Professional Documents
Culture Documents
Ana A. Paulo and Luis S. Pereira, Agricultural Engineering Research Center, Institute
of Agronomy, Technical University of Lisbon, Tapada da Ajuda, Lisbon, Portugal
Abstract: A better knowledge of droughts is required to improve water management in water scarce
areas. To appropriately cope with droughts, there is the need to adopt adequate concepts relative to
droughts and water scarcity, to properly use drought indices that help characterize them, including
ones relative to their severity, and to develop prediction tools that may be useful for early warning and
that may reduce the respective lead time needed for appropriate response. In this paper, concepts
relative to drought and other water scarcity regimes are discussed aiming both to distinguish droughts
from other water scarcity regimes and to base a common understanding of the general characteristics of
droughts as hazards and disasters. Three main drought indices are described aiming at appropriate
characterization of droughts: the theory of runs, the Palmer Drought Severity Index (PDSI), and the
Standardized Precipitation Index (SPI). Their application to local and regional droughts in the region
of Alentejo, Portugal is presented focusing on the respective comparison and possible adequateness for
drought monitoring. Results indicate some difficulties in using the theory of runs, particularly because
it requires a subjective definition of thresholds in precipitation and does not provide a standardized
classification of severity. Results show that drought characterization with the PDSI and the SPI produce
coherent information, but the PDSI is limited relative to the SPI because it requires more data to perform
a soil water balance while the SPI needs only precipitation data, which are more easily available in
numerous locations. It is concluded that adopting the SPI is appropriate, but there is advantage in
combining different indices to characterize droughts.
Keywords: Palmer Drought Severity Index, Portugal, Standardized Precipitation Index, theory of
runs, water scarcity
as in relation to water requirements for natural and manmade ecosystems. The concept of scarcity also embraces
the quality of water because degraded water resources
are unavailable or at best only marginally available for
use in human and natural systems (Pereira et al., 2002).
The sustainable use of water implies resource conservation, environmental friendliness, technological appropriateness, economic viability, and social acceptability of
development issues. The adoption of these sustainability
facets is a priority for using water in every human, economic and social activity human and domestic consumption, agriculture, industry, energy production, recreational
and leisure uses particularly in water scarce regions as
analyzed by Pereira et al. (2002). However, the causes
for water scarcity have to be well understood and identified in order to adopt the measures and practices that are
appropriate for each case. Moreover, global change may
highly impact water resources availability by increasing
38
Permanent
Temporary
Nature produced
Man induced
Aridity
Drought
Desertification
Water shortage
the variability of rainfall in arid regions, the climate extremes and, particularly, the frequency and severity of
droughts (Kabat et al., 2002). Therefore, the understanding of drought phenomena, their identification and prediction constitute a current challenge to mitigate the drought
impacts in many areas around the world.
Water scarcity may result from a range of phenomena. These may be produced by natural causes, may be
induced by human activities, or may result from the interaction of both, as indicated in Table 1.
In a recent study for UNESCO, these water scarcity
regimes were defined and analyzed in view of developing
the appropriate means to cope with water scarcity. The
following definitions and concepts (Pereira et al., 2002),
based upon those originally proposed by Yevjevich et al.
(1983), are adopted:
Aridity is a natural permanent imbalance in the water
availability consisting in low average annual precipitation, with high spatial and temporal variability, resulting
in overall low moisture and low carrying capacity of
the ecosystems. Aridity may be defined through climatological indices such as the Thornthwaite moisture index, the Budyko radiation index of dryness, or the
UNESCO precipitation/evapotranspiration index
(Sanderson, 1992). Under aridity, extreme variations
of temperatures occur, and the hydrologic regimes are
characterized by large variations in discharges, flash
floods and large periods with very low or zero flows.
