Introduction

Soil erosion poses enormous threats to natural ecosystems and livelihoods worldwide because of its adverse in-situ and ex-situ consequences (FAO 2019; Bamutaze et al. 2021). The UN’s Food and Agricultural Organization (FAO) reports that soil erosion is escalating above tolerable thresholds in diverse geographic settings (FAO 2019). Approximately 30% of the planet’s land is severely degraded, with soil erosion widely known as the eminent cause of the debacle (UNCCD 2017). Estimates elucidate annual soil losses at global scales with enormous magnitudes ranging between 24 and 35.9 Petagram (Pg)/year (UNCCD 2017). Globally, soil erosion adversely affects terrestrial ecosystems, sometimes triggering gullying (Islam et al. 2020) and the subsequent land degradation posing an important threat to the environment (Islam et al. 2020; Bekele and Gemi, 2021). Various international development agendas are targeting addressing such threats posed on the world ecosystems including the Sustainable Development Goal (SDG) 15 of the United Nations Global Development Agenda 2030, which advocates for life on Land, aiming at the protection, restoration and sustainable management of terrestrial ecosystems (UN 2016). Regardless of water being the world’s most valuable asset (Clothier, 2015), aquatic ecosystems can be detrimentally impacted by erosion as the eroded sediments impair the water quality (Karamage et al. 2017), further jeopardizing the integrity of the earth’s natural systems (UNCCD 2017; Bamutaze et al. 2021).

Lake Victoria is the biggest natural water resource in East Africa, serving diverse roles, from fishing, navigation, eco-conservation to supplying fresh water for various purposes such as irrigation, industrial use and domestic use. According to the African Great Lakes Initiative, the Lake Victoria basin population in 2019 was over 40 million people, with a population growth rate of 3.5% per year, one of the highest rates in the world (AGLI 2020). The gigantic water body has suffered deteriorating water quality attributed mainly to human factors (Bongomin 2011; MWE 2018; Kolding et al. 2014). The water quality has been impaired by nutrient and sediment inputs into the lake, a manifestation of degradation in its catchment (Banadda et al. 2009; MWE 2018). The lake’s ecological health has been adversely affected by the rapidly increasing human population in the catchment, clearance of natural vegetation, and emissions of untreated effluent (Banadda et al. 2009). With the country’s population projected over 104 million by 2050, this explosion is poised to pose enormous pressure on natural resources, augmenting the risk of soil erosion by water if inadequate conservation practices are applied (Karamage et al. 2017).

The Murchison Bay, consisting of the inner Bay and the outer Bay, is located on the northern shores of Lake Victoria in Uganda. It is the primary water source for Kampala, the Capital City of Uganda, in addition to being the recipient of the runoff and discharged wastewater from the city (Banadda et al. 2009; World Bank 2015). The IMB is suffering from deteriorating water quality and increasing sedimentation (Bongomin 2011; Kayima et al. 2010; Banadda et al. 2009; Akurut et al. 2017; MWE 2018), attributed to both point and non-point pollution sources in its catchment. Even in minute amounts, sediment-adsorbed pollutants such as heavy metals pose grave ecological and human health concerns (Jaishankar et al. 2014; Sankhla et al. 2019; Huang et al. 2020). The pollution impacts are visible in the drainage network and the lake, as evidenced by the frequent algal blooms in Murchison Bay (Ssebiyonga et al. 2013; Haande et al. 2011). As a result, the treatment cost of potable water has increased (World Bank 2015) and the Murchison Bay aquatic ecosystem is at the peril of adverse pollution effects (Lake Victoria Environmental Management Project 2005; Tebandeke et al. 2020). The IMB catchment has two major factions; the eastern, primarily rural, and the western, predominantly semi-urban. Though several studies reckon urban areas as non-erodible zones (Karamage et al. 2017), there is evidence of soaring sediment yields and alarming sedimentation levels, especially in the western Bay (Banadda et al. 2009; MWE 2018). For instance, the sediment yield of only the Nakivubo sub-catchment of IMB was estimated at a rate of 26.5 t ha−1 year−1 and 26.4 t ha−1 year−1 for years 1995 and 2003, respectively (Anaba et al. 2017). There has been a significant loss of forest, wetland and other natural vegetation covers in the catchment due to anthropogenic activities (Anaba et al. 2017) with statistics showing declining forest cover in the Bay from 1274 ha (31%) in 1995 to 569 ha (14%) in 2003 (Anaba et al. 2017). The trend has been worse in recent decades making the catchment more prone to soil erosion and depriving the Bay of its natural buffering effect (Banadda et al. 2009; MWE 2018).

