AgriMetSoft

Formulas of Drought Indices

1- Meteorological drought
A: Rain Based-drought indices (Salehnia et al., 2017):

• SPI (Standardized Precipitation Index, McKee et al. 1993, 1995

• DI (Deciles Index), Gibbs and Maher, 1967
• PN (Percent of Normal Index), Willeke et al. (1994)
• CZI (China-Z Index), Wu et al. (2001)
• MCZI (Modified CZI), Wu et al. (2001)
• EDI (Effective drought Index), Byun and Wilhite (1999)
• RAI (Rainfall Anomaly Index), van Rooy (1965)
• ZSI (Z-score Index), Palmer (1965)

B: Other meteorological drought indices:

• PDSI (Palmer Drought Severity Index), Palmer (1965) and Dehghan et al.,
2020
• PHDI (Palmer Hydrological Drought Index), Palmer (1965)
• SPEI (Standardized Precipitation Evapotranspiration Index), Vicente-Ser-
rano et al., 2010 and Salehnia et al., 2020
• RDI (Reconnaissance Drought Index), Tsakiris and Vangelis, 2005.

2- Agricultural drought indices

• ARI (Agricultural Rainfall Index), Nieuwolt, 1981

• SMDI (Soil Moisture Deficit Index), Narasimhan and Srinivasan, 2005
• ETDI (Evapotranspiration Deficit Index), Narasimhan and Srinivasan,
2005
 AgriMetSoft

3- Hydrological drought indices

• SWSI (Surface Water Supply Index), Garen, 1993

• SDI (Streamflow Drought Index), Nalbantis and Tsakiris, 2009

1.1 SPI (Standardized Precipitation Index)

The SPI stands as the predominant drought index on a global scale,

utilized for monitoring and comprehensive analysis, as highlighted

by Karabulut (2015). Widely acknowledged, the SPI serves as a cru-

cial instrument for delineating meteorological drought patterns, as

evidenced by the works of Hayes et al. (1999) and Deo (2011). The

pioneering work of McKee et al. (1993, 1995) encompassed the es-

tablishment of SPI across multiple temporal scales (1, 3, 6, 12, 24,

and 48 months). The resultant SPI values span from +2.0 to −2.0.

Given the possibility of precipitation data conforming to a gamma

distribution, the calculation of SPI entails the utilization of the prob-

ability density function inherent to the gamma distribution:

http://www.youtube.com/AgriMetSoft
 AgriMetSoft

1 −𝑥
𝑔(𝑥) = 𝛼 𝑥 𝛼−1 𝑒𝛽 𝑓𝑜𝑟 𝑥 >0
𝛽 𝛤(𝛼)

In this context, the gamma function is denoted by Γ, where x repre-

sents the quantity of precipitation (x > 0), α (> 0) stands as the shape

parameter, and β (> 0) represents the scale parameter. For further

insights, refer to the comprehensive studies conducted by Edwards

and McKee (1997) as well as Dogan et al. (2012).

1.2 PN (Percent of Normal Precipitation Index)

The PN index, delineated by Willeke et al. (1994), represents the

percentage of normal precipitation attributed to a particular loca-

tion. This metric, PN, holds the versatility to be computed over var-

ious temporal scales such as monthly, seasonal, and annual periods.

Notably, when applied to a singular geographic region or specific

season, PN has demonstrated its efficacy in elucidating drought

conditions, as demonstrated by the research of Hayes (2003). The

calculation of PN is as follows:

http://www.youtube.com/AgriMetSoft
 AgriMetSoft

𝑃
𝑃𝑁 = 𝑖⁄𝑃 × 100

Here, Pi denotes the precipitation during the time increment la-

beled as "i," while P represents the mean normal precipitation for

the duration of the study period.

1.3 DI (Deciles Index)

The deciles index (DI) was defined by Gibbs and Maher (1967).

Monthly historical precipitation data are sorted from lowest to

highest and divided into ten equal categories or deciles, so a given

month’s precipitation is placed into historical context by decile.

To compute the deciles index in DMAP, we first calculate the deciles

for each month and subsequently identify any specific month-year

in which the deciles meet certain criteria.

𝐷𝐸𝐶𝑛 = 𝑃𝑒𝑟(𝑃𝑚 , 𝑛)

𝐷𝐼𝑖,𝑗 = 𝑛 𝑖𝑓 𝐷𝐸𝐶𝑛−1 < 𝑃𝑖,𝑗 ≤ 𝐷𝐸𝐶𝑛

The term DECn represents the deciles associated with the value of n,

http://www.youtube.com/AgriMetSoft
 AgriMetSoft

where n takes on values such as 10, 20, ..., 90. Here, Per signifies the

percentile function. The Pm corresponds to the rainfall observed in

month m.

