Abstract

Normalized Bidirectional Adjusted Reflectance (NBAR) is a key parameter for a consistent time series monitoring over non-lambertian surfaces. The Sen2like is a Virtual Constellation (VC) which harmonizes and fuses Landsat 8 / Landsat 9 & Sentinel 2 dataset giving out a higher spatial and temporal resolution surface reflectance. However, for adequate monitoring of land surface is necessary the correction of sun and sensor angle view across the VC acquisitions. In this context, the High resolution Adjusted BRDF Algorithm (HABA) provides up to 10m NBAR product retrieved from the disaggregation of the Bidirectional Reflectance Distribution Function (BRDF) parameters based on the VJB method applied to MODIS M{O,Y} D09 Climate Model Grid (CMG) at 1km resolution. HABA downscales this product to Sen2Like resolution inverting BRDF parameters (V & R) using the k-means unsupervised classification for each dataset. In order to compensate for the impact on images that do not present sufficient data representativeness due to cloud coverage, the disaggregated parameters are stabilized computing linear trends of time series of Normalized Difference Vegetation Index (NDVI) versus V & R. The model was evaluated on stable sites, such as Sahara Desert (Libya) and Amazonian Forest (Brazil) by comparing the impact of View Zenith Angle (VZA) and Solar Zenith Angle (SZA) of directional reflectance, a static NBAR model and HABA for Near InfraRed (NIR) and red spectrum. Also, the Sen2Like performance was assessed on dynamic sites with a mosaic of land covers across the Belgium tiles, calculating the absolute difference per tile in a 5-day window. The results of stable sites show a decline of linear dependency on the Amazon VZA from R² 0.57 (directional) to 0.37 (HABA) in NIR and R² 0.04 (directional) to 0.0 (HABA) in red. The Sahara Desert showed a correction of 4% of linear dependency of SZA versus reflectance. Finally, in Belgium, HABA corrected up to 12,74 % the directional effect on the time series. This work contributes to develop a dynamic and operationalization of NBAR correction method based on pixel scale for high resolution datasets.