Soil organic carbon mapping using remote sensing techniques and multivariate regression model

Abstract

Soil organic carbon (SOC) is an importance aspect of soil quality and plays an imperative role in soil productivity in the agriculture ecosystems. The present study was applied to estimate the SOC stock using space-borne satellite data (Landsat4-5 Thematic Mapper (TM)) and ground verification in the Medinipur Block, Paschim Mednipur District, and West Bengal in India. In total, 50 soil samples were collected randomly from the region according to field surveys using a handheld Global Positioning System (GPS) unit to estimate the surface SOC concentrations in the laboratory. Bare Soil Index (BSI) and Normalized Difference Vegetation Index (NDVI) were explored from TM data. The satellite data derived indices were used to estimate spatial distribution of SOC using multivariate regression model. The regression analysis was performed to determine the relationship between SOC and spectralindices (NDVI and BSI) and compared observed SOC (field measure)to predict SOC (estimated from satellite images). Goodness fit test was performed to determine the significance of the relationship between observed and predicted SOC at p ≤ 0.05 level. The results of regression analysis between observed SOC and NDVI values showed significant relationship (R2 = 0.54; P<0.0075). A significant statistical relationship (r=-0.72) was also observed between SOC and BSI. Finally, our model showed nearly 71% of the variance of SOC distribution could be explained by SOC and NDVI values. The information from this study has advanced our understanding of the on-going ecological development that affects SOC dissemination and might be valuable for effective soil Downloaded by [University of Connecticut] at 08:32 19 September 2017 management.