socio-economic and environmental concerns [3]. It is caused by an increasing human population, industrialization, and urbanization [4]. They are a continuous and interlinked process that needs to be understood from different perspectives. Meanwhile, land cover and land use represent the assimilating elements for the resource base [[5], [6], [7], [8]]. In Africa, dramatic land use and land cover changes (LULCC) were observed from 1975 to 2000 [3]. It mostly indicated the transformation of grassland, woodland, bushland, and other vegetation into agriculture and settlement areas [9,10]. As a result, land use and land cover (LULC) refer to a combination of existing alternative land uses and the type of ground cover under interaction of natural environment and human activities which has a crucial impact on global environmental changes and sustainable development [[11], [12], [13]]. In Ethiopia, LULCC and its dynamics have also followed a similar trend, with significant implications for natural resource degradation and the loss of ecosystem services [[14], [15], [16]]. In the country, LULCC is mostly subjected to the conversion of natural vegetation cover to agricultural land [17,18].

In addition, for sustainable land management, apart from historical and existing LULCC the scenarios prediction also needs more attention [19]. It helps to understand the hydrologist's and environmentalist's perspective on past changes and potential future dynamics in order to make informed decisions at various levels. Some of the models that can be used for LULCC simulation and prediction include the Markov chain, the cellular automata model, and neural networks [[20], [21], [22]]. Cellular-automata-based models use statistical analysis for modeling, Markov chain models use artificial neural networks, and agent-based modeling has been widely utilized in land use change modeling [21,23]. According to Ref. [24]was the top researcher who proposed Markov chain analyses for modeling land-use change. Such analyses are based on the core principle of the continuation of historical development.

The major drivers for LULCC are topography, population density, government policy, infrastructural expansion, and water resource availability [19,22,[25], [26], [27], [28]]. Earlier studies have emphasized that the availability of a better market and road infrastructure were the major reasons underlying LULC modifications [29,30]. Those require additional consideration when simulating changes in land use and land cover. Using the concepts gained from the base year, it attempts to predict the spatial distribution of the particular land cover and land use classes in a subsequent year [8].

The study area, Wabe-Shebele river basin, is the most water-scarce in the country, and agriculture is the basis of the economy [31]. It is characterized by small-scale and rainfed agriculture that is reliant on inconsistent rainfall. Hence that to address the needs of a growing population, extensive agriculture is being practiced on the expense of forest, water body and grass land. It is mostly related to the fact that growing of crops twice or more in a year is difficult due to water scarcity and erratic rainfall [31,32]. Traditional and small-scale irrigation, particularly in the Adaba, Ticho, and Arsi Robe districts, is replacing the ecosystems of riparian regions along the Wabe-shebele main river, its tributaries, and wetlands [31]. Similarly, agricultural investments and town expansions are the major factors influencing the land use land cover in the study area. Access to land, including land availability and individual plot size induced the decision to outmigration to towns which directly bring town expansion. As a result of the aforementioned pressure, land use land cover change is prevalent in the studied area. In Malka Wakana catchment, there is dwindling forest coverage, while major agricultural expansions have a key impact on dramatic LULCCs [32].

An investigation of the LULC changes in the basin observed agricultural expansion as the main environmental challenge [31]. But within population growth, water scarce, settlement and agricultural expansion, LULC changes in the basin have also received little attention yet. Similarly, no study has attempted to predict the LULCC dynamic using CA-Markov chain models, which can take into account both physical and socioeconomic factors. It is needed in order to explore practical answers to issues related to LULC changes. It assists in calculating both the rate of conversion between different classes and the temporal change of conversion between LULC classes [20,23,27]. Consequently, it is a stochastic process in a discrete space (i.e., mainly based on probabilities, not certainties). The prediction of future changes also generates a transition probability matrix of LULCC using the base and later periods [8].

Such a study can be used to understand how the local community, government, and other stakeholders could make land-use decisions in alignment with specific biophysical and socioeconomic factors that interact to influence the decisions [29]. For the implementation of land management, detailed information on LULC is needed for modeling and monitoring environmental dynamics. It can also support decision-makers in the preparation of short-, medium-, and long-term plans for the conservation and sustainable use of natural resources [8,10,33,34]. Furthermore, it serves as a baseline for understanding trends and potential causes of LULCCs, as well as their implications. Therefore, the objective of the study was to quantify historical trends, simulate, and predict LULCC in the Upper Wabe-Shebel River Basin.

2. Materials and methods

2.1. Descriptions of the study area

2.1.1. Location and map of the study area

The study area is a sub-basin of the Wabeshebele River Basin, in southeast Ethiopia. It has a total area of 10,259.03 km². The study area is located between 39^o24′30.65″to 39^o29′49.13″E and 6^o53′44.63″ to 7^o27′37.75″N (Fig. 1). It comprises six districts from the West Arsi zone: Adaba, Dodola, Asassa, Kofale, Kokosa, and Kore; eight districts from the Arsi zone: Limu Bilbilo, Inkolo Wabe, Shirka, Tana, Arsi Robe, Hamda Diksis, Amigna, and Bele Gesgar; and four districts from Bale Dinsho and Agarfa [35].

Fig. 1.

Map of the study Area.

2.1.2. Biophysical characteristics

Agroecology: The minimum and maximum annual temperatures of the study area are found between 2 and 15 °C in the higher altitude areas and between 16 and 24 °C in the lower plateau areas. However, there is variation in temperature; October to May is the hottest month, while June to September is the coldest month [32]. The study area is categorized into the agro-climatic regions of Dega, Weina Dega, and Kolla. From Dega to Weina Dega, the frequency and length of rainy days and seasons vary, and this number slightly declines as one descends to the kolla regions along the Wabe river course after the dam to Amigna district. The majority of the time, the main rainy season starts in June and lasts through July and August. In certain higher elevations, it also lasts through September. The mean annual rainfall of the study area is 425 mm, ranging from 900 to 1300 mm.

Topography and soil/geology: According to a geological assessment, the research area's terrain is comprised of plains, hilly valleys, gorges, the highest peaks, and dissected plateaus. The Melka-Wakena dam sub-station is located close to the catchment's minimum elevation of 2350 m above sea level and a maximum elevation of 4322 m (Kaka Mountain peaks). The basin is provided naturally with numerous rivers and streams as well as one man-made lake. Totolamo and Ashoka from Kofale; Ukuma, Lensho, Kesa, and Negesso from Dodola; Maribo, Furuna, Nanisha, and Ashiro from Adaba; and Wekentera, Geredela, Uruba, Ubulto, and Debara from Gedeb-Assass and Melka Wekena are the intermittent and perennial rivers that drain into the study catchment from each district [31,32].