Drought is a natural but temporary imbalance of water
availability, consisting of a persistent lower-than-average precipitation, of uncertain frequency, duration and
severity, of unpredictable or difficult to predict occurrence, resulting in diminished water resources availability, and reduced carrying capacity of the
ecosystems. Many other definitions of drought exist
(e.g. Yevjevich, 1967; Dracup et al., 1980; Wilhite and
Glantz, 1987; Tate and Gustard, 2000). Generally, these
definitions clearly state that drought is mainly due to
the break down of the rainfall regime, which causes a
series of consequences, including agricultural and hydrological hazards that result from the severity and
duration of the lack of rainfall. The less predictable
characteristics of droughts with respect to their initiation and termination, frequency, and severity make
drought both a hazard and a disaster: a hazard because
it is a natural event of unpredictable occurrence but of
recognizable recurrence; a disaster because it causes
the disruption of the water supply to the natural and
agricultural ecosystems as well as to other human activities.
produce the conservation and perhaps the seasonal augmentation of the available resource. On the other hand
droughts require the development and implementation of
preparedness and emergency measures.
Differences in the perception of drought lead to the
adoption of different definitions, which do not have general acceptance nor have worldwide applicability as reviewed by Wilhite and Glantz (1987) and Tate and Gustard
(2000). The perception of the hazardous nature of dependence on precipitation and water availability depends on
the climatic, meteorological and hydrological regimes of
the affected region, as well as on the severity of the effects. Some authors prefer to adopt an operational definition that distinguishes between meteorological, agricultural,
and hydrological droughts. These usually focus on the indicator variable of prime interest, which could be the precipitation (meteorological drought), soil moisture
(agriculture drought), stream flow discharges or groundwater levels (hydrological drought and groundwater
drought). Alternative definitions result from the complexity of the hydrological process that controls the temporal
and spatial distribution of rainfall via the various paths within
the large-scale hydrological cycle and the global circulation of the atmosphere. In certain regions, where water
supplies mainly depend upon river diversions, when dealing with a regional drought it may be necessary to consider not only precipitation but also stream flow. However,
stream flow is also a dependent variable, controlled by the
current and antecedent precipitation. In many cases,
where the river discharges are regulated by dams and other
hydraulic structures, the drought definition may need to be
more a reflection of the river management decisions than
of the natural supply.
The controversy over perceptions of drought and the
consequent defining of them and their characteristics do
not help decision and policy makers to plan for droughts
and is one of the central problems of water management
for drought (Grigg and Vlachos, 1990). Lack of clearly
agreed definitions makes it difficult to implement preparedness measures, to apply timely mitigation measures when
a drought occurs or to adequately evaluate drought impacts. Using appropriate drought indices to monitor climate and hydrological variables related with drought is
extremely helpful to identify the occurrence of droughts,
their onset, evolution and dissipation, so to adopt first preventive measures and, later, appropriate mitigation measures. As for definitions, there is no common agreement
on drought indices, their adaptation to local climates or
about their use in water management. However, adopting
appropriate drought indices supports water management
decisions in line with dominant water scarcity problems
and should help not only coping with droughts but with
water scarcity. Further analysis is provided in the abovereferred UNESCO publication by Pereira et al. (2002).
Several studies were developed in Portugal relative to
drought indices and their use for local and regional droughts.
39
40
formed after testing all data sets used for quality. The randomness of the annual data sets was investigated through
tests for homogeneity, absence of trends and autocorrelation
using a set of non-parametric tests (Helsel and Hirsch, 1992).
Theory of Runs
A run is a succession of the same kind of observations
preceded and succeeded by one or more observations of
different kind. The theory of runs is based on the choice
of a critical threshold level, yc (Guerrero-Salazar and
Yevjevich, 1975). Considering a discrete time series, x1,
x2, , xi,, xn, a negative run occurs when xi is consecutively less than yc, during one or more time intervals. Negative runs in rainfall time series are related to drought and
the difference between yc and xt is referred as a deficit. A
run can be characterized by its length (L), its cumulated
deficit (D), and its intensity (I), as described in Figure 1.
Therefore, any drought, s, can be characterized by its:
duration, L(s), the number of consecutive time intervals
where rainfall remains below the critical level; cumulated
deficit, D(s), the sum of consecutive deficits, and; intensity, I(s), given by the ratio D(s)/L(s).
The number of droughts and their characteristics depend upon the choice of the critical rainfall threshold level
yc. After analyzing several threshold levels, the threshold
yc = , i.e. a rainfall below the mean by the quantity
corresponding to the standard deviation, was selected for
the case study presented herein.
