Improved retrieval of land surface biophysical variables from time series of Sentinel-3 OLCI TOA spectral observations by considering the temporal autocorrelation of surface and atmospheric properties

• DOI: 10.1016/j.rse.2021.112328
• Publication Date: 2021-02-11T06:09:15Z
• Authors: Peiqi Yang; Wout Verhoef; Egor Prikaziuk; Christiaan van de...
• Keywords: Time series; UT-Hybrid-D; Radiative transfer model; 04 agricultural and veterinary sciences; 15. Life on land; LAI; ITC-HYBRID; 13. Climate action; ITC-ISI-JOURNAL-ARTICLE; Temporal autocorrelation; SPART; 0401 agriculture, forestry, and fisheries; Sentinel-3; OLCI; Model inversion; SDG 15 - Life on Land

Estimation of essential vegetation properties from remote sensing is crucial for a quantitative understanding the Earth system. Ill-posedness model inversion problem leads to multiple interpretations one satellite observation, and using prior information promising way reduce ill-posedness increase accuracy land surface products. Tobler's first law geography states that "everything related everything else, but near things are more than distant things". Likewise, it expected state an object at single moment every other moment, temporally attributes ones. This temporal autocorrelation vital source can be used improve retrieval accuracy. In this study, we develop framework makes use dependence atmospheric properties. We apply algorithm Sentinel-3 Ocean Land Colour Instrument (OLCI) data derive biophysical variables with focus on leaf area index (LAI) top-of-atmosphere (TOA) radiance observations. The results both synthetic dataset real show continuity as priori improves estimation properties, such chlorophyll content LAI. Compared MODIS LAI products, much less unrealistic short-term fluctuations found in retrievals OLCI time-series approach across different cover types including cropland, forest savannah. Field measurements two sites quantitatively confirm estimated reasonably accurate R2 > 0.65 RMSE < 1.00 m2m−2. Overall, time series robust smoother standard individual scenes, stable product, values match better reported field. present make spectral observations potentially multi-sensor