Standardized Precipitation Index (SPI) forecast

and its relevance

P. Guhathakurta
 SPI and Drought Monitoring

What is Drought?
Drought is a deficiency in precipitation over an extended period, usually a month
or season or more, resulting in extensive damage to crops, loss of yield,
resulting in a water shortage causing adverse impacts on vegetation, animals,
and/or people.

Meteorological Agricultural
Hydrological
drought drought
drought

CAUSE IMPACT

It is a normal, recurrent feature of climate that occurs in

virtually all climate zones, from very wet to very dry.
Type of Drought and time response

1 to 2-months

1 to 4 months

4 to 12 months or more
Interrelationships between meteorological, agricultural, hydrological and socio-economic
drought. Source: National Drought Mitigation Center, University of Nebraska–Lincoln, USA)
 Standardized Precipitation Index (SPI) in Drought Monitoring

 The Standardized Precipitation Index (SPI) is a tool which was developed

primarily for defining and monitoring drought (McKee et al 1993). It allows an
analyst to determine the rarity of a drought at a given time scale (temporal
resolution) of interest for any region with historic data. It can also be used to
determine periods of anomalously wet events.

 In 2009, the participants at the Inter-Regional Workshop on Indices and Early

Warning Systems for Drought held at the University of Nebraska-Lincoln issued
“The Lincoln Declaration on Drought Indices” (Hayes et al., 2011). There were
fifty-four experts from all regions agreed on the use of a universal
meteorological drought index for more effective drought monitoring and climate
risk and gave recommendation to WMO

 The World Meteorological Organization (WMO) recommends, that all national

meteorological and hydrological services should use the SPI for monitoring of
dry spells (Press report December 2009, WMO No. 872).

 The World Meteorological Organization (WMO) releases “Standardized

Precipitation Index User Guide”, WMO-No. 1090 in 2012 giving details and some
key points: about using SPI. It is desirable one should go through this before
applying SPI.
 Interpretation:
 For each time step, precipitation of the preceding t months is accumulated, where t is
referred to as the time scale. The time series is first fitted with a model distribution to
the data (for precipitation series, the Gamma distribution is typically used)
Subsequently it is transformed to values of the standard normal distribution for each
calendar month separately.

 The resulting time series has no seasonality and takes values of the standard normal
distribution, where negative values indicate below average water availability. Positive
values indicate greater than median rainfall; negative values indicate less than median
rainfall.

 The interpretation of SPI is strictly probabilistic. As SPI has units of the standard
normal distribution, its values can be directly related to probability of occurrence. The
standardization implies also that SPI is independent of the mean conditions, it do not
differs in magnitude between different regions and is thus suitable to analyse the
synchronicity of drought events. This enables an easy and direct comparison between
locations with different climate. The possibility to compute SPI for different time
scales (t) allows for an adaptation of the index to slowly or fast evolving
environmental or societal systems.

 SPI is straight forward to interpret as negative (positive) values are directly related to a
shortage(surplus) of water availability at a given location relative to the normal
conditions. As SPI is probabilistic index it has a straight forward relation to occurrence
frequencies. For example, extreme droughts (SPI ≤ -2), are expected to occur with a
chance of 2.3%.
Example of equiprobability transformation from fitted
gamma distribution to the standard normal distribution.
 Positive SPI values indicate greater than median precipitation and negative values
indicate less than median precipitation. Because the SPI is normalized, wetter and
drier climates can be represented in the same way; thus, wet periods can also be
monitored using the SPI.
 A drought event occurs any time when the SPI is continuously negative
and reaches an intensity of -1.0 or less. The event ends when the SPI
becomes positive. Each drought event, therefore, has a duration defined
by its beginning and end, and an intensity for each month that the event
continues.
India Meteorological Department monitors meteorological drought in India using two
well defined indices viz.
1. Aridity Anomaly Index (AAI)
2. Standardized Precipitation Index (SPI)
Both the indices can also identifies agricultural drought while the second one can also be
considered for monitoring hydrological drought if period consider is 6 –12 months.
 IMD has started monitoring drought situations in district scale using SPI since January,
2013 in monthly scale as a part of climate monitoring.
 To meet the demand from Agricultural Division of IMD for its National level Weekly
agricultural Advisory, we have started producing SPI district map in every week based
on cumulative recent 4 weeks since June 2013.
 Standardized Precipitation Index (SPI) for the cumulative four weeks period is being
generated operationally in every week (Thursday/Friday) to identify the districts
experiencing moisture stress situation for preparation of appropriate agromet
advisories.
 We have also started generating Standardized Precipitation Index weekly forecast
since the 1st week of Southwest monsoon 2014 using IMD GFS/MME weekly district
rainfall forecast as input.
 Drought situation during the SW monsoon season 2014 and 2015 : Monthly District
SPI maps
JUNE JULY AUGUST SEPTEMBER