The dominant soil types in the study area are Vertisols, Chernozems, Cambisols, Luvisols, and Nitosols. Particularly, dominance Cambisols and chernozems in Adaba and Gadeb-Hassassa; Luvisols and chernozems in Dodola; and Utric Nitosols and Pellic Vertisols in Kofale district [32].

Natural Vegetation: The nature and distribution of the nature of vegetation of this district range from wooded grassland to afro-alpine. Alpine, Afro, and sub–Afro Alpine vegetation is found in the area above 3100 m sea level of the area. Abundant low-growing bush grasses and lichens are common species on the top of the mountain, where the temperature is very low. Below the Afro-Alpine and sub-Alpine broad-leaved forests, which are dominated by Juniperus, Podocarpus, and Hagena Abyssinica tree species, as well as shrub and bush communities that are highly dominated by “Asta” species, are found parts of the Adaba and Dodola districts. Pockets of scrattered woodlands and mixed with eucalyptus trees can also be found in the Kofale district's western and northwestern parts. Very limited natural bush, shrubs, and spare forest (remnant tree species) are found at the highest altitude in the north and northwest parts of the Gadeb-Assassa district [36,37]. The diverse climate and topographic phenomena have provided a wide range of natural environments, which form favorable habitats for a wide variety of fauna in the study catchment. The local inhabitants rely on the forest to supply most of their needs, mainly fuel wood, pasture, timber, wild fruits, and medicinal herbs [37].

2.2. Methods of data collection

2.2.1. Data acquisition and processing

The study included both primary and secondary data collection. Primary data was produced by the analysis of field observations, Ground Control Points (GCPs), and 1992, 2007, and 2022 satellite images (Table 1). The main justification for selecting these years is that, following the change in government in 1991, new proclamations on land administration began to be implemented in 1992. Additionally, the Wabe-Shebele River basin master plan implementation from 2004 was started by various stakeholders in 2007 [31]; the 2022 year was selected since it is the most recent data. The USGS website was utilized to download these satellite images. Then satellite images were processed, classified, and analyzed using QGIS 3.16. As a result, each tile with a different path and row was re-projected, and then merge tiles were added. Then part of the mosaic was clipped by the shape file of the study area for further analysis. In order to remove disturbances and radiometric variation between acquisition dates, sub-setting of image preprocessing was done in order to Ortho-rectify the satellite images into Universal Transverse Mercator (UTM) (WGS, 1984) (coordinates -WGS 37 N; Spheoid - Clarke 1880); and datum -Adindum) [[38], [39], [40]].

Table 1.

Description of the satellite images used for the study.

Satellite images	Path/Row	Acquisition date
Landsat TM for 1992	168/055, 168/054, 167/055,167/054	January 26, 1992
Landsat ETM+ for 2007	168/055, 168/054, 167/055,167/054	February 14, 2007
Landsat-8 OLI for 2022	168/055, 168/054, 167/055,167/054	January 02, 2022

One FGD per district (with 12 participants in each of the 17 districts) and three key informant interviews per district Based on the field observations, FGD, and key informant interviews, water bodies, forests, grazing, cultivating, and settlement land were identified as major LULC types in the study area (Table 2). In addition during the desk review these land uses and land cover predominated in Ethiopia's highlands [[41], [42], [43]].

Table 2.

Major LULC classes and their descriptions in the study area.

No	LULC Class	Description
1.	Water bodies	Rivers, streams, permanent open water, lakes, ponds, wetlands, reservoirs
2.	Forest Land	Protected forests, plantations, deciduous forests, mixed forest lands, and forests on customary land
3.	Grazing Land	This class refers to an area covered with grass that is used for grazing
4.	Cultivated land	cultivated and uncultivated agricultural lands areas, such as farmlands, crop fields including fallow lands/plots, and horticultural lands
5.	Settlement	Residential, commercial and service, industrial, socioeconomic infrastructure, and mixed urban and other urban, transportation, roads

Sources: [31,[41], [42], [43]]; FGD, KII and Field observation

A hybrid classification system that involves unsupervised classification followed by supervised classification methods was used to reduce confusion. During unsupervised classification, the ISODATA (Iterative Self-Organizing Data Analysis Technique) algorithm was used to generate a natural cluster composed of multiple classes, with the maximum number of iterations set to 10 at a 95% confidence level. The easily identifiable classes were categorized into land cover units to create appropriate signatures for the supervised classification, whereas unidentifiable clusters were excluded from the signature categorization. In this case, unsupervised classification was only employed for identifying LULC classes and preparing signatures for prior or historical land use land cover. This was accomplished with combination of key informant and elders interview result on land use land cover classes and literatures data. These approaches have significant similarity with procedures followed by Refs. [44,45].

Subsequently, in supervised classification, GCPs were used for signature preparation. The district -level topo-map also helped with training site selection and georeferencing. In order to define AOI and signatures for known classes, combinations of unsupervised classification, GCPs, Google Earth images, and topo-maps were used. After a number of band combinations, the final signatures were demarcated for the supervised classification. The images were finally categorized using maximum likelihood.

2.2.2. Accuracy assessment of the classification

The most important step for image classification is the analysis of the accuracy assessment for LULC change. A total of 162, 176, and 200 GCP were collected for the years 1992, 2007, and 2022, respectively, using a random sampling approach. In order to accomplish this, GCP for all five LULC classes was triangulated using signatures produced from unsupervised classification, archival Google Earth photos, FGD, and key-informant interviews. Plus, this direct observation was used for the 2022 satellite image. Based on the GCPs, the classified LULC was compared with the ground truth/reference image. This was done to investigate how the result reflects the reality on the ground [46]. The accuracy of the classification was assessed by the producer's accuracy (PA), the user's accuracy (UA), and overall accuracy (OA). An error matrix was generated for the identified LULC types. Finally, the LULC accuracy level was indicated. The same strategy was used [47].

2.2.3. Land use and land cover change accuracy assessment and analysis

The overall LULC classification accuracy for the years 1992, 2007, and 2022 is 0.81, 0.88, and 0.91, respectively (Table 3).

