A comparative performance analysis of three standardized climatic
drought indices in the Chi River basin, Thailand
Department of Water Resources Engineering, Faculty of Engineering, Kasetsart University, Bangkok 10900, Thailand
Department of Civil Engineering, Tokyo Institute of Technology, M1-6, 2-12-1, O-okayama, Meguro-ku, Tokyo, 152-8552, Japan
Received 2 November 2015
Accepted 4 February 2016
Available online 25 June 2016
Standardized precipitation index (SPI)
evapotranspiration index (SPEI)
Standardized precipitation actual
evapotranspiration index (SPAEI)
Drought indices are generally used as a tool for monitoring changes in drought conditions. This paper
evaluated the performance of three climatic drought indices to characterize drought trends in the Chi
River basin in Northeast Thailand. Initially, the drought assessment was conducted using the Stan-
dardized Precipitation Index (SPI), a precipitation-based index, and the Standardized Precipitation
Evapotranspiration Index (SPEI), an index taking into account the difference between precipitation and
potential evapotranspiration (PET). Then, this study simply applied an index called the Standardized
Precipitation Actual Evapotranspiration Index (SPAEI), similar to the commonly used SPEI, with the
difference being in the use of actual evapotranspiration (AET) instead of PET. Time series of the three
indices were compared with observed droughts. The results indicated that various indicators of different
indices can have diverse effects on drought conditions. The simple SPI, considering only precipitation,
can be used to identify characteristics of droughts with certain restrictions. Being multivariate indices,
the SPEI and the SPAEI were able to clearly detect the temporal variability of droughts to a greater extent
than the SPI index. Moreover, the different results derived from using P-AET instead of P-PET made a
substantial difference to temporal drought severity. Thus, climatic water demand had important aspects
in determining the drought conditions for this area.
Copyright ©2016, Kasetsart University. Production and hosting by Elsevier B.V. This is an open access
article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Drought is a recurring natural hazard characterized as having
below normal precipitation over an extended period of time
ranging from months to years (for example, Dai, 2011). Drought
occurs as a result of numerous variables acting on multiple time
scales and varies with spatial location and temporal season; thus,
drought can have impacts on various sectors especially on agri-
culture and ecosystems (for example, Heim Jr., 2002; Hao and
Singh, 2015). Due to its complex nature and widespread occur-
rence, it is difﬁcult to deﬁne drought and to identify its character-
istics. A drought initially occurs when there is a deﬁcit of
precipitation for a prolonged period and it may even lead to further
deﬁciencies of other hydrological parameters depending on
different time scales. Consequently, the impacts of drought can be
categorized into four typesdmeteorological drought, agricultural
drought, hydrological drought and socio-economic drought (for
example, Wilhite and Glantz, 1985). Thus, the classiﬁcations of
droughts may differ depending on perspectives and stakeholders
(for example, Mo, 2008; Intergovernmental Panel on Climate
Various types of data sources, including in-situ observations,
land surface model simulations and remote sensing, can be used to
characterize droughts. Generally, drought indices have been
developed and used to capture a drought's physical characteristics
such as its frequency, duration, severity and spatial extent (Wilhite
and Glantz, 1985). In addition, the selection of drought indicators e
a broad term for hydrological parameters (for example, precipita-
tion, temperature, streamﬂow) eis an important factor in devel-
oping a drought index. Construction of drought indices can
therefore be formed in several ways; for example, by using a single
aspect, combining multiple variables or mixing drought indices
(Hao and Singh, 2015). Examples of drought indices are often used,
such as the Palmer Drought Severity Index (PDSI; Palmer,1965), the
standardized precipitation index (SPI; McKee et al., 1993) and the
E-mail address: firstname.lastname@example.org (T. Homdee).
Contents lists available at ScienceDirect
Agriculture and Natural Resources
journal homepage: http://www.journals.elsevier.com/agriculture-and-
2452-316X/Copyright ©2016, Kasetsart University. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://
Agriculture and Natural Resources 50 (2016) 211e219
Standardized Precipitation Evapotranspiration Index (SPEI;
Vicente-Serrano et al., 2010). Recently, major drought indices were
comprehensively reviewed and their different development
methods together with limitations and strengths were also speci-
ﬁed (Hao and Singh, 2015). Currently, there is no particular index
that is capable of adequately characterizing drought conditions for
every place and every time period (Svoboda et al., 2015). Thus, the
selection of drought indices depends on the type(s) of droughts,
region(s) of interest, objective(s) and information available to each
drought user (Smakhtin and Schipper, 2008). Furthermore, the
challenge of drought assessment lies in the selection of the proper
indicators and index to quantify changes in drought conditions
(Tsakiris and Vangelis, 2005).
Some drought indices have proven popular The SPI (McKee
et al., 1993) is a precipitation-based drought index, so it is less
complicated as it requires only simple inputs. This index keeps
track of accumulative precipitation deﬁcits which can be computed
at various time scales and places. Such a method corresponds to the
multi-scalar nature of droughts taking place across multiple tem-
poral scales (Hou et al., 2007). However, apart from precipitation,
other variables such as temperature (Shefﬁeld and Wood, 2008),
atmospheric demand (Vicente-Serrano et al., 2015) and heat waves
(Beguería et al., 2014) may also increase the variability of droughts.