Erosion risk assessments and application of RUSLE in peri-urban and urban catchments have been understudied globally (Shikangalah et al. 2016; Marondedze and Schütt 2020), with urban settings often regarded as non-erodible zones (Karamage et al. 2017). Nonetheless, practically there are complex detachment and sediment transport mechanisms, attributed to the portions of built-up area and the large runoff volumes which trigger erosion and sediment transport within such catchments (Marondedze and Schütt 2020). Similarly in the IMB catchment, there has been no sufficient information on erosion hotspots, with systematic soil erosion assessments lacking in the catchment. However, the development and implementation of soil and water conservation (SWC) measures require a foundation of knowledge and the understanding of soil erosion risk (Bamutaze et al. 2021). Erosion risk assessments serve as a basis to give key insights for the design of appropriate interventions and environmental risk management (Almasalmeh et al. 2022). Thus potential erosion hotspots, which are zones with a high potential to contribute to the overall erosion from the catchment (Karamage et al. 2017), needed to be mapped and their potential erosion rates envisaged. To bridge existing knowledge gaps, the contribution from this research is therefore the assessment of the spatial patterns of erosion risk and identification of erosion hotspots in the IMB catchment to lay a knowledge base for tailored mitigation and conservation measures in this peri-urban catchment to ameliorate sediment pollution into Lake Victoria.

Materials and methods

Description of the study area

The study was for the Inner Murchison Bay (IMB) catchment, located on the northern shores of Lake Victoria, with a size of 282 km2 with drainage basins (sub-catchments): Nakivubo (40 km2) in the northwest, Kansanga (15.9 km2) and Ggaba (Kyetinda) (7.0 km2) in the west, Kinawataka (35.9 km2) and Portbell (2.3 km2) in the north, Namanve (86.7 km2) in the northeast, Nakiboga (17.5 km2) and Zirimiti (78.3 km2) in the East (Fig. 1). The IMB is a semi-enclosed small water body of the main Lake Victoria with an area of 18.4 km2 and length of 5.6 km (Fig. 1).

Fig. 1
figure 1

Map showing the location IMB catchment in Uganda

Full size image

Mapping erosion hot spots in the catchment using the RUSLE model

Revised universal soil loss equation (RUSLE) (Renard et al. 1997) was used to estimate erosion risk and the spatial distribution of the erosion hotspots within the IMB catchment. RUSLE model was built in a GIS environment using ArcGIS 10.5 software and a GISUS-M plugin (Oliveira 2015), focusing on several factors that influence erosion in the catchment. These included: land use/cover, soil data (soil type), topography (DEM), erosion control practices and rainfall data. The RUSLE model was run, producing a sediment hotspot map for the entire IMB catchment. This model was selected because of its minimal data requirements making it more suitable for data-scarce regions (Tamene et al. 2017) like the IMB catchment. The RUSLE model is described in Eq. 1 (Renard et al. 1997):

$$A=R\times K\times \mathrm{LS}\times C\times P,$$
(1)

where A = estimated average soil loss in t ha−1 year−1; R = rainfall erosivity factor (MJ mm ha−1 h−1 yr−1); K = soil erodibility index (t ha h ha−1 MJ−1 mm−1); L = slope length factor (dimensionless); S = slope steepness factor (dimensionless); C = land use/cover factor (dimensionless) and P = conservation/management factor (dimensionless).

The RUSLE parameters in Eq. 1 were computed using the input datasets in the ArcGIS software environment. GIS layers for all the contributing factors were generated and then the Raster Calculator in ArcGIS 10.5 software was deployed in computing the RUSLE model to produce the erosion hotspot map for the entire IMB catchment (Fig. 2).

