Characterizing Climate-Change Impacts on the 1.5-yr Flood Flow in Selected Basins across the United States: A Probabilistic Approach

Transcription

1 Paper No. 18 Page 1 Copyright Ó 2011, Paper ; 5106 wors, 5 Figures, 0 Animations, 3 Tables. Characterizing Climate-Change Impacts on the 1.5-yr Floo Flow in Selecte Basins across the Unite States: A Probabilistic Approach John F. Walker* U.S. Geological Survey, Mileton, Wisconsin Lauren E. Hay an Steven L. Markstrom U.S. Geological Survey, Denver, Colorao Michael D. Dettinger U.S. Geological Survey, La Jolla, California Receive 21 August 2010; accepte 12 December 2010 ABSTRACT: The U.S. Geological Survey Precipitation-Runoff Moeling System (PRMS) moel was applie to basins in 14 ifferent hyroclimatic regions to etermine the sensitivity an variability of the freshwater resources of the Unite States in the face of current climate-change projections. Rather than attempting to choose a most likely scenario from the results of the Intergovernmental Panel on Climate Change, an ensemble of climate simulations from five moels uner three emissions scenarios each was use to rive the basin moels. Climate-change scenarios were generate for PRMS by moifying historical precipitation an temperature inputs; mean monthly climate change was erive by calculating changes in mean climates from current to various future * Corresponing author aress: John F. Walker, U.S. Geological Survey, 8505 Research Way, Mileton, WI aress: jfwalker@usgs.gov DOI: /2010EI379.1

2 Paper No. 18 Page 2 ecaes in the ensemble of climate projections. Empirical orthogonal functions (EOFs) were fitte to the PRMS moel output riven by the ensemble of climate projections an provie a basis for ranomly (but representatively) generating realizations of hyrologic response to future climates. For each realization, the 1.5-yr floo was calculate to represent a flow important for seiment transport an channel geomorphology. The empirical probability ensity function (pf) of the 1.5-yr floo was estimate using the results across the realizations for each basin. Of the 14 basins stuie, 9 showe clear temporal shifts in the pfs of the 1.5-yr floo projecte into the twenty-first century. In the western Unite States, where the annual peak ischarges are heavily influence by snowmelt, three basins show at least a 10% increase in the 1.5-yr floo in the twenty-first century; the remaining two basins emonstrate increases in the 1.5-yr floo, but the temporal shifts in the pfs an the percent changes are not as istinct. Four basins in the eastern Rockies/central Unite States show at least a 10% ecrease in the 1.5-yr floo; the remaining two basins emonstrate ecreases in the 1.5-yr floo, but the temporal shifts in the pfs an the percent changes are not as istinct. Two basins in the eastern Unite States show at least a 10% ecrease in the 1.5-yr floo; the remaining basin shows little or no change in the 1.5-yr floo. KEYWORDS: Climate change; Hyrology; Probability ensity function; 1.5-yr floo 1. Introuction General circulation moel (GCM) simulations through 2099 project a wie range of possible future climate changes in response to increasing greenhouse-gas concentrations in the atmosphere (Solomon et al. 2007). To etermine the sensitivity an potential impact of long-term climate change on the freshwater resources of the Unite States, the U.S. Geological Survey (USGS) global change stuy An integrate watershe scale response to global change in selecte basins across the Unite States was unertaken in 2008 an 2009 (Markstrom et al. 2010). The long-term goal of this national stuy is to provie the founation for hyrologically base climate-change stuies across the nation. Fourteen river basins for which Precipitation-Runoff Moeling System (PRMS; Markstrom et al. 2010) moels previously ha been calibrate an evaluate were selecte as stuy sites (Figure 1; Table 1). PRMS is a process-base, istributeparameter watershe moel evelope to evaluate the effects of various combinations of precipitation, temperature, an lan use on streamflow an general basin hyrology. The PRMS moels for this national stuy were evelope, calibrate, an evaluate for previous or current hyrologic investigations. Outputs from five GCMs responing to three greenhouse-gas emission scenarios were input to PRMS moels to simulate an ensemble of hyrologic responses to climate changes for each watershe. The hyrologic impact an sensitivity of the simulations to climatechange scenarios were etermine by comparisons to PRMS simulations of baseline (historical) conitions (Hay et al. 2011). One approach to analyzing results from numerous simulations is to compute mean or meian statistics across the various GCMs an emissions scenarios. Although this may provie a picture of hyrologic response to climate change, it oes not fully represent ranges an uncertainties regaring the future climate-change scenarios an associate hyrologic responses. As an alternative, we apply a