The SPEI (Vicente-Serrano et al., 2010) provides a further devel-
opment of the drought index by taking into account monthly cli-
matic water balances (P-PET). The SPEI is similar to and ﬂexible like
the SPI but it includes in addition, the PET demand, a key component
of water cycle (Oki and Kanae, 2006). The original formulation of the
SPEI employs the Thornthwaite (Th) equation (Thornthwaite, 1948)
for PET calculation since it only requires temperature and latitudinal
data. However, the use of Th equation leads to an underestimation of
PET in arid and semiarid regions (Jensen et al., 1990), and an over-
estimation of PET in humid and tropical regions (Van der Schrier
et al., 2011). The errors can be diminished by using the Penman
Monteith equation (Allen et al., 1989) which accounts for additional
climatic data. In addition, the issue of using actual evapotranspira-
tion (AET) instead of PET has been evaluated and suggested for
calculating drought indices (Joetzjer et al., 2013). There are still high
uncertainties and much disagreement on global drought trends due
to inconsistencies in the techniques and indicators used (for
example, Seneviratne et al., 2012). Hence, more research is needed
to bridge the gap, especially on local scales.
The objective of this paper was to evaluate the performance of
three climatic drought indices to characterize the drought trends at
the local scale. Focusing on a tropical monsoon climate, the
Northeast Thailand was chosen as it is repeatedly subjected to
drought events. In particular, the study considered temporal
drought trends compared to observed droughts.
Materials and methods
Thailand, in Southeast Asia, is located close to the equatorial
Indo-Paciﬁc basins; thus monsoon precipitation has a strong cor-
relation with the Southern Oscillation index (SOI), a sea level
pressure-based El Ni~
no-Southern Oscillation (ENSO) index
(Singhrattna et al., 2005). This region is characterized by a monsoon
tropical climate with distinctive dry and rainy seasons causing
droughts and ﬂoods to repeat alternatively.
The Chi basin is located in the central part of Northeastern
Thailand, within the range 15.3e17. 3
N and 101.3e104. 3
total catchment area of the Chi basin is approximately 49,476 km
which more than 60% is rain-fed agricultural area growing rice as the
major crop. The basin is surrounded by mountains and high plateau
ranges from the north to the west making the river basin of the lower
part look like a ﬂat bowl, as shown in Fig. 1. Deciduous and evergreen
forests are the main forest types covering 20% of the entire area. The
spatial dataset including a digital elevation map (DEM) and land use
data were provided by the Land Development Department (LDD),
Thailand. In terms of climate observation, the area is not greatly
inﬂuenced by human activities (dams and irrigation) compared to
the central region of Thailand (Kim et al., 2005).
The daily precipitation datasets used in this study were the
V1003R1 datasets at 0.25
gridded data provided by the Asian
Precipitation- Highly-Resolved Observational Data Integration To-
wards Evaluation of Water Resources (APHRODITE) project (Yatagai
et al., 2009). The dataset of APHRODITE is freely available online
(http://www.chikyu.ac.jp/precip/) in various formats. The daily,
observed weather datasets consisting of the maximum and the
minimum temperatures, relative humidity, wind speed and solar
radiation were obtained from the Royal Irrigation Department (RID)
of Thailand and the Thai Meteorological Department (TMD) for the
period between January 1978 and December 2007.
The average annual precipitation over the Chi River basin is
approximately 1150 mm per year, and the range in precipitation
varies between 900 and 1700 mm. Fig. 2A shows a box plot for
mean monthly precipitation for 30 year period. Fig. 2B shows a
decreasing trend of annual precipitation which is consistent with
Shefﬁeld et al. (2012) who noted decreases in regional precipitation
over Asia. The rainfall pattern in this area plays an important role in
the occurrence of drought. The dry season always dries up the
rivers causing serious problems for people during severe droughts.
Fig. 2B shows an increasing trend of PET and evapotranspiration
(ET) in the Chi River basin. There is evidence demonstrating that
the frequency, severity and duration of droughts in the North-
eastern region are higher than in other regions (Suwanabatr and
This study used the AET data obtained by the study conducted by
Homdee et al. (2012) which used the Soil Water Assessment Tool
(SWAT) model to simulate water balance. SWAT is a spatially
distributed, hydrological model developed to mainly predict the
impacts of land management practices on water balance (Arnold
et al., 1998). Conceptually, the SWAT model overlays a DEM, land
use and soil map and slope classes and further calculates the water
balance for each subdivision. The Penman-Monteith method was
used in the model to calculate the daily reference ET andthe potential
plant transpiration was calculated using the actual daily crop height
and leaf area index. Potential soil evaporation is an exponential
function of the reference ET and the soil cover, and is signiﬁcantly
reduced during periods with high plant water use. Actual soil evap-
oration is limited by the soil water content (
) and is reduced
drops below ﬁeld capacity. AET is the sum of
interception, actual soil evaporation and actual plant transpiration.
