Statistical correlation analysis between multispectral survey and agricultural soil physics variables

Abstract

This study explores the use of multispectral surveys as a tool to acquire information on productive soils through flights with unmanned aerial vehicles (UAVs) equipped with cameras that capture images in specific ranges of visible and invisible light. Its objective is to observe the correlations between the Normalized Difference Vegetation Index (NDVI) and variables obtained in soil analysis, such as nitrogen, phosphorus, potassium and iron. The analysis seeks to identify the behavior of these physical variables by means of statistical models and to establish a rapid inspection methodology to characterize the soil and optimize agricultural management. Sampling was carried out on 40 properties in the municipality of Corozal, Sucre, Colombia, using two methods: autonomous overflight with UAVs to capture NDVI and manual collection of soil samples for nutrient analysis. The results include linear and polynomial correlation models between NDVI and physical variables, thus facilitating the characterization of the plots in the region. The main conclusion indicates that the third order polynomial model is the one that best fits the real behavior of the properties evaluated in the study area, with coefficients of determination higher than 0.4. This allows ruling out the nullity of the model and confirming its usefulness in the characterization of the soil conditions of the sampled properties.