Soil Erosion Susceptibility Mapping of Imo River Basin Using Modified Geomorphometric Prioritisation Method

The Imo River Basin in South-Eastern Nigeria, West Africa has a total area of 2494.68 km². Geographically, it is located between latitudes 4°50′–6°00′N, and longitudes 7°05′–7°35′E. The Imo River originating from Okigwe in Imo State is the major river draining the basin (Amangabara 2015). The maximum elevation is found in the upper part of the basin with a height of 417 m above mean sea level (MSL), and the minimum height is in the lower part of the basin with a height of 5 m above MSL. The basin extent is shared by three dominant states namely – Imo, Abia and Rivers. Figure 1 presents a map showing the location of the Imo River Basin. The surrounding states have a hot, semi-hot equatorial type of climate. The area experiences heavy rainfall, with a mean annual precipitation of 2000–2400 mm and an average number of 152 rainy days especially during the wet season (April–October). The distribution of superficial precipitation is skewed, with highs in July and September and a 2-week break in August. The rainy season starts in March and lasts until October or early November (NIMET 2019). Rainfall also hits its highest at night and during the early morning hours. The greater annual levels of rainfall and days of rainfall contribute to significant amounts of runoff. Differences in the amount of rainfall happen from year to year, however, usually between 1990 and 2200 mm (NIMET 2019). As a result, agricultural practices require little irrigation. This also promotes luxuriant forests beyond agriculture and recharging water sources from the soil and surface. Relative humidity during the dry and rainy seasons alternates between 75 and 90%. January to March are the hottest months, with annual average temperatures >20 °C.

Fig. 1

The Imo River Basin is predominantly occupied by the Igbo-speaking tribe. With good soil for agriculture and water for fishing, the main occupation is farming and fishing as it provides a source of livelihood to the populace. Extraction of minerals such as crude oil by both skilled and unskilled labourers are also important activities within the basin. Manufacturing industries abound in this region's metropolitan areas. The whole South-East zone falls within Nigeria's forest and derived savanna belt. Much of the forestland has been transformed into oil palm tree-dominated farmlands and palm bushes. The region's major environmental concerns are deforestation, soil erosion and biodiversity depletion (Okali et al. 2001).

This study utilised the Shuttle Radar Topography Mission (SRTM) v3.0 Global DEM acquired from the United States Geological Survey (USGS) web portal (earthexplorer.usgs.gov/). The SRTM DEM is provided in 1 × 1 tiles at 1 arc-second (30 m) and is defined on WGS 84 datum. It is also referenced to the EGM96 geoid model. Hence its heights are orthometric with vertical units in metres. The SRTM DEM's global coverage provides tremendous advantages for wide-area environmental modelling, especially when data availability is a problem. SRTM v3.0 which was released in September 2014 reveals the total detail of the world's landforms as originally captured by SRTM in the year 2000 (NASA JPL 2014). However, in providing reasonably reliable data to define drainage basin morphometry, several studies have shown that SRTM DEM is far better than the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) DEM (Sun et al. 2003, Farr et al. 2007, Forkuor, Maathuis 2012, Patel et al. 2016). Figure 2 presents a workflow of the methodology.

Fig. 2

DEM pre-processing is the first stage of the processing carried out on the DEM to make it more hydrologically correct and to improve on its terrain characterisation. The pre-processing steps to be implemented are dependent on the type of terrain, and on the type of analysis to be carried out. The following pre-processing steps were executed within the ArcGIS 10.3 software environment – fill sinks, flow direction and flow accumulation. Thereafter, the watershed was delineated. The delineated watershed boundary was further segmented into SWs using the ArcHydro extension tools for ArcGIS 10.3. The drainage network of the SWs was extracted from the pre-processed DEM using the ArcHydro tools. ArcHydro uses a rational, efficient and reliable algorithm. In the next phase, the stream ranking (stream ordering) was carried out following the stream ranking system initially introduced by Horton (1945) and later amended by Strahler (1964). According to Strahler (1964), all rivers and streams of the fingertips are designated as streams of the first order and where two of them join; they form a stream of second order. Similarly, two streams of the second order join to form a stream of the third order, and so forth. Stream order increases only when streams of the same order converge. The convergence of a first-order and second-order link will therefore remain a second-order link, instead of creating a third-order link. The categorisation of rivers and streams within drainage basins can be accomplished by conventional methods such as using topographic maps and ground observations, or by using advanced techniques such as remote sensing and GIS (Macka 2001, Maidment 2002, Altaf et al. 2014).