AET is one of the standard outputs of the SWAT model which
commonly generates daily stream ﬂows during the period
19782007. Thus, this paper focused on changes in past drought
events over the same period due to the limited, available AET data.
Also, Thailand's monsoon rainfalls were more closely related to El
no during 1980e1999 (Singhrattna et al., 2005); thus the obser-
vation of such drought variability over this period is valuable.
Climatic drought indices calculations
Standardized precipitation index
The Standardized Precipitation Index (SPI) (McKee et al., 1993)is
an index based on probability of the long term precipitation
T. Homdee et al. / Agriculture and Natural Resources 50 (2016) 211e219212
representing either abnormal wetness or dryness conditions. The
SPI was designed to quantify precipitation deﬁcits for different time
scales such as 3 mth, 6 mth, 9 mth, 12 mth or 24 mth of cumulative
precipitation. This allows the SPI index to reﬂect impacts resulting
from the availability of different hydrological parameters. The
initial step for calculating the SPI is to ﬁt long time series of
monthly precipitation measurements to determine the cumulative
probability density function (PDF). Then, the given PDF distribution
of the observed precipitation is transformed into a normal distri-
bution with a mean of zero and standard deviation of one. The
original concept of the SPI applied the two-parameter gamma
distribution. The complete calculation procedure for the SPI can be
Fig. 2. Climatic data in the Chi river basin: (A) box plot for mean monthly precipitation; the solid line in the solid squares within the box plot depicts the mean monthly value of
precipitation (B) time series of annual precipitation [P(Y)], potential evapotranspiration (PET) and evapotranspiration (ET).
Fig. 1. Topography and river tributaries of the Chi River basin in Northeastern Thailand.
T. Homdee et al. / Agriculture and Natural Resources 50 (2016) 211e219 213
found in McKee et al. (1993), and some details are provided in
is a scale parameter,
is a shape parameter, g(x) is the
gamma probability density function, e is Euler's number for expo-
) is the ordinary gamma function of
. The esti-
can be found in more detail in McKee et al. (1993).
Standardized precipitation evapotranspiration index
The Standardized Precipitation Evapotranspiration Index (SPEI)
has been developed to measure drought conditions (Vicente-
Serrano et al., 2010). This index is based on a monthly climatic
water balance driven by precipitation and PET. The procedure of the
SPEI computation relies on the original SPI calculation but uses the
monthly difference between precipitation (P) and PET as shown in
Here D provides a simple measurement of water deﬁcits or
surpluses aggregated at different time scales. The values are then
ﬁtted to several parametric statistical probability distributions to
transform the original values to standardized units. Unlike Vicente-
Serrano et al. (2010), the PET calculation in this paper used the
Penman-Monteith (PM) equation (Allen et al., 1998) recommended
by World Meteorological Organization (2009) which was
conﬁrmed as having the best performance over Northeastern
Thailand (Haruethaithip, 2003).
Furthermore, this study applied the use of actual evapotrans-
piration (AET) instead of PET for the SPEI calculation and it was
named the SPAEI index, where SPAEI stands for the Standardized
Precipitation Actual Evapotranspiration Index. Similar to the SPEI
method, the procedure for the SPAEI computation requires monthly
differences between precipitations (P) and AET as shown in Equa-
In order to measure the accuracy of the drought indices, dis-
tribution ﬁtting was tested by applying various probability distri-
butions such as gamma, log normal and normal distributions to ﬁta
time series of accumulated precipitation. Based on validation for
the distribution selection, the KolmogoroveSmirnov (KeS) test is a
test for the goodness of ﬁt. The nonparametric test is calculated as
the maximum difference between the empirical cumulative dis-
tribution of sampled points and the theoretical cumulative distri-
bution. This paper used the Pearson r coefﬁcient to measure the
linear relationship between the time series of each index.
Results and discussion
The results of the KeS test (not displayed) illustrated that the
two-parameter gamma distribution produced the best ﬁt for all
accumulation periods except for 1 mth SPI. This was consistent with
some studies in India (Kumar et al., 2009) and the USA (McKee
et al., 1993). However, it was less capable of ﬁtting the data in
isolated, coastal regions (Beguería et al., 2014). Apparently, the
distribution function at the short timescale (1 mth SPI) does not
quite ﬁt with the empirical distribution compared to the longer
timescale. This agreed with the ﬁndings by Vicente-Serrano et al.
(2010). This may have been due to numerous zero values of
precipitation in the short timescale of the SPI; hence, the evaluation
of SPI can be misleading particularly in dry climates. Consequently,
poorly ﬁtting probability distributions have the potential to bias
drought index values, either overstating or minimizing the
perceived severity of the drought index (Beguería et al., 2014).
Therefore, the proper selection of the probability distribution of
precipitation over different timescales is necessary. With the SPEI
distribution ﬁtting, GEV distribution consistently performed the
best goodness of ﬁt across accumulation periods which was
consistent with Stagge et al. (2015) reporting that the GEV distri-
bution consistently produces the best goodness of ﬁt, but which
contrasts with (Vicente-Serrano et al., 2010).
