+ All Categories
Home > Documents > Markov Chain Model for probability of weekly rainfall in ... · observed in Bangladesh by Sayedur...

Markov Chain Model for probability of weekly rainfall in ... · observed in Bangladesh by Sayedur...

Date post: 19-Jul-2018
Category:
Upload: dangdan
View: 214 times
Download: 1 times
Share this document with a friend
13
INTERNATIONAL JOURNAL OF GEOMATICS AND GEOSCIENCES Volume 3, No 1, 2012 © Copyright by the authors - Licensee IPA- Under Creative Commons license 3.0 Research article ISSN 0976 – 4380 Submitted on May 2012 published on July 2012 191 Markov Chain Model for probability of weekly rainfall in Orathanadu Taluk, Thanjavur District, Tamil Nadu Senthilvelan.A 1 , Ganesh.A 2 , Banukumar.K 3 1-Assistant Professor, Department of Geography, Government Arts College (Autonomous), Kumbakonam 2- Professor, Department of Geography, Bharathidasan University, Tiruchirappalli. 3- Assistant Professor, Department of Geography, Periyar EVR College (Autonomous), Tiruchirappalli [email protected] ABSTRACT It is necessary to know the sequence of dry and wet periods for successful agricultural management and planning of soil and water conservation measures. Probability analysis is a very useful tool for making important decisions in agricultural operation. In this study, Markov Chain Model has been extensively used to study spell distribution. For this purpose a week period was considered as the optimum length of time. The present study has been carried out to find the probabilities of occurrence of wet week (W), wet week preceded by wet week (W/W) at different threshold limits of 10 and 20 mm. On the basis of the analysis the following conclusions are made: (a) the 3 and 3 1/2 month varieties are best suited for Vettikkadu region; (b) 3 1/2 and 4 months paddy varieties can be successfully grown in Neivasal Thenpathi area and (c) 4 and 4 1/2 months paddy varieties are favourably grown in Orathanadu region. 1. Introduction The yield of crops particularly in rainfed condition depends on the rainfall pattern. Simple criteria related to sequential phenomena like dry and wet spells could be used for analyzing rainfall data to obtain specific information needed for crop planning and for carrying out agricultural operations (Srinivasa Reddy et al.,). Markov Chain probability model has been recognized as a suitable model to explain the long term frequency behaviour of wet or dry weather spells. In this model, the conditional probability has been accepted as fully justified in the analysis of weekly rainfall data. Several research scholars have demonstrated its practical utility in agricultural planning for both long and short term periods. This model enables to determine the probability of occurrence of dry and wet weather during a particular week (5, 10, and 13). The agriculture is the mainstay of the economy of the people of Orathanadu Taluk. It is one of the major paddy growing taluks of Thanjavur District, Tamil Nadu (Figure 1). It is located between 10°24’40.84”N and 10°45’20.33”N latitudes and between 79°8’10.27”E and 79°24’42.85”E longitudes. The total geographical area of the study area is 585.74 Km 2 . Many authors have used Markov chains to model the daily occurrence of precipitation. Gabriel and Neumann analyzed the occurrence of rain at Tel Aviv, Israel, by fitting a two- sate, first-order Markov chain. The two states corresponded to ‘rain’ and ‘no rain’. They used Markov chain probability model to study the data of daily rainfall occurrence at Tel Aviv.
Transcript
Page 1: Markov Chain Model for probability of weekly rainfall in ... · observed in Bangladesh by Sayedur Rahman . Markov ... Markov Chain Model for probability of weekly rainfall in Orathanadu

INTERNATIONAL JOURNAL OF GEOMATICS AND GEOSCIENCES

Volume 3, No 1, 2012

© Copyright by the authors - Licensee IPA- Under Creative Commons license 3.0

Research article ISSN 0976 – 4380

Submitted on May 2012 published on July 2012 191

Markov Chain Model for probability of weekly rainfall in Orathanadu

Taluk, Thanjavur District, Tamil Nadu Senthilvelan.A

1, Ganesh.A

2, Banukumar.K

3

1-Assistant Professor, Department of Geography, Government Arts College (Autonomous),

Kumbakonam

2- Professor, Department of Geography, Bharathidasan University, Tiruchirappalli.

3- Assistant Professor, Department of Geography, Periyar EVR College (Autonomous),

Tiruchirappalli

[email protected]

ABSTRACT

It is necessary to know the sequence of dry and wet periods for successful agricultural

management and planning of soil and water conservation measures. Probability analysis is a

very useful tool for making important decisions in agricultural operation. In this study,

Markov Chain Model has been extensively used to study spell distribution. For this purpose a

week period was considered as the optimum length of time. The present study has been

carried out to find the probabilities of occurrence of wet week (W), wet week preceded by

wet week (W/W) at different threshold limits of 10 and 20 mm. On the basis of the analysis

the following conclusions are made: (a) the 3 and 31/2

month varieties are best suited for

Vettikkadu region; (b) 31/2

and 4 months paddy varieties can be successfully grown in

Neivasal Thenpathi area and (c) 4 and 41/2

months paddy varieties are favourably grown in

Orathanadu region.

