Evolution of meteorological drought characteristics in Vietnam during the 1961–2007 period

ORIGINAL PAPEREvolution of meteorological drought characteristics Hang Vu-Thanh&Thanh Ngo-Duc&Tan Phan-Van Received: 5 April 2013 / Accepted: 9 December 2013 # Springer-Verlag Wien 2013

ORIGINAL PAPER

Evolution of meteorological drought characteristics

Hang Vu-Thanh&Thanh Ngo-Duc&Tan Phan-Van

Received: 5 April 2013 / Accepted: 9 December 2013

# Springer-Verlag Wien 2013

Abstract The drought conditions over the seven

sub-climatological regions in Vietnam are examined using three

meteorological drought indices: de Martonne J, PED, and

Standardized Precipitation Index (SPI) According to the

season-al probabilities of drought occurrence estimated by the de

Martonne index, droughts mainly occur between November

and March in all the sub-regions The PED index and the SPI

index generally show high probabilities of drought occurrence

from April to August and from May to October, respectively In

the southern sub-regions of Vietnam, droughts more frequently

occur in El Niño years and wet conditions are more frequently

observed in La Niña years However, such El Niño–Southern

Oscillation influences are not clearly observed in the northern

sub-regions During 1961–2007, droughts significantly increased

in the northern part of Vietnam In the southern regions, PED

shows increasing drought conditions while J and SPI show

decreasing drought trends for almost all the stations

1 Introduction

Drought is a temporary natural disaster, which originates from a

lack of precipitation over an extended period compared to

long-term average conditions and can bring significant economic

losses Generally, droughts can be classified into four types: (1)

meteorological drought; (2) agricultural drought; (3) hydrological

drought; and (4) socioeconomic drought (Heim2002) Among

them, meteorological droughts are usually related to the

deficien-cies of precipitation over a specific region Although it is still

impossible to avoid meteorological droughts, they can be

predicted and monitored so that their impacts can be alleviated Assessing and monitoring droughts are normally performed using drought indices

A number of drought indices have been developed and widely used to date Drought indices are usually defined as functions of rainfall and/or river discharge or other measurable hydro-meteorological variables Among them, rainfall is the main factor controlling the drought formation and its duration (Palmer1965) Oladipio (1985) showed that the indices that were based only on precipitation data performed well when compared to more com-plex hydrological ones Friedman (1957) identified four basic criteria that any drought index should meet: (1) the timescale should be appropriate to the problem at hand; (2) the index should be a quantitative measure of large-scale, long-continuing drought conditions; (3) the index should be applicable

to the problem; (4) a long accurate past record of the index should be available or computable In operational drought mon-itoring, one should add a fifth criterion: (5) the index should be able to be computed on a near real-time basis Several reviews of drought indices can be found in Alley (1984), Wu et al (2001), and Smakhtin and Hughes (2004) The inherent complexity of drought phenomena implies that no drought index is ideal for all regions (Morid et al 2006) Therefore, to assess the drought conditions of a specific region, it is useful to consider different indices The choice of drought indices is normally based on the climate data availability and the index ability in detecting drought characteristics (Morid et al.2006)

During an El Niño–Southern Oscillation (ENSO) event, droughts can occur virtually anywhere in the world in varying degrees of magnitude The impacts of ENSO on meteorological quantities and drought indices around the globe have been reported in several previous studies (e.g., Ropelewski and Halpert1987; Rajagopalan et al.2000; Lyon2004) For example, Ropelewski and Halpert (1987) pointed out different regions in the world that are prone to droughts during ENSO They con-cluded that ENSO events seem to have a stronger influence on

Department of Meteorology, Hanoi College of Science,

Vietnam National University, 334 Nguyen Trai, ThanhXuan,

Hanoi, Vietnam 10000

e-mail: hangvt@vnu.edu.vn

DOI 10.1007/s00704-013-1073-z

regions in the lower latitudes, especially in the equatorial Pacific

and bordering tropical areas Juneng and Tangang (2005) showed

that the Southeast Asia rainfall anomalies depend strongly on

phases of ENSO, and those anomalies are linked with the

evo-lution of sea surface temperature anomalies as well as with the

anomalous low-level circulation Lyon (2004) also showed that

severe droughts developed during El Niño events in the tropics

and the severity and extent of tropical droughts clearly have a

relationship to the strength of the El Niño events

The climate of Vietnam is strongly affected by

mon-soons and its complex topography Based on the differences

in radiation, temperature, and rainfall conditions, Vietnam is

classified into seven sub-regions, consisting of the North

West (B1), North East (B2), North Plain (B3), North

Central (B4), South Central (N1), Central Highland (N2),

and the South Plain (N3) regions (Fig 1) (Phan et al

2009) Over the sub-regions, the rainy season usually starts

in May and ends in December The total precipitation

amount in the rainy season accounts for approximately

75–85 % of the annual precipitation (Ngu and Hieu

2004) The average annual rainfall over the country is

approximately 1,400–2,400 mm and can vary from 700

up to 5,000 mm, depending on each region

The linkages between ENSO and some climate elements in

Vietnam have been documented Yen et al (2011) showed that

Central Vietnam has less (more) rainfall in El Niño (La Niña)