Evaluation of climatic drought indices
The temporal variations of observed droughts were examined.
Fig. 3 presents the time series of the SPI calculated for 1 mth, 3 mth,
6 mth, 9 mth, 12 mth and 24 mth, respectively. Drought events are
deﬁned whenever the SPI values are lower than 1, the threshold
value. The degree of drought can be computed as a factor of the
highlighted area size. The different timescales of the SPI demon-
strated differences in magnitude and duration of droughts. Longer
timescales showed a higher severity and longer duration of
droughts than the short timescales. For long timescales (>6 mth
SPI), water scarcity was less frequent, but lasted longer. Clearly,
drought events were likely take place every year based on the 1-
month timescale. This may have been due to the limitation of the
probability distribution being ﬁtted with zero values of precipita-
tion causing the SPI to fail to detect drought conditions in the re-
gion with a distinct dry season, particularly at short timescales (Wu
et al., 2007).
There is still no consensus on the most suitable techniques for
the validation of the drought indices (Hao and Singh, 2015). Several
studies conﬁrmed a relationship between drought events and the
ENSO variability. Their associations with El Ni~
no conditions were
also linked to the variability of monsoon precipitation causing
drought variability in this region including India (Niranjan Kumar
et al., 2013) and Thailand (for example, Buckley et al., 2007;
Singhrattna et al., 2012). Moreover, many studies validated the
performance of drought indices through drought events reported
or compared with well-accepted drought indices (for example,
Fig. 4 illustrates evidence of El Ni~
no years in Thailand conducted
by anomalies of the 3-month moving average precipitation given
by Thai Meteorological Department (2011). The ﬁgure demon-
strates dry periods occurring many times during the 30 yr of this
study scope. Mainly, droughts dominated during the early and late
1980s in 1982e1983 and 1986e1987; and were tended to be severe
during 1991e1993, 1997e1998 and 2002e2003. A severe drought
was experienced in 2004e2005. This was in accordance with the
report by the National Weather Service (National Oceanic and
Atmospheric Administration, 2015) as years of El Ni~
spondingly, the annual observed precipitation shows a slightly
decreasing trend (Fig. 2B) and rising temperature (not displayed)
over the region which causes the increasing trend in PET (Fig. 2B).
Singhrattna et al. (2005) also found an increasing drought trend
since the 1980s as a result of a warmer climatic trend in Thailand.
This was conﬁrmed by Dai (2011) and Shefﬁeld et al. (2012) that
there had been increasing evaporation driven by warmer weather
and decreased precipitation from 1950 to 2008.
Compared with the time series of 3 mth SPI, drought events
could be detected and they conformed to El Ni~
no years. Moreover,
the results were compared with the recorded natural disasters in
Thailand provided by the Center for Research on the Epidemiology
of Disaster (Guha-Sapir et al., 2011). The recorded droughts sorted
T. Homdee et al. / Agriculture and Natural Resources 50 (2016) 211e219214
by total number of affected people are rather similar to the number
of drought episodes compared to the SPI time series. However, the
index is unsuitable to be used to quantify drought severity based on
the original criteria of the SPI calculation. For example, McKee et al.
(1993) deﬁned the category moderately to severely dry when the
SPI values were found in the rangee1to1.99 for the period of
1e3 mth SPI. In fact this was simply dry spells in the summer
monsoon season in this region which may not cause damage to
agricultural ﬁelds. However, longer timescales of the SPI (>12 mth
SPI) are able to broadly identify the characteristics of drought du-
rations with the exception of their intensity level. This corre-
sponded to an earlier study that analyzed the spatial patterns of
droughts using remote sensing data in Northeastern Thailand
which found that the worst dry years in 1979, 1981, 1997e1998,
2001 and 2003 affected widespread areas for 6e12 mth SPI
(Wattanakij et al., 2006). However, the interpretation and utiliza-
tion of the SPI over the tropical monsoon region with distinct
seasonal precipitation should be carefully carried out to avoid any
Fig. 3. Evaluation of different months of the Standardized Precipitation Index (SPI). The gray shaded areas mean duration and magnitude of drought deﬁned by its beginning and
end. Drought events (gray shaded areas) occur any time an SPI value reaches an intensity of 1 or less.
T. Homdee et al. / Agriculture and Natural Resources 50 (2016) 211e219 215
misleading interpretations when being applied to the short
Fig. 5A and B show the performance of multivariate indices of
the SPEI (P-PET) and the SPAEI (P-AET) at multiple timescales
(1e24 mth). Both indices indicated extreme drought severity in the
early 20th century. The drying values ﬂuctuated between þ2
and 2 at short timescales (3e6 mth) except for the period of
1 mth. There was a downward trend toward the value of 4at
longer timescales (9e24 mth) which indicates a severe dry period
in this region. However, slight problems of overstating the fre-
quency of droughts at the 1-mth timescale still appear as it still uses
the SPI index. The performances of both climatic indices seem to be
able to capture the main characteristics of drought conditions. Both
indices recognized signiﬁcant severe droughts (12 mth period) in
1986e1987 and 1993e1994, especially the worst droughts in
1997e1998 and 2003e2004. More consecutive episodes of
droughts and higher intensity of multi-year droughts (12e24 mth)
were observed since the early 1990s. The ﬁndings are consistent
with Guha-Sapir et al. (2011) who recorded natural disasters in
Thailand conﬁrming that the years mentioned earlier experienced
extreme droughts. These prolonged droughts had adverse effects
on the agricultural and water resource sectors across the country.
