Evaluation of maximum entropy (Maxent) machine learning model to assess relationships between climate and corn suitability

Authors: FITZGIBBON, ABIGAIL - Esri; PISUT, DAN - Esri; Fleisher, David

Submitted to: Land
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/19/2022
Publication Date: 8/23/2022
Citation: Fitzgibbon, A., Pisut, D., Fleisher, D.H. 2022. Evaluation of maximum entropy (Maxent) machine learning model to assess relationships between climate and corn suitability. Land. 11. https://doi.org/10.3390/land11091382.
DOI: https://doi.org/10.3390/land11091382

Interpretive Summary: Crops are grown on land that is suitable for their cultivation, and this suitability is typically linked with the local climate and soil characteristics. Due to projected changes in water availability and weather, however, it is important to predict if and how cropland that is currently suitable for production may change in the future. Two different mathematical models were constructed to identify which parcels of land in the United States Midwestern region were currently most suitable for corn production, and what specific variables, such as soil texture or temperature, were associated with that suitability. Results showed that one particular model approach, called Maxent, was more accurate in terms of identifying units, or parcels, of land in which corn was observed to be currently grown. Using this approach, the most important bioclimatic variables included the highest temperature of the warmest month of the growing season, the lowest temperature for the coolest month of the growing season, and the seasonality of rainfall. This modeling approach was shown to be useful for identifying additional units of land area which could be suitable for corn both in the Midwest and other areas of the country. The approach will also be useful for developing geographic maps of cropland suitability under current and future climate conditions and identifying management methods that farmers may be able to use to adapt to climate change. The method can also be easily extended to other important crops for additional studies associated with U.S. food security.

Technical Abstract: Given the impact that climate change is projected to have on agriculture, it is essential to understand the mechanisms and conditions that drive agricultural land suitability. The Maximum Entropy model (Maxent) is a correlative machine learning model often used to model cropland suitability. In this study, we used Maxent to model land suitability for corn production in the contiguous United States under current bioclimatic conditions. We evaluated Maxent's predictive ability through three comparisons: (i) classification of suitable land units and comparison of results with another similar species distribution model (Random Forest Classification), (ii) with real-world corn location suitability and yield data, and (iii) existing literature on corn suitability thresholds associated with the Maxent formulation. We determined that Maxent was superior to Random Forest, especially in its modeling of areas of land suitable absence in which land was likely suitable for corn but was not currently associated with observed corn presence data. We also determined that Maxent's predictions correlated strongly with observed yield statistics and were consistent with existing literature regarding the range of bioclimatic variable values associated with suitable production conditions for corn. Maxent was also used to evaluate areas that were deemed unsuitable but were still observed to have significant corn presence. Maxent associated these areas with the need for more intensive crop management and climate adaptation. We concluded that Maxent was an effective method for modeling current cropland suitability and could be applied to broader issues of agriculture-climate relationships.