Fig. 2
figure 2

Schematic layout of the procedure for assessing erosion risk using RUSLE model

Full size image

Rainfall erosivity factor (R)

R-factor is a key parameter for soil erosion studies and represents the erosive nature of rainfall for a given geographical location using the amount of rainfall and its intensity as inputs (Mekonnen and Melesse 2011). The general formula (Eq. 2) for computing rainfall erosivity factor R that was proposed by Wischmeier and Smith (1978) is:

$$R=\frac{{\mathrm{EI}}_{30}}{1000},$$
(2)

where R = rainfall erosivity factor, E = rainfall kinetic energy (J m–3) and I30 = 30 min rainfall intensity (mm h−1). Estimating the R-factor using the above equation required average kinetic energy intensity (EI) values for at least 20 years. However, that data was unavailable for IMB catchment. Instead, R was estimated with an alternative equation proposed by Hurni (1985) and commonly used in studies in East Africa and the Nile basin (Tamene and Le 2015; Njiru et al. 2018; Eniyew et al. 2021; Tsegaye and Bharti 2021; Jothimani et al. 2022) (Eq. 3).

$$R=0.562\times P-8.12,$$
(3)

where P is the long-term mean annual precipitation. In this computation, 39 years of CHIRPS rainfall data were downloaded from the CHIRPS website (https://www.chc.ucsb.edu) to produce the mean annual rainfall map in Fig. 3, with values ranging between 1198 to 1570 mm.

Fig. 3
figure 3

IMB catchment mean annual rainfall for 39 years (1980–2019)

Full size image

Soil erodibility factor (K)

K-factor is a quantitative description and measure of the susceptibility of soil particles to detachment and transport by rainfall and runoff (Karamage et al. 2017). The contributing soil properties for the K value include soil structure, texture, organic matter and permeability. Global FAO soil datasets were downloaded from the FAO website (https://data.aps.fao.org/map/catalog/), where the IMB catchment soils were extracted for use in this study (Fig. 4). The K-factor for the IMB catchment was computed from FAO soil data using Eq. 4 (Wang et al. 2021).

Fig. 4
figure 4

a Map of major soil types in the IMB catchment. b Map of IMB catchment DEM

Full size image
$$K={f}_{\mathrm{csand}}\times {f}_{\mathrm{cl}-\mathrm{si}}\times {f}_{\mathrm{orgc}}\times {f}_{\mathrm{hisand}}.$$
(4)

With ms, msilt, mc and orgc the percentage (%) compositions of sand, silt, clay and organic carbon, respectively. The factors in Eq. 4 are described as:

fcsand factor; gives low soil erodibility factors for soils with high coarse sand contents and high values for soils with little sand.

$${f}_{\mathrm{csand}}= \left(0.2+0.3\times \mathrm{exp}\left[-0.256\times \mathrm{ms}\times \left(1-\frac{\mathrm{msilt}}{100}\right)\right]\right)$$
(5)

fcl-si factor; gives low soil erodibility factors for soils with high clay to silt ratios.

$${f}_{\mathrm{cl}-\mathrm{si}}=\left(\frac{\mathrm{msilt}}{\mathrm{mc}+\mathrm{msilt}}\right)$$
(6)

forgc factor; that reduces soil erodibility for soils with high organic carbon content.

$${f}_{\mathrm{orgc}}=\left(1-\frac{0.0256\times \mathrm{orgC}}{\mathrm{orgC}+\mathrm{exp}\left[3.72-2.95\times \mathrm{orgC}\right]}\right)$$
(7)

fhisand factor; reduces soil erodibility for soils with extremely high sand contents.

$${f}_{\mathrm{hisand}}=\left(1-\frac{0.7\times \left(1-\frac{\mathrm{ms}}{100}\right)}{\left(1-\frac{\mathrm{ms}}{100}\right)+\mathrm{exp}\left[-5.51+22.9\times \left(1-\frac{\mathrm{ms}}{100}\right)\right]}\right)$$
(8)

The K-factors were multiplied with 0.1317 value to convert the K-factor from the American system to the metric system unity/International System of Unit (Karamage et al. 2017; Bamutaze et al. 2021). The primary soil types in the catchment were gleysols and ferralsols (acric and lixic). Soil erodibility values were categorized as per IWR (2002) classes (Table 1).