1. Introduction

The yield of crops particularly in rainfed condition depends on the rainfall pattern. Simple

criteria related to sequential phenomena like dry and wet spells could be used for analyzing

rainfall data to obtain specific information needed for crop planning and for carrying out

agricultural operations (Srinivasa Reddy et al.,). Markov Chain probability model has been

recognized as a suitable model to explain the long term frequency behaviour of wet or dry

weather spells. In this model, the conditional probability has been accepted as fully justified

in the analysis of weekly rainfall data. Several research scholars have demonstrated its

practical utility in agricultural planning for both long and short term periods. This model

enables to determine the probability of occurrence of dry and wet weather during a particular

week (5, 10, and 13). The agriculture is the mainstay of the economy of the people of

Orathanadu Taluk. It is one of the major paddy growing taluks of Thanjavur District, Tamil

Nadu (Figure 1). It is located between 10°24’40.84”N and 10°45’20.33”N latitudes and

between 79°8’10.27”E and 79°24’42.85”E longitudes. The total geographical area of the

study area is 585.74 Km2.

Many authors have used Markov chains to model the daily occurrence of precipitation.

Gabriel and Neumann analyzed the occurrence of rain at Tel Aviv, Israel, by fitting a two-

sate, first-order Markov chain. The two states corresponded to ‘rain’ and ‘no rain’. They used

Markov chain probability model to study the data of daily rainfall occurrence at Tel Aviv.

Page 2: Markov Chain Model for probability of weekly rainfall in ... · observed in Bangladesh by Sayedur Rahman . Markov ... Markov Chain Model for probability of weekly rainfall in Orathanadu

Markov Chain Model for probability of weekly rainfall in Orathanadu Taluk, Thanjavur District, Tamil

Nadu

Senthilvelan.A et al

International Journal of Geomatics and Geosciences

Volume 3 Issue 1, 2012 192

This accounts for the form of distribution of dry and wet spells. Further aspects of rainfall

occurrence patterns are derived. In particular, the distribution of the number of rainy days per

week is obtained.

The fitting and the use of models for daily rainfall observation are discussed by Stern and

Coe . They found that the non-stationary Markov chains are fitted to the occurrence of rain,

gamma distribution with parameters vary with time of year, are fitted to rainfall amount.

They also concluded that numerical methods are used to derive results from these models that

are important in agricultural planning. A rainfall simulation model based on a first-order

Markov chain has been developed to simulate the annual variation in rainfall amount that is

observed in Bangladesh by Sayedur Rahman . Markov chain models have been used to

evaluate probabilities of getting a sequence of wet and dry weeks during South-west

monsoon period over the districts Purulia in West Bengal and Giridih in Bihar state and dry

farming tract in the state of Maharashtra of India by Pabitra Banik .

Srinivasa Reddy, et.al employed Markov chain model to analyze probability of dry and wet

weeks for agricultural planning at Bangalore. An application of stochastic process for

describing and analyzing the daily rainfall pattern at Aduthurai is presented by Samuel

Selvaraj, et.al. They suggest that first order Markov chaining with two parameters gamma

distributions were found to be adequate to generate daily rainfall sequences at Aduthurai.

Pandharinath (10) applied Markov Chain model in Andhra Pradesh, India and identified the

wet and dry weeks during monsoon period. He considered 20mm/week as the threshold limit,

to distinguish between dry and wet weeks. Rajendram and Sivasamy applied Markov Chain

model at 10,20,40 and 80mm threshold levels to analyse probability of weekly rainfall and its

agro-climatic implications to paddy crop over Batticaloa, Sri Lanka.

Figure 1: Study area map

Page 3: Markov Chain Model for probability of weekly rainfall in ... · observed in Bangladesh by Sayedur Rahman . Markov ... Markov Chain Model for probability of weekly rainfall in Orathanadu

Markov Chain Model for probability of weekly rainfall in Orathanadu Taluk, Thanjavur District, Tamil

Nadu

Senthilvelan.A et al

International Journal of Geomatics and Geosciences

Volume 3 Issue 1, 2012 193

2. Data and Methodology

The daily rainfall data have been collected for three stations of Orathanadu Taluk , Thanjavur

District, viz., Neivasal Thenpathi, Vettikkadu and Orathanadu from the Department of

Economics and statistics, Chennai. Daily rainfall data for Neivasal Thenpathi and Vettikkadu

stations were available for 30 years (1981-2010) whereas Orathanadu station has 18 years of

data (1993-2010). The long term mean and its maximum and minimum, standard deviation

and coefficient of variation of the rainfall in meteorological standard weeks have been

calculated (Table 1). For the purpose of agricultural planning, Reddy applied Markov Chain

model in Combolcha station, Ethiopia by choosing 10,20,40 and 80mm/week as threshold

limits. These threshold levels were considered as adequate for the crop activities such as land

preparation (10mm), crop planting or sowing (20mm), and application of fertilizer and/or

weeding (40mm). According to Reddy , in a given week i of a given year received more than

20mm/week at more than 50%(W/W) threshold level, then week i is the right time for

planting. If weeding/fertilizer application is to be carried out in week i then the week should

have at least 75%(W/W) probability at 40mm/week. if the interest is when not apply

fertilizer/pesticides, then one can use the probability estimate at 80mm/week. If fertilizer

and/or insecticides/pesticides are applied on week i then W should not exceed 25%

probability level at 80mm/week. Rajendram and Sivasami applied the same threshold limits

to estimate the weekly rainfall probabilities over Batticaloa, Sri Lanka. In the present study,

Markov Chain model has been used and the same threshold limits have been employed for

estimating the weekly rainfall probability.

