Erosion Susceptibility Modelling Research Articles

Derivation of Coastal Erosion Susceptibility and Socio-Economic Vulnerability Models for Sustainable Coastal Management in Senegal

Coastal erosion has posed significant challenges to sustainability and socio-economic stability along Senegal’s coastline, leading to substantial infrastructure losses. Using GIS multi-criteria decision analysis (MCDA), two sub-indices were derived for Senegal’s coastal departments: the physical susceptibility (PSI) and the social-economic vulnerability (SVI) to coastal erosion. The integrated coastal erosion vulnerability (ICER) model was derived by their aggregation. A total of 26 criteria were used, 18 for PSI and 8 for SVI. The criteria weighting coefficients of the sub-indices were determined using the analytic hierarchy process (AHP). Validation of the model accuracy was performed using receiver operating characteristic (ROC) curves that were calculated based on a created coastal erosion cadaster and true positive (TP) sites and manually acquired true negative (TN) sites. The accuracy assessment confirmed the consistency of the physical susceptibility model (PSI) and proved that existing coastal erosion sites are within (5) very high susceptibility areas. Through the generated ICER, the coastal departments were divided into areas of (1) very low, (2) low, (3) medium, (4) high and (5) very high vulnerability to coastal erosion. Very high (5) and high (4) classes cover around 31% of the coastal departments, mostly encompassing a narrow coastal strip and low river valleys and mouths. The presented coastal susceptibility and vulnerability maps, with a spatial resolution of 30 m, identified problematic areas in Senegal’s coastal departments and can help decision-makers in the construction of effective coastal zone management and sustainable development.

Application of a Soil Erosion Susceptibility Model Using Unmanned Aerial Vehicle Photogrammetry in a Timber Harvesting Area, South Korea

Application of a Soil Erosion Susceptibility Model Using Unmanned Aerial Vehicle Photogrammetry in a Timber Harvesting Area, South Korea

Multi-criteria-based decision making approach for soil erosion susceptibility modelling of Sali River Basin, Bankura, India

Multi-criteria-based decision making approach for soil erosion susceptibility modelling of Sali River Basin, Bankura, India

A novel problem-solving method by multi-computational optimisation of artificial neural network for modelling and prediction of the flow erosion processes

This research aims to forecast, using various criteria, the flow of soil erosion that will occur at a particular geographical location. As for the training dataset, 80% of the dataset from the sample sites, four hybrid algorithms, namely heap-based optimizer (HBO), political optimizer (PO), teaching-learning based optimization (TLBO), and backtracking search algorithm (BSA) combined with artificial neural network (ANN) was used to create an erosion susceptibility model and establishes a unique and original approach. After it was confirmed to be successful, the algorithms were applied to create a susceptibility map for this area, demonstrating the integrity of the results. The AUC values were computed for every optimisation algorithm used in this study. The optimal estimated accuracy indices for populations of 450 were determined to be 0.9846 using the BSA-MLP training databases. The maximum AUC value for the HBO-MLP training databases with different swarm sizes was 0.9736. A swarm size of 350–300 is considered optimal for forecasting erosion susceptibility mapping in hybrid models. With the same swarm size constraints, the AUC value for training in the TLBO-MLP scenario was 0.996. After 150 swarm size conditions were used to train the PO-MLP model, the AUC values were 0.9845. According to these findings, the TLBO-MLP and PO-MLP algorithms worked best with populations of 50 and 150, respectively.

Field based inventory of river bank erosion susceptibility model (BESI) of Raidak-II river in the Himalayan foreland basin