Table 1 Thresholds for categorization of soil erodibility (K-factor) as per IWR (2002)
Full size table

Slope factor (LS)

Soil erosion is highly correlated with slope and therefore increases rapidly with slope (David 1988). L stands for the effect of slope length on soil erosion. It is the ratio of field soil loss to the corresponding soil loss from a 22.13 m length on the same soil type (Uddin et al. 2016). The S represents the effect of slope steepness on erosion (Uddin et al. 2016). An SRTM DEM (Fig. 4) was downloaded from the USGS website (https://earthexplorer.usgs.gov), and the IMB catchment was delineated for computations to determine the slope factor (LS).

The L-factor was computed using Eq. 9 developed by Desmet and Govers (1996) while the S-factor was estimated using the McCool et al. (1987) method in Eq. 12.

$${L}_{ij-\mathrm{in}}=\frac{\left[{\left({A}_{i,J-\mathrm{in}}+{D}^{2}\right)}^{m+1}-{\left({A}_{i,j-\mathrm{in}}\right)}^{m+1}\right]}{\left({D}^{m+2}\right)\times \left({{x}_{i,j}}^{m}\right)\times {\left(22.13\right)}^{m}}$$
(9)
$$m=\frac{\beta }{1+\beta }$$
(10)
$$\beta =\frac{\mathrm{sin}\theta /0.0896}{3{\left(\mathrm{sin}\theta \right)}^{0.8}+0.56}$$
(11)
$${S}_{i,j}=\left\{\begin{array}{c}10.8sin{\theta }_{i,j}+0.03, tan{\theta }_{i,j}<9\%\\ 16.8sin{\theta }_{i,j}-0.50, tan{\theta }_{i,j}\ge 9\%\end{array}\right\},$$
(12)

where Li,j-in = slope length for grid cell (i, j); Ai,j-in = contributing area at the inlet of the grid cell with coordinates (i, j) (m2); D = grid cell size (m); m = length exponent of the USLE L-factor; xij = (sin α i, j + cos α i, j). β is the ratio of rill to inter-rill erosion for conditions when the soil is moderately susceptible to both rill and interrill erosion; θ is the slope angle in degrees (Oliveira 2015; Karamage et al. 2017). For efficiency, computing the LS-factor from the DEM of IMB catchment was accomplished using the ArcMap GISUS-M plugin (Oliveira 2015), whose LS Tool has inbuilt algorithms for automatically computing the LS factor from DEM (Oliveira 2015).

Cover-management factor (C)

C-factor represents the effect of ground and vegetation cover on the reduction of soil loss by reducing rainfall and runoff (Bekele and Gemi 2021). Land cover is the second most important factor next to topography controlling soil erosion risk (Uddin et al. 2016). The land cover intercepts rainfall, increases infiltration, and reduces rainfall energy. The IMB catchment land use map was generated in ArcGIS 10.5 software through supervised classification of a Landsat 8 satellite image of the IMB catchment downloaded from the USGS website (https://earthexplorer.usgs.gov), dated 12 December 2021. Averaged literature values for C-factors from similar studies in Sub-Saharan Africa were then assigned to particular land cover classes as illustrated in Table 2.

Table 2 Land use and land cover (LULC) descriptions and their averaged C values
Full size table

Support practice factor (P)

P-factor represents the impacts of support practices such as contouring or strip cropping on the erosion rate (Bekele and Gemi 2021; Panditharathne et al. 2019). Literature values for the P-factor, (Table 3) for the different land uses in IMB catchment were assigned to the land use Arc Map layer. The major crop cover was in the eastern catchment, which is relatively flat and hence the average C-factor for agricultural lands (0.5) (David 1988; Naqvi et al. 2012) was assigned and a 30 m resolution P-factor raster map for the catchment was generated.

Table 3 P values for the land use in IMB catchment
Full size table

Erosion risk

RUSLE predicted soil loss was categorized using the following thresholds: very slight, slight, moderate, high, severe and very severe, as shown in Table 4 (Bamutaze et al. 2021). The descriptive statistics encompassing the minimums, maximums, means, and standard deviations were obtained for each of RUSLE factors and the modelled erosion risk map. To envisage the spatial disparities of the input RUSLE layers and modelled erosion, mean and standard deviation values were utilised to compute the coefficients of variation (CV).