With the following notations, the initial and conditional probabilities for the above said

threshold limits have been computed using the following formula. Under initial probabilities,

the probability of a given week as wet or dry is estimated, whereas in the case of conditional

probabilities, if a given week ‘i’ is wet, then the chances of (i+k)th

period as wet, wet/wet or

dry/dry are estimated. A period is said to be wet when the parameter of that period exceeds a

threshold limit and to be dry when under the limit.

P(W) = F(W)/n

P(W/W) = F(W/W)/F(W)

P(W) – Probabilities of the week being wet

P(W/W) – Conditional probability of wet week preceded by a wet week

n= number of years of data

F(W) – Frequency of wet weeks

F(W/W) – Frequency of wet weeks preceded by another wet week

3. Rainfall probabilities of Orathanadu Taluk

The major crops cultivated in Orathanadu Taluk are paddy, pulses, groundnut and sugarcane.

Paddy is the principal crop growing in three seasons, viz., kuruvai (Sornavari), samba/

Thaladi (Pishanam) and Kodai (Navarai). The first crop is known as ‘kuruvai’, the short term

crop with a duration of 100-115 days from June/July to October/November. The paddy crop

cultivated after harvesting the kuruvai crop is called ‘Thaladi’, which has a duration of 115-

120 days from October/November to February/March. The more important season for paddy

crop in the study area is samba season. Samba is a long term crop with a duration of 135 days

to 150 days from August /September/October to Jan/Feb/March.

Page 4: Markov Chain Model for probability of weekly rainfall in ... · observed in Bangladesh by Sayedur Rahman . Markov ... Markov Chain Model for probability of weekly rainfall in Orathanadu

Markov Chain Model for probability of weekly rainfall in Orathanadu Taluk, Thanjavur District, Tamil

Nadu

Senthilvelan.A et al

International Journal of Geomatics and Geosciences

Volume 3 Issue 1, 2012 194

Table 1: Descriptive Statistics on weekly rainfall (mm) for Vettikkadu (1981-2010)

Week Month Date Total Mean Max Min SD CV

W23 June 04 - 10 363.90 12.13 74.40 0.0 20.96 172.82

W24 11 - 17 478.40 15.95 149.00 0.0 30.42 190.76

W25 18 - 24 321.20 10.71 123.00 0.0 24.79 231.51

W26 25 - 01 270.40 9.01 46.50 0.0 13.55 150.29

W27 July 02 - 08 385.60 12.85 100.00 0.0 22.60 175.84

W28 09 - 15 516.30 17.21 136.00 0.0 26.77 155.55

W29 16 - 22 406.90 13.56 65.00 0.0 19.88 146.57

W30 23 - 29 389.30 12.98 81.00 0.0 20.25 156.08

W31 30 - 05 462.00 15.40 86.70 0.0 21.09 136.98

W32 August 06 - 12 304.00 10.13 83.00 0.0 18.21 179.75

W33 13 - 19 506.70 16.89 94.60 0.0 24.68 146.12

W34 20 - 26 918.60 30.62 133.60 0.0 36.91 120.54

W35 27 - 02 670.60 22.35 92.00 0.0 22.08 98.79

W36 September 03 - 09 455.20 15.17 90.40 0.0 23.12 152.38

W37 10 - 16 865.30 28.84 113.60 0.0 30.99 107.45

W38 17 - 23 692.90 23.10 174.00 0.0 34.99 151.49

W39 24 - 30 957.20 31.91 185.90 0.0 37.99 119.07

W40 October 01 - 07 767.50 25.58 161.20 0.0 31.83 124.40

W41 08 - 14 718.90 23.96 154.30 0.0 34.74 144.99

W42 15 - 21 1134.10 37.80 137.00 0.0 38.73 102.46

W43 22 - 28 1374.10 45.80 148.40 0.0 40.19 87.75

W44 29 - 04 1712.80 57.09 224.60 0.0 60.73 106.37

W45 November 05 - 11 2229.40 74.31 258.20 0.0 77.63 104.46

W46 12 - 18 1770.60 59.02 620.00 0.0 119.52 202.51

W47 19 - 25 1615.90 53.86 319.40 0.0 82.56 153.28

W48 26 - 04 1565.00 52.17 493.30 0.0 102.79 197.05

W49 December 03 - 09 1306.40 43.55 168.00 0.0 48.84 112.17

W50 10 - 16 831.80 27.73 267.00 0.0 54.19 195.45

W51 17 - 23 1389.20 46.31 494.00 0.0 105.19 227.17

W52 24 - 31 382.20 12.74 96.40 0.0 23.38 183.55

Table 2: Descriptive Statistics on weekly rainfall (mm) for Neivasal Thenpathi (1981-2010)