The Palmer Drought Severity Index
The Palmer Drought Severity Index (PDSI) was developed by Palmer (1965) as a meteorological index to
identify and assess the severity of a drought event. The
purpose of the index was to measure the departure of
the moisture supply. The Palmer Index has been used
since 1965 as a useful tool in drought assessment.
The PDSI is derived from the soil water balance, usually on a monthly basis. The input data consist of historical
records of precipitation and evapotranspiration, the latter
computed either by the Thornthwaites method, or by any
other method like the FAO-PM method (Allen et al., 1998)
adopted in this research. The soil is conceptually divided
in two layers: the surface layer, where the total available
zi = k j d i
(1)
where ETPi is potential evapotranspiration, PRi is potential recharge of soil moisture, PROi is potential runoff and
PLi is potential soil moisture loss, all referring to the time
interval i. The coefficients j, j, j, and j are the ratios
between the average of each of the actual values (ET, R,
RO, and L) to the average of the corresponding potential
values (ETP, PR, PRO, and PL) over a calibration period.
Their subscript j refers to the month of the year. These
coefficients are called water balance coefficients and they
were conceived to adjust the potential values according to
monthly changes.
There are three intermediate indices, X1, X2 and X3
relative to the severity of a wet spell that may or may not
be developing, the severity of a dry spell that may or may
not be developing, and the severity of the current spell,
respectively. These indices are computed at each interval
i from the following recursive equation:
X i = 0.897 X i 1 + zi / 3
(3)
PDSI values
4.00
3.00 to
2.00 to
1.00 to
0.50 to
0.49 to
0.50 to
1.00 to
2.00 to
3.00 to
4.00
3.99
2.99
1.99
0.99
0.49
0.99
1.99
2.99
3.99
41
SPI values
Drought category
Time in category
Extremely wet
Very wet
Moderately wet
Slightly wet
Incipient wet spell
Near Normal
Incipient drought
Mild Drought
Moderate drought
Severe drought
Extreme drought
0 to 0.99
1.00 to 1.49
-1.50 to 1.99
2.00
Total
Mild drought
Moderate drought
Severe drought
Extreme drought
34.0%
9.2%
4.4%
2.3%
50.0%
sults obtained for the SPI, which were previously computed using precipitation data sets starting by 1932 (Paulo
et al., 2003a). Care is therefore required to select the values for the initial soil moisture.
Some limitations of the PDSI were pointed out by Alley (1984), namely the arbitrary rules to quantify the intensity, the beginning and the end of a drought event and
the backtracking procedure to select the drought index.
Limitations on the use of PDSI to Mediterranean regions
have been shown by Cancelliere et al. (1996), therefore it
was required to perform an appropriate analysis for its
application in southern Portugal and compare it with other
indices. A former study (Paulo et al., 2003b) identified some
limitations but demonstrated the robustness of the PDSI.
The severity of droughts by the PDSI show it to be sensitive to the soil water characteristics, particularly to the
value selected for the total available soil water (TAW),
and to the evapotranspiration computation procedure.
However, results have shown that the FAO-PM method
(Allen et al., 1998) could be appropriate and was therefore adopted in this study.
Standardized Precipitation Index
The Standardized Precipitation Index (SPI) was developed by McKee et al. (1993) with the purpose of identifying and monitoring local droughts. Multiple time scales
(McKee et al., 1995), from three-month to 24-month, may
be used. Shorter or longer time scales may reflect different lags in the response to precipitation anomalies. The
initiation of a drought event is identified by a backtracking
procedure: the drought event has its confirmation only when
in a series of continuously negative values of SPI the value
-1 or less is reached. A drought ends when the SPI becomes positive. Each drought event is then characterized
by its:
lead-time, which is the number of months within a
drought event before SPI 1 is reached;
duration, defined by the time between its beginning and
end;
severity, given by the monthly SPI value (Table 3);
magnitude, calculated by the sum of the SPI for every month
from the initiation to the end of each drought event; and
The SPI is an index based on the probability distribution of precipitation. This index depends on the distribution
function, on the sample used to estimate the parameters
of the distribution, and on the method of estimation. The
two-parameter gamma distribution function is often used
but statistical goodness-of-fit tests are required before its
adoption. The entire period of records is generally used to
estimate the parameters of the gamma distribution function, e.g. by the method of maximum likelihood (Kite, 1988).
