Advances and Applications in Statistics
Volume 49, Issue 5, Pages 387 - 398
(November 2016) http://dx.doi.org/10.17654/AS049050387 |
|
CLIMATE CHANGES DETECTION USING BAYESIAN CHANGE POINT METHODS AND EMPIRICAL ORTHOGONAL FUNCTION FOR IMPROVING THE TRADITIONAL CROPPING CALENDAR IN GUNUNG KIDUL REGENCY OF INDONESIA
Fajar Adi-Kusumo, Gunardi, Herni Utami, Aluicius Irwan Endrayanto, Mulyono Nitisapto, Tony Agus Wijaya, Ardhasena Sopaheluwakan and Titus Christiawan
|
Abstract: “Pranoto Mongso (PM)” is one of the traditional cropping calendars in Indonesia used to guide farming activities since last two hundred years. The calendar was created in combination with the traditional knowledge of the weather prediction and the natural sign of plants and animals. Due to the climate change issues, there are some weather anomalies which have an implication to the accuracy of the PM. In this paper, we consider the climate change detection based on the daily rainfall data at Gunung Kidul regency in Daerah Istimewa Yogyakarta (DIY) Province Indonesia. The Bayesian Change Point (BCP) and the Empirical Orthogonal Function (EOF) are the methods which are useful to detect the change points and the dominant pattern of the rainfall. Our analysis is focused to find the transition between the rain and the dry periods in such area for several years. The rainfall events are assumed to follow the Poisson process and we calculate the rainfall probability when the intensity is more than 8 mm per day. By using the combination between BCP and Gibbs sampling algorithm, the seasonal classifications of the weather which can be compared with the original PM are determined. The dominant patterns of the rainfall in such areas are analyzed by EOF. The EOF method which is based on the largest eigenvalue of the rainfall-data covariant matrix at some stations is important for reducing the dimension of the system without losing its significant information in order to figures the whole data. In this case, we will show the zones in the regency which have the similar rainfall patterns and conclude our study by some recommendations of the cropping strategies that sample area. |
Keywords and phrases: Poisson process, Bayesian change point, MCMC methods, EOF, rainfall characteristics, dominant pattern. |
|
Number of Downloads: 391 | Number of Views: 1405 |
|