1. Assimilation of Sentinel‐1 Backscatter to Update AquaCrop Estimates of Soil Moisture and Crop Biomass.
- Author
-
de Roos, Shannon, Bechtold, Michel, Busschaert, Louise, Lievens, Hans, and De Lannoy, Gabrielle J. M.
- Abstract
This study assesses the potential of regional microwave backscatter data assimilation (DA) in AquaCrop for the first time, using NASA's Land Information System. The objective is to assess whether the assimilation setup can improve surface soil moisture (SSM) and crop biomass estimates. SSM and crop biomass simulations from AquaCrop were updated using Sentinel‐1 synthetic aperture radar observations, over three regions in Europe in two separate DA experiments. The first experiment concerned updating SSM using VV‐polarized backscatter and the corrections were propagated via the model to the biomass. In the second experiment, the DA setup was extended by also updating the biomass with VH‐polarized backscatter. SSM was evaluated with local in situ data and with downscaled Soil Moisture Active Passive (SMAP) retrievals for all cropland grid cells, whereas crop biomass was compared to SMAP vegetation optical depth and the Copernicus dry matter productivity. The assimilation showed mixed results for root mean square error and Pearson's correlation, with slight overall improvements in the (anomaly) correlations of updated SSM relative to independent in situ and satellite data. By contrast, the biomass estimates obtained with backscatter DA did not agree better with reference data sets. Overall, the SSM evaluation showed that there is potential in using Sentinel‐1 backscatter for assimilation in AquaCrop, but the present setup was not able to improve crop biomass estimates. Our study reveals how the complex interaction between SSM, crop biomass and backscatter affect the impact and performance of DA, offering insight into ways to optimize DA for crop growth estimation. Plain Language Summary: This study evaluates if using observations from a microwave satellite, Sentinel‐1 (S1) can improve model simulations of a crop model AquaCrop, specifically for both soil moisture and crop biomass, over different regions in Europe. For each day in which S1 observations were available over the region, the modeled soil moisture and biomass were "updated" based on these observations, which over time is expected to reduce the model error and uncertainty. This iterative process is called data assimilation (DA) and was executed in a model framework called NASA's Land Information System. Two DA experiments were held. In the first DA experiment, only soil moisture was updated by S1 observations, but the changes in soil moisture were expected to also affect the biomass simulations compared to no DA model runs. In the second DA experiment, both the soil moisture and biomass were updated with S1 data. When comparing the results with independent data sets, the assimilation showed mixed results. The soil moisture showed slight improvements after DA, but the biomass estimates did not improve. Given the complexity of S1 data over agricultural areas, more research is required to optimally perform DA before this setup is able to improve crop growth estimation. Key Points: First assessment of Sentinel‐1 backscatter data assimilation in a crop model integrated into NASA's Land Information SystemCo‐ and cross‐polarization backscatter observations were used to update regional AquaCrop soil moisture and biomass estimates, respectivelySentinel‐1 data assimilation resulted in improved soil moisture estimates, but further research is needed for optimal vegetation updating [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF