Conference Agenda

This project carries out research on cross calibration technology based on the RadCalNet site for high resolution optical satellites. As some satellite can provide a radiometric calibration reference with high accuracy and good stability, in this project, they are selected to reduce the uncertainty of calibration caused by the temporal differences between the reference and to-be-calibrated payload, which effectively promote the accuracy of the on-orbit radiometric calibration and also modify the accuracy of cross-calibration model based on the TOA reflectance of RadCalNet sites. Additionally, TOA reflectance conversion models are optimized in site of Baotou, China and constructed in sites of Railway Vally Plaza, US and Gobabeb, Namibia. Based on that, the validation of the promoted method based on the RadCalNet sites is then conducted and the uncertainty in each step can be evaluated quantitatively.

Leaf area index (LAI) is an important parameter in quantitative terrestrial models describing carbon and water cycles at regional and global scales. Verification of the accuracy and reliability of land products of leaf area index (LAI) from Earth observation is an important issue currently facing the remote sensing community. Ground measurements, as the "true value", are an important indicator in the verification process of remote sensing products. Direct verification methods based on actual ground measurement points often require a large area of homogeneous ground material to correspond to satellite product pixels. However, in actual remote sensing observations and applications, a large homogeneous target is often difficult to find. Therefore, how to verify satellite products under the condition of uneven surface is a problem that needs to be solved. On the other hand, quantitative remote sensing product inversion models are based on certain spatial scale effects. Ground and satellite scales will also affect the quality of LAI products, and the uncertainty of these data will be transferred to the data to be processed during the verification process.

UAS imagery, with its higher resolution and more flexible revisit period, is considered a powerful verification tool that can replace ground surveys. However, LAI mapping based on UAS imagery is usually determined by vegetation indices (VIs) that enhance spectral characteristics. The traditional empirical model does not consider the radiation mechanism of the complex canopy. On the other hand, the LAI products and inversions from the Sentinel-2 and Gaofen-6 satellites all include radiative transfer modules, and the different inversion methods of different platforms also contribute to the instability of LAI verification.

This study conducts experiments on scale conversion for LAI product verification and implements a radiative transfer model (RTM)-based UAS LAI inversion. First, the ground measurement "point" data and the VIs-based UAS LAI "area" data are upscaled to the Sentinel-2 satellite pixel scale. We compared three methods for upscaling UAS LAI maps to 10 m, namely the regional mean method, the point spread function (PSF) and the PSF improvement (PSFI). We compared three methods for upscaling UAS LAI maps to 10 m, namely the regional mean method, the point spread function (PSF) and the PSF improvement (PSFI). We verified the LAI products obtained by the Sentinel-2 inversion and compared them with the traditional direct comparison of satellite product and ground measurement. In the RTM-based canopy index inversion, we applied a look-up table (LUT) approach based on RTM PROSAIL model and perform pixel-wise LAI inversion of UAS images and compared the result with the previous VIs-based LAI mapping.

The results show that the VI-based UAS LAI map upscale to 10m is more consistent with the satellite LAI product than the traditional ground measurement. The R² increased from 0.55 to 0.64, and the RMSE decreased from 0.49 to 0.44, with an error of 15%. The number of pixels within the area increased from 56.6% to 60.9%, proving that LAI maps based on UAS VIs estimation can be used as true values to verify other satellite products such as Gaofen-6. Compared with the best VIs-based LAI retrieval from Atmospherically Resistant Vegetation Index (ARVI), the results show that the UAV LAI inversion based on radiative transfer model has a nicer match with GF-6 LAI product with R² of 0.68, the RMSE is reduced from 0.76 to 0.7 and the MAE is reduced from 0.59 to 0.55.

In the next step, we will focus more on testing the temporal consistency of VIs based UAS LAI products with satellite products and explore the possible way to constrain the RTM based UAS LAI inversion.