3 Paper No. 18 Page 3 Table 1. General characteristics of each stuy basin. Basin name USGS gauge No. Drainage area (km 2 ) Elev range (m) Western Unite States Feather River, CA Flathea River, MT Naches River, WA Sagehen Creek, CA Sprague River, OR Eastern Rockies an central Unite States Black Earth Creek, WI Clear Creek, IA East River, CO Starkweather Coulee, ND Trout River, WI Yampa River, CO Eastern Unite States Cathance Stream, ME Flint River, GA Pomperaug River, CT methoology for estimating probability istributions of hyrologic responses by resampling the information containe in the ensemble of hyrologic preictions. The resulting istributions provie useful estimates of the probability that various magnitues of hyrologic response will result (in the PRMS moels) from any subset of the climate-change projection ensemble use here to force the watershe moels (Dettinger 2006). For this paper, the changing magnitue of estimates of flow magnitues recurring with 1.5-yr return intervals was chosen to represent changes in flows important for seiment transport an channel geomorphology, with associate implications for stream habitat (Leopol et al. 1964; Dunne an Leopol 1978; Castro an Jackson 2001; Simon et al. 2004). Using the component resampling approach, empirical probability ensity functions (pfs) of the 1.5-yr floo were etermine for the 14 stuy watershes at each of 3 ecaes spanning the twenty-first century. These pfs provie temporally varying probabilistic characterization of climatechange impacts on floo flows at locations across the nation. Following these results, limitations of the analysis are escribe, an conclusions rawn from the exercise are presente. 2. Methos A brief escription of the evelopment of climate-change emission scenarios an processing of the PRMS outputs is given here; a etaile escription of the methos is given in Hay et al. (Hay et al. 2011). Given the uncertainty in climate moeling, it is esirable to use more than one GCM to explore a range of potential future climatic conitions. Monthly precipitation an temperature output from five GCMs constitute the climate-projection ensemble use here. The five GCMs were selecte because they prouce output neee for the PRMS moel (aily precipitation an maximum an minimum aily temperature). The GCM output was

4 Paper No. 18 Page 4 obtaine from the Worl Climate Research Programme s Couple Moel Intercomparison Project phase 3 multimoel ataset archive, which was reference in the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report Special Report on Emission Scenarios (SRES; Solomon et al. 2007). For each GCM, one baseline an three future carbon emission scenarios were analyze. Climate-change emission scenarios were erive by calculating mean change in climate from baseline to future conitions in the simulations from each GCM. The IPCC historical simulation for water years was use to represent the baseline climatic conitions of each GCM. This 12-yr perio of recor was selecte to match the perio of overlap share by the available historic recors from the 14 basins inclue. Mean monthly climate-change estimates (percentage changes in precipitation an egree changes in temperature) were compute for 12-yr moving winow perios (from 2001 to 2099) using the IPCC historical conitions (20C3M; ) an the SRES A2, B1, an A1B scenarios. Climatechange scenario files for input to PRMS were generate by applying these future (mean) changes uniformly to the aily PRMS precipitation an temperature inputs ( ), base on historical observations. The first year of each 12-yr simulation was use as PRMS initialization an is not inclue in the analysis, resulting in 11 years available for analysis. The analysis presente in this paper is for three specific perios, centere aroun the years 2030, 2060, an 2090 ( , , an , respectively), with simulations base on each of the three future emissions scenarios an each of five GCMs, for a total of 45 scenarios analyze in each river basin Component resampling approach The component resampling approach use in this paper (Dettinger 2006) will be briefly escribe here. The PRMS moel was run with the ensemble of 15 climatechange input files for each perio of interest. This resulte in an ensemble of aily moel outputs, which were harveste to etermine the maximum aily ischarge for each year in each simulation. For a particular simulation perio, this results in an m 3 n matrix X containing the ensemble of moel forecasts, representing m years of annual maximum aily ischarges (11 years) an n GCM scenario combinations (15 members). A log transformation of the annual maximum ischarges was use to ensure that ischarges generate by the resampling approach woul be nonnegative. Thus, the term x i,j represents the natural log of the maximum aily ischarge for year i corresponing to moel j. The ensemble of moel forecasts is ecompose into an empirical orthogonal function (EOF) matrix E an a coefficient matrix P such that X 5 EP T, (1) where the superscript T signifies the transpose operation. The EOF matrix E is etermine by a principal component analysis (PCA) of the correlation matrix ZZ T, where Z is a stanarize zero-mean version of the matrix of moel forecasts, with expectations taken across the ensemble of n moel forecasts for each year in the ensemble,