2014 58% 90% 91% 108%

78
2015 116% % AIND RF 84%
78% 76 %
%
District SPI for the cumulative months during the SW monsoon season 2014 and 2015
2 MONTHS JUNE-JULYY 3 MONTHS JUNE TO AUGUST 4 MONTHS JUNE TO SEPTEMBER

2014
88%
79% 83%

2015 95 %
89%
86 %
 Categories(SPI range) Percentage Area No. of Districts

Extremely wet(2.00 or more) 3.40% 13

JUNE Severely wet(1.50 to 1.99) 3.96% 24

Moderately wet(1.00 to 1.49) 15.84% 72

Mildly wet(0 to 0.99) 46.77% 250

Mildly dry(0 to -0.99) 25.48% 192

No. of Districts and Moderately dry(-1.00 to -1.49) 3.53% 23

percentage of area of Severely dry(-1.50 to -1.99) 0.63% 7

the country under Extremely dry(-2.00 or less) 0.35% 4

different categories in
July June-July
2015
Percentage No. of Percentage No. of
Categories(SPI range)
Area Districts Area Districts
Extremely wet( 2.00 or more) 5.07% 18 5.33% 22

Severely wet(1.50 to 1.99) 3.6% 19 3.78% 14

Moderately wet(1.00 to 1.49) 8.19% 37 8.89% 39

CUMMULATIVE 2 MONTHS Mildly wet(0 to 0.99) 14.18% 107 21.56% 149

JUNE-JULYY Mildly dry(0 to -0.99) 33.66% 227 40.87% 246

Moderately dry(-1.00 to -1.49) 13.86% 81 10.17% 61

Severely dry(-1.50 to -1.99) 9.34% 52 6.5% 35

Extremely dry(-2.00 or less) 12.09% 45 2.91% 19

 August June-August
AUGUST Categories(SPI range) Percentage Area
No. of
Percentage Area
No. of
Districts Districts
Extremely wet( 2.00 or more) 0.38% 4 0.93% 6

Severely wet(1.50 to 1.99) 0.66% 7 2.33% 14

Moderately wet(1.00 to 1.49) 2.81% 22 8.32% 34

Mildly wet(0 to 0.99) 23% 122 24.58% 130

No. of Districts
Mildly dry(0 to -0.99) 281 36.21% 245
and percentage 48.24%

of area of the Moderately dry(-1.00 to -1.49) 14.97% 90 14.54% 82

country under Severely dry(-1.50 to -1.99) 6.2% 35 10.54% 52

different Extremely dry(-2.00 or less) 3.74% 25 2.55% 22

categories in September June- September

2015 Categories(SPI range)
Percentage No. of Percentage No. of
Area Districts Area Districts
Extremely wet( 2.00 or more) 0% 0 0.45% 3
Severely wet(1.50 to 1.99) 0.75% 4 1.22% 8
Moderately wet(1.00 to 1.49) 4.69% 22 6.60% 27
SEPTEMBER Mildly wet(0 to 0.99) 30.54% 142 24.19% 114
Mildly dry(0 to -0.99) 35.76% 225 37.12% 234
Moderately dry(-1.00 to -1.49) 13.72% 88 14.62% 99
Severely dry(-1.50 to -1.99) 8.7% 59 11.02% 63
Extremely dry(-2.00 or less) 5.84% 44 4.77% 35
Comparative Statistics
for the Southwest SW monsoon 2015 SW monsoon 2014
monsoon 2014 & 15

