Spatiotemporal Dynamic of Land Use/Land Cover Changes and Their Drivers in the Fincha’a - Neshe Sub - Basin, Southeastern Blue Nile Basin, Ethiopia

It is paramount to evaluate the spatiotemporal dynamics of land use and land cover (LULC) changes and their drivers. This is because it helps generate information on biodiversity, land productivity, ecology, and livelihoods for decision - making. Due to land degradation, deforestation, shifting cultivation, high population pressure, and the three national megaprojects (NMPs), the Fincha'a Neshe sub - basin (FNSB) LULC, changes may be unprecedented. This study aimed to investigate the spatiotemporal dynamics of LULC changes and their drivers using remote sensing (RS) data and geographic information systems (GIS). Landsat images 5, 7


Introduction
Ethiopia is Africa's second-most populated country and has experienced significant LULC changes (Genet, 2020). However, the problem of land cover dynamics is more severe in the Ethiopian highlands. These highlands account for 44% of the country's landmass and have been cultivated for millennia (Hurni et al., 2005), as in the study sub-basin (Ayana et al., 2014). These highlands have greater potential for cultivation and are preferred for settlements. The greatest potential of highland areas for agricultural production and their suitability for settlement may accelerate the transformation of natural vegetation into agricultural land and human settlements. Population growth, resettlement programs, and climate change are the primary causes of the fast-changing LULC in the Ethiopian highlands (Regasa et al., 2021). Furthermore, other human and natural driving forces may also play a significant role (Yesuph & Dagnew, 2019). Therefore, changes in LULC affect life support functions and human lives (Othow et al., 2017;Regasa et al., 2021). It has diverse environmental impacts, negatively affecting the water supply, reservoir storage capacity, agricultural productivity, and regional ecology (Chen et al., 2021;Tadesse et al., 2017). Changes resulting from anthropogenic (human activities) and natural (earthquakes, landslides, droughts, and floods) factors may have influenced the LULC dynamics (Solefack et al., 2018) and are principal drivers. Anthropogenic activities are main season extends from June to September, with peaks occurring from July to September. It is dry from November to March. The sub-basin frequently experiences storms and flash floods during the rainy season. The annual minimum and maximum temperatures for the FNSB are 11.9 °C and 24.2 °C, respectively. The major parts of the sub-basin are cultivated intensively. Teff, maize, barley, and wheat are the principal crops grown in this sub-basin. FNSB has a wide range of soil types dominated by clay-loam, clay, and loam textural classes. Clay soil associated with swamps and temporary wetlands on plains has good-to-moderate fertility. These are characteristics of the most significant parts of the sub-basins. The rest of the catchment is continuously cultivated with low soil fertility ( Ayana et al., 2014).

Data Sources and Method of Acquisition
In this study, two primary data sources were used to assess the LULC changes and their drivers. These data sources included remote sensing and population data. This section describes the data sources, collection methods, and processing procedures.

Population Data
Population data were compiled from census reports of the Central Statistical Agency (CSA) of Ethiopia (CSA, 1991(CSA, , 1998(CSA, , 2007. Population data were collected and prepared: a) following the administrative boundary in which the sub-basin was embedded; and b) census data for 1984, 1994, and 2007 were projected to the study periods of 1986, 2000, and 2016 by calculating the growth rate as stated by CSA (2007). The projected population size for each study period was estimated based on the growth rate using equation 2 (Bewket, 2002;Gashaw et al., 2014b).
where P 1 is the present population, P 2 is the future population, r is the growth rate in percentage, and t is the number of years between the future and present censuses.