Table 4 Categories for RUSLE predicted soil loss
Full size table

Field walks for identifying other erosion sources

The walks were centered on field observations to further understand the model results and identify other erosion hotspots other than those captured by the RUSLE model. As a recognized ecological monitoring tool, systematic walks (FAO 2015; Rojas and West 2021) and field investigations (Islam et al. 2020) were applied in this study to aid the further understanding of the erosion hotspots and sediment connectivity, given the complex drainage network of the IMB catchment in addition to the diverse land uses. The walk design was guided by the mapped erosion hotspots from the RUSLE model and the drainage network of the catchment since proximity to the drainage channels accelerates the transfer of the eroded soil downstream as sediment (Ali et al. 2021). Another consideration for these walks was accessibility, and therefore the road network was given utmost consideration, especially in the swampy and remote eastern part of the catchment. The map in Fig. 5 illustrates the field walks in the catchment.

Fig. 5
figure 5

Map showing the field walks in the IMB catchment

Full size image

Results

RUSLE factors

R-factor

The catchment exhibits a distinct rainfall erosivity gradient (Fig. 6) influenced by a wetter eastern side and gradually drying towards the west. The rainfall gradient directly correlates to the erosivity distribution. The R-factor results for the study area ranged from 665 to 874 MJ mm ha−1 h−1 year−1 (Fig. 6), with a mean value of 774 MJ mm ha−1 h−1 year−1. The Zirimiti sub-catchment is shown to have the highest rainfall erosivity whereas the lowest erosivity was calculated for the Kansanga sub-basin. The Namanve sub-catchment is located in the transition between the high and low erosivity zones. With a standard deviation of 49 MJ mm ha−1 h−1 year−1, the catchment had a low computed coefficient of variation (CV) for the R-factor of 6.33%.

Fig. 6
figure 6

The R-factor map for IMB catchment

Full size image

K-factor

The predominant soil types in IMB catchment are Gleysols and Ferralsols, covering 45.4% and 54.6% of the catchment area, respectively. Both soil types exhibited low erodibility values as per Institute for Water Resources, IWR (2002) erodibility classes (Table 1). The computed catchment K-factors ranged from a minimum of 0.018 to a maximum of 0.019 t ha h ha−1 MJ−1 mm−1 for Gleysols to Ferralsols respectively, as summarised in Table 5 and Fig. 7. The central catchment characterised by gleysols and lower erodibility, in contrast to the Western parts of Ggaba, Kansanga, Nakivubo and Kinawataka; Northern Namanve and Eastern Zirimiti sub-catchments (Fig. 7). The catchment erodibility had a mean value of 0.0185 t ha h ha−1 MJ−1 mm−1, a standard deviation of 0.0007 and a coefficient of variation (CV) of 3.81%, signifying a low variation in erodibility of the catchment according to Bamutaze et al. (2021).

Table 5 Calculated K-factor values
Full size table
Fig. 7
figure 7

K-factor map for IMB catchment soils

Full size image

Slope length and steepness (LS-factor)

Computed LS-factor values ranged from as low as 0.01, shooting to a maximum of 30.5 for certain hills as shown in the map in Fig. 8. Spatially, the Western sub-catchment of Ggaba, Kansanga and Kinawataka exhibited high LS values. Central IMB catchment had the lowest LS-factors, attributed to the flat nature of the predominantly wetland central sub-catchments including Namanve, Nakiboga and Western Zirimiti. The LS-factor eventually rises in the South Eastern Zirimiti sub-catchment (Fig. 8). LS-factor had a mean value of 1.33, a standard deviation of 1.93 and a coefficient of variation (CV) of 145.11%, thus portraying a high variance in the slope of the catchment.

Fig. 8
figure 8

LS-factor map for IMB catchment

Full size image

Cover-management (C-factor) and support practices (P-factor)

From the land cover map (Fig. 9), 23.2% of the area was built-up, 16.2% semi-built up (25–50% green), 8.6% forest area, 21.9% wetland cover, 28.1% cropland and 2% bare surfaces (bare soils). The western sub-catchments of Ggaba, Nakivubo, and Kinawataka were predominantly built-up, while the East was majorly covered with croplands. The central catchment is a transition from built-up to crop and forest cover, but also with a large wetland cover (Fig. 9). The C-factors in the catchment ranged from zero for open water and wetlands, to a maximum of 1 for bare surfaces (soils) (Fig. 9), as per assigned literature values in similar Sub-Saharan Africa studies (Table 2). The cover factor of IMB catchment had a mean value of 0.131 and a standard deviation of 0.387.

