Nationwide monthly burned area monitoring in Indonesia using Sentinel-2

David L A Gaveau; Adrià Descals; Mohammad A Salim

doi:10.1371/journal.pone.0331831

. 2026 Apr 8;21(4):e0331831. doi: 10.1371/journal.pone.0331831

Nationwide monthly burned area monitoring in Indonesia using Sentinel-2

David L A Gaveau ^1,^2,^*, Adrià Descals ³, Mohammad A Salim ¹

Editor: Bijeesh Kozhikkodan Veettil⁴

PMCID: PMC13061323 PMID: 41950337

Abstract

Wildfires pose a major challenge for many nations. Rapid mapping of their extent is key to evaluating their impacts. We present the first operational monthly burned-area processing chain for Indonesia, based on largely automated processing of Sentinel-2 imagery in Google Earth Engine. Our approach uses a Random Forest applied to Sentinel-2 imagery and integrates FIRMS fire hotspots to reduce false positives. The resulting 20-m monthly burned-area maps cover the entire country. From January 2019 to December 2024, fires burned a cumulative 5.62 million hectares (Mha), including 2.92 Mha that burned once and 1.12 Mha that burned multiple times. This represents a total burned extent of 4.04 Mha. Compared to the MCD64A1 product, our dataset detects more burns with higher spatial detail and accuracy. In total, 122,164 hectares of primary humid forest burned, representing 2.2% of the burned area. In 2019 and 2023, fire activity accelerated around July and peaked in September–October, coinciding with Oceanic Niño Index (ONI) values ≥ +0.5 °C and Indian Ocean Dipole (IOD) values ≥ +1.5°C. In contrast, neutral or negative phases from 2020 to 2022 corresponded with minimal burning. The year 2024 recorded intermediate fire activity without strong climatic anomalies. These findings confirm that climatic anomalies are associated with fire activity in Indonesia, reaffirming the importance of ONI and IOD for early warning. Our results suggest that prevention efforts are limiting forest fires, as burns in 2019 and 2023 remained lower than during earlier events. Monthly burn-scar updates are available on Nusantara Atlas (www.nusantara-atlas.org), an open-access platform for monitoring deforestation in Southeast Asia.

Introduction

Wildfires damage terrestrial ecosystems, threaten human lives, and disrupt economies [1]. They often spread beyond control, especially in fire-prone regions with degraded ecosystems [2,3]. This contrasts with traditional controlled burns, which have long been used by farmers to manage farmland. Rapid and accurate mapping of burned areas is necessary for quantifying the extent of the damage, understanding the causes, and guiding response strategies.

Indonesia is a major wildfire hotspot in the equatorial region [4]. Recurring forest fires have become an international concern because the emitted smoke regularly blanket Southeast Asia in toxic haze for weeks. Their frequency and intensity have sharply increased since the late 20th century, particularly in Kalimantan after 1980, due to large-scale land-use change [5,6]. While El Niño and the Indian Ocean Dipole create dry conditions that favour fire spread [7,8], humans activities intensified risk and triggered more fire events [5]. In particular, large-scale degradation and deforestation of the humid forests of Equatorial Asia have accentuated wildfire occurrence. This began with large-scale timber extraction in the 1970s followed by conversion to industrial plantations [9,10]. Logging roads, peatland drainage canals, Food Estates, and industrial plantations have further fragmented vast ‘otherwise forest-resistant’ humid forests, and lowered water tables in peat soils, making them highly flammable during droughts [11,12]. Even well-resourced oil palm and pulpwood estates, where fire risk declines once land conversion ends, regularly suffer significant losses of productive plantations to wildfire [13].

Land burning is illegal in Indonesia under Law No. 32 of 2009 on Environmental Protection and Management, with limited exceptions for traditional smallholder practices. Despite this ban, fire use remains widespread because enforcement is weak and land disputes persist [13]. Fire is employed by a range of actors, including small farmers, farmer associations, land speculators, and plantation companies as a low-cost method to clear land, dispose of woody debris, and fertilize soils [14,15]. In this context, detailed and rapid burned area assessments, when combined with contextual spatial data such as concession boundaries, cadastral maps, administrative boundaries, peatlands extent and drainage canals provide evidence for forensic investigations, help authorities enforce legislation and impose fines more effectively [13].

