1.  INTRODUCTION

Drought is a natural climatic phenomenon characterized by a deficiency of precipitation which has an adverse effect on biota, including plants, animals, and humans. A drought at a given location or region is a period of time, lasting weeks, months or years, during which the actual moisture supply consistently falls short of the climatically expected moisture supply (Palmer, 1965).

The impact of drought is wide ranging; therefore, it is interpreted in many ways. Meteorological drought refers to a precipitation anomaly which leads to a marked deficiency of precipitation in a given period. Hydrological drought is associated with a long-term lack of water in soil profiles and fresh water bodies, in an amount sufficient to affect their normal use. Agricultural drought refers to a short-term soil moisture deficiency in an area which normally receives sufficient moisture to sustain crops and livestock (Karl and Young, 1987). Socioeconomic drought is associated with a deficiency of water needed to meet the demand for industrial and urban activities.

Droughts are better documented in semiarid and subhumid regions, where humans tend to concentrate (Lugo and Morris, 982). Drought data in extremely arid regions is scant, since very few people are actually affected (Campbell, 1968). Likewise, droughts in very humid regions go largely unnoticed, since the supply of water usually exceeds the actual demand.

Here we relate drought characteristics to climatic parameters across the climatic spectrum. The latter is defined in terms of mean annual precipitation and cross-referenced to annual potential evapotranspiration. A characterization in terms of either of these two parameters provides an appropriate framework for the systematic analysis of droughts. Rasmusson (1987) has noted that the ability to predict regional, seasonally averaged patterns of precipitation anomalies can be of significant value to society.

2.  DROUGHT CHARACTERISTICS

Droughts are usually characterized by their duration, intensity, and recurrence interval. Duration is the period of time when there is a marked deficiency of precipitation, preceded and followed by periods when there is no such deficiency. Intensity measures the extent to which actual precipitation is different than the expected quantity. Recurrence interval is the number of years that it would take a drought to recur.

Droughts are cyclic natural phenomena. Coping with droughts is possible through proper forecasting and planning. To reduce the impact of drought, it is necessary to develop a capability to forecast its characteristics, i.e., its duration (How long will it last?), its intensity (How severe will it be?), and its recurrence interval (How often will it recur?). Once these characteristics are known for a given region, they can be used in programs of drought mitigation.

3.  THE CLIMATIC SPECTRUM

Droughts are regional in nature; their occurrence is related to prevailing climatic parameters. A readily available climatic parameter is mean annual precipitation, which depends on several factors, including: (1) Latitude; (2) orographic factors; (3) mesoscale ocean currents; (4) atmospheric wind circulation; (5) proximity to oceans and large lakes; (6) atmospheric pressure; (7) character of the Earth's surface, including color and texture; and (8) presence of atmospheric particles; both natural and human-induced. Closely related to mean annual precipitation is annual potential evapotranspiration, which is a function of: (1) Net solar radiation; (2) vapor-pressure deficit; (3) surface roughness; and (4) leaf-area index (Monteith, 1965).

For our purposes, we define the climatic spectrum solely in terms of mean annual precipitation. This type of characterization is particularly useful for midlatitudinal regions, where droughts are shown to be more intense (Karl, 1983). We characterize the climatic spectrum in terms of the ratio of mean annual precipitation P_ma to annual global terrestrial precipitation P_agt.

The amount of moisture stored in the atmosphere is a function of latitude and climate, varying typically from 2-15 mm in polar and arid regions, to 45-50 mm in humid regions (UNESCO, 1978). A global terrestrial mean value of 25 mm is assumed for the purpose of estimating annual global terrestrial precipitation. The atmospheric moisture recycles every eleven days on the average, for a total of 33 cycles per year (L'vovich, 1979), which results in the annual global terrestrial precipitation P_agt = 825 mm. For comparison, L’vovich (1979) has estimated 910 mm for exorheic drainages, which constitute 78.4 percent of the total terrestrial area, and 238 mm for the remaining 21.6 percent (endorheic drainages). This amounts to a weighted value P_agt = 765 mm. Here we assume P_agt = 800 mm.