5 Paper No. 18 Page 5 z i, j 5 (x i, j 2 m i ) s i for i 5 1, m an j 5 1, n, (2) where m i 5 1 n X n j51 x i, j an (3) " 1 s i 5 n X n j51 (x i, j 2 m i ) 2 # 1/2. (4) The PCA on the correlation matrix ZZ T is performe, an all eigenvectors are retaine as the EOF matrix. Because the correlation matrix is symmetric, the eigenvectors form an orthogonal set that can be use to ecompose an recreate the original time series (Blyth an Robertson 2002). Retaining all of the eigenvectors allows for a complete recreation of the first two moments of the original ensemble matrix X. The coefficient matrix P is compute by projecting the stanarize matrix Z onto the EOFs; the kth vector p k is thus given as p k 5 Z T e k, (5) where e k is the kth eigenvector from the PCA analysis. For a particular realization r, a column in the stanarize forecast matrix Z is given as z i,r 5 Xm k51 e k i pk j. (6) In this expression, the inex j is ranomly rawn from 1 to n for each k in the summation. Because of the orthogonal nature of the eigenvectors, each eigenvector is inepenent of every other eigenvector; thus, the coefficient vectors are also inepenent of one another an hence can be selecte at ranom without changing the first two moments of the original time series. This provies for a much larger set of ranom permutations an a more realistic representation of the pf of an outcome calculate from the resample time series. The results are then rescale using the ensemble mean an stanar eviations for each time step; thus, x i,r 5 m i 1 s i z i,r. (7) Each resample ischarge is then exponentiate to return the ischarge to natural units. A Monte Carlo experiment was esigne to implement the component resampling approach escribe above. For each basin, three future perios were consiere, centere aroun the years 2030, 2060, an The annual maximum ischarge for each year was extracte from the PRMS output files for these three perios an the 15 GCM emission scenario combinations. This resulte in three ensemble forecast matrices for each basin, which were stanarize following

6 Paper No. 18 Page 6 Table 2. Number of realizations resulting in high outliers ientifie by Bulletin 17B proceures base on total realizations. Realizations with high outliers Basin Western Unite States Feather River, CA Flathea River, MT Naches River, WA Sagehen Creek, CA Sprague River, OR Eastern Rockies/central Unite States Black Earth Creek, WI Clear Creek, IA East River, CO Starkweather Coulee, ND Trout River, WI Yampa River, CO Eastern Unite States Cathance Stream, ME Flint River, GA Pomperaug River, CT Equations (2) (4). A principal component analysis was performe, an the collection of eigenvectors was use to form each EOF matrix. The coefficient vector p was then compute using Equation (5). The result for each basin was a set of three EOF (E) an coefficient (P) matrices. For each set of EOF an coefficient matrices, a set of N real realizations of possible future annual maximum flow series was forme by proviing N real sets of ranom j inices to Equation (6). Each annual maximum flow series was use to estimate the 1.5-yr floo following the proceures escribe in the next section. This resulte in N real estimates of the 1.5-yr floo for each of the three future perios. To etermine an appropriate number of realizations for estimating pfs of 1.5-yr floo flows, N real was increase until various statistics of the 1.5-yr floo approache asymptotic values. In the application here, N real realizations prove to be aequate Floo frequency analysis For each realization, an estimate of the 1.5-yr floo was obtaine by fitting the log-pearson type-3 istribution (LP3) to the generate 11-yr series of annual maximum ischarges using proceures outline in Bulletin 17B (Interagency Avisory Committee on Water Data 1982). Station skewness was weighte with general skewness obtaine from the map in Bulletin 17B using the proceure escribe in the report. Source coe for the USGS software application PeakFQ was obtaine ( an repackage as a subroutine for the program executing the Monte Carlo experiment escribe in the previous section. A sequence of annual maximum ischarges was generate using