Where; P 1 is the present population, P 2 is the future population, r is the growth rate in percentage, and t is the number of years between the future and present censuses.  Table 1 provides details of the Landsat data used in this study. In this study, LULC change analysis used ancillary enrr.ccsenet.org Environment and Natural Resources Research Vol. 12, No. 1;2022 and satellite data. The ancillary data included ground control points, Google Earth data, and topographic maps of the sub-basin at a scale of 1:50000. The satellite data included Landsat 5, Landsat 7, and Landsat 8 for the discrete study periods of 1986, 2000, and 2016, respectively. The study period dates closer to cloud-free imagery accessed from the UGS Earth Explorer used for the study. Field observations, informal interviews, and discussions that helped to generate information in line with the drivers of LULC changes were used. In this study, image pre-processing techniques, such as image enhancement and geometric correction, were performed using the method proposed by Lillesand et al. (2015). This method includes assigning the coordinate system, stacking layers of the separate bands of the datasets, and sub-setting the images. The software packages ERDAS IMAGINE 2014 and ArcGIS10.4.1 were used to complete the pre-processing and post-classification analyses. Hybrid classification methods involving unsupervised and supervised techniques algorithms were used for image classification. The unsupervised classification was performed before the fieldwork to obtain general insights into the land cover classes in the sub-basin. It is an automated classification system that does not require knowledge of the study site. This enabled us to determine the features that showed the overall LULC clusters of pixels. A supervised classification with a maximum likelihood algorithm was used to classify the images based on the training sets (signatures). The training data set provided by the user directs the software to detect pixels for known land cover types. Thirty to forty training samples were used for each classification. The re-coding work was applied to merge spectrally distinct classes to redevelop the final information classes. Following this classification, eight LULC classes were identified: bare land, cropland, forest, grassland, settlement, shrublands, water bodies, and wetlands. Table 2 presents a detailed description of each LULC type in the sub-basin. Visual interpretation elements and reflectance characteristics of the features in the satellite images from 1986, 2000, and 2016 were used to select the LULC types.

Accuracy Assessment
The accuracy of thematic maps created using remotely sensed data is vital in most mapping projects. It provides a guide for the quality of a map and its suitability for specific purposes (Congalton & Green, 2019). Therefore, the decision to use a map significantly depends on its accuracy. Collecting reference data (RD) is the first and most important step for any accuracy assessment. Thus, Google Earth and GPS field data from the study site were the data sources for the RD. The RD, independent of the ground truth, was the data source used in the classification scheme. Furthermore, RD should be reliable in making meaningful assessments. Error matrices were used to validate the quality and reliability of the RD. Four tools: Kappa statistics, overall accuracy, user accuracy, and producer accuracy were used.
Kappa statistics and overall accuracy were used to assess the overall classification performance and the user and producer accuracies were used to assess the individual LULC classes (Congalton & Green, 2019). The descriptions and calculations of these terms have been explained by several researchers (Congalton, 2018;Gu & Congalton, 2019). This study determined the accuracy of the three-time steps (1986, 2000, and 2016) of the classified maps. The final maps were produced to achieve the lowest overall accuracy of 85% (Anderson, 1976). A comparison of map features and matrix analysis was performed to define LULC change detection (Lu et al., 2014). In this case, the areas of the LULC categories converted from each class to any other class, and the changes in direction and magnitude were determined.

Land Use/Land Cover Change Detection Analysis
After the accuracy assessment, maps showing detailed LULC changes in FNSB were produced. The maps produced from the image classification periods of 1986, 2000, and 2016 were used to analyze LULC changes. The various steps developed and used to analyze, quantify, and interpret the map are presented in Fig.2. To validate the data collected on LULC changes in the subbasin, focus group discussions, and informal interviews were conducted. Elderly people believed to have a better historical understanding of the changing LULC trends in the sub-basin were selected. After classifying the images, the spatial distribution of the LULC classes was calculated for each period and the extent of change in land use within and between periods was compared. Changes in the different LULC classes were analyzed using ArcGIS10.4.1, TerrSet-20, and ERDAS IMAGINE2014. The rate of change (km 2 /year) and the percentage change in each class during the study period were determined using the formula described by Gashaw et al. (2017).
Where R is the Rate of change; Y 1 is the earlier area of LULC features in km 2 , Y 2 is the most current area of LULC features in km 2 , and N is the Time interval between Y 2 and Y 1 in Years ΔA= (At 2 -At 1 )*100 where ΔA is the percentage change in the area of the LULC class between the initial (t 1 ) and final (t 2 ) periods; At 1 is the Area of the LULC class at the initial time; and At 2 is the Area of the LULC class at the final time.
