Chapter-5 Development of Modified Drought...

Modeling of rainfall variability and drought assessment in Sabarmati basin, India _____________________________________________________________________________________

51

Chapter-5 Development of Modified Drought Index

5.1 Introduction

Drought is generally considered to be occurring when the monsoon fail or are

deficient or scanty. Failure in the monsoon may create crop failure, shortage of

drinking water and affecting the socio-economic life of the rural and urban

community. Generally, the amount of rainfall received by any area/region gives idea

about occurrence of drought in any area/region. The occurrence of drought can be

identified using various drought indices. Drought indices give quantitative

assessment of climatic conditions and can be used as a tool for early warning system,

drought monitoring.

5.2 Relationship between drought indices and micro-climate

Drought is a normal, recurrent feature of climate and it is observed in

all the climate zones, with different characteristics in different regions. Drought is a

climatic anomaly, characterized by deficient supply of moisture resulting either from

sub-normal rainfall, erratic rainfall distribution, higher water need or a combination

of all the three factors. About two thirds of the geographic area of India receives low

rainfall (less than 1000 mm), which is also characterized by uneven and erratic

distributions. While considering drought, it is important to take into account the

onset of rainy season, delay in start of monsoon, breaks in the monsoon, rainfall

intensity, severity, etc. The severity of drought can also be affected by other factors

like temperature, wind velocity, humidity, etc. Assessment of Regional Drought is


52

based on drought indices for identifying drought characteristics. Drought Indices

simplify the complex relationships and provide good communication tool. It gives

quantitative assessment of anomalous climatic conditions (intensity, duration, and

spatial extent).Drought Indices provides a basis for drought management.

5.3 Description of Various Drought Indices:

A drought index value is typically a single number, far more useful than raw

data for decision making. There are several indices that measure how much rainfall

for a given period of time has deviated from historically established norms. Although

none of the major indices is inherently superior to the rest in all circumstances, some

indices are better suited than others for certain uses. The drought indices which use

meteorological observations recorded at meteorological stations are called

meteorological drought indices. The drought index which uses rainfall data are

percent of normal (PN), deciles, standardized precipitation index (SPI), effective

drought index (EDI), etc. The indices based on only rainfall data are not only simple

to compute, it has also been shown that these indices perform better compared to

more complex hydrological indices (Oladipio, 1985). In India also there have been

studies related to drought using drought indices based on the rainfall data only.

When the seasonal rainfall received over an area is less than 75% of its long-term

average, it is called meteorological drought (Ray-2000). These studies were mainly

for the monsoon season (June to September) which contributes about 75-90% of the

total annual rainfall over most parts of the country.

5.3.1 Standardized Precipitation Index (SPI):

It is an index based on the probability of rainfall for any time scale. Many

drought planners appreciate the SPI’s versatility. The SPI can be computed for

different time scales, can provide early warning of drought and help to assess

drought severity, and is less complex than the Palmer. It is developed by T.B. McKee,


53

N.J. Doesken, and J. Kleist, Colorado State University, 1993. The understanding that a

deficit of rainfall has different impacts on groundwater, reservoir storage, soil

moisture, snow pack, and stream flow led McKee, Doesken, and Kleist to develop the

Standardized Precipitation Index (SPI) in 1993. The SPI was designed to quantify the

rainfall deficit for multiple time scales. These time scales reflect the impact of

drought on the availability of the different water resources. Soil moisture conditions

respond to rainfall anomalies on a relatively short scale. Groundwater, stream flow,

and reservoir storage reflect the longer-term rainfall anomalies. For these reasons,

McKee et al. (1993) originally calculated the SPI for 3,6,12, 24 and 48 month time

scales. The SPI calculation for any location is based on the long-term rainfall record

for a desired period. Standardized precipitation is the difference of precipitation

from the mean for a specified time divided by the standard deviation, where the

mean and standard deviation are determined from the climatological record. The fact

that precipitation is not normally distributed is overcome by applying a

transformation (i.e., gamma function) to the distribution.