1.4 EDI (Effective Drought Index)

The Effective Drought Index (EDI) is computed at a daily time step

and follows a standardization process akin to SPI. Byun and Wilhite

(1999) pioneered the development of EDI, aiming to address certain

shortcomings observed in other indices like SPI. The EDI's numerical

range spans from -2.5 to 2.5. When the EDI falls within the range of

-1 to 1, it signifies near-normal conditions, whereas values below -

2 indicate severe drought. The calculation of EDI is accomplished

through the application of the subsequent equation:

http://www.youtube.com/AgriMetSoft
 AgriMetSoft

𝑛=1
(∑𝑚=1
𝑛 𝑃(𝑖−𝑚) )⁄
𝐸𝑃𝑖 = ∑ [ 𝑛]
𝐷𝑆

𝐷𝐸𝑃𝑖 = 𝐸𝑃𝑖 − 𝑀𝐸𝑃

𝐷𝐸𝑃𝑖
𝐸𝐷𝐼𝑖 = ⁄𝑆𝐷(𝐷𝐸𝑃)

Where DS represents the count of days used for summing precipi-

tation in the calculation of drought severity (usually 365 or 15), SD

stands for standard deviation, and MEP stands for the mean of EP.

The DMAP calculates EDI on daily scale then you can use Monthly,

Yearly, or seasonally to convert daily EDI to this time scales.

1.5 CZI (China-Z Index)

In 1995, the National Climate Center of China introduced the CZI as

a substitute for the SPI (Ju et al., 1997) specifically designed for sit-

uations where mean precipitation adheres to the Pearson Type III

distribution. The CZI is computed as follows:

1⁄
6 𝐶𝑠𝑖 3 6 𝐶𝑠𝑖
𝐶𝑍𝐼𝑖𝑗 = × ( × 𝜙𝜄𝑗 + 1) − +
𝐶𝑠𝑖 2 𝐶𝑠𝑖 6

http://www.youtube.com/AgriMetSoft
 AgriMetSoft

∑𝑛𝑗=1(𝑋𝑖𝑗 − 𝑋̅𝑖 )3
𝐶𝑠𝑖 =
𝑛 × 𝜎𝑖3

𝑋𝑖𝑗 − 𝑋̅𝑖
𝜑𝑖𝑗 =
𝜎𝑖

𝑛
1
𝜎𝑖 = √ ∑(𝑋𝑖𝑗 − 𝑋̅𝑖 )2
𝑛
𝑗=1

where i is the time scale of interest, j is the current month, Csi is the

coefficient of skewness and 𝜑𝑖𝑗 is the standardized variation. Fur-

ther details can be found in Wu et al., 2001.

1.6 MCZI (Modified CZI)

The adjusted MCZI employs the aforementioned formula but re-

places the mean precipitation with the median precipitation.

1.7 ZSI (Z-Score Index)

The Z-Score is sometimes mistaken for SPI. However, it bears a

closer resemblance to CZI, although it doesn't necessitate fitting

http://www.youtube.com/AgriMetSoft
 AgriMetSoft

precipitation data into either Gamma or Pearson Type III distribu-

tions:

𝑃𝑖 − 𝑃̄
𝑍𝑆𝐼 =
𝑆𝐷
Here, P represents the average monthly precipitation (mm), Pi

stands for the precipitation in a particular month, and SD denotes

the standard deviation across any time scale.

1.8 RAI (Rainfall Anomaly Index)

The RAI involves two phases: positive anomalies and negative

anomalies. Initially, precipitation data are sorted in descending or-

der. The top ten and bottom ten data points are then averaged

separately to create these phases. The positive and negative

anomalies are computed using equations 6 and 7, respectively:

(𝑝 − 𝑝̄ )
𝑅𝐴𝐼 = 3 × [ ]
(𝑚̄ − 𝑝̄ )

(𝑝 − 𝑝̄ )
𝑅𝐴𝐼 = −3 × [ ]
(𝑚̄ − 𝑝̄ )

In this context, RAI stands for Rainfall Anomaly Index. Here, p

http://www.youtube.com/AgriMetSoft
 AgriMetSoft

represents the actual precipitation (mm) for each year, p̅ is the

long-term average precipitation, and m̅ is the mean of the top ten

values of p for the positive anomaly, as well as the mean of the

bottom ten values of p for the negative anomaly.

