Assessing vulnerability of peat fire in the Indonesian Peat Hydrological Unit (PHU) areas

Prevention of fires in peatlands must be based on a holistic landscape approach in the Peat Hydrological Unit (PHU), which involves mapping the areas vulnerable to peat fires. This study employed regression modeling, incorporating land cover, soil (peatland), concession area, and tree cover loss, which were linked to hotspots verified by burnt areas derived from datasets obtained from 2015 to 2019. The results of this study in Indonesian PHUs showed that 3.8 million hectares were vulnerable in the high-class area, 12.6 million hectares in the medium-class area, and 7.7 million hectares in the low-class area. The PHU of the Kahayan River-Sebangau River in Central Kalimantan Province was the largest area with high vulnerability to fires (covering an area of 190 thousand hectares). This model was verified using the fire hotspot approach; out of 38 fire spots that occurred from January to June 2023, 33 locations (86.84%) were detected according to the predicted vulnerability of the peat fires category (high and medium class).


Introduction
Tropical peatlands are unique wetland ecosystem that characterized by organic-rich soil formed through an anaerobic decomposition process.Indonesia is renowned for its incredible diversity of flora and fauna.It houses 10% of the world's tropical rainforests [1] as well as a staggering 36% of tropical peatlands [2].This underscores the vital importance of protecting these ecosystems in Indonesia, to ensure their continued sustainability.
Natural and human factors are causing changes in Indonesia's peatlands, resulting in a reduction of vegetation cover, alterations to water balance, and biodiversity shifts within the ecosystems.The majority of these anthropogenic influences stem from an increase in large-scale monoculture plantations that have caused Indonesia to lose approximately 498,000 hectares of forest yearly between 2000-2015 [3].This has led to Indonesia holding second place for highest deforestation rates worldwide following Brazil as ranked in 2015 [4].
Indonesia's government is trying to save the existing peatlands by stipulating new regulations for using peat areas based on the Peat Hydrological Unit (PHU) as a basis for development.PHU is a peat ecosystem bordered by rivers and/or the sea, which has specific properties and may contain peat domes.Peat domes are parts of the peat ecosystem with a convex shape and are higher than their surroundings, controlling water balance and movement, and should be protected areas [5].
However, peat destruction is still ongoing, as some areas of peat with more than 3 meters of depth have been utilized, and licenses have been granted before the PHU's regulations were in place.Largescale monoculture plantations are the primary cause of destruction, as they often drain peatlands 1315 (2024) 012057 IOP Publishing doi:10.1088/1755-1315/1315/1/012057 2 extensively.Land clearing through burning is also common to reduce opening costs but contributes to emitting greenhouse gases into the atmosphere.The extensive forest and land fires that occurred in 2015 and 2019 were closely linked with high levels of greenhouse gas emissions in Indonesia, with at least half of burned areas comprising peatlands.
Out of a total area burnt from these years (4.4 million hectares), approximately 18% or 789,600 hectares experienced repeated burning [6].Fires also were largely fueled by concession areas, with a notable contributing factor.From 2015 through to 2019, palm oil and pulpwood concessions accounted for approximately thirty percent (30%) or up to 1.3 million hectares of the delineated fire-affected zones [6].This research aims to raise awareness among key stakeholders by sharing information about the vulnerable areas of peat fire based on PHU areas using a vulnerability fire map, with particular emphasis on the anticipated impact of the upcoming El Nino climate anomaly in 2023.

Study Site
The vulnerability model of peat fires was carried out across the entire Indonesian PHU territory [7], totaling 865 PHUs.The implementation of restoration on peatlands will be most effective when done in landscape areas [8].The principle of restoration on a landscape basis in the PHU area, in addition to regaining ecological functions, also aims to improve community welfare through collaborative and participative restoration around the damaged peatlands [9].Therefore, preserving peatlands as PHU ecosystems through the vulnerability of peat fire maps becomes exceedingly crucial.