In practice, the computation of the SPI index (Edwards,
2000) in a given year i and calendar month j, for a k time
scale requires:
1. calculation of a cumulative precipitation series X ik, j
(i=1,,n) for that calendar month j, where each term
is the sum of the actual monthly precipitation with precipitation of the k-1 past consecutive months;
2. fitting of a gamma distribution function F(x) to the
monthly series;
3. computing the non-exceedence probabilities corresponding to the cumulative precipitation values;
4. transforming those probabilities into the values of a standard normal variable, which actually are the SPI values.
The SPI is a z score and represents an event departure from the mean, expressed in standard deviation units.
SPI is a normalized index in time and space. This feature
allows comparisons of SPI values between different locations. Drought severity is arbitrarily defined according to
the SPI values as listed in Table 3 (McKee et al., 1993),
where the expected time in each drought category was
based on an analysis of a large number of rainfall stations
across Colorado, USA. The percent of time in moderate,
severe, and extreme droughts correspond to those expected
from a normal distribution of the SPI.
42
A finer analysis may be produce using specific mathematical tools as used in Mediterranean regions (e.g. Rossi et
al., 1992; Alvarez and Estrela, 2003; Rossi and Cancelliere,
2003), including in Portugal (Santos, 1983; Santos et al.,
1988; Henriques and Santos, 1996). This approach is particularly useful for regional planning of water resources at
the river basin scale or within the boundaries of an administrative region, because when adopting this analytical
methodology, it is possible to understand how severe and
which area is affected. This information can be useful for
drought monitoring and drought contingency planning as a
component of a water resources plan aimed at a water
scarcity and drought prone region. In addition, this approach
may be particularly important in the national context because drought studies in Portugal have often focused regional droughts.
A regional drought is identified when a significant fraction of the total area of the region is under drought conditions or, in other words, when the sum of the areas Aj
affected by drought reaches a selected critical areal threshold Ac , as explained in Figure 2. The most common areal
threshold is Ac is 50 percent of the total area.
Applying the theory of runs, a regional drought may
be defined by its: (1) duration, the length of the consecutive time intervals in which drought affects an area that
equals or exceeds the pre-defined areal threshold; (2) cumulated areal deficit, the sum of the cumulated deficits at
each site, weighted by the corresponding areas of influence; (3) regional drought intensity, the ratio between the
cumulated areal deficit and the duration; (4) mean regional
coverage, the mean fraction of the area in the region that
is affected by drought.
Standardized Precipitation Index
The SPI may be extended to identify regional droughts
and the respective regional coverage and severity. At the
regional scale, a month is classified to be under drought if
the percent area of the region having SPI -1 during that
month exceeds the adopted areal threshold. A regional
80
60
40
20
Oct
Nov Dec
Jul
Aug Sep
Figure 4. Monthly mean areal precipitation in Alentejo from 49 rainfall stations for 1931/1932 to 1998/1999
43
contrarily to the twelve-month time scale, the gamma distribution function adopted for computing the SPI did not
pass satisfactorily the chosen goodness of fit tests (e.g.
Kolmogorov-Smirnov, Cramr-von Mises and AndersonDarling) for several data sets when using the three-month
time-scale.
When comparing the results for the theory of runs for
the critical threshold yc = with those for the SPI, it
was observed that both indices are compatible but yield a
different identification of droughts (Table 4), in a larger
number for the SPI. This reflects difficulties in appropriately selecting the critical rainfall threshold level yc when
droughts are identified with negative runs. Departing from
the mean by only a fraction of the standard deviation could
give better results but that fraction would be different from
a location to the other. Defining the critical threshold by
the median produces an exaggerated number of droughts
identified (Paulo et al., 2002). Moreover, the severity of
droughts using the theory of runs would require a subjective classification by combining the cumulative deficits and
the average intensity of droughts.