June- September 2015 June- September 2014

Percentage No. of Percentage No. of

Categories(SPI range)
Area Districts Area Districts
Extremely wet( 2.00 or more) 0.45% 3 3
0.27% Total 30.4% area of
Severely wet(1.50 to 1.99) 1.22% 8 1.27% 11 the country was
Moderately wet (1.00 to 1.49) 6.60% 27 2.84% 18
under dry condition
during SW monsoon
Mildly wet(0 to 0.99) 24.19% 114 26.61% 137
2015 compare to
Mildly dry(0 to -0.99) 37.12% 234 40.89% 228 28.11% that of 2014.
Moderately dry(-1.00 to -1.49) 14.62% 99 14.02% 88

Severely dry(-1.50 to -1.99) 11.02% 63 7.78% 54

Extremely dry(-2.00 or less) 4.77% 35 6.31% 48

National Agricultural Drought Assessment and Monitoring
(NADAM) Reports of Mahalanobis National Crop Forecast
Centre, Department of Agriculture & Cooperation (NCFC)
are also using SPI products of IMD for comparison of drought
assessment.
 Districts experienced two years consecutive drought
during 2014 and 2015
There were 190 and 197 districts were under drought
conditions ( all categories i.e. moderate, severe and extremely
severe together) during the year 2014 and 2015. List of the
districts which have experienced drought conditions in
consecutive two years i.e. 2014 and 2015 is presented in next
slide.

Total 105 number of districts which contribute 15 % of country

area were affected by consecutive two years of drought
situations
 State No. of districts District State No. of districts District State No. of districts District
AGRA BEED
HYDERABAD ALLAHABAD NANDED
Telangana 4 MEDAK AMBEDKAR NAGAR OSMANABAD
Maharashtra 6
NIZAMABAD AURAIYA PARBHANI
RANGAREDDY AZAMGARH SATARA
EAST KAMENG BALLIA LATUR
Arunachal Pradesh 2 BANDA
TIRAP
BARABANKI Meghalaya 1 RI-BHOI
CHANDAULI
Assam 1 SIBSAGAR
DEORIA FEROZEPUR
ARARIA ETAH HOSHIARPUR
Punjab 4
GOPALGANJ ETAWAH JALANDHAR
NAWADA FAIZABAD MANSA
Bihar 5
PURNEA FARRUKHABAD CHATRA
FATEHPUR DHANBAD
SAHARSA
GONDA GARHWA
DANGS GORAKHPUR Jharkhand 5 PALAMU
Gujarat 3 NARMADA HAMIRPUR
RANCHI
TAPI MAHAMAYA NAGAR
AMBALA JALAUN
Uttarakhand 1 UDHAM SINGH NAGAR
JAUNPUR
MAHENDRAGARH JHANSI JANJGIR
Uttar Pradesh 44
ROHTAK JYOTIBA PHULE NA Chhathisgarh 3 JASHPUR
BHIWANI KANNAUJ SURGUJA
Haryana 9 KURUKSHETRA KANPUR CITY
KAITHAL KANPUR DEHAT
PANCHKULA KAUSHAMBI
FATEHABAD KUSHI NAGAR
LUCKNOW
PALWAL
MAU
KINNAUR MAINPURI
Himachal Pradesh 2 MAHARAJGANJ
SIRMAUR
MEERUT
BIDAR MIRZAPUR
Karnataka 2 PILIBHIT
YADGIR
PRATAPGARH
BALAGHAT RAE BAREILLY
BHIND RAMPUR
CHHATARPUR SHAHJAHANPUR
DAMOH SITAPUR
JABALPUR SULTANPUR
MORENA SONBHADRA
NARSINGHPUR UNNAO
Madhya Pradesh 13
PANNA
MAHOBA
REWA
SAGAR
SEONI
SHAHDOL
SINGRAULI
Area weighted SPI values of the States for Jun-Sep 2014 & 2015
calculated from the district SPI values
AREA WEIGHTED SPI AREA WEIGHTED SPI
STATE STATE