The computation of SPI requires long term rainfall data. It is found by Thom

(1966) that the gamma distribution function fit to the rainfall time series. The long-

term record is fitted to a probability distribution, which is then transformed into a

normal distribution so that the mean SPI for the location and desired period is zero

(Edwards and McKee, 1997). The rainfall series was fitted to the gamma distribution.

It is defined by its frequency or probability density function. The gamma probability

distribution function (pdf) is given as follows. The alpha and beta parameters of the

gamma probability density function are estimated for each station.

xexxg

11

for x >0

where, x = rainfall amount

= shape parameter =

3411

41 AA

A = n

xx lnln

n = number of rainfall observations

β = scale parameter = ^

x


54

Г( ) = gamma function

The gamma distribution function is fitted to the data for estimating the

parameters and β. The gamma cumulative distribution function (cdf) is computed

at each value of x by integrating pdf with respect to x and inserting the estimated

values of and β. The cdf is then transformed into the standard normal distribution

to find SPI. To compute SPI corresponding to a rainfall amount, mark the rainfall

amount on x-axis (Fig.5.1). From this point draw a line parallel to y-axis till it

intersects with the theoretical gamma cdf line. From this point of intersection, extend

a line perpendicular to y-axis till it intersects the standard normal cdf graph. Draw a

line parallel to y-axis from this point upwards to the secondary axis to have the SPI

value.

Figure 5.1 Illustration of computation of SPI for the seasonal rainfall over the country

as a whole obtained through equiprobability transformation from fitted gamma cumulative probability

distribution to standard normal cumulative probability distribution. (Source:NCC Research report

2/2010)

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, and wet periods can also be


55

monitored using the SPI. McKee et al. (1993) used the classification system shown in

the SPI values table to define drought intensities resulting from the SPI. McKee et al.

(1993) also defined the criteria for a drought event for any of the time scales. A

drought event occurs any time 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 intensity for

each month that the event continues. The positive sum of the SPI for all the months

within a drought event can be termed the drought’s “magnitude”. Table 5.1 Drought Classification as per SPI (Source: http//drought.uni.edu)

SPI Values Drought Criteria As per SPI Drought Class As per SPI

2.0+ Extremely wet G

1.5 to 1.99 very wet F

1.0 to 1.49 moderately wet E

-.99 to .99 near normal D

-1.0 to -1.49 moderately dry C

-1.5 to -1.99 severely dry B

-2 and less Extremely dry A

5.3.2 Percent of Normal: The percent of normal is one of the simplest measurements of rainfall for a

location. It is useful for analyzing a single region or a single season. Percent of

normal is also easily misunderstood and gives different indications of conditions,

depending on the location and season. It is calculated by dividing actual rainfall by

normal rainfall—typically considered to be a 30-year mean—and multiplying by

100%. This can be calculated for a variety of time scales. Usually these time scales

range from a single month to a group of months representing a particular season, to

an annual or water year. Normal rainfall for a specific location is considered to be

100%.

One of the disadvantages of using the percent of normal is that the mean, or

average, rainfall is often not the same as the median rainfall, which is the value

exceeded by 50% of the rainfall occurrences in a long-term climate record. The reason

for this is that rainfall on monthly or seasonal scales does not have a normal


56

distribution. Use of the percent of normal implies a normal distribution where the

mean and median are considered being the same. Because of the variety in the

rainfall records over time and location, there is no way to determine the frequency of

the departures from normal or compare different locations. This makes it difficult to

link a value of a departure with a specific impact occurring as a result of the

departure, inhibiting attempts to mitigate the risks of drought based on the

departures from normal and form a plan of response (Willeke et al., 1994).