ABSTRACT River bank erosion is a fluvio-hydrological hazard, and sometimes, it turns into disaster in the human-encroached river bank and flood plain. The principal objective of this study is to detect the river bank erosion potential zone using new Bank Erosion Susceptibility Index (BESI) model and spatiotemporal shifting of river Raidak-II (1980–2020). Sedimentary bank facies (SBF) analysis was conducted to identify the nature of cohesiveness of bank materials. The result showed that the maximum average lateral shifting (213.20 m) was recorded in the year 1990 (right bank), whereas the minimum average shifting (77.32 m) was measured in the year 2020 (right bank). The result also showed that the right bank of Raidak-II was mostly oscillated due to poorly sorted quaternary non-cohesive bank materials. The Bank Erosion Hazard Index (BEHI) model showed that 46.15% and 50% bank erosion sites are fallen in high-to-extreme bank erosion susceptibility zones in the years 2020 and 2022. In the case of BESI model, 69.23% and77.77% bank erosion sites are confined in high-to-extreme bank erosion susceptibility zones in the years 2020 and 2022. Therefore, BSEI model gives more precise results compared with BEHI due to three additional factors. The poorly sorted non-cohesive quaternary sediments stimulate high rate of bank erosion within Himalayan foreland basin.

Mapping the impacts of coastal erosion on the heritage assets of Ynys Enlli (Bardsey Island), North Wales, UK

Coastal erosion and sea level rise has become a significant threat to coastal heritage due to climate change. Management bodies are monitoring the risks posed by undertaking studies, ranging from Rapid Coastal Zone Assessments to much larger government and research council funded studies. These studies have incorporated varying techniques and datasets such as environmental sampling, LiDAR, GPS, and UAV surveys. However, these approaches rarely combine study of historic shoreline change with physical susceptibility modeling. This paper presents a combined methodology, utilizing LiDAR and survey data collected by the Climate, Heritage, Environments of Reefs, Islands and Headlands (CHERISH) project, along with historic mapping and aerial photography to produce a risk assessment for the heritage assets located on Ynys Enlli (Bardsey Island), North Wales. By constructing an Historic Erosion Model and an Erosion Susceptibility Model using ESRI’s ArcGIS 10.5.1. with the DSAS Plugin and combining them with the LiDAR-derived spatial extents of archaeological features on the island, this project has shown that using both methodologies alongside each other can create a more complete overview of coastal environment change for the creation of vulnerability indexes.

AHP GIS-Based Gully Erosion Susceptibility Modelling in Tropical Volcano Environment

Gully erosion is the most destructive form of erosion and contributes significantly to the accumulation of water sediment in an area. This study aims to map the gully erosion susceptibility in the Kodil watershed, Central Java, Indonesia. The base data used in this study are from Sentinel-2 with a resolution of 10 m and DEMNAS (National Digital Elevation Model of Indonesia) with a resolution of 8 m. The Analytical Hierarchy Process (AHP) was used to assess the weight of each parameter that influences the gully erosion susceptibility. Overall, there were ten parameters used in this study, namely a) elevation (m); b) slope gradient (0); c) slope aspect; d) slope curvature; e) topographic wetness index; f) distance from the river (m); g) distance from the road (m); h) land use; i) soil texture; and j) soil aggregate. An accuracy test used the AUC (Area under Curve). The results of spatial data analysis proved that 855 ha (6.34%) of land in the Kodil DAS has a high susceptibility to gully erosion. Testing with AUC shows that the AUC training value was 0.742, the AUC test was 0.737, and the accumulative AUC value was 0.73, so the model results fall into the good category. The novel aspect of this study is the use of texture and soil aggregate parameters that have never been studied before. The decision to employ soil parameters is inextricably linked to the characteristics of the research site, which has extremely thick soil layers with high clay content. The gully erosion susceptibility model helps plan the development of land resources in its sustainability aspect. Keywords: gully erosion, analytical hierarchy process, susceptibility modeling, tropical volcano environment. https://doi.org/10.55463/issn.1674-2974.50.2.20

Identifying sensitivity of factor cluster based gully erosion susceptibility models.

