Remote sensing has emerged as a valuable tool for monitoring peatland water table depth (WTD), offering local to global-scale observations and high-frequency monitoring capabilities. We present the first attempt to adapt the trapezoidal model concept to SAR, developing the SAR trapezoid model (SARTRAM) for peatland WTD remote sensing. The model is based on the interpretation of the pixel distribution within the surface backscatter-radar vegetation index (RVI) trapezoid space. We used Sentinel-1 and Sentinel-2 data to evaluate and compare the performance of SARTRAM, surface backscatter (Formula: see text) and the traditional optical trapezoid model (OPTRAM) in 29 restored peatlands and 98 WTD logging sites in Scotland and Lithuania. We employed five different model value extraction methods: value at logger point, average buffer value, best-pixel (BP) within the peatland area, BP within 100 m (BP100m) and 500 m (BP500m) search distance buffers. To enhance performance, a moving average window (MAW) filter was applied to SARTRAM. This study demonstrated that SARTRAM exhibited substantial potential, particularly when temporal MAW filtering was applied, resulting in a 10–19% improvement in correlation with WTD compared to the initial SARTRAM results. SAR backscatter showed an average Spearman correlation coefficient of 0.52 and 0.57 with WTD for BP500m and BP methods (VV polarization), respectively. The model with MAW-filtered SARTRAM achieved average Spearman correlation values of 0.65 and 0.62 with the VH and VV polarization data for the BP and BP500m methods, respectively. OPTRAM outperformed SARTRAM across all value extraction method cases, resulting in average correlation coefficients of 0.85 and 0.83 for the BP and BP500m methods, respectively. Nevertheless, SARTRAM demonstrated slightly better overall results compared to traditional WTD prediction methods based on SAR backscatter signal analysis.
Jukna et al. (Fri,) studied this question.