Moreover, dry spells and droughts have threatened widespread
areas of crops. For example, after the spring season in 1997, drought
brought disaster to the farm sector and cost an estimated USD 290
million loss in farm income through the crop year 1998e1999. In
addition, these drought episodes were the most spatially extensive
covering more than 8 million km
from eastern China to central
Asia (Shefﬁeld and Wood, 2008).
Similarly, the SPEI and the SPAEI showed substantial higher
severity and longer durations of droughts than the SPI. This may
have been due to the temperature rise which was conducive to an
increase in the water demand of PET and AET that triggered the
severity of droughts especially on longer timescales. This has been
supported by McCarthy (2001) who found that temperatures
measured in Thailand in the spring during 1976e1999 displayed an
increasing trend, and thereby, the rainfall showed a decreasing
trend. Moreover, both indices clearly detected notable drought
episodes during 2003e2004 consistent with El Ni~
no years (Thai
Meteorological Department, 2011). The serious, strong drought
episodes resulting in a scarcity of water for consumption and
agricultural use continued to the following year. Several provinces
experienced serious droughts and were declared as droughte-
stricken areas. During 2004, the Department of Disaster Prevention
and Mitigation of Thailand declared serious water shortages in
major dams which lasted until 2005. Thus, these droughts can be
referred to as agricultural and hydrological droughts in this region.
However, the SPI could not clearly identify the episode of drought
severity. Therefore, both indices detected agricultural drought
better than the SPI; and appeared to be more suitable to capture
characteristics of droughtsdspeciﬁcally the degree of drought.
Interestingly, the SPEI failed to represent extreme droughts
during 1993e1994 where it only showed a slightly laggard dry
period, while the SPAEI based on P-AET clearly identiﬁed an
intensiﬁcation of drought related to observed droughts. Further-
more, Joetzjer et al. (2013) conﬁrmed that the use of P-AET shows
higher consistency with the hydrological drought index. In contrast
to our study, Beguería et al. (2014) preferred to use P-PET as evi-
denced by the detection of a strong heat wave in Central Europe
and Russia. Nevertheless, the inclusion of precipitation and PET or
AET contains diverse substantial variability of drought severities in
this study. This supports the ﬁnding of previous studies indicating
that PET plays an important role in the hydrological cycle in
Thailand (for example, Kanae et al., 2001); and its variation is pri-
marily affected by precipitation, land use and land cover change
(Kim et al., 2014).
Fig. 6 shows differences in the Pearson r coefﬁcient of the three
indices for 1e24 months. The correlation between the SPI and the
SPEI was relatively close at shorter timescales (1e6 mth) and
dramatically decreased at longer timescales (9e24 mth). The cor-
relation between the SPI and the SPAEI also followed a similar
pattern with a slightly lower correlation than the former. These
indicate that precipitation is a dominant aspect driven by drought
conditions conforming to the characteristics of a seasonal monsoon
climate in this region.
Undoubtedly, the correlation between the SPEI and the SPAEI
was expected to give a high degree of similarity through all time
series due to the use of rather similar variables. The differences
were very close with the maximum correlation value of 0.98 at
6 mth timescales. The consistent pattern implied a robust rela-
tionship between PET and AET which can be generally found in
monsoon periods. These results emphasized that changes in pre-
cipitation were not the only dominant driver causing the long term
drought variations but that rising temperature in terms of PET and
AET was also responsible for the drought severities. However, Ma
et al. (2014) argued that no consistent association between the
SPEI and PET in humid areas existed.
There are many factors leading to the diverse performances of P-
PET and P-AET, especially in land use and land cover in agricultural
areas. Correspondingly the recent study by Kim et al. (2014) noted
that the variation of ET in Thailand was the result of precipitation
variation and adverse land change. Moreover, differences in AET
and PET are derived not only from the process of transpiration of
land cover, but are also attributable to the fact that AET was
Fig. 4. Anomaly of 3-mth moving average precipitation (base period 1971e2000) in Thailand. El Ni~
no (shaded with no outline) and La Ni~
na (shaded with dashed outline) events are
an El Ni~
no-Southern Oscillation index based on the SST anomalies in the Ni~
no 3.4 region at 120We170 W and 5Se5N(Thai Meteorological Department, 2011).
T. Homdee et al. / Agriculture and Natural Resources 50 (2016) 211e219216
obtained from a hydrological simulation model which included
direct evaporation from large reservoirs that may play an important
role in the available water cycle. There are complex interactions
among precipitation, soil moisture and evapotranspiration by
vegetation (Intergovernmental Panel on Climate Change, 2012); for
example, dry weather in summer certainly increases the amount of
soil water available for evapotranspiration. Thus, land use and land
cover change may have substantial inﬂuences on drought trends.