Materials
The vulnerability of peat fires in Indonesian PHUs is assessed using a regression model created using data from 2015 to 2019.The independent variables include land cover maps, peatland distribution, concession area boundaries, and the loss of tree cover maps, while the dependent variables consist of fire hotspots and fire area maps.The period of datasets was based on the significant historical fires in Indonesia, especially the El Nino anomaly in 2015 and 2019 [10].The independent variable datasets were then used as predictive data for modeling the vulnerability map in 2020-2021.However, only the data on peatland distribution (2019), concession area boundaries (2019), and PHU (2017) are on the same year basis.
The materials used in this study include the PHU Area, burned area, and land cover maps from the Ministry of Environment and Forestry (MoEF), peatland distribution map from the Ministry of Agriculture (MoA), boundaries map of oil palm and pulp paper concession areas from Pantau Gambut, and tree cover loss from Global Forest Watch (GFW) [11].This study used Indonesian hotspot data, specifically the VIIRS-SNPP hotspot sensor with a high confidence level for model development, and VIIRS-SNPP, VIIRS-NOAA, and MODIS data from the NASA FIRMS catalog for hotspot analysis in 2023.

Data Processing Procedure
The mechanism applied in the preparation of this map uses bio-physical factors that affect the level of fire vulnerability.Furthermore, this study uses concession data and a 1 km buffer zone to address issues related to water governance, expansion of plantation areas, and the impact of companies on the surrounding areas in terms of social and physical environmental aspects.The vulnerability fire classification in PHU is carried out through scoring and weighting analysis (Figure 1).Scoring is done quantitatively using the Composite Mapping Analysis (CMA) method based on the relationship of each factor to the percentage of hotspots that have been verified with a burnt area map.The hotspots used in the CMA method are the results of the VIIRS-SNPP sensor recordings with a high confidence level.The classification of each factor along with the formula for calculating scores and weights is presented in Table 1.The scoring formula with the CMA method is [12]:

Data Analysis
Following the modelling of the vulnerability classification map, the analysis determines which PHU rankings are fire-prone by evaluating the PHUs with the greatest high-class vulnerability.Apart from assigning a fire risk rating to PHUs, this research also keeps an eye on hotspots and assesses the model's accuracy in 2023 using the fire hotspot method.Hotspot data from two satellite image sensors are used for the hotspot analysis: the Visible Infrared Imaging Radiometer Suite (VIIRS) sensor on S-NPP and NOAA 20 (formerly known as JPSS-1) and the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor on Aqua and Terra satellites.In order to maximise the study of site attributes suggesting fire occurrences, this hotspot analysis makes use of multiple hotspot sensors [14].The examination of 2023's fire hotspots is restricted to the months of January through June.

Vulnerability Fire Map Model
The history of fires from 2015 to 2019 is a dataset used to provide scoring for each sub-factor and build a prediction model for fire vulnerability in Indonesian PHUs.where tree cover loss occurred (score 99.21) are identified as the most frequently associated sub-factors with burned areas.This highlights the crucial importance of monitoring and safeguarding these subvariables in each factor to secure the PHUs from recurrent peatland fires.
The scores of the sub-factors show that deforestation occurring on peatlands increases the risk of fires on peatlands.Bare land and areas with tree cover loss are the sub-factors most frequently associated with burnt areas.Deforestation of peatlands disrupts the balance of the ecosystem, particularly due to massive drainage [15].This study also found that estate crop plantations and fires inside the concession area + 1km buffer from the outer boundary are sub-factors related to burned areas.Extensive monoculture planting operations lead to excessive drainage, which increases the risk of fire in peatlands.Their qualities are changed by draining, making it impossible for them to revert to their previous wetland state and increasing their fire vulnerability [16].The equation models the historical fire incidents based on the dataset's regression analysis from 2015 to 2019: Tree cover loss In the resulting regression model, only the significance value of X1, the land cover factor, is smaller than 0.05 (i.e., the significance level value used in the multiple linear regression that was performed).The significance value of the land cover factor is 0.003, indicating that the land cover factor has a significant influence on fire vulnerability compared to other factors.From this study, the negative value on the TCL factor (tree cover loss) is not by the research hypothesis.It is suspected that the dominance of peatlands is characterized by swamp shrub land cover.This is also evident through the highest sub-factor scores on land cover, namely bare land and swamp shrub, which do not consist of dominant tree cover.
In addition, due to massive deforestation in Indonesia, especially the conversion of forests into monoculture plantations, some locations are no longer considered as forests.There are two possible reasons for the observed relationship between fire and forest loss [17].Either the fire causes forest loss, or forest loss makes the landscape more susceptible to fire.Since the fire frequency is lower after forest loss compared to the rate before forest loss occurs, it suggests that forest loss may make the landscape more susceptible to fire.Another possibility is that economic factors contribute to this situation, as the practice of burning land to clear it may be more economical than using heavy equipment for land clearing [18].
The final model serves as the computation formula for the 2020-2021 data, which is utilised to provide location categorization and forecast data for the PHU area's sensitivity to peatland fire levels.To categorise the vulnerability levels, the average and standard deviation are computed once the prediction data has been entered into the model.The following is how the classification is established: High vulnerability fire class : Values above 44 Medium vulnerability fire class : Values range from -2 to 44 Low vulnerability fire class : Values below -2