5.4 Need of modified drought index

Earlier drought indices developed and used by many researchers uses

weather parameters like rainfall, temperature, evapo-transpiration and remote sense

data to compute drought indices. Although none of drought indices is superior to

other in all situation. Suitability of a drought index depends on its use. Rainfall data

are used to calculate drought index as long term rainfall data are available. Rainfall

data alone may not reflect the range of drought related conditions, but they can work

as one of the practical solution in data-poor regions. Therefore, a new definition of

drought index has been proposed based on standardized precipitation index (SPI)

and number of rainy days to compute modified drought index. SPI represents the

probability of rainfall while the proposed modified drought index considers the

rainfall amount and the duration. This modified index will be superior as it considers

rainfall duration also on which crop growth depends. Therefore, for the purpose of

drought planning this modified drought index will offer better insight on drought

definition.


57

5.5 Reconstruction and classification of modified drought index

The SPI developed by McKee et al (1993) is based on the probability of rainfall

for a particular time scale. The basic data used in the study is rainfall data of 20

rainfall stations of Sabarmati basin of Gujarat for the period 1976 – 2007 for the

monsoon period May to October. Based on this data, Standardized Precipitation

Index (SPI) was computed for the districts of Sabarmati basin. After calculating the

SPI values for individual stations, a modified drought index is developed by

multiplying the SPI values with average number of rainy days for individual

stations.

Modified Drought Index= SPI * Number of Rainy Days ……….. (5.1)

The drought classes have been defined considering average SPI and average

number of rainy days for a given station over long time period. Although none of the

drought indices are superior, hence the need of new drought index was felt. Based on

the new drought index calculated, drought years are identified and compared with

the past records of available data.

Based on the modified drought index calculated, drought years are identified

and compared with the past records of drought years in Sabarmati basin. Seven MDI

classes has been defined as class A (< -72), B (-54 to -72), C (-36 to -54), D (-36 to 36), E

(36 to 54), F (54 to 72), and G (>72). The quantitative estimates on MDI for 20 stations

have been done for year 1976-2007.

5.6 Analysis and results

Using the daily rainfall data of 20 stations of Sabarmati basin for the period of

1976-2007, two drought indices; Percent of Normal (PN) and Standardized

Precipitation Index (SPI) were calculated. The analysis is carried out to compute SPI

and PN for the period of 1976 – 2007. Fig: 5.2(a) & (b) shows the time series rainfall

expressed in terms of SPI & PN for the period 1976 to 2007. It is clearly seen that the

year to year variation in both plots are nearly same. The percent of Normal (PN) was

calculated by dividing the actual rainfall by average rainfall and multiplying by


58

100%. Based on the available rainfall record, the average rainfall for 20 stations in

Sabarmati basin is 770mm. When the rainfall is between 90-110% of normal rainfall, it

is termed as normal. The seasonal rainfall over a district is classified as moderate

drought when the rainfall over an area is 50-74% of average rainfall and a severe

drought when the rainfall is less than 50% of average rainfall. Drought years for the

20 stations in Sabarmati basin were identified based on PN criteria,. It is observed

that during the period of 1976 to 2007, there were 12years with Normal condition (90-

110%of Normal Rainfall). In the year 1986, 1987, 1995, 1999, 2000 and 2002, all the

stations received less then normal rainfall. In the year 1987, all the stations faced

severe drought situations. The year 1976, 1977, 1994, 1997 and 2006 were received

more than average rainfall for almost all stations.

The SPI value has been calculated as per the methodology explained in section

5.3. The classification of the drought intensities based on the SPI value is given in

Table 5.1. Drought classification based on SPI criteria shows that all the stations

considered for analysis suffered moderate to severe drought conditions in the year

1986, 1987 1999 and 2002. The year 1995, 2000, and 2001 shows around 50% stations

were affected by moderate drought. In the year 1979, 1982 and 1985 around 25%

stations were affected by moderate drought. The year 1976, 1977, 1994, 1997 and 2006

were received more than average rainfall for almost all stations. The year 1978, 1980,

1981, 1983, 1984, 1988, 1989, 1991, 1992, 1993, 1996, 1998, 2003, 2004 and 2007 were

received normal rainfall. One peculiar observation was seen from this analysis that

the year which was suffered by severe drought conditions, the previous year and

next year shows extreme situations.

