Modelling the spatial distribution of soil organic carbon, case study: lyantonde town council.

Abstract

Soil organic carbon (SOC) is the carbon content that remains in the soil after the partial decomposition of any material produced by living organisms including decaying plant matter, soil organisms and microbes, and carbon compounds. It is a key representative of soil fertility and an essential parameter for controlling the dynamics of atmospheric carbon dioxide thereby reducing impacts of climate change. Studies that have been done to determine the spatial distribution of SOC are at a global and continental level and there is no data at a local scale. As a result, these SOC representations are unreliable i.e. they have a low spatial resolution, are not up to date, and were not verified with ground truth data. This study combined the potential of GIS and remote sensing techniques to model the spatial distribution of soil organic carbon using Lyantonde town council as a case study. Soil samples were collected following a stratified random sampling technique based on land use as strata and it was done while recording the sample locations coordinates using a handheld GPS and then split into training and validation sets using 80% for training and 20% for validation. A SOC model map and model equation were created using the Regression Kriging Technique in which the training set was used together with eight explanatory variables which included; Bulk Density, Brightness Index, Aspect, Clay Content, Slope, Land Surface Temperature, Topographic Position Index and Rainfall. According to the coefficient of determination (R2), Bulk Density, Brightness Index, Aspect, Clay content, Slope, Land Surface Temperature, Topographic Position Index and Rainfall caused explainable variabilities of 26.8%, 7.5%, 5.2%, 3.5%, 2.7%, 2.3%, 1.1% and 1% respectively with Bulk Density explaining the highest variance. This means that as the Bulk Density changes, it causes a 26.8% variability in the spatial distribution of SOC followed by the rest of the variables respectively. Collectively, the explanatory variables accounted for 99% of SOC variation in Lyantonde town council, with an R2 value of 0.996431. The developed SOC model accurately represents the SOC distribution, as validated by a low Root Mean Square Error value of 0.318. According to the research findings, this model is able to produce accurate SOC data, verified by ground-truth data, at a high spatial resolution and on demand.