Limitations on Seasonal Snowmelt Forecast Accuracy

Hydrologic model calibration is usually a central element of streamflow forecasting based on the ensemble streamflow prediction (ESP) method. Evaluation measures of forecast errors such as root-meansquare error (RMSE) are heavily influenced by bias, which in turn is readily reduced by calibration. On the other hand, bias can also be reduced by postprocessing (e.g., "training" bias correction schemes based on retrospective simulation error statistics). This observation invites the question: How much is forecast error reduced by calibration, beyond what can be accomplished by postprocessing to remove bias? The authors address this question through retrospective evaluation of forecast errors at eight streamflow forecast locations distributed across the western United States. Forecast periods of length ranging from 1 to 6 months are investigated, for forecasts initiated from 1 December to 1 June, which span the period when most runoff occurs from snowmelt-dominated western U.S. rivers. ESP forecast errors are evaluated both for uncalibrated forecasts to which a percentile mapping bias correction approach is applied, and for forecasts from an objectively calibrated model without explicit bias correction. Using the coefficient of prediction (C p ), which essentially is a measure of the fraction of variance explained by the forecast, the authors find that the reduction in forecast error as measured by C p that is achieved by bias correction alone is nearly as great as that resulting from hydrologic model calibration.

show abstract

“…We evaluated forecast skill using the coefficient of prediction, C p , as defined by Lettenmaier (1984):…”

Section: E Assessment Of Forecast Skillmentioning

confidence: 99%

How Essential is Hydrologic Model Calibration to Seasonal Streamflow Forecasting?

Shi

Wood

Lettenmaier

2008

Journal of Hydrometeorology

Self Cite

123

View full text Add to dashboard Cite

show abstract

“…There are three sources of error inherent in rainfall-runoff models [Lettenmaier, 1984]' model error, input error, and parameter error. Model error results in the inability of a rainfallrunoff model to predict runoff accurately, even given the correct estimates and input.…”

Section: (• Mean Value Over the Verification Events I = N + 1 Tomentioning

confidence: 99%

A Comparison of Rainfall‐Runoff Modeling Techniques on Small Upland Catchments

Loague¹,

Freeze²

1985

Water Resources Research

291

192

View full text Add to dashboard Cite

This paper reports a set of model performance calculations for three event‐based rainfall‐runoff models on three data sets involving 269 events from small upland catchments. The models include a regression model, a unit hydrograph model, and a quasi‐physically based model. The catchments are from the Washita River Experimental Watershed, Oklahoma; the Mahantango Creek Experimental Watershed, Pennsylvania; and the Hubbard Brook Experimental Forest, New Hampshire. Model performance was assessed for a verification period that is carefully distinguished from the calibration period. Performance assessment was carried out both in forecasting mode and in prediction mode. The results show surprisingly poor forecasting efficiencies for all models on all data sets. The unit hydrograph model and the quasi‐physically based model have little forecasting power; the regression model is marginally better. The performance of the models in prediction mode is better. The regression model and the unit hydrograph model showed acceptable predictive power, but the quasi‐physically based model produced acceptable predictions on only one of the three catchments. We believe that the primary barrier to the successful application of physically based models in the field lies in the scale problems that are associated with the unmeasurable spatial variability of rainfall and soil hydraulic properties. The fact that simpler, less data intensive models provided as good or better predictions than a physically based model is food for thought.

show abstract

“…Notwithstanding practical complications that in some cases can constrain the forecast accuracy of ESP methods (see, e.g., Lettenmaier 1984;Day 1985, for a discussion), there is an ongoing trend away from the traditional regression-based forecasts for some of the reasons noted above. The NWS AHPS initiative has motivated implementation of ESP for seasonal streamflow forecasts at an expanded set of forecast points in the western U.S. RFCs.…”

mentioning

confidence: 99%