Geoscience Reference
In-Depth Information
6.4 Relationship between radar data resolution, catchment characteristics and
hydrological model performance
Various studies have examined the impact of the spatial and temporal resolution of
remotely sensed precipitation observations on hydrological model performance (Krajewski
et al., 1991; Pessoa et al., 1993; Obled et al., 1994; Ogden & Julien, 1994; Ball, 1994; Faurès et
al,, 1995; Shah et al., 1996; Winchell et al., 1998; Bell & Moore, 2000b; Carpenter et al., 2001).
Ogden and Julian (1994) explored the relationship between catchment size, and the
correlation length and horizontal resolution of the radar derived precipitation fields input to
a two-dimensional, physically-based hydrological model. They defined two, dimensionless
length parameters and considered their impacts on the accuracy of predicted run-off.
Storm
smearing
describes a reduction in the horizontal gradient of precipitation rate as the
horizontal resolution of the radar data approaches its correlation length.
Watershed smearing
occurs when the horizontal resolution of the radar data approaches the characteristic length
scale of the catchment (square root of the catchment area). Watershed smearing was shown
to be the main source of error in predicting river flow over small catchments. Berenguer et
al. (2005) point out that the sensitivity of hydrological models to biases in mean areal rainfall
are due to the fact that river catchments act as integrators of the precipitation falling on
them.
Bell and Moore (2000b) emphasized the need to calibrate hydrological models with rainfall
data for a given resolution. They found that the most skilful distributed rainfall-run-off
model predictions were made with lower resolution rainfall data. This finding was
interpreted as evidence for the need to improve distributed hydrological model formulation.
6.5 Impact of radar intensity resolution
The impact of the intensity resolution in radar data on hydrological forecast errors was
investigated by Cluckie, Tilford & Shepherd (1991). They demonstrated that 8 intensity
levels were adequate for the majority of rural and urban catchments in the majority of UK
precipitation events. This is because the bulk of the relevant information content is
concentrated at the low frequency end of the power spectrum. Nonetheless, in convective
precipitation events, a reduction in intensity resolution may have an effect similar to that of
the storm smearing described by Ogden and Julian (1994).
6.6 Benefits of precipitation nowcasts to hydrological forecasting
In the absence of precipitation forecasts, the lead time of flood warnings is limited by the
catchment response time, a quantity dependent on catchment size, morphology and land
use. Skilful precipitation forecasts offer the prospect of some forewarning of flash floods in
small, fast responding catchments, and of extending the lead time of flood warnings in other
catchments (Roberts et al., 2009).
Although numerous authors have evaluated the impact of QPE algorithms on the utility of
radar for hydrological forecasting there have been relatively few investigations of the
benefits of precipitation nowcasts in this area. Cluckie and Owens (1987) compared the
performance of stream flow forecasts made using a linear transfer function model and radar
extrapolation nowcasts from FRONTIERS (Browning, 1979) against similar flow predictions
