|
<br />JARRETT AND TOMLINSON: REGIONAL INTERDISCIPLINARY PALEOFLOOD METHOD
<br />
<br />little bucket data have been collected in recent years, severa)-,;
<br />extreme rainstorms have been documented in northwestern
<br />Colorado since the early 1900s.
<br />
<br />4.5. Streamflow Data
<br />
<br />Streamflow data for 218 sites in the Yampa River and White
<br />River basins were compiled and used to assess extreme flood-
<br />ing in the region. U.S. Geoiogical Survey and Colorado State
<br />Engineer streamflow records through 1998 are available from
<br />the U.S. Geological Survey's NWIS-W Data Retrieval system
<br />(http://waterdata.usgs.gov). These data were then used to de.
<br />velop an envelope curve of peak discharge versus drainage
<br />area for northwestern Colorado. Unit discharge can be used to
<br />infer both maximum rainfall intensities and spatial extent of
<br />rainstorms [Jamtt, 1990b]; the data were used to develop an
<br />envelope curve for unit discharge versus elevation.
<br />
<br />4.6. Flood-Frequency Relations
<br />
<br />To help facilitate risk assessments of rare floods for dam
<br />safety officials and floodplain managers, flood-frequency rela-
<br />tions were developed from an analysis of annual peak flows
<br />through 1998 for selected streamflow-gaging stations in north-
<br />western Colorado where paleoflood data are available. A va-
<br />riety of distribution functions and estimation methods are
<br />available for estimating a flood-frequeocy distribution [NRC,
<br />1999]. Flood-frequency relations normally are developed using
<br />a log-Pearson type III frequency distribution [Interagency Ad-
<br />visory Committee on Waler Data, 1981] referred to as Bulletin
<br />17B (BI7B) and the expected moments algorithm (EMA)
<br />[Cohn el aI., 1997; England, 1998J. B17B guidelines were es-
<br />tablished to provide consistency in federal flood-risk manage-
<br />ment for handling low and high outliers, for recognizing the
<br />need for regionaiized skew, and for zero-flow adjustment, for
<br />example.
<br />EMA is an efficient approach for incorporating historical
<br />and paleoflood data and uses the iog-Pearson III distribution
<br />[Cohn el aI., 1997; NRC, 1999]. The NRC [1999] recognized the
<br />need to follow the spirit of the guideiines such as when using
<br />EMA. The EMA is used as the generalization of the conven-
<br />tionallog.space method of moments and makes more effective
<br />use of historical and paleoflood data in a censored-data frame-
<br />work [England, 1998; NRC, 1999]. EMA explicitly iocorpora1es
<br />the number of known and unknown discharges above and
<br />below a threshold, number of years in the historic/paleoflood
<br />period, and knowiedge of the number of years wben no large
<br />floods have occurred [Cohn el al., 1997; England, 1998J. The
<br />difference between BI7B and EMA is the treatment of historic
<br />and paleoflood data [England, 1998]. For B17B the gage record
<br />is used to fill in the censored (unknown) floods, whereas EMA
<br />computes the expectations for flow data below the threshold
<br />and weights this value by the number of censored values [En.
<br />gland, 1998]. A comprehensive review of the EMA, censoring
<br />thresholds and analyses of contemporary and paleoflood data
<br />is provided by England [1998].
<br />For this study, several alternative flood-frequency distribu-
<br />tion estimates were determined for the study sites in north-
<br />western Colorado using EMA to better use the long paleoflood
<br />records. This analysis was done for various combinations of
<br />gage and paleoflood data available at each site. Low outliers
<br />were adjusted using the B17B procedure to externally elimi-
<br />nate low outliers that affected the fit of the upper end of the
<br />CUlVe to the data, which is similar to discharge threshold cen-
<br />soring [Cohn el al., 1997; Levish el al., 1994; NRC, 1999].
<br />
<br />2967
<br />
<br />Censoring beiow a threshold can account for an assumed dis-
<br />tribution not fitting the "true" distribution at a site [NRC,
<br />1999J. Low outliers in Colorado often result from modest
<br />streamflow diversions for irrigation of hay meadows during
<br />low-flow years, but in normal years these diversions have min-
<br />imai (<5%) effect on peak flows [Jamll, 1987]. Historicai and
<br />paleoflood data also are censored sampies because only the
<br />largest floods are recorded. Paleoflood data (magnitude and
<br />ages) were incorporated into the flood-frequency analysis to
<br />extend the gaged record. In the EMA analysis the paleoflood
<br />discharge was specified as a range, and EMA runs were made
<br />for 1he range in age (Table 2) for a site.
<br />
<br />5. Results
<br />
<br />5.1. Paleoflood Investigations
<br />
<br />Paieoflobd data from on-site studies of 88 sites throughout
<br />the study area are provided in Table 2. Although not all trib-
<br />utary streams in the northwestern Colorado were documented
<br />(because of inaccessibility to private property), sites were se.
<br />iected such that paleoflood data were collected downstream
<br />from tributaries on the main stream or similar to "nested"
<br />sites. For each site, drainage area, elevation, channel slope,
<br />type of PSI (either flood bar (FB) or noninundation surface
<br />(NI), width, depth, veiocity, flood discharge corresponding to
<br />the maximum PSI (e.g., Figure 6), and sediment-size data,
<br />where available, are presented. Site selection was dependent
<br />on finding good PSIs (FB and NI) throughout a 50 to 300 m
<br />length (length dependent on size of channel). Ideally, each site
<br />would have a FE and a NI surface. Sites were primarily in
<br />straight, uniform reaches where local aggradation or degrada-
<br />tion would be least. General scour appears to have been small
<br />during the Holocene. According to Madnle [1991aJ, there has
<br />been about a meter of degradation along most the Yampa
<br />River during the Hoiocene. Paleoflood evidence along the
<br />upper Yampa River primarily was midchannei bars. These bars
<br />(islands) were interpreted as erosional remnants of the late
<br />Pleistocene channel rather than depositional features, and thus
<br />maximum paleofloods using present channel geometry may be
<br />slightly overestimated. Some streams have had littie bed ma-
<br />terial movement (Table 2), and only NI evidence exists, but if
<br />the NI surface is consistent. in the reach of river, then the
<br />maximum paleoflood discharge was considered reliable. If a
<br />reach of the same channel had similar paleoflood discharges at
<br />several sites, the discharge estimate was considered more re-
<br />liable. For alluvial channels the discharge was considered less
<br />reliable and reflected in an assigned sbort 8ge of 100 years,
<br />which is when much of the channel arroyos deveioped in Col-
<br />orado [Patton and Schumm, 1975, 1981; Womack and Schumm,
<br />1977].
<br />RD data using the criteria shown in Table I also are sum-
<br />marized in Table 2. For each site the weathering characteristics
<br />for each of the RD methods used, their numerical rating,
<br />estimated age, and reliability (range) of age estimates, are
<br />provided. Assigning a numerical rating, an age, and range in
<br />age for each RD method used is subjective; however, if all
<br />methods used suggest similar numerical ratings, then the com-
<br />posite age estimate is iikely more reliable. Although individual
<br />RD ages are rather crude and may provide different relative
<br />ages of a surface, a composite relative age using several meth-
<br />ods clearly enables one to distinguish deposits of various ages.
<br />Although the error of individual RD ages can be :!:50% [Bir-
<br />keland, 1990], composite age is likely more accurate. For use
<br />
<br />
|