*4.1. Performance of the Precision Index in the Rasterization Process without Sacrificing Spatial Resolution*

The specular point vector data derived from the Lv. 1 product were rasterized to form the Lv. 2 product, as shown in Figure 4. All the specular points depicted from southwest to northeast had relatively fine spatial resolutions due to the low effective-scattering-area values. In contrast, all the specular points depicted from northwest to southeast had relatively coarse spatial resolutions due to the relatively large effective scattering areas. It is known that the effective scattering area and footprint size/shape are mainly controlled by the Delay and Doppler effects [32]. This indicates that the relationship between the velocity (particularly in the advancing direction) of the GNSS receiver and the transmitter is the main factor controlling the spatial resolution of each specular point rather than the difference in the incidence angle or land surface roughness over lowlands such as the Mekong Delta (which has an elevation approximately 2 m above the sea surface) [21]. In this context, further GNSS-R receivers are expected to flexibly choose/adjust their transmitters to continuously receive only fine-spatial-resolution GNSS signals. The future use of geostationary GNSS transmitters or quasi-zenith-satellite-system-boarded transmitters (QZSSs) is also expected to be selected occasionally in specific regions.

The precision index developed in this study was designed to be maximized at the centers of the specular points, as the maximum analog power was detected at the center of the DDM (i.e., the neutral Delay/Doppler position), as shown in Figure 3. Since highdelay specular points are occasionally found in the Mekong Delta, hollow-ring-shaped Gaussian kernels might be appropriate for such unique specular points [32]. To further improve the index, such spatial localization regarding the *dst.centerSP*/*semidiameter.SP* ratio in the denominator of Equation (2) should be implemented for the further development of specular points with relatively high-delay chips. Considering the unique specular points for which the maximum analog power is located in the non-neutral Doppler bin, performing spatial localization while considering the specular advancing direction and Doppler effect would also be desirable, although such specular points were rarely found over the delta in this study.

The temporally Kalman-smoothed product (Lv. 3) clearly visualized the spatial pattern throughout the year, even without spatial interpolation/filtering/smoothing. This indicated that spatial inundation mapping can be accomplished even without performing a bias correction, depending on ad hoc parameter tuning to deal with incidence angle differences or even without depending on external NDVI data to deal with vegetation interactions. Although noise associated with relatively high Γ values is occasionally detected with small effective scattering specular points (Figure 5a), such specular noise was seemingly found to be accompanied by high DDM 3D skewness/kurtosis values (Figure 5c,d) and could thus be denoised naturally, as shown in Figure 5b.
