Error analysis in soil urea prediction based on RF spectroscopy

With the growing concern for environmental pollution and shrinking land resources available for agriculture, the need for sustainable agriculture is increasing. Soil sensing plays an important role in sustainable agriculture as it provides an insight into the various soil properties thus enabling the farmer to adjust the inputs accordingly. The aim of the study is to design a soil sensor and analyze the errors in the prediction of a soil nutrient. The manuscript describes a new method for soil nutrient sensing using RF spectroscopy. The technique can predict soil urea content and is based on multivariate analysis using the PLSR (Partial Least Square Regression) mathematical tool. Eight different combinations of five important soil nutrients (Urea, Potash, Phosphate, Salt, and Lime) at varying concentration were used to develop multivariate block. The Urea prediction algorithm takes into account the effect of various other soil nutrients present in the sample. The results obtained show that the percentage error in prediction of urea is within the tolerable limits of +/-5% of the actual value, when other soil nutrient concentrations are varied below and above their normal values. The method can be extended for sensing multiple nutrients simultaneously by modifying the algorithm.