Canadian RCM projected changes to high flows for Québec watersheds using regional frequency analysis

Abstract

Information related to changes in streamflow characteristics is important in the management and future planning of various water resources-related projects in the context of a changing climate. In this study, projected changes to selected return levels of high flows for 21 watersheds, located mainly in the Québec province of Canada, are assessed following regional frequency analysis (RFA). This assessment is based on a ten-member ensemble of Canadian regional climate model (CRCM) transient climate change simulations, of which five correspond to current 1970-1999 period and the remaining five are the matching simulations for the future 2041-2070 period. This is the first comprehensive study on high-flow characteristics for the entire 21 watersheds in the context of a changing climate using the RFA approach, which allows better estimation of return levels, particularly for higher return periods, than single site/grid-cell based analysis. Validation of the European Centre for Medium-Range Weather Forecasts Re-Analysis-driven CRCM simulated 5-, 10-, 30-and 50-year return levels against those observed at selected gauging stations suggests a general overestimation, with average differences of 9 and 12% (3-5%) for the 5-and 10-year (30-and 50-year) return levels, respectively. Results of ensemble averaged projected changes to regional return levels show increases for most of the watersheds. In particular, changes to 5-and 10-year (compared to 30-and 50-year) regional return levels are found to be statistically significant more often for northern watersheds compared to the rest. An evaluation of confidence in projected changes to selected regional return levels, based on five current/future period simulation pairs, reveal a higher level of confidence for northern watersheds. When regional level projected changes are developed at the CRCM grid-cell level, a pattern similar to that for regional level changes is noticed, though with slightly increased range of changes.

Hydrologic extremes – an intercomparison of multiple gridded statistical downscaling methods

“…Approaches to converting coarsescale GCM simulations to project changes to peak flows and low flows vary. Some examples include direct downscaling of streamflow extremes by sparse Bayesian learning and multiple linear regression (Joshi et al, 2013), weather generators combined with hydrologic models (Cunderlik and Simonovic, 2007), regional frequency analysis of regional climate model (RCM) projections (Clavet-Gaumont et al, 2013), and, most commonly, statistical downscaling of GCM or RCM projections run through a physically based hydrologic model (Elsner et al, 2010a;Maurer et al, 2010;Shrestha et al, 2012;Bürger et al, 2011). The uncertainty in hydrologic projections from GCMs is greater than that from emissions scenarios or model parameterizations (Bennett et al, 2012;Prudhomme and Davies, 2008) and all GCMs represent the climate imperfectly in different ways (Gleckler et al, 2008;Knutti et al, 2008); therefore, to fully characterize the uncertainty in projected hydrological extremes, an ensemble of GCMs is required.…”

A copula-based multivariate analysis of Canadian RCM projected changes to flood characteristics for northeastern Canada

“…As for precipitation, the global average is projected to increase with increasing water holding capacity of the atmosphere in a warmer climate (Meehl et al 2007). For North America, Christensen et al (2007) reported projected decreases in future snow season length and snow depth, but increases in future precipitation in winter and spring for southern Canada. Increase in temperature and precipitation can significantly affect flood dynamics in Canadian river basins where high flows are primarily generated due to spring snowmelt (Mareuil et al 2007).…”

“…Based on projections from 21 global climate models that participated in the AR4 (IPCC 2007), Christensen et al (2007) predicted that the annual mean temperature will increase by 3.6°C (with a range of 2.3-5.6°C) and precipitation by 7 % (with a range of -3 to 15 %), over eastern North America including middle and southern parts of Québec, for the 2080-2099 period with respect to the 1980-1999 period; these results consider the IPCC's (2001) Special Report on Emissions Scenarios (SRES) AlB scenario. The largest increase in mean temperature (3.8°C) and precipitation (11 %) is expected in winter, while the smallest increase in mean temperature (3.3°C) and precipitation (1 %) is expected in summer.…”

Adaptation of the L-Moments Method for the Regionalization for Maximum Annual Temperatures in Colombia

“…The approach of RFA is to use the attainable data from gauged sites within a homogeneous region, to compute quantiles of several return periods in ungauged sites [1]- [3]. A regional approach to frequency analysis in poorly gauged areas makes possible a significant reduction of uncertainty associated with the assessment of hydrological events with a high return period [2], [4]- [8].…”

“…The most important step in the RFA procedure is the selection/delineation of homogeneous regions [4], in other words, the process of gathering the selected stations-which have enough record length and other suitable propertiesinto groups that obey the hypothesis of statistical homogeneity. Whenever the grouped stations fulfill the assumption of statistical homogeneity, then RFA will take advantage of all the available information and will make reliable assessments of extreme quantiles using the fitted regional pdf [7], [8].…”