Beruflich Dokumente
Kultur Dokumente
International Journal of ComputerComputerand Technology (IJCET), ISSN 0976 6367(Print), International Journal of Engineering Engineering ISSN 0976 6375(Online) Volume 1, Number 1, May - June (2010), IAEME
IJCET
IAEME
MODELING AND PREDICTING THE MONTHLY RAINFALL IN TAMILNADU AS A SEASONAL MULTIVARIATE ARIMA PROCESS
M. Nirmala Research Scholar, Sathyabama University Rajiv Gandhi Road, Jeppiaar Nagar, Chennai 19 Email ID: monishram5002@gmail.com S. M. Sundaram Department of Mathematics, Sathyabama University Rajiv Gandhi Road, Jeppiaar Nagar, Chennai 19 Email ID: sundarambhu@rediffmail.com
ABSTRACT:
Amongst all weather happenings, rainfall plays the most imperative role in human life. The understanding of rainfall variability helps the agricultural management in planning and decision- making process. The important aspect of this research is to find a suitable time series seasonal model for the prediction of the amount of rainfall in Tamilnadu. In this study, Box-Jenkins model is used to build a Multivariate ARIMA model for predicting the monthly rainfall in Tamilnadu together with a predictor, Sea Surface Temperature for the period of 59 years (1950 2008) with a total of 708 readings.
INTRODUCTION:
Time series analysis is an important tool in modeling and forecasting. Among the most effective approaches for analyzing time series data is the model introduced by Box and Jenkins [1], Autoregressive Integrated Moving Average (ARIMA). Box-Jenkins ARIMA modeling has been successfully applied in various water and environmental
103
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976 6367(Print), ISSN 0976 6375(Online) Volume 1, Number 1, May - June (2010), IAEME
management applications. The association between the southwest and northeast monsoon rainfall over Tamilnadu have been examined for the 100 year period from 1877 1976 through a correlation analysis by O.N. Dhar and P.R. Rakhecha [2]. The average rainfall series of Tamilnadu for the northeast monsoon months of October to December and the season as a whole were analyzed for trends, periodicities and variability using standard statistical methods by O. N. Dhar, P. R. Rakhecha, and A. K. Kulkarni [3]. Balachandran S., Asokan R. and Sridharan S [4] examined the local and teleconnective association between Northeast Monsoon Rainfall (NEMR) over Tamilnadu and global Surface Temperature Anomalies (STA) using the monthly gridded STA data for the period 19012004. The trends, periodicities and variability in the seasonal and annual rainfall series of Tamilnadu were analyzed by O. N. Dhar, P. R. Rakhecha, and A. K. Kulkarni [5]. The annual rainfall in Tamilnadu was predicted using a suitable Box Jenkins ARIMA model by M. Nirmala and S.M.Sundaram [6].
104
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976 6367(Print), ISSN 0976 6375(Online) Volume 1, Number 1, May - June (2010), IAEME
105
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976 6367(Print), ISSN 0976 6375(Online) Volume 1, Number 1, May - June (2010), IAEME
same variable number of autoregressive delays xt1,xt2,. . . xtp and include a random disturbance et. The moving average models MA(q) generate predictions of a variable xt based on a number q of past disturbances of the same variable prediction errors of past values et1,et2,. . ., etq. The combination of the auto regressive and moving average models AR(p) and MA(q) generates more flexible models called ARMA(p,q) models. The stationarity of the time series is required for the implementation of all these models. In 1976, Box and Jenkins proposed the mathematical transformation of the non-stationary time series into stationary time series by a difference process defined by an order of integration parameter d. This transforms ARMA (p,q) models for non stationary transformed time series as the ARIMA (p,d,q) models. The ARIMA model building strategy includes iterative identification, estimation, diagnosis and forecasting stages [7]. Identification of a model may be accomplished on the basis of the data pattern, time series plot and using their autocorrelation function and partial autocorrelation function. The parameters are estimated and tested for statistical significance after identifying the tentative model. If the parameter estimates does not meet the stationarity condition then a new model should be identified and its parameters are estimated and tested. After finding the correct model it should be diagnosed. In the diagnosis process, the autocorrelation of the residuals from the estimated model should be sufficiently small and should resemble white noise. If the residuals remain significantly correlated among themselves, a new model should be identified estimated and diagnosed. Once the model is selected it is used to forecast the monthly rainfall series. Time series analysis provides great opportunities for detecting, describing and modeling climatic variability and impacts. Ultimately, to understand the meteorological information and integrate it into planning and decision making process, it is important to study the temporal characteristic and predict lead times of the rainfall of a region. This can be done by identifying the best time series model using Box Jenkins Seasonal ARIMA modeling techniques. The Seasonal ARIMA (p,d,q)(P,D,Q)s model is defined as
p ( B) P ( B s ) d s D yt = Q ( B s ) q ( B) t
[1]
106
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976 6367(Print), ISSN 0976 6375(Online) Volume 1, Number 1, May - June (2010), IAEME
Where
p ( B ) = 1 1 B ........... p B p , q ( B ) = 1 1 B ............ q B q
P ( B s ) = 1 1 B s ........ P B sP , Q ( B s ) = 1 1 B s ....... Q B sQ
..[2]
t denotes the error term, s and s are the non seasonal and seasonal autoregressive parameters and s and s are the non seasonal and seasonal moving average parameters.