Jun-Sep 2014 Jun-Sep 2015 Jun-Sep 2014 Jun-Sep 2015

ANDAMAN & NICOBAR -0.21 0.02 MADHYA PRADESH -0.78 -0.47
ANDHRA PRADESH -0.71 0.34 MAHARASHTRA -0.60 -1.07
ARUNACHAL PRADESH -0.17 -0.03 MANIPUR -1.61 -0.21
ASSAM -0.21 -0.43 MEGHALAYA -0.14 -0.23
BIHAR -0.71 -1.11 MIZORAM -1.12 -0.53
CHHATHISGARH -0.37 -0.75 NAGALAND -0.94 -0.45
DELHI -1.34 0.27 ORISSA 0.42 -0.55
GOA 0.52 -0.68 PUNJAB -1.30 -0.64
GUJARAT -0.05 -0.11 RAJASTHAN 0.13 0.46
HARYANA -1.59 -0.85 SIKKIM 0.54 0.76
HIMACHAL PRADESH -1.18 -0.53 TAMILNADU -0.14 -0.17
J&K 0.79 0.88 TELANGANA -1.04 -0.70
JHARKHAND -0.73 -0.66 TRIPURA 0.08 0.18
KARNATAKA 0.36 -0.39 UTTAR PRADESH -1.74 -1.58
KERALA 0.29 -1.18 UTTARAKHAND -0.37 -0.43
LAKSHADWEEP -0.19 -0.53 WEST BENGAL -0.50 0.27
All India Drought years based on SPI
EXTREMELY SEVERE DROUGHT YEAR (4) : 1965, 1972, 2002 AND 2009
SEVERE DROUGHT YEAR (5) : 1901, 1905, 1911, 1918, 1979 AND 1987
MODERATE DROUGHT YEAR (14) : 1902, 1904,1907,1913, 1915, 1920, 1941,
1951, 1966,1974, 1982, 2004, 2014, 2015
Consecutive two years : 1901-02 ; 1904 - 05; 1965 -66; 2014-15.
 WEEKLY SPI MONITORING AND FORECAST

• In this forecast we have used Actual district rainfall for the

past three weeks and the district cumulative rainfall forecast
based on IMD Global Forecasting system (GFS)/MME model
for the coming week.
• The base period of the SPI computation is 1901-2000.
• Thus SPI forecast is generated every Thursday/Friday for the
cumulative four weeks period consisting of one advance week
along with past three weeks to identify the regions with
prevailing or beginning/ending of the
extremely/severely/moderately dry/wet conditions.

• The SPI forecast maps were uploaded in IMD Pune website in

every Friday.
Whether one week can have significant contribution in
changing drought situation?
In the year 2015 on several occasions dry/wet conditions largely
changes in a week. For example from the week 32 to 33,
number of extremely severe dry conditions districts decreased
from 42 to 20 and severely dry condition from 56 to 45; from
week no. 36 to 37 opposite happened as number of extremely
severe dry conditions districts increased from 24 to 72 and
severely dry condition from 84 to 102. Again next week i.e. in
week no. 38 it decreased to 24 and 52 respectively. The SPI
forecast in all these cases has indicated the changes in
drought situation one week advance. Thus this type of SPI
forecast is particularly helpful in these cases to identify the
districts well in advance (one week) for the starting or cessation
of drought conditions.
 Week No. 22 ( Input rainfall forecast GFS)