Because the SPI is obtained from cumulated values
referring to the month in analysis and the k-1 antecedent
months, it provides more continuous information on the
variation of dryness conditions in time and takes into consideration the lag in response to the identified lack of precipitation. Therefore, it produces information that may be
more consistent than that resulting from the theory of runs.
In addition, it does not require a subjective decision on
defining a threshold and the SPI severity classification is
clear and objective.
Comparing the PDSI and the SPI
The Palmer Index was computed with ET estimated
from both the Thornthwaite and the Penman-Monteith
equations for the locations referred in Figure 3. Three different soil types referring to total available water of 250,
200, and 150 mm were also used. The SPI was computed
for the time scales three-, six- and twelve-month. The
PDSI was compared with the SPI for the same locations
relative to the time period 1965-2000. Best results were
obtained when the PDSI was calculated with the ET Penman-Monteith method and the soil water characteristics
for TAW equal to 150 mm, and the SPI was computed
with the twelve-month time scale. Selected results for the
comparison are shown in Table 5.
The linear correlation coefficient between the PDSI
and the SPI is higher for the twelve-month time scale,
ranging from 0.814 in Beja to 0.765 in vora, denoting
that the PDSI has some built-in memory. Svoboda (2004)
compared the linear correlation between the PDSI and
the SPI computed in several time scales for Hilo, USA,
concluding that the highest value was achieved for the
nine-month time scale, around 0.85.
An example when the PDSI was computed with reference ET using the FAO Penman-Monteith equation
44
Table 4. Comparing the local droughts identified for Alentejo using the theory of runs and the 12-month SPI for the period 1931/1932 to 1998/1999
Theory of runs
# of years
# of
Mean
Station under
drought
cumulated
code
drought
events
deficit (mm)
16L03
17J01
17L02
17M01
18G01
18M01
19G02
19J03
19L01
19L02
19M01
19N01
20I01
20L01
20O01
21F01
21G02
21J02
21K01
21M01
21M02
22E01
22H01
22J01
23E01
23G01
24F01
24H01
24I01
24J03
24N01
25J02
25P01
26F01
26G01
26J01
26L01
27E01
27G01
27H01
27I01
27K01
28H01
28I01
29G01
29G02
30E01
30F01
30H03
11
11
12
14
11
8
13
11
11
10
11
13
12
13
7
12
12
7
10
9
12
13
10
12
9
11
13
10
9
13
12
10
10
11
12
11
11
13
12
10
10
8
12
10
10
10
9
12
12
10
10
11
12
10
7
11
9
10
9
7
11
11
10
6
10
10
7
7
8
10
12
6
10
7
10
11
9
8
9
9
8
8
8
10
8
8
12
10
9
8
8
10
9
8
8
8
11
10
73.2
87.8
81.1
66.1
80.7
98.4
97.3
102.7
70.0
86.0
97.0
81.0
66.6
64.5
30.2
103.7
84.1
128.0
94.1
94.8
89.3
91.7
146.2
75.6
107.9
73.5
70.8
71.2
102.1
108.0
67.6
90.5
78.3
120.4
74.2
87.7
