WATER RESOURCES RESEARCH, VOL. 36, NO. 10, PAGES 2957-2984, OCTOBER 2000 Regional interdisciplinary paleofiood approach to assess extreme flood potential Robert D. Jarrett U.S. GeologicalSurvey, Denver, Colorado Edward M. Tomlinson Applied Weather Associates,Monument, Colorado Abstract. In the past decade,there has been a growinginterest of dam safetyofficials to incorporatea risk-basedanalysis for design-floodhydrology. Extreme or rare floods,with probabilitiesin therange of about10 -3 to 10-7 chanceof occurrenceper year, are of continuinginterest to the hydrologicand engineeringcommunities for purposesof planningand designof structuressuch as dams [NationalResearch Council, 1988]. The National ResearchCouncil stressesthat as much information as possibleabout floods needsto be usedfor evaluationof the risk and consequencesof any decision.A regional interdisciplinarypaleoflood approach was developedto assistdam safetyofficials and floodplainmanagers in their assessmentsof the risk of large floods.The interdisciplinary componentsincluded documenting maximum paleofloods and a regionalanalyses of contemporaryextreme rainfall and flood data to complementa site-specificprobable maximumprecipitation study [Tomlinson and Solak, 1997]. The cost-effectiveapproach, which can be usedin many other hydrometeorologicsettings, was applied to Elkhead Reservoirin ElkheadCreek (531 km 2) in northwesternColorado; the regional study area Was10,900 km 2. Paleoflood data using bouldery flood deposits and noninundation surfaces for 88 streamswere used to documentmaximum flood dischargesthat have occurred during the Holocene.Several relative datingmethods were usedto determinethe age of paleoflooddeposits and noninundationsurfaces. No evidenceof substantialflooding was foundin thestudy area. The maximum paleoflood of 135m 3 s-1 for ElkheadCreek is about13% of thesite-specific probable maximum flood of 1020m 3 s-1. Flood-frequency relationsusing the expectedmoments algorithm, which better incorporatespaleoflood data, were developedto assessthe risk of extremefloods. Envelope curves encompassing maximumrainfall (181 sites)and floods(218 sites)were developedfor northwestern Coloradoto help define maximumcontemporary and Holoceneflooding in Elkhead Creek and in a regionalfrequency context. Study results for Elkhead Reservoirwere acceptedby the ColoradoState Engineerfor dam safetycertification. 1. Introduction tion of structures such as dams has included an estimate of the probablemaximum flood (PMF) [Cudworth,1989]. The PMF Worldwide,floods are amongthe mostdestructive events is an estimate of the maximum flood potential for a given related to meteorologicalprocesses. In the United Statesthe drainagebasin and is derivedfrom an analysisof the probable averageannual death toll of 125is accompaniedby about$2.4 maximum precipitation (PMP) [Cudworth,1989]. Prior to billion in damagesfrom floods [FederalEmergency Manage- about 1950, a variety of methods were used to obtain the ment Agency(FEMA), 1997]. Poor understandingof floods magnitudeof the designflood. The NRC [1988,1994] recog- contributesto unnecessaryloss of life and increasedflood nized (1) the limited hydroclimaticdata availableto estimate damagein somecases and, conversely,leads to costlyoverde- PMP/PMF values for mountain basins less than about 1050 sign of hydraulic structureslocated in floodplainsfor other km2, (2) thesubjectivity and variation of PMPestimates among situations[Jarrett, 1991, 1993; Baker, 1994]. Extreme or rare experiencedmeteorologists, (3) the critical need for regional floods,with probabilities in the rangeof about10 -3 to 10-7 chanceof occurrenceper year, are of continuinginterest to the analysesof extremeprecipitation and flooding, (4) theneed to use historicand paleoflooddata, and (5) the potentialuse of hydrologicand engineeringcommunities for purposesof plan- probability-basedmethods for providingan alternativeto the ning and designof structuressuch as dams[National Research PMP/PMF approach.In an evaluationof extreme floods and Council(NRC), 1988].However, estimating the magnitude and frequencyof extreme floods is difficult becauseof relatively the useof the PMF methodologyto estimatedesign floods, the short streamflow-gagingstation record lengths. InteragencyAdvisory Committee on WaterData [1986] raiseda For about the past 50 yearsthe designcriteria for construc- major concernabout PMFs being either dangerouslysmall or wastefullylarge, and they emphasizedthe importanceof accu- Copyright2000 by the AmericanGeophysical Union. rately estimatingthe risk of extremeflooding. Thus the objec- Paper number 2000WR900098. tive of a flood studyshould be to generateas much information 0043-1397/00/2000WR900098509.00 as practicableabout the flood potential at a site [NRC, 1988]. 