EF5 · CREST US Parameter Explorer

EF5/CREST water balance

CREST is a single-soil-layer derivative of the Xinanjiang variable-infiltration model with an impervious-area partition. The six configurable parameters are mapped on the Parameter Maps tab.

Where EET is effective ET, EP effective precipitation, DP/SP the direct and soil portions, I infiltration, and ER excess rainfall split into overland and subsurface flow. (Eqs. 1–14 in the paper.)

CREST parameters (Table 1)

Soil texture → b exponent (Table 2, Cosby et al. 1984)

EF5/SAC-SMA water balance

SAC-SMA is the most complex option: an upper and a lower soil zone, each split into tension and free water, generating runoff by a saturation-excess process. It uses 21 parameters (11 distributed grids shown here, the rest uniform).

SAC-SMA parameters (Table 3)

Routing

Kinematic-wave routing approximates the 1-D Saint-Venant equations on steep slopes. Channel discharge follows a power law; overland flow uses a Manning-derived conveyance. Subsurface flow is routed separately as linear reservoirs.

Kinematic-wave parameters (Table 4)

Overland roughness — Manning's n by land cover (Table 5)

How well do the uncalibrated parameters perform?

The paper ran EF5 from 2002–2011 against 4,366 USGS gauges with basin area < 1000 km² (the flash-flood scale), forced by 5-minute MRMS radar rainfall. Skill is summarized with the Nash–Sutcliffe efficiency (NSE), Pearson correlation (CC), and normalized bias.

Performance summary (Table 6)

• CREST and SAC-SMA give comparable skill, with NSE up to 0.71 and 0.76 respectively — on some basins matching what a calibrated model would achieve.
• Skill drops in the intermountain West: radar beam blockage and unmodeled snow (no Snow-17 in this run) degrade the precipitation forcing.
• The Hydrophobic (HP) model is an upper-bound end-member — large positive bias (median 248%), useful for burn scars, saturated soils, and diagnosing rainfall under-estimation.

Toward CREST Global

This US dataset is the template for a global parameter set. The model physics are resolution- and location-independent; what must change are the source datasets used to derive each a-priori grid and the projection of the domain.

What carries over vs. what must be re-derived

Candidate global source datasets

• Terrain (DEM, flow dir/acc, α₀ slope): MERIT Hydro / MERIT-DEM (global, hydrologically conditioned, 3″≈90 m).
• Soils (Wm, Fc, b, SAC capacities): SoilGrids 2.0 and/or HWSD v2 for texture, available water capacity, and saturated conductivity.
• Impervious area & roughness (Im, Manning's n): global land-cover (ESA WorldCover, Copernicus GLC) and global impervious/built-up layers (GHSL).
• PET climatology: a global monthly PET product (e.g. derived from ERA5 / CRU / WorldClim) replacing the CONUS Koren climatology.
• Channel routing (α, β): re-fit the Vergara et al. (2016) geomorphology regression to global basins, or transfer with regional adjustment.

Practical steps

1. Reproject the working domain from NAD83 (EPSG:4269, US) to a global geographic grid (EPSG:4326).
2. Regenerate flow direction/accumulation from a global hydro-DEM at the target resolution.
3. Resample each soil/land-cover source onto the grid and apply the same lookup tables (Cosby b; Manning's n).
4. Re-derive scale-dependent routing parameters at the chosen resolution.
5. Validate against a global gauge network (e.g. GRDC) following the same NSE/CC/bias protocol.

This explorer is v1: once global grids exist in the same GeoTIFF layout, the exact same processing pipeline and interface visualize them with no code changes.