The present study has attempted to address the issue of sensitivity of different clusters of factors towards gully erosion in the Mayurakshi river basin. Firstly, the gully erosion susceptibility of the basin area has been mapped by integrating using 18 parameters divided into four factor-cluster, viz. erodibility, erosivity, resistance, and topographical cluster, with the help of four machine learning (ML) models such as random forest (RF), gradient boost (GBM), extreme gradient boost (XGB), and support vector machine (SVM). Results show that almost 20% and 25% of the upper catchment of the basin belongs to extreme and high gully erosion susceptibility. Among the applied algorithms, RF is appeared as the best performing model. The spatial association of factor cluster-based models with the final susceptibility model is found the highest for the erosivity cluster, followed by the erodibility cluster. From the sensitivity analysis, it becomes clear that geology and soil texture are dominant contributing factors to gully erosion susceptibility. The geological formation of unclassified granite gneiss and geomorphological formation of denudational origin pediment-pediplain complex is dominant over the entire upper catchment of the basin, and therefore, can be considered regional factors of importance. Since the study has figured out the different grades of susceptible areas with dominant factors and factor cluster, it would be useful for devising planning for gully erosion check measures. From economic particularly food security purpose, it is very essential since it is concerned with precious soil loss and negative effects on agriculture.

Assessment of the gully erosion susceptibility using three hybrid models in one small watershed on the Loess Plateau

Gully erosion is globally considered a severe environmental problem that causes great damage to arable lands, roads, and forests; therefore, it is crucial to evaluate the gully erosion susceptibility. The objective of this study was to establish three hybrid models based on VlseKriterijumska optimizacija I Kompromisno Resenje (VIKOR), frequency ratio (FR), random forest (RF), gradient boosting decision tree (GBDT), and randomized tree (ET) data to assess the susceptibility of gully erosion in a small watershed of the Loess Plateau of China. We extracted 540 gullies with 24, 919 gully pixels and 11 conditioning factors were extracted and utilized to establish a gully inventory database. Subsequently, FR was applied to determine the relationships between the conditioning factors and gully pixels, and machine learning methods were used to quantify the relative importance of these conditioning factors. Three hybrid gully erosion susceptibility models, that is, VIKOR-FR-RF, VIKOR-FR-GBDT, and VIKOR-FR-ET, were established for gully erosion susceptibility mapping (GESM). The receiver operating characteristic curve (ROC) and area under the curve (AUC) were utilized to evaluate the performance of these models. The results showed that elevation, distance to road, and the normalized difference vegetation index significantly contributed to the gully occurrences. VIKOR-FR-ET had the highest performance, with an AUC of 0.83, followed by VIKOR-FR-RF (AUC = 0.81) and VIKOR-FR-GBDT (AUC = 0.70). Therefore, we concluded that VIKOR-FR-ET was the most efficient approach for predicting gully erosion susceptibility in a semi-arid region, such as the Loess Plateau. Our results will help design soil and water conservation measures for gully erosion control at small watershed scales.

Analysis of gully erosion susceptibility and spatial modelling using a GIS-based approach

Gully erosion, a dominant limiting factor in promoting ecosystem services and functions, is caused by a wide range of geo-environmental factors. However, limited information is available on the gully erosion mechanisms at a regional scale. Herein, susceptibility of a particular type of large-scale gullies across southern China were analyzed based on the 2005 National Survey on Soil and Water Loss. Specifically, frequency ratio model was used to determine the thresholds of 18 geo-environmental factors, including topography (elevation, slope gradient, slope aspect, slope length, hypsometric integral), distance from primary and tertiary rivers, climate (average annual temperature, precipitation, and rainfall erosivity), lithology, soil texture, soil moisture, vegetation index (NDVI, FVC, and EVI), as well as human activity (LULC and population density), followed by evaluating their relative importance by Boruta algorithm, and modelling the gully susceptibility modelling by random forest. The gully susceptibility was showed to be determined by comprehensive effect of all the 18 factors, with the top three important factors including average annual precipitation (9.97%), rainfall erosivity (9.60%), and distance from primary rivers (9.43%), whose thresholds were 1701 ∼ 1872 mm, 5967 ∼ 15676 MJ·mm·(ha·h·a)-1, and 20000 ∼ 50000 m, respectively. Moreover, the random forest model with its AUC value ranging from 0.96 to 0.99 performed excellently in the spatial modelling of gully susceptibility. Collectively, the very low, low, moderate, and high susceptibility to these gullies occupied 65.30%, 21.36, 10.61% and 2.73% of southern China, in which the moderate and high susceptibility level were mainly clustered in and around Guangdong Province.