Pixel-based comparisons were conducted to obtain information on changes within each pixel. The most common approaches proposed methodologies were utilized to interpret the changes more, taking "from-to-to" information (Alshari & Gawali, 2021;Olmanson & Bauer, 2017;Singh et al., 2015). By coding the classification results for discrete study periods, the analysis produced a change map showing a complete matrix of changes (e.g., from wetlands to cultivated land and grasslands to forests). In addition, a post-classification approach that reduces the problem of radiometric calibration between dates has been used (Ban & Yousif, 2016;Mishra et al., 2017). A change matrix (Zhang & Weng, 2016) was made with the help of the TerrSet v-20 (Geospatial Monitoring and Modeling System) software. Quantitative data for each LULC were collected in 1986, 2000, and 2016 to determine the gains and losses for each LULC category and this information is an indispensable tool for management decisions. Projecting the Census data to study periods (1986, 2000 and 2016) Analyzing the population dynamics of the sub-basin

Land Use/land Cover Maps Accuracy Assessment
This section evaluates the classification accuracy of the LULC class maps to determine their applicability for land use planning and development. Consequently, for each LULC class over distinct study periods, the assessment estimated the user and producer accuracies. The grassland cover of the sub-basin had the lowest accuracy values for both users and producers. This lower accuracy assessment result might be attributed to the reflectance similarity of grasslands with other land cover classes. For example, Sahalu (2014) reported that other land cover classes appear as agricultural land and vice versa. Václavík and Rogan (2009) reported that agricultural land was the most challenging to classify because it contained a mix of crops at different phenological stages. In contrast, as shown in Supplementary Tables (Table S1, S2, and S3), barren land, settlements, and wetlands had the highest user and producer accuracy. Table 3 presents the overall accuracy assessment values and Kappa statistics of the classified maps for each study period. The overall classification accuracy showed that the categories were well-aligned, and the Kappa Statistics ranged from very good to excellent. As the developed LULC maps satisfied the bare minimum requirement of 85% accuracy for LULC analysis, the findings of the overall accuracy assessment showed that the generated maps could serve well. The resulting maps can then be used for change detection analysis because they show a strong agreement between the categorized LULC classes and geographical data (ground truths).  It has increased at the expense of forests, shrublands, and grasslands. During the first distinct study period, settlement and bare land increased by 184.8% and 203.4%, respectively. The increase in bare land might be attributed to the decline in vegetative cover of the sub-basin. As a result, the bare land expands, making the sub-basin vulnerable to soil erosion, leading to severe land degradation.
Wetlands are one of the principal land cover types of the sub-basin, accounting for 10.10% of the total land cover. It showed a consistent decreasing trend from 10.1% (334.7 km 2 ) in 1986 to 8.0% (265.1 km 2 ) in 2000 and further decreased to 7.0% (232 km 2 ) in 2016 ( Fig. 3 and 4 and Table 4). According to the analysis of LULC changes over the past three decades (1986-2016), 102.9 km 2 (30.7%) of wetlands in the sub-basin were impaired. The decline in wetlands is likely attributable to the increased desire for more land for cultivation, its potential use as grazing land during the dry season when cattle feed is limited, and its unique production potential.   Forest land cover was another dominant LULC class in the sub-basin, covering 6% (199.3 km 2 ) of the total land cover (Fig 3). However, this land cover declined by 58.2% (116 km 2 ) over the past three decades (Table 4). This conversion of forest cover to other LULC classes requires immediate intervention that otherwise disrupts ecological balance. Shiferaw and Singh (2011) reported a dramatic expansion of agricultural land and a reduction in forest cover between 1972 and 2003 in the southern Wollo highlands. According to Ayana et al. (2014), the forest cover in the sub-basin decreased by 36,225ha (50.48%) between 1985 and 2005. Shrubland is the first dominant LULC class of the sub-basin and accounted for 41.2% (1364.6 km 2 ) in 1986 and declined to 36.8% (502.6 km 2 ) in 2016. Quantitative evidence gathered by interpreting the satellite images showed that the study site experienced significant LULC changes. Shrublands, grasslands, wetlands, and forest land classes showed a significant decline. In contrast, croplands, barren lands, settlements, and water bodies increased during the study period (Table 4; Fig 4, Fig 3). Generally, the conversion in LULC types was vigorous and common across the landscapes. This was because of the wider demands of the growing population, which required more cropland to fulfill the food demand. Meshesha et al. (2016) reported that population growth had transformed the land cover over time.    