2957 2958 JARRETT AND TOMLINSON: REGIONAL INTERDISCIPLINARY PALEOFLOOD METHOD The NRC stresses that this should be the basis for evaluation haveoccurred in the regionsince the generalizedhydrometeo- of the risk and consequencesof any decision. rology report was published.Site-specific hydrometeorologic Efforts have been made to assigna frequencyto the PMF, studiesare being conductedbecause dam safetyofficials rec- but they are subjective[Interagency Committee on WaterData, ognizethe difficultproblems inherent in PMP estimatesin the 1986;NRC, 1988]. There is disagreementabout the range of Rocky Mountains.Utilization of an interdisciplinaryregional frequencyfor the PMF and whether any frequencyshould be paleofloodstudy providesadditional supportinginformation assignedto a deterministicPMF estimate.Extensions of flood- for understandingthe magnitudeof the largestcontemporary frequencyrelations to rare floods(e.g., 10,000-yearrecurrence floodsand paleofloodswith estimatesof the PMF potentialfor interval) are tenuousfor short streamflowrecords but are a particularbasin. enhanced when paleoflood data are included in flood- frequencyanalysis [Kochel and Baker,1982; NRC, 1988;Jarrett, 1987; Jarrett and Costa, 1988; Levish et al., 1994; Ostenaaand 2. Background Levish,1995]. Jarrett and Costa[1988] proposed using regional Substantialuncertainty and controversyexists in estimating flood-frequencyestimates, w,hich incorporated paleoflood in- flood magnitudesand frequencies,particularly those of ex- formation, to estimate recurrence intervals for rare floods in- treme floods in the Rocky Mountains. This results from a cludingPMF values.These extensions provide an approachto misunderstandingof the complexhydrometeorological pro- place PMF estimatesin perspectivewith regionalgaged and cesses involved and a lack of data on extreme rainstorms and p•aleofloodestimates of flood potential [Jarrettand Costa, flooding.Jarrett [1993] made a systematicevaluation of flood- 881. frequencyestimates for 25 long-term streamflow-gagingsta- Paleofioodhydrology is the studyof past or ancientfloods tions throughoutColorado where publishedFederal Emer- [Baker,1987]. Kochel and Baker [1982],Gregory [1983],Baker et gencyManagement Agency (FEMA) floodplainreports, which al. [1988], Costa[1987c], Stedinger and Baker [1987],Stedinger were derivedby variousmethods, also were available.On av- and Cohn [1986],Hupp [1988], andJarrett [1987, 1990b,1991] erage,the publishedFEMA 100-yearflood is 35% greaterthan provide summariesof paleofioodhydrology. Although most the gaged100-year flood obtainedby fitting a log-Pearsontype studiesinvolve prehistoric floods, the methodologyis applica- III distribution to streamflow data for the 25 stations. Similar ble to historic or modern floods [Baker et al., 1988; Jarrett, differenceswere identifiedfor the 10-, 50-, and 500-yearfloods 1990b].Paleofiood studies provide important information that and demonstratethe needto improveflood-frequency estimates can be used in risk assessments and in the assessment of cli- usedfor floodplainmanagement and other usesin Colorado. matic changeon floodingand droughts[Jarrett, 1991]. Paleo- Interdisciplinaryflood researchin Colorado [Jarrettand flood data are particularlyuseful in providingupper limits of Costa, 1983, 1988; Jarrett, 1987, 1990a, b; Grimm, 1993; Pitlick, the largestfloods that have occurredin a river basinin long 1994; Waythomasand Jarrett, 1994; Pruess, 1996; Capesius, time spans[Jarrett, 1990b; Enzel et al., 1993]. 1996;Pruess et al., 1998;McKee and Doesken,1997] and in the A regionalinterdisciplinary paleoflood study was conducted Rocky Mountains [Hertz, 1991; Jarrett, 1993; Jensen,1995; in northwesternColorado (Figure 1) to help assessthe flood Buckley,1995; Eastwood, 1995; Brien, 1996; Parrerr, 1997, 1998] hydrologyfor Elkhead Reservoirin Elkhead Creek basinnear providesnew insight into the hydrometeorologyof extreme Craig. The objectiveof the paleofloodstudy was to estimate flooding in the Rocky Mountains. In an analysisof USGS prior maximumflooding during the Holocene from evidence streamflow-gagingstation and paleoflooddata in the Rocky preservedin the floodplain.The interdisciplinarycomponents Mountain region,which included stations in New Mexico,Col- included documentingmaximum paleofloodsand regional orado, Wyoming, Idaho, and Montana, envelope curvesof analysesof contemporary(---155 years in Colorado)extreme maximumunit discharges(maximum peak flow dividedby rainfall and flood data in and near Elkhead basin.The major drainagearea) and gageelevation identified elevation limits drainageswithin the regionalstudy area are the Yampa River for largeunit discharges[Jarrett, 1993]. In Coloradothe eleva- and White River basins.Hydroclimatic conditions