Fig. 9
figure 9

a IMB catchment 2021 land use and cover map, b C-factor map for IMB catchment and c P-factor map for IMB catchment

Full size image

Croplands comprised 28.1% of the IMB catchment area, while other land uses claimed 71.9% of the catchment area. Thus as per the guiding literature (Table 3), P-factor values ranged from 0.5 for agricultural croplands to 1 for other land uses (Fig. 9).

Erosion risk (RUSLE based)

RUSLE model results indicated the percentage catchment areas for the various erosion rates as; 66.7% for 0–2 t ha−1 year−1, 10.8% for 2–5 t ha−1 year−1, 10.1% for 5–10 t ha−1 year−1, 9% for 10–50 t ha−1 year−1 and 3.3% for 50 and 100 t ha−1 year−1 erosion rates (Fig. 10). The average watershed erosion rate was 7 t ha−1 year−1, ranging in moderate levels as per the guiding categorization in Table 4 (Bamutaze et al. 2021). Average erosion rates for the sub-catchments are summarized in Table 6.

Fig. 10
figure 10

IMB catchment modeled soil erosion by the RUSLE model

Full size image
Table 6 Table showing the average erosion of each sub-catchment
Full size table

Major erosion sources

Prone zones: Generally, RUSLE predicted erosion-prone areas comprised of bare surfaces, croplands and steep areas (LS-factor > 20) as the major erosion hotspots (> 5 t ha−1 year−1), with erosion rates up to 100 t ha−1 year−1 in a few worst case scenarios (Fig. 11).

Fig. 11
figure 11

Images from field walks for validating RUSLE predicted erosion-prone areas in IMB catchment

Full size image

Drainage channels: Field walks revealed that unlined drainage systems in the catchment are major erosion hotspots (Fig. 12), yet the RUSLE model does not consider that. Therefore, the modelled erosion values could be significantly higher than those reported from the RUSLE model, especially for several urban sub-catchments, given the high runoff values generated from the large paved surfaces.

Fig. 12
figure 12

Other identified erosion hot spots from field walks

Full size image

Exposed/bare soils: Bare soils have increased erosion risks since the erosive energy of rainfall can detach and transport the soil particles with relative ease on such surfaces (Uddin et al. 2016). Such surfaces were very common in the sub-catchments and included: unpaved (marram/gravel roads), excavated areas, fresh developments/construction sites, football pitches, home compounds and exposed workspaces such as garages and parking spaces (Fig. 12).

Discussion

RUSLE factors

The increasing R-factor values eastwards of the IMB catchment are attributed to the higher rainfall values in the East (1570 mm) as compared to the Western side with rainfall lowering to 1198 mm. The attained R-factors in this study are comparable to results from similar studies in East and Central Africa in countries like Ethiopia, Kenya and the DRC (Bekele and Gemi 2021; Jothimani et al. 2022; Tsegaye and Bharti 2021; Njiru et al. 2018; Eisenberg and Fabrice 2020), where values ranged between 201 and 1066 MJ mm ha−1 h−1 year−1. The computed R values were as well comparable to the ranges of EU Soil Data Centre (https://esdac.jrc.ec.europa.eu) global R-factors, where East African values generally range between 700 to 1700 MJ mm ha−1 h−1 year−1.

From the IWR (2002) classification as shown in Table 1. Low soil erodibility (0.006585–0.02634 t ha h ha−1 MJ−1 mm−1) is for clay soils that are not easily detached or sandy soils that have a low runoff. Moderate erodibility (0.02634–0.05268 t ha h ha−1 MJ−1 mm−1) signifies medium-textured soils, while high erodibility (> 0.05268 t ha h ha−1 MJ−1 mm−1) denotes soils with high silt content and prone to more runoff and detachment. The spatial distribution of erodibility values is thus determined by the soil type, and with the predominant soil types in the IMB catchment being Gleysols and Ferralsols; both exhibited low erodibility, implying that the K-factor is not a major factor in the IMB catchment erosion risk, as compared to other RUSLE factors. For slope, the range of LS-factor is a reflection of the complex topographical variation from the West, Central to the East IMB catchment. Thus the high coefficient of variation (CV) of 145.11% is a representation of the heterogeneity in slopes within the catchment.