Vulnerability Fire Classses in PHU
The PHU in Indonesia serves as the foundation for the vulnerability study to forest and land fires.This method emphasises how crucial it is to protect and observe peatlands in relation to other landscapes or ecosystems.Figure 2   In comparison to PHU areas without peatlands, Table 2 shows that high vulnerability predominates in peatland-rich PHU areas.This is demonstrated by the fact that 2.5 million hectares of the 3.8 million hectares of peatland are considered to be highly vulnerable.Put differently, 65.9% of the peatland in the PHU areas needs to receive more attention when it comes to ecosystem-or landscape-based restoration projects.With a total size of 1.2 million hectares, Central Kalimantan Province emerges as the province with the greatest high-vulnerability PHU fire in Indonesia based on the integration of data from PHU areas and provincial administrative regions.This is because many PHUs in Central Kalimantan province are controlled by concessions, comprising around 1.5 million hectares of the 4.3 million hectares of PHUs that are present only in that province.The high class of vulnerability to fire in PHUs is also influenced by the existence of concessions on peat or PHU areas in Indonesia.Based on the analysis, approximately 54% of the 3.8 Mha of Indonesian PHUs with high-class vulnerability to fire are located within concession areas and their surroundings.

The Largest PHU with High-Class Vulnerability Fire
This study found that the PHU of Kahayan River-Sebangau River is classified as the largest high-class vulnerability to fire, covering an area of approximately 190 thousand hectares.Sebangau National Park, a conservation area with unique qualities that desperately need protection, lies next to this PHU.The Sebangau Area had Forest Concession Rights (Hak Pengusahan Hutan/HPH) and was in use from the early 1970s to the mid-1990s until it was designated as a national park [19].After being named a national park in 2004, it was anticipated that this region would be recovered and turned back into a viable natural peat forest.This location is among the top 10 PHUs with the highest susceptibility since it is prone to fires and has a history of fires.However, in 1996/1997, this PHU was part of the Mega Rice Project (MRP) program.That program destroyed around one million hectares of peatland forest in Central Kalimantan with drainage canals for conversion to agricultural land [20].These failed ex-MRP locations need to have been repaired rather being abandoned or put to new purposes.In the nine years following the Mega Rice Project (1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005), 72% of the peat swamp forest experienced fire-related losses.The majority of these areas were replaced with non-woody vegetation, primarily ferns or tree mosaics with non-woody flora, as opposed to cultivated land [21].In 2020, however, the locations were re-designated for the food estate programme under President Joko Widodo's administration.The agricultural programme implemented on these areas has not produced any harvest to yet and has only resulted in deforestation.