Week Month Date Total Mean Max Min SD CV

W23 June 04 - 10 269.60 8.99 97.60 0.0 19.58 217.88

W24 11 - 17 372.20 12.41 125.20 0.0 26.99 217.56

W25 18 - 24 299.70 9.99 75.00 0.0 19.64 196.60

W26 25 - 01 124.00 4.13 46.50 0.0 10.44 252.61

W27 July 02 - 08 358.10 11.94 81.20 0.0 18.67 156.41

W28 09 - 15 353.40 11.78 86.00 0.0 19.20 163.03

W29 16 - 22 273.50 9.12 53.30 0.0 14.38 157.69

W30 23 - 29 418.70 13.96 129.40 0.0 26.22 187.88

W31 30 - 05 467.30 15.58 107.70 0.0 28.93 185.72

W32 August 06 - 12 471.40 15.71 99.40 0.0 26.07 165.91

W33 13 - 19 704.00 23.47 135.20 0.0 31.84 135.68

W34 20 - 26 824.00 27.47 121.20 0.0 32.95 119.97

W35 27 - 02 1016.70 33.89 152.00 0.0 40.77 120.32

W36 September 03 - 09 584.40 19.48 110.60 0.0 28.05 144.00

Page 5: Markov Chain Model for probability of weekly rainfall in ... · observed in Bangladesh by Sayedur Rahman . Markov ... Markov Chain Model for probability of weekly rainfall in Orathanadu

Markov Chain Model for probability of weekly rainfall in Orathanadu Taluk, Thanjavur District, Tamil

Nadu

Senthilvelan.A et al

International Journal of Geomatics and Geosciences

Volume 3 Issue 1, 2012 195

W37 10 - 16 965.30 32.18 147.60 0.0 37.90 117.79

W38 17 - 23 765.60 25.52 183.00 0.0 36.90 144.58

W39 24 - 30 804.10 26.80 166.90 0.0 35.31 131.73

W40 October 01 - 07 910.00 30.33 120.70 0.0 32.09 105.78

W41 08 - 14 768.30 25.61 97.50 0.0 30.48 119.01

W42 15 - 21 932.60 31.09 130.80 0.0 33.36 107.30

W43 22 - 28 1577.50 52.58 255.60 0.0 56.57 107.57

W44 29 - 04 1788.20 59.61 248.80 0.0 72.55 121.71

W45 November 05 - 11 2472.10 82.40 292.70 0.0 88.71 107.66

W46 12 - 18 1151.90 38.40 197.60 0.0 54.19 141.13

W47 19 - 25 1620.50 54.02 291.90 0.0 72.60 134.41

W48 26 - 04 1812.20 60.41 526.40 0.0 111.14 183.98

W49 December 03 - 09 1650.00 55.00 302.70 0.0 79.41 144.39

W50 10 - 16 1049.40 34.98 289.00 0.0 67.58 193.19

W51 17 - 23 1462.10 48.74 549.00 0.0 109.61 224.91

W52 24 - 31 431.50 14.38 131.70 0.0 28.82 200.35

It is pointed out that the definite dates of opening and closing of Mettur Dam water for

irrigation is not known. This reflects in planning cropping sequence of farmer’s choice. This

constraint not only affects kuruvai, thaladi and samba rice growing seasons, but also has an

impact on the growing of the summer season crops. The torrential rain during Northeast

monsoon hinders both kuruvai harvest as well as Thaladi transplanting. It also causes larger

patches of low lying lands inundated. So, it is essential to know the probability of receiving

rainfall at a particular week for sowing operation. According to Reddy (18), at 20mm/week

threshold limit, the conditional probability (W/W) of a given week ‘i’ is >50% (under k=1),

then week ‘i’ is appropriate for sowing/planting . If not, ‘i’ is not the right time for planting.

Table 3: Descriptive Statistics on weekly rainfall (mm) for Orathanadu (1993-2010)

Week Month Date Total Mean Max Min SD CV

W23 June 04 - 10 365.50 20.31 138.70 0.0 36.57 180.09

W24 11 - 17 683.20 37.96 138.20 0.0 44.75 117.89

W25 18 - 24 351.50 19.53 243.70 0.0 57.55 294.72

W26 25 - 01 72.00 4.00 42.60 0.0 10.32 258.01

W27 July 02 - 08 330.00 18.33 78.00 0.0 28.49 155.38

W28 09 - 15 300.60 16.70 115.40 0.0 29.27 175.27

W29 16 - 22 351.20 19.51 144.60 0.0 35.66 182.77

W30 23 - 29 180.40 10.02 58.00 0.0 17.85 178.14

W31 30 - 05 459.90 25.55 124.20 0.0 39.29 153.78

W32 August 06 - 12 184.20 10.23 74.30 0.0 18.29 178.68

W33 13 - 19 863.30 47.96 164.40 0.0 58.98 122.98

W34 20 - 26 1077.30 59.85 198.40 0.0 72.92 121.83

W35 27 - 02 637.60 35.42 149.20 0.0 45.30 127.89

W36 September 03 - 09 503.30 27.96 157.40 0.0 40.04 143.20

W37 10 - 16 534.70 29.71 112.00 0.0 37.78 127.19

W38 17 - 23 532.80 29.60 91.60 0.0 28.88 97.56

W39 24 - 30 859.30 47.74 186.00 0.0 52.41 109.79

W40 October 01 - 07 632.70 35.15 180.00 0.0 52.99 150.74

W41 08 - 14 666.50 37.03 131.80 0.0 40.90 110.45

W42 15 - 21 998.90 55.49 131.00 0.0 36.08 65.01

W43 22 - 28 1451.20 80.62 308.40 0.0 72.55 89.99

W44 29 - 04 1344.40 74.69 214.80 0.0 69.69 93.31

Page 6: Markov Chain Model for probability of weekly rainfall in ... · observed in Bangladesh by Sayedur Rahman . Markov ... Markov Chain Model for probability of weekly rainfall in Orathanadu

