The PDSI, based on the Palmer Drought Model (Palmer, 1965), has been one of the most commonly used drought indicators in the United States. One reason for its popularity is that its development in 1965 preceded other indices and resulted in its widespread use and wide-ranging application. The PDSI is derived from a moisture balance model, using historic records of precipitation, temperature, and the local available water capacity of the soil. The PHDI uses a modification of the PDSI to assess longer term moisture anomalies that affect streamflow, groundwater, and water storage. A primary difference between the PDSI and the PHDI is in the calculation of drought termination, using a ratio of moisture received to moisture required to definitely terminate a drought. With the PDSI, a drought ends when the ratio exceeds 0%, if it remains greater than 0% until reaching 100%. With the PHDI, a drought does not end until the ratio reaches 100% (Karl, 1986; Karl et al., 1987).
The PDSI and PHDI are calculated for climate divisions, typically on a monthly basis, with cumulative frequencies representing all months and all climate divisions (Karl, 1986):
PDSI/PHDI Drought Category Cumulative Frequency (approx.)
-3.00 to -3.99 Severe drought -4.00 or less Extreme drought
These PDSI/PHDI values, however, are not spatially and temporally invariant. Cumulative frequencies vary, depending on the region and time period under consideration (Guttman et al., 1992; Karl et al., 1987; Nkemdirim and Weber, 1999; Soulé, 1992). For example, the category of "extreme drought," with the overall frequency of 4%, varies in frequency from less than 1% in January in the Pacific Northwest to more than 10% in July in the Midwest (Guttman et al., 1992; Karl et al., 1987). As another example, the probability of "extreme drought" in Virginia varies as follows: Climate Division 1 for January, 4.17%; Climate Division 1 for July, 2.08%; Climate Division 6 for January, 3.21%; Climate Division 6 for July, 1.04% (Lohani et al., 1998).
As regional drought indices, the PDSI and PHDI permit comparisons of drought events over relatively large areas. The Palmer indices also offer a long-term historic record, going back more than 100 years. Yet the Palmer indices and their water balance model have several limitations (Alley, 1984; Guttman et al., 1992; Karl, 1986; Karl and Knight, 1985). The Palmer indices are not particularly suitable for droughts associated with water management systems, because they exclude water storage, snowfall, and other supplies. They also do not consider human impacts on the water balance, such as irrigation. The values for determining the severity of the drought, and the beginning and end of a drought, were arbitrarily selected based on Palmer's studies of central Iowa and Kansas. The water balance model has been critiqued on several grounds; for instance, soil moisture capacities of the two soil layers are independent of changes in vegetation. The methodology used to normalize the values is only weakly justified on a physical or statistical basis. For instance, for climatic regions with a large interannual variation of precipitation, the statistical measure of normal is less meaningful than other measures, such as the range, median, or mode of the precipitation distribution (Wilhite and Glantz, 1985). The indices are based on departures from climate normals, with no consideration of precipitation variability, so they tend not to perform well in regions with extreme variability in rainfall or runoff (Smith et al., 1993). Although the Palmer indices are widely applied within the United States, they have little acceptance elsewhere (Kogan, 1995).
Was this article helpful?