Developing Peak Discharges for Future Flood Risk Studies Using IPCC's CMIP5 Climate Model Results and USGS WREG Program

Abstract

Extreme climate events, floods, and drought, cause huge impact on daily lives. In order to produce society resilient to extreme events, it is necessary to assess the impact of frequent and high intensity storm events on design parameters. This article describes a methodology to develop future peak "design discharges" throughout the United States that can be used as a guidance to map future floodplains. In order to develop a lower and upper limit for anticipated peak flow discharges, two future growth scenarios -Representative Concentration Pathways (RCPs)-RCP 2.6 and 8.5 were identified as the weak and strong climate scenario respectively based on the output from the global climate models. The Generalized Least Square technique in United States Geological Survey's Weighted Multiple Regression (WREG) program was used to develop regression equations that relate peak discharges to basin and climate parameters of the contributing watershed. The design discharges reflect the most recent climate model results. Number of frost days, heavy rainfall days, high temperature days, and snow depth were found to be the common extreme climate parameters influencing the regression equations. This methodology can be extended to other flood frequency events if rainfall data is available. The future discharges can be utilized in hydraulics models to estimate floodplains that can assist in resilient infrastructure planning and outline climate change adaptation strategies.(KEY TERMS: climate variability/change; design discharges; representative concentration pathways; weighted multiple regression.)

A Framework to Develop Nationwide Flooding Extents Using Climate Models and Assess Forecast Potential for Flood Resilience

“…Regression equations for estimating flows for future climate scenarios were developed for 18 Hydrologic Unit Code 2 (HUC-2) regions within the contiguous U.S. (Selvanathan et al, 2016). Table 1 lists the basin and climate parameters that significantly influenced flow in each HUC-2 region based on these established equations.…”

Generating Regional Models for Estimating the Peak Flows and Environmental Flows Magnitude for the Bulgarian-Greek Rhodope Mountain Range Torrential Watersheds

“…A total of 25 watershed explanatory variables were calculated and used for the development of the peak discharge and Q env regional models. To the best of the authors' knowledge, among them, the length of the straight line from the source to gauging-station (L), the Total stream length (Tl), the circulatory ratio (Rc), the elongation ratio (Re), and the lemniscate ratio (R l ) [44] have been considered for the first time for such purposes, whereas the remaining 20 characteristics have been used in previous regional flood frequency studies and they include the following: Drainage area (A) [1,7,41,[45][46][47], Main-channel length (Ls) and Main channel slope (S), [1,7,41,46], Basin length (BL) [7,41,45], Mean basin slope (BS) [46,47], Mean basin elevation (E) [1,7,45,47], Basin Perimeter (B) [41], Mean basin width (BW), Form factor (Rf ), Curve factor (Cr), or main-channel sinuosity, accounting the main rivers meandering, gaging-station latitude (GLT) and longitude (GLN) [7], Drainage density (D) [45,47], Compactness ratio (CR), Maximum basin elevation (E max ), Minimum basin elevation (E min ) and Relative basin relief (RH) [47], Basin Relief (H) [45], Basin centroid elevation (CE) [46], and gaging-station elevation (GE) [1].…”

“…A total of 25 watershed explanatory variables were calculated and used for the development of the peak discharge and Qenv regional models. To the best of the authors' knowledge, among them, the length of the straight line from the source to gauging-station (L), the Total stream length (Tl), the circulatory ratio (Rc), the elongation ratio (Re), and the lemniscate ratio (Rl) [44] have been considered for the first time for such purposes, whereas the remaining 20 characteristics have been used in previous regional flood frequency studies and they include the following: Drainage area (A) [1,7,41,[45][46][47], Main-channel length (Ls) and Main channel slope (S), [1,7,41,46], Basin length (BL)…”

“…The relationships between each dependent variable (Q 25 . Q 50 , Q 100 , and Q env ) with the 14 explanatory variables derived from the PCA analysis (B, L, Ls, BL, A, Tl, Emax, BW, S, E, BS, CE, GE, and Emin) for the four different model types were calculated through application of the stepwise multiple regression analysis, [40,46], which reduces the number of explanatory variables to those significant at the 95% confidence level [35]; Table 5 illustrates the relevant results, where it can be observed that that the linear models and logarithmic models exhibit the worst and the second worst prediction efficiency, respectively, for the peak discharge values prognosis concerning the selected recurrence intervals with low R 2 and high PE, in comparison to the power or exponential models. Although the power and exponential model yielded similar results for the Q 25 forecast in terms of the R 2 , RMSE, NSE, and d statistical measures, the exponential model was found to have a smaller PE value, and it should be selected for the discharge value with 25 years recurrence interval assessment.…”

“…Notably, the four regional equation models' performance marked a high forecasting accuracy for the dependent variables by the explanatory variables, since the R 2 , the NSE, and the d were found to range from 0.91-0.96, 0.88-0.95, and 0.97-0.99, respectively. Mimikou [40] also reported R 2 values that were equal to 0.96 and 0.98 for the maximum observed floodflows for the western and north western region of Greece, respectively, whereas Selvanathan et al [46] developed regression equations that relate peak discharges to basin and climate characteristics with an accuracy reaching an R 2 values equal to 0.86 for certain regions of the U.S. El-Jabi and Caissie [69] developed regional models for mean, median, high, as well as low flows of different recurrence intervals in New Brunswick with high R 2 values ranging between 0.799 and 0.989. On the other hand, Zhang et al [35] developed a regression equation for computing environmental flow with an adjusted R 2 value of 0.96.…”