2.4 CREST-iMAP model
Hydrologic modeling is so far a common approach to deliver timely flood information for the sake of scalability and efficiency (Gourley et al., 2017). Yet, conventional hydrologic models bear large uncertainties in such developed regions, which is mainly due to 1) simplified representation of terrain (Dullo et al., 2021) and 2) one-dimensional routing that raises issues in flat regions (Flamig, Vergara, & Gourley, 2020; Getirana & Paiva, 2013; Li et al., 2021b). On the other hand, hydraulic models do not excel in representing hydrologic processes. In light of these issues, the newly developed Coupled Routing and Excess STorage inundation MApping and Prediction (CREST-iMAP) model is used to investigate the importance of the re-infiltration scheme in flood inundation models. The CREST-iMAP integrates CREST V2.1 for the hydrologic part that simulates vertical water distribution by land surface and ANUGA V2.1 for the hydraulic routing that distributes spatial water over terrain by solving 2D shallow water equation. Its performance has been evaluated in this region against the non-coupled hydrologic models and other popular coupled models – WRF-Hydro+HAND and LISFLOOD FP (Chen et al., 2021; Li et al., 2021b). However, the previous version of CREST-iMAP V1.0 does not include the re-infiltration scheme, meaning that surface running water is not allowed to re-enter the soil. Here, we release the CREST-iMAP V1.1, an upgrade version, which considers two-way coupling via exchanging surface water between the hydraulic and hydrologic module and re-infiltration. Two different schemes are illustrated schematically in Figure 2, where the left panel represents the re-infiltration scheme, and the right does not. The CREST-iMAP V 1.0 and V1.1 are openly accessible from https://github.com/chrimerss/CREST-iMAP.
[INSERT FIGURE 2 HERE]
CREST-iMAP inherits the previous version of the CREST model, which simulates saturation excess runoff as the primary runoff generation process (Wang et al., 2011; Xue et al., 2013; Flamig, Vergara, & Gourley, 2020). The schematic model structure is depicted in Figure 2. The study area is discretized in variable triangular meshes which allow higher density in river channels to resolve high-resolution river flow. Each modeling unit receives excess rainfall (rainfall minus evaporation) from forcing data. Then surface water is divided into overland flow and soil water according to the impervious area ratio through linear weighting. Overland flow is generated once soil water exceeds its holding capacity; otherwise, soil water is separated into the remaining amount and interflow based on the Variable Infiltration Curve (VIC) concept. The VIC model is a widely recognized infiltration model that has been applied in several classic hydrologic models (Liang, Lettenmaier, Wood, & Burges, 1994; Zhao, 1995). Overland flow, combined with the impervious area and saturation excess flow, is eventually fed into the 2D shallow water equation solver – the Finite Volume Scheme. It solves water depth and momentum distributed at each grid cell and propagates across boundaries. The outputs of the model include water depth, velocity, discharge, and soil moisture at a desired time step. In the current setting, re-infiltration, termed as water moves from subsurface to surface, is not considered because surface flow is the dominant process in major flood events (Freeze, 1974). The flexibility of the unstructured mesh in CREST-iMAP allows dense meshes in regions that reflect high terrain variability (e.g., river channel) and sparse meshes in other regions (e.g., flood plain). This study simulates the extreme flood events at 10-m resolution using the embedded unstructured mesh generator.
There are five hydrologic parameters and one hydraulic parameter for the CREST-iMAP, which are listed in Table 2 along with parameter ranges. It is noteworthy that all these parameters are spatially distributed to account for the spatial heterogeneity of land cover and soil types. The mean soil saturated hydraulic conductivity, Ksat from 0 to 20 mm/d , indicates the soil infiltration capability. Higher Ksat values imply higher infiltration rates if soils are not saturated while reaching plateau for the saturated soils. The mean soil water capacity, WM from 10.4 to 365.4 mm, measures the total water content the soil can hold with lower value representing the impermeable soils. The exponent of the Variable Infiltration Curve (VIC), B, determines soil water partitioned to saturation excess runoff or interflow, with a higher B value corresponding to higher infiltration rates. KE is the ratio of the potential evapotranspiration to actual evapotranspiration, similar to the concept of pan coefficient. These soil-related a-priori parameters can be approximated from a look-up table at an individual grid cell basis (Chow, Maidment, & Mays, 1988). There are also CONUS-wide optimized parameter sets that are configured for operational flood monitoring systems (Flamig, Vergara, & Gourley, 2020). The impervious area ratio, IM from 0% to 100%, is obtained directly from the NLCD dataset; the manning’s n coefficient is derived from the LULC via a look-up table. Both parameters determine water conveyance capacity, meaning that higher values relate to faster and larger flood peaks. The hydrologic parameters are configured at their optima based on previous study, but for the hydraulic parameter – manning coefficient, we manually adjusted it in a preceding event to ensure generating timely and accurate possible flood peaks.
[INSERT TABLE 2 HERE]