Reference for 1.1 up to 1.8: Estimation of meteorological drought indices

based on AgMERRA precipitation data and station-observed precipitation

data

1.9 KBDI (Keetch-Byram Drought Index)

Calculating the index requires data for daily maximum tempera-

ture and precipitation. Following that, the KBDI was computed us-

ing the subsequent methodology:

[800 − 𝐾𝐵𝐷𝐼𝑡−1 ] × [0.968 × 𝐸𝑥𝑝(0.875𝑇𝑚𝑎𝑥 + 1.5552) − 8.30]

𝐷𝐹 = × 10−3
1 + 10.88 × 𝐸𝑥𝑝(−0.001736𝑅)

𝐾𝐵𝐷𝐼𝑡−1 𝑖𝑓 𝑃 = 0 𝑎𝑛𝑑 𝑇𝑚𝑎𝑥 ≤ 6.78 𝐶

𝐾𝐵𝐷𝐼𝑡−1 + 𝐷𝐹 𝑖𝑓 𝑃 = 0 𝑎𝑛𝑑 𝑇𝑚𝑎𝑥 > 6.78 𝐶
𝐾𝐵𝐷𝐼𝑡 = 𝐾𝐵𝐷𝐼𝑡−1 + 𝐷𝐹 𝑖𝑓 𝑃 > 0 𝑎𝑛𝑑 ∑ 𝑃 ≤ 5.1 𝑚𝑚

{ 𝐾𝐵𝐷𝐼𝑗 + 𝐷𝐹 𝑖𝑓 𝑃 > 0 𝑎𝑛𝑑 ∑ 𝑃 > 5.1 𝑚𝑚

http://www.youtube.com/AgriMetSoft
 AgriMetSoft

𝐾𝐵𝐷𝐼𝑗 = 𝐾𝐵𝐷𝐼𝑡−1 − 39.37 × ∑ 𝑝

The drought factor (DF) on a given day in the metric system, as pre-

sented by Janis et al. in 2002, is calculated based on several factors.

Specifically, it depends on the daily maximum temperature (Tmax t)

in degrees Celsius, the average annual rainfall (R) in centimeters for

the region, the Keetch-Byram drought index for the previous day

(KBDIt-1), and the daily precipitation (P t) in millimeters (Janis et al.

2002). The calculation involves intricate interplays between these

variables. Notably, the initialization of the Keetch-Byram drought

index (KBDI) is of paramount importance. Traditionally, this initiali-

zation occurs after a series of consecutive rainy days when soil sat-

uration is attained. However, it's important to acknowledge that the

field capacity of arable land, which signifies its ability to absorb wa-

ter, can differ significantly due to variations in soil composition and

changes in vegetation.

http://www.youtube.com/AgriMetSoft
 AgriMetSoft

Referenc: Predictive value of Keetch-Byram Drought Index for cereal yields

in a semi-arid environment

1.10 PDSI

1.11 PHDI
Calculating these two indices involves a substantial and intricate

process. It's advised to consult the original paper (Meteorological

drought Palmer 1965 Washington DC ) for detailed information.

Additionally, our calculation of the indices is directly based on this

paper: A tool for calculating the Palmer drought indices

1.12 SPEI
In the computation of SPEI, we employed both the classical approx-

imation by Abramowitz and Stegun (1965) and the method pro-

posed by Vicente-Serrano et al. (2009). The SPEI, accumulated over

various time scales, is founded upon a monthly climatic water bal-

ance derived from the difference between precipitation and

http://www.youtube.com/AgriMetSoft
 AgriMetSoft

Evapotranspiration (ETO). This balance is then refined through the

application of a three-parameter log-logistic distribution. The accu-

mulation of deficit or surplus in the climatic water balance (D) at

different time scales is ascertained by evaluating the dissimilarity

between daily precipitation (P) and ETO for a given day "i":

Di = Pi - EToi

The probability density function of a three-parameter log-logistic

distributed variable is expressed as below:

−2
𝛽 𝑥 − 𝛾 𝛽−1 𝑥 −𝛾 𝛽
𝑓(𝑥) = ( ) × [1 + ( ) ]
𝛼 𝛼 𝛼

Here, α, β, and γ represent the scale, shape, and origin parameters,

respectively, for D values within the range (γ > D < ∞). The parame-

ters for the Pearson III distribution can be acquired using the fol-

lowing method:

2𝑤1 − 𝑤0
𝛽=
6𝑤1 − 𝑤0 − 6𝑤2

(𝑤0 − 2𝑤1 )𝛽
𝛼=
Γ(1 + 1⁄𝛽 )Γ(1 − 1⁄𝛽 )

http://www.youtube.com/AgriMetSoft
 AgriMetSoft

1+1 1−1
𝛾 = 𝑤0 − 𝛼Γ ( )Γ( )
𝛽 𝛽

Here, Γ(β) denotes the gamma function of β. The probability distri-

bution function of the series "D" is expressed as follows:

