Nexus of land use land cover dynamics and extent of soil loss in the Panjkora River Basin of eastern Hindu Kush

Since the beginning of human history, numerous civilizations have used the surface of the Earth extensively, resulting in changes. The rate of erosion is predicted to grow in the 21st century due to intensified anthropogenic activities like changes in LULC, which is linked to a higher potent hydrological process characterized by extreme rainfall amplitude (Panagos et al. 2012, 2017). Water-induced erosion of soil is influenced by a variety of natural and human processes, and as a result, it has an indirect or direct effect on natural ecosystems (Leh et al. 2013; Erol et al. 2015). In addition to the natural processes, human alterations have increased erosion rates above normal levels. Boardman et al. (2003) and Tağıl (2007) emphasized that the disturbance of human history (such as deforestation, industrialization, urbanization, and agriculture) has been crucial because of the effects of the surface conditions of the soil. Human activities are therefore the main cause of the degradation of the land (Williams 1991; Morris 2020).

After the population explosion, the second-biggest issue for the world's environment is considered to be erosion-induced soil degradation (Kiassari et al. 2012; Nikkami 2012). The issue has become rather serious as a result of improper and excessive agricultural area usage resulting in soil erosion (Pradhan et al. 2012). This soil erosion results in loss in agriculture productivity and ultimately land destruction (Pimentel 2006; Kidane & Alemu 2015). The ability of soil to retain water is changed by soil erosion. Eroded soil is transported by water flow and dumped in the river system where the flow of water is interrupted (Ganasri & Ramesh 2016; Rawat et al. 2016). The carrying capacity of the river is reduced by runoff from the erosion of top land that is dumped there, which causes siltation in irrigation canals, and damage to turbines and water management systems. Changes in land use and land cover (LULC) and ineffective changes in LULC management strategies have all contributed to an increase in erosion (Sharma et al. 2011).

Numerous research studies have shown that the erosion of soil affects the changes in LULC at various spatial and temporal balances (Sharma et al. 2011; Tadesse et al. 2017). A higher risk of landslides exists in the region with highly erodible soil (Abidin & Abu Hassan 2005; Khosrokhani & Pradhan 2014). Therefore, it is vital to understand LULC and its impact on the dynamics of erosion of soil. To manage LULC more effectively and lower the danger of a major landslide, it would be helpful to understand the pattern and dispersion of change in LULC and the risk of soil erosion.

The soil loss models can be divided into empirical, regression and conceptual models. While models vary their sole purpose is to assess water-induced erosion of soil. The USLE and RUSLE are examples of empirical or regression models (Alkharabsheh et al. 2013). Among the many models available, the RUSLE model is frequently used to access and recognize soil loss due to its low requirement of data easy understanding (Perović et al. 2013). The model is helpful in determining the soil loss distributed spatially over a broad area (Ganasri & Ramesh 2016). It can determine the risk of erosion on a regular grid (Shinde et al. 2010). Integration of geospatial techniques and these models have produced good results when used to assess soil loss (Renard et al. 1991; Uddin et al. 2016). To address the problem of soil erosion, this integration has made evaluation and forecasting simpler and more effective (Tadesse et al. 2017).

This study holds importance as it offers insights into a region characterized by elevated erosion rates. These insights could prove helpful to urban planners and conservationists in their efforts to mitigate and analyze soil erosion. The mitigation and regulation of soil erosion have the potential to enhance agricultural yield and prolong the functional longevity of dams and reservoirs by preventing sediment accumulation. The discoveries of this research will serve as a valuable resource for soil experts, policymakers, irrigation divisions and the soil survey department within the Panjkora Basin, aiding them in implementing effective watershed and sediment control measures.

Image classification is a vital tool for processing and analysis of satellite imagery. Acquired Landsat images were classified.

Five classes, including the vegetation area, barren land, built-up area, snow cover and water body training samples, were taken. More than 200 training samples of each class were taken. The classified images were assessed for reliability by performing an accuracy assessment in which the classified images were compared with the other source of data that was recognized as reliable, validation data and data retrieved by interpreting satellite imagery.

The assessment included calculating:

• (i)

  Commission error: The number of incorrectly classified pixels in a row.

• (ii)

  Omission error: The incorrectly classified pixels in a column.

