Statistical models for climate change in northern and central Kerala

Abstract

STATISTICAL MODELS FOR CLIMATE CHANGE IN NORTHERN KERALA The research work entitled “Statistical models for climate change in northern Kerala” was carried out at college of agriculture, Vellayani during 2018-2020. The objective was to develop statistical models to evaluate climate change over time across different regions of northern Kerala and to determine the effect of climate change on paddy production. The secondary data of rainfall, maximum temperature and minimum temperature was collected from RARS Pattambi for a period of 37 years (1982-2018) and from RARS Pilicode for a period of 36 years (1983-2018). The secondary data on paddy production from Palakkad district of Kerala was collected from the report of agricultural statistics, GOK for a period of 23 years (1995-2017). The descriptive statistics for the monthly data of weather parameters include mean, range, coefficient of variation, skewness and kurtosis. The analysis of weather data was done with the help of R programming and open software Gretl. The results of descriptive statistics and box plot of weather parameters showed that heavy rains was received in some months during several years at Pilicode as compared to Pattambi. Moreover, the Skewness for all the weather parameters of both Pilicode and Pattambi showed positive and negative skewness which indicated the absence of normal distribution among the weather data. The climate change over the years of different weather parameters was analysed in terms of trends and its presence and direction was determined. Shapiro-wilks test was initially conducted for the weather parameters which showed that all the parameters didn’t follow normal distribution for both Pilicode and Pattambi. The trend was detected using non parametric Mann-Kendall (MK) test and trend was estimated using Sen's slope estimator. The results of MK test and Sen's slope estimator revealed positive nonsignificant trend for annual, summer and monsoon rainfall and maximum temperature in all the seasons at Pilicode suggests a nonsignificant increase in both parameters overtime. However a significant increase in summer rainfall and significant decrease in annual maximum temperature was recorded in Pattambi. Moreover the annual rainfall of Pilicode was more as compared to Pattambi with low maximum temperature in Pilicode. The deseasonalized rainfall (Z value is 2.00, P value is 0.04) of Pilicode also showed a significant positive trend. The classification of weather data in to different seasons also helped to identify season wise significant trend in different weather parameters Modeling of weather parameters was done in order to develop best model which is suitable for determining the climate change. Seasonal ARIMA model was selected for modelling the weather data since the weather data consist of seasonality. The best model was estimated with the help of X12 ARIMA in Gretl open source software. The best model was selected on the basis of least AIC value, BIC value and Hannan-Quinn criterion value. The X12 ARIMA automatically detect the best model and then the best model was confirmed by trial and error method that no other models have the least value for the criterion. The best estimated models were respectively SARIMA ((0,1,1)(0,1,1)12) for rainfall, SARIMA ((1,0,1)(0,1,1)12) for Maximum Temperature and SARIMA ((1,0,1)(0,1,1)12) for minimum temperature for Pilicode. Similarly the best models identified for rainfall, maximum temperature and minimum temperature were respectively SARIMA ((0,0,0)(0,1,1)12), SARIMA ((1,0,1)(0,1,1)12) and SARIMA ((0,1,1)(0,1,1)12) for Pattambi. The validation of the model was done with the help of forecasting and comparing for 2018 and mean absolute error for forecasted was calculated. The mean absolute error was low for forecasted maximum temperature and minimum temperature in both Pilicode and Pattambi but the rainfall had high standard error and mean absolute error which was due to the fluctuations in rainfall at both Pilicode and Pattambi respectively. The multiple linear regression was done to analyse the impact of weather parameters on the production of Paddy uder unirrigated area in virippu and mundakan season from 1995 to 2017 at Palakkad district. The results of the analysis reported no significant influence of weather parameters on virippu season but maximum and minimum temperature in October and maximum temperature in November had significant negative influence on mundakan Paddy production. While an increase in maximum temperature during December was favourable for mundakan Paddy production.

The results of the analysis on climate change indicated an increase in rainfall and decrease in maximum temperature over time at Pilicode as compared to Pattambi. SARIMA models were found to be the best model for weather parameters for prediction or forecasting. An increase in maximum temperature during December was favourable while increase in maximum and minimum temperature during October was unfavourable to Paddy production under unirrigated area in mundakan season.