The study aimed to analyze the temporal drying trends using
three climatic drought indices and to provide useful insights on
how different drought indices can be effectively used. The various
indicators of different indices can have diverse effects on drought
conditions. The simple SPI (considering only precipitation) can be
used to identify characteristics of droughts with certain re-
strictions. Even though the SPI has been put forward by the World
Meteorological Organization (2009) as a universal index, a single
indicator may not be sufﬁcient to characterize a complicated
drought (for example, Hao and Singh, 2015).
Similar to the procedure for the SPI calculation, the SPEI con-
siders the difference between P and PET, while the SPAEI takes into
Fig. 5. Comparison of climatic drought indices for varying monthly periods: (A) Standardized Precipitation Evapotranspiration Index (SPEI); (B) Standardized Precipitation Actual
Evapotranspiration Index (SPAEI). Drought events (gray shaded areas) occur any time an SPEI/SPAEI value reaches an intensity of 1 or less.
T. Homdee et al. / Agriculture and Natural Resources 50 (2016) 211e219 217
account the difference of P-AET. Certainly, the use of multivariate
indices is more effective in terms of detecting temporal changes in
drought conditions than using the stand-alone indicator, the SPI.
The results suggested that although precipitation had a key role in
explaining the temporal variability of droughts, increasing tem-
perature also had a crucial inﬂuence on water stress in the region,
especially over the longer timescales. Moreover, different aspects
also affect diverse drying such as the replacement of PET with AET
produces a substantially different result when capturing the du-
rations and magnitudes of the droughts that occurred in 1993.
Thus, the use of AET may be a useful parameter for better under-
standing drought trends in that kind of basin. The indicator not only
explicitly considers the actual effects of vegetation changes that
substantially inﬂuence evaporation, but also includes alteration of
land atmosphere feedback processed through simulation of the
Thailand is an agricultural country; hence, metrological and agri-
cultural drought indices are important. The Thai Meteorological
Department is the national weather agency carrying out the opera-
tional monitoring of drought conditions using the Generalized
Monsoon Index (GMI) that considers only monsoon precipitation in
rainy season (Thai Meteorological Department, 2014). However,
monsoon precipitation varies, and the factors affecting its variations
differ in different seasons and regions. For example, precipitation in
the southern region is slightly inﬂuenced by ENSO but it correlated
well with local wind (Tsai et al., 2 015). Therefore, these present a big
challenge to the prediction of monsoon droughts on seasonal-to-
decadal timescales (for example, Niranjan Kumar et al., 2013). Thus,
the use of the GMI index is difﬁcult to connect to other parameters
since it only considers monsoon precipitation in the rainy season,
unlike SPEI's long time series which can be directly usedto account for
the inter-annual variability of annual stream ﬂow (Joetzjer et al., 2013).
Even though drought episodes cannot be immediately detected
like ﬂoods, their consequences are somehow greater than the
damages caused by ﬂoods. Thus, effective monitoring of drought
conditions needs to be implemented in order to mitigate drought
impacts. There is still no consensus on global drought trends, thus
more research is strongly encouraged, especially at local and
regional scales (for example, Shefﬁeld et al., 2012). In addition,
drought experts should play an important role in providing infor-
mation using several different indices at multiple time steps that
are most suitable for their respective application (for example,
Svoboda et al., 2015).
Conﬂict of interest
The authors declare that there are no conﬂicts of interest.
The authors would like to thank the Thai Meteorological
Department and the Royal Irrigation Department, Thailand for
providing data. The authors gratefully acknowledge the Program of
Strategic Scholarships for Frontier Research Network for the Joint
Ph.D. Program Thai doctoral degree, CHE Thailand andProf. Shinjiro
Kanae, Tokyo Institute of Technology, Japan for supporting this
Allen, R.G., Jensen, M.E., Wright, J.L., Burman, R.D., 1989. Operational estimates of
reference evapotranspiration. Agron. J. 81, 650e662.
Allen, R., Pereira, L., Raes, D., Smith, M., 1998. Crop Evapotranspiration: Guidelines
for Computing Crop Water Requirements FAO Irrigation and Drainage Paper 56.
FAO, Rome, pp. 377e384.
Arnold, J.G., Srinivasan, R., Muttiah, R.S., Williams, J.R., 1998. Large area hydrologic
modeling and assessment part I: model development. J. Am. Water Resour.
Assoc. 34, 73e89.
Beguería, S., Vicente-Serrano, S.M., Reig, F., Latorre, B., 2014. Standardized precipi-
tation evapotranspiration index (SPEI) revisited: parameter ﬁtting, evapo-
transpiration models, tools, datasets and drought monitoring. Int. J. Climatol.
Buckley, B., Palakit, K., Duangsathaporn, K., Sanguantham, P., Prasomsin, P., 2007.
Decadal scale droughts over Northwestern Thailand over the past 448 years:
Links to the tropical paciﬁc and Indian ocean sectors. Clim. Dyn. 29, 63e71.
