Corn, Phenology, Landsat, Random Forest, Support Vector Machine


Existing methods to report phenology are expensive, labor-intensive, time-consuming, and often not very accurate, especially at some specific crop growth stages. The objec­tive of this study was to develop large-scale phenology models via utilization of satellite imagery data and machine learning techniques for the southwest (SW) agricultural crop reporting district of Kansas. Different satellite images collected from Landsat were utilized as the main input to obtain different vegetation indices (normalized difference vegetation index, NDVI; enhanced vegetation index, EVI; green chlorophyll vegeta­tion index, GCVI; normalized difference water index, NDWI; and global vegetation moisture index, CVMI). Vapor Pressure Deficit (VPD), temperature, precipitation, and growing degree units (GDU) were evaluated for improving phenology prediction models. A large set of ground truth data with information about day of the year, crop phenology, and field location was provided by Crop Quest Inc. (Dodge City, KS) from 2014–2018 and utilized to train two different statistical models (Random Forest and Support Vector Machine) to catalog corn fields, and build a phenology evolution model for this crop.

Creative Commons License

Creative Commons Attribution 4.0 License
This work is licensed under a Creative Commons Attribution 4.0 License.