• (iii)
  Producer's accuracy: The number of pixels correctly classified in a particular class as a percentage of the total number of pixels actually belonging to that class in the image. Producer's accuracy measures how well a certain area has been classified. It was calculated using the following formula:

  
• (iv)
  User's accuracy: The sum of correctly classified pixels is divided by the total number of pixels. It is calculated by using the following formula:

  

The important initiating water erosion factor is the intensity and quantity of rainfall (Foster et al. 2002). According to EI30 (the sum of energy of kinetic and the rainfall of 30 min intensity), erosion would increase as rainfall intensity and amount increase (Renard 1997). For the R factor calculation, continuous data of long-term rainfall are required; however, in most countries, these EI30 data are not accessible. It is a challenging and time-consuming task even if there are still enough data available. To calculate the factor of R from the data of the annual average rainfall, certain straightforward approaches have been applied in different regions of the world. The availability and dependability of monthly rainfall data is the key benefit of this streamlined method (Lee & Heo 2011). Numerous experts have recognized a strong association between monthly data from various regions of the world and rainfall erosivity (Teh 2011).

For this study, the erosivity factor of R was computed from the annual average rainfall data. Data on rainfall were obtained from the Pakistan Meteorological Department, Peshawar for the two met stations close to Panjkora Basin (Timergara and Dir metrological stations) to estimate the erosivity factor R. This data was used to create a point map, which was then interpolated in ArcMap.

Erodibility is the erosion of soil resistance caused by runoff and rainfall. The vulnerability of different types of soil to erosion of water exhibits varying degrees (Thomas et al. 2018). Numerous soil physical and chemical characteristics have an impact on it. However, the RUSLE model only takes into account the physical characteristics of the soil, such as its structure, particle size, organic content, and permeability, which are the key factors affecting soil erodibility. The soil type map of the PRB was obtained from the soil survey in Pakistan, and soil types were categorized into four classes, namely, Glaciers, Haplic Xerosols, Lithosols and Eutric Cambisols. This was then used to calculate the factor of erodibility by giving the values of K obtained from various sources (Table 2).

The LS is a mixture of two topographical variables: slope length (L) and slope steepness (S), which greatly influences the erosion of the soil process (Gashaw et al. 2018). On a steep slope, the water rushes faster, causing increased pressure on the surface and, as a result, increasing the transport of many sediments. Slope length, which refers to the distance from the point where streamflow originates to either the point where the slope drops and sedimentation takes place or the point where the water can flow into discrete channels, also contributes to erosion (Haile & Fetene 2012).

The slope percentage suggests that the value of exponent m ranges from 0.2 to 0.5 (Table 3). Because the majority of the area in the study region has a steeper slope of 5%, 0.5 was taken as the m value from Table 3 for Equation (3).

Factor C takes into account the effects of the combination of cover and activity of management on the loss of soil. Vegetation can significantly slow down the runoff thus protecting soil from erosion. Anthropogenic activities are the major factors of change (Karaburun et al. 2010). Plant cover significantly decreases soil erosion because it intercepts rainwater, slows down rainfall, runoff and speeds up infiltration (Nedd et al. 2021). Remote sensing provides up to date and accurate data of earth surface on a large scale, which is extremely useful for understanding the earth surface dynamics (Karaburun et al. 2010).

The factor of P differs from the factor of C in that it describes the impact of strategic planning on the runoff by changing its direction, pattern and speed (Panagos et al. 2015). Some scientists suggested that the factor of P values is largely influenced by the gradient's slope (Nagendra et al. 2004), while others suggested calculating the value of factor P using agricultural production. The factor P can be calculated in various ways, such as by directly examining the type of land usage in the field and by identifying particular farming practices that are especially time- and money-consuming. Landsat-classified images were used in this study to develop land cover classes and compute the practice support factor (Table 4).

The results show that, in the last three decades, LULC in the study area changed significantly from 1990 to 2020 (Table 5). The built-up areas increased up to 19% while a decrease of 3% in barren land and 13% in vegetation was recorded in the study area.

The accuracy assessment and Kappa coefficient calculation for the years 1990, 2005 and 2020. The 1990 image showed producer's and user's accuracy scores above 80% in all categories. Overall accuracy and Kappa coefficient index were calculated at 81.16% and 0.81, respectively. In the 2005 image, producer's and user's accuracy scores exceeded 85% in all categories. Overall accuracy and Kappa coefficient index were calculated at 85.91% and 0.85, respectively. The 2020 classified image showed producer's and user's accuracy scores above 90% in all categories with a high accuracy index, with overall accuracy recorded at 90.12% and a Kappa coefficient index of 0.91, respectively, as shown in Table 6.