In the second step of analysis, an attempt has been made to develop a modified

drought index. For quantitative assessment of anomalous climatic conditions

(Intensity, Duration, and spatial extent), Standardized precipitation Index (SPI) is

calculated for 20 stations in Sabarmati basin. A modified Drought Index is developed

and calculated by multiplying the SPI values with average number of rainy days for

individual stations for identifying the drought years and shown in Figure 5.3. The


59

value of modified drought index for each year of all station has been plotted in Arc

GIS map (Figure 5.4). The analysis shows that during the time period of 1976 to 2007,

majority of rain gauge stations considered for analysis in Sabarmati basin faced

drought in the year of 1986, 1987 and 2002. In the year 1987, out of 20 stations, 5

stations faced moderate drought condition, 14 stations faced severe drought and one

station extreme drought condition. For year 1999, out of 20 stations, only seven

stations were under normal condition, and 13 faced moderate drought condition. In

the year 2002, 50 % stations were under moderate drought conditions. Out of 32

years, nine years (1978, 1980, 1984, 1988, 1989, 1991, 1992, 1993, and 1996) were

normal years for all stations and for years 1977, 1981, 1983, 1990, 1998, 2003, 2004 and

2007 the stations were received normal/above normal rainfall. The modified drought

index has been developed to identify and classify the drought at regional level. The

year 1976 has been considered as a year with more then average rainfall, in the S-W

region of basin, majority of stations received excess rainfall then average rainfall

while in N-W region; all the stations received normal rainfall. In the region N-E, two

stations received excess rainfall while other with normal rainfall. The modified

drought index derived has been found to have strong relationship with severe

drought and wet years.

5.7 Discussion

The variability of rainfall at spatial and temporal level is high in Sabarmati

basin as discussed in previous chapter. The number of rainy days has also been

found to be variable at spatial and temporal scale. The temporal variability found to

be more then spatial variability. So, in the first step of analysis, two drought indices

(SPI and PN) were computed for analyzing the drought condition during a period of

1976 to 2007 for 20 stations in Sabarmati basin. The analysis shows that SPI is better

as compared to PN as it shows dry as well as wet conditions of an area. PN is useful

for analyzing a single region or a single season as it considers only rainfall amount

and use of the PN implies a normal distribution, while the rainfall for a season does

not have a normal distribution. The SPI was designed to quantify the rainfall deficit


60

for multiple time scales, which reflect the impact of drought on the availability of the

different water resources. The analysis shows that SPI is better as compared to PN as

it shows dry as well as wet conditions of an area.

5.8 Conclusions

An attempt has been made to develop a drought index for identifying the

drought situations based on SPI and number of rainy days in Sabarmati basin. Based

on SPI, a modified drought index has been developed and results were verified with

past data record on drought and has been found good coherence. The analysis shows

that when there is a variability of rainfall and rainy days at spatial and temporal

scale, MDI may be used for identification of drought at station level. MDI

considers rainfall amount and distribution, which may be used at regional level

drought classification. The overall analysis leads to the following conclusions.

SPI is better as compared to PN as it shows dry as well as wet conditions of an

area.

A new drought index (MDI) based on SPI and number of rainy days for

identifying drought has been developed.

-2.50

-1.50

-0.50

0.50

1.50

2.50

3.50

1976

1978

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

AHD BDL BRJ BYD BHL SJT CDL CPW DHR HMT

IDR LML MHD MNS RSP RPW TTI VRP VJA VSI

Figure 5.2(a) Standardized precipitation Index (SPI) for 20 stations in Sabarmati basin