Globally, the middle of the climatic spectrum, i.e., the division between semiarid and subhumid climates, corresponds to P_ma/P_agt = 1. Regions with P_ma/P_agt < 1 have less-than-average moisture; conversely, regions with P_ma/P_agt > 1 have greater-than-average moisture. Mean annual terrestrial precipitation varies typically in the range 100-6000 mm, with a few isolated cases falling outside this range (Baumgartner and Reichel, 1975). This justifies the division of the climatic spectrum into the following eight regions:

1. Superarid, with P_ma/P_agt < 0.125

2. Hyperarid, with 0.125 ≤ P_ma/P_agt < 0.25

3. Arid, with 0.25 ≤ P_ma/P_agt < 0.5

4. Semiarid, with 0.5 ≤ P_ma/P_agt < 1

5. Subhumid, with 1 ≤ P_ma/P_agt < 2

6. Humid, with 2 ≤ P_ma/P_agt < 4

7. Hyperhumid, with 4 ≤ P_ma/P_agt < 8

8. Superhumid, with P_ma/P_agt ≥ 8

Table 1 shows the climate types, with mean annual precipitation P_ma and corresponding P_ma/P_agt ratios. This climatic classification, applicable to midlatitudinal regions, extends previous classifications of a similar nature (Bull, 1991).

Since our model relates drought characteristics to climate, it is expected that climatic changes will lead to changes in drought characteristics. Climatic changes attributable to anthropogenic activities are currently being examined throughout the world (IPCC, 1996). Climatic changes may intensify floods ard droughts even before the changes in temperature are severe enough to be noticed (Bruce, 1994; Houghton, 1994).

4.  GLOBAL DATA ON DROUGHTS AND CLIMATE

The scientific and engineering literature is quite extensive on the subject of droughts. However, systematic documentation across a diversity of climatic regions is lacking. Drought characteristics and related climatic data from the United States, India, Australia, Russia and Brazil is reported in this section.

Legates and DeLiberty (1993a; 1993b) and Groisman and Legates (1994) have shown that the long-term annual spatially averaged precipitation in the continental United States is approximately 800 mm. Thus, it can be surmised that the central part of the United States is close to the middle of the climatic spectrum.

Droughts in the United States have a tendency to be more persistent in the interior of the country (High Plains and Rocky Mountain States) than in areas farther east or west. For these regions, durations for moderate and severe droughts varied between 3 and 5 yr (Karl, 1983). Greater drought persistence in the Great Plains of central North America than in any other part of the United States has been documented by Karl et al. (1987) and Laird et al. (1996). Johnson and Kohne (1993) have shown that drought persistence is greater in the interior of the United States, including the states of Wyoming, Colorado, North Dakota, and Montana.

Horn (1989) studied the spatial variability of droughts in Idaho, where the mean annual precipitation varies between 250 and 1500 mm, with an average of about 800 mm (NOAA, 1980). He concluded that the median drought durations varied between 5.6 to 6.4 yr throughout the state. Diaz (1983) has shown that the interior and western areas of the United States are more likely to endure protracted periods of dry weather, i.e., persistent droughts.

Klugman (1978) has studied droughts in the Upper Midwest of the United States from 1931 to 1969. His analysis showed that while the 1930's and the 1950's were decades of drought, the 1940's and the 1960's were wet periods. This indicates a recurrence interval of 20 yr for this subhumid region, with mean annual precipitation of 1500 mm (NOAA, 1980).

Karl and Young (1987) have reported on drought recurrence interval in the Southeastem United States, ranging from 20 to 60 yr. These locations are in the states of North Carolina, Georgia, and Tennessee, which have humid climates (NOAA, 1980).

The Indian subcontinent, the northern half of which lies above the tropic of Cancer, has a wide variability of climatic regimes, from superarid to superhumid. High rates of evapotranspiration prevail over hyperarid Rajastan, in western India, with annual rates exceeding 2500 mm and reaching 3500 mm in some regions of Northwest Rajastan. Conversely, low rates of evapotranspiration prevail over hyperhumid Assam and the Himalayan Bengal, in northeastem India, with annual rates in the range of 1200 to 1500 mm. Over central India, which is semiarid, evapotranspiration rates vary in the range 1400-1800 mm (Abbi, 1974). Gregory (1989) has reported that droughts in hyperhumid regions of India such as Assam are very infrequent, with the last documented drought in the region dating back to 1900.

The Australian continent, more than half of which lies below the tropic of Capricorn, has a wide variability in climatic regimes. For instance, in hyperarid William Creek, in South Australia, precipitation is 127 mm and potential evapotranspiration is more than 2540 mm. In arid Alice Springs, in the Northern Territory, precipitation is 250 mm and potential evapotranspiration is 2460 mm. In semiarid/subhumid Perth, in Western Australia, precipitation is 890 mm and potential evapotranspiration is 1670 mm. In subhumid Sydney, in New South Wales, precipitation is 1200 mm, and evapotranspiration exceeds 1020 mm (Kendrew, 1961). French (1987) has analyzed long-term series of annual rainfall for semiarid Georgetown, in North Central South Australia, where the mean annual rainfall is 475 mm. The records from 1874 to 1985 show 20 drought events, i.e, a mean recurrence interval of 5.6 yr.