Table 3.

Accuracy assessment and performance criteria for three periods of LULCC.

LULC Classes	1992		2007		2022
	PA	UA	PA	UA	PA	UA
Water Body	0.93	0.88	0.93	0.82	0.88	0.93
Forest Land	0.9	0.88	0.92	0.92	0.9	0.90
Grazing Land	0.83	0.83	0.82	0.84	0.94	0.96
Cultivated Land	0.86	0.93	0.83	0.89	0.88	0.93
Settlement	0.89	0.84	0.91	0.87	0.93	0.87
OA	0.81		0.88		0.91

2.2.4. Post-processing and change analysis

Using the overlay functions in ArcGIS, the PCC approach produced the LULCC transition matrix. Gross gains and losses were also calculated for six periods: 1992–2007, 2007–2022, 1992–2023, 2037–2022, and 2022–2022. Then the area covered, the percentage change, the annual rate of change, gain, and losses were calculated for each of the identified LULC types (Eqs. (1), (2), (3), (4), (5))). The following equation was used to determine the LULCC [48].

$$\mathrm{T}\mathrm{e}\mathrm{m}\mathrm{p}\mathrm{o}\mathrm{r}\mathrm{a}\mathrm{l}\mathrm{L}\mathrm{U}\mathrm{L}\mathrm{C}\mathrm{c}\mathrm{h}\mathrm{a}\mathrm{n}\mathrm{g}\mathrm{e}=\left(\frac{\mathrm{A}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{o}\mathrm{f}\mathrm{f}\mathrm{i}\mathrm{n}\mathrm{a}\mathrm{l}\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}-\mathrm{A}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{o}\mathrm{f}\mathrm{i}\mathrm{n}\mathrm{i}\mathrm{t}\mathrm{i}\mathrm{a}\mathrm{l}\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}}{\mathrm{A}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{o}\mathrm{f}\mathrm{i}\mathrm{n}\mathrm{i}\mathrm{t}\mathrm{i}\mathrm{a}\mathrm{l}\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}}\right)\times 100$$ (1)

$$\mathrm{R}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{o}\mathrm{f}\mathrm{c}\mathrm{h}\mathrm{a}\mathrm{n}\mathrm{g}\mathrm{e}\left(\mathrm{R}\right)\left(\mathrm{h}\mathrm{a}/\mathrm{y}\mathrm{r}\right)=\frac{1}{\left({\mathrm{t}}_2-{\mathrm{t}}_1\right)}\ln \left(\frac{{\mathrm{A}}_2}{{\mathrm{A}}_1}\right)$$ (2)

$$\mathrm{G}\mathrm{a}\mathrm{i}\mathrm{n}=\mathrm{A}\mathrm{g}\mathrm{y}-\mathrm{A}\mathrm{s}\mathrm{y}$$ (3)

$$\mathrm{L}\mathrm{o}\mathrm{s}\mathrm{s}=\mathrm{A}\mathrm{i}\mathrm{i}\mathrm{y}-\mathrm{A}\mathrm{s}\mathrm{y}$$ (4)

$$\mathrm{T}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}\mathrm{g}\mathrm{a}\mathrm{i}\mathrm{n}/\mathrm{l}\mathrm{o}\mathrm{s}\mathrm{s}\mathrm{o}\mathrm{r}\mathrm{N}\mathrm{e}\mathrm{t}\mathrm{C}\mathrm{h}\mathrm{a}\mathrm{n}\mathrm{g}\mathrm{e}=\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{o}\mathrm{f}\mathrm{f}\mathrm{i}\mathrm{n}\mathrm{a}\mathrm{l}\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}-\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{o}\mathrm{f}\mathrm{i}\mathrm{n}\mathrm{i}\mathrm{t}\mathrm{i}\mathrm{a}\mathrm{l}\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}$$ (5)

$$\mathrm{A}\mathrm{g}\mathrm{f}\mathrm{y}=\mathrm{T}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{o}\mathrm{f}\mathrm{s}\mathrm{p}\mathrm{e}\mathrm{c}\mathrm{i}\mathrm{f}\mathrm{i}\mathrm{c}\mathrm{c}\mathrm{l}\mathrm{a}\mathrm{s}\mathrm{s}\mathrm{i}\mathrm{n}\mathrm{t}\mathrm{h}\mathrm{e}\mathrm{f}\mathrm{i}\mathrm{n}\mathrm{a}\mathrm{l}\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}$$

$$\mathrm{A}\mathrm{i}\mathrm{i}\mathrm{y}=\mathrm{T}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{o}\mathrm{f}\mathrm{s}\mathrm{p}\mathrm{e}\mathrm{c}\mathrm{i}\mathrm{f}\mathrm{i}\mathrm{c}\mathrm{c}\mathrm{l}\mathrm{a}\mathrm{s}\mathrm{s}\mathrm{i}\mathrm{n}\mathrm{t}\mathrm{h}\mathrm{e}\mathrm{i}\mathrm{n}\mathrm{i}\mathrm{t}\mathrm{i}\mathrm{a}\mathrm{l}\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}$$

$$\mathrm{A}\mathrm{s}\mathrm{y}=\mathrm{U}\mathrm{n}\mathrm{c}\mathrm{h}\mathrm{a}\mathrm{n}\mathrm{g}\mathrm{e}\mathrm{d}\mathrm{a}\mathrm{r}\mathrm{e}\mathrm{a}\mathrm{o}\mathrm{f}\mathrm{t}\mathrm{h}\mathrm{e}\mathrm{s}\mathrm{a}\mathrm{m}\mathrm{e}\mathrm{c}\mathrm{l}\mathrm{a}\mathrm{s}\mathrm{s}\mathrm{i}\mathrm{n}\mathrm{t}\mathrm{h}\mathrm{e}\mathrm{s}\mathrm{a}\mathrm{m}\mathrm{e}\mathrm{y}\mathrm{e}\mathrm{a}\mathrm{r}$$

Where: A₁ and A₂ are land cover at initial (t₁) and next time step (t₂), respectively.

2.2.5. Land use and land cover change modeling

With the help of TerrSet software, a CA-Markov model was used to simulate and project the LULCC. It is a methodology and one of the tools used in LULCC modeling. It is expressed as mainly focusing on both spatial and temporal changes [49]. It is one of the planning supports tools for the analysis of temporal changes and the spatial distribution of LULC [50]. It is also important for land use policy design and planning and the objectives of sustainable land use development [20,27]. Therefore, studying the historical LULCC is necessary to comprehend how humans interact with the environment [51].