7 Paper No. 18 Page 7 Figure 1. Location of stuy basins in the Unite States with shae topography inicating relief across the basins. Site name colors inicate the egree of change in the 1.5-yr floo resulting from climate change: black inicates relatively little change (less than 10% ifference between current conitions an the 2090 projection), re inicates ecreasing 1.5-yr floos, an blue inicates increasing 1.5-yr floos. Note that basins are not to scale. the component resampling approach escribe in the previous section, an a particular sequence was iscare if the application of Bulletin 17B proceures ientifie one or more low outliers. Because the Bulletin 17B proceure rops the ata corresponing to low outliers, it was felt that reucing the sample size below 11 years woul result in an unreasonable estimate of the 1.5-yr floo. For cases

8 Paper No. 18 Page 8 Figure 2. Change in (a) maximum an (b) minimum temperature an (c) percent change in mean aily precipitation (for the five GCMs an three emissions scenarios for the three perios examine in this paper (2030 in green; 2060 in tan; 2090 in blue).

9 Paper No. 18 Page 9 where high outliers were ientifie, they were inclue in the analysis, an a count of the percent of realizations resulting in high outliers was tabulate (Table 2). For all basins, less than 0.5% of the realizations resulte in high outliers, which was consiere to a reasonable percentage of high outliers that were inclue in the analyses. With the exception of Sagehen Creek, the percentages of high outliers generate were extremely small. It is unclear why the EOF for the Sagehen Creek basin results in a larger number of high outliers. However, because the number of outliers oes not exhibit a tren as the climate-change scenarios progress through time, the conclusions concerning the change in the istribution of the 1.5-yr floo over time shoul still be vali Empirical probability ensity function Each of the N real resample flow-series realizations results in its own 1.5-yr floo flow estimate. To estimate the probability ensity function of the 1.5-yr floo, we use a proceure base on a kernel estimate (Parzen 1962), ^f h (x) 5 1 hn real X Nreal i51 x 2 xi K h, (8) where N real is the number of iscrete values comprising the empirical pf, h is the with of the kernel winow, K() is the kernel, an ^f h (x) is the resulting empirical pf for a particular value of x. The Epanechnikov kernel was use, K(u) 5 3 p 1 4 ffiffiffi 2 u2 p for juj ffiffi 5. (9) 5 5 The following kernel winow prouce reasonably smooth pfs: h 5 1:06s x n 1/5, (10) where s x is the sample stanar eviation of the N real 1.5-yr floo preictions. Empirical pf values were thusly estimate for 1000 values of flow, istribute uniformly between the minimum an maximum 1.5-yr floo from the N real scenarios. 3. Results A etaile escription of the temporal changes in the inputs to the PRMS moel (temperature an precipitation) for the five GCMs an three emission scenarios is given in Hay et al. (Hay et al. 2011). A summary of the changes in maximum an minimum aily temperature an percent change in precipitation across the 14 basins is given in Figure 2 for reference. In general, the central tenencies of the three emissions scenarios inicate a 28 68C increase in maximum an minimum temperature, with a positive tren through time. The most extreme increases in

10 Paper No. 18 Page 10 Figure 3. Estimate probability ensity functions for the 1.5-yr floo for basins in the western Unite States: (a) Feather River, California; (b) Flathea River, Montana; (c) Naches River, Washington; () Sagehen Creek, California; an (e) Sprague River, Oregon. Squares represent the meian value. temperature are somewhat higher for maximum temperature compare to minimum temperature. Further, the variability of the changes in temperature increases as the simulations progress through time. The percent changes in precipitation show less istinct patterns, with central tenencies ranging from a ecrease of less than 5% to an increase on the orer of 20%. In general, most of the basins show an increase in precipitation aroun 5%, with one basin showing a ecrease an two basins showing increases between 10% an 15%. For some basins, there is a slight positive tren with time, but for the most part precipitation is roughly stationary across the three future perios. As with temperature, the variability of the changes