Table 5 shows the matrices of the LULC transformation observed in the sub-basin during the study period. A transformation matrix table was generated for each discrete study period to show the area of each LULC class converted to another. The CROSSTAB module generated using TerrSet-v 20 Geospatial Modelling and Monitoring System software was used for cross-tabulation analysis. During the first study period (1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000), shrublands constituted a significant portion of the study area (1355.5 km 2 ) but later underwent a rapid transformation to other LULC classes, mainly croplands. As a result, cropland cover increased from 36.3% (1201.2 km 2 ) in 1986 to 43.9% (1453.6 km 2 ) in 2000. In contrast, the shrubland, grassland, wetland, and forest land cover declined by 14.9%, 33%, 17.5%, and 6.10%, respectively. Croplands gained from shrubland, wetland, forests, bare land, and grassland, respectively, by 435.3, 48.7, 15.4, 4.1, and 19.7 km 2 between 1986 and 2000. In the same period, 8.4, 15.4, 218.7, 3.6, and 12.8 km 2 of cropland cover were transformed into bare lands, forests, shrublands, settlements, wetlands, and grasslands, respectively. In contrast, shrublands has decreased to 202 km 2 (14.9%) from 1986 to 2000, while gaining, 3.4, 11.15, 50.42, 52.80, and 218.7 km 2 , respectively, from bare land, grassland, wetland, forest, and cropland in the same period. Thus, the conversion of other LULCs to shrublands did not compensate for the shrinkage during this study period (It lost 539.9 km 2 of its cover to other LULCs and only gained 337.9 km 2 from other LULCs). In general, during this study period, 1070.2 km 2 (32.3%) of the sub-basin experienced an overall LULC change.

Land Use/land Cover Change Detection
The trends in the second and third study periods are presented in Table 5. During the second study period, one exciting event was the decline in bare land from 13.3 km 2 in 2000 to 8.6 km 2 in 2016. The decline in bare land cover may be attributed to the implementation of soil and water conservation practices in the FNSB by the local government. In general, during the entire study period, 1325.5 km 2 (40%) of the sub-basin LULC experienced a change. The conversion of natural forests and shrublands to cultivated land may have resulted in considerable losses in natural forests and shrubland cover. Croplands gained most of their cover from the shrublands after conversion. Over the entire study period, cropland has gained a total of 852.2 km 2 from shrubland 657 km 2 (77.1%), wetland 88.5 km 2 (10.40%), and forest 55.2 km 2 (6.1%), and grassland 42 km 2 (4.9%  Minta et al., 2018;Wondie & Mekuria, 2018). They reported that the cultivated land expanded at the expense of forests, bushes, and grazing. Relatively, cropland cover increased by 694.4 km 2 (57.81%) over the entire study period, with the greatest increase being registered in the second distinct study period accounting for 30.4% (Fig 5;  Table 5).
This study showed that wetlands were converted into croplands and grasslands across the sub-basin. It lost its cover to shrublands, contributing to 47.4% of wetland loss, whereas conversion to water bodies, croplands, and forests accounted for 24.4%, 18%, and 6%, respectively. Studies conducted in other parts of the country and elsewhere have shown that wetlands have declined enormously and have been lost to other LULC classes (Bezabih & Mosissa, 2017;Gebresllassie et al., 2014;Hussien et al., 2018;Moges et al., 2018;Rebelo et al., 2015). Table 6 shows the rate of change in LULC in the FNSB over the entire study period. During the first study period, croplands, settlements, and bare lands expanded at rates of 19.5, 0.3, 0.3, and 0.8 km 2 year -1 . In contrast, grassland, forest, shrubland, and wetland cover decreased by 2.2, 2.5, 4.8, and 11.5 km 2 year -1 . Similarly, the cropland cover increased at a rate of 29.1 km 2 year -1 between 2000 and 2016. The analysis showed that shrublands, forests, grasslands, and wetlands had decreased at a rate of 21.3. 5.1, 1.4, and 2.3 km 2 year -1 , respectively, in the second study period (Fig 6; Table 6).

