Two New Radar Vegetation Indices (RVVI and RSVI) for Reconstructing NIRv as an Indicator of Tree Growth

Based on the latest research findings, structural vegetation indices such as the Near Infrared of Vegetation Index (NIRv) are better suited for determining tree growth than greenness indices like the Normalized Difference Vegetation Index (NDVI). Since the tree thickness growth-Vegetation Index (VI) relationship depends on the time of the growing season, continuous and cloud-independent datasets are necessary for operational applications. Consequently, we introduce two Sentinel-1 SAR-based VIs, the Radar Structure Vegetation Index (RSVI) and the Radar Volume Vegetation Index (RVVI), enabling operational modelling of structural photosynthetic capacity indicators.

Since there are no continuously cloud-free datasets suitable for operational applications, an operationally usable, cloud-free NIRv dataset was modelled using pairs of Sentinel-1 and Sentinel-2 data. Seven common radar VIs and the two newly developed RSVI and RVVI were calculated and tested for their tree species-specific correlation with NIRv over the entire growing season. To show the potential of RSVI and RVVI, a simple random forest model with forward feature selection (FFS) was trained using the local incidence angle, tree species, date within the growing season and Radar VIs as input variables.

NIRv's model results for reconstruction achieved an R² of 0.82 and MAE of 0.03. A total of seven variables were selected by the FFS. RSVI and RVVI showed highest increase of model explanation and were found to be the most important Radar VIs for modelling NIRv.

The introduced Sentinel-1 radar VIs, RSVI and RVVI, show great potential for modelling NIRv. The findings can help to identify early harvest damage in forestry and are a useful tool on the path to climate-resilient forests.