Conference Agenda

In 2022, this project continues to carry out research on space radiation benchmark transfer and calibration technology based on the RadCalNet site according to the plan. The accuracy of on-orbit satellite radiation calibration was improved, and the transfer calibration method based on RadCalNet TOA reflectivity products was improved by using a high accuracy and stability reference satellite. The TOA reflectance conversion model of Baotou site was optimized. The TOA reflectance model of the American site and Namibian site was constructed. Based on the above, an uncertain transfer link was constructed that connects a spatial radiation reference to each of the RadCalNet ground sites. This link can provide measurements of the contributions to total uncertainty caused by each factor.

Surface reflectance is the fundamental quantity required in the majority of optical Earth Observation analyses, and as an essential input to derive biophysical products. In addition to parameters such as the fraction of vegetation cover (FCOVER) and Canopy Chlorophyll Content (CCC), these products also include essential climate variables (ECVs) such as leaf area index (LAI). LAI is an integral plant canopy attribute and critical indicator of plant growth status. Currently several satellite derived LAI products exist, covering local to global scales with various spatial resolutions. In turn, they are crucial in understanding vegetation productivity/yield, biogeochemical cycles, and the weather and climate systems. Therefore, validation of such products is of great importance to ensure they meet the accuracy requirements for specific applications. However, ground measurements are not always match reflective of the spatial resolution of the satellite imagery, and contribute to uncertainty in the validation of LAI products. The key to reducing this source of uncertainty is upscaling from ground-measured LAI values to data representative of a satellite pixel.

In the study, high-spatial-resolution UAV remote sensing images were used as an intermediary for upscaling processing. We applied this approach to validate LAI retrievals based on Sentinel-2 and Gaofen-6 imagery (in which the Sentinel-2 Level-2 Prototype Processor (SL2P) was used to retrieve LAI from Sentinel-2, whilst a look-up-table (LUT) method was used to retrieve LAI from Gaofen-6). UAV images can well connect ground data and satellite data, thereby reducing the error caused by the mismatch of spatial resolution.

In the mid-term of the project, this study collects field LAI data and UAV images in Taizishan Forest Park (30.91-30.92°N, 112.87-112.88°E), China. Very high spatial resolution LAI reference maps were derived from the UAV imagery using four vegetation indices (VIs). In order to verify the LAI products retrieved by Sentinel-2 and Gaofen-6, we upscaled the UAV LAI map to 10m and 16m resolution. Finally, we compared the UAV-based upscaling approach to the direct comparison between LAI retrievals and ground measurements. Our results revealed improved correspondence between the satellite retrievals and UAV-based reference map when compared to direct comparison with the ground measurements (RMSE reduced from 1.02 to 0.59 for Sentinel-2 and 1.49 to 0.89 for Gaofen-6). Compared to SL2P, larger MAE(≥0.59) and RMSE(≥0.76) values were obtained for the Gaofen-6 LAI retrievals, indicating a need for further algorithm refinement.