years Nguyen-Thi et al (2012) analyzed tropical cyclone

rainfall ratio and concluded that during El Niño (La Niña)

years, the tropical cyclone rain ratio in Central Vietnam has a significant decrease (increase) in October–November Presently, there are few publications dealing with the drought characteristics in Vietnam This study focuses on analyzing variations and trends of droughts in the seven sub-regions in Vietnam during 1961–2007 Performance of differ-ent indices in represdiffer-enting drought conditions is examined Besides, the influences of ENSO on droughts in Vietnam are discussed

2 Data and drought indices

2.1 Data

Daily near surface temperature and rainfall data of 50 meteo-rological stations during 1961–2007 were provided by the National Hydro-Meteorological Service of Vietnam Locations of the stations are displayed in Fig.1 A list of the seven sub-regions and their stations are described in Table1 Missing data is replaced by−99.00 and is excluded from the calculation

2.2 Drought indices

In this study, three drought indices based on temperature and precipitation data are used Details of the indices are described below

(1) The de Martonne J index (de Martonne1926) The de Martonne index (mm/°C) for monthly values

is defined as

J¼ 12P

Tþ 10

where P and T are monthly total precipitation (mm) and monthly mean near surface air temperature (°C), respectively

To estimate the annual values of J, the below expression is applied:

J¼ Pa

Taþ 10

where Paand Taare annual total precipitation (mm) and annual mean near surface air temperature (°C), respectively

The J index was introduced by de Martonne (1926) to characterize the aridity of a given climate Low values of J represent dry conditions while higher values represent wet conditions When J is less than 30, moderate drought condi-tions can be observed A value of less than 20 for J typically indicates severe drought

(2) The Standardized Precipitation Index (SPI) (McKee et al

1993)

B1

B2

B3 B4

N1 N2

N3

102E 104E 106E 108E 110E 112E

8N

10N

12N

14N

16N

18N

20N

22N

24N

0 500 1000 1500 2000 2500

3000 m

Fig 1 Spatial distribution and topography (m) of stations included in the

analysis in the seven sub-regions in Vietnam

The formula of SPI is described as

Table 1 List of the meteorological stations in the seven sub-regions in

Vietnam

Sub-region 1: North West (B1)

Sub-region 2: North East (B2)

Sub-region 3: North Plain (B3)

Sub-region 4: North Central (B4)

Sub-region 5: South Central (N1)

Sub-region 6: Central Highland (N2)

Table 1 (continued)

Sub-region 7: South Plain (N3)

0.0 20.0 40.0 60.0 80.0 100.0

Jan Feb Ma

Apr Ma

Jun Jul Aug Sep Oc

B2 B3 B4 N1 N2 N3

15.0 20.0 25.0 30.0 35.0

Jan Feb Ma

Apr Ma

Jun Jul Aug Sep Oc

B2 B3 B4 N1 N2 N3

0.0 5.0 10.0 15.0 20.0

Jan Feb Ma

Apr Ma

Jun Jul Aug Sep Oc

B2 B3 B4 N1 N2 N3

a J index

b Ped index

c SPI index

Fig 2 Seasonal occurrence probability of drought conditions estimated

by the a J index, b PED index, and c SPI index in the seven sub-regions in

where P and are precipitation and average precipitation (mm)

in a given time period, respectively, and σ is the standard

deviation of precipitation According to Eq (3), SPI can quantify

the degree of wetness in different time periods (e.g., 1 month,

3 months, 6 months, 1 year, 2 years, and so on) depending on

user application The SPI value normally ranges from (−2.0)

to (+2.0) A value of (+2.0) indicates extremely wet

condi-tions; (1.5) to (1.99) indicates very wet; (1.0) to (1.49)

mod-erately wet; (0.99) to (−0.99) near normal; (−1.0) to (−1.49)

moderately dry; (−1.5) to (−1.99) severely dry; and an SPI

value of (−2.0) or (less) indicates extremely dry conditions

(McKee et al.1993)