The spatial distribution of C-factors in the IMB catchment is a depiction of the underlying determinants, which are the different land uses and land covers. Since the eastern catchment was predominantly an agricultural area, it exhibited the highest magnitude of C-factor (0.3), attributed to the susceptibility of croplands to erosion (Eisenberg and Fabrice 2020; Bamutaze et al. 2021; Marondedze and Schütt 2020). Nonetheless, significant forest cover in the eastern area can limit excessive erosion, hence a C-factor of 0.011 (Eisenberg and Fabrice 2020; Eniyew et al. 2021; Ligonja and Shrestha 2013; Tsegaye and Bharti 2021). Since the central IMB catchment was predominantly wetland covered, it reduces the erosion risk, thus the lowest C-factor (0), (Armour and Lait 2014; Rozos et al. 2013). The western catchment is majorly built-up, which reduces the soil exposure to detachment. Similar studies in Sub-Saharan Africa settings have highlighted such built-up and semi-built zones as low-prone areas to erosion although not necessarily resistant, with average C values of 0.035 and 0.11 for built-up and semi-built settings, respectively (Eisenberg and Fabrice 2020; Eniyew et al. 2021; Ligonja and Shrestha 2013; Marondedze and Schütt 2020). Non-agricultural land covered 71.9% of the IMB catchment with no agricultural erosion control practices where the P-factor for such areas is 1 (David 1988; Naqvi et al. 2012). The major crop cover being in the eastern catchment (Fig. 9), the average C-factor (0.5) for agricultural croplands (David 1988; Naqvi et al. 2012) was thus assigned.

Modelled erosion

RUSLE results exhibited a wide range of variations in erosion risk for the different sub-catchments of IMB watershed and according to Bamutaze et al. (2021), Uganda soil loss tolerance value is 5 t ha−1 year−1. The erosion proneness of the Eastern catchment (Nakiboga and Zirimiti sub-catchments) was attributed to the higher rainfall values, with erosivity values surging as high as 871 MJ mm ha−1 h−1 year−1. Furthermore, these sub-catchments are predominantly agricultural areas with high C values as compared to the urban western sub-catchments. RUSLE predicted erosion-prone areas in these eastern sub-catchments as croplands on the slopes, with erosion rates up to 100 t ha−1 year−1 in worst scenarios. However, the overall effect of this potential erosion on IMB ecological health is anticipated limited, as confirmed by the field walks. This is due to the large buffering effect of the Zirimiti and Nakiboga wetlands in addition to the limited effluents in these rural sub-catchments. The middle catchment is relatively flat with LS-factor as low as 0.01, in addition to a predominant wetland cover whose C-factor is 0. This contributed to the lowest erosion risk in the central IMB catchment (Fig. 10).

Nonetheless, most western sub-catchments such as Nakivubo, Kansanga, Gaba (Kyetinda), Kinawataka, Portbell and Namanve are predominantly urban with most surfaces paved or semi-built up. RUSLE model calculated erosion risk with the soil K value cancelled out by the corresponding C value in such paved areas. However, these western sub-catchments are hilly with high LS-factors of up to 30.5. Thus in those catchments, RUSLE predicted higher erosion values on hills and semi-built up as well as agricultural areas in comparison to paved areas such as the Kampala city center. Predicted erosion-prone areas included Mbuya hill, Biina, Banda-Kyambogo, Kireka (Kiganda hill zone), Luzira farmlands and Makindye Lukuli (Fig. 11). Nevertheless, RUSLE modelled values for specific areas such as Kololo seemed slightly higher than the practical realities from field walks, attributed to their semi-built up and hilly nature with high LS-factor values.