Satellite remote sensing provides consistent observations over large areas and, thus, has become an important tool for mapping burned areas [16]. Previous studies have used MODIS sensors for detecting burn scars [17]. The most used burned area products are the MCD64A1, FireCCI51, and C3SBA10 due to their global coverage [17–19]. However, these datasets miss small fires particularly in heterogeneous landscapes and smallholder agricultural areas [20]. Consequently, this limitation has driven the development of more detailed burned area maps using higher-resolution satellites (10–30 m) [21,22].

Research has shown that Sentinel-2 (10–20 m resolution) and Landsat (30 m) improve the detection of small burn scars in complex tropical landscapes. In Indonesia, Landsat has been extensively used by the Directorate of Forest and Land Fire Control [23], which is under the Directorate General for Forestry Law Enforcement of the Ministry of Forestry and is responsible for forest-fire control policy. Similarly, Sentinel-2 data are increasingly used to detect burned areas in parts of Indonesia [24–27] and to generate national-level burned-area datasets in sub-Saharan Africa [20].

Despite these advancements, mapping burned areas on monthly basis remains challenging in tropical regions. Persistent cloud cover and heterogeneous land cover types complicate the detection of burn scars and reduce the consistency and accuracy of the maps [28,29]. Moreover, tropical vegetation can regrow rapidly after burning and burn scars may become undetectable within a few weeks. This further complicates the accurate monitoring of fires, especially over areas where satellite observations are frequently obscured by clouds.

Our previous study addressed these challenges by developing a methodology based on a machine learning model (Random Forest) that classified Sentinel-2 data [30]. The workflow was implemented in Google Earth Engine [31] and produced national-scale annual burned area maps. The method represented a major advancement because it produced improved burned area maps that were able to detect small fires and outperformed the MCD64A1 product [30]. Despite the improvements, the method was primarily designed for retrospective annual analysis. The increasing demand for timely geospatial data requires approaches that can deliver accurate burn-scar maps in near-real time. This can be achieved primarily by using fine-scale imagery from Sentinel-2, which offers a higher revisit frequency than Landsat. This operational approach would offer finer detail than the MCD64A1 product, at 500-m resolution, which is the most accurate global monthly burned area product currently in operation [17].

Recent studies have proposed methods for monthly burned-area mapping in Indonesia, using Sentinel-2 and other remote sensing inputs [26,32]. However, none have yet delivered an operational monthly burned-area mapping system for Indonesia. To address this gap, we present the first automated monthly burned-area processing chain for Indonesia. Our method integrates high-frequency Sentinel-2 time series imagery (every 2–5 days) with FIRMS (Fire Information for Resource Management System) daily fire hotspots [33]. These daily hotspots are derived from MODIS (Moderate Resolution Imaging Spectroradiometer) sensors onboard NASA’s Terra and Aqua satellites and the VIIRS (Visible Infrared Imaging Radiometer Suite) sensors onboard the Suomi NPP and NOAA-20 satellites [34]. These instruments detect daily thermal infrared anomalies associated with active combustion on the surface of the earth. In this study, the fire hotspot detections serve as spatial and temporal filters for identifying candidate burned areas detected using Sentinel-2 imagery. The workflow is implemented in Google Earth Engine [31], where Sentinel-2 imagery is processed to generate monthly burn-scar maps at 20-meter resolution, with a minimum mapping unit of 6.25 hectares — aligned with the official minimum burn-scar size adopted by Indonesian authorities [35].

Methods

Summary of methods

A burned area is identified by alteration of vegetation cover and structure along with deposits of char and ash. We mapped such areas using a change-detection approach, i.e., by comparing Sentinel-2 top-of-canopy reflectance before and after a burning event [36].

The monthly burned-area processing chain is composed of three consecutive steps. Step 1 involves the processing of Sentinel-2 images into monthly cloud-free pre- and post-fire composites, which reflect the Earth’s surface before and after the first day of the reference month. In this study, the month being processed in the processing chain is called the reference month. Step 2 entails the classification of the Sentinel-2 composites into monthly burned area maps using a ‘Random Forest’ classifier. Step 3 includes a post-classification procedure consisting of four sub-steps: a) removal of small patches of pixels (<6.25 ha) that were detected as burned by the ‘Random Forest’ classification, aligning with the Indonesian official minimum mapping unit requirements, b) morphological operations to refine the visual appearance of the classification layer, c) removal of ‘burned’ false positives using FIRMS data, and d) identification of burned areas that were also detected in preceding months. Steps 1, 2, 3a,b are modifications of the annual burned-area processing chain published previously [30], whereas steps 3c,d are additional steps specific for the monthly burned-area processing chain. Fourth, we assessed our monthly burned-area map, against the monthly MCD64A1 burned-area product [17], and against our previously published reference annual dataset [30] to gauge the reliability and accuracy of the three burned-areas products. Details of the operational execution, automation, scheduling, and integration of the processing chain with the Nusantara Atlas platform are described in S1 Text and illustrated in Figure 1 in S1 Text.