The CA-Markov model was selected because it combines the modeling techniques of Markov chains with cellular automata. Firstly, important vector data were converted into raster data. Secondly, the spatial distribution of land use is achieved by the Markov model analysis of land use trends and the application of CA modeling. This was done based on the CA Markov and TerrSet 2020 software, a GIS, and an image processing module. The two primary operations of the CA-Markov model used in this study were the computation of the transition matrix and the prediction of the LULC mappings. On the basis of the previous state, the transition rules of Markov chain analysis for future land use changes were established and predicted. The observed transition probabilities between maps in the 1992, 2007, and 2022 matrices were used to express this. These steps were used before by Refs. [39,46].

Spatial overlay analysis is used to calculate the transition probability matrix and the transfer area of the matrix. The baseline map for the prediction was the superimposed land use maps of 2022 and 2007. The estimated transition probability matrix-assisted transformation rules are used to implement CA-Markov model simulations. The next step was to determine CA filters that can clearly produce the space weighting factor and can be adjusted in accordance with the current nearby cellular status. The neighborhood is defined in this study using the common 5*5 contiguity filter. A matrix space made up of 5*5 cellular units surrounds each cellular center in order to significantly influence cellular changes.

The starting point (2022) and the number of iterations for the CA were finally established. In order to simulate the landscape spatial pattern for the study area, the number of CA iterations was increased to 15 based on the time between the first and second LULCC analyses.

2.2.6. Period of prediction, validation, and integration of the model

The baseline LULC map of 2022 was simulated using the LULC map from 2007, which was used for future projection. The performance of LULCC predictions was then assessed using the Kappa Agreement Index (KAI) principles in IDRISI Andes' VALIDATE module (Eq. (6)). Then, the simulated 2022 LULC map and the actual 2022 LULC map were compared using IDRISI's CROSSTAB Module and Kappa Agreement Index. The components of KAI that were finally taken into account for the study were Kappa for no information (Kno), Kappa for location (Klocation), and Kappa for standard (Kstandard), with the following statistics [4].

$$K_{no},K_{location},K_{s\tan dard}=\frac{\left(M_m*N_n\right)}{P_p-N_n}$$ (6)

Whereas, N (n), M (m) and P (p) define as no information, medium grid cell-level information and perfect grid cell-level information across the landscape respectively

The Kappa agreement indices between 0 and 1. A value of Kappa below 0.4 indicates less precision and less consistency; when 0.4 ≤ Kappa ≤0.75 the accuracy is moderate; and when Kappa is greater than 0.75, there are small differences and a high level of consistency between the two LULC maps and Kappa = 1 for perfect agreement [4,47].

In order to predict future LULC maps for 2037 and 2052, the simulated LULC maps of 2022 were used as a baseline map. This was accomplished through the figure of merit (FOM) feature of Terrset2020 Software's and CA-Markov algorithm (Eq. (7)). The intersection of the simulated and reference changes forms the numerator of the ratio FOM, and the union of the two constitutes the denominator. The FOM is equal to 0 if there is no overlap between actual and expected changes and 1 if there is total overlap, according to Refs. [52,53] studies.

$$FOM=\frac{H}{\left(H+M+FA+WH\right)}*100$$ (7)

Whereas M = Misses or area of error due to reference change simulated as persistence; H= Hits or area of correct due to reference change simulated as change; WH= Wrong Hits or area of error due to reference change simulated as a change to the wrong category; FA = False Alarms or area of error due to reference persistence simulated as change.

2.2.7. Driver variables for land use and land cover change

The primary contributing variables to the change should be taken into account during LULCC simulation. The LULCC is impacted by proximity to rivers and/or streams, highways, and towns, which creates a welcoming and accessible environment for the local community. Along with population density, topography, especially elevation and slope, is dramatically changing the types of land that are used and covered. There is an expansion of agriculture, settlement, industries, and towns as the population grows. As a result, the existing state of land use and cover is modified and converted. Hence that medium elevation and slope are ideal for human activities due to a number of determining variables. Because of this, land use and cover within certain areas were required to adjust in accordance with the needs of land users.

The result of the model was validated using a simulated 2022 LULC map. The CA-Markov procedure can be used to continue the validation test and provide prediction of LULC map for the years 2037 and 2052. The validation test is quantified by the Kappa Index of Agreement (KIA). The CA-Markov model's input drivers for land-use change took into account elevation, slope, and distance from roads, rivers or streams, and towns (Fig. 2, Fig. 3, Fig. 4, Fig. 5, Fig. 6). The evidence likelihood is shown as an empirical probability of change in the LULC categories between an earlier and a later map [54]. It is used to convert LULC class categorical variables into numerical values that belong to a different class.

Fig. 2.

Euclidean distance of river in UWSRB

Fig. 3.

Euclidean distance of roads in UWSRB

Fig. 4.

Euclidean distance of towns in UWSRB

Fig. 5.

Slope gradients in UWSRB

Fig. 6.

Elevation distribution in UWSRB.

The DEM 12.5 m resolution, which was obtained from the Ethiopian Geospatial Information Agency, was used to calculate elevation and slope. The data of roads, streams, and towns were obtained from Ethiopian Road Authority, the Ministry of Water and Energy, and the Ministry of Urban Development and Construction, respectively. The study area shapefile was used to clip them, and topographic and proximity variables were prepared as suit to CA-Markov model. For additional analysis, procedures consist converting a shapefile to a raster, projecting the output raster, resampling it, measuring its Euclidean distance, extracting it using a mask, reclassifying it, and finally converting the raster to Ascii to import the file into the TerrSet 2020 program. Cramer's V Test in Terrset software was used to determine the importance of the driving forces' influence on land-use change. The significance level and correlation between the two groups of variables are assessed and evaluated. For variables association measurements, a total Cramer's V result of 0.15 to 0.4 is considered acceptable; if it is larger than 0.4, noble consideration is given [27].

Population data and growth rate were taken from reports at the national level and computed in accordance with the Ethiopian Central Statistics Service report [55]. The population growth rates were calculated using exponential growth, and between 1992 and 2007, 2007–2022, and 2022–2037, they were 2.97, 2.85, and 3.65%, respectively (Table 4). Using the year 2022 as a baseline, the estimated population for the year 2037 was 4,700,734. The rate between two periods, which is 15 years, is set.