87.3
80.2
99.8
85.9
123.3
76.0
98.6
90.7
134.1
115.9
113.3
174.4
102.5
Intensity
(mm/yr)
66.5
79.8
74.4
56.7
73.4
86.1
82.4
84.0
63.6
77.4
61.7
68.5
61.0
49.6
25.9
86.4
70.1
128.0
65.8
84.3
74.4
84.6
87.7
63.0
83.9
66.8
59.9
64.1
90.8
74.8
50.7
72.4
62.6
87.6
61.8
63.8
63.5
74.0
83.1
77.3
98.6
76.0
82.1
81.6
107.3
92.7
100.7
159.9
85.4
# of
drought
events
Average
duration
(months)
16
16
16
16
16
16
15
13
17
18
14
13
15
14
13
17
18
15
13
17
14
18
15
13
17
15
18
17
18
15
13
17
15
11
15
15
15
13
15
15
17
15
16
13
12
16
14
16
17
21.3
19.0
18.7
22.7
19.2
16.9
18.6
25.2
20.8
19.7
22.6
23.9
21.1
25.1
26.0
18.9
16.4
20.5
23.9
22.3
21.4
17.3
19.2
24.0
18.0
17.2
17.1
17.6
18.3
20.3
24.7
18.7
19.0
24.2
20.3
19.9
21.4
22.5
19.6
20.9
17.4
22.7
19.7
24.1
24.0
21.6
22.1
18.9
18.9
(Allen et al., 1998) and a soil with TAW equal to150 mm/
m is shown in Figure 5. Results for all locations are similar
and show a generally good agreement between these two
indices (Paulo et al., 2003b). However, because the SPI
requires only precipitation data while the PDSI needs other
weather data to compute ET, the SPI can be computed
for more locations, which constitutes a great advantage
over the PDSI. Meanwhile, since results obtained so far
show the appropriateness for using the SPI, stochastic
SPI
Average time (months) in
classes of severity
Extreme
Severe
Moderate
1.0
1.4
0.7
0.3
0.5
1.1
1.2
1.1
0.4
0.7
1.0
1.2
1.2
1.0
0.6
0.6
0.9
2.0
1.3
0.3
1.2
1.2
0.7
0.9
1.7
1.6
0.8
1.0
0.9
1.5
0.7
1.1
0.8
1.3
1.3
1.7
1.3
1.3
1.1
0.5
1.5
1.5
0.8
1.6
2.3
0.8
1.5
0.8
0.9
1.9
2.4
3.3
2.6
3.3
1.6
3.8
2.9
2.6
3.1
3.4
3.7
2.7
2.9
1.8
3.1
2.2
1.5
3.2
2.8
3.6
2.8
3.9
3.6
1.7
2.7
2.8
2.5
2.1
2.5
3.4
1.9
3.0
5.6
2.9
2.3
2.3
3.5
3.9
4.2
3.0
2.3
3.3
3.1
3.4
2.9
2.9
3.4
2.2
6.1
4.5
4.9
6.6
5.1
4.2
4.4
7.5
5.9
4.6
5.4
6.8
5.0
6.3
7.2
5.2
4.8
4.5
5.1
4.8
4.9
3.5
4.9
6.9
3.7
4.1
4.3
4.8
4.8
5.7
6.9
4.4
5.4
6.1
5.1
5.0
5.3
6.4
4.2
4.7
2.5
4.6
4.6
4.4
4.3
5.3
4.3
4.6
5.1
Mild
Lead
time
(months)
12.3
10.7
9.8
13.2
10.3
10.0
9.2
13.7
12.0
11.3
12.9
12.3
12.3
14.9
16.5
10.1
8.5
12.4
14.4
14.5
11.8
9.8
9.6
12.6
11.0
8.8
9.2
9.4
10.5
10.6
13.8
11.4
9.8
11.2
10.9
10.9
12.5
11.3
10.3
11.5
10.4
14.2
11.1
15.0
13.9
12.6
13.4
10.1
10.8
3.3
3.2
2.3
4.9
4.6
4.1
5.1
3.4
3.9
3.8
3.3
4.5
4.5
3.8
6.5
2.8
2.6
4.7
7.3
5.3
6.0
3.2
4.3
5.6
5.1
3.5
3.7
3.5
4.3
3.5
5.1
4.2
3.9
3.5
3.6
5.1
5.5
2.4
4.9
4.0
5.3
5.5
3.9
2.9
7.1
4.9
4.4
3.7
3.5
models are being applied to the SPI series aiming at predicting the drought class transitions, as exemplified in a
companion paper.
Results obtained for the three indices indicate the advantage for using more than one index when operational
identification of droughts is aimed. This is the option for
several drought watch systems as analyzed by Rossi
(2003).
