An assessment and prediction of soil erosion risk using modified fournier index and machine learning algorithm: an external agricultural project risk

Abstract

Soil erosion, defined as a naturally occurring process that adversely affect all landform leads to increased pollution and sedimentation in rivers and streams which causes decline in fish and other forms of aquatic life. Suitable land use guided by scientific research findings can help reduce these impacts. The current study therefore aimed at characterisation and prediction of soil erosion by water using Modified Fournier Index methodology. Prior to final data analysis, data quality checks were deployed where outliers were detected, removed and replace by expectation maximum algorithm aided by SPSS. A machine learning algorithm, Neural Network was applied to forecast probable annual values of the Modified Fournier Index (Cp). Major findings exhibited a significant decreasing trend implying a high likelihood of drought events in the area. This phenomenon provides an insight for possible shift in the type of soil erosion risk to prevail in the near future, where soil particles will be prone to wind erosion. The Neural Network forecasted Fournier index values were seen diminishing annually. From these results it is therefore recommended that more studies be undertaken on drought risk analysis since Fournier index values are diminishing giving way to drought events. This information will provide details necessary for informed decision in the protection and sustainability of the agricultural activities in the study area.