Sequential Pattern Analysis of Hotspots based on Peatland Characteristic in Sumatra and Kalimantan

Abstract

Peat fires become a yearly tradition, so that necessary precautions are needed in order to minimize the destruction of peatlands. Peatland fires can be detected through hotspots occurrence in the region, but not all hotspots indicate fires. However, hotspots consecutively found in a certain area in the period of two to five days are strong indicators for forest and land fires. This study aims to obtain characteristics of hotspot sequential patterns in peatland using data mining approach. The algorithm used is Sequential pattern Discovery using Equivalent Class (SPADE) that applies a vertical format in sequential pattern mining to extract frequent sequences on datasets. This study involves spatial data including hotspot data from FIRMS NASA, peatland characteristics from Wetland International to enrich our analysis, and the district border layer to determine the location of hotspot sequences. Data transformation was performed to prepare datasets before applying SPADE. There are 316 sequences found in Kalimantan and Sumatra in 2014 and 2015 with a minimal of two days occurrences and minimum support of 1%. This study shows that hotspot sequences in Kalimantan 2014 and 2015 are frequently occurred in peatland that has the type of Hemists/Fibrists (60/40) and depth of 400-800 cm. In addition, hotspot sequences in Sumatera 2014 and 2015 are typically occurred in peatland that has the type of Hemists/Saprists (60/40), landuse of swamp forest and depth of 100-200 cm.