Markov Chain Model for probability of weekly rainfall in Orathanadu Taluk, Thanjavur District, Tamil

Nadu

Senthilvelan.A et al

International Journal of Geomatics and Geosciences

Volume 3 Issue 1, 2012 196

W45 November 05 - 11 2165.30 120.29 362.80 0.0 122.79 102.07

W46 12 - 18 763.90 42.44 195.00 0.0 53.99 127.23

W47 19 - 25 2083.90 115.77 370.80 24.0 108.85 94.02

W48 26 - 04 1994.00 110.78 1103.80 0.0 261.38 235.95

W49 December 03 - 09 1050.00 58.33 256.20 0.0 82.98 142.25

W50 10 - 16 1075.20 59.73 280.70 0.0 83.59 139.93

W51 17 - 23 1216.40 67.58 532.00 0.0 150.24 222.32

W52 24 - 31 255.10 14.17 148.00 0.0 35.24 248.64

Figure 2: Conditional probabilities of wet week preceded by wet week P (w/w) at 10 mm

threshold limit

Page 7: Markov Chain Model for probability of weekly rainfall in ... · observed in Bangladesh by Sayedur Rahman . Markov ... Markov Chain Model for probability of weekly rainfall in Orathanadu

Markov Chain Model for probability of weekly rainfall in Orathanadu Taluk, Thanjavur District, Tamil

Nadu

Senthilvelan.A et al

International Journal of Geomatics and Geosciences

Volume 3 Issue 1, 2012 197

Figure 3: Conditional probabilities of wet week preceded by wet week P (w/w) at 20 mm

threshold limit

3.1.Vettikkadu: The weekly mean rainfall and its descriptive statistics for this station are

shown in table 1. The mean monthly or weekly rainfall data gives only trends of certain

climatic pattern. They can be useful tool to indicate agro-climatic homogenous zone. But they

do not give any information on the temporal rainfall variability. The maximum rainfall of

620mm has been recorded during the study period for the standard meteorological week 46,

which comes under Samba crop season. The samba season is a major rice growing season in

Vettikkadu. The estimation of co-efficient of variation (CV) of rainfall is more suited for

agricultural purposes. The higher the CV, the lesser the dependability of rainfall and vice

versa. In orathanadu taluk, the planting/sowing of Kuruvai crop depends on the opening of

Mettur Dam for irrigation. Because the planting is done in the month of June/July which falls

during the standard weeks of 23rd

-26th

week /27th

-30th

week . During the Kuruvai season the

co-efficient of variation of rainfall ranges from 150.29% to 172.82% which reflect the higher

Page 8: Markov Chain Model for probability of weekly rainfall in ... · observed in Bangladesh by Sayedur Rahman . Markov ... Markov Chain Model for probability of weekly rainfall in Orathanadu

Markov Chain Model for probability of weekly rainfall in Orathanadu Taluk, Thanjavur District, Tamil

Nadu

Senthilvelan.A et al

International Journal of Geomatics and Geosciences

Volume 3 Issue 1, 2012 198

variability of rainfall during this period. The initial and conditional probabilities occurrences

of wet week and wet week followed by wet week at 10 mm and 20 mm threshold limits for

the same period also shows less than 50%. Therefore rainfed crops cannot be successfully

grown during Kuruvai crop season.

The planting/sowing Samba crop starts in the month of August/September/October. The co-

efficient of rainfall during the sowing period of Samba crop varies from 98.79% to 179.75%.

According to the crop planning-climate Atlas (2003), the threshold limit for Co-efficient of

Variation for weekly rainfall should be less than 150%. During the Samba crop season the

CV is less than threshold limit of 150 percent (except week no. 32). This indicates higher

dependability of rainfall during this period. Hence, agricultural operations like

planting/sowing can be undertaken successfully during this period. The initial and conditional

probabilities of rainfall during the Samba crop season at 10 mm and 20 mm threshold limits

have been shown in Tables 4 and 5 for this station. During this season particularly between

week 32 and week 44, the occurrence of probability of wet week ranges from 13 percent to

67 percent at 20mm/week limit. The probability of occurrence of a wet week preceded by a

wet week (W/W) is ranging from 13 percent to 74 percent during the same season.

3.2. Neivasal Thenpathi: The mean weekly rainfall and other statistics of this station have

been shown in Table 2. The maximum rainfall of 549mm has been received in week 51. The

Samba crop season is the major paddy growing season in this area also. The Co-efficient of

Variation of rainfall for Kuruvai season varies from 156.41 percent to 252.61 percent which

clearly indicate higher variability of rainfall during this crop season. The values of rainfall

variability are higher than the threshold limit of 150 percent. Therefore, rainfed crops cannot

be grown in this period. The probability of occurrence of wet week at 10mm threshold limit

for the same season varies from 20 to 37 percent. So, it is highly impossible for farmers to

undertake land preparation activities during Kuruvai season.

