National Drought Atlas Developed

A National Drought Atlas for the United States is nearing completion. The idea for the atlas was inspired by the severe droughts that plagued parts of the country from 1987 to 1989, covering much of the Great Plains states, the Ohio Valley, and the southeast. In California, the drought lasted 6 years. In general, water managers were caught off-guard by the drought's severity and duration.

The atlas is the first nationwide study of drought frequency that planning groups can use to answer a large number of "what-if" questions. They can determine the most likely durations of critical drought, when such droughts are likely to occur, how much drier than "normal" a given drought may be, and what droughts have occurred in a given region.

A compendium of maps, graphs, and tables provide frequency analysis of precipitation for durations of 1 month through 5 years and historical summaries of data on streamflow and the Palmer Drought Severity Index. The atlas derives its power from recently developed statistical techniques. Because the questions most likely to be asked about drought pertain to some aspect of risk, the atlas presents information in frequency terms. A typical question might be, "What is the probability that a 3-month period beginning in June will bring 60% or less of normal precipitation to southwestern Ohio?"

The atlas is a product of the National Study of Water Management During Drought, which took place from November 1989 to October 1993. The study's principal objective was to develop strategies for improving water management to reduce the nation's vulnerability to drought. The study was funded and managed by the U.S. Army Corps of Engineers. Other federal agencies, non-federal governments, universities, environmental and public interest groups, and IBM Research Division provided additional expertise.

Using consistent terminology and methodology, the atlas can provide planners with a reliable base of information for risk assessment. It uses monthly precipitation totals from 1119 sites in the National Climatic Data Center's Historical Climatology Network. The average length of record at each site is 85 years. Frequency analysis of the precipitation data used two recently developed techniques, regionalization and L-moments.

Regionalization is the use of data from several sites, rather than just one, to estimate drought frequencies at a site. L-moments [Hosking, 1990; Wallis, 1993] are statistics that enable extreme frequencies to be estimated reliably without being too severely affected by individual erratic observations.

The 1119 gaging sites were grouped into 111 regions [Guttman, 1993] within which the drought risk was judged to be uniform, using a statistical test [Hosking and Wallis, 1993] based on L-moments. Drought frequencies were estimated for each region [Guttman et al., 1993]. Figure 1 illustrates the results for precipitation for durations of 12 months starting in January (that is, annual totals); as in the atlas itself, rainfall amounts are shown as ratios of the median rainfall. For example, for the southernmost Texas region, only once in 50 years would the annual rainfall total be as little as 51% of the median annual rainfall. The information on precipitation frequency contained in the atlas is equivalent to 468 versions of Figure 1 for different durations, starting months, and recurrence intervals.

Fig. 1. Annual precipitation totals, as a percentage of the median annual precipitation, for a drought with recurrence interval of 50 years. Percentages apply to each site in a region. Gaging sites are indicated by lines joining the site location to the center of the region.

The Palmer Drought Severity Index (PDSI) is calculated from observed temperature and precipitation values and is a surrogate for soil moisture. A number of states rely on it for initiating and discontinuing drought emergencies. However, a close look at the index showed it to be misleading in some respects.

In preparing the National Drought Atlas, PDSI values were recalculated at monthly intervals for 1036 sites. These calculations revealed that the PDSI index overstates the severity and variability of droughts within a given region. The PDSI is supposed to indicate of the severity of a drought. A PDSI value of -3 to -3.9 is supposed to indicate a "severe" drought. Figure 2 shows the frequency of occurrence of PDSI values of -3 or less for the month of July. The frequencies vary widely from state to state, and many sites in the midwest and west show a surprisingly high frequency, as high as 31% at some sites, for this "severe" drought.

Fig. 2. Observed probabilities of a July Palmer Drought Severity Index value of -3 or less. Adapted from Figure 1(a) of Guttman et al. [1992].

This suggests that states using the PDSI for declaration of drought emergencies should, at a minimum, consider adjusting the trigger values of the index to maintain a consistent frequency.

References

Wallis, J. R., Regional frequency studies using L-moments, in Concise Encyclopedia of Environmental Systems, edited by P. C. Young, pp. 468-476, Pergamon, Oxford, 1993.

Return to Science and Society
Return to Starting Point