Improving Forecasts of Flood Risk by Incorporating Climate Variability into Bulletin 17B LP3 Model

The current techniques for flood frequency analysis presented in Bulletin 17B assume annual maximum floods are stationary; meaning the distribution of flood flows is not significantly affected by climatic trends or long-term cycles (i.e. decadal variations). In light of growing evidence that streamflows are nonstationary and are impacted by climate variability, Bulletin 17B techniques should be modified. The effects of climatic cycles occurring over a shorter time frame, such as El Niño-Southern Oscillation (ENSO), are averaged into flood risk estimates obtained using the procedures of Bulletin 17B. However, the effects of ENSO are likely to affect the magnitude of annual maximum streamflows, and thus would impact flood risk in a given year. Estimates/forecasts obtained using the Bulletin 17B LP3 model can be improved by incorporating the effects of climate variability associated with ENSO events into updated estimates of the mean, and perhaps the standard deviation, by regressing the LP3 parameters on a climatic index such as sea surface temperature anomalies. In this study, a regression model is applied to annual maximum streamflow records for unimpaired gauging stations across the contiguous United States to obtain a one-year ahead forecast of the mean. For stations where the regression analysis yields significant results, the forecasted flood risk is compared with estimates obtained using the existing Bulletin 17B LP3 model.