The Co-efficient of variation of rainfall for Samba season varies from 119.97 percent to

185.72 percent. The variability of rainfall is less than 150 percent during the sowing period of

this season (except week 32). The initial and conditional probabilities of rainfall during the

Samba crop season at 10 mm and 20 mm threshold limits have been shown in Tables 4 and 5

for this station.The probability of occurrence of wet week during the sowing period of Samba

season (i.e., from week 34 to week 44) ranges from 33 percent to 70 percent at 20mm/week

threshold limit. The conditional probability of W/W at 20mm/week during the same season

varies from 31 percent to 71 percent. So, the week having 50 percent probability level should

be carefully chosen for planting. In this analysis, even though the samba crop season starts in

the month of August (week 32), the standard week 40 has the conditional probability of 65

percent and the consecutive weeks also has the level of probability above the threshold limit

of >50 percent. Hence, planting/sowing of rainfed crops can be successfully undertaken from

40th

week.

3.3. Orathanadu: The descriptive statistics of this station is given in Table 3. The maximum

rainfall recorded during the study period was 1103.8mm in week 48. The major paddy

growing season for this area is Samba crop season. The Co-efficient of Variation of rainfall

during kuruvai season varies from 117.89% (week 24) to 294.72% (week 25). The variability

is above the threshold limit of 150 percent except for week 24. It indicates the dependability

of rainfall during the season is lesser. The probability of occurrence of wet week for the

kuruvai crop season at 10mm/week threshold limit varies 7 percent to 33 percent. It is also

less than 50 percent at 20mm/week threshold limit. The initial and conditional probabilities of

Page 9: Markov Chain Model for probability of weekly rainfall in ... · observed in Bangladesh by Sayedur Rahman . Markov ... Markov Chain Model for probability of weekly rainfall in Orathanadu

Markov Chain Model for probability of weekly rainfall in Orathanadu Taluk, Thanjavur District, Tamil

Nadu

Senthilvelan.A et al

International Journal of Geomatics and Geosciences

Volume 3 Issue 1, 2012 199

rainfall during the Samba crop season at 10 mm and 20 mm threshold limits have been shown

in Tables 4 and 5. for this station. Therefore agricultural activities like land preparation,

sowing/planting cannot be taken up during this season.

The Co-efficient of Variation of rainfall during the Samba sowing period

(August/September/October) varies from 89.99% (week43) to 150.74% (week40). The lesser

variability of rainfall during this season indicates higher dependability of rainfall. The

probability of occurrence of wet week during samba season varies from 10% to 50% at

20mm/week threshold limit. The conditional probability of W/W at the same threshold limit

is ranging from 13 percent to 87 percent (Table 13 and Fig.11). Therefore the week having

the conditional probability of 50 percent and above at 20mm/week threshold limit should be

carefully chosen for samba crop planting/sowing. In present analysis, 35th

week records 75

percent probability level and the consecutive weeks also have the conditional probability

above the threshold limit of 50 percent and above. Therefore rainfed agricultural activities

can be initiated starting from 35th

standard week.

3.4. Spatial Distribution of Rainfall Probabilities in Orathanadu Taluk

The spatial distribution map of rainfall probability has been prepared using spline method,

one of the interpolation techniques of the spatial analyst tools in Arc GIS environment. The

conditional probabilities of wet week preceded by wet week P(W/W) at 10 mm and 20 mm

threshold limit have been summarised in tables 4 and 5 respectively. The resultant map

( figures 2 & 3) shows that the north, northeast, east, southeast and northwestern part of the

study area experience 50 – 70 percent probability at 10mm threshold level in 34th

, 35th

and

39th

standard meteorological weeks whereas the entire Orathanadu taluk enjoys the same

probability level in 36th

,38th

and 40th

meteorological weeks. It is concluded from fig.3 that the

conditional probability of 50 percent and above can be identified in 35th

and

Table 4: Conditional probabilities of wet week preceded by a wet week P(W/W)at 10mm

threshold limit

Rainfall Stations

Weeks Vettikkadu

Neivasal

Thenpathi Orathanadu

W23 0.14 0.17 0.57

W24 0.45 0.22 0.40

W25 0.50 0.25 0.75

W26 0.27 0.67 0.50

W27 0.44 0.00 0.14

W28 0.36 0.50 0.38

W29 0.55 0.44 0.63

W30 0.50 0.36 0.40

W31 0.33 0.33 0.25

W32 0.38 0.10 0.80

W33 0.31 0.38 0.33

W34 0.67 0.56 0.50

W35 0.56 0.68 0.89

W36 0.67 0.71 0.73

W37 0.37 0.47 0.75

Page 10: Markov Chain Model for probability of weekly rainfall in ... · observed in Bangladesh by Sayedur Rahman . Markov ... Markov Chain Model for probability of weekly rainfall in Orathanadu

Markov Chain Model for probability of weekly rainfall in Orathanadu Taluk, Thanjavur District, Tamil

