A nonlinear model to estimate nitrogen level in agricultural soil using Gaussian kernels

Nitrogen fertilizers are commonly used to improve agricultural productivity. However, its excessive use may cause or lead to environmental problems. Therefore, technologies capable of monitoring and measure levels of nitrogen in agricultural soil in-situ and in real time are required in order to make efficient the use of fertilizers. Nitrogen levels are usually measured by direct and indirect methods. Direct methods can be conducted in-situ or in laboratory, but they are really expensive and/or little resistant to soil conditions. Otherwise, indirect methods can estimate nitrogen levels in-situ and in real time, based on the measure of other parameters, and at the expense of accuracy. This paper proposes an indirect method to estimate the nitrogen level in agricultural soil through the measurement of the levels of electrical conductivity, temperature and humidity. The proposed model uses a nonlinear estimator based on Gaussian kernels. The results after training the model with real data showed values very close to the actual measured values.