Wind erosion susceptibility modelling along the Eastern Cape Wild Coast, South Africa

Wind erosion is a major driver of land degradation, impacting more than a third of all land areas in recent centuries. It has been estimated that approximately 19% of lands in Europe are moderately to highly susceptible to wind erosion, whilst up to 25% of East and Southern Africa are at moderate to elevated risk of soil erosion. Along the eastern seaboard of South Africa, accelerated wind erosion has led to severe, localised land degradation in parts of the Wild Coast.In data sparse regions where conventional wind erosion modelling is difficult, mitigation of the devastating impacts associated with accelerated wind erosion could be facilitated by the identification of highly susceptible areas. Therefore, the study objectives were to map the regional wind erosion susceptibility conditions and identify vulnerable areas. In this study, general concepts of the Wind Erosion Equation were adopted. Two regional susceptibility methods were implemented and compared. Model 1 employed a geostatistical approach, based on erosion factor class frequency ratio data and Analytical Hierarchy Process importance weights. Model 2 utilised the data driven Weights of Evidence modelling technique. Model 1 has classified large areas of the study area as having low susceptibility (46%) whilst Model 2 classed more than 90% of the areas as very low susceptible zones. Both models show that less than 4% of the study region has a high to very high susceptibility to wind erosion. In general, areas associated with a higher wind erosion susceptibility are poorly vegetated, wind exposed coastal zones that are characterized by unconsolidated, erodible sandy soils. Model 1 and Model 2 are associated with area under the receiver operating characteristic curve values of 0.987 and 0.946 respectively, displaying satisfactory average performances.The employed modelling approaches provide an effective means to assess regional wind erosion susceptibility in coastal environments worldwide.

Assessing gully erosion susceptibility and its conditioning factors in southeastern Brazil using machine learning algorithms and bivariate statistical methods: A regional approach

Despite being a very common phenomenon worldwide, in Brazil, the factors and mechanisms that control gully erosion on a regional scale are still little known, which leads to the neglect of this environmental hazard by territorial and environmental management policies. In order to reducing this gap, we explored the potential of four common supervised machine learning algorithms, named random forest (RF), logistic regression (LR), naïve Bayes (NB) and artificial neural network (ANN) to produce gully erosion susceptibility models for two gullied watersheds, located in the state of Minas Gerais, southeastern Brazil. The modeling was based on the construction of a solid gully inventory and a database consisting of fifteen geo-environmental factors (GEF), whose influence was determined from the information gain ratio (IGR) and two bivariate statistical methods, named frequency ratio (FR) and modified information value (MIV). The predictive performance of the models was evaluated by the area under the receiver operating characteristic curve (AUC), overall accuracy (ACC) and sufficiency analysis. The results revealed that random forest achieved the highest overall performance in correct prediction of gullies and produced the most realistic gully susceptibility maps. The IGR data indicated that all GEF considered in the analysis contributed to the predictive model, although lithology, elevation and rainfall are the most influential variables. From an integrated analysis between the gully inventory, field observations, FR and MIV values, we found that gullies seem to be triggered by high annual average rainfall, but only develop where a set of specific geo-environmental conditions occur simultaneously. Finally, despite the limited land use data available, anthropogenic activities do not seem to affect the regional distribution pattern of gullies, although we have not excluded their local influence in triggering some erosive features.

Risk and Resources: An Evaluation of the Ability of National Soil Datasets to Predict Post-Depositional Processes in Archaeological Sites and Heritage at Risk

