Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 11
Estimating Bounds on Extreme Precipitation Events: A Brief Assessment Discussion A very conservative version of PMP could be created through the use of Figure 1. For every point on the Earth's surface, we could assume that in any given amount of time the greatest precipitation would be bounded by the greatest precipitation accumulation observed anywhere in the world for that duration. However, designing all structures to survive such conditions would be prohibitively expensive. In practice, we know that the greatest precipitation that can be expected at high elevations or high latitudes is much less than what occurs in most tropical coastal regions. Other local processes also limit or enhance precipitation. DETERMINATIONS OF PMP The Office of Hydrology of the National Weather Service (NWS) and its predecessors in the U.S. Weather Bureau (USWB) have carried out assessments of PMP for nearly 50 years. These serve as a standard for engineering design of high-hazard structures in the United States. The methodologies currently used by the NWS have been documented in a series of technical reports. Major elements of NWS analyses of PMP were established in 1956 with the report Seasonal Variation of PMP East of the 105th Meridian (USWB 1956, revised in 1978). The standard PMP products from the NWS are “generalized” maps of PMP estimates for the United States, which allow variation only on a coarse scale. The NWS has updated its PMP estimates periodically.
OCR for page 12
Estimating Bounds on Extreme Precipitation Events: A Brief Assessment In the standard approach to PMP outlined by the World Meteorological Organization, observed precipitation accumulations for extreme storms are used as indicators of maximum values of convergence and precipitation efficiency. Implicit in standard PMP procedures are a number of assumptions concerning the physical processes that control extreme storms. These assumptions are imbedded in two major components of the standard approach to PMP computation: storm transposition and moisture maximization. Both of these procedures call for judgment; competent professionals can come to different conclusions. The storm transposition step is based on the notion that an extreme event similar to one under consideration could have happened at a different location in a region judged to be meteorologically similar. Determination of the geographic limits for the transposition of a particular storm is difficult and subject to debate. Transposition is generally inappropriate to regions with greatly different orography or greatly different typical moisture levels. Some of the most difficult problems that have arisen with PMP assessments center around storm transposition and the implicit transposition that can result from the objective analysis step in which PMP estimates are mapped onto a national grid and contour lines are smoothed. Storm transposition is a regionalization procedure analogous to those used in a wide range of hazard assessment problems (see NRC 1988 for a detailed discussion of estimating probabilities of extreme floods). PMP estimates are often strongly controlled, through the storm transposition step, by the few largest observed rainstorms. The dramatic influence of the Smethport, Pennsylvania, storm of 1942 on PMP estimates for the northeastern United States is one prominent example. A second example is the influence of rainfall from Hurricane Camille in 1969 on PMP estimates throughout much of the southeastern United States. For both storms; orographic features have been cited as important elements in determining the location of extreme rainfall (e.g., Schwarz 1970). Nevertheless, these storms have been used to determine PMP in areas with greatly different terrain. In moisture maximization the goal is to increase the storm total amounts to reflect the maximum possible moisture availability. For a storm transposed, or even at the location of the actual event, greater humidities are often observed than was the case during an observed storm. Moisture maximization allows for the possibility that a storm of similar structure could have occurred during the same season but during a time of greater absolute humidity. This procedure increases the rainfall of the observed storm by the ratio of maximum observed moisture in the location under consideration to the observed moisture that flowed into the storm being examined. The calculations are sensitive to measurements of dewpoint for both the storm inflow and the location where PMP
OCR for page 13
