Investigation of Long Short-Term Memory Artificial Neural Networks as Estimators of Nitrate Concentrations in Soil from Measured Electrical Impedance Spectra

Abstract

Monitoring the nitrate concentration in soil is crucial to guide the use of nitrate-based fertilizers. This study presents an investigation of long short-term memory (LSTM) recurrent artificial neural networks with regard to their suitability to extract nitrate concentrations from electrical impedance spectra of soil samples. Based on measured impedance spectra and physical properties of various synthetic sandy soils, the importance of different features for model training was investigated first. Both Random Forests and LSTM were tested as feature selection methods. Then numerous LSTM networks were trained to predict the nitrate concentration in sandy soils. The resulting regression models showed coefficients of determination between true and predicted nitrate concentrations as high as 0.95.