The morphometric analysis includes linear character estimation, channel network slope and contributing drainage basin ground slopes (Nautiyal 1994). The morphometric analysis for the different sub-basins was carried out by measuring the linear, shape and relief morphometric parameters of the basin. Areas and perimeters of the SWs were assessed by measuring the geometry of the derived sub-watershed polygons in the ArcGIS environment. Morphometric analysis is based on the overall drainage system activity. The drainage pattern refers to the spatial correlation between streams or rivers which may be impacted by the region's steepness, soil, rock rigidity, framework and geological history inequalities in their erosion. The morphometric research uses some very important elements (linear, shape and relief parameters) to prioritise the SWs in water supply management. In this study, extraction and evaluation of the morphometric parameters in each of the SWs were ranked using their relationship to soil erodibility. The overall number of ranks assigned in this method is based upon the number of watersheds. Ranks were numbered from 1 to 25, based on the number of SWs delineated. Rank 1 was assigned in such a way that the parameter value represents the maximum contribution to erodibility and rank 25 is a minimum contribution. The formulas used to derive the necessary morphometric parameters pertaining to linear, areal, and relief aspects of the watershed are presented in Table 1.

Erodibility prioritisation is the hierarchical ranking of various SWs in relation to soil erodibility, and the required soil and water conservation measures are taken according to the priority order. The WSA approach is effective and ideally suited to prioritizing watersheds. In the present study, sub-watershed wise delineation was done by using a newly developed prioritisation methodology where geospatial approaches are coupled with a statistical approach to allow the correct sub-watershed ranking, by avoiding the individual bias associated with several influencing parameters. Various morphometric parameters were rated in this method in terms of their relationship with erodibility, and correlation analysis among these parameters was performed. Therefore, it is hypothesised that an incentive exists to rectify the incorrect findings in order to increase the prioritisation accuracy. This method was useful in generating and assigning weights to each input variable according to their significance.

Secondly, the correlation between erosion assessment (morphometric) parameters using Pearson's product-moment correlation (r) was identified. The sum of all correlations in the columns was calculated and recognised as ‘sum of correlation’. The totals of the ‘sum of correlation’ yielded the ‘grand total’. Also, the final weights (Fw) of the morphometric parameters were calculated by dividing the ‘sum of the correlation coefficient’ of each parameter by the grand total of correlations (Table 6).

The Fw were calculated using (Eq. (1).

Finally, using Eq. (2), the compound parameter constant (Cpc) values of all SWs in the Imo-River Basin were calculated based on the weights of each morphometric parameter (metrics). A final erosion susceptibility map of the study area was prepared using the Cpc values. (2) Cpc=∑(Pri×Fwi) {\rm{Cpc}} = \sum \left( {\Pr i \times {\rm{Fw}}i} \right) where: –

Pri is the preliminary rank of ith parameter for sub-watershed, and

To systematically and qualitatively evaluate the erosion susceptibility metrics, the method of receiver operating characteristics (ROC) curves was used to evaluate the model's prediction accuracy. ROC curves are useful in evaluating the performance of binary classifiers (Zhang et al. 2015) and can discriminate between the observed and predicted variables based on the logistic regression S-shaped curve. This validation technique is also known as relative operating characteristics (Pontius, Schneider 2001). The ROC curve technique follows a binary classification rule which results in four potential outcomes: true positive, true negative, false positive, and false negative. If a correct outcome is properly identified as positive, a true positive exists, while a false positive defines a negative outcome as being incorrectly positive. On the negatives, if a negative outcome is properly recognised as negative, a true negative exists, and a false negative wrongly recognises a true result as negative (Carter et al. 2016). The relationship between the sensitivity and 1 – specificity is graphically represented as the ROC curve and the area under the curve (AUC) is the accuracy of the predicted results. Sensitivity is characterised as the measure of positives correctly identified as such or the likelihood of a positive test, often referred to as the true-positive rate Eq. (3). Specificity is the proportion of negatives correctly classified as such or the possibility of a negative test. This is also referred to as the true-negative rate. The false positive rate (1 – specificity) is the proportion of incorrect positive results that are actually negative Eq. (4). The false negative rate (1 – sensitivity) is the proportion of incorrect results that are actually positive (Nelson et al. 2005). By plotting the true-positive rate (sensitivity) on the y-axis against the false-positive rate (1 – specificity) on the x-axis, ROC curves are formed.