Estimating Bounds on Extreme Precipitation Events: A Brief Assessment is being estimated. The validity of the assumptions that underlie moisture maximization is questionable in many cases. For example, precipitation efficiency may vary nonlinearly with available moisture, in contrast with moisture maximization assumptions. Surface dewpoint temperature may also be a poor indicator of precipitable water for some events. Notable controversy has also surrounded PMP estimates for mountainous regions of the western United States. Transposition of the 31 July 1976 Big Thompson storm in Colorado, which killed 139 people, has drawn attention to a wide range of issues. The use of paleohydrologic data in the Rocky Mountain region (e.g., Jarrett 1989) has played a major role in the PMP debate. Paleohydrologic analyses have found no stratigraphic evidence for large floods in high-elevation regions of the Rocky Mountains. Orographic adjustment procedures for PMP have been proposed for mountainous regions (see WMO 1986); however, physical justification for these procedures has not been established clearly. It is important to distinguish site-specific from national-scale generalized PMP estimates. The procedures used for constructing the two are not necessarily the same. The NWS has developed generalized map estimates for the continental United States. Site-specific PMP estimates are carried out for specific drainage basins or regions. One distinction between the two types of PMP estimates is that generalized PMP estimates entail an additional smoothing step in the objective analysis procedure used to map local PMP estimates onto national (or regional) maps. Generalized estimates of PMP are determined by the higher values in a region to ensure that they are sufficient for all points in a region, even though at specific points within the region topographic features would lead to smaller PMP values from a site-specific analysis. Furthermore, there have not been sufficient resources, nor are there sufficient data, to perform high-resolution (i.e., site-specific) analyses for the entire country. Site-specific and regional PMP analyses have been conducted by nonfederal entities, ranging from state governments to private consulting firms working for industry consortiums, as well as by the NWS. One of the best-known cases of disagreement on PMP calculations was for the Deerfield River drainage behind the Harriman dam in New England. For a period of 24 hours and a 200-square-mile basin size, estimates ranged from a low of 14.3 inches by the utility company up to 23.2 inches by the NWS; a board assembled by FERC finally settled on 17.3 inches. Much of the argument centered around the appropriateness of transposing a single storm (the Smethport, Pennsylvania, 1942 storm mentioned in the Introduction) and the effects of terrain on storm development. A study of PMP for Wisconsin and Michigan sponsored by the Electric Power Research Institute (EPRI 1993a) found that major differences
OCR for page 14
Estimating Bounds on Extreme Precipitation Events: A Brief Assessment among competing procedures arise principally for small basins (less than 1000 square kilometers) and short durations (less than 6 hours). The EPRI study suggested values of PMP that were about 20 percent lower than the NWS values* for these situations. (For larger basins and longer durations, the values in the EPRI study were quite close to the NWS values.) The climatological raingauge network, which serves as the backbone of PMP analyses, is poorly suited for resolving rainfall accumulation gradients for small basins during short-duration precipitation events. High-resolution radar data may be of particular importance in developing PMP estimates for these events. Although PMP is defined as an absolute upper bound, the estimated PMP values may have a nonzero probability of being exceeded. Using sophisticated statistical techniques, the probability of exceeding established values of PMP has been estimated to vary significantly across the country. This variation causes a problem of equity because PMP is a legal standard. The annual exceedance probability has been estimated to range from 10−5 down to 10−9 within the United States (M. Schaefer, State of Washington Department of Ecology, presentation to the Committee). Some areas of the country already have seen storms that exceeded 90 percent of PMP, while other areas have not recorded a storm exceeding 50 percent of PMP (Riedel and Schreiner 1980). However, there is no simple relationship between how close a storm gets to PMP and the exceedance probability for that PMP estimate. Some PMP estimates developed by the NWS have tended to increase over time (NRC 1985, p. 47f), as more extreme storms have been observed, and are likely to do so in the future. Structures built to standards derived from PMP may eventually be considered substandard because of this trend. The resulting dam relicensing difficulties have economic repercussions and cause inequities. However, there is no meteorological solution to this problem, which occurs in many situations as our society becomes more risk conscious. It is likely that regional estimates of PMP will continue to differ from the national ones. Furthermore, unscrupulous practitioners could design regional PMP studies to yield lower values, saving dam owners millions of dollars, but increasing the risk that a dam spillway may be inadequate. This puts tremendous pressure on the regulatory agencies. NWS is pursuing methodological improvements for PMP analyses in a number of areas, including enhancements to moisture maximization procedures, use of statistical methods, and use of WSR-88D (Weather * FERC has decided to allow dam owners to use the values in EPRI 1993a for determining PMP in the areas covered.