Rate of Land Use/Land Cover Changes
When examining forest conversion, shrublands constituted a significant portion, which couldn't be considered an absolute degradation of forest cover. For example, forest, shrubland, and grassland cover losses may have profoundly affected other LULC classes, resulting in erosion-induced land degradation in the sub-basin.
This loss in vegetative cover implies the susceptibility of the NFSB to soil erosion and land degradation and a consequent decrease in the productivity potential of the sub-basin. The continuous decline in the shrubland, forest, and grassland categories may be inversely related to the overall expansion rate in bare land in the sub-basin (Table 6). This expansion rate suggests that the shrinkage of these LULCs directly affected the increase in bare land cover, as observed in this study. Shiferaw and Singh (2011) reported that barren fields increased by 256 ha/year between 1972 and 1985 at the expense of shrubland, forest, and grassland cover. Assefa and Bork (2014); Gashaw et al. (2014a) and Gebrehiwot et al. (2014) reported the same and stated that farmland and settlements expanded at the expense of forests, bare lands, and grazing lands.

Implications of Land Use/land Cover Changes in Fincha'a-Neshe Sub-basin
Changes in LULC class may not necessarily result in soil erosion or land degradation. However, expanding cultivated land at the expense of other LULC has had undesirable consequences. This study demonstrated that shrublands, wetlands, forests, and grasslands decrease as croplands increase. There might have been a decline in grasslands, which would have left less fodder and resulted in a decline in the number and quality of livestock. This decline in grassland cover may affect livestock productivity by decreasing the food and income from livestock and their products. During the study period, there was a steady decline in grassland cover, limiting the number of animals available for land preparation and transportation. The subsequent decrease in the number of animals resulted in a reduction in the quantity of manure used as a soil fertility amendment, thereby decreasing crop yield. According to studies in the upper Blue Nile basin and elsewhere, the increase in cultivated land at the expense of other LULC categories has detrimental effects on animal productivity and production (Berihun et al., 2019;Hurni et al., 2005;Miheretu & Yimer, 2018;Mulugeta & Legese, 2015).
This study showed that there has been a rapid conversion of LULC classes to croplands. These conversions are associated with the demand for land for crop production to fulfill the food requirements of the growing human population. As land resources are limited, many locals are forced to shift their cultivation to marginal areas. This shift of the cropland towards marginal land might result in intensive erosion, massive degradation, and siltation of water bodies, waterborne diseases, and flood in the sub-basin. In the sub-basin, the LULC class expanded to croplands, grasslands, and barren lands. This shift of LULC class to croplands, grasslands, and barren lands may make fertile soils more prone to erosion and land degradation. Alemu et al. (2015) and Mesene (2017) reported that extensive farming practices coupled with unwise land management practices and limited improved input provision and support for agriculture have brought persistent environmental challenges to watersheds.
The decline in vegetative cover (shrublands, wetlands, grasslands, and forest cover) has caused land degradation and food-insecure subsistence agriculture. During fieldwork, informal discussions with informants confirmed that the productive potential of their land plots had fallen. Reports from agricultural offices have confirmed this finding. Rapid LULC transformations and expansions in croplands and settlements in the FNSB may have caused hydrological, socioeconomic, and environmental problems. Population growth and established NMPs have contributed to the rapid shift in LULC classes in the sub-basin. This shift in LULC might cause a significant loss of natural vegetation and later conversion to croplands and waterbodies in the sub-basin. Tefera and Sterk (2010) reported that 5-8% of the forest cover was devastated because of the resettlement and reallocation of the local population in the sub-basin. Ganaie et al. (2021); Gebrehiwot et al. (2014); Genet (2020) and Kim et al. (2015) reported similar motives for the decrease in forest resources in northwestern Ethiopia due to population growth.