Russian experience shows that over the past 1000 yr, catastrophic droughts have occurred with a frequency of 8-12 per century, i.e., every 10 years on the average. In Kazakhstan, which is mostly hyperarid and arid (Zonn ef al, 1994), around 35 severe droughts have occurred in the last 100 years, i.e., every thre years the average. In the Ukraine, where climate and soils are more favorable for agricultural production than in Kazakhstan, droughts affect the area every 4-5 yr (Kogan, 1997).

Ponce (1995a) has documented the drought events in the drought polygon of Northeastern Brazil during the twentieth century: 1903-04, 1915, 1919, 1930-32, 1942, 1953, 1958, 1970, 1979-83, and 1990-93. The polygon contains a diversity of biogeographical regions, ranging from the seridó (arid) to the caatinga (semiarid) to the agreste (subhumid) to the mata (humid). The data suggests a recurrence interval from 4 to 12 yr, with an average value of about 10 yr. The drought duration varied between 1 and 5 yr. The mean annual precipitation varied between 395 mm and 1831 mm.

Ponce (1995b) has reported that hydrological droughts in the Upper Paraguay river basin, in central western Brazil, recur every 28 to 30 yr on the average, and last 3 to 6 yr. The mean annual precipitation in the Upper Paraguay river basin varies between 900 and 2000 mm, with an average of 1380 mm. This shows that this subhumid region is subject to persistent droughts at approximately 30-yr intervals.

The aforementioned studies are not strictly comparable, since drought may have been defined in different ways. However, the data does provide an empirical basis on which to formulate a conceptual model of drought characterization.

5.  POTENTIAL EVAPOTRANSPIRATION

The climatic spectrum can also be characterized by the ratio of annual potential evapotranspiration to mean annual precipitation. Annual potential evapotranspiration E_ap is the amount of evapotranspiration that would take place in a terrestrial ecosystem under the assumption of an ample supply of moisture at all times (Thornthwaite et al., 1944).

To determine suitable E_ap/P_ma ratios, we have estimated potential evapotranspiration for subtropical and midlatitudinal regions across the climatic spectrum (Vysotskii, 1905; Ivanov, 1948; WMO, 1975). For instance, we estimate E_ap = 3000 mm at the limit between superarid and hyperarid regions. Corresponding estimates of annual potential evapotranspiration for other climatic regions led to the E_ap/P_ma ratios shown in Table 1. These limits are to be taken as indication of general conceptual trends, and not necessarily as precise values.

6.  DROUGHT PERSISTENCE AND SEVERITY

A meteorological drought can last one or more weeks, months, or years. Our conceptual model deals specifically with meteorological droughts lasting at least one year (Dracup et al., 1980a; 1980b). By definition, a drought year is one with markedly less-than-average precipitation. One or more consecutive drought years constitute a drought event. Persistence is the property of a drought event to last more than one year. For a given drought event, intensity refers to the extent of the precipitation deficit.

To determine drought intensity, the moisture deficiency is accumulated over the drought duration. Therefore, the longer the duration, the greater the intensity. Since dry periods are generally followed by corresponding wet periods, it follows that recurrence interval is always greater than duration. Thus, for meteorological droughts lasting at least one year, the recurrence interval is at least two years.

The preceding concepts of drought intensity and duration resemble those of Dracup et al.'s severity, duration, and magnitude, wherein magnitude is defined as the ratio of severity over duration (Dracup. et al., 1980a). In our case, intensity is akin to severity and, therefore, intensity is equal to magnitude times duration.

7.  CONCEPTUAL MODEL OF DROUGHT CHARACTERIZATION

A conceptual model works in the mean, i.e., it describes general trends and not necessarily specific events. It is meant to aggregate the deterministic and stochastic components of the precipitation anomalies. Its value is that it provides an appropriate framework for interpreting the regional variability of drought phenomena.

Given a drought year with precipitation P, where P < P_ma, we classify the precipitation deficiency into three types: (1) moderate, with P/P_ma = 0.75, (2) severe, with P/P_ma = 0.5, and (3) extreme, with P/P_ma = 0.25 (Central Water Commission, 1982; National Institute of Hydrology, 1990). As an example of a moderate to severe drought, Groisman and Legates (1994) have reported a value P/P_ma = 0.6 during the Dust Bowl of the 1930's in the U.S. Great Plains.