45
Table 5. Comparing the number of months in the classes of no-drought (N), mild drought (1), moderate drought (2), severe drought (3), and
extremely severe drought (4) computed with the PDSI and the SPI for 3, 6, and 12-month time scales
PDSI
Severity
classe
SPI 3-month
SPI 6-month
# of
Severity classes and number of months in each class
months
N
1
2
3
4
N
1
N
1
2
3
4
261
83
55
18
15
432
N
1
2
3
4
268
94
40
21
9
432
N
1
2
3
4
245
94
63
24
6
432
N
1
2
3
4
257
92
56
23
4
432
Alvalade
213
37
12
4
0
266
Beja
214
27
19
3
0
263
Elvas
204
46
17
5
0
272
vora
214
23
19
8
2
266
SPI 12-month
2
30
27
21
4
6
88
16
13
14
5
3
51
1
3
3
2
2
11
1
3
5
3
4
16
229
25
5
1
0
260
22
44
17
3
1
87
8
10
19
4
3
44
0
4
10
7
2
23
2
0
4
3
9
18
216
26
8
0
0
250
36
48
16
1
0
101
6
8
21
3
0
38
2
1
10
9
0
22
1
0
0
5
15
21
42
36
7
13
4
102
9
18
7
1
3
38
2
8
5
2
2
19
1
5
2
2
0
10
228
39
12
0
0
279
34
32
13
4
0
83
5
18
14
6
0
43
0
5
1
6
8
20
1
0
0
5
1
7
218
37
0
0
0
255
44
50
19
0
0
113
6
7
13
8
0
34
0
0
8
7
5
20
0
0
0
6
4
10
27
28
21
4
0
80
9
10
11
6
2
38
5
7
6
6
3
27
0
3
8
3
1
15
213
30
9
0
0
252
28
47
18
3
0
96
2
14
21
7
0
44
2
2
9
13
5
31
0
1
6
1
1
9
190
29
9
0
0
228
52
53
19
0
0
124
3
11
25
8
1
48
0
1
8
12
2
23
0
0
2
4
3
9
27
33
15
7
1
83
10
21
13
6
0
50
4
7
5
1
0
17
2
8
4
1
1
16
203
18
8
1
0
230
45
40
17
8
2
112
8
25
16
10
0
59
0
9
12
3
2
26
1
0
3
1
0
5
194
28
3
0
0
225
58
47
11
1
0
117
5
12
19
8
0
44
0
5
23
11
3
42
0
0
0
3
1
4
Notes: N= No-drought; 1 = Mild drought; 2 = Moderate drought; 3 = Severe drought; 4 = Extreme drought
End
1
1934/35
1934/35
2
1943/44
1944/45
3
1952/53
1952/53
4
1956/57
1956/57
5
1964/65
1964/65
6
1980/81
1980/81
7
1982/83
1982/83
8
1991/92
1991/92
9
1994/95
1994/95
10
1998/99
1998/99
Average drought characteristics
Duration
Areal
Intensity
(year) coverage (%) (mm/year)
1
2
1
1
1
1
1
1
1
1
1.1
60
84
39
41
51
98
97
82
92
84
73
34.6
92.9
28.4
9.3
19.2
90.8
94.6
45.2
77.9
72.9
56.6
46
1
Nov-1933
2
Oct-1938
3
Dec-1943
4
Jan-1949
5
Apr-1953
6
Jan-1957
7
Dec-1964
8
Apr-1967
9
Jan-1971
10
Feb-1973
11
Feb-1980
12
Dec-1982
13
Apr-1991
14
Aug-1994
15
Nov-1998
Average
Time in category (%)
End
Jan-1936
Feb-1939
Apr-1946
Apr-1950
Jan-1955
Nov-1958
Oct-1965
Feb-1968
Jan-1972
Oct-1976
Oct-1982
Dec-1983
Oct-1993
Nov-1995
Sep-1999
27
5
29
16
22
23
11
11
13
45
33
13
31
16
11
20.4
12
2
13
5
2
1
6
1
2
7
2
6
12
5
5
5.4
10.1
1
0
6
1
0
0
0
0
0
1
11
5
4
7
5
2.7
5.1
0
0
1
0
0
0
0
0
0
0
1
0
1
0
0
0.2
0.4
42.1
36.0
61.9
37.8
21.2
21.3
42.6
17.5
15.7
28.0
44.2
71.8
48.1
64.4
77.5
42.0
-0.92
-0.72
-1.18
-0.75
-0.57
-0.60
-0.83
-0.57
-0.52
-0.72
-0.97
-1.28
-0.89
-1.18
-1.40
-0.87
6) are considered. One single and one long duration regional drought event is identified when the SPI is used for
Ac equal to 30 percent, or two drought events are considered when using Ac equal to 50 percent (Events 11 and 12
in Table 7). As shown in Figure 6, the areal coverage
changes with the severity of drought taken as threshold.