Table 4.

Population statistics in the study area.

Year	1992	2007	2022
Population number	1,471,908	2,128,149	3,038,059
Growth Rate (%)	2.97	2.85	3.65

The growing rates were designed on the bases of the work of [56]with the assumption of exponential growth in Equation [8]:

$$P=P_0*e^{rt}$$ (8)

Where P = Total population after time “t”; P₀ = starting population; r = % rate of growth (Mean annual growth rate); t = in Years (15 years) and e = Euler number (2.71828).

3. Result and discussions

3.1. Analysis of land use land cover change from 1992 to 2022

The results of this study showed that five major LULC classes, i.e., forest land, grazing land, water bodies, cultivated land, and settlements, were identified in the study area. The spatiotemporal LULC classification maps for the years 1992, 2007, and 2022 are shown in Fig. 7, Fig. 8, Fig. 9. In 1992, cultivated land made up 6570.20 km² (64.04%) of all land covered, followed by 2471.84 km² of forest land, 908.58 km² of grazing land, 205.79 km² of settlements, and 102.61 km² of water bodies. Comparable to 1992, in 2007 there were 7691.53 km² (74.97%) of land that was under cultivation, followed by 1554.58 km² of forest land, 690.42 km² of grazing land, 241.83 km² of settlements, and 80.67 km² of water bodies. Between 1992 and 2007, each of the areas under cultivation and settlement increased by 1121.33 km² (14.58%) and 36.04 km² (14.9%), respectively (Table 5) [57]. reported a significant decline in forest cover in the South-Central Rift Valley between the late 1800s and the 1930s. On the contrary [58], reported an increase in forest cover and a decrease in cultivated land between 1973 and 2015. The result of this finding is also in agreement with [29], who reported that large-scale agriculture is the primary global source of forest degradation. Particularly in Ethiopia, recent studies revealed a substantial expansion of the area used for settlement and agriculture at the expense of forests, water bodies, and grazing areas [7]. In the Northwest lowlands of Ethiopia, it was stated that between 1985 and 2010, agricultural and settlement lands increased while forest land and water bodies declined [14].

Fig. 7.

LULC map of 1992 in UWSRB

Fig. 8.

LULC map of 2007 in UWSRB

Fig. 9.

LULC map of 2022 in UWSRB

Table 5.

Historical Land Use Land cover Changes trend and Annual rate between 1992, 2007 and 2022.

LULC Classes	1992		2007		2022		Net Change (km2)			Percentage Change (%)			Rate of change per year
	km2	%	km2	%	km2	%	1992–2007	2007–2022	1992–2022	1992–2007	2007–2022	1992–2022	1992–2007	2007–2022	1992–2022
Water Body	102.6	1	80.67	0.79	74.72	0.73	−21.94	−5.95	−27.89	−21.38	−7.38	−27.18	−1.60	−0.51	−2.11
Forest Land	2472	24.1	1554.6	15.2	1415	13.7	−917.26	−139.7	−1057	−37.11	−8.99	−42.76	−3.09	−0.63	−3.72
Grazing Land	908.6	8.86	690.42	6.73	422.1	4.11	−218.16	−268.3	−486.47	−24.01	−38.9	−53.54	−1.83	−3.28	−5.11
Cultivated L.	6570	64	7691.5	75	8080	79.2	1121.3	388.01	1509.34	17.07	5.04	22.97	1.05	0.33	1.38
Settlement	205.8	2.01	241.83	2.36	267.8	2.36	36.04	25.95	61.99	17.51	10.73	30.12	1.08	0.68	1.76

The West Arsi zone and the center of the study area have seen an increase in cultivated land and settlement (Fig. 7, Fig. 8, Fig. 9). Better land productivity and flat to moderate topography may be the reasons for this change. Between 1992 and 2022, the amount of cultivated and settlement land increased by 1509.34 km² (22.97%) and 61.99 km² (30.12%), correspondingly, at rates of 1.38% and 1.76% each year. On the other hand, between these years, the area of water bodies, forests, and grazing land declined by 27.18, 42.76, and 53.54%, respectively. This showed that water bodies, forests, and grazing land were primarily replaced by the cultivated and settlement area (Table 5, Table 6, Table 7). According to the FGD, key informant interviews, and personnel observations, land use and land cover type has changed as a result of increased demand for farmland, settlement, firewood, charcoal production, and building materials. It is consistent with [8,59] in which cultivated land increased at rates of 726.81 ha/year between 1979 and 1984 and 1515.7 ha/year between 2000 and 2018, respectively. According to Ref. [32], the Malka Wakana watershed has experienced a considerable amount of cultivated land expansion and a decline in forest land. Similar to how [60] noted an increase in cultivated land of 91.5 ha/year [61], found an increase in cultivated land of 7.13% between 1964 and 2014.

Table 6.

LULCC (km²) Transitional area matrix between 1992 and 2007.

	LULC classes	2007
		Water Body	Forest land	Grazing Land	Cultivated land	Settlement	Total	Loss
1992	Water Body	49.47	13.77	7.61	28.97	2.79	102.61	53.14
	Forest land	3.82	631.58	306.3	1434.2	95.95	2471.84	1840.26
	Grazing Land	10.41	129.61	55	696.52	17.05	908.58	853.58
	Cultivated land	16.25	744.22	303.27	5487.11	19.35	6570.2	1083.1
	Settlement	0.71	35.39	18.25	44.74	106.69	205.79	99.1
	Total	80.67	1554.57	690.42	7691.53	241.83	10259.03
	Gain	31.2	922.99	635.42	2204.42	135.14

Note: The bold numbers indicate the unchanged LULC proportions from 1992 to 2007.

Table 7.

LULCC (km²) Transitional area matrix between 2007 and 2022.

