9:15 AM - 9:30 AM
[AAS03-02] Investigation of Key Role in Hotspot Occurrence using Multivariate Principal Component and Distribution Analysis (Study case: Kalimantan and South Sumatra, Indonesia)
Keywords:Forest Fire, Principal Component Analysis, Distribution Analysis, Extreme event, Drought
For the last decades, forest fire have been increases around the world, especially as result of increasing risk of severe and longer drought condition caused by climate change. Due to complexity of the phenomena correspond to local climate condition, forest fire have high variability which makes it hard to determined the key role of the event. Moreover, high influence of human in terms of land uses further increases the variability of hotspot and climate variables relation. While many research has been done to address this issue, there is still high variance of result in determining key factors of the hotspot occurrence which is very important to prevent further risk of fires event around the world.
Objective of this reseatch is to investigate key roles of land and forest fires in analysed region (Study case: Kalimantan and South Sumatra, Indonesia / 6°S - 4°N, 100°E - 119°E) . Investigation is provided using principal component analysis, multivariate distribution analysis. Region will be separated based on spatial information of the region. The analysis is done through local climates (7-8 days hotspot, 7-8 days total precipitation, number of days since last rain, 7-8 days normalized difference vegetation index, 7-8 days 2m air temperature, 7-8 days wind speed, 7-8 days soil moisture, 30 days standardized precipitation evaporation index) as well as global climate indicators (3 months El-Niño Southern Oscillation (ENSO) index and 3 month Indian Ocean Dipole (IOD) index). Analysis was done through weekly time period of data from January 2001-December 2020 with spatial grid data of 0.25°×0.25°. Multiple different time period accumulation of climate variable were done to reduce auto correlation that happened between each variable. Moreover,all of the variables were normalized to avoid domination among each variables during multivariate principal component analysis. In order to negate the downside of normalization, distribution analysis were conducted to investigate changes of distribution correspond to number of hotspot occurrence.
Result of principal component analysis shows that there multiple (3-5) pattern for each analysed region that related to high hotspot occurrence with high variability of contribution between each pattern. Consistency of 1st pattern (22-25% variance explain) for all region, shows high interaction for most of the climate variables with 7-8 days accumulation, except vegetation index. However, hotspot contribution in the 1st pattern never been the highest one which resemble the behavior climate/drought condition when short period of fires event happened in the peak dry season. Another consistency was occurred related to long time period climate variables (30 days SPEI, 3 months ENSO and IOD) with varying contribution for each region related to how strong influence ENSO and IOD in prolong respective dry season. The third consistency across all region were there are some different behavior between vegetation index and other climate variables. For example in West Kalimantan, 1st pattern have lower contribution (relative to other variables) compared to its 2nd pattern (relative to other variables). This result were obtained due to different characteristic of land type that contribute to high/low risk of hotspot occurrence as well as different causes of fires in respective region.
In conclusion, forest fires were complex phenomena with varying characteristic spatially and temporally. This research provide novel analysis of interaction between crucial climate variables that could influence condition of high risk hotspot occurrence(Study case: Kalimantan and South Sumatra, Indonesia . Result from this research is important to be help understanding high variability of events in effort of developing estimation, prediction, and projection of forest fire risk in the future.
Objective of this reseatch is to investigate key roles of land and forest fires in analysed region (Study case: Kalimantan and South Sumatra, Indonesia / 6°S - 4°N, 100°E - 119°E) . Investigation is provided using principal component analysis, multivariate distribution analysis. Region will be separated based on spatial information of the region. The analysis is done through local climates (7-8 days hotspot, 7-8 days total precipitation, number of days since last rain, 7-8 days normalized difference vegetation index, 7-8 days 2m air temperature, 7-8 days wind speed, 7-8 days soil moisture, 30 days standardized precipitation evaporation index) as well as global climate indicators (3 months El-Niño Southern Oscillation (ENSO) index and 3 month Indian Ocean Dipole (IOD) index). Analysis was done through weekly time period of data from January 2001-December 2020 with spatial grid data of 0.25°×0.25°. Multiple different time period accumulation of climate variable were done to reduce auto correlation that happened between each variable. Moreover,all of the variables were normalized to avoid domination among each variables during multivariate principal component analysis. In order to negate the downside of normalization, distribution analysis were conducted to investigate changes of distribution correspond to number of hotspot occurrence.
Result of principal component analysis shows that there multiple (3-5) pattern for each analysed region that related to high hotspot occurrence with high variability of contribution between each pattern. Consistency of 1st pattern (22-25% variance explain) for all region, shows high interaction for most of the climate variables with 7-8 days accumulation, except vegetation index. However, hotspot contribution in the 1st pattern never been the highest one which resemble the behavior climate/drought condition when short period of fires event happened in the peak dry season. Another consistency was occurred related to long time period climate variables (30 days SPEI, 3 months ENSO and IOD) with varying contribution for each region related to how strong influence ENSO and IOD in prolong respective dry season. The third consistency across all region were there are some different behavior between vegetation index and other climate variables. For example in West Kalimantan, 1st pattern have lower contribution (relative to other variables) compared to its 2nd pattern (relative to other variables). This result were obtained due to different characteristic of land type that contribute to high/low risk of hotspot occurrence as well as different causes of fires in respective region.
In conclusion, forest fires were complex phenomena with varying characteristic spatially and temporally. This research provide novel analysis of interaction between crucial climate variables that could influence condition of high risk hotspot occurrence(Study case: Kalimantan and South Sumatra, Indonesia . Result from this research is important to be help understanding high variability of events in effort of developing estimation, prediction, and projection of forest fire risk in the future.