OCR for page 15
Estimating Bounds on Extreme Precipitation Events: A Brief Assessment Surveillance Radar - 1988 Doppler design) radar rainfall data (discussed further below). The NWS plans to use WSR-88D rainfall products informally in PMP analyses. However, there are no formal plans for incorporating rainfall products from WSR-88D data directly into a PMP storm catalog database. EXTREME PRECIPITATION EVENTS To those who have not tried it, measuring rainfall might seem simple. It is not. Groisman and Legates (1994) provide a good summary of some of the difficulties with raingauges. Most of the problems lead to underestimates. Windy conditions significantly reduce the catch efficiency of raingauges (sometimes by as much as half), and extreme precipitation rarely occurs with light winds. This bias is reduced with windshields at only about 200 of the less than 3000 U.S. stations that record hourly accumulations. Another 5000 stations record only daily or twice-daily accumulations. During extreme events, tipping-bucket-type gauges can miss significant rainfall while resetting. Other errors, such as evaporation, contribute to a lesser degree. Most extreme precipitation events are associated with mesoscale systems, with spatial scales of 10 to 1000 kilometers. Precipitation often varies significantly on spatial scales of a few kilometers, especially during extreme events. The observing network is not nearly fine enough to capture this variation. Observing stations are also quite scarce at high elevations. Many stations have moved over time, and even for those stations that have remained largely in place over a century, the gauge type and surrounding conditions have often changed. All of these problems limit our ability to obtain accurate precipitation statistics. Radar estimates of precipitation have the advantage of very high spatial and temporal resolution. Complete scans of the air volume around an antenna can be performed within several minutes with a spatial resolution on the order of a kilometer by the new WSR-88D Doppler radars now being deployed by the NWS as part of the NEXRAD system. These new instruments will provide precipitation information for the United States with unprecedented coverage. Comparisons of new radar data with concurrent traditional-style point measurements of precipitation, combined with the existing record of about a century 's worth of station-based time series, may allow greatly improved inferences concerning the probabilities of extreme precipitation events. Although the potential for radar is great, significant questions concerning the accuracy of radar rainfall estimates must still be resolved. The absolute accuracy of radar precipitation estimates is questionable, although, as noted above, the standards against which they are com-
OCR for page 16
Estimating Bounds on Extreme Precipitation Events: A Brief Assessment pared are poor. Heavy-precipitation estimates from radar can suffer from problems of either overestimation or underestimation (Austin 1987). At extreme ranges, radars suffer from decreases in spatial resolution and an increase in elevation of the radar beam above ground level. There have been well-documented examples of underestimation of heavy rainfall, including the Big Thompson storm in 1976 and the Shadyside, Ohio, flash flood in 1989, because the radars could detect only the upper parts of the distant storms. The conversion from reflectivity to rainfall is not a solved problem even in the absence of the beam geometry problems discussed above. The radar reflectivity of rain increases linearly with the number of drops but with the sixth power of drop size. Therefore, there is strong dependence on drop size distribution, which varies from storm to storm. Radars that provide polarization diversity information can better estimate drop size and therefore better estimate the relationship between reflectivity and rainfall. The WSR-88D radars do not now have this capability. If polarization diversity information were available, there would be still greater potential for radar estimates of rainfall, especially for heavy rain rates. Visible and infrared satellite data have been utilized qualitatively for PMP analyses in recent years. Satellite imagery has proven especially useful in documenting the occurrence and characteristic life cycles of mesoscale convective complexes. This information has been helpful in storm transposition analyses. Satellite imagery also is of utility in lifecycle studies of a broad range of mesoscale convective systems. However, the difficulty (and cost) of obtaining data is a significant obstacle to broader utilization of operational satellite data. Major advances in rainfall measurement through radar and satellite will be of least utility in mountainous regions. Mountainous regions have the most severe problem of inadequate observations from standard raingauge networks as well. The NEXRAD radar network has major gaps in the mountainous western United States. Even in mountainous regions that are to be covered by NEXRAD radars, observations will be of limited value for quantitative applications because of ground clutter and obscuration. Interpretation and quantitative utilization of satellite data also are most difficult in the complex terrain of mountainous regions. Paleohydrologic data can be used to assess the frequency of very rare events and are most likely to be available in mountainous regions, where the data discussed above are particularly inadequate. Recent research has shown that for the Colorado River, gauged streamflow during the past 100 years approaches the envelope curves of paleoflood values for the past 10,000 years (Enzel et al. 1993). These analyses provide the most