−1
𝛼 𝛽
𝐹(𝑥) = [1 + ( ) ]
𝑥−𝛾

In the final stage, by utilizing the value of F(x), the SPEI can be de-

termined as the standardized value of F(x). The calculation of the

SPEI equation is as follows:

𝐶0 + 𝐶1 𝑊 + 𝐶2 𝑊 2
SPEI = W −
1 + 𝑑1 𝑊 + 𝑑2 𝑊 2 + 𝑑3 𝑊 3

Here, for cases where P ≤ 0.5, W is defined as the square root of -2

times the natural logarithm of P, where P represents the probabil-

ity of surpassing a specific value of "D." The constants involved are

as follows: C0 = 2.515517, C1 = 0.802853, C2 = 0.010328, d1 =

1.432788, d2 = 0.189269, and d3 = 0.001308.

Reference: Rainfed wheat (Triticum aestivum L.) yield prediction

http://www.youtube.com/AgriMetSoft
 AgriMetSoft

using economical, meteorological, and drought indicators through

pooled panel data and statistical downscaling

1.13 RDI

RDI (Reconnaissance Drought Index), developed by Tsakiris and

Vangelis in 2005, compares two cumulative measures: precipita-

tion and potential evapotranspiration. For a given month "m", this

calculation for a yearly index is represented by the following equa-

tion:
∑𝐷𝑒𝑐
𝑖=𝐽𝑎𝑛 𝑃𝑖
𝐷𝑚 = 𝐷𝑒𝑐
∑𝑖=𝐽𝑎𝑛 𝑃𝐸𝑇𝑖

The Reconnaissance Drought Index (RDI) is characterized by two

variations: the normalized RDI and the standardized RDI. The nor-

malized RDI serves as the default option within DMAP.

The Normalized Reconnaissance Drought Index (RDI) is computed

using the following equation:

𝐷𝑚
𝑅𝐷𝐼𝑚 = −1
𝐷̅

http://www.youtube.com/AgriMetSoft
 AgriMetSoft

̅ is average of D for all years. The Standardized Reconnais-

The 𝐷

sance Drought Index (RDI) is computed using the following equa-

tion:

̅
𝐷𝑚 − 𝐷
𝑅𝐷𝐼𝑚 =
𝜎
̅ is average of D for all years and 𝜎 is standard deviation of
The 𝐷

D for all years.

Reference: Regional Drought Assessment Based on the Reconnais-

sance Drought Index (RDI)

1.14 SMDI

For the computation of SMDI, it is necessary to employ the long-

term median, maximum, and minimum monthly (or seasonal or

yearly) soil water values. The following formulas are utilized for

this purpose:
𝑆𝑊𝑖,𝑗 − 𝑀𝑆𝑊𝑗
× 100 𝑖𝑓 𝑆𝑊𝑖,𝑗 ≤ 𝑀𝑆𝑊𝑗
𝑀𝑆𝑊𝑗 − 𝑚𝑖𝑛𝑆𝑊𝑗
𝑆𝐷𝑖,𝑗 =
𝑆𝑊𝑖,𝑗 − 𝑀𝑆𝑊𝑗
× 100 𝑖𝑓 𝑆𝑊𝑖,𝑗 > 𝑀𝑆𝑊𝑗
{𝑚𝑎𝑥𝑆𝑊𝑗 − 𝑀𝑆𝑊𝑗
𝑆𝐷𝑗
𝑆𝑀𝐷𝐼𝑗 = 0.5 × 𝑆𝑀𝐷𝐼𝑗−1 +
50

http://www.youtube.com/AgriMetSoft
 AgriMetSoft

The SMDI values will range from -4 to 4 for comparison with PDSI

values.

Here, MWSj represents the long-term median of water stress for

month j, maxMWSj stands for the long-term maximum water

stress of month j, minWSj signifies the long-term minimum water

stress of month j, and WS corresponds to the monthly water

stress. The subscripts i and j are employed for years and months

respectively. The water stress anomaly varies from -100 to +100

for each month, denoting exceedingly dry to exceptionally wet

conditions.

