First Validation of GEDI Vegetation Structure Metrics in South African Savannas.

Abstract:

Savannas have complex vegetation structure that varies greatly in its vertical and spatial arrangement and researchers are eager to determine how well the new GEDI data characterises them. This study validated GEDI structure metrics by comparing the GEDI footprint data metrics (Relative Height - RH) to the simulated GEDI metrics derived from Airborne Laser Scanning (ALS) data across a range of forest types (savanna, thicket, sand forest) in South Africa. From the first 7 months of GEDI (v.1) data released to date, 10 orbits intersect 7 sites with recent ALS data (2018, 2019), providing a total of 14 test cases (each orbit per site = 1 test case) after applying degrade flags and quality flags to filter bad quality data. Only RH95 results are presented here. Ten out of 14 test cases had strong relationships between GEDI metrics (RH95) and simulated GEDI metrics, with R2 > 0.65, biases between -1.16m to +1.42m and small RMSEs 0.35-1.55m. Four test cases had moderate relationships, with R2 0.3-0.45, biases between -0.61m to + 2.63m and RMSEs between 0.6-1.3m. Furthermore, the GEDI power beam data collected during nighttime performed better than the coverage beam, and test cases with weak relationships were more likely to have a big difference in phenology (i.e. leaf-on vs. leaf-off) between the dates of the ALS and GEDI data. Overall, the strong relationship between GEDI and ALS metrics demonstrated that GEDI data have very good potential to measure complex vegetation structures in diverse forest types, including savannas.