Sentinel-2 pre- and post-fire Sentinel-2 monthly composites

Here, we describe our procedure to create pre- and post-fire composites. Prior to creating the composites, we removed non-valid pixels using the Sentinel-2 imagery quality flag (this flag provides information about clouds, cloud shadows, and other non-valid observations) produced by the ATCOR processing chain and included in the atmospherically-corrected surface reflectance multispectral images of the Sentinel-2 A and B satellites Surface Reflectance products (Level 2A product) [37].

The pre-fire composite is built using imagery from a three-month window preceding the reference month (i.e., the month being analysed by the processing chain). For example, if September is the reference month, the pre-fire composite includes all valid observations from June 1st to August 31st. This relatively long compositing window was chosen to ensure sufficient cloud-free observations while capturing a stable representation of the vegetation and land cover prior to fire events. In tropical regions like Indonesia, where cloud cover is frequent, this approach reduces the risk of missing data due to persistent atmospheric interference. Moreover, since vegetation typically undergoes minimal short-term changes before a fire, this extended temporal window does not compromise the characterization of pre-fire conditions but instead increases compositing robustness.

The post-fire composite, by contrast, includes only observations acquired during the reference month (e.g., September 1st to 30th), ensuring that it captures fire-related spectral changes as close as possible to the time of burning. For both pre- and post-fire composites, we apply a median aggregation of all valid pixel values within the compositing window to reduce noise from outliers and residual artifacts.

This approach differs from our previous annual-burned-area processing chain [30], which used a pixel-wise moving window to determine the compositing period based on the minimum Normalized Burn Ratio (NBR) value observed each year. As a result, annual composites tend to capture burn scars accumulated over the entire fire season, while the monthly composites are temporally constrained to the reference month and are therefore more suitable for near-real-time monitoring (Figs 1 and 2).

Fig 1 — The example corresponds to a single pixel (20 × 20 m) in Kalimantan, where the reference month being processed is September. The pre- and post-fire compositing windows used in the monthly burned-area processing chain are indicated.

Fig 2 — Comparing Sentinel-2 pre- and post-fire monthly composites (August-November 2019) with the annual composites published previously [30]. Sentinel-2 composites are displayed in false colours (RGB: short-wave infrared, band 11; near infrared, band 8; blue: red, band 4). The images reveal large burn scars, visible as areas that have transitioned from green to dark brown/red tones. The Monthly sequence reveals the progression of the burn scars.

Supervised burned/unburned classification

We used the ‘Random Forest’ [38], a supervised classification algorithm available in the Google Earth Engine, to determine whether the spectral changes observed by the monthly pre- and post-fire composites corresponded to a fire event, and subsequently classify burned areas. ‘Random Forest’ is a popular algorithm because it is a non-parametric model that can process large datasets and yield high accuracies without requiring a large amount of training data or rigorous hyper-parameter optimization. The ensemble nature of random forests also reduces the risk of overfitting.

The input variables used in the ‘Random Forest’ are the spectral bands of Sentinel-2 in the pre- and post-fire composites plus their respective NBR index. We excluded the bands at 60-meter spatial resolution (bands B1, B9, and B10) since these bands present a low spatial resolution for the aim of the study. Therefore, we used a total of 22 features: the NBR and bands B2, B3, B4, B5, B6, B7, B8, B8A, B11, and B12 and the NBR index of the pre- and post-composites. The 10-meter bands (B2, B3, B4, and B8) were resampled to 20 meters to align with the spatial resolution of B5, B6, B7, B8A, B11, and B12. The class predictions of the ‘Random Forest’ were 0 (class ‘burned’) and 1 (class ‘unburned’).

