B34C-07 – Spatiotemporal Analysis of Corn Phenoregions in the Continental United States

Authors

Venkata Shashank Konduri
Oak Ridge National Laboratory
Northeastern University
Jitendra Kumar
Oak Ridge National Laboratory
Forrest M. Hoffman (forrest at climatemodeling dot org)
Oak Ridge National Laboratory
Auroop R. Ganguly
Northeastern University
William W. Hargrove
USDA Forest Service

Session

Vegetation Phenology as Forcing and Response Across Diverse Biomes: Detection, Attribution, Prediction, and Implications III
Wednesday, December 13, 2017 17:30–17:45
New Orleans Ernest N. Morial Convention Center – 383–385

Abstract

The delineation of regions exhibiting similar crop performance has potential benefits for agricultural planning and management, policymaking and natural resource conservation. Studies of natural ecosystems have used multivariate clustering algorithms based on environmental characteristics to identify ecoregions for species range prediction and habitat conservation. However, few studies have used clustering to delineate regions based on crop phenology. The aim of this study was to perform a spatiotemporal analysis of phenologically self-similar clusters, or phenoregions, for the major corn growing areas in the Continental United States (CONUS) for the period 2008–2016. Annual trajectories of remotely sensed normalized difference vegetation index (NDVI), a useful proxy for land surface phenology, derived from Moderate Resolution Spectroradiometer (MODIS) instruments at 8-day intervals and 250 m resolution was used as the phenological metric. Because of the large data volumes involved, the phenoregion delineation was performed using a highly scalable, unsupervised clustering technique with the help of high performance computing. These phenoregions capture the spatial variability in the timing of important crop phenological stages (like emergence and maturity dates) and thus could be used to develop more accurate parameterizations for crop models applied at regional to global scales. Moreover, historical crop performance from phenoregions, in combination with climate and soils data, could be used to improve production forecasts. The temporal variability in NDVI at each location could also be used to develop an early warning system to identify locations where the crop deviates from its expected phenological behavior. Such deviations may indicate a need for irrigation or fertilization or suggest where pest outbreaks or other disturbances have occurred.


Forrest M. Hoffman (forrest at climatemodeling dot org)