Mapping Extreme Rain Conditions in Sumatra by Influence Global Conditions

Determination of the extreme value category with a specific threshold value for a region with diverse climate characteritics may lead to incorrect extreme characterization. This study aims to determine thresholds of extreme rainfall in Sumatra, identify its spatial patterns and occurence in Sumatra and examine the link between extreme weather events to global climate phenomena which are the ENSO and the Dipole Mode. The data used in this study are daily rainfall data from 38 stations in Sumatra and indices of global climate phenomena, the dipole mode index (DMI) and the Southern Oscillation Index (SOI). The method used to determine the threshold for extreme value is with a probability of 95% based on their corresponding distribution, cluster analysis and correlation analysis. The expected result is an extreme value for all parts of Sumatra, the extreme character of all clusters by time and region, as well as its relationship with DMI and SOI. Preliminary results obtained that extreme daily rainfall values ranged from 42.4 to 114.7 mm per day. Extreme daily rainfall for western region of Sumatra has an average value higher and more varied than the eastern region. Extreme daily rainfall over northern region has an average value higher and more varied than the southern region. Clustering analysis of 38 abservation stations with ward method gets 5 (five) clusters. At least 20 years long periods of ideal data are needed to find the threshold of extreme rainfall over Sumatra region. Threshold of extreme rainfall with long periods of different data showing different values for each station are sometimes larger or smaller that may be caused by different spread of the data for each station.


Introduction
Sumatra Island is geographically an island which is above equatorial region surrounded by ocean. Rainfall in Indonesia in general is influenced by the interaction of land, atmosphere and the oceans around it. The phenomenon of positive anomaly of sea surface temperatures in the Indian Ocean to the west and negative anomalies in sea surface temperatures occurred west of Sumatra, resulting in increased rainfall in Sumatra's west coast region known as the Indian Ocean dipole mode (IOD) [1]. There is a negative IOD in most areas of Sumatra causing increased rainfall. In Indonesia the events of floods and severe droughts are always associated with El Niño-Southern Oscillation (ENSO) [2]. The results of research related to extreme rainfall with cumulative distribution function method for Sumatra conducted by Marpaung et al [3] using rainfall data from TRMM and observations indicate that extreme rainfall events with the number of incidence days of 1 to 2 days per annum dominant place. In 1998 and 1999 it appears to increase in extreme rainfall events for a number of days per year Year 2010 is rainfall intensity of 50 mm/day or 20 mm/hour. The determination of threshold values with the extreme categories that limit values for a region as diverse character of the climate was not able to provide characterization of the extreme right. Therefore, this study aims to determine the threshold of extreme rainfall in Sumatra and group them by similarity.

Research Locations
Location of the study is the island of Sumatra in the form of 18 points stations in Sumatra and surrounding areas.

Data
The data used in this study are the daily rainfall data of observation stations in the island of Sumatra in the period of 1983-2012.

Cluster analysis
Cluster analysis is part of a multivariate analysis with the purpose to categorize objects based on the characteristics of its observations, where each object in the same group have a level of closeness and similarity between one another [8]. The result of cluster analysis is the formation of groups that have a degree of similarity and heterogeneity of high inter-cluster. Nazaripour

Analysis of Opportunities
In determining the threshold of extreme rainfall that contains zero millimeters of rainfall, the chances are calculated by the method of distribution of mixture [10], namely: where : H (x) : probability distribution for a mixture of rain recesses value that exceeds x for x ≥ 0 q : zero chance of rain events p : probability of rain events greater than zero, p = 1 -q Px (X) : probability of rain events by gamma cumulative distribution opportunities (x > 0)

Data Length Determination
The steps in determining the length of the data that is used to find the value extreme are: 1. Find a station that has a data length more than 30 years 2. Divide data into several periods where each period is a multiple of five years 3. Determine the extreme value of each station 4. Create scatter plot graph 5. Determine the value that is closest to the ticket    Figure 4 shows that the extreme daily rainfall values range from 42.4 to 114.7 mm. Extreme value of daily rainfall in western region of Sumatra has an average value higher and more varied than the eastern region. While the northern region of Sumatra has an extreme value of daily rainfall is higher than in the south.

Analysis of Data Period
Minimum length of data period is very important as a reference in determining the length of data that is ideal for calculating the thresholds of extreme rainfall. Determination of a minimum length of the data period was done by taking a sample of rainfall observation data over 30 years. After sorting the data from 38 stations in Sumatra, seven stations that have long data period for 35 years are Raden Inten II Lampung, Fatmawati Bengkulu, Kenten South Sumatra, Sutan Mahmud Badarudin II South Sumatra, Depati Amir Bangka Belitung, Sampali North Sumatra, Kualanamu North Sumatra.

Summary and Concluding Remarks
Clustering analysis of 38 abservation stations with the ward method results 5 (five) clusters. Each cluster has a different pattern every month. Extreme daily rainfall in the region of Sumatra has a value ranging from 42.4 to 114.7 mm. Extreme daily rainfall over western region of Sumatra has an average value higher and more varied than the eastern region. Extreme daily rainfall over the northern region has an average value higher and more varied than the southern region. Threshold of extreme rainfall with long periods of different data showing different values for each station are sometimes larger or smaller that may be caused by different spread of the data for each station.