Dai, A., 2011. Drought under global warming: a review. Wiley Interdiscip. Rev. Clim.
Change 2, 45e65.
Guha-Sapir, D., Below, R., Hoyois, Ph, 2011. EM-DAT: the CRED/OFDA International
Disaster Database. Universit
e Catholique de Louvain, Brussels-Belguim. www.
Hao, Z., Singh, V.P., 2015. Drought characterization from a multivariate perspective:
a review. J. Hydrol. 527, 66 8e678.
Haruethaithip, P., 2003. Estimating Potential Evapotranspiration Using Remote
Sensing and Meteorological Data of Mun and Chi Watershed, Northeastern of
Thailand. M.Sc. Thesis. Faculty of Watershed Management, Kasetsart University,
Heim Jr., R.R., 2002. A review of twentieth-century drought indices used in the
United States. Bull. Am. Meteorol. Soc. 83, 1149e1165.
Homdee, T., Pongput, K., Kanae, S., 2012. Impacts of land cover changes on hydro-
logic responses: a case study of Chi river basin, Thailand. J. Jpn Soc. Civil Eng.
Ser. B1 Hydraul. Eng. 67, I_31eI_36.
Hou, Y.Y., He, Y.B., Liu, Q.H., Tian, G.L., 2007. Research progress on drought indices.
Chin. J. Ecol. 26, 892e897.
Intergovernmental Panel on Climate Change, 2012. Managing the Risks of Extreme
Events and Disasters to Advance Climate Change Adaptation. A Special Report of
Fig. 6. Pearson correlation between the 1e24 mth of Standardized Precipitation Index (SPI), Standardized Precipitation Evapotranspiration Index (SPEI) and Standardized Pre-
cipitation Actual Evapotranspiration Index (SPAEI).
T. Homdee et al. / Agriculture and Natural Resources 50 (2016) 211e219218
Working Groups I and II of the Intergovernmental Panel on Climate Change.
Cambridge University Press, Cambridge, UK, and New York, USA, p. 582.
Jensen, M.E., Burman, R.D., Allen, R.G., 1990. Evapotranspiration and Irrigation
Water Requirements. Am. Soc. Civil Engrs., New York. ASCE Manuals and Re-
ports on Engineering Practices No. 70.
Joetzjer, E., Douville, H., Delire, C., Ciais, P., Decharme, B., Tyteca, S., 2013. Hydrologic
benchmarking of meteorological drought indices at interannual to climate
change timescales: a case study over the Amazon and Mississippi river basins.
Hydrol. Earth Syst. Sci. 17, 4885e4895.
Kanae, S., Oki, T., Musiake, K., 2001. Impact of deforestation on regional precipita-
tion over the indochina peninsula. J. Hydrometeorol. 2, 51e70.
Kim, W., Kanae, S., Agata, Y., Oki, T., 2005. Simulation of potential impacts of land
use/cover changes on surface water ﬂuxes in the Chaophraya River basin. Thail.
J. Geophys. Res. 110, 1e10 .
Kim, W., Komori, D., Cho, J., Kanae, S., Oki, T., 2014. Long-term analysis of evapo-
transpiration over a diverse land use area in Northern Thailand. Hydrological
Res. Lett. 8, 45e50.
Kumar, M.N., Murthy, C., Sai, M.S., Roy, P., 2009. On the use of standardized precipi-
tation index(SPI) for drought intensity assessment. Meteorol. Appl. 16, 381e389.
Ma, M., Ren, L., Singh, V.P., Yang, X., Yuan, F., Jiang, S., 2014. New variants of the
palmer drought scheme capable of integrated utility. J. Hydrol. 519, 1108e1119 .
McCarthy, J.J., 2001. Climate Change 2001: Impacts, Adaptation, and Vulnerability:
Contribution of Working Group II to the Third Assessment Report of the
Intergovernmental Panel on Climate Change. Cambridge University Press,
McKee, T.B., Doesken, N.J., Kleist, J.,1993. The relationship of drought frequency and
duration to time scales. In: Proceedings of the 8
Conference on Applied
Climatology Boston, MA, USA, pp. 179e183.
Meza, F.J., 2013. Recent trends and ENSO inﬂuence on droughts in northern Chile:
an application of the standardized precipitation evapotranspiration index.
Weather Clim. Extrem. 1, 51e58.
Mo, K.C., 2008. Model-based drought indices over the United States.
J. Hydrometeorol. 9, 1212e1230.
Niranjan Kumar, K., Rajeevan, M., Pai, D.S., Srivastava, A.K., Preethi, B., 2013. On the
observed variability of monsoon droughts over India. Weather Clim. Extrem. 1,
National Oceanic and Atmospheric Administration, 2015. Historical El Nino/La Nino
Episodes (1950epresent), Climate Prediction Center. http://www.cpc.ncep.
noaa.gov/products/analysis_monitoring/ensostuff/ensoyears.shtml, 19 October
Oki, T., Kanae, S., 2006. Global hydrological cycles and world water resources. Sci-
ence 313, 1068e1072.
