Skip to main content
ARS Home » Midwest Area » Columbia, Missouri » Cropping Systems and Water Quality Research » Research » Publications at this Location » Publication #378533

Research Project: Sustainable Intensification of Cropping Systems on Spatially Variable Landscapes and Soils

Location: Cropping Systems and Water Quality Research

Title: Developing an innovative in-season and site-specific nitrogen recommendation strategy with machine learning for US Midwest corn production

item MIAO, YUXIN - University Of Minnesota
item LI, DAN - University Of Minnesota
item FERNANDEZ, FABIAN - University Of Minnesota
item Kitchen, Newell
item Ransom, Curtis
item CAMBERATO, JAMES - Indiana University-Purdue University
item CARTER, PAUL - Retired Non ARS Employee
item FERGUSON, RICHARD - University Of Nebraska
item FRAZEN, DAVID - North Dakota State University
item LABOSKI, CARRIE - University Of Wisconsin
item EMERSON, NAFZIGER - University Of Illinois
item SAWYER, JOHN - Iowa State University
item SHANAHAN, JOHN - Soil Health Institute

Submitted to: ASA-CSSA-SSSA Annual Meeting Abstracts
Publication Type: Abstract Only
Publication Acceptance Date: 10/9/2020
Publication Date: N/A
Citation: N/A

Interpretive Summary:

Technical Abstract: The objective of this research was to develop a machine learning-based in-season N recommendation strategy by incorporating soil information, early season weather conditions, and preplant and sidedress N application information. Plot research data from 48 site-year N rate trials conducted over three years (2014-2016) eight US Midwest states were used. At each site-year, there were a total of 16 N rate treatments with different preplant and sidedress combinations. The RapidSCAN CS-45 sensor was used to collect canopy reflectance at V6-V10 stages before a sidedress N application. Two machine learning methods (Random Forest and Support Vector Machine) were used to predict corn yield on a randomly selected portion of the data (75%) and validated on the remaining unused observations (25%). The preliminary results indicated that NDVI-based in-season estimate of yield (INSEY) could only explain 5% of corn yield variability at best, with root mean square error (RMSE) being 2452 kg/ha for validation. Using NDVI-based INSEY and sidedress N rate information could explain 9% of corn yield variability at best, with RMSE being 2396 kg/ha for validation. The validation results of Random Forest and Support Vector Machine models were similar (R2=0.78-0.79), with the Support Vector Machine model having slightly lower RMSE (1164 kg/ha) than the Random Forest model (1192 kg/ha). The developed and successfully validated models can then be used before sidedressing (around the V9 stage) to predict corn yield responses to a series of sidedress N rates at small increments (e.g. 5 kg/ha). This predicted yield response can then be used to determine the in-season site-specific agronomic optimal sidedress N rate or EONR. This research demonstrated the use of machine learning methods for integrating readily available data with remote sensing information to guide in-season site-specific N management.