Paper ID: 2403.10730

Counterfactual Analysis of Neural Networks Used to Create Fertilizer Management Zones

Giorgio Morales, John Sheppard

In Precision Agriculture, the utilization of management zones (MZs) that take into account within-field variability facilitates effective fertilizer management. This approach enables the optimization of nitrogen (N) rates to maximize crop yield production and enhance agronomic use efficiency. However, existing works often neglect the consideration of responsivity to fertilizer as a factor influencing MZ determination. In response to this gap, we present a MZ clustering method based on fertilizer responsivity. We build upon the statement that the responsivity of a given site to the fertilizer rate is described by the shape of its corresponding N fertilizer-yield response (N-response) curve. Thus, we generate N-response curves for all sites within the field using a convolutional neural network (CNN). The shape of the approximated N-response curves is then characterized using functional principal component analysis. Subsequently, a counterfactual explanation (CFE) method is applied to discern the impact of various variables on MZ membership. The genetic algorithm-based CFE solves a multi-objective optimization problem and aims to identify the minimum combination of features needed to alter a site's cluster assignment. Results from two yield prediction datasets indicate that the features with the greatest influence on MZ membership are associated with terrain characteristics that either facilitate or impede fertilizer runoff, such as terrain slope or topographic aspect.

Submitted: Mar 15, 2024