Comparing the Conceptual Functional Equivalent to the National Water Model (CFE) versions 1.0 and 2.0, and the CAtchment Model for Arid and Semiarid regions (CAMAS) Against the National Water Model 3.0 Retrospective Simulation: The Value of Appropriate Conceptualizations

This study investigates the relative performance of different model formulations running within the Next Generation Water Resources Modeling Framework (NextGen) and comparing their streamflow outputs against both stream gauge observations and the National Water Model 3.0 retrospective simulation outputs. The first model tested is the Conceptual Functional Equivalent (CFE) to the WRF-Hydro based National Water Model (Version 3.1 and earlier). The CFE model is a catchment-based conceptual hydrologic model that identically applies methods as used in the WRF-Hydro based NWM for (1) rainfall partitioning; (2) soil moisture thresholds for lateral subsurface flow and percolation to deep groundwater; and, (3) conceptual groundwater non-linear reservoir to simulate base flow. CFE differs from the WRF-Hydro based NWM in that it replaces the 250 m routing grid with appropriate conceptualizations from the hydrologic science literature. The use of conceptual routing functions is justified by the fact that a 250 m grid often significantly distorts the cumulative density function of land surface slope, effectively “smoothing” the land surface and eliminating regions of higher slope. Grid size can be an important source of model errors because land surface slope is the driving variable in surface and some subsurface routing processes. Furthermore, in terms of surface flow routing a 250 m grid size presumes that overland flow persists at scales where it clearly does not. Microtopographic studies in the literature indicate that typical overland flow persistence length rarely exceeds 10 m before flow becomes channelized in a rille. The original version (1.0) of the CFE model applies a geomorphic instantaneous unit hydrograph (GIUH) to route surface water to the catchment outlet, which omits runon infiltration and retention depth processes. Version 2.0 of the CFE model uses the GIUH to calibrate parameters of a Nash-Cascade, which allows inclusion of retention depth and runon/channel infiltration loss conceptualizations. Both versions of the CFE model use a Nash Cascade to route lateral subsurface flow and similarly avoid grid size effects. We also evaluate streamflow outputs from a catchment-based infiltration-excess model that is formulated to accurately simulate catchments where stormflow generation is sensitive to rain rate and layered soils have a dominant effect such as tilled agricultural fields. This model employs the Layered Green & Ampt Infiltration with Redistribution (LGAR) scheme, that provides equivalent performance to the numerical solution of the Richardson/Richards’ equation (Hydrus-1D) on a continuous basis, without runtime, convergence, and mass conservation issues. Results show that the performance of the conceptual models often exceeds those from the NWM 3.0 retrospective simulation as determined using a wide range of continuous and event-based metrics. This presentation will discuss the performance and computational advantages of using appropriate conceptualizations for streamflow modeling in a standards-based model interoperability framework, and how they advance the NOAA/NWS mission to protect lives and property and promote the national economy.