Predicting State-Wide Cotton Yield using Geospatial Data

The focus of this study was to use Geospatial Linear Regression to predict cotton yield.

Forecasting cotton yield has immense interest in the agriculture world considering how valued the crop is world wide. However, current methods of predicting cotton yield remain significantly unreliable since they rely on human assumption. Additionally, these forecasts are significantly impacted by climate change. This paper aimed to use spatial models to better predict cotton yield. Prior papers on predicting cotton production have yielded fair results utilizing non-spatial models [1]. Therefore, the focus of this study was to use Geospatial Linear Regression to predict cotton yield. Through this study, it was found that using climate, temporal, and geographic data cohesively produces the most accurate predictions. Additionally, the Random Forest model, with an MSE of 2.67, performed better than the Geospatial model, with an MSE of 8.22. This concludes that non-spatial models are better for predicting state-wide cotton yield when compared to spatial models. This research provides valuable results to the agriculture setting as well as different aspects to further this research.

Explore More!

Research Paper