| Prediction of soil parameters using laboratory spectral data of Kurnool area, India |
| Paper ID : 1031-ICRSSSA-FULL |
| Authors |
|
Ali Refaat Ali Moursy *1, Rabi Narayan Sahoo2, Nayan Ahmed3, Shalini Gakhar2, Bhabani P Mondal2, Abdel-Rahman A Mustafa4 1Soil and water department, Faculty of agriculture, sohag University, Egypt 2Division of Agricultural Physics, Indian Agricultural Research Institute, New Delhi, India, 110012. 3Division of Soil science and Agricultural Chemistry, Indian Agricultural Research Institute, New Delhi, India, 110012. 4Soil and water Department, Faculty of Agriculture, Sohag University, Sohag, Egypt, 82524. |
| Abstract |
| Diffuse Reflectance Spectroscopy (DRS) as a contemporary technique has evolved as a highly efficient method for predicting a multitude of soil properties. On the contrary, is less expensive and time-saving to collect and analyze a vast number of samples. Current study exploits the treasure of spectral and spatial information extracted from hyperspectral data integrated with multivariate modeling to characterize and quantify soil parameters. As part of the data acquisition campaign, a total of 96 soil surface samples were collected from the Kurnool and Andhra Pradesh regions of India. Overall, 14 soil physical and chemical parameters, namely, clay, silt, sand, pH, electrical conductivity (EC), soil organic carbon (SOC), available nitrogen (Av.N), available phosphorous (Av.P), available potassium (Av.K), available sulfur (Av.S), available iron (Av.Fe), available manganese (Av.Mn), available copper (Av.Cu), available zinc (Av.Zn) were analyzed using conventional methods. Further, soil spectral signatures were captured using ASD Field spectroradiometer in a controlled environment for the Vis-NIR region of the electromagnetic spectrum in the range of 350nm to 2500nm. Later, pre-treated spectral data were fed as an input to the Partial Least Square Regression (PLSR) model for the prediction of considered soil parameters. Subsequently, an accuracy assessment is done, wherein, soil parameters i.e. clay, sand, pH, EC, SOC, Av.N, Av.Fe, and Av.Zn were predicted with a R2 ≥ 0.5 and RPD ≥ 1.4. Besides, poor performance has been recorded for estimation of silt, Av.P, Av.K, Av.S, Av.Mn, Av.Cu. Overall, the approach showed promising performance for predicting soil parameters. |
| Keywords |
| Spectroradiometer, Vis-NIR, soil prediction, PLSR, hyperspectral |
| Status: Accepted (Oral Presentation) |