Nadu

Senthilvelan.A et al

International Journal of Geomatics and Geosciences

Volume 3 Issue 1, 2012 200

W38 0.80 0.65 0.75

W39 0.48 0.59 0.62

W40 0.80 0.65 0.73

W41 0.63 0.69 0.50

W42 0.59 0.60 0.67

W43 0.75 0.78 0.88

W44 0.81 0.78 0.93

W45 0.71 0.77 0.80

W46 0.88 0.80 0.90

W47 0.63 0.50 0.56

W48 0.65 0.58 1.00

W49 0.60 0.67 0.70

W50 0.67 0.60 0.73

W51 0.42 0.33 0.57

W52 0.50 0.56 0.60

Table 5: Conditional probabilities of wet week preceded by a wet week P(W/W)at 20mm

threshold limit

Rainfall Stations

Week Vettikkadu

Neivasal

Thenpathi Orathanadu

W23 0.14 0.00 0.20

W24 0.43 0.33 0.25

W25 0.40 0.00 1.00

W26 0.50 0.33 0.00

W27 0.00 0.00 0.20

W28 0.30 0.33 0.40

W29 0.83 0.14 0.33

W30 0.29 0.14 0.25

W31 0.25 0.33 0.17

W32 0.00 0.13 0.67

W33 0.13 0.33 0.13

W34 0.43 0.57 0.44

W35 0.67 0.60 0.75

W36 0.43 0.60 0.86

W37 0.38 0.31 0.67

W38 0.70 0.62 0.30

W39 0.25 0.46 0.70

W40 0.69 0.65 0.63

W41 0.27 0.64 0.33

W42 0.37 0.65 0.64

W43 0.65 0.60 0.87

W44 0.74 0.71 0.85

W45 0.70 0.70 0.77

W46 0.71 0.69 0.88

W47 0.53 0.50 0.44

W48 0.46 0.64 1.00

Page 11: Markov Chain Model for probability of weekly rainfall in ... · observed in Bangladesh by Sayedur Rahman . Markov ... Markov Chain Model for probability of weekly rainfall in Orathanadu

Markov Chain Model for probability of weekly rainfall in Orathanadu Taluk, Thanjavur District, Tamil

Nadu

Senthilvelan.A et al

International Journal of Geomatics and Geosciences

Volume 3 Issue 1, 2012 201

W49 0.44 0.53 0.50

W50 0.63 0.67 0.63

W51 0.50 0.30 0.33

W52 0.17 0.43 0.67

40th

weeks in the entire Orathanadu taluk. The same probability level is experienced in almost

entire study area in 36th

week except Vettikkadu area.

4. Conclusion

The initial crop activities start with the land preparation. According to Reddy (20), the initial

and conditional probabilities at 10mm/week threshold limit is considered for the land

preparation. The conditional probability of occurrence of rainfall at 20mm/week threshold

limit can be adopted for the crop sowing/planting. The following conclusions emerge from

the above analysis.

(a) Vettikkadu: The initial and conditional probabilities at 10mm/week threshold level is

calculated as more than 50% in the 40th

week (W = 67% and W/W = 80%). Therefore, land

preparation can be taken up in 40th

standard week and can be extended up to 43rd

week. The

chances of occurrence of 20mm/week is above 50% in 43rd

week which is the optimum time

for planting 3 months paddy varieties and 42nd

week for sowing 31/2

months varieties.

Considering the above probabilities, it is found that the 3 and 31/2

month varieties are best

suited for this area.

(b) Neivasal Thenpathi : The initial and conditional probabilities at 10mm/week threshold

limit is reported as above 50 percent in the 38th

week (W = 57% and W/W = 65%). Hence,

land preparation can be initiated from 38th

to 40th

week. The likelihood of occurrence of

20mm/week is more than 50% in 40th

week which will be the favourable time for sowing 4

months paddy varieties and 31/2

months varieties can be sown in 42nd

week. It is concluded

that 31/2

and 4 months paddy varieties can be successfully grown in this region.

(c) Orathanadu : The initial and conditional probabilities for land preparation is estimated at

more than 50% in the 36th

week (W = 61% and W/W = 73%). So, the land preparation can be

undertaken from 36th

to 39th

week. The conditional probability of occurrence of rainfall for

planting is above 50% in 39th

standard week (W/W = 70%) which is best suited for sowing

41/2

months paddy variety. The 4 month variety can also be sown in 40th

week. It is therefore

concluded that 4 and 41/2

months paddy varieties are favourably grown in this area.

5. References

1. Alaguraja P. Nagarathinam S.R. et al (2010), Rainfall Distribution Study in

Coimbatore District TamilNadu Using GIS, I.K International Publication, New Delhi,

5, pp 92-115.

2. Banukumar.K, Aruchamy.S, (2007), Climatic Types of Tamil Nadu, India, , Journal

of Spatial Science, 1(1&2), pp 1-8.

3. Banukumar K., Rajamanickam G.V., and Aruchamy, S, (2005), Study of Drought

Prone Areas in Pudukkottai Taluk, Tamil Nadu -A

Page 12: Markov Chain Model for probability of weekly rainfall in ... · observed in Bangladesh by Sayedur Rahman . Markov ... Markov Chain Model for probability of weekly rainfall in Orathanadu

Markov Chain Model for probability of weekly rainfall in Orathanadu Taluk, Thanjavur District, Tamil

Nadu

Senthilvelan.A et al

International Journal of Geomatics and Geosciences

Volume 3 Issue 1, 2012 202

Hydrogomorphological Approach, Indian Journal of Geomorphology, 10(1&2), pp

23-36.