(3) The PED index (Ped1975) The PED index is defined as

PED¼ΔTσ

T −ΔPσ

whereΔT and ΔP are anomalies of near surface air tempera-ture (°C) and precipitation (mm) in a given time period andσT andσPare standard deviations of temperature and precipita-tion, respectively A negative PED value represents a wet period An insignificant drought may occur if PED is between

1 and 2 When PED is between 2 and 3, a moderate drought

0.0 20.0 40.0 60.0 80.0 100.0

0.0 20.0 40.0 60.0 80.0 100.0

0.0 20.0 40.0 60.0 80.0 100.0

Fig 3 Occurrence probability of

drought conditions estimated by

the annual a J index, b PED index,

and c SPI index in the seven

2007

occurs A greater value of PED (PED>3) indicates severe

drought conditions

3 Results and discussions

3.1 Seasonal drought conditions

The probabilities of drought occurrence estimated with the

three indices in the seven sub-regions in Vietnam during

1961–2007 are represented in Fig.2

According to the J index (Fig.2a), the occurrence

proba-bilities of drought conditions over Vietnam are highly

con-centrated from November to March, which is in accordance

with the dry seasons in almost all sub-regions The high

drought probabilities by more than 90 % are found in

January in N2 and N3 It is due to little rainfall at some stations

during this month Except for B4 and N1, drought occurrence

has a low probability from May to September compared to

other months During summer, Central Vietnam, including the

B4 and N1 sub-regions, experiences the foehn effect leading

to more frequent hot days and aridity

According to the PED index (Fig 2b), the occurrence

probabilities of drought conditions are smaller compared to

those of the J index These values are in the range of over 15 %

to more than 30 % It can be seen that the high probabilities of

drought occurrence are found in April except for B2 and B3

sub-regions During the period of June–August, the

probabil-ities of drought occurrence vary from 25 to 30 % and have no

big differences between sub-regions The lower values are

normally seen in November and December in almost all

sub-regions

The probabilities of drought occurrence according to the

SPI index are smaller to those of the J and PED indices

(Fig.2c) The high probabilities of drought occurrence are

from May to October in almost sub-regions The lower

prob-abilities of drought occurrence are often found from

November to January Droughts do not occur in December

in B1, from November to January in B3, and from January to

March in N3

A comparison of the three indices shows that the J index,

which is calculated directly from monthly precipitation and

temperature, indicates high probabilities of drought

occur-rence mostly in dry seasons in almost all sub-regions

However, the PED and SPI indices are examined

includ-ing the variations of precipitation (and temperature) durinclud-ing

the period so that the high probabilities of drought

occur-rence are even found in wet seasons In general, the high

probabilities of drought occurrence are from November to

March according to the J index, from April to August

according to the PED index and from May to October

according to the SPI index

3.2 Annual drought conditions The occurrence probabilities of all conditions estimated by the three indices on an annual timescale are represented in Fig 3 The conditions are grouped only in wet (W), moderate drought (MD), and severe drought (SD) con-ditions It can be seen in Fig 3a that the annual J index often shows wet conditions over the country because of the high total annual precipitation in the tropics, which leads to high values in the J index according to Eq (2) MD is more frequently observed in N1 and B3 while SD rarely occurs over the country

According to the PED index (Fig.3b), one can note that the

MD condition can occur in all sub-regions in which the highest probability of MD is 24.4 % in N1 and the lowest probability of MD is 17.5 % in B1 SD can occur with low probabilities in all sub-regions except N1

The occurrence probability of drought conditions es-timated by the SPI index (Fig 3c) shows that MD may occur with lower probabilities but SD occurs with higher probabilities compared to those of the PED index

in all sub-regions during 1961–2007 The probabilities

of MD and SD occurrences do not have significant differences among the sub-regions

Generally, the J index is not useful to determine drought conditions on annual timescales due to the high annual pre-cipitation in the tropics The occurrence probabilities of MD estimated by the PED index are generally greater than those estimated by the SPI index are, while it is vice versa with the occurrence probabilities of SD

Table 2 List of the ENSO years and their intensities in the period

3.3 Linkages with ENSO

In this section, the relationships between drought features and

ENSO will be examined Information on ENSO years and

intensities based on the Oceanic Niño Index (ONI) (Kousky

and Higgins 2007) are available on the Climate Prediction

Center's (CPC's) website (http://www.cpc.ncep.noaa.gov)

(Table2)