Extreme erosion of soil decreases not only the fertility and land productivity but also the capacity and effectiveness of dams and reservoirs by adding a significant amount of sediment to them. It takes a lot of time and effort to calculate and map the spatial extent of erosion of soil hazards, but the approach of RUSLE and GIS combination is a highly beneficial tool for measuring and mapping the soil erosion of a region. This research is significant because it provides personal knowledge of an area with high rates of erosion, which could assist planning experts and conservationists in reducing and examining the erosion of soil. The prevention and control of the erosion of soil will contribute to a rise in cultivated land output, and the removal of sediments will lengthen the lifespan of dams and reservoirs.

A study conducted by Maqsoom et al. (2020) employed the RUSLE model to assess the annual erosion of soil within the Chitral River Basin. The findings indicated a computed loss of soil of 58 tons/ha/year from the Chitral River Basin. Notably, the study site corresponds to the northern border of the PRB. Another study by Batool et al. (2021) in Pakistan's Potohar region suggested an average sediment yield of 4.3 tons/ha/year due to soil erosion. Moreover, a separate study undertaken by Ullah et al. (2018) in Punjab province's Chakwal district estimated an annual average soil removal of 18 tons/ha/year. This estimation pertained specifically to areas adjacent to rivers and hilly terrain.

Gilani et al. (2022) conducted an assessment of the yearly soil erosion in Pakistan using the RUSLE model. The study provided estimations for annual soil loss across various administrative divisions of the country. According to their findings, the annual soil loss in Khyber Pakhtunkhwa is approximately 11.78 ± 21.97 tons/ha/year. In a separate study, Nasir et al. (2023) determined that the average annual soil loss from the PRB is 10.25 tons/ha/year. It's noteworthy that the Panjkora Basin, the study area, falls within the geographical scope of Khyber Pakhtunkhwa. The current study's evaluation of annual soil loss stands at 16.1 tons/ha/year, which closely corresponds to the figure reported by Gilani et al. (2022) for Khyber Pakhtunkhwa and the 10 tons/ha/year estimation for the Panjkora Basin as suggested by Nasir et al. (2023).

The PRB is selected as the study area due to its fragile soil and having shown regular soil erosion. The increasing population, deforestation and conversion of agricultural land to the built-up areas are causing continuous pressure on land resources. Soil loss has posed adverse consequences to natural ecosystems and sources of livelihood. This study holds importance as it offers insights into a region characterized by high erosion rates. This study will prove helpful to urban planners, agriculturists and conservationists to mitigate soil erosion, aiding them in implementing effective watershed and sediment control. These measures will enhance agricultural yield and prolong the functional longevity of dams and reservoirs by preventing sediment accumulation. The finding reveals that in the study area water occupied 3.09% in 1990 while in 2020 it has reduced to 1.85%. In the same way, vegetation and barren areas have significantly reduced; in 1990, the vegetation covered 63.68% of the study area while, in 2020, it has reduced to 50.89%. Like the vegetation cover, the barren land in 1990 was 17.14%, and it decreased to 14.75% in 2020. The area under snow cover was 9.74% in 1990, which reduced to 6.31% in 2020. The built-up area has increased from 1990 to 2020 from 6.34 to 26.38%. The maximum actual annual soil loss per hectare was 5,195 tons in 1990, which gradually increased to 6,247 tons in 2005. Similarly, in the study region, during 2020, the actual annual soil loss was further accelerated and marked the figure of 8,297 tons/ha/year, whereas the average annual loss of soil per hectare is also increasing. It was found from the analysis that, in 1990, the average soil loss per hectare was 9.2 tons, which gradually increased to 12.4 tons in 2005 and in 2020 it was 16.1 tons/ha/year. This demands that government line agencies and policymakers start watershed management and conservation in the basin. Many of the limitations and uncertainties in our research could be attributed to existing RUSLE formulations, including uncertainty associated with the model's simple empirical nature and numerous subcomponents; data availability issues; and the model's inability to account for soil loss due to gully erosion, mass sliding events, or the prediction of soil erosion. For large-size basins, these results are significant for baseline data preparation but for accurate and comprehensive analysis, it is recommended to use high-resolution satellite images and DEM. The sub-basin study will also lead toward more appropriate results.

The paper is not submitted to any other journal.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.