61

0

50

100

150

200

250

1976

1978

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

% o

f Nor

mal

AHD BDL BRJ BYD SJT CDL CPW DHR HMT LML

MHD RSP VRP RPW TTI VJA BHL MNS VSI IDR

Figure 5.2 (b) % of Normal (PN) for 20 stations in Sabarmati basin

-125

-75

-25

25

75

125

1976

1978

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

MDI

AHD BDL BYD BRJ SJT BHL CDL CPW DHR IDR

HMT LML MHD MNS RSP VRP RPW TTI VJA VSI

Figure 5.3 Modified Drought Index derived for Sabarmati basin


62

-3

-2

-1

0

1

2

3

1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006

YEAR

SPI

-3

-2

-1

0

1

2

3

MD

I

SPI_AHD SPI_BRJ SPI_CDL SPI_MHD SPI_SJT SPI_RPW SPI_VSI SPI_LMLSPI_MNS SPI_RSP SPI_VRP SPI_VJA SPI_VDL SPI_DHR SPI_BYD SPI_BHLSPI_CPW SPI_IDR SPI_HMT SPI_TTI MDI_CDL MDI_MHD MDI_SJT MDI_RPWMDI_VSI MDI_LML MDI_MNS MDI_RSP MDI_VRP MDI_BDL MDI_DHR MDI_BHLMDI_CPW MDI_IDR MDI_AHD MDI_BRJ MDI_VJA MDI_BYD MDI_HMT MDI_TTI

Figure 5.3-a Comparison of SPI and MDI for twenty stations of Sabarmati basin for a

period of 1976 to 2007


63

Table 5.2 Performance of various drought indices for twenty stations


64


65

Figure 5.4 Modified drought index for 20 stations during 1976-2007 plotted in Arc-

GIS


66

Chapter-6 Prediction of Monsoon onset

6.1 Introduction

In arid and semi-arid regions like Sabarmati basin, where rainfall is

limited to a few months per year only, is the most critical factors for the ecological

and environmental processes,. The amount of water available strongly depends on

the rainy season’s on-set, length and end in the areas where most of the agricultural

production depends on rainfall (Ati et al 2002). According to Steward (1991), the

onset is the most important variable to which all other seasonal variables are related.

Rainfall data is most important to hydrologists as it forms basis of many hydrological

studies. The critical problem is the uneven distribution of rainfall during rainy

season and the gap between the successive rainfall events. In order to get maximum

yield, it is necessary to supply optimum quantity of water at right time which may

not be possible every time. The total amount of rainfall in a particular area may be

not sufficient or is not in time. The rainfall may be non-uniform over the crop period.

There are sensitive periods of crop where proper amount of water is required. If

sufficient moisture is not available, yield may be reduced. Due to very high spatial

and temporal variability of rainfall and non uniform distribution of rain during rainy

season, farmers have problems to decide when to start with sowing of plants. Some

of the strategies adopted by farmers to cope with the problems are re-sowing, dry

seeding, crop rotation, exchanging information about rainfall with local workers and

steps for sustaining soil fertility.

It is necessary to predict the onset of rainy season which is the most important factor

and which coincide with the growing season of crops. Information of onset of rainy

season, length and end of season imparts significant information in timely


67

preparation of farmland, planting, preparation of equipments and manpower and

also help in contingency planning to the government in the situation of drought.

Due to random distribution of local convection events and potential shifts of

onset dates on site scales, this study will concentrate on determination and prediction

of arrival of monsoon on basin scale. It is focused on onset of monsoon for the

different regions of Sabarmati basin.

6.2 Methodology on prediction of onset

The onset of rainy season (monsoon) is defined in different ways at

present. The principal research areas are India, Australia and West Africa i.e. areas

having water scarcity and rainfall is limited to rainy season. In general two types of

definitions can be distinguished. The definition of monsoon onset is generally based

on the parameters measured on the surface which may be used for agro-

climatologically purposes on local scale or on the basis of atmospheric dynamics by

analyzing the appearance of large-scale circulation patterns in combination with start

of rains. Most researchers refer to rainfall itself in order to determine the onset

and/or end of rainy season. The benefit of this approach is that precipitation tools

are readily available and it exhibits the most direct relationship rather than some

other related factors. For rainfall-alone definition, two sub groups can be found in

literature, a definition based on certain threshold value (e.g. Stern et al.,1981) and a

relative definition using a proportion relative to the total amount (Ilesanmi,1972)

The overall objectives of this investigation are:

1) To develop a reasonable onset definition for Sabarmati basin

2) To predict arrival of monsoon in a region based on arrival of monsoon in

neighboring region.