Drivers of Land Use/land Cover Changes
This study developed two scenarios in line with the drivers of the LULC changes. These scenarios included population growth and NMPs (FHEPP, NHEPP, and FVSE). The first scenario focused on population growth and its direct impact on LULC changes. Here, the sub-basin population growth may be due to the natural growth and establishment of NMPs. Due to the developed NMPs, individuals from every corner of the country and around the globe may migrate to these project areas substantially contributing to the population increase. Consequently, the sub-basin population increased from 257,932 in 1986 to 576,868 in 2016 at an average growth rate of 3.372%, 1.75%, and 2.75% in the distinct study period of 1986, 2000, and 2016 (Table 8).
Subsequently, the population density of the sub-basin was compared to that of its (Ethiopian) counterpart. As a result, in 2016, the population density of the sub-basins was 174.1 persons per square kilometer, compared with 103.6 persons per km 2 in the country (Tables 7 and S4). The increase in population density and the associated land scarcity may have resulted in intense competition over land resources. To fulfill the growing demand of the population, an increase in agricultural production and housing construction is mandatory. Consequently, croplands and settlements have expanded at the expense of forests, shrublands, and grasslands. For example, cropland and settlement increased from 1163.1 km 2 and 4.6 km 2 in 1986 to 1902.3 km 2 and 13.1 km 2 in 2016, respectively ( Table  4). The increase in croplands and settlements further pressurized the other LULC classes. Mhawish and Saba (2016) stated that population growth contributed to the expansion of cultivation and degradation. Thus, the growing population may continue to pressure the land cover of the sub-basin. Findell et al. (2017); Gashaw et al. (2014a); Hurni et al. (2005) and Wondie et al. (2016) stated that population growth caused the intensification of land use in rain-fed highlands. This increase in population size has prompted the shortening of fallow periods and the ultimate abandonment of land resources, the conversion of cultivated land into grazing land, and ongoing deforestation. Growing more crops on grazing land reduces the amount of feed available to livestock, which in turn reduces the number of animals and the income from their products.
Furthermore, this study investigated the relationship between population increases, cropland, and settlement expansion. To this end, simple regression and correlation tests were performed. Croplands had a correlation coefficient r 2 of (+) 0.985 and a p-value of less than 0.005 when compared to the population size, indicating a strong association (Table S5). Similarly, the settlement had a correlation coefficient r 2 of (+) 0.96 and a p-value of less than 0.013, indicating a significant association with population size (Table S6). The high positive value of r 2 and small p-value indicate a strong relationship among the competing parameters. These statistical indices demonstrate how population growth affects increases in settlements and agricultural land. As the population increased, more land was required for habituation and agriculture. Therefore, population growth was one of the driving forces behind the LULC changes in the sub-basin. Similar results have been reported by Ouedraogo et al. (2010), Hailemariam et al. (2016), andFenta et al. (2020), who showed that cropland expansion rates have increased worldwide at the expense of forests, grasslands, and shrublands.
Furthermore, to satisfy the growing population demand, LULC classes (e.g., Forests and forest products) are vital. Population size can, directly or indirectly, influence forest and forest products. This study investigated the relationship between urban and rural housing units and their associated forest and forest product demands. Thus, the growing population demands forests and forest products for infrastructure, housing construction, and cooking. Between 1994 and 2007, 90.4% and 95% of the urban housing units and 94.5% and 95% of the rural housing units used forests and forest products for wall construction (Table S7). Similarly, 96% and 97% of the urban population and 92% and 91% of the rural population, used forests and forest products, respectively, as major sources of cooking fuel in 1994and 2007. CSA (1998 reported that a higher proportion (98.2%) of housing units in Jimma and East Wollega use firewood/leaves as a major fuel source for cooking. Godfray and Garnett (2014) and Tendaupenyu et al. (2017) reported that a growing population and consumerism are pressurizing agriculture and natural resources.
