Modelling Approaches for Determining Relationships between Crop Yield and Landscape Features.

Sound agronomic recommendations are crucial for today’s agronomists as they strive for improved yields, profits, and sustainability. Determining the spatial relationships between yield and landscape variation including soil properties and terrain attributes may improve management decisions, particularly with regards to proper N application for minimizing both costs to farmers and environmental impacts. Here we investigate relationships between landscape features, soil texture, soil properties and corn yield as part of a preliminary study to model corn yield with variations in landscape attributes, soil properties, and weather. We used yield monitor data collected from 2010 – 2015 at a 12 ha field at the Davis Purdue Agricultural Center in Randolph County, IN, USA. We obtained 21 digital elevation-based models of terrain attributes that describe morphometric and hydrologic characteristics of the field. We also sampled the field to obtain soil texture and soil property estimates. For each year we used the random forest method to select the variables that were most important for predicting corn yield across the field and used these in spatial error models. The variables with the most significant relationship with corn yield each year were soil texture and topographic wetness index. These results demonstrate that models for predicting corn yield in Indiana need to include landscape features for increased model performance. This analysis met one objective of a larger investigation that will incorporate soil properties, terrain attributes, and weather patterns into models of corn yield across Indiana landscapes.