11 Paper No. 18 Page 11 in precipitation increase somewhat with time, although it is less pronounce than the increase variability in temperature. The estimate pfs for the western Unite States, eastern Rockies/central Unite States, an the eastern Unite States are presente in Figures 3 5, respectively. The meian values for the 2090 future perio along with the simulate values for the historic perio ( ) are presente in Table 3. Of the 14 basins stuie, 9 showe clear temporal shifts in the pfs of the 1.5-yr floo in response to the ensemble projections of climate change (Figure 1, re an blue basin names). Three basins in the western Unite States show an increase of at least 10% in the 1.5-yr floo, four basins in the eastern Rockies/central Unite States show a ecrease of at least 10% in the 1.5-yr floo, an two basins in the eastern Unite States show a ecrease of at least 10% in the 1.5-yr floo (Table 3). For the remaining five basins without clear temporal shifts in the pfs, three show a slight increase in the 1.5-yr floo an two show a slight ecrease in the 1.5-yr floo. For most of the pfs presente in Figures 3 5, the moe of the istribution changes an the variability increases as the simulations progress through time ( ), inicating that the ensemble range of 1.5-yr floo estimates increases with time. This increasing ensemble range reflects increasing ifferences among the three emission scenarios an between the various GCMs as time progresses, but it may also reflect some increase year-to-year variability later in the projections. Exceptions to this tenency for increasing variability coul likely be ue to interactions between precipitation an temperature effects on the simulation of peak ischarge at those sites. In the western Unite States, three basins show clear temporal shifts in the pfs trening towar an increase in the 1.5-yr floos throughout the simulation perio (Figures 3a,,e). For these basins, the largest annual peak ischarge occurs uring unusual rainfall events in December an January; in years where there is not a winter storm, the annual maximum ischarge typically occurs uring snowmelt in the spring. Because the smaller annual peak ischarges woul more strongly impact the lower-frequency floos such as the 1.5-yr floo, changes in the snowmelt peaks an the frequency of winter storms woul likely impact estimates of the 1.5-yr floo. From the projecte changes in hyrologic buget components, for these basins surface runoff increases an the snowmelt peak occurs earlier (Markstrom et al. 2010), which is likely resulting in higher annual maximum ischarges uring these years. For the two basins without clear temporal shifts in the pfs (Figures 3b,c), snowmelt ominates the annual ischarge hyrograph, an the 1.5-yr floos uner climate-change conitions are slightly larger than the corresponing baseline ( ; Table 3). This is consistent with the fining for the other basins in this region. In the eastern Rockies/central Unite States, five basins show clear temporal shifts in the pfs trening towar a ecrease in the 1.5-yr floos throughout the simulation perio (Figures 4a,f). For all of these basins, the projecte changes in hyrologic buget components inicate an increase in infiltration an evapotranspiration with a corresponing ecrease in soil moisture an a ecrease in surface runoff (Markstrom et al. 2010), which is consistent with a ecrease in the annual maximum ischarge. The projecte changes in hyrologic buget components also inicate a ecrease in snowmelt runoff, which woul also contribute to ecreasing annual peak ischarges. For the basin without a clear temporal shift in the pfs (Figure 4e), the trens in the hyrologic buget components are generally weaker compare to other basins (Markstrom et al. 2010); however, this basin

12 Paper No. 18 Page 12 Figure 4. Estimate probability ensity functions for the 1.5-yr floo for basins in the eastern Rockies/central Unite States: (a) Black Earth Creek, Wisconsin; (b) Clear Creek, Iowa; (c) East River, Colorao; () Starkweather Coulee, North Dakota; (e) Trout River, Wisconsin; an (f) Yampa River, Colorao. Squares represent the meian value. exhibits a slight ecrease in the 1.5-yr floo compare to the corresponing baseline ( ; Table 3), which is consistent with the other basins in this region. In the eastern Unite States, one basin shows a clear temporal shift in the pfs trening towar a ecrease in the 1.5-yr floos throughout the simulation perio (Figure 5b). For this basin, the projecte changes in hyrologic buget components inicate an increase in infiltration an evapotranspiration with a corresponing ecrease in soil moisture an a ecrease in surface runoff (Markstrom et al. 2010),

