Multi-satellite Data Fusion for agricultural applications in the Midwestern United States
There is tradeoff between spatial and temporal resolutions for any single satellite mission. Multi-sensor data fusion is a promising approach to achieve both high spatial and temporal resolutions. Here we will use Blue Waters to scale up the MODIS-Landsat fusion project to the 12 States of the U.S. Midwest through our newly developed SaTellite DAta IntegRation (STAIR) algorithm.
MODIS has a daily revisit frequency but relatively lower spatial resolution (500m for surface reflectance product MCD43A4). Landsat has a 30-m spatial resolution but lower revisit frequency (16 days). The expected product from this fusion project is a 30-m resolution, daily, cloud-free and spatio-temporally continuous surface reflectance dataset for the midwestern United States, which will enable novel agricultural applications in both academia research and industry, such as estimations of surface downward radiation, evapotranspiration, crop phenology stage, crop yield, and crop type classification, at unprecedentedly high resolutions in both space and time.
The computational demand of producing such high resolution and daily satellite images through fusion is huge and the Blue Waters facility offers us the best solution for large processing element demands, high-frequency I/O, and output post-processing and visualization.