Supervised classification models require training data, that is, exemplary spectral signatures of ‘burned’ and ‘unburned’ lands in the present case, to guide the to reliably classify the target classes. In this study, we used the training dataset published previously [30], and collected additional training data to improve the classification of the monthly Sentinel-2 composites. The additional training points were collected to reduce false positives (classified as ‘burned’) over moist, vegetation-free soils, particularly in peatland areas during wet years (2020 and 2021) when mechanically cleared peatlands with exposed, water-saturated soils exhibit spectral signatures like those of burn scars. The final training dataset consists of 2,343 points; 852 points for the class ‘burned’ and 1,491 points for the class ‘Unburned’ (Fig 3).

Fig 3 — The map above shows the 2,343 training points used to train the Random Forest. The map below shows the 1,042 reference points used to validate the burned area map.

Post classification procedure

Masking of small patches.

The first step in the post-classification involved removing small patches of pixels categorized as ‘Burned.’ The threshold was set to 6.25 ha, equivalent to 156 pixels in the 20-meter classification images. Small patches below 6.25 ha, including isolated pixels classified as ‘Burned,’ were removed to stay in line with Indonesia’s ministry of forestry who adopted a minimum mapping unit for national burned area assessments of 6.25 ha.

Morphological operations.

Some detected burned areas had unnaturally rough or blocky shapes that didn't match the true outlines of the fire scars. To correct this, morphological operations were applied to smooth and round the shapes of the burn patches and improving their visual realism. The process involved two dilation steps: i) opening using a 3x3 square kernel, which helps remove small noise and detach weak connections; and ii) closing using a 3x3 circular kernel, which fills small gaps and further smooths the shape. Additionally, a rule was applied to avoid misclassifying vegetated areas as burned: If a pixel's post-fire NBR (Normalized Burn Ratio) was greater than −0.1, it was kept as ‘unburned’. This threshold filtered out vegetated pixels, where burning is unlikely.

Masking false positives with FIRMS daily fire hotspots.

Despite using an improved training dataset, false positives persisted, particularly during wet years, when peatlands with exposed, water-saturated soils exhibit spectral signatures like those of burn scars. To correct this, we used a post-classification procedure that filtered the false positives of class ‘Burned’ using daily fire hotspot data from FIRMS. FIRMS data comprises points where MODIS or VIIRS satellites detected thermal anomalies, indicating the presence active fires [33,39]. A hotspot point represents the centroid of a pixel in which a thermal anomaly was detected and, thus, the location of the fire is within the perimeter of the pixel. As a result, hotspot points have an uncertainty that is associated to the spatial resolution of the satellite sensors. The spatial resolution is approximately 1 km for MODIS and approximately 375 m for VIIRS. In the first step of this procedure, we buffered by 100 meters the pixels that we detected as burned in our monthly burned area classification. Then, we re-classified the ‘Burned’ pixels to ‘Unburned’ if no fire hotspot was present within the burn scar or in the buffering area. This rule ensured that only the patches of pixels where there is evidence of active fire were considered in the final burned area layer. The masking procedure uses daily FIRMS hotspot data taken in the three months before the reference month and the reference month. For instance, if September is the reference month, the hotspot points include FIRMS data detected from July 1^st until September 30^th.

Masking repeated burning detection.

The burned area processing chain produces monthly burned area layers intended to capture fire activity within a specific reference month. However, because burn scars can remain visible in Sentinel-2 imagery for months, the same area may be repeatedly detected as ‘Burned’ across consecutive months, leading to potential double counting. To prevent this, we applied a filtering rule that reclassifies a pixel as ‘Unburned’ if it had already been detected as ‘Burned’ in any of the previous three months. For example, if a pixel was detected as ‘Burned’ in February 2019, and remained visible in March, April and May, it would not be counted again in those subsequent months (Fig 4).

Fig 4 — Pre- and post-fire Sentinel-2 monthly composites, and corresponding burned-area classifications for February, March, April, and May 2019 are shown over an area (*Pulau Rupat*) in Riau province, Sumatra where peatland burn scars remained visible several months after fire. Pixels that were detected as ‘Burned’ for the first time are shown in red, while pixels that were repeatedly detected as ‘Burned’ are shown in orange.

Burned-area map validation

To assess the spatial accuracy of our monthly burned-area product, we validated it against a randomly distributed reference dataset originally developed for the annual burned-area product published previously [30]. The validation follows best practices guidelines and employs a balanced sampling approach to minimize bias [40]. The reference dataset was created using stratified-random sampling of ‘Burned’ and ‘Unburned’ sites across seven fire-prone provinces in Indonesia. Each site (20 × 20 m) was visually interpreted using the original time-series Sentinel-2 imagery by three independent interpreters to detect fire activity based on the presence of flaming fronts, smoke, or charred vegetation.

