 |
| Titel |
Rainfall-runoff model calibration at an ungauged catchment using the map-correlation method |
| VerfasserIn |
Stacey Archfield, Richard Vogel, Thorsten Wagener, Riddhi Singh |
| Konferenz |
EGU General Assembly 2010
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 12 (2010) |
| Datensatznummer |
250036895
|
|
|
|
|
|
| Zusammenfassung |
| The International Association of Hydrological Sciences ten-year Prediction in Ungauged
Basins (PUB) initiative encourages the development of approaches to estimate streamflow at
ungauged catchments. One such approach is to transpose calibrated rainfall-runoff model
parameters from a gauged, reference catchment to an ungauged catchment. Central to this
approach is the selection of the reference catchment from which to transpose the model
parameters to the ungauged catchment. Previous studies have found the selection
of the reference catchment to be problematic and different selection criteria have
shown little success. We introduce the map-correlation method, which selects a
reference catchment whose logarithms of daily streamflow are most correlated with the
ungauged catchment. This is achieved by first kriging the cross-correlations, r,
between the logarithms of daily streamflow at each reference catchment and all
other reference catchments in the study area. Then, at an ungauged catchment, the
map-correlation method yields an r value for each reference catchment. The reference
catchment resulting in the highest r value is selected. To determine if r is related to
model goodness-of-fit, 34 sets of simulation model parameters were obtained by
calibrating rainfall-runoff models at 34 gauged catchments in the mid-Atlantic United
States. At each of the 34 study catchments, the other 33 calibrated sets of model
parameters were transposed to the study catchment, resulting in 33 goodness-of-fit
values between the observed and estimated daily streamflows. These goodness-of-fit
values were compared to the r values estimated from the observed, concurrent daily
streamflow between each study catchment and the other 33 catchments. This comparison
was repeated for each of the 34 catchments to obtain 1,122 (34 multiplied by 33)
goodness-of-fit and r values. The relation between r and goodness-of-fit will be
presented and contrasted with the use of other criteria to choose a reference catchment. |
|
|
|
|
|
|