Crop Soil Mapping Using Machine Learning In Tiaret Region

Abstract

Soil is a main key for land use management, in agriculture soil management is an important factor that determines production, it is fundamental to know the soil quality in order to advance in any agricultural practice. Soil properties are affected by land use practices and climatic factors. In order to collect and determine the right soil data a soil mapping is unavoidable, however, there are many mapping techniques that aren't always accurate or representable of the instantaneous situation. Digital soil mapping is a suitable approach as a decision support, based on weighted factors to approach the real soil properties using different geostatistics, remote sensing, machine learning models and other digital tools to estimate the unavailable data. In this study we perform a Multi linear regression model in python based on 113 sampling points where we predict the (N P K) values based on the laboratory analysis data, we used 5 covariates in the model NDVI, NDMI, BI, Slope and Texture, the predicted values where near to the real NPK values with an R-square at approximately 0.2