Early Season Planted Acreage Estimates Using Machine Learning

• Activity Number: 268 - Extreme Machine Learning Methods and Applications: Domestic and International
• Type: Topic Contributed
• Date/Time: Tuesday, August 4, 2020 : 1:00 PM to 2:50 PM
• Sponsor: Section on Statistical Consulting
• Author(s): Jake Abernethy* and Claire Boryan and Kevin Hunt and Luca Sartore
• Companies: NASS and USDA and USDA and National Institute of Statistical Sciences
• Keywords: machine learning; big data; agriculture; supervised learning

Abstract

The United States Department of Agriculture National Agricultural Statistics Service (NASS) provides timely and accurate statistics in service to U.S. agriculture. An example is planted acreage estimates for key crops across the major growing regions. Toward this goal, NASS conducts the Prospective Planting Survey (PPS) in March and the June Area Survey in June. While these surveys provide unbiased estimates at collection time, NASS needs to anticipate the planted acreage before March and adjust these estimates based on influencers such as inclement weather. New modeling techniques are being developed to predict planted acreage using large, geospatial datasets. The objective of this study is to employ machine learning models that use data from different sources to obtain the estimated acreages planted to corn and soybeans for the state of Illinois before the PPS is conducted. The models will be updated during the early growing season to incorporate precipitation, temperature, survey, and other data. The accuracies of the model estimates for 2014-2019 are measured based on relative error with respect to the NASS final official planted acreage estimates.