In our previous publication, we sampled reference points to validate three burned area layers: our annual burned-area layer, the MCD64A1 annual burned-area product, and an annual burned-area map produced by the Indonesian Ministry of Environment and Forestry [30]. The validation dataset contained 1,168 reference points, including 280 ‘burned’ and 888 ‘unburned’ points. For this study, we adjusted the number of reference points to validate monthly burned-area layer. This correction was necessary to ensure a proper stratified random sampling, in which all pixels mapped with the same class have the same probability of being sampled. The correction consisted of i) randomly sampled 20 points within the area classified as ‘Unburned’ in the monthly burned-area layer and ‘Burned’ in the annual burned-area layer (area in light red in Fig 5); ii) randomly reduce the number of points classified as ‘Burned’ in the monthly burned-area layer and ‘Unburned’ in the annual burned-area layer to 4 points (area in cyan), and iii) randomly reduce the number of points classified as ‘Burned’ in the monthly and annual burned-area layer to 150 points (area in red). The final reference point dataset used for validating the monthly burned-area layer consisted of 1,042 points: 150 ‘Burned’ points and 892 ‘Unburned’ points (Fig 3, lower panel).

Fig 5 — From the original 1,168 points, we adjusted the sample to ensure proper stratified random sampling in the monthly burned-area layer in 2019.

We computed overall accuracy (OA), producer’s accuracy (PA), and user’s accuracy (UA) with 95 % confidence interval. Validation was conducted on the cumulative monthly burned-area map for 2019, which is the same year as the reference dataset. The validation metrics were also calculated for the annual burned-area layer and the MCD64A1 product.

To evaluate the temporal accuracy of the monthly burned-area product, we compared the month of detection against the independent reference dataset derived from visual interpretation of Sentinel-2 time series. This reference dataset consists of burn dates for 259 points across seven fire-prone provinces in Indonesia. These 259 points include all reference points classified as ‘burned’ in the burned area layer and with a true class of ‘burned’. We did not apply the point correction used in the spatial validation illustrated in Fig 5 as a stratified sampling is not required for the temporal validation. For each site, we recorded the first month in which the burn scar was visible in the Sentinel-2 imagery. We then constructed a confusion matrix comparing the month assigned by the burned-area processing chain with the month determined by visual inspection. This comparison allowed us to assess how accurately the processing chain captures the timing of burn events at monthly resolution and to identify potential lags in the detection of the burn events.

We also evaluated burn size distributions across burned-area datasets to examine potential biases in fire detection. We compared the frequency distributions of burn scars among our Sentinel-2-derived burned-area maps, and the MCD64A1 dataset. Differences in burn size distributions were tested using the Kruskal-Wallis H test, followed by Kolmogorov-Smirnov and Mann-Whitney U tests to assess whether specific datasets over- or under-represented certain burn size classes.

Results

Burned-area detection

In six years, between 2019 and 2024, fires burned a cumulative 5.62 million hectares (Mha), including 2.92 Mha burned once and 1.12 Mha burned multiple times (Fig 6). This represents a total burned extent of 4.04 Mha.

The spatial distribution of burned area from 2019 to 2024 reveals a highly uneven pattern across Indonesia’s regions and provinces (Fig 7-8). At the regional scale, Kalimantan, Nusa Tenggara, and Sumatra experienced the largest extent of burning, with each surpassing 1 million hectares (Mha). These three regions consistently recorded substantial fire activity across multiple years, particularly during peak fire seasons in 2019 and 2023. At the provincial level, Nusa Tenggara Timur, Kalimantan Tengah, Papua Selatan, and Sumatera Selatan stand out as the most affected, collectively accounting for 52.2% of total burned area during the 2019–2024 period.

Fig 8 — Burned area is shown in million hectares (Mha), with bars stacked by year to illustrate interannual variability. The highest levels of fire activity occurred in Kalimantan, Bali & Nusa Tenggara, and Sumatra, with the provinces of Nusa Tenggara Timur, Kalimantan Tengah, Papua Selatan, and Sumatera Selatan accounting for 52.2% of the national total across the six-year period. Peak fire years (2019 and 2023) dominate most provinces, while lower activity is seen in 2021 and 2022.