For the validation of the erosion susceptibility result, the locations of prevalent erosional occurrences (shown in Table 2) were obtained from the literature and records of recent events from online sources (Amangabara et al. 2015, Amangabara et al. 2018, Channels Television 2018). With this technique, the ROC curve tool embedded in the SPSS software environment was deployed and the ROC curve plot was produced where the AUC was determined. The coordinates of the curve are shown in Table 3.

The watershed processing resulted in the delineation of the Imo River Basin with 25 SWs. Figure 3 presents a map showing the basin's SWs, with SW5 (656.12 km²) being the largest sub-watershed and SW6 (35.53 km²) being the smallest sub-watershed in terms of areal extent. The Imo River which is the major river in the basin has its source in the Okigwe upland located in SW16, and flows north to south through SW15, SW14, SW10, SW8, SW7, SW5, SW3 downwards to the basin outlet (SW1). The river flows across some major towns such as Okigwe, Abekenta, Ugwuaku, Umuzi, Ezinachi, Ife, Okpuala and Abiama in each SW, having its longest length of flow (105 km) in SW5. It is joined at SW3 by its major tributary (Otamiri River) flowing from the north and exits the basin at SW1.

Fig. 3

The results of the computation of basic morphometric parameters are shown in Table 4. The drainage networks in the Imo River Basin transport water and sediments of the basin through a single outlet (Imo River), which is marked as the maximum order of the basin and conventionally the highest order stream (6th order) available in the basin considered as the order of the basin. Based on stream order, Imo River Basin was graded as 6th order with an area of 2494.68 km² and a perimeter of 746.72 km. Figure 4 presents a map showing the basin's stream order. The cumulative number of streams in the basin is 1954. The total length of the streams in various orders is 2746.06 km.

Fig. 4

Some of the morphometric parameters are directly considered as soil erosion metrics and thus are referred to as erosion risk assessment parameters (Biswas et al. 1999). These include the linear, shape and relief morphometric parameters as shown in Table 5. Linear and relief morphometric parameters have a direct/positive association with erodibility, i.e. the higher the values of these parameters, the greater the erodibility in the area and vice versa (Javed et al. 2009). In contrast, shape morphometric parameters have an inverse relation to erodibility i.e. the lower the values of these parameters, the higher the erodibility in the field of analysis and vice versa (Ratnam et al. 2005).

These parameters include Drainage density (Dd), Stream frequency (Fs), Mean Bifurcation ratio (Rbm), and Drainage texture (Dt). A lower Dd in any watershed implies a porous sub-surface soil, strong vegetation-cover, low-relief and vice-versa (Harlin, Wijeyawickrema 1985, Nag, Chakraborty 2003, Prasad et al. 2008, Vaidya et al. 2013, Farhan, Anaba 2016, Farhan et al. 2017, Ali et al. 2018). A study by Asfaw and Workineh (2019) on Gumara and Ribb watershed, Ethiopia, shows a Dd of 0.26 km⁻¹ and 0.30 km⁻¹ which is low and implies that the basin experience highly permeable sub-surface, low erosion exposure and better groundwater potential. Similarly, in the Imo River Basin, the lowest volume of Dd was observed in SW10 (0.94 km⁻¹), which indicates that this sub-watershed has a high degree of porosity among the other SWs, or it has the highest tendency to resist erosion if only the drainage parameter is taken as a metric for susceptibility to erosion.