Robust approaches were adhered to during modelling in this research and as a detachment model, RUSLE estimated the erosion risk with reliable results as confirmed from the field walks (Fig. 11). The field walks served a comparison purpose for further understanding of the RUSLE model results and ground realities as well as identification of other erosion hotspots than those captured by the model. Generally, most soil erosion studies in Sub-Sahara African countries including Uganda concentrated majorly on hilly (mountainous), agricultural areas as well as countryside catchments (Eisenberg and Fabrice 2020; Karamage et al. 2017; Tsegaye and Bharti 2021; Bamutaze et al. 2021), with minimum focus on urban and peri-urban settings. Karamage et al. (2017) conducted an erosion risk assessment in Uganda, identifying soil erosion hotspot areas that were high priorities. RUSLE model predictions showed that the mean rate of soil loss risk in Uganda’s erosion‐prone lands was 3.2 t ha−1 year−1, with a total annual soil loss of about 62 million tons in the year 2014. The study revealed that out of 112 Ugandan districts, 66 were found having unsustainable estimated soil loss rates > 1 t ha−1 year−1 (Karamage et al. 2017). Six districts were found to have mean annual soil loss rates of > 10 t ha−1 year−1 including Bududa (46.3 t ha−1 year−1), Kasese (37.5 t ha−1 year−1), Bundibugyo (28.9 t ha−1 year−1), Bulambuli (20.9 t ha−1 year−1), Sironko (14.6 t ha−1 year−1) and Kotido (12.5 t ha−1 year−1) (Karamage et al. 2017). In the findings, however, the estimated soil erosion risk categorised Kampala district and IMB catchment under non-erosive lands, attributed to its built-up and peri-urban nature. Thus, findings from Karamage et al. (2017) do not precisely align with the unveiled erosion threat in IMB catchment investigated further in this study.

Though often regarded as non-erodible zones with significant portions of paved surfaces (Karamage et al. 2017), peri-urban catchments are associated with high runoff volumes and complex detachment mechanisms which instigate erosion and sediment transport (Marondedze and Schütt 2020). Globally, erosion risk assessments have majorly focused on agricultural and rural landscapes (Islam et al. 2020; Bekele and Gemi 2021; Wang et al. 2021), with RUSLE model application in peri-urban catchments not exhaustively investigated (Shikangalah et al. 2016; Marondedze and Schütt 2020). From similar research in a peri-urban setting of Harare Metropolitan Province in Zimbabwe, 40% of the Epworth district was threatened by increased soil erosion risk within the built-up areas leading to unsustainable soil loss (Marondedze and Schütt 2020). The predicted mean potential soil erosion rate was 13.2 t ha−1 year−1 between the years 1984–2018 and vulnerable areas to erosion were foot slope areas with direct tributaries to the major streams and steep sloping zones (Marondedze and Schütt 2020). Such erosion patterns from Marondedze and Schütt (2020), are in agreement with the findings of this study from an identical setting of IMB catchment. Findings from this research suggest that RUSLE model is applicable in peri-urban settings, with proper consideration of local catchment characteristics prior to the utilization of the model outputs. A high erosion risk from the catchment is a significant ecological threat to the ecosystem with sediment deposits alongside their adsorbed pollutants such as heavy metals, nutrients and microplastics (Banadda et al. 2009), posing grave health concerns to humans and the ecosystem flora and fauna. Literature findings from earlier studies confirm this increasing sedimentation of IMB and the deteriorating water quality in Lake Victoria as a whole (Bongomin 2011; Kayima et al. 2010; Banadda et al. 2009; Akurut et al. 2017; MWE 2018).

Conclusions and recommendations

Results from this study unveiled a wide range of soil erosion hotspots, which are potential contributors to the overall sediment yield from the catchment into the Inner Murchison Bay (IMB) of Lake Victoria. The two ecological monitoring tools/approaches (RUSLE model and field walks) used in the catchment study unveiled unique attributes of erosion dynamics in the fragile IMB catchment. Where the RUSLE model identified croplands in hilly zones (LS-factor > 20) as the major erosion risk hotspots (> 5 t ha−1 yr−1), while the field walks identified bare surfaces, murram (gravel) roads and unlined drainage channels as other major erosion sources, especially in the semi-urban catchment. This serves as a basis for identifying mitigation priorities and its recommended that tailored soil and water conservation measures be integrated into physical planning, focusing on identified non-conventional hotspots to ameliorate sediment pollution in Lake Victoria. Nonetheless, this study encountered some limitations as some areas in the predominantly rural eastern catchment were not easily accessible due to the poor road network and multiple wetlands. Although RUSLE is regarded as the leading model in soil erosion risk assessment, further studies should focus on deploying multiple approaches in erosion mapping to envisage sediment connectivity and yield. Furthermore, the short temporal scales of erosion such as monthly or event basis were not addressed and should be studied more, since the annualized RUSLE erosion does not explicitly take it into consideration.