Conclusions

The aim of the reported study was to predict woody vegetation FTCC at 20 m resolution for floodplain vegetation and evaluate predictions using LiDAR data. This study has shown that a combining predictor model was able to explain up to 91% of FTCC variation, returning an acceptable RMSE at our study sites. Individual models (RFYanga and RFBarmah) displayed weaker correlations and larger errors when compared to the combined model. Analysis of sensor band importance suggests SWIR is the most important band which contributes mostly to model training as it is sensitive to variation in leaf area index and leaf water content. Additionally, Sentinel-1 (radar) band contributions cannot be ignored for Random forest model training. Our presented approach will prove useful in expanding knowledge of remote sensing ET related directly to tree ET, improving estimations at a finer spatial resolution. This study will be significant to further our collective understanding of floodplain vegetation response to climatic conditions and catchment water management.