Palmer, W.C., 1965. Meteorological Drought. US Department of Commerce, Weather
Bureau Washington, DC, USA.
Seneviratne, S.I., Nicholls, N., Easterling, D., Goodess, C.M., Kanae, S., Kossin, J.,
Luo, Y., Marengo, J., McInnes, K., Rahimi, M., 2012. Changes in Climate Extremes
and Their Impacts on the Natural Physical Environment. Intergovernmental
Panel on Climate Change Special Report on Managing the Risks of Extreme
Events and Disasters to Advance Climate Change Adaptation. Cambridge Uni-
versity Press, Cambridge, New York, USA, pp. 109e230.
Shefﬁeld, J., Wood, E.F., 2008. Projected changes in drought occurrence under future
global warming from multi-model, multi-scenario, IPCC AR4 simulations. Clim.
Dynam. 31, 79e105 .
Shefﬁeld, J., Wood, E.F., Roderick, M.L., 2012. Little change in global drought over the
past 60 years. Nature 491, 435e438.
Singhrattna, N., Babel, M.S., Perret, S.R., 2012. Hydroclimate variability and long-
lead forecasting of rainfall over Thailand by large-scale atmospheric variables.
Hydrol. Sci. J. 57, 26e41.
Singhrattna, N., Rajagopalan, B., Kumar, K.K., Clark, M., 2005. Interannual and
interdecadal variability of Thailand summer monsoon season. J. Clim. 18,
1697e170 8 .
Smakhtin, V.U., Schipper, E.L.F., 2008. Droughts: the impact of semantics and per-
ceptions. Water Policy 10, 131e14 3.
Stagge, J.H., Tallaksen, L.M., Gudmundsson, L., Van Loon, A.F., Stahl, K., 2015.
Candidate distributions for climatological drought indices (SPI and SPEI). Int. J.
Climatol. 35, 4027e4040.
Suwanabatr, B., Mekhora, T., 2002. Stabilization of Upland Agriculture under El Nino
Induced Climate Risk: Impact Assessment and Mitigation Measures in Thailand.
United Nations Centre for Alleviation of Poverty Through Secondary Crops'
Development in Asia and the Paciﬁc (CAPSA), p. 46.
Svoboda, M.D., Fuchs, B.A., Poulsen, C.C., Nothwehr, J.R., 2015. The drought risk
atlas: enhancing decision support for drought risk management in the United
States. J. Hydrol. 526, 274e286.
Thornthwaite, C.W., 1948. An approach toward a rational classiﬁcation of climate.
Geogr. Rev. 38, 55e94.
Thai Meteorological Department, 2011. Thailand Rainfall Anomaly El NinoeLa Nina
Thai Meteorological Department, 2014. Generalized Monsoon Index (GMI). www.
tmd.go.th/climate/climate.php?FileID¼15, 19 October 2015.
Tsai, C., Behera, S.K., Waseda, T., 2015. Indo-china monsoon indices. Sci. Rep. 5, 8107.
Tsakiris, G., Vangelis, H., 2005. Establishing a drought index incorporating evapo-
transpiration. Eur. Water 9, 3e11.
Van der Schrier, G., Jones, P.D., Briffa, K.R., 2011. The sensitivity of the PDSI to the
thornthwaite and penman-monteith parameterizations for potential evapo-
transpiration. J. Geophys. Res. 116, 1984e2012.
Vicente-Serrano, S.M., Begueria, S., Lopaz-Moreno, J.I., 2010. A multiscalar drought
index sentitive to global warming: the standardized precipitation evapotrans-
piration index. J. Clim. 23, 1696e1718 .
Vicente-Serrano, S.M., Van der Schrier, G., Beguería, S., AzorineMolina, C.,
LopezeMoreno, J.I., 2015. Contribution of precipitation and reference evapo-
transpiration to drought indices under different climates. J. Hydrol. 526,
Wattanakij, N., Thavorntam, W., Mongkolsawat, C., 2006. Analyzing spatial pattern
of drought in the Northeast of Thailand using multietemporal standardized
precipitation index (SPI). In: Proceedings of the 27
ASEAN Conference on
Remote Sensing. Thailand, Bangkok.
Wilhite, D.A., Glantz, M.H., 1985. Understanding the drought phenomenon: the role
of deﬁnitions. Water Int. 10, 111e120.
World Meteorological Organization, 2009. Lincoln declaration on Drought Indices.
World Meteorological Organization. http://www.wmo.int/pages/prog/wcp/
22 April 2011.
Wu, H., Svoboda, M.D., Hayes, M.J., Wilhite, D.A., Wen, F., 2007. Appropriate appli-
cation of the standardized precipitation index in arid locations and dry seasons.
Int. J. Climatol. 27, 65e79.
Yatagai, A., Arakawa, O., Kamiguchi, K., Kawamoto, H., Nodzu, M.I., Hamada, A.,
2009. A 44eyear daily gridded precipitation dataset for asia based on a dense
network of rain gauges. Sci. Online Lett. Atmos. SOLA 5, 137e140 .
T. Homdee et al. / Agriculture and Natural Resources 50 (2016) 211e219 219