Mapping Crop Phenological Metrics at Field Scales by Fusing Time Series of VIIRS and HLS over the United States Corn Belt

Crop phenology is one of the most essential parameters for predicting crop productivity. It has been commonly mapped from moderate or coarse resolution (>500m) satellite data at a national or global region, where each pixel frequently represents the mixture of difference crop types or between crop and natural vegetation. Recently, crop phenological progress at filed scales (<30m) has been investigated using various different methods, but the methods were only applied in a local region, such as a single county or state level. We in this study for the first time mapped crop phenological progress at a field scale (30m) over entire Corn Belt in the Midwestern United States using a new approach that fuse time series of Visible Infrared Imaging Radiometer Suite (VIIRS) and harmonized Landsat and Sentinel-2 (HLS) data. VIIRS observes crop growth daily with a moderate spatial resolution (500m) while HLS provides observations approximately every 3 days at a 30m field. To generate high spatial (30m) and temporal (daily) crop observations without cloud contamination, we used the shape of temporal VIIRS observations to fuse HLS time series. The fused HLS-VIIRS time series was applied to map crop phenological metrics of greenup onset, mid-greenup phase, maturity onset, senescence onset, mid-senescence phase and dormancy onset in 2018. These metrics were evaluated by comparing with corn and soybean crop progress (CP) reports of ground observations from the National Agricultural Statistics Service (NASS) of the United States Department of Agriculture (USDA). The result indicated that the HLS-VIIRS crop phenological metrics are capable of tracing the crop progress very well. The HLS-VIIRS crop phenological metrics differ with NASS CP stages: less than 3 days for corn and soybean emergence, 10 days for corn silking, 6 days for corn dough, 9 days for corn mature, 9 days for soybean blooming, 7 days for soybean setting pods, and 1 day for soybean dropping leaves. The relatively large difference is associated with the fact that the HLS-VIIRS crop phenological metrics and NASS CP stages are defined differently. Based on their significant correlation (p<0.01), the HLS-VIIRS crop phenological metrics can predict NASS CP in a much higher accuracy.