An empirical model for the seasonal prediction of southwest monsoon rainfall over Kerala, a meteorological subdivision of India

Abstract

ABSTRACT:There are several studies showing a skillful empirical prediction of the All India Summer Monsoon Rainfall (AISMR) based on various combination of parameters as the predictors. However, the southwest monsoon rainfall over Kerala, a meteorological subdivision of India, bears a considerably low correlation coefficient with the AISMR. This implies that the existing predictors in the long-range forecast models of the AISMR do not have much influence on the Kerala Summer Monsoon Rainfall (KSMR). This study attempts to examine the relationship of some ocean and atmospheric parameters with the rainfall and to formulate a linear multiple regression model for the long-range forecast over a small area like the Kerala. Parameters having significant correlation (significant at 1% level) with the KSMR were identified for the period 1961-1994. The consistency of the relationship between these parameters and the KSMR was checked by doing a 21-year sliding window correlation (significant at 5% level). Using a stepwise regression method, seven predictors, explaining a significant amount of variance in the KSMR were selected and a linear multiple regression model was developed. The parameters that explain the high inter-annual variability of the KSMR are specific humidity, sensible heat net flux, relative humidity, zonal wind at 70 and 10 hPa, meridional wind and geopotential height. The characteristics of the forecast and its reliability were studied by various statistical techniques such as, Durbin Watson statistics and variance inflation factor. The model has a multiple correlation of 0.943 and coefficient of determination of 88.8%. The root mean square error (RMSE) was 6.60% (15.80%), bias (BIAS) was −0.26% (6.20%), absolute error (ABSE) was 5.33% (13.15%) of mean rainfall for the training (test) period respectively. Climatological predictions were also made and the RMSE was (17.90%), BIAS (−5.40%) and ABSE (15.16%) of mean rainfall. The selected parameters were at least 2 months prior to the monsoon season and hence have predictive value.

Annual precipitation forecast for west, southwest, and south provinces of Iran using artificial neural networks

“…It is obvious that accurate forecasting of precipitation amount is a hard task, since precipitation is one of the most complex elements of the hydrological cycle to forecast (French et al 1992; Gwangseob and Ana 2001;Rogers and Yau 1989). Different studies have been conducted for forecasting the quantity of precipitation by applying different methods such as numerical weather prediction models and remote sensing observations (Sheng et al 2006;Diomede et al 2008), statistical models (Munot and Kumar 2007;Li and Zeng 2008;Nayagam et al 2008), support vectors regression, and fuzzy inference system (Cheng et al 2002;Lin et al 2006Lin et al , 2009Li et al 2006;Muttil and Chau 2007;Chattopadhyay and Chattopadhyay 2007;Guhathakurta 2008). However, they have their own difficulties, complexity, and large data requirements.…”

Enhanced Monthly Precipitation Forecasting Using Artificial Neural Network and Singular Spectrum Analysis Conjunction Models

“…Nevertheless, for water resources planning purposes, a long-term rainfall series is needed in many hydrological and simulation models (Tantanee et al 2005;Venkata Ramana et al 2013). There have been many attempts to address the hydrological processes in general and the quantitative precipitation forecasting in particular through different techniques including numerical weather prediction models and remote sensing observations (Yates et al 2000;Ganguly and Bras 2003;Diomede et al 2008;He et al 2013), statistical models (Chu and He 1994;Chan and Shi 1999;DelSole and Shukla 2002;Munot and Kumar 2007;Li and Zeng 2008;Nayagam et al 2008), and soft computing methods including artificial neural network (ANN), self-organizing map (SOM), support vector machine (SVM), fuzzy inference system and extreme learning machine (ELM) (French et al 1992;Navone and Ceccatto 1994;Pongracz et al 2001;Freiwan and Cigizoglu 2005;Marzano et al 2006;Kalteh and Berndtsson 2007;El-Shafie et al 2011;Chen et al 2015;Gholami et al 2015;Taormina and Chau 2015). For example, an ANN model for forecasting rainfall intensity fields at a lead-time of 1 h was applied by French et al (1992).…”

Oceanic and atmospheric linkages with short rainfall season intraseasonal statistics over Equatorial Eastern Africa and their predictive potential

“…A good MLR model requires that the residuals (the difference between the actual observations and the forecasted values) are independent and have a normal distribution (Nayagam et al , ). The Durbin–Watson (DW) statistic which checks the significance of the assumption that the residuals for successive observations are uncorrelated/independent was used to determine whether the residuals were independent (Makridakis et al , ).…”