Reference: Development and evaluation of Soil Moisture Deficit

Index (SMDI) and Evapotranspiration Deficit Index (ETDI) for agri-

cultural drought monitoring

1.15 ARI

ARI serves as a dependable indicator of monthly water equilib-

rium, specifically designed for assessing agricultural droughts. The

calculation of ARI follows this methodology:

http://www.youtube.com/AgriMetSoft
 AgriMetSoft

𝑃
𝐴𝑅𝐼 =
𝑃𝐸𝑇
Here, P represents the monthly precipitation (mm/month), and

ET0 signifies the reference evapotranspiration (mm/month). The

presence of drought within a specific month is denoted by ARI val-

ues of 40. A range of ARI values between 40 and 200 indicates fa-

vorable conditions for vegetation growth and agricultural produc-

tivity. An ARI exceeding 200 for a month signifies a period of ele-

vated moisture, indicating wet conditions.

Reference: Understanding Dry and Wet Conditions in the Vietnam-

ese Mekong Delta Using Multiple Drought Indices: A Case Study in

Ca Mau Province

1.16 ETDI

The ETDI is computed similarly to the SMDI, yet it relies on the de-

viation of water stress from its long-term mean. Here, the monthly

water stress is established through the interplay of potential and

actual evapotranspiration.

𝑃𝐸𝑇 − 𝐴𝐸𝑇
𝑊𝑆 =
𝑃𝐸𝑇

http://www.youtube.com/AgriMetSoft
 AgriMetSoft

𝑊𝑆𝑖,𝑗 − 𝑀𝑊𝑆𝑗
× 100 𝑖𝑓 𝑊𝑆𝑖,𝑗 ≤ 𝑀𝑊𝑆𝑗
𝑀𝑊𝑆𝑗 − 𝑚𝑖𝑛𝑊𝑆𝑗
𝑊𝑆𝐴𝑖,𝑗 =
𝑊𝑆𝑖,𝑗 − 𝑀𝑊𝑆𝑗
× 100 𝑖𝑓 𝑊𝑆𝑖,𝑗 > 𝑀𝑊𝑆𝑗
{𝑚𝑎𝑥𝑊𝑆𝑗 − 𝑀𝑊𝑆𝑗
𝑊𝑆𝐴𝑗
𝐸𝑇𝐷𝐼𝑗 = 0.5 × 𝐸𝑇𝐷𝐼𝑗−1 +
50
The ETDI values will range from -4 to 4 for comparison with PDSI

values.

Reference: Development and evaluation of Soil Moisture Deficit

Index (SMDI) and Evapotranspiration Deficit Index (ETDI) for agri-

cultural drought monitoring

1.17 SWSI

A prevalent hydrological drought index consists of a series of prac-

tical individual measures, known as the Surface Water Drought Se-

verity Index (SWSI). The formula used to compute this index, as

per Garen's investigations, can be described as follows:

100 × 𝑃𝑖,𝑗 − 50
𝑆𝑊𝑆𝐼𝑖,𝑗 =
12

http://www.youtube.com/AgriMetSoft
 AgriMetSoft

𝑃𝒊,𝒋 = 𝐺𝑎𝑚𝑚𝑎𝐶𝐷𝐹(𝑆𝑡𝑟𝑒𝑎𝑚𝐹𝑙𝑜𝑤𝒊,𝒋 )

The GammaCDF represents the cumulative distribution function of

the gamma distribution fitted onto month j across all years. In this

context, StreamFlowi,j denotes the streamflow during month j of

the year i.

Reference: Comparing surface water supply index and streamflow

drought index for hydrological drought analysis in Ethiopia

1.18 SDI

The Streamflow Drought Index can be computed using the follow-

ing formula when applied to monthly data. However, if you pro-

vide daily data, the tool will automatically aggregate it into

monthly values before performing the calculation.

3𝐾

𝑉𝑖,𝑘 = ∑ 𝑆𝐹𝑖,𝑗 𝑖 = 1,2, . . , 𝑒𝑛𝑑𝑌𝑒𝑎𝑟 𝑗 = 1,2, . . ,12 𝑘

𝑗=1

= 1,2,3,4

𝑉𝑖,𝑘 − 𝑉̅𝑘
𝑆𝐷𝐼𝑖,𝑘 =
𝑆𝐷𝑘

The premise is that a time series of monthly streamflow volumes

is available, denoted as SFi,j. Here, i indicates the hydrological year,

http://www.youtube.com/AgriMetSoft
 AgriMetSoft

and j represents the month within that hydrological year (where j

= 1 corresponds to October and j = 12 corresponds to September).

In this context, 𝑉̅𝑘 and SDk stand for the mean and standard devi-

ation of cumulative streamflow volumes observed during the ref-

erence period k, established over a substantial time span. Within

the framework of DMAP, the resulting output will encompass

these four categories: Oct-Dec, Oct-Mar, Oct-Jun, and Oct-Sep

(Yearly).

Reference: Assessment of Hydrological Drought Revisited

http://www.youtube.com/AgriMetSoft