3.4.Fire Hotspot Analysis in 2023
A quick study called a "hotspot analysis" is used to pinpoint areas that are thought to be vulnerable to land and forest fires.Satellite imaging measures the pixel images that show the presence of hotspots, which are places that have higher temperatures than their surrounds [22].Hotspots are regions that may be at risk of fire, although the existence of one hotspot does not guarantee that a fire will break out.Hotspots with particular traits, however, may be taken into account as fire warning signs.If a hotspot expands in groups or clusters in one area and/or lasts for three days or longer, it is classified as a forest and land fire.The likelihood of forest and land fires increases with the amount and frequency of hotspots in a given area [23].This study found a total of 6,842 hotspots in the Indonesian PHU during January-June 2023.When separated into vulnerability fire classes in Table 4, around 25% of hotspots were found in the high class, 38% in the medium class, and 37% in the low class.This means that 63% of hotspots were found in predicted locations that are prone to burning (the high and medium classes) in 2023.
From January to June of 2023, the PHU of Rokan River-Siak Kecil (Riau Province) has the greatest total number of hotspots among all Indonesian PHUs. Figure 4(a) illustrates the 850 hotspots that our investigation found in that PHU.Though it is Indonesia's largest PHU, large-scale fires occur there frequently.Based on the clustered hotspots approach in Figure 4(b), this study identified 38 locations that experienced fires in the PHU from January to June 2023.These locations consist of 23 locations in the high class, 10 locations in the medium class, and 5 locations in the low class.This means that 33 out of 38 locations are prone to burning (the high and medium classes).In percentage terms of the location by hotspot approach, the accuracy of the vulnerability fire map in 2023 is 86.84%.This also verifies that the results of the studies made are close to accurate, but these results need to be verified by the end of 2023.
Every month from January to June 2023 shows an increase in the number of hotspots in PHU areas.The biggest increase in the number of hotspots was in April.This increase is strongly suspected to be related to the occurrence of the El Nino anomaly in 2023.The El Nino anomaly is highly associated with fires in Indonesia; major fires in 2015 and 2019 occurred during the El Nino period and were dominant in peatlands [24].In March, the condition of the Pacific Ocean strengthened to ENSO Neutral, and in May, the Sea Surface Temperature (SST) anomaly at Nino-3.4 was +0.8°C.Therefore, it is suspected that the number of hotspots and locations of burnt areas will continue to increase, especially during the dry season in Indonesia.

Conclusion
The history of fires from 2015-2019 showed that land cover type of bare land (score 21.99), peat areas (score 79.44), areas inside the concession area + 1km buffer from the outer boundary (score 55.02), and areas where tree cover loss occurred (score 99.21) were the most frequent sub-factors associated with burned areas.Approximately 16.4 Mha of Indonesia's 24.2 Mha PHU area are classified as high or mediumly vulnerable.With a total size of over 190 thousand hectares, the PHU of the Kahayan River-Sebangau River is categorised as having the highest high-class vulnerability to fire.
In 2023, this study found a total of 6,842 hotspots in the Indonesian PHU area from January to June.Using the hotspot approach, there were 23 locations in the high class, 10 locations in the medium class, and 5 locations in the low class.In percentage terms of the location by hotspot approach, the accuracy of the vulnerability fire map in 2023 is 86.84%.The El Nino in Indonesia during 2023 will have an impact on increasing the number of monthly hotspots.

Figure 1 .
Figure 1.Framework of Vulnerability Fire Modeling Method displays the visualisation of the land fire and forest fire vulnerability analysis results in the PHU zones for the year 2023.

Figure 2 .
Figure 2. Vulnerability of Peat Fire Map in the Indonesian PHU Areas

Figure 3 .
Figure 3. Fire Vulnerability Map in The PHU of the Kahayan River -Sebangau River

Figure 4 .
Hotspot Analysis (a) Highest Total of Hotspots in PHU from January to June 2023; (b) Total Hotspots are Indicated as Areas Potentially to Fire

Table 2 .
The degree of vulnerability to forest and land fires in PHU areas across Indonesia in 2023

Table 3 .
PHUs with the Largest High-Class Vulnerability Fire

Table 4 .
Hotspots Total from January to June 2023