An attempt has been also made to forecast the rainfall using Box-Jenkins approach of

time series using ARIMA model. The description of model, methodology, data and

performance model has been given in annexure-II.


68

6.3 Description of Fuzzy Logic model

For the statistical analysis performed, daily rainfall time series of 20

stations were applied. The meteorological data were obtained from State water data

centre. The data were checked for continuity by calculating monthly and annual

totals. Table 6.1 represents the details of rain gauge stations used for analysis.

A Fuzzy logic approach has been developed to facilitate modeling.

Fuzzy Logic (FL) is not a control methodology, but it is a way of processing data by

allowing partial set membership rather than crisp set membership or non-

membership. This approach to set theory was not applied to control systems until the

70's due to insufficient small-computer capability prior to that time. Fuzzy logic is a

problem-solving control system methodology that lends itself to implementation in

systems ranging from simple, small, embedded micro-controllers to large,

networked, multi-channel PC or workstation-based data acquisition and control

systems. It can be implemented in hardware, software, or a combination of both.

Fuzzy logic provides a simple way to arrive at a definite conclusion based upon

vague, ambiguous, imprecise, noisy, or missing input information. The approach of

fuzzy logic to control problems mimics how a person would make decisions, only

much faster. Fuzzy logic incorporates a simple, rule-based “if x and y then z”

approach to a solving control problem rather than attempting to model a system

mathematically. The fuzzy logic model is empirically-based, relying on an operator's

experience rather than their technical understanding of the system. Fuzzy logic

requires some numerical parameters in order to operate such as what is considered

significant error and significant rate-of-change-of-error, but exact values of these

numbers are usually not critical unless very responsive performance is required in

which case empirical tuning would determine them. Fuzzy logic was conceived as a

better method for sorting and handling data but has proven to be an excellent choice

for many control system applications since it mimics human control logic. It can be

built into anything from small, hand-held products to large computerized process

control systems. It uses an imprecise but very descriptive language to deal with input


69

data more like a human operator. It is very robust and forgiving of operator and data

input and often works when first implemented with little or no tuning.

Fuzzy logic is a form of multi-valued logic derived from fuzzy set theory to

deal with reasoning that is approximate rather than precise. In contrast with "crisp

logic", where binary sets have binary logic, the fuzzy logic variables may have a

membership value of not only 0 or 1 – that is, the degree of truth of a statement can

range between 0 and 1 and is not constrained to the two truth values of classic

propositional logic. Furthermore, when linguistic variables are used, these degrees

may be managed by specific functions.

The onset of monsoon has been defined considering three constraints

namely, total amount of rainfall, number of rainy days and percentage of stations

receiving rainfall. An artificial intelligence based fuzzy logic approach has been used

to model the onset of rainy season. A fuzzy logic approach is important as it can

incorporate the sternness of constraints which have to be fulfilled simultaneously. In

this approach, each constraint is attached to a fuzzy membership function using

triangular (subscript T) fuzzy numbers. The first two definition constraints are

attached to a fuzzy membership function using triangular fuzzy numbers while the

third constraint considers the threshold limit. For the first constraint dealing with

total amount of rainfall within a 10 days spell, the triangular fuzzy numbers are (18,