4. Chandler E. Richard and Howard S. Wheater (2002), Analysis of Rainfall Variability

using generalized linear models: A case study from the west of Ireland, Water

Resources Research, 38(10), 1192-1193.

5. Chaudhary J.L. (2006), Rainfall Analysis using Markov Chain Modeling for Bastar,

Geographical Review of India, 60(1), pp 74-83.

6. Dhar, O.N et.al. (1982), Trends and fluctuations of seasonal and annual rainfall of

Tamil Nadu, Proceedings of Indian Academy of Science,(Earth Planet Science), 91(2),

pp 97-104.

7. Di Giuseppe, E. et.al. (2005), Analysis of dry and wet spells from 1870 to 2000 in

four Italian sites, Geophysical Research Abstracts, 7(6).

8. Kirupalani, R.H. and Aswini Kulkharni (1997), Rainfall Variability over South East

Asia – Connections with Indian Monsoon and ENSO Extremes: New Perspectives,

International Journal of Climatology, 17, pp 1155-1168.

9. Maria Mimimou (1983), Daily precipitation occurrences modeling with Markov

Chain of seasonal order, Hydrological Sciences, 28, pp 221-232.

10. Nicholson, S.E. (1999), An Analysis of Recent Rainfall Conditions in West Africa,

Including the Rainy Seasons of the 1997 El Niño and the 1998 La Niña Years, Journal

of Climate, Article: pp 2628–2640.

11. Pabitra Banik, et.al., (2000), Markov Chain Analysis of Weekly Rainfall data in

determining drought proneness, Discrete Dynamics in Nature and Society, 7, pp 231-

239.

12. Raghavendra Ramanan. S (2006), Water Balance Modeling of Rain fed crops: A

study of sorghum and groundnut in Salem and Namakkal Districts, Unpublished

Ph.D., Thesis, Department of Geography, University of Madras, Madras.

13. Rajeevan, M. et.al. (2006), High Resolution daily gridded rainfall data for the Indian

Region: Analysis of break and active monsoon spells, Current Science, 91,3, pp 296.

14. Rajendram, K and Sivasami, K.S. (2005), Markov Chain Model for Probability of

Weekly Rainfall and its Agro-Climatic Implications to Paddy Crop over Batticaloa,

Sri Lanka,The Indian Geographical Journal, 79(2),pp 83-97.

15. Ramamurthy. K (1943), A Study of Rainfall Regime at Vellore, Indian Geographical

Journal, 18, pp 197-203.

16. Ramamurthy. K (1948), Some aspects of the Regional Geography of Tamilnad,

Climate, Indian Geographical Journal, 23, pp 20-64.

17. Ramamurthy. K (1972), A Study of Rainfall Regimes in India, Text Book, University

of Madras, Madras.

Page 13: Markov Chain Model for probability of weekly rainfall in ... · observed in Bangladesh by Sayedur Rahman . Markov ... Markov Chain Model for probability of weekly rainfall in Orathanadu

Markov Chain Model for probability of weekly rainfall in Orathanadu Taluk, Thanjavur District, Tamil

Nadu

Senthilvelan.A et al

International Journal of Geomatics and Geosciences

Volume 3 Issue 1, 2012 203

18. Rao, S.S. (1992), Upland cultivation in Bastar (Constraints-Solutions), Technical

Bulletin No. 14, ZARS, Jagdalpur, M.P., India .

19. Robertson, G.W. (1976), Dry and wet spells, UNDP/FAO, Tan Razok Agri. Res.

Center. Project Field Report, Agrometeorology, A-6, pp 15.

20. Samuel Selvaraj, R. and Tamil Selvi, S. (2010), Stochastic modeling of daily rainfall

at Aduthurai, International Journal of Advanced Computer and Mathematical

Sciences, 1(1), pp 52-57.

21. Sayedur Rahman, M. (1999), A Rainfall Simulation Model for Agricultural

Development in Bangladesh, Discrete Dynamics in Nature and Society, 5, pp 1-7.

22. Srikanthan, R and McMahon, T.A. (2001), Stochastic generation of annual, monthly

and daily climate data : A review, Hydrology and Earth System Sciences, 5(4),653-

670.

23. Srinivasa Reddy, G.V. et.al., (2008), Markov Chain Model Probability of Dry, Wet

Weeks and Statistical Analysis of Weekly Rainfall for Agricultural Planning at

Bangalore, Karnataka journal of agricultural sciences, 21(1),pp 12-16.

24. Stern, R. D. and R. Coe (1982), Fitting Models to Daily Rainfall Data, Journal of

Applied Meteorology, 21(1), pp 1024-1031.

25. Stern, R. D. and R. Coe (1984), A Model Fitting Analysis of Daily Rainfall Data,

Journal of the Royal Statistical Society. Series A (General), 147(1), , pp 1-34.

26. Subramaniam, A. R. and Sanjeeva Rao, P. (1989), Dry spell sequences in South

coastal Andhra, Mausam, 40(1), pp 57- 60.

27. Yukiko Hirabayashi, et.al., (2008), A 59-year (1948-2006) global near-surface

meteorological data set for land surface models. Part I : Development of daily forcing

and assessment of precipitation intensity, Hydrological Research Letters 2, pp 36-40.


Recommended