Assessment of the ability of Sentinel-1 SAR data and Sentinel-2 NDVI product to identify and monitor phenological changes in agricultural crops.

Abstract

Effective monitoring of agricultural crops is of importance for food security, yield predictions and effective land management. Rising human populations and shifting climate patterns are increasing pressures on the food system, augmenting the need to monitor crops. Remote sensing offers an approach to monitor crops over large areas and at regular temporal intervals. The Copernicus programme provides recent advancements in both the temporal and spatial resolution of Optical and Synthetic-Aperture Radar instruments. This research provides a comparative analysis between the use of Sentinel-1 and Sentinel-2 NDVI data for identifying crop type and monitoring crop phenology over the 2018 to 2019 growing season in Northern England. The results determine that Sentinel-1 and Sentinel-2 NDVI data are each effective at varying points in the growing season for monitoring crop growth at the field level. Sentinel-2 was more effective at identifying the emergence of crops through the soil, whilst Sentinel-1 is more effective at determine the more mature Spring phenological phases. The variance of the Sentinel-1 backscatter in each field was also found to be an important metric, with increased variance coinciding with crop ripening. For identifying crop types, stacked NDVI time-series produced higher classification accuracies earlier in the growing season than stacked Sentinel-1 data. The research concludes that both Sentinel-1 and Sentinel-2 had merits and limitations at different stages in the growing season. Future research should continue the growing body of literature around integrating Synthetic- Aperture Radar and Optical datasets to maximise the benefit of these two datasets within the agricultural industry.