25, +∞) т. This means that membership grade of rainfall totals less than 18 mm is

attached to zero and total larger than 25 mm to one. Between 18 and 25 mm the

membership grade is linearly interpolated. For the second constraint dealing with

number of rainy day, the triangular fuzzy numbers are (1, 3, +∞) т. This means that

membership grade of rainy days less than 1 is attached to zero and more than 3 to

one. The membership grades between 1 and 3 are linearly interpolated and

appropriate values are assigned. If 1 , 2 , 3 are membership grades for amount

of rainfall, number of rainy days and percentage of stations receiving rainfall then,


70

the onset date is defined as the first day of year where the product

321 ** exceeds a defined threshold value.

onset date = 321 ** --------------------------(6.1)

Where = Onset date,

,01 if ∑ Rf =18 and 11 , if ∑ Rf =25

02 if ∑ Rd =1 and 12 if ∑ Rd =3

13 if 60% stations in a region met criteria 1 and 2

Based on the Fuzzy logic algorithm, software for calculating the values of

membership grades 1 and 2 has been developed in FOXPRO. A filter was applied

on onset date to make prediction relevance, in case the onset definition is beyond the

stipulated monsoon duration.


71

Figure 6.1 Fuzzy logic algorithm for membership grade 1

EOF: End of File


72

Figure 6.2 Fuzzy logic algorithm for membership grade 2

EOF: End of File


73

0

0.2

0.4

0.6

0.8

1

10 15 20 25 30

rainfall amount within 10-day spellM

embe

rshi

p gr

ade

0

0.2

0.4

0.6

0.8

1

0 1 2 3 4

number of wet days within 10-day spellM

embe

rshi

p gr

ade

Condition 1 Condition 2 Condition 3

Figure 6.3 Membership functions representating onset conditions 6.4 Analysis of Results

The daily rainfall values of all rainfall stations were used as input in

order to derive past onset dates. The monsoon onset is the time when first spell of

rains are being received. Therefore it is important to predict arrival of monsoon for a

basin. A Fuzzy logic based approach was used to develop a reasonable onset

definition. Three constraints namely, one- amount of rainfall in ten days, two-

successive number of rainy days and three- percentage of stations receiving rainfall

has been considered for defining onset. Fig 6.3shows the three conditions used for

identifying the onset of monsoon using fuzzy logic approach. The arrows represent

the direction for predicting the onset of rainy season of one region using the current

onset date of another region. The entire basin is divided into three regions based on

similar rainfall characteristics. The analysis of rainfall shows that the variability of

rainfall is very high. The average daily rainfall distribution for regions has been

shown in Figure 6.4 It shows that the all the regions received few amount of rainfall

in the beginning of month of May. The region1 received around 10mm rainfall daily

after mid of June which lasts up to first week of September, while region2 and region

3 receives average daily rainfall of around 10mm after third week of June which lasts

up to end of September. The basin has been classified into three regions based on

terrain, amount of rainfall received and


74

cluster of stations. The model results showed monsoon onset for

region-1 on 14th June, region-2 on 17th June, and region-3 on 21st June. The model

captures periodic variability in monsoon onset for various years. The model output

was verified with Indian Meteorological Department’s dates for monsoon arrival

which shows good coherence. The performance of model may be improved using

filter(s).

The Linear Regression models have been developed to forecast

monsoon arrival in region 2 based on onset in region 1 and for region 3 based on

onset in region 2 (Figure 6.6). The model parameters for targeted and independent

region have been found to be 0.40 and 0.35 respectively. The application of onset

model will help farmers to decide crop to be sown and cropping pattern. Such type

of simple tools may be used for advisory services.

Region1

0

6

12

18

0 50 100 150 200Days

Avg

.Rai

nfal

l

Region2

0

5

10

15

20

0 50 100 150 200DayAv

g.R

f

Region3

0

5

10

15

20

0 50 100 150 200Day

Avg.