13 Paper No. 18 Page 13 Figure 5. Estimate probability ensity functions for the 1.5-yr floo for basins in the eastern Unite States: (a) Cathance Stream, Maine; (b) Flint River, Georgia; an (c) Pomperaug River, Connecticut. Squares represent the meian value. which is consistent with a ecrease in the annual maximum ischarge. For the remaining basins without a clear temporal shift in the pfs (Figures 5a,c), the trens in the hyrologic buget components are generally weaker compare to other basins (Markstrom et al. 2010). For these basins, one showe a slight increase (Figure 5a) an one showe a ecrease (Figure 5c) in the 1.5-yr floo compare to the corresponing baseline ( ; Table 3). This is likely ue to the balance between slightly increase precipitation an reuce soil moisture an surface runoff for these basins (Markstrom et al. 2010). 4. Limitations The floo frequency analysis escribe in Bulletin 17B is normally applie to the recor of annual maximum instantaneous ischarge. For many streams, the annual maximum instantaneous peak is consierably larger than the aily maximum ischarge for a particular year. Because the PRMS moels use in this stuy are base on aily ata, the peaks use in the analysis were base on the largest aily ischarge for each year. The ifferences between instantaneous an aily ischarges are generally small for large streams, which typically have floo hyrographs that span many ays. However, for small streams the ifferences between the maximum instantaneous an aily peaks can be consierable. For the 14 stuy

14 Paper No. 18 Page 14 Table 3. Comparison of meian values for the 11-yr perio centere aroun 2090 relative to conitions for the stuy basins. Moele value for the 1.5-yr floo (ft 3 s 21 ) Basin meian Western Unite States Feather River, CA Flathea River, MT Naches River, WA Sagehen Creek, CA Sprague River, OR Eastern Rockies/central Unite States Black Earth Creek, WI Clear Creek, IA East River, CO Starkweather Coulee, ND Trout River, WI Yampa River, CO Eastern Unite States Cathance Stream, ME Flint River, GA Pomperaug River, CT basins liste in Table 1, the ifference between instantaneous an maximum aily flows for the perio was less than 10% for the 7 basins with an area greater than 700 km 2 an for the grounwater-ominate Trout River basin. The ifference between instantaneous an maximum aily flows for the two basins with areas between 200 an 700 km 2 an Sagehen Creek was less than 50%. For the remaining three basins, the ifference between instantaneous an maximum aily flows was greater than 50%. Base on the ownscaling proceure use, the future climate recors represent climate average over a 12-yr winow; the temporal patterns of precipitation are essentially the same as the patterns containe in the historic perio ( ). Hyrologic preictions from this average climate will result in reasonable estimates of annual an monthly average ischarges. For floos, the annual peaks are typically generate by extreme conitions that are not represente by the ownscale ata. It shoul be note that the component resampling methos set forth in this paper woul be vali for alternative ownscale climate projections an results from other hyrologic moels. Further, the aitional information provie by estimating the pf of a given hyrologic moel response enhances the utility of the future projections for resource managers. Finally, the ensemble of climate-change scenarios use to rive the PRMS moels represents a relatively small sampling of variations an uncertainties regaring how the five GCMs respon to emissions scenarios, ifferences among the three emissions scenarios use to rive each of the GCMs, the ownscaling approach use, an variations ue to the whole range of simulate climate processes from less-than-hourly to century scales. The ensemble is limite, however, an oes not span the full range of possible future climates (incluing long-term natural influences), the full range of possible GCM configurations an errors, or the full