The J index in the seven sub-regions for the El Niño, La

Niña, and non-ENSO years are displayed in Fig 4a It is

shown that all annual J values in the four sub-regions from

B1 to B4 indicate wet conditions (i.e., J≥30) because of a

relatively high annual precipitation in these regions The

ENSO years do not have a clear effect on J especially in the

B1 and B2 sub-regions J had a slightly decreasing tendency

during 1961–2007 In B3 and B4, it is found that there is less precipitation in some El Niño years resulting in less wet conditions Some small values of J mainly occur in the El Niño years such as J=35 in 1969, J=32.8 in 1991 in B3; and J=35.7 in 1968, J=39.1 in 1969 in B4 Figure4aalso shows

an insignificant decreasing trend of J in B3 and an increasing trend in B4 This means that dry conditions increase in B1, B2, and B3 while they decrease in B4 In the N1 and N2 sub-regions, the de Martonne index significantly increases while there is almost no change in N3 In N1, the J index is rather small in some strong El Niño years such as J=19.68 in 1965, J=24 in 1972 and high in the strong La Niña year such as J=90.93 in 1999 In N2, J is also small in the El Niño years such as in 1963 (J=26.11) and in 1976 (J=19.76) The min-imum value of J is equal to 28.83 in the weak El Niño year of

a J index

b PED index

c SPI index

Fig 4 The a J index, b PED

index, and c SPI index in ENSO

years in the seven sub-regions of

years are marked with the tiny

square pattern; La Niña years are

marked with the up left to down

right pattern Areas without

patterns represent non-ENSO

years

1977 and was at a maximum of 68.09 in the strong La Niña

year of 1999 in N3 In general, the southern sub-regions are

more clearly affected by the ENSO events, in which lower/

higher values of J occur more frequently in the El Niño/La

Niña years, corresponding to the drier/wetter conditions over

the regions, respectively

The inter-annual changes of the PED index and the

link-ages with the ENSO years are shown in Fig.4b It is noted that

there are homogeneous increasing trends of the PED values

during 1961–2007 in all sub-regions, indicating an increase of

drought conditions over the whole country The effects of the

ENSO years on the PED index do not show a systematic

behavior In B1, some relatively high values of PED occur

in El Niño years such as PED=2.62 in 1987, PED=2.35 in

1991, and PED=2.16 in 2006 However, the maximum of

PED (PED=3.19) is found in the moderate La Niña year of

1998 The same situations as in B1 are also realized in B2, B3,

and B4 In the southern sub-regions, drought conditions

main-ly occur in the El Niño years and wet conditions are frequentmain-ly observed in the La Niña years Nevertheless, some exceptions exist here and there, for example, the significant negative values of PED found in the non-ENSO year of 1996 and the high positive PED values found in the La Niña year of 1998 Similar to the impacts on the J index, the ENSO years also have significant influences on SPI in the sub-regions from B4

to N3 (Fig.4c) SPI is rather small (i.e., drought conditions may occur) in the El Niño years and rather high in the La Niña years In the B1, B2, and B3 sub-regions, such ENSO influ-ences are not clearly observed

3.4 Long-term drought trends The trends of the three indices for the 1961–2007 period are computed using Sen's method (Sen 1968) and the

thick-black contours indicate that the trends are significant at a 10 % level Warm/cold colors indicate more/less severe drought conditions

Fig 6 Number of stations listed

its linear trend, estimated by using

the J index (black line), the PED

index (red line), and the SPI index

(blue)

parametric Mann–Kendall test (Kendall1975) is used to

ex-amine the trend-significant level (Fig.5) Briefly, Sen's trend

of a data series (x1, x2,…, xn), where xirepresents the value at

the time i, is the median of the series composed of n(n−1)/2

elements {, k=1,2,…,n−1; j>k}

Figure5shows that all the three indices indicate a

signif-icant increase of drought conditions in the north of Vietnam

with the highest rate occurring in the B3 sub-regions In the

southern sub-regions, PED shows increasing conditions while

J and SPI show decreasing drought trends for almost all the

stations This difference can be explained by the fact that the

increase of temperature over the regions can well compensate

the increase of precipitation (Endo et al.2009) in the PED

formula, leading to an increase of PED during 1961–2007

One should note that using the daily precipitation data of

the second half of the twentieth century, Endo et al (2009)