When examined by year, national cumulative burn trends from 2019 to 2024 show notable interannual variability: peak burning occurred in 2019 and 2023, moderate levels were observed in 2020 and 2024, and fire activity was lowest in 2021 and 2022 (Fig 9). Our method detects significantly more burns than the MCD64A1 across all years. For example, in 2019, the burned area reached 2.64 Mha in our dataset compared to 2.09 Mha in MCD64A1; in 2023, the difference was more pronounced, with 1.70 Mha recorded in our study versus 1.11 Mha in MCD64A1. Even during low-fire years such as 2021–2022, MCD64A1 continued to underestimate burned area, reporting 0.05–0.07 Mha, compared to 0.06–0.11 Mha in this study. Fire activity in 2020 and 2024 was of intermediate intensity: in 2024, we recorded 0.49 Mha versus only 0.15 Mha in MCD64A1, and in 2020, the difference between both datasets was largest, with 0.62 Mha detected by our method compared to 0.18 Mha in MCD64A1.

Fig 9 — This graph compares results from this study (left panel, red lines) and the MCD64A1 burned area product (right panel, black lines). Each line represents the cumulative burned area for a given year, with final annual totals indicated on the right. The sharp increase in burned area from June to October reflects the seasonal pattern typical of fire activity in Indonesia, driven by dry conditions during these months. The consistent underestimation by MCD64A1 highlights the improved detection capacity of this study’s approach.

Using a forest area mask representing conditions in January 2019, which we developed previously [41], we detected 122,164 ha of burned area in primary forests between 2019 and 2024, corresponding to 2.2% of the total burned area (5.62 Mha). MODIS reported a higher burned forest area (180,928 ha), which represents 5.0% of its total burned area. Regarding peatlands, the MCD64A1 product also reported a higher proportion of burning on peat (27.4%), whereas our estimate showed lower, but still substantial share of 16.1%. These results indicate that MCD64A1 consistently underestimates burned area across all years, which stems from its coarse spatial resolution and limitations in detecting small fires, leading to an incomplete representation of fire extent in Indonesia, and suggest that our maps capture a broader extent of burning, over both peatland and mineral soils, indicating improved detection across diverse land types.

Fire activity and climatic anomalies

Analysis of monthly burned area across Indonesia from January 2019 to December 2024 reveals a strong association between fire activity and large-scale climatic anomalies (Fig 10). Two major fire seasons occurred in late 2019 and 2023, with monthly burned areas peaking in September and October. These peaks in fire activity closely align with positive phases of both the Indian Ocean Dipole (IOD) and the El Niño Southern Oscillation (ENSO), the latter indicated by positive values of the Oceanic Niño Index (ONI), which reflect El Niño conditions. In 2019, a strong positive IOD (+1.8°C) and a moderate El Niño (ONI reaching +0.5°C) coincided with the highest burned area recorded during the study period. Similarly, in 2023, a strong El Niño (ONI > 1.5°C from September to December) alongside a positive IOD (+1.5 °C), preceding the second largest fire season observed. In contrast, the period from 2020 to 2022 was characterized by neutral to negative IOD and ONI phases, coinciding with minimal fire activity, with peak monthly burned areas consistently below 0.2 Mha and approaching zero in several months. During positive IOD and El Niño phases, reduced convection and rainfall lead to drier and warmer conditions across much of Indonesia, increasing fuel flammability and extending the dry season. Conversely, La Niña and negative IOD phases enhance rainfall and humidity, reducing fire occurrence. These results highlight the critical role of El Niño and positive IOD events in driving severe fire seasons in Indonesia, whereas neutral and La Niña conditions are associated with substantially reduced burned area.

Fig 10 — The top panel shows total burned area across all provinces (in million hectares, Mha). The bottom panel displays the Indian Ocean Dipole (IOD; black line, left axis, °C) and Oceanic Niño Index (ONI; blue line, right axis, °C), representing major climate anomalies. Peaks in burned area during late 2019 and mid to late 2023 align with positive IOD and El Niño conditions, while reduced burned area between 2020 and 2022 corresponds to neutral or negative phases of both indices. Vertical lines mark the start of each calendar year.