If a mild drought is considered, it starts by Feb 1980 with
a coverage of 70 percent of the region, reduces coverage
below 50 percent for only one month (Nov 1982) and returns immediately to 100 percent coverage the next month
100
90
80
70
Area (%)
Initiation
60
50
40
30
20
10
0
Jan-80
Jul-80
Jan-81
Jul-81
Jan-82
Jul-82
Jan-83
Jul-83
Jan-84
Jan-82
Jul-82
Jan-83
Jul-83
Jan-84
-3,00
-2,00
SPI-12m
Drought
Number
Areal coverage
for moderate,
severe, and extreme
drought (%)
SPIreg
-1,00
0,00
1,00
2,00
Jan-80
Jul-80
Jan-81
Jul-81
Date
until falling below 50 percent by Jan 1984. When a moderate drought is considered, the area coverage starts to be
above 50 percent some months later, by Oct 1980, and
falls below this threshold between Nov 1981 and Dec 1982;
then it overcomes that threshold and falls below it by Nov
1983, thus corresponding to two separate droughts. However the SPIreg is always negative, except in the month of
Nov 1982, where it has the value of 0.09, indicating that a
recovery may have not been produced and, probably, only
one but long drought has occurred. Therefore, based on
the analysis exemplified above, variable areal thresholds
should be used with the SPI in relation to the severity of
drought: Mild, area affected by mild drought, 50 percent
total; Moderate, area affected by moderate drought, 40
percent total; Severe, area affected by severe drought, 30
percent total; Extreme, area affected by extreme drought,
20 percent total.
When adopting variable areal thresholds as above it
results that the percent time in mild, moderate, severe,
and extreme drought categories is respectively 29.5, 11.8,
6.9, and 1.6 percent (Paulo et al., 2003a), thus in agreement with the data produced by McKee et al. (1993) and
given in Table 3.
Conclusions
The analysis presented above confirms that drought
indices greatly help to identify and characterize local and
regional droughts. The indices relative to the theory of
runs and the SPI are easier to use because they require
rainfall data only, while the PDSI needs additional weather
data to estimate the evapotranspiration. Because such data
are available in only few locations, they are less useful for
a regional scale analysis.
When comparing the theory of runs with the SPI, the
latter seems to be more appropriate to characterize droughts
because, on the one hand, the theory of runs does not
provide an objective classification of severity and the selection of an appropriate critical threshold is also not objective. On the other hand, contrary to the theory of runs,
the SPI time scales are not simple computation time scales
but represent the memory of the precipitation observed
for the antecedent periods relative to the calculation month,
i.e., time scales may reflect different lags in the response
to precipitation anomalies. Adopting a regional SPI, the
SPIreg as proposed in this study, and variable areal thresholds, regional drought studies may be conducted using SPI
instead of the theory of runs, which was dominant in former
regional drought studies. This allows to overcome the limitations inherent to the theory of runs and to adopt at the
regional scale a more powerful drought index.
The PDSI is a robust index, which considers the
memory of past precipitation through the soil water balance, thus somewhat similarly to lag responses corresponding to the SPI time scales. However, the PDSI is influenced
by the soil TAW value selected for the water balance. A
47
Acknowledgements
Data used in this study were made available by the
Institute for Water (INAG), Portugal. This study was funded
through the national project PEDIZA 1999.64.006326.1, and
is now part of the research contract INTERREG III B
MEDOC 2002-02-4.4-1-084.
References
Al-Salihi A. 2003. Drought identification and characterization
in Jordan. In G. Rossi, A. Cancelliere, L.S. Pereira, T. Oweis,
M. Shatanawi, and A. Zairi, eds. Tools for Drought Mitiga-
48
49