3.1.2 Spatiotemporal relationships
To explore the spatiotemporal differences of scenarios with and without re-infiltration, we set three parameters to their default values, namely an initially dry soil condition and normal soil infiltration rates (i.e., a-priori setting). Figure 4 shows the difference in flood extent for the two schemes. Despite a considerable amount of grid cells showing positive agreements (i.e., both detect floods; TP=16.0%), there are still 1.7% of the grid cells issuing false positives, amounting to 15.6k grid cells (~1.56 km2) in this model configuration. Specifically, those false positives accumulate around upstream floodplains while the downstream such as areas near the basin outlet does not present discernable differences, as the flood depth there due to accumulation is not sensitive to inundation thresholds for flooded cells. Moreover, Figure 5 portrays the spatial distribution of the differences with respect to maximum depth, initial inundation timings, and total inundation duration. Figure 5a depicts the major differences that are situated in floodplains and river channels where surface water is accumulated via routing, and the maximum depth difference is up to 3 meters in the river channel, especially downstream of Halls Bayou. However, the initial inundation timings and durations are scattered sparsely over the study area, with a majority of the grid cells showing earlier and longer inundations for the case without re-infiltration scenario. Therefore, it is likely that the flood is over-predicted by models without the re-infiltration scenario.
[INSERT FIGURE 4 HERE]
Meanwhile, we notice that there are some samples in the opposite distribution, indicating delayed and/or shorter inundation time. Arguably, this could be some local effects when the soil reaches earlier saturation in the re-infiltration scenario, thereby leading to earlier flooding. A supporting material is found in Figure 5d, in which the basin-average soil moistures of two schemes are compared. Notably, evapotranspiration is not considered in this ideal test, so the soil moisture does not deplete with time. Within the storm lifetime, soil moisture surges from completely dry to 85% saturation for the scenario without re-infiltration and to 95% saturation for scenario with re-infiltration. Early saturation reduces infiltration rates later on and thus has pronounced effects on local flooding when surface water is not routed timely. Figure 5e presents the evolution of surface water volume by integrating surface water depth along with grid cells. Although both scenarios similarly reach the maximum surface water volume concurrently, their recession limbs show considerable differences. The re-infiltration scenario apparently has a steep exponential decay, as both still water and running water infiltrates into the soil; for the scenario without re-infiltration, in contrast, there is a mild decay and even levels out at the end of the simulation. The difference between the two increases with time, as shown in the shaded area, up to\(0.4\times 10^{8}\) m3 volume difference, which equates to almost half of the total surface water volume.
[INSERT FIGURE 5 HERE]
In summary, re-infiltration scheme indeed influences flood magnitude and timings via surface water-soil interaction, and it possibly reduces flood magnitude and delays (shortens) flood timing (duration). Flood magnitude differences are pronounced downstream or in depressions, while flood timings are scattered. Such results are markedly tied to soil condition (wet or dry) and soil characteristics (infiltration capacity).