Multisource Image Kalman Filtering for Rapid Phenological Monitoring and Forecasting

The availability of images from Earth observing satellites continues to grow as new sensors become operational, and existing archives are expanded and made more accessible. This wealth of data, along with computational advances that enable large archive processing, provides a valuable opportunity to investigate large-scale phenological dynamics and the climate drivers of phenological variability. Fusion methods capable of exploiting complementary aspects of spatial, temporal, and spectral resolutions from multiple sources are necessary to fully realize the potential of this asset. However, existing multisensor fusion approaches are designed for specific applications and/or sensor combinations, and most cannot incorporate knowledge of physical drivers of variability. Here, we explore the potential of the Kalman Filter to address these shortcomings, and demonstrate its applications to landscape-scale phenological monitoring. We demonstrate the fusion of Landsat, Sentinel, and MODIS data for the purposes of rapid phenological state estimation and forecasting, and extend the methodology to incorporate climate drivers of phenological variation. We conclude by discussing some outstanding challenges related to process and error model specification, along with potential solutions.