	LULC Classes	2022
		Water Body	Forest Land	Grazing Land	Crop Land	Settlement	Total	Loss
2007	Water Body	46.18	8.6	9.63	15.77	0.49	80.67	34.49
	Forest Land	5.65	654.76	94.06	793.26	6.84	1554.57	899.81
	Grazing Land	6.41	191.62	39.16	449.25	3.97	690.42	651.26
	Cultivated land	14.75	553.47	269.85	6790.99	62.46	7691.53	900.54
	Settlement	1.73	6.43	9.41	30.25	194.02	241.83	47.81
	Total	74.72	1414.88	422.11	8079.53	267.78	10259.02
	Gain	28.54	760.12	382.95	1288.54	73.76
	LULC Classes	2022
		Water Body	Forest Land	Grazing Land	Crop Land	Settlement	Total	Loss
2007	Water Body	46.18	8.6	9.63	15.77	0.49	80.67	34.49
	Forest Land	5.65	654.76	94.06	793.26	6.84	1554.57	899.8
	Grazing Land	6.41	191.62	39.16	449.25	3.97	690.42	651.3
	Cultivated land	14.75	553.47	269.85	6790.99	62.46	7691.53	900.5
	Settlement	1.73	6.43	9.41	30.25	194.02	241.83	47.81
	Total	74.72	1414.88	422.11	8079.53	267.78	10259.1
	Gain	28.54	760.12	382.95	1288.54	73.76

Note: The bold numbers indicate the unchanged LULC proportions from 2007 to 2022.

Besides that, between 1992 and 2007, cultivated land increased by 1121.3 km² (17.07%) at a rate of 1.05% per year, while settlement increased by 36.04 km² (17.51%) at a rate of 1.08% per year. Meanwhile, within the stated years, there was a decline in the area of forest, grazing, and water bodies by 917.27 km² (59%), 218.16 km² (31.6%), and 21.94 km² (27.19%), respectively (Table 5, Fig. 7, Fig. 8). Finally, out of the entire study area, 2314.74 km² (22.56%) were converted to a different land use and land cover, while the remaining 77.44% remained unchanged. Therefore, cultivated and settlement land had shown substantial expansion relative for the past 30 years, while forest and grazing land had declined. The result of LULCC and participatory data analysis revealed the expansion of cultivated and settlement areas at the expense of forest and grazing land (Table 5). It is in line with the [9] report that, in the northeastern part of Ethiopia, cultivated land increased due to the shrinkage of other land uses and land cover classes. Dense forests, open forests, and water bodies had been converted to agricultural land and settlement areas in western Ethiopia [62]. Between 1973 and 2019, cultivated land, settlement area, and water bodies increased, whereas forest land, bare land, and wetlands decreased [1]. Similarly [5], reported that cultivated land increased by 36.70% for 60 years (1957–2017). In agreement with this finding, a significant loss of woodland cover was reported in the Huluka watershed of Oromia Regional State between 1979 and 2017 [59]. Another study found a major decline in areas of forest by 64% whereas shrub grassland decreased by 6% in the Gish Abay watershed of the Sekella district of the West Gojjam Zone between 1957 and 2001 [18].

Similar trends in land use and extent of land cover for 1992 and 2007 were observed for 2022 as well, except for the magnitude of change in specific land use and land cover. Thus, cultivated land, forest land, grazing land, settlements, and water bodies were occupied on 8079.54 km² (79.15%), 1414.88 km² (13.66%), 422.11 km² (4.11%), 267.78 km² (2.36%), and 74.72 km² (0.73%), respectively. Besides, the forest land, grazing land, and water bodies declined from 2007 to 2022 by 139.69 km² (8.99%), 268.31 km² (38.9%), and 5.95 km² (7.38%), respectively (Table 5 and Fig. 8). The annual rate of change from 2007 to 2022 was found to be 0.63%, while the percent of change between these years showed a decline of 8.99%. According to Ref. [63], Ethiopia's forest cover decreased from 13.3% to 11.4% of the country's total area between 1993 and 2016, with an expected annual rate of change of 0.8%. This might have happened because of rapid population growth; people are forced to convert other land uses and land cover to cultivated land. The degradation of grazing and forest lands was mostly driven by the growing demand for cropland and wood products. Also, this particular study is similar to the earlier study, which found a large increase in built-up area coverage over a 32-year period (1986–2018) [11]. According to Ref. [10], Ethiopia's expansion of built-up areas is supported by the country's rapid population growth and economic development.

3.2. Land use and land cover change transitional matrices

The transitional area matrix from 1992 to 2007 showed that 1434.20 km² of forest land and 696.52 km² of grazing land had been converted into cultivated land, while 95.95 km² of forest land and 19.35 km² of cultivated land had altered into settlement by 2007 (Table 6). Likewise, the land use/cover change transition between 2007 and 2022 indicated that 793.26 km² of forest land and 449.25 km² of grazing land had changed into cultivated land, whereas 62.46 km² of cultivated land and 6.84 km² of forest land had been converted into settlements by 2022 (Table 7). The differences between gains and losses could be decreased or increased (Fig. 10) [64]. revealed that there was a decrease in vegetation cover and a large increase in built-up areas between 1992 and 2022. The recent 2022 LULC classification accuracy assessment showed excellent results. This was done using overall accuracy. It agrees with the findings of [65,66], who discovered overall accuracy scores higher than 86%.

Fig. 10.

Gains and losses of LULC classes between 1992 and 2007.

3.3. Implications of the change

For the study area, the impact of agricultural expansion on LULC has significant environmental implications. The increase in cultivated land was at the expense of forest land, grazing land, and water bodies. Gain and loss differences for cultivated land between 1992 and 2007 were calculated using the transitional area matrix and became 1121.33 km² and 36.04 km² for settlements, respectively. Similar trends were seen in the LULCC transition matrix between 2007 and 2022, which showed that some of the forested and grazing lands were replaced by cultivated land that could alter the natural ecosystem. This suggests that ecological services, the availability of locally used resources, and local community livelihoods were all adversely affected by LULCC. Participants in the FGD were told that the spatial and temporal decline of forest and grassland resources has an impact on the collection of traditional medicines, wild edible fruits, vegetables, and non-timber forest products. Moreover, it results in poor water potential, increases climate variability [10,34,53], high soil erosion and sedimentation [16] and decreases the availability of resources for animal feed. This could be happened in the Adaba, Dodola, Kokosa, Kofele, Shirka, Tana, and Robe district. Changes in LULC have a significant impact on biodiversity loss, forest cover loss, and climate change [67]. Furthermore, runoff, soil loss, and stream flow are all impacted by LULCC [68].

