Investigating the influence of drought on Sentinel-1 C-band SAR data over agricultural crops

Abstract

The stress on global food security is expected to increase. Hence, crop and drought monitoring will become increasingly important in the future. Synthetic Aperture Radars (SAR) are able to penetrate clouds and thus reliably provide data. Despite the extensive literature that can be found on the use of high spatio-temporal resolution SAR data, a study investigating the influence of drought on Sentinel-1 data over agricultural crops has yet to be conducted. This research aims to bridge this knowledge gap by utilizing Sentinel-1 data. The Sentinel-1 data is acquired and processed in Google Earth Engine and afterwards, data analysis is performed using Python. This results in parcel level SAR (VV, VH and VH/VV) data. This research is focused on maize, sugar beet, potato, onion and barley parcels in study areas in the Netherlands during the 2017, 2018 and 2019 agricultural summer season, of which 2018 and 2019 were impacted heavily by drought. 
This research demonstrates that phenological changes are reflected in Sentinel-1 data with increasing backscatter intensities during leaf development and stem elongation phases. Subsequently, saturation occurs which halts the rapid increase of backscatter. During harvest, the VH/VV ratio decreases rapidly. Time series of barley behave differently due to its unique vertical structure.The results show that VV and VH backscatter values are 2.5, 2 and 1 dB lower during the 2018 drought compared to 2017 for maize, sugar beet and potato parcels, respectively. Furthermore, the seasonal VH/VV ratio cycle for maize, onion and barley is shorter in a drought year and shortest in 2018. The VH/VV ratio cycle in 2018 was 30, 10 and 20 days shorter compared to 2017 for maize, onion and barley, respectively. Lastly, significantly lower VH/VV ratio values are observed during the vegetative stages in 2019. The percentage of individual parcels that show responses similar to aggregated responses ranges from 68% to 100%. Moreover, the results show that the overpass time has a large influence on drought response. Morning passes show significant increase in the magnitude of the VV and VH backscatter drop during the drought periods, especially for sugar beet and potato.The regional variability was assessed by comparing parcel backscatter from the northern part of the Vechtstromen water board, the Scheldestromen water board and the Flevopolder. Generally, drought impact is found to be most extreme in Vechtstromen. However, onions in 2018 were impacted most in Scheldestromen according to yield data. This clearly translated into lower VH backscatter and VH/VV ratio values during and after the drought period. Also, regional differences in maize time series caused by irrigation are observed. The results show that areas in which irrigation was allowed with ground and open water had a longer VH/VV ratio cycle in 2018, compared to areas in which irrigation was allowed only with groundwater.Overall, the usage of Sentinel-1 data for drought monitoring purposes shows tremendous potential. This gives a promising outlook on the use of dense C-band SAR data for the detection of crop drought stress.