The Fs has an inverse correlation with porosity and a positive correlation to watershed relief (Montgomery, Dietrich 1992). High Fs indicates that there is rocky terrain in the watershed and low infiltration ability which correlates to further erosion and vice versa (Altaf et al. 2014). In the Imo River Basin, the Fs ranges from 0.49 km⁻² for SW13 to 0.95 km⁻² for SW12. Based on the Fs, these results indicate that SW12 has the least infiltration capacity and hence, the highest susceptibility to erosion while SW13 has the least erosion susceptibility in terms of Fs only. The Rbm is a measure of the terrain's structural complexity and porosity and is thus negatively associated with the watershed's porosity. High Rbm suggests early hydrographic optimum with capacity for flood events during heavy rains leading to the deterioration of the soil surface (Howard 1990). Rbm's connection to an area's susceptibility to erosion is the same as that conveyed by Dd and Fs. Hence in this instance, the classification was the same as in the case of Dd and Fs. In the Imo River Basin, the Rbm ranges from 3.11 for SW11 to 6.44 for SW22. SW22 has the highest susceptibility to erosion in the study area according to Rbm results. The final linear morphometric parameter is the Drainage texture (Dt) that is strongly affected by the infiltration ability (Horton 1945). Low-infiltration regions will produce higher Dt and thus result in more erosion. The sub-watershed with the highest value of Dt in the Imo River Basin is SW5 (1.56 km⁻¹), indicating that it is the most susceptible to erosion due to a low infiltration rate. The sub-watershed with the lowest Dt is SW23 (0.46 km⁻¹), and this suggests that it is less vulnerable to erosion based on the study of the drainage texture.

The basin shape mainly regulates the rate at which the water is delivered to the secondary channel. These parameters include Elongation ratio (Re), Circulatory ratio (Rc), Form factor (Rf) and Compactness coefficient (Cc). In general, the Re ranges from 0.6 to 1.0 and is correlated with climate and geology (Schumm 1956, Strahler 1964). Re values close to 1.0 are transmitted to very low relief areas, while those values of 0.6–0.8 are correlated with high relief and steep slopes on the ground (Dar et al. 2013). The study found that SW12 with the highest Re (0.92) was the least susceptible to erosion and, in comparison, SW22 with the lowest Re (0.41) was the most susceptible to erosion. The Rc is influenced primarily by characteristics such as stream length and frequency, geological structures, climate, relief, and gradient of the basin (Miller 1953). Rc values range from 0 (in line) to 1 (in one circle). In a related study of the SWs of the Shivganga River Basin India by Kadam et al. (2019), a low Rc (i.e. 0.19) indicated rapid discharge from the sub-watershed and consequently, a higher susceptibility to erosion (Kadam et al. 2019). Higher values reflect more basin-shaped circularity and vice-versa. The sub-watershed SW5 in the Imo River Basin with the lowest amount of Rc (0.07) indicates the highest vulnerability to erosion with Rc as the only criterion.

In this way, the Rf has an inverse correlation to erosion; the lowest Rf values are most susceptible to erosion. The Rf values are within the range of 0.1–0.8 (Horton 1932). High-shaped basins have high peak flows of shorter length whereas elongated drainage basins with low-shaped influences have lower peak flows of longer duration (Singh, Singh 2011, Farhan, Anaba 2016, Asfaw, Workineh. 2019). The results of the Rf derived from the SWs of the Imo River Basin differ from the minimum (0.13) for SW22 up to the maximum (0.66) for SW12. Thus, the Rf analysis places SW22 as being most susceptible to erosion. A watershed's Cc is directly related to the potential for watershed penetration. Therefore, the rating adopted was identical to that adopted by Rf. The lowest Cc was recorded for SW20 with a value of 1.79, which means that it has the minimum absorption potential and thus the highest erosion sensitivity. The highest was noticed in SW5 with a value of 3.90 implying that it has the highest capacity for infiltration.