3.4. Drivers of land use and land cover change

Rapid population growth, topography, proximity to roads, streams, and towns; illegal logging; fuel wood collection; soil erosion; high cost and limited access to agricultural inputs; resettlement policies; and the weaknesses of institutions revealed during FGD and by key informants were identified as a major cause for LULCC. As population growth increased, the demand for cultivated and settlement land increased. This might have occurred due to the dominance of extensive agriculture practices in the study area. Population growth and the expansion of cultivated land were the major drivers for LULC changes [25,69,70]. Population growth has an impact on land resources because of the need to produce more food as well as the rise in demand for settlement and fuel wood. In other words, the local community was obliged to clear forests on steep slopes due to a lack of farmland caused by population development [71].

The settlement area increased from 205.8 km² in 1992 to 267.8 km² in 2022, an increase of 30.13%. This could be accelerated by the rural-urban movement. whereas the study area land holdings, which include grazing land, forest land, and water bodies, show a significant decline from 1992 to 2022. The LULC expansion and shrinkage in extent between 1992 and 2022 were shown by gains and losses graphs (Fig. 10, Fig. 11). It revealed a drastic decrease in the area of forest cover, grazing land, and water bodies, whereas cultivated land and settlement increased. This could be found due to irrigation agriculture spread to water bodies, such as rivers, marshes, and streams, in Kokosa, Adaba, Shirka, and Arsi Robe District, whereas rainfed agriculture considerably replaced the forest and grazing land. In line with past findings [23,27,72,73], the 2037 and 2052 estimates showed that cultivated land and settlements would increase at the expense of grazing land, forest land, and water bodies throughout the forecasted period.

Fig. 11.

Gains and losses of LULC classes between 1992 and 2007.

It was feasible to study the land use change evaluated by gains and losses experienced by different classes using LULC maps from 1992, 2007, and 2022, as well as the projected LULC maps of 2037 and 2052, utilizing the change analysis tool available in Terrset's Land Change Model (LCM). The results of the change analysis in LULC between 1992 and 2052 showed a considerable change. In comparison to grazing land, water bodies, and forest land in the study's area, cultivated and settlement land increased by 0.75 and 19.28%, respectively, from 2022 to 2037 and 2037 to 2052. This might be a function of the increase in population, the expansion of agriculture, the collection of wood for charcoal and firewood, weak land use policy, and deforestation operations, which significantly contributed to the shift in land use [74]. The CA-Markov model may be useful in predicting LULCCs, as shown by the similar concepts of agreement and disagreement that were achieved [27].

3.5. Analysis of predicted and simulated land use and land cover change

The predicted LULCC for 2037 and 2052 also showed a similar trend to the previous three periods. Most of the recent LULC classes used for simulation, namely 2022, bear a strong resemblance to the predicted years in trend analysis. In predicting LULCC, the major variables drawn during FGD, key informant interviews, and field observation were used. In that case, the most influential variables were selected for prediction based on qualitative survey and desk review results. In the problem, prioritization participants selected population density, topography, and distance from towns, roads, and streams as the top driving factors for LULCC in the study area. Desk review literature [68] in the Chemoga watershed, Blue Nile Basin [75]; in the Lake Hawassa watershed [43]; in southern Ethiopia [76]; in the Lake Tana catchment; and [20]) in agreement with this study on problem identified.

3.6. Interpretation of land use and land cover change driving variables

The spatial analysis of LULCC is considered a driver variable that can generate variations. It should be modeled as either static or dynamic components [54]. A driver's impact, population density, topography, and proximity factors were selected in the study area to predict LULCC and tested for explanatory value using Cramer's V results variable transition. Population density and proximity to towns provide the greatest influence on the LULC conversion process (Table 8). Slope variables have low Cramer's V values, so they have an insignificant effect on LULCC. Evidence likelihood is used for the determination of the relative frequency of pixels of different LULC types within the areas of change. It is recommended in cases where there are low Cramer's V values. The evidence likelihood result obtained is considered good. Similar results were obtained by Ref. [22] as deforestation increases with the increase in population. LULCC were influenced by population numbers, the expansion of irrigation agriculture to the wetlands, the development of industrialization, and urbanization [[77], [78], [79]]. Elevation and slope are recognized as critical topographic factors influencing LULC change. Topography has effects on the spread and extent of urban distribution, forest cover, and range land conversion to agricultural land [17].

Table 8.

Cramer's V for each of the LULCC variables.

Driver Variables	Cramer's V
Population density	0.4678
Distance from Towns	0.3004
Distance from Streams	0.1908
Distance from Roads	0.1483
Elevation	0.2616
Slope	0.0735
Evidence Likelihood	0.4936

3.7. Model validation for simulated land use and land cover

A LULCC area transition possibility matrix was made for 1992–2007 and 2007–2022 (Table 9, Table 10). Then the LULC map of 2022 was predicted through an integrated CA–Markov model in TerrSet 2020 (Fig. 12). The kappa index agreement results for the predicted LULC of 2022 and the actual LULC of the same year showed that there were small differences and a high level of consistency between the two LULC maps. The values 0.8443, 0.8312, 0.8312, and 0.7997 of kappa coefficients were observed for $K_{no},K_{location},K_{locationstrata}and{K}_{standard}$, respectively (Table 9). Thus, all the KIA results showed that there was a high level of agreement, which confirmed that the accuracy was reasonable for future land use prediction.

Table 9.

Kappa index agreement for predicted LULC map of 2022.

KIA components	Description	Value
$K_{no}$	Kappa index agreement for no information	0.8443
$K_{location}$	Kappa index agreement for location	0.8312
$K_{locationstrata}$	Kappa index agreement for location stratification	0.8312
$K_{standard}$	Kappa index agreement standard	0.7997

Table 10.

The difference between the actual and predicted LULCC in 2022.

LULC category	Actual Area	Predicted Area	Difference (P-A)	Percentage Difference (%)
Water Body	74.72	75.1136	0.393604	0.526773
Forest land	1414.88	1263.804	−151.076	−10.6777
Grazing Land	422.11	445.4513	23.34132	5.529677
Cultivated land	8061.53	8149.32	87.79032	1.089003
Settlement	285.78	325.3312	39.55121	13.83974

NB: P = predicted area in SKM; A = Actual area in SKM.

Fig. 12.

Simulated LUCC for 2022.

The comparison results of actual and predicted LULCC in 2022 showed differences in percentages of 0.53, −10.68, 1.09, and 13.84 for a water body, forest land, grazing land, cultivating land, and settlement, respectively (Table 10). Thus, around 86% of predicted and actual LULCC in 2022 were in agreement on land use type. Finally, based on KIA results and the difference between actual and predicted LULCC, the model was acceptable for making predictions for 2037 and 2052.