MONTHLY RAINFALL
MONTH
Figure 2 Time Series Plot of Monthly Rainfall in Tamilnadu (1950 - 2008) The visual plot of the time series plot (Figure 2) is often enough to convince a statistician that the series is stationary or non stationary [6]. The visual inspection of the sample autocorrelation function shows that the rainfall series is stationary. If the Autocorrelation function dies out rapidly, that is, it reaches zero within one or two lag periods then it indicates that the time series is stationary. From the pattern of ACF and PACF plots, the monthly rainfall series is stationary but with seasonality.
107
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976 6367(Print), ISSN 0976 6375(Online) Volume 1, Number 1, May - June (2010), IAEME
Lag 1 2 3 4 5
The ACF and PACF correlograms (figure 3) and the coefficients are analyzed carefully and the tentative multivariate ARIMA model chosen is ARIMA (1,0,1)(1,1,1)12. The parameters estimates are tabulated in table 2.
108
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976 6367(Print), ISSN 0976 6375(Online) Volume 1, Number 1, May - June (2010), IAEME
After fitting the appropriate ARIMA model, the goodness of fit can be examined by plotting the ACF of residuals of the fitted model. If most of the autocorrelation coefficients of the residuals are within the confidence intervals then the model is a good fit.
Figure 4 Residual plots of ACF and PACF of ARIMA (1,0,1)(1,1,1)12 model Since the coefficients of the residual plots of ACF and PACF are lying within the confidence limits, the fit is a good fit and the error measures obtained through this model is tabulated in the Table 3. The graph showing the observed and fitted values is shown in Figure 5. Table 3: Error Measures obtained for the model ARIMA (1,0,1)(1,1,1)12 Error Measures RMSE MAPE R squared Value 1.082 16.081 0.547
109
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976 6367(Print), ISSN 0976 6375(Online) Volume 1, Number 1, May - June (2010), IAEME
Figure 5: Graph showing the observed and the fitted values obtained in ARIMA (1,0,1)(1,1,1)12 model
CONCLUSION:
In the modern world, there is an ever-increasing demand for more accurate weather forecasts. Accurate weather forecasting happens to be the ultimate target of atmospheric research. In this article, an attempt was made to predict the monthly rainfall in Tamilnadu with a predictor, sea surface temperature through Box Jenkins Multivariate ARIMA model. The error measures show the model fitted is a good model.
REFERENCE:
1. George E. P. Box, Gwilym M. Jenkins and Gregory C. Reinsel (1994). Time Series Analysis Forecasting and Control, 3rd Edition, Pearson Education, Inc. 2. O. N. Dhar and P. R. Rakhecha (1983). Foreshadowing Northeast Monsoon Rainfall over Tamilnadu, Monthly Weather Review, Vol.3, Issue 1, pp. 109 112. 3. O. N. Dhar, P. R. Rakhecha, A. K. Kulkarni (1982). Fluctuations in northeast monsoon rainfall of Tamilnadu, International Journal of Climatology
Volume 2, Issue 4, pp. 339 345. 4. Balachandran S., Asokan R. and Sridharan S (2006). Global surface temperature in relation to northeast monsoon rainfall over Tamilnadu, Journal of Earth System Science, vol. 115, no.3, pp. 349-362. 5. O. N. Dhar, P. R. Rakhecha, A. K. Kulkarni (1982). Trends and fluctuations of seasonal and annual rainfall of Tamilnadu, Proceedings of Indian Academy of Sciences (Earth planetary Sciences), Vol. 91, No.2, pp. 97 104. 110
International Journal of Computer Engineering and Technology (IJCET), ISSN 0976 6367(Print), ISSN 0976 6375(Online) Volume 1, Number 1, May - June (2010), IAEME
6. Nirmala M and Sundaram S.M. (2009), Rainfall Predicting Model: An Application to Tamilnadu Rainfall Series, Proceedings of National Conference on Effect of Climatic Change and Sustainable Resource Management, SRM University, and Meteorological Department, Chennai India. 7. Damodar N. Gujarati and Sangeetha (2007). Basic Econometrics, 4th edition, the Tata Mcgraw Hill Publishing Company Limited. 8. Meaza Demissie (2003). Characterizing trends and identification of rainfall predicting model: An application to Melksa observatory Nazreth rainfall series, Proceedings of Biometric society, pp.132 134. 9. Nirmala M and Sundaram S.M. (2010), Comparison of Learning Algorithms with the ARIMA Model for the Forecasting of Annual Rainfall in Tamilnadu, International Journal on Information Sciences and Computing, pp.64 71. Vol.4, No.1,
111