An innovative phenology-matching model to estimate crop growing stages

Abstract:

Large-scale remote monitoring of crop phenological development plays an essential role in modeling seasonal agroecosystem carbon, water, and energy exchanges, assessing biomass accumulation and net primary production, and scheduling farm management practices (e.g., irrigation, fertilizer and other chemical applications). Over the past decade, a variety of curve-fitting based phenological methods have been developed to estimate critical crop phenological transition dates. Despite the promising results, the crop growing stages are mostly characterized in terms of satellite time series curve properties (e.g., inflection points), the potential of which to extend to other physiological growing stages may be limited. Detecting the growing stages (e.g., planting stage) that do not maintain distinct curve properties may be challenging. To overcome the challenge, we will develop a novel phenology-matching model that can robustly retrieve a range of crop physiological growing stages. With its integrative landmark and reference design, the phenology model demonstrates enhanced capabilities in characterizing the phenological stages without distinct curve properties. The further scenario design of reference shapes and transition dates indicates that the model can effectively conduct phenology matching with varying levels of publicly accessible phenological information. In Illinois, the phenology-matching model with year- and region-adjusted phenological reference can identify the median transition dates of most phenological stages of corn and soybean with R squares higher than 0.9 and RMSEs less than 5 days. Together with crop progress report-enabled phenological reference calibration, the developed model holds large potential to improve the applicability of phenology matching models in revealing spatiotemporal patterns of crop phenology over extended geographical regions.