15 Paper No. 18 Page 15 range of possible emissions. The ensemble also is not weighte accoring to which moels might be more accurate (however, see Pierce et al. 2009) or accoring to which emissions trajectories might be more likely. Thus, as note previously, the pfs of 1.5-yr flows evelope here represent the variations an uncertainties as capture by the ensemble of climate-change scenarios use to force the PRMS moels; they o not represent completely all of the uncertainties surrouning future climate changes. Thus, the results presente herein can be consiere a heuristic exercise examining the potential impacts of climate change on the 1.5-yr floo. 5. Conclusions Of the 14 basins stuie, 9 showe istinct temporal shifts in the pfs of the 1.5-yr floo projecte into the future using results from five GCM projections for three emissions scenarios. Three snowmelt-ominate basins in the western Unite States show at least a 10% increase in the 1.5-yr floo in the twenty-first century. The other two snowmelt-ominate basins in the western Unite States also inicate increases in the 1.5-yr floos, but the temporal shifts in the pfs an the percent changes are not as istinct. Four basins in the eastern Rockies/central Unite States show at least a 10% ecrease in the 1.5-yr floo. Two basins in the eastern Rockies/central Unite States inicate a ecrease in the 1.5-yr floo, but the temporal shifts in the pfs an the percent changes are not as istinct. Two basins in the eastern Unite States show at least a 10% ecrease in the 1.5-yr floo. For other basins in the eastern Unite States, the 1.5-yr floo shows little or no change in the twenty-first century. The results presente in this paper emonstrate a technique for estimating the probability ensity function of hyrologic change resulting from climate change. The estimate pf provies resource managers with vital information on the variability of future projections. Because of limitations impose by the use of previously evelope hyrologic moels an the ownscaling technique use, the results presente here shoul be consiere a heuristic exercise, showing potential results an techniques that can be use to provie valuable information to resource managers struggling to plan for the future in the face of climate change. Acknowlegments. This work was supporte by the U.S. Geological Survey through the Global Change Research an Development Program. Ken Potter an Faith Fitzpatrick provie thoughtful reviews that greatly improve this manuscript. References Blyth, T. S., an E. F. Robertson, 2002: Basic Linear Algebra. Springer, 232 pp. Castro, J. M., an P. L. Jackson, 2001: Bankfull ischarge recurrence intervals an regional hyraulic geometry relationships. J. Amer. Water Resour. Assoc., 37, Dettinger, M. D., 2006: A component-resampling approach for estimating probability istributions from small forecast ensembles. Climatic Change, 76, Dunne, T., an L. B. Leopol, 1978: Water in Environmental Planning. W.H. Freeman, 818 pp. Hay, L. E., S. L. Markstrom, R. S. Regan, an R. L. Viger, 2011: Integrate watershe-scale response to climate change through the twenty-first century for selecte basins across the Unite States. Earth Interactions, in press.

16 Paper No. 18 Page 16 Interagency Avisory Committee on Water Data, 1982: Guielines for etermining floo flow frequency. Hyrology Subcommittee Bulletin 17B, 194 pp. Leopol, L. B., M. G. Wolman, an J. P. Miler, 1964: Fluvial Processes in Geomorphology. W.H. Freeman, 522 pp. Markstrom, S. L., an Coauthors, 2011: Integrate watershe scale response to climate change for selecte basins across the Unite States. U.S. Geological Survey Scientific Investigations Rep Parzen, E., 1962: On estimation of a probability ensity function an moe. Ann. Math. Stat., 33, Pierce, D. W., T. P. Barnett, B. Santer, an P. J. Gleckler, 2009: Selecting global climate moels for regional climate change stuies. Proc. Natl. Aca. Sci. USA, 106, Simon, A., W. Dickerson, an A. Heins, 2004: Suspene-seiment transport rates at the 1.5-year recurrence interval for ecoregions of the Unite States: Transport conitions at the bankfull an effective ischarge? Geomorphology, 58, Solomon, S., D. Qin, M. Manning, M. Marquis, K. Averyt, M. M. B. Tignor, H. L. Miller Jr., an Z. Chen, Es., 2007: Climate Change 2007: The Physical Science Basis. Cambrige University Press, 996 pp. Earth Interactions is publishe jointly by the American Meteorological Society, the American Geophysical Union, an the Association of American Geographers. Permission to use figures, tables, an brief excerpts from this journal in scientific an eucational works is hereby grante provie that the source is acknowlege. Any use of material in this journal that is etermine to be fair use uner Section 107 or that satisfies the conitions specifie in Section 108 of the U.S. Copyright Law (17 USC, as revise by P.IL ) oes not require the publishers permission. For permission for any other from of copying, contact one of the copublishing societies.