Previous studies have demonstrated the vast range of physical, chemical and biological processes that influence the preservation of archaeological sites, yet characterisation at the site-level remains largely unexplored. National datasets on soil type, land use and erosion modelling have the potential to predict localised impacts but remain an untapped resource in the evaluation of heritage at risk. Using early medieval Scotland as a case study, this paper explores in detail some of the primary factors which have impacted the archaeological record and the degree to which site-based evidence contained in excavation reports compares with national datasets (Land Cover Map 2015, Soil Information for Scottish Soils and Soils of Scotland Topsoil pH) and coastal erosion models (Dynamic Coast National Coastal Change Assessment and Coastal Erosion Susceptibility Model). This provides valuable information on the preservation of Scotland’s early medieval settlement, as well as a methodology for using national datasets in the remote assessment of post-depositional factors across the broader archaeological landscape. Results indicate that agriculture, bioturbation and aggressive soil conditions are among the most significant factors impacting Scotland’s archaeological remains. While the national datasets examined have the potential to inform heritage management strategies on these processes, their use is limited by a number of theoretical and methodological issues. Moving forward, site-specific studies that characterise the preservation environment will be crucial in developing baseline assessments that will advance both local and global understandings of destructive factors and soil-mediated decay.

Robustness analysis of machine learning classifiers in predicting spatial gully erosion susceptibility with altered training samples

The present research intended to assess the robustness of three popular machine learning models, i.e. random forest (RF), boosted regression tree (BRT) and naïve bayes (NB) in spatial gully erosion susceptibility modelling in Jainti River basin, India. A gully inventory map of 208 gullies was prepared through field survey and Google earth imageries. Following the 70/30 ratio, three randomly sampled groups of altered training and validation gully sets G1, G2 and G3 were prepared for modelling gully erosion susceptibility. Using information gain ratio and multi-collinearity analysis, 14 gully conditioning factors (GCF) were selected. The discrimination ability and reliability of the models were measured through Kappa coefficient, efficiency, receiver operating characteristic curve, root-mean-square-error (RMSE) and mean-absolute-error (MAE). The stability of the machine learning models was estimated by comparing the accuracy statistics and the departure in areal outcomes among intra-model and inter-model. RF model was found as the most consistent. With the highest mean AUC (0.903), efficiency (91.17), Kappa coefficient (0.835) and lowest RMSE (0.192) and MAE (0.081), RF was found to be more consistent when the training and validation data sets were altered. The effectiveness of each input GCFs was determined using map removal sensitivity analysis technique. This study could be supportive in ascertaining model deployment for mapping gully erosion and managing the land resource.

Integration of artificial intelligence with meta classifiers for the gully erosion susceptibility assessment in Hinglo river basin, Eastern India

The main aim of this study is to evaluate the gully erosion susceptibility coupling the artificial intelligence and machine learning ensemble approaches. In the present study, the multilayer perceptron neural network (MLP) was used as the base classifier and the hybrid ensemble machine learning methods i.e. Bagging and Dagging were used as the functional classifiers. The Hinglo river basin, an important tributary of the Ajay River was selected as the study area, consists with the parts of Chhotonagpur plateau and Rarh lateritic region. The study area is facing the gully erosion problems which are interrupted the growth of the agriculture. The gully erosion susceptibility maps (GESMs), prepared by MLP, MLP-Bagging and MLP-Dagging were classified into four classes such as low, moderate, high and very high susceptibility classes with the help of natural break method (NBM) in GIS environment. The very high susceptibility class covered 19.41% (MLP), 13.52% (MLP-Bagging) and 15.30% (MLP-Dagging) areas of the basin. For the evaluation and comparison of the models, receiver operating characteristics (ROC), accuracy, mean absolute error (MAE) and root mean square error (RMSE) were applied. Overall, all the gully erosion susceptibility models were performed as excellent. Integration of hybrid ensemble models with MLP has increase the accuracy of the MLP models. Among these models MLP-Dagging has achieved the highest accuracy in compare to the other models. The importance of the selected factors in the present study was assessed by the Relief-F method. The results show that the soil type factor has the highest predictive performance. Sensitivity analysis also showed soil type as most important factor. The gully erosion susceptibility maps (GESMs) are considered as the efficient tool which could be used to take the necessary steps for mitigating and controlling the soil erosion problem and sustainable environmental management and development.

