Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Titre : New Statistical Approaches in Modeling and Forecasting Indian Rainfall Time Series

Auteur : Kokila Ramesh

Université de soutenance : Jain University

Grade : Doctor of Philosophy (PhD) in Mathematics 2017

Résumé partiel
The annual rainfall of India has three seasons per year accounting for about 11% each in the pre-monsoon (January-May) and the northeast monsoon (October-December) and 78% in the southwest monsoon season also known as summer monsoon (JuneSeptember). The maximum amount of the rainfall occurs during southwest monsoon (SWM), which governs the agricultural economy of India and hence for administrative purposes. While the season recurs annually, the variation about the long term expected value can be as high as 40-50% in some parts of the country. Variability during SWM season is an uncertain quantity which India faces every year. This uncertainty can be year to year, season to season (within year), month to month (within season and within year) and so on depending on the requirement in the practical purposes. The huge variation in the rainfall causes droughts and floods. The distress caused by droughts and floods due to extreme variations of the monsoon can be mitigated to some extent if the rainfall time series can be modeled efficiently for simulation and forecasting of SWM data. Hence this becomes the primary reason to develop new models for Indian monsoon rainfall. Rainfall data is a strongly non-Gaussian time series exhibiting non-stationarity. The main objective of the present thesis is to develop new statistical approaches to model and forecast Indian monsoon rainfall data. By considering only the rainfall and its past relation, models have been constructed. These models are without the usage of any atmospheric and oceanic related data parameters. The major areas touched upon in this thesis include (a) Empirical mode decomposition analysis of Indian SWM rainfall and seasonal data (three seasons per year data), (b) Modeling and Forecasting of Indian monsoon rainfall using a new ANN model, (c) Inclusion of within year variation in the ANN to model and forecast Indian SWM data with higher efficiency, (d) developing a new non-Gaussian probability distribution model for the SWM data distribution of core monsoon region and its subdivisions and (e) new non-Gaussian time series model for simulation of Indian monsoon rainfall. The present thesis is structured into eight chapters.

Présentation et version intégrale (Shodhganga)