3.8. Land use and land cover change analysis between 2022, 2037, and 2052

As stated earlier, from 2022 to 2037, the area under cultivated land, grazing land, and settlement increased by 15.44 km² (0.19%), 16.04 km² (3.8%), and 83.63 km² (31.23%), respectively (Fig. 13). On the other hand, forest land and water bodies decreased by 110.94 km² (7.84%) and 4.16 km² (5.57%), respectively (Table 11). Estimates of changes in LULC will be influenced by population and economic growth rates by 2050; developed areas of all densities and the area covered by impervious surfaces (roads and towns) expand more quickly in scenarios that assume ongoing high growth [80]. Socioeconomic dynamics and livelihood activities influence the conversion of natural ecosystems to easily useable forms by the local community is expected [77,[81], [82], [83]].

Fig. 13.

Predicted LULC map of 2037.

Table 11.

Predicted Land Use Land cover and its Changes from 2022 to 2037.

LULC Classes	2022		2037		2052		Net Change (km2)			Change (%)			Rate of change km2/year
	km2	%	km2	%	km2	%	2022–2037	2037–2052	2022–2052	2022–2037	2037–2052	2022–2052	2022–2037	2037–2052	2022–2052
Water Body	74.72	0.73	70.56	0.69	65.56	0.64	−4.16	−5	−9.16	−5.57	−7.09	−12.26	−0.38	−0.49	−0.87
Forest Land	1414.9	13.8	1303.9	12.7	1191.02	11.61	−110.9	−112.9	−223.86	−7.84	−8.66	−15.82	−0.54	−0.60	−1.15
Grazing L.	422.1	4.11	438.15	4.27	425	4.14	16.04	−13.15	2.89	3.80	−3.00	0.68	0.25	−0.20	0.05
Cultivated L.	8079.5	78.8	8091	78.9	8200.31	79.93	15.44	105.33	120.77	0.19	1.30	1.49	0.01	0.09	0.10
Settlement	267.78	2.61	351.41	3.43	377.14	3.68	83.63	25.73	109.36	31.23	7.32	40.84	1.81	0.47	2.28

From 2037 to 2052, the LULCC prediction indicated that the area under cultivated land increased by 1.3%, which represented 105.33 km², whereas the settlement area increased by 7.32%, which represented 25.73 km² (Fig. 14). Forest land, grazing land, and water bodies decreased by 112.92 km² (8.66%), 13.15 km² (3%), and 5 km² (7.09%), respectively (Table 11). Markov chain analysis described LULCC from one period to the next in order to project future change. It gives a probability matrix for each land use transition to another land use type, particularly for water bodies that have a high likelihood of being converted into cultivated land, as presented in Table 12, Table 13. Between 2022 and 2052, the area of forest land, grazing land, and settlements are likely to be taken over by cultivated land. The probability of 51.61 and 63.97% of forest land and grazing land being converted to cultivated land from 2007 to 2037, respectively, the forest and grazing land had the same probability of being converted to cultivated land from 2037 to 2052: 18.86% and 18.5%, respectively [25]. observed the expansion of cultivated land cover from 1985 to 2015, with similar predicted trends for 2030 and 2045. This study's findings are consistent with [6,42,80], which demonstrated that future agricultural expansion and water body conversion will be expected. As the population increases the demand for infrastructure and pressure on natural resources will be increased by the exceptional circumstance [13,84]. In comparison to 2018, predicted LULCC for 2050 shows a larger fall in the forest (16.32%) and agricultural (0.12%) [85].

Fig. 14.

Predicted LULC map of 2052.

Table 12.

Markov matrix probability of each land use land cover from 2022 to 2037.

	LULC Category	2037
		Water Body	Forest Land	Grazing Land	Cultivated land	Settlement
2022	Water Body	0.7778	0.0139	0.0139	0.1898	0.0046
	Forest Land	0.0012	0.4111	0.0622	0.5161	0.0094
	Grazing Land	0.0063	0.2868	0.0589	0.6397	0.0082
	Cultivated land	0.0026	0.0667	0.0381	0.8727	0.0163
	Settlement	0.0032	0.0522	0.0393	0.1985	0.9679

Table 13.

Markov matrix probability of each land use land cover from 2037 to 2052.

	2052
	LULC classes	Water Body	Forest Land	Grazing Land	Cultivated land	Settlement
2037	Water Body	0.9962	0	0	0.0038	0
	Forest Land	0	0.7892	0.0222	0.1886	0
	Grazing Land	0.0025	0.1622	0.6496	0.185	0.0006
	Cultivated land	0	0.0095	0.0135	0.9456	0.0314
	Settlement	0	0.1108	0.0185	0.2709	0.8997

4. Conclusion

Land use and land cover change is one of the top agenda of the current world. In the upper Wabe-Shebele river basin the growing in cultivated and settlements area found at the expense of water body, forest and grazing land decline. Hence that water body, grasslands and forest were susceptible to these expansion because of high population density and expansion of agricultural mechanization. With this between 1992 and 2022, cultivated land increased at annual rate of more than 0.33% whereas settlement areas increased by more than 0.68% annual rate. The predicted land use land cover change also shows that the cultivated and settlement areas will be increased whereas water body, grazing, and forest land shows decline. In prediction result the annual rate of change pronounced the significant expansion of settlements by 0.47 to 1.81% whereas annual rate of change for cultivated land failed between 0.01 and 0.09%. The underlying forces for land use land cover change in the study area are population growth, climatic change, town growth, expansion infrastructure, and agriculture. Natural and human factors have the greatest impact on the LULC of the study area, while proximity factors have a relatively small impact. It causes ecosystem services to be lost, crop yield decline, poor water potential, and land degradation. Hence, integrated land use planning based on the findings of the land evaluation is needed to prevent the disintegrant of land use and land cover. Planning and conducting projects should give consideration to community-based resource management and participatory rural appraisal.

Author contribution statement

Siraj Beshir Sheko: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data; Wrote the paper.

Awdenegest Moges: Conceived and designed the experiments; Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data; Wrote the paper.

Mihret Dananto: Conceived and designed the experiments; Contributed reagents, materials, analysis tools or data.

Data availability statement

Data will be made available on request.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.