These parameters include Total Basin relief (H), Relief ratio (Rh), and Ruggedness number (Rn). The Total Basin relief (H) is the difference in altitude of the highest and lowest point of a basin valley floor. A strong correlation exists between the hydrological characteristics and a drainage basin's H (Schumm 1956). The H is an indicator of a drainage basin's total slope, as well as the ferocity of the erosion processes functioning on the basin slopes. In this case, the ranking system implemented was equivalent to the one implemented in the case of D, Fs and Rbm, since they all infuse the same erosive traits in any landscape. The lowest basin relief (i.e. 0.57) in a study conducted by Altaf et al. (2014) on the Rembiara Watershed, India was identified in one of the SWs, indicating its low steepness and hence low susceptibility to erosion. Similarly, the lowest value of H in the Imo River Basin was observed in SW2 (23 m) which demonstrated its low slope and therefore low risk to erosion. H's highest value was identified in SW24 (308 m), which is due to its high steepness and therefore higher sensitivity to erosion. The next parameter is the Rh that is specifically linked to the slope of the streams and the surface of the earth; which influences the hydrological processes and the watershed's erosion. The Rh basin has a direct correlation with the watershed's erodibility. In the Imo River Basin, the largest rate of Rh was in SW25 (0.020) and its lowest is for SW3 (0.002). Based on this analysis, SW25 is the most vulnerable to erosion compared with other SWs. The final parameter, Rn, is used to compute stream flood potential (Patton, Baker 1976). It symbolises the geometrical characteristics of the sub-basins of the Imo River Basin. Rn has a direct relationship with erodibility so that the erodibility also increases with increasing values of that parameter. The lowest value of Rn was observed for SW2 (0.024) which indicates that it has the lowest susceptibility to erosion. The highest value of Rn was obtained for SW24 (0.335) which means that it has the greatest susceptibility to erosion. In a study of the Dhidhessa River Basin in Ethiopia by Kabite and Gessesse (2018) the Rn of 3.1 indicated rough topography and susceptibility to soil erosion. Similarly, Kumar and Joshi (2016) derived an Rn value of 0.78 in another watershed and this was interpreted as having a steep slope and soil erosion susceptibility in the region.

After allocating ranks to the morphometric erosion potential parameters with respect to the relationship with erodibility, the correlation matrix was constructed using Pearson's product-moment correlation (r). Several parameters are very highly correlated with each other, and this high level of interdependence is observed between parameters in similar categories (i.e. linear, shape or relief). However, there is a strong negative correlation (r = −0.96) between two shape parameters, the Cc and Rc, and also a strong positive correlation (r = 0.99) between two relief parameters, total basin relief (H) and Rn. The Fw of the parameters were calculated using the WSA approach (Table 6). The correlation matrix showed that the Rbm has a negative correlation with most of the morphometric parameters except the Cc which have a positive correlation (0.53). The Re has a strong positive correlation with Rc and Rf and a strong negative correlation with Cc (−0.70). Also, Fs has the highest correlation with drainage texture (0.63).

In the WSA approach, the sum of correlation is the summation of the columns (Table 6) for Dd, Fs, Rbm, Dt, Re, Rc, Rf, Cc, H, Rh, Rn as: Sum of correlation for Rn= [0.38 + (−0.29) + (−0.07) + (−0.07) + (−0.35) + 0.05 + (−0.34) +0.00 + 0.99 + 0.80 + 1.00] = 2.10 \matrix{ {{\rm{Sum\, of\, correlation\, for\, Rn}} = [0.38 + \left( { - 0.29} \right) + } \cr {\left( { - 0.07} \right) + \left( { - 0.07} \right) + \left( { - 0.35} \right) + 0.05 + \left( { - 0.34} \right) + } \cr {0.00 + 0.99 + 0.80 + 1.00] = 2.10} \cr }

Further, the grand total obtained from the sum of correlations (14.71) was used for calculating the final weight of each morphometric parameter using Eq. (1). This was used in the final prioritisation ranking. By assigning those weights to each parameter, a model was formulated to assess the priority. For example, the final weight of Rn is the sum of correlation i.e. 2.10 divided by the grand total of 14.71 (Table 6). The resulting value (0.143), was used for the calculation of the Cpc, and further used in the SWs’ prioritisation and final ranking.