Gully erosion susceptibility modelling to support avoided degradation planning

ABSTRACT Restoration resources are usually limited and strategic information on and erosion susceptible areas are required to avoid further degradation. This study has potential in the Mzimvubu River Catchment, South Africa, where two large reservoirs are planned on the Tsitsa tributary. The Tsitsa River Catchment, however, consists of highly erodible soils with widespread gully erosion evident. It is important to prevent further gully erosion in the catchment due to the presence of duplex and dispersive soils. Therefore, this study modelled areas that are susceptible to gully development in the Tsitsa River Catchment, as well as estimated the sediment yield potential from the susceptible areas if gully development occurs. This was achieved by mapping gully-free areas in a GIS that have the same DEM-derived topographical variables, soil associations and land cover as gullied areas, followed by scenario analysis of the potential sediment yield. More than 30 000 ha (7%) of the catchment is intrinsically susceptible to further gully development, consisting of drainage paths with a large contributing area and erodible duplex soils. If not protected, these susceptible areas could contribute an additional 300 million m3 of sediment to the river network, reducing the volumes of both reservoirs by more than 50%.

Novel Ensemble Approaches of Machine Learning Techniques in Modeling the Gully Erosion Susceptibility

Gully erosion has become one of the major environmental issues, due to the severity of its impact in many parts of the world. Gully erosion directly and indirectly affects agriculture and infrastructural development. The Golestan Dam basin, where soil erosion and degradation are very severe problems, was selected as the study area. This research maps gully erosion susceptibility (GES) by integrating four models: maximum entropy (MaxEnt), artificial neural network (ANN), support vector machine (SVM), and general linear model (GLM). Of 1042 gully locations, 729 (70%) and 313 (30%) gully locations were used for modeling and validation purposes, respectively. Fourteen effective gully erosion conditioning factors (GECFs) were selected for spatial gully erosion modeling. Tolerance and variance inflation factors (VIFs) were used to examine the collinearity among the GECFs. The random forest (RF) model was used to assess factors’ effectiveness and significance in gully erosion modeling. An ensemble of techniques can provide more accurate results than can single, standalone models. Therefore, we compared two-, three-, and four-model ensembles (ANN-SVM, GLM-ANN, GLM-MaxEnt, GLM-SVM, MaxEnt-ANN, MaxEnt-SVM, ANN-SVM-GLM, GLM-MaxEnt-ANN, GLM-MaxEnt-SVM, MaxEnt-ANN-SVM and GLM-ANN-SVM-MaxEnt) for GES modeling. The susceptibility zones of the GESMs were classified as very-low, low, medium, high, and very-high using Jenks’ natural break classification method (NBM). Subsequently, the receiver operating characteristics (ROC) curve and the seed cell area index (SCAI) methods measured the reliability of the models. The success rate curve (SRC) and predication rate curve (PRC) and their area under the curve (AUC) values were obtained from the GES maps. The results show that the ANN model combined with two and three models are more accurate than the other combinations, but the ANN-SVM model had the highest accuracy. The rank of the others from best to worst accuracy is GLM, MaxEnt, SVM, GLM-ANN, GLM-MaxEnt, GLM-SVM, MaxEnt-ANN, MaxEnt-SVM, GLM-ANN-SVM-MaxEnt, GLM-MaxEnt-ANN, GLM-MaxEnt-SVM and MaxEnt-ANN-SVM. The resulting gully erosion susceptibility models (GESMs) are efficient and powerful and could be used to improve soil and water conservation and management.

A methodological comparison of head-cut based gully erosion susceptibility models: Combined use of statistical and artificial intelligence

A GIS-based hybrid approach for gully erosion susceptibility mapping (GESM) in the Biarjamand watershed in Iran is presented. A database comprised of 15 geo-environmental factors (GEFs) was compiled and used to predict the spatial distribution of 358 gully locations; 70% (251) of which were extracted for training and 30% (107) for validation. A Dempster-Shafer (DS) statistical model was employed to map susceptibility. Next, the results of four kernels (binary logistic, reg logistic, binary logitraw, and reg linear) of a boosted regression tree (BRT) model were combined to increase the efficiency and accuracy of the mapping. Area under receiver operating characteristics (AUROC), true skill statistic (TSS) and efficiency (E) metrics were used to rank the five validated models. The results show that elevation and distance to road play crucial roles in gullying. Integrating BRT and DS enhanced prediction accuracy. Among the four BRT kernels, binary logistic performed best (AUROC of 0.886, TSS of 0.854 and E equal to 0.880). The worst results were produced by the individual DS model (AUROC = 0.849, TSS = 0.774 and E = 0.834). The hybrid binary logistic-BRT and DS map categorized 14.50% of the study area as having very-low susceptibility, 16.99% low susceptibility, 22.77% moderate susceptibility, 24.12% high susceptibility, and 21.59% very-high susceptibility.