OCR for page 17
Estimating Bounds on Extreme Precipitation Events: A Brief Assessment direct support for the assumption of an upper bound for either streamflow or precipitation accumulations. Therefore, paleoflood data may be useful for assessing PMF, and indirectly in assessing PMP. Numerical models also have potential for contributing to enhancements and modifications of PMP methodology in the absence of sufficient data. Weisman and Klemp (1984) have demonstrated that atmospheric dynamics can have a significant effect on precipitation efficiency for severe convective storms. These model results have direct bearing on assessing the validity of moisture maximization assumptions that are part of PMP calculation procedures. Numerical models also can provide a way to examine orographic effects on extreme storms and to test the appropriate range for transposition of storms used for determining PMP. We expect that the use of high-resolution mesoscale numerical models will improve understanding and forecasting of precipitation in mountainous regions. Although precipitation research and mesoscale meteorology are subjects of ongoing work, they have not been an especially high priority in recent years. The earlier recommendations of this Committee (NRC 1990) regarding the STORM (Storm-scale Operational and Research Meteorology) program (STORM Program Office 1990) have not been implemented. In fact, STORM has been abandoned. Research on similar topics of mesoscale weather have been incorporated into the plans for the U.S. Weather Research Program. These plans include an emphasis on hydrometeorological linkages with mesoscale research, but they have not yet been fully implemented. There is a sense in the hydrology community that problems associated with heavy rainfall have not received the attention they warrant from the atmospheric sciences community. A striking example is found in the field experiment known as PRE-STORM (Preliminary Regional Experiment for STORM-Central; see Blanchard 1990), from which a large body of literature has been produced concerning a severe squall line that passed through Oklahoma and Kansas on 11 June 1985 (summarized in Houze 1993). On the other hand, little research has focused on the storm of 10 June 1985 that produced extremely heavy rainfall. Although significant emphasis has been put on heavy rainfall within the operational forecasting community, problems persist there as well. In describing field operations during the Austin, Texas, flash flood of May 1981, Maddox and Grice (1986) note that “greater attention is given to the more classic manifestations of severe thunderstorms (hail, wind and tornadoes) [than to heavy rainfall] in field operations. During the Austin storm, on this particular night it is possible that a flash flood warning might have been issued sooner had not this particular psychology been operating.”
OCR for page 18
Estimating Bounds on Extreme Precipitation Events: A Brief Assessment ALTERNATIVES TO PMP Alternatives to PMP have been proposed. Most of these are statistical procedures that seek to describe the probability distribution of extreme precipitation events. The question of probability-based analyses of extreme hydrometeorological events has been examined in detail in the report Estimating Probabilities of Extreme Floods (NRC 1988). These statistical analysis techniques are a mature set of methodological procedures, and there is significant implementation experience. However, these procedures are designed principally for frequency analysis of rainfall data from gauged stations (measurements at single points). An important issue to consider in development and utilization of statistical procedures is the choice between storm-based and station-based analyses. The standard PMP procedures are based on analyses of individual storms (storm-based), whereas many frequency-analysis procedures for precipitation apply only to data from specified points (station-based). There are important arguments for maintaining storm-based analyses. Storm-based approaches provide greater potential for links with meteorological analyses. Advances in numerical modeling and from radar and satellite analyses are more readily integrated into a storm-based framework. Furthermore, total accumulation over a drainage basin, rather than at a single point, is much more useful for engineering practice. Also, the time variability of rainfall during a storm is increasingly recognized as playing a central role in PMF analyses. The stochastic storm-transposition procedures of Fontaine and Potter (1989) and Foufoula-Georgiou (1989) provide a model for transforming statistical procedures to a storm-based context, but these techniques are not mature.
Representative terms from entire chapter: