Informasi Detil Paper

Judul: Rainfall and Rainy Days Prediction in Ambon Island Using Vector Autoregression Model
Penulis: Lexy Janzen Sinay & Salmon Notje Aulele  || email:
Jurnal: Prosiding FMIPA 2015 Vol. 1 no. 1 - hal. 90-98 Tahun 2015  [ MIPA ]
Keywords:  Ambon island, time series, rainfall, rainy days, vector autoregression, 𝑉𝐴𝑅(7), lag order selection criteria, 𝑅2, autocorrelations, 𝑀𝑆𝐸, forecasting.
Abstract: Ambon Island is one of the islands in Maluku Province, located in the eastern part of Indonesia. Ambon island is a tropical area with two seasons, namely rainy season and dry season. Geographically, Ambon island has a seasonal rainfall patterns with the high rainfall. The highest rainfall on the Ambon island occurred between May until July. In the last decade, the pattern of rainfall on the Ambon island become irregular. One of the indicator of this phenomenon was the extreme rainfall occured on the Ambon island. Extreme rain is the one of the factors that cause natural disasters in the Ambon island such as floods and landslides. Therefore, this study aimed to estimate rainfall model on the Ambon island using vector autoregression model (𝑉𝐴𝑅) and to forecast rainfall and rainy days based on the best model of 𝑉𝐴𝑅. The data used in this research is time series data from January 2005 to December 2014. The data is monthly rainfall data (𝑟𝑎𝑖𝑛𝑓𝑎𝑙𝑙) and rainy days data for every month (𝑑𝑎𝑦𝑠) on the Ambon island. The best of 𝑉𝐴𝑅 models are based on the following analysis procedures: 𝑉𝐴𝑅 lag order selection criteria, 𝑅2, residual test for autocorrelations, and 𝑀𝑆𝐸. In this study, we’ve gained the (7) as the best model. (7) derived from logarithmic transformation of the actual data. From this model, we’ve predicted the situation in 2015 that the mean value of monthly rainfall and rainy days is 171 mm and 16 days per month, respectively. 𝑀𝑆𝐸 of this model was 0.499 for (𝑟𝑎𝑖𝑛𝑓𝑎𝑙𝑙) and 0.141 for (𝑑𝑎𝑦𝑠). It is accurate because this model has the smallest 𝑀𝑆𝐸 values.
File PDF: Download fulltext PDF PDF

<<< Previous Record Next Record >>>