Similarly, the Cpc for SW1 was calculated as follows: Cpc for SW1= (0.120×23) + (0.111×18) +(−0.048×2) + (0.151×23) + (0.102×14) +(0.115×5) + (0.111×14) + (−0.092×21) +(0.130×24) + (0.157×22) + (0.143×24) = 19.78 \matrix{ {{\rm{Cpc\, for\, SW}}1 = \left( {0.120 \times 23} \right) + \left( {0.111 \times 18} \right)\, + } \cr {\left( { - 0.048 \times 2} \right) + \left( {0.151 \times 23} \right) + \left( {0.102 \times 14} \right)\, + } \cr {\left( {0.115 \times 5} \right) + \left( {0.111 \times 14} \right) + \left( { - 0.092 \times 21} \right)\, + } \cr {\left( {0.130 \times 24} \right) + \left( {0.157 \times 22} \right) + \left( {0.143 \times 24} \right) = 19.78} \cr }

In addition, the Cpc values of the 25 SWs in the Imo River Basin were calculated using the WSA method, and prioritization ranking allocation was incorporated as presented in Table 7. The first priority (1) was given to sub-watershed 5 with a Cpc value of 6.25, with the second priority (2) being sub-watershed 21, with a Cpc value of 7.51. The Imo River Basin's SWs were further classified into four (4) priority groups based on natural breaks (Jenks method) of the Cpc values – very high priority (6.25–8.52) covering 1035.64 km² or 41.51% of the basin area (2494.68 km²), high priority (8.52–10.50) covering 262.01 km² or 10.50% of the basin area, medium priority (10.50–17.56) covering 1105.82 km² or 44.33% of the basin area and low priority (17.56–21.77) covering 91.21 km² or 3.66% of the basin area. Of the 25 SWs in the Imo River Basin, ‘SW5, SW21, SW25, SW24, SW15, and SW23’ fall in very high priority; ‘SW14, SW19, SW22, and SW16’ fall in high priority; ‘SW17, SW4, SW18, SW11, SW8, SW13, SW3, SW20, SW6, SW7, SW9, SW12 and SW10’ fall in medium priority; whereas ‘SW1 and SW2’ fall in the low priority category. The highest priority suggests a greater degree of erosion exists in the sub-watershed and is a potential area for soil conservation measures to be implemented. Figure 5 shows the final erosion susceptibility map of the study area based on the Cpc values.

Fig. 5

These results show an improved accuracy in prioritising sub-basins than other methods such as the multi-criteria evaluation approach (Altaf et al. 2014, Alvarado et al. 2016, Vollmer et al. 2016), compound factor analysis (Patel et al. 2012, Mohammed et al. 2018) and compound parameter technique (Hlaing et al. 2008, Rawat et al. 2014).

In the past, governmental intervention efforts were initiated in some of these communities to arrest the spate of erosion. For example, gully sites at Umueshi, Amanato and Ntueke in the very high susceptibility area were within the zones marked for rehabilitation by the Imo State Nigeria Erosion and Watershed Management Project (IMO-NEWMAP) in 2014. One of the objectives of IMO-NEWMAP was to reduce the longer-term erosion vulnerability of the communities. In recent media reports on gully erosion in South-East Nigeria, the Isiokpo, Okpulu and Obidioukwu communities of Ideato North Local Government Area (LGA) of Imo State within the very high erosion susceptibility area have received a lot of attention. Residents of these three communities fled their homes on the 1st of September 2018, amid significant gully erosion after heavy rainfall (Channels Television 2018). Chiemelu et al. (2013) also described the Okigwe upland, which falls within the highly susceptible region, as a trouble spot for gully erosion.

The validation points are overlaid on the erosion susceptibility map in Figure 6. The ROC curve plot shown in Figure 7 has the false-positive rate on the X-axis and the true-positive rate on the Y-axis. The AUC value computed is 0.810 which implies a prediction accuracy of 81% of the morphometry-based model.

Fig. 6

Fig. 7