Or a given drought year, we define drought intensity as the ratio of the deficit (P_ma-P) to the mean P_ma. For a given drought event lasting at least one year, intensity is defined as the summation of the annual intensities I = Σ [(P_ma-P)/P], in which I = drought-intensity index.

Therefore, average annual drought intensity is the total drought intensity divided by the duration. Average annual drought intensity has been referred to as drought magnitude by Dracup et al. (1980a, 1980b), and as run intensity by Frick et al. (1990).

Our conceptual model of drought characterization is based on the following premises:

1. Drought duration varies across the climatic spectrum, reaching a maximum around the middle and decreasing toward the extremes. A typical duration in both superarid and superhumid regions is one year. Long- duration, i.e., more persistent droughts, prevail around the middle of the climatic spectrum, at the limit between semiarid and subhumid regions (Karl, 1983; Laird et al., 1996).

2. Since drought intensity is directly related to duration, intensity also reaches a maximum around the middle of the climatic spectrum, and decreases toward the extremes (Johnson and Kohne, 1993).

3. Drought recurrence interval increases gradually from the dry to the wet side of the climatic spectrum, from 2 to 100 yr (Karl and Young, 1987; Gregory, 1989; Zonn et al., 1994; Ponce, 1995a).

Table 1 summarizes our conceptual model of drought characterization. For drought duration, the expected values vary between 1 and 6 yr, with larger values toward the middle of the climatic spectrum (6 yr), decreasing toward either extreme (1 yr).

This choice of durations clarifies the issue of persistence, i.e., the tendency of droughts to last longer in the middle of the climatic spectrum (WMO, 1975; Karl, 1983; Johnson and Kohne, 1993). The changes in drought duration across the climatic spectrum point to the regional, rather than local, nature of persistence (UNESCO-WMO, 1985).

The longer durations toward the middle of the climatic spectrum are due to greater interannual precipitation variability within the semiarid and subhumid regions (Rasool, 1984). Within these regions, drought duration is likely to be the longest, approaching 4-6 yr.

The shorter durations toward both extremes of the climatic spectrum are justified because of smaller interannual precipitation variability. In superarid regions, variability is reduced because the precipitation amounts are small; in superhumid regions, variability is reduced because of the long length of the rainy season, approaching twelve months.

The Sub-Saharan drought of 1968-73, which does not match this drought duration pattern, is apparently the result of anthropogenic effects on the regional climate. In this case, an active desertification process led to regional climatic changes, decreases in mean annual precipitation, decreases in drought retum period, and a succession of drought events (Charney et al., 1975; Nicholson, 1983; Rasool, 1984; Hulme, 1989). The unusual persistence of the Sub-Saharan drought has been documented by Lamb (1982).

Drought recurrence interval varies between 2 yr on the extreme dry side and 100 yr on the extreme wet side, increasing in a geometric progression. Since recurrence interval decreases from wet to dry climates, and since recurrence interval must always exceed duration, it follows that duration must decrease toward the dry side of the climatic spectrum (Table 1). Thus, in hyperarid regions, droughts are short and recur once every 2-3 yr; in semiarid and subhumid regions, droughts are long and recur once every 6-25 yr; in hyperhumid regions, droughts are short and recur once every 50-100 yr.

The rainy season is shortest (one month) on the extreme dry side, longest (twelve months) on the extreme wet side, and lasts four months around the middle (Sinha et al., 1987).

To use Table 1 for a semiarid region while the duration is 5 yr, the intensity ranges from 1.25 (moderate) to 3.75 (extreme), and the recurrence interval is 9 yr. For a subhumid region while the duration is 5 yr, the intensity ranges from 1.25 (moderate) to 3.75 (extreme), and the recurrence interval is 18 yr. For these two examples, while the duration and intensity have remained the same, the recurrence interval has doubled from semiarid to subhumid.

8.  SUMMARY

We have daveloped a conceptual model of drought characterization across the climatic spectrum, primarily suited to midlatitudinal regions. The model estimates expected values of drought duration, intensity, and recurrence interval as a function of the ratio of mean annual precipitation to annual global terrestrial precipitation.

Drought duration varies between 1 and 6 years across the climatic spectrum, and reaches a maximum toward the middle. Drought intensity varies directly with duration. Drought recurrence interval increases geometrically, from 2 yr on the extreme dry side to 100 yr on the extreme wet side.

Comparison of our conceptual model with drought data from the United States, India, Australia, Russia, and Brazil, shows that the model is suitable as a framework for drought analysis and planning. The model's singular strength is its climatic basis, as evidenced by its ability to depict regional variability. The model may also be used to describe changes in drought characteristics which could be attributed to anthropogenic climatic change.

REFERENCES