Head-cut gully erosion susceptibility modelling based on ensemble Random Forest with oblique decision trees in Fareghan watershed, Iran

Gully erosion is the most active hydro-geomorphological phenomenon in the continental areas due to the high erosion rates triggered by the gully system. Monitoring and modelling gully development and gully distribution will contribute to understand landforms evolution and risk assessment. The purpose of the current research is to model head-cut gully erosion susceptibility (HCGES) using support vector machine (SVM), random forest (RF) and novel ensemble model of random forest with four Oblique methods (Logistic Regression, Ridge Regression, Partial least squares (PLS) and Support vector machine (SVM)) (hereafter called ensemble ORF) data mining models in Fareghan watershed, Hormozghan province, Iran. For this purpose, 14 variables influencing the gully development, were prepared and 145 head-cut gully erosion locations were identified in the study area. The efficiency of SVM, RF, ensemble ORF were evaluated based on receiver operating characteristic (ROC), the results have shown that all these three models are highly accurate and robust in predicting the head-cut gully erosion susceptibility zones. The results of the models were evaluated based on the area under the receiver operatic characteristic curve (AUC) in the validation stage presented that the efficiency of these models are 0.91, 0.94, and 0.96, respectively. Altitude and distance from the road in all three models were more important than other variables. The findings of this research will contribute to develop gully control strategies and to prevent the gully initiation where gully erosion is more susceptible.

Gully Erosion Susceptibility Mapping Using Multivariate Adaptive Regression Splines—Replications and Sample Size Scenarios

Soil erosion is a serious problem affecting numerous countries, especially, gully erosion. In the current research, GIS techniques and MARS (Multivariate Adaptive Regression Splines) algorithm were considered to evaluate gully erosion susceptibility mapping among others. The study was conducted in a specific section of the Gorganroud Watershed in Golestan Province (Northern Iran), covering 2142.64 km2 which is intensely influenced by gully erosion. First, Google Earth images, field surveys, and national reports were used to provide a gully-hedcut evaluation map consisting of 307 gully-hedcut points. Eighteen gully erosion conditioning factors including significant geoenvironmental and morphometric variables were selected as predictors. To model sensitivity of gully erosion, Multivariate Adaptive Regression Splines (MARS) was used while the Area Under the Receiver Operating Characteristic (ROC) Curve (AUC), drawing ROC curves, efficiency percent, Yuden index, and kappa were used to evaluate model efficiency. We used two different scenarios of the combination of the number of replications, and sample size, including 90%/10% and 80%/20% with 10 replications, and 70%/30% with five, 10, and 15 replications for preparing gully erosion susceptibility mapping (GESM). Each one involves a various subset of both positive (presence), and negative (absence) cases. Absences were extracted as randomly distributed individual cells. Therefore, the predictive competency of the gully erosion susceptibility model and the robustness of the procedure were evaluated through these datasets. Results did not show considerable variation in the accuracy of the model, with altering the percentage of calibration to validation samples and number of model replications. Given the accuracy, the MARS algorithm performed excellently in predictive performance. The combination of 80%/20% using all statistical measures including SST (0.88), SPF (0.83), E (0.79), Kappa (0.58), Robustness (0.01), and AUC (0.84) had the highest performance compared to the other combinations. Consequently, it was found that the performance of MARS for modelling gully erosion susceptibility is quite consistent while changes in the testing and validation specimens are executed. The intense acceptable prediction capability of the MARS model verifies the reliability of the method employed for use of this model elsewhere and gully erosion studies since they are qualified to quickly generating precise and exact GESMs (gully erosion sensitivity maps) to make decisions and management edaphic and hydrologic features.