FORECAST
ACTUALL

Forecast
Categories
1 2 3 4 5 6 7 8 Total
1 2 1 0 0 1 0 0 0 4
2 1 6 4 2 0 0 0 0 13
3 0 3 11 10 0 0 1 0 25
4 0 4 11 198 18 3 1 1 236 Percentage of Forecast within + 1
Actual

5 0 0 3 61 136 9 3 2 214 Category : 95.0%

6 0 0 0 4 16 16 3 0 39
7 0 0 0 0 2 4 11 1 18
8 0 0 0 1 0 0 2 4 7
Total 3 14 29 276 173 32 21 8 556
Correc In + 2 In + 4 In + 5 In + 6 In + 7
In +1 cat In + 3 cat
t F/C cat cat cat cat cat
Freq 384 144 21 3 4 0 0 0
% 69.1 25.9 3.8 0.5 0.7 0.0 0.0 0.0
 Week No. 23 GFS rainfall Forecast was not available

Week No. 24 GFS rainfall Forecast was

not available
From week 32 to 33, number of extremely severe dry conditions districts decreased from 42 to 20
and severely dry condition from 56 to 45
From week no. 36 to 37 opposite happened as number of extremely severe dry conditions districts
increased from 24 to 72 and severely dry condition from 84 to 102.
Performance of SPI forecast was best for the last week i.e. week ending 30th September as the correct
forecast was more than 86% and forecast within + 1 Category was more than 98%. Not only that even
out of 43 districts experienced extremely severe drought in that week, 40 districts were predicted one
week back when there was only 11 districts under extremely severe drought condition.
Over all performance district SPI forecast for the complete SW monsoon season 2015

Forecast
Categories
1 2 3 4 5 6 7 8 Total
1 266 60 34 26 8 1 3 1 399
2 61 284 129 81 23 2 1 3 584
3 35 99 414 320 37 9 7 7 928
Actual

4 12 58 223 1770 476 46 22 18 2625

5 3 10 28 386 1183 187 68 24 1889
6 3 1 3 32 165 170 70 39 483
7 0 0 0 7 29 52 94 35 217
8 1 0 0 1 10 13 26 100 151
Total 381 512 831 2623 1931 480 291 227 7276

The following table obtained from the above contingency table gives the number and
percentage of correct forecast, within + 1 category, within + 2 category etc.

Correct In +1 cat In + 2 cat In + 3 cat In + 4 cat In + 5 cat In + 6 cat In + 7 cat

F/C
Freq 4281 2289 500 146 40 12 6 2
% 58.8 31.5 6.9 2.0 0.5 0.2 0.1 0.0

90.3 % of forecast was within + 1 Category

 Conclusions:
Southwest monsoon rainfall for two consecutive years was below 90 % of LPA ( 88% in
2014 and 86% in 2015) but their monthly distributions were completely opposite in
nature which may have significant effect on crop production activities.
Also there was spatial variation in the dry regions in this two years. Except UP, no other
states experienced consecutive two years drought situations but there are many districts
which experienced consecutive two years drought like situation. The SPI values computed
from all India rainfall for both of these years were in moderate drought category ( -1.2 in
2014 and -1.4 in 2015).

The performances GFS /MME based SPI forecast is fully depended on the skill of GFS
model. Instead of district rainfall forecast , SPI forecast has better advantage in
monitoring dry/wet condition over the region as SPI better represent dry conditions or
water stress over a region than than actual/percent departure of rainfall.
The performances of the SPI forecast model was quite good for the season (90.3 % of
forecast was within + 1 Category whereas 59% was correct forecast). Even in some week
it was more than 98% within + 1 Category while more than 86% was correct forecast.
The SPI forecast has rightly detected the sudden changes in the dry/wet conditions along
with identifying the regions/districts also.
Thanks