Rf

Figure 6.4 Average daily rainfall plots for regions during monsoon season

Region1

0

20

40

60

80

1977 1982 1987 1992 1997 2002 2007Year

Region2

0

20

40

60

80

1977 1982 1987 1992 1997 2003Year

Ons

et d

ay

Region3

0

20

40

60

80

1977 1982 1987 1992 1997 2002 2007Year

Ons

et d

ay

Figure 6.5 Onset dates computed based on fuzzy logic approach for 1976 -2007


75

6.5 Discussion

As the onset of rainy season is of major interest for managing farming strategies, it is

required that before sowing of plants, farmers should be aware of the onset of rainy

season .In this paper, for Sabarmati basin, based on the past data of daily rainfall

(time period: year 1976 to 2007), analysis of past onset dates and mean rainfall

amount is carried out. For this, Fuzzy logic approach is considered. The entire basin

is divided into three regions based on similar rainfall characteristics. Region1

receives the rainfall in the second week of June and region 2 and Region 3 in the third

week of June. This information is useful to the farmers for growing the seeds. The

survival of seeding is key points for agriculturists (Sultan and Janicot, 2003). For

sowing, it is important to know, whether, (1) the rain is continuous and sufficient to

provide soil moisture during planting time, (2) level of soil moisture will be

maintained or is there any change during growing period to avoid crop failure

(Walter, 1967). The data dependent model for defining monsoon onset using FL for

Sabarmati basin has been developed.

Table 6.1 Details of rain gauge stations Serial Number

NAME OF STATION

RAINFALL REGION

1 AHD 780 2 RPW 709 3 CDL 737

S-W

ModelR1

10

20

30

40

50

60

70

80

0 5 10 15 20 25 30 35Year

Ons

et d

ay

Predicted Y Observed y

ModelR1

10

20

30

40

50

60

70

80

0 5 10 15 20 25 30 35Year

Ons

et d

ay

Predicted Y Observed y

Figure 6.6 observed and predicted onset days using linear Regression Analysis


76

4 VSI 749 5 BRJ 811 6 MHD 761 7 SJT 738

8 LML 726 9 MNS 718 10 RSP 671 11 VJA 800 12 BDL 712 13 VRP 675 14 DHR 758

N-W

15 BHL 876 16 CPW 906 17 HMT 779 18 IDR 875 19 BYD 836 20 TTI 784

N-E

Table 6.2 Mean onset days and standard deviation

Region Mean onset date

Mean Onset day Std. Dev.

Region 1(S-W) 14-Jun 45 17 Region 2(N-W) 17-Jun 48 15 Region 3(N-E) 21-Jun 52 14

Table 6.3 Linear regression models

Model Target Region

Independent Region

γ threshold

target region

Γ threshold

independent region

Regression Equation

M1 Region 2 (N-W) Region1 (S-W) 0.35 0.40 Y = -0.3683x + 53.795

M2 Region3 (N-E) Region 2 (N-W) 0.35 0.40 Y = -0.3438x + 57.597


77

6.6 Conclusion

There are methods to predict the rainfall using meteorological and /or

atmospheric data. In this study, Linear Regression Analysis is used to find out onset

dates of regions using the onset dates of one region. Here, region 1 is considered as

the first region to receive onset and based on that linear regression models has been

developed between regions, which can be useful for predicting inter-region monsoon

onset. (Table 6.3) The analysis leads to following points.

A data dependent model for defining monsoon onset using FL for Sabarmati

basin has been developed.

A linear regression models between regions which can be useful for predicting

inter-region monsoon onset has been developed.

ARIMA methodology of time-series has been attempted to predict rainfall

where the select models fails to catch the trend of series.

As the Sabarmati basin is a semi arid area and most of the agriculture depends on

rainfall in the basin, the analysis of past onset dates helps the farmers in planning

their crop season. Knowing the start of rainy season, they can plan for type of the

crop to be sown and date of sowing. The analysis also helps the farmer for drought

contingency planning if the quantity of rain is less for a particular crop. An attempt

has been made to predict rainfall using ARIMA methodology of time-series.

Date post:	20-Mar-2018
Category:	Documents
Upload:	lamnga
View:	214 times
Download:	1 times

Chapter-5 Development of Modified Drought...

Documents