The second scenario assessed the direct and indirect effects of NMPs on the LULC. Thus, the impact of NMPs was analyzed in line with the LULC dynamics of the sub-basin. Established NMPs can cause LULC changes in two ways: by increasing local population increase and land demand for the NMPs. Because of the established megaprojects, people came to the project site from different corners of the country and subsequently contributed to local population growth. Thus, local population growth and newcomers put excessive pressure on land resources, transforming the land cover into settlements. Additional land was required for the setup and development of these projects. Establishing these NMPs involves tremendous earthmoving and shaping activities that convert the land cover into other classes, thereby triggering soil erosion-induced land degradation. Furthermore, to fulfill their objectives, these projects require extra land. For example, hydropower plants demand more land to accumulate water upstream of the dams for hydroelectric power generation. Two hydropower plants swept 190 km 2 (other LULCs were converted to water bodies) of other land cover classes to spread water upstream of the dam (Table 4). Similarly, the FSVE has swept over 262 km 2 of other LULCs in the sugar cane plantations (SCP) and continues to expand. Source: a= Own work adapted from CSA; b= FAO and World Bank report, 2019 Source: a= Own work, calculated

Conclusions and Recommendations
The Fincha'a sub-basin is located upstream of the GERD in the southeastern Abbay Basin. The sub-basin underwent significant spatiotemporal changes in LULC, as indicated by a sharp increase or decrease between 1986 and 2016. Croplands were the dominant LULC class over the entire study period and increased by 694.4 km 2 (57.81%) at a rate of 24.6 km 2 per year across the study period. Similarly, the settlement increased from 8 km 2 to 14 km 2 at a rate of 0.3 km 2 per annum. The cropland classes have increased at the expense of grasslands, shrublands, forests, and wetlands. In contrast, shrublands, grasslands, wetlands, and forests declined at rates of 16.8, 1.7, 3.4, and 3.9 km 2 year -1 from 1986 to 2016. Compared to other LULC classes, the overall decline in shrub and grassland cover may have affected livestock and crop production. This decline in crop and animal production could have resulted from the reduced livestock feed and the amount of manure required to improve soil fertility. The overall expansion of cropland into shrub and forest lands may have further degraded the vegetative cover of the sub-basin, resulting in new erosion-prone sites. Furthermore, this study assessed the impact of population growth on the natural vegetation in both rural and urban housing units. Of the housing units assessed, both in rural and urban centers, over 90% had used forest and forest products for household energy consumption and construction purposes and hence for natural vegetation (forest and shrubland) degradation.
This study attempted to identify the primary drivers of changes in LULC in the sub-basin during the study period and found that population growth and established NMPs were the main drivers. Due to population growth cropland and settlement expanded significantly to meet the growing demand. The expansion of cropland and settlement caused excessive pressure on the LULC of the sub-basin and triggered changes in the LULC. On the other hand, the three established NMPs (Two hydroelectric power plants (FHEPP and NHEPP) and Sugar Estate (FVSE)) both transformed approximately 452 km 2 (13.8%) of the total land cover of the sub-basin. Established NMPs have displaced and moved the local population to nearby villages, districts, forests, and marginal land. This displacement of the local people may enhance deforestation and the cultivation of marginal lands, thereby increasing soil loss and affecting the hydrology of the sub-basin.
This study suggests the mitigation of the adverse effects of population growth on natural resources by adopting other income-generating off-farm activities (fishing on the two dams) and soil fertility amendments. Furthermore, local governments (GOs) and non-governmental organizations (NGOs) should grant improved land management practices. Land management practices such as soil and water conservation and providing improved agricultural inputs should be implemented to boost agricultural production and productivity. Moreover, local involvement in natural resource management and implementation of land tenure policies are indispensable for sustainable land use. Introducing and adopting alternative energy sources for household energy consumption is vital for reducing the pressure on forests and shrublands. In addition, devising forest development, protection, watershed management, and use strategies is essential to counteract the deteriorating shrubs and bushes and avoid further degradation. Wetlands, the dominant LULC class with unique productivity potential and diverse biodiversity, should have legal ownership to maintain biodiversity and ecosystem balance.