Development of artificial neural network model for soil nitrate prediction

Abstract

Nitrate is the main form of nitrogen absorbed by plants. Leaching of nitrate can contaminate groundwater. The measurement of soil nitrate with conventional methods is less practical, takes a long time, and requires a lot of costs. Measurement of variables that affect the presence of soil nitrate can be an alternative solution. The application of prediction models is proven to save time and cost. Complexity problems can use the ANN model. This study aims to developed prediction models for soil nitrate use the ANN model. The measurable parameters such as solution volume, soil moisture, and soil electrical conductivity were used as input parameters for the model prediction development. The samples use oven-dry soil that was added nitrate solution with several variations. The measurement of parameters was carried out in three replications. The training and validation of the ANN model resulted in RMSE values of 1,0840029 and 1,000646 then R2 values were 0.973 and 0.970. The ANN model can be an alernative to predict soil nitrate at different monitoring volumes.