showed that precipitation increased in the south while it

de-creased in the north of Vietnam Therefore, the opposite

drought trends between the north and south of Vietnam

indi-cated by the J and SPI indices will follow the precipitation

trends

The number of meteorological stations listed in Table1

affected by drought during 1961–2007 is represented in Fig.6

According to the J and SPI indices, the number of affected

stations has a slightly increasing linear trend while a

decreas-ing linear trend is detected accorddecreas-ing to the PED index As an

affected station represents a certain area that has drought

conditions, an increase in the number of affected stations

(estimated by using the J and SPI indices) represents an

increase of the drought-affected area over time

4 Conclusions

This study has examined the drought conditions over the

seven sub-regions of Vietnam using the J, PED, and SPI

indices According to the de Martonne J index, droughts can

occur with high probabilities from November to March, from

April to August according to the PED index, and from May to

October according to the SPI index in almost all sub-regions

On annual timescales, the J index is not useful to determine

drought conditions because of the high total annual

precipita-tion in the tropics, which leads to high values in the J index

Moderate drought (MD) occurrence probabilities estimated by

the PED index are generally greater than those estimated by

the SPI index and vice versa in the case of severe drought

(SD)

The inter-annual analysis of the three indices in the

south-ern sub-regions of Vietnam revealed that drought conditions

mainly occur in the El Niño years and wet conditions are

frequently observed in the La Niña years However, such

ENSO influences are not clearly observed in the northern

sub-regions

During the period of 1961–2007, drought conditions sig-nificantly increase in the northern part of Vietnam In the southern regions, PED shows an increasing trend while J and SPI show decreasing drought trends for almost all the stations An increase in the number of stations affected by drought estimated by using the J and SPI indices represents an increase of the drought-affected area over time

Due to the lack of socioeconomic and agricultural drought records over the seven sub-regions, this study has not aimed to identify which meteorological drought indices would be the best to represent the drought conditions in Vietnam This question remains unanswered and is an important challenge for a further study in order to better understand the relation-ships between the different drought types in Vietnam

Min-istry of Science and Technology Foundation (DT.NCCB-DHUD.2011-G/ 10), the 11-P04-VIE DANIDA project, and the Vietnam National Foun-dation for Science and Technology Development (NAFOSTED, code 105.06-2013.03).

References

Alley WM (1984) The Palmer Drought Severity Index: limitations and

Endo N, Matsumoto J, Lwin T (2009) Trends in precipitation extremes

Friedman D.G (1957) The prediction of long-continuing drought in south and souhtwest Texas Occational Papers in Meteorology, No.1, The Travelers Weather Research Center, Hartford, CT, 182 pp Heim RR (2002) A review of twentieth-century drought indices used in

Juneng L, Tangang FT (2005) Evolution of ENSO-related rainfall

Kendall MG (1975) Rank Correlation Methods Charles Griffin, London,

272 pp Kousky VE, Higgins RW (2007) An alert classification system for

371 Lyon B (2004) The strength of El Niño and the spatial extent of

2004GL020901

McKee T.B., Doesken N.J., Kleist J (1993) The relationship of drought frequency and duration to time scales Preprints, Eighth Conference

Morid S, Smakhtin VU, Moghaddasi M (2006) Comparision of seven meteorological indices for drought monitoring in Iran Int J Climatol

Ngu ND, Hieu NT (2004) Vietnam Climate and Climatic Resources Agriculture Publisher, Hanoi, 296 pp (in Vietnamese)

Nguyen-Thi HA, Matsumoto J, Ngo-Duc T, Endo N (2012) A

044 Oladipio EO (1985) A comparative performance analysis of three

Palmer W.C (1965) Meteorological drought Research Paper 45, U.S.

Department of Commerce, Weather Bureau, Washington D.C., 58 pp.

Ped DA (1975) On indicators of droughts and wet conditions (in

Phan VT, Ngo-Duc T, Ho TMH (2009) Seasonal and interannual

Rajagopalan B, Cook E, Lall U, Ray BK (2000) Spatiotemporal

variabil-ity of ENSO and SST teleconnections to summer drought over the

Ropelewski CF, Halpert MS (1987) Global and regional scale

precipita-tion patterns associated with the El Niño-Southern Oscillaprecipita-tion

Sen PK (1968) Estimates of the Regression Coefficient Based on

Smakhtin V.U., Hughes D.A (2004) Review, Automated Estimation and Analyses of Drought Indices in South Asia IWMI

Colombo; 24.

Wu H, Hayes MJ, Welss A, Hu Q (2001) An evaluation the standardized precipitation index, the china-Z index and the statistical Z-score Int

Yen MC, Chen TC, Hu HL, Tzeng RY, Dinh DT, Nguyen TTT, Wong CJ (2011) Interannual variation of the fall rainfall in Central Vietnam J