Accuracy assessment

In the seven provinces for which we assessed accuracy, our monthly burned area estimate, aggregated over one year (2019) presents a high user’s accuracy (UA) for the ‘burned’ class, at 98.7% (CI: 97.7%−99.6%) indicating a low 1.3% (CI: 0.4%−2.3%) commission-error rate (Table 1). This result is on par with our previously published Annual estimate, which reported a similar UA for the ‘burned’ class, at a 97.9% (CI: 97.1%−98.8%) [30]. By contrast, the MCD64A1 data had a much lower UA for the burned class, at 76.0% (CI: 73.3%−78.7%).

Table 1. Accuracy assessment of each of the three burned-area maps performed in seven Indonesian provinces (87.60 Mha) targeted for peatland restoration. The accuracy metrics were estimated with a 1,042 points randomly distributed using stratified sampling. The reported metrics are (1) the overall accuracy (OA), the user accuracy (UA), and the producer accuracy (PA) with their 95 % confidence intervals.

		Annual (Gaveau et al. 2021)	Monthly (This study)	MCD64A1
OA (%)		99.3 (99.0, 99.6)	99.0 (98.7, 99.3)	98.4 (98.0, 98.7)
	Unburned	99.3 (99.1, 99.6)	99.0 (98.7, 99.3)	98.8 (98.4, 99.1)
UA (%)	Burned	97.9 (97.0, 98.7)	98.7 (97.7, 99.6)	76.0 (73.3, 78.7)
	Unburned	100.0 (99.3, 100.0)	100.0 (99.0, 100.0)	99.6 (98.7, 99.6)
PA (%)	Burned	75.6 (68.3, 83.0)	65.3 (57.9, 72.7)	53.1 (45.8, 60.5)

Open in a new tab

The producer’s accuracy (PA) for the ‘burned’ class in our Monthly estimate was 65.3% (CI: 57.9%–72.7%). This PA is lower than the metric reported for the Annual burned-area estimate, at 75.6% (CI: 68.3%–83.0%), although the difference is not statistically significant. This translates to an omission error rate of 34.7% (CI: 27.3%–42.1%) for the Monthly dataset. While the confidence intervals overlap slightly, our monthly approach shows a higher PA than the MCD64A1 product, which reported a PA of 53.1% (CI: 45.8%–60.5%), suggesting improved sensitivity to burned areas. This highlights that while the Monthly burned-area product has a higher omission error than the annual product, it still offers a notably improved detection rate compared to MCD64A1. Both burned-area datasets underestimate the true burned area extent, as per their respective PA figures, but our Monthly burned-area processing chain strikes a stronger balance between producer and user accuracy, while delivering timely, monthly updates at fine spatial resolution.

The confusion matrix (Fig 11) shows the count of reference sites (n = 259) classified as burned in a given month by the monthly processing chain (y-axis) and the month in which the same sites were confirmed as burned by visual inspection (x-axis) in the original point dataset. The results demonstrate strong temporal alignment, with 206 out of 250 sites (80%) correctly assigned. 68 sites were correctly assigned to September and 56 sites to October — the two peak fire months in 2019. Only 2% of the points showed detection mismatches longer than one month. This indicates that our method assigns fire timing accurately in most cases, while also reflecting the known difficulty of detecting fires that span multiple months.

Burn size comparison

To assess how fires of different sizes contribute to total burned area, we examined the cumulative burned area as a function of burned patch size for each year from 2019 to 2024 (Fig 12). Burned patch sizes are presented on a logarithmic scale to accommodate the wide range observed, from small burn patches of just a few hectares to very large scars spanning tens of thousands of hectares. The results reveal consistent differences between our monthly burned area maps (Sentinel-2, this study) and those from the MCD64A1 product. Across all years, our dataset (red line) captures a larger cumulative burned area across the full spectrum of patch sizes. This discrepancy is especially marked in major fire years like 2019 and 2023, where our processing chain detects many larger, burned patches (>1,000 ha) than MCD64A1. Even in low-fire years such as 2021 and 2022, our approach identifies more burned area in small patches (<100 ha), underscoring MCD64A1’s limited ability to detect fine-scale fires. Notably, the red curves from our maps show a near-linear trend on the log-scale axis, indicating a size-frequency pattern that approximates a power-law distribution — characteristic of scale-invariant fire regimes. In contrast, the MCD64A1 curves are more irregular and consistently fall below ours, confirming their systematic underestimation of burned area, particularly for small and fragmented fires. These results highlight the value of high-resolution, monthly Sentinel-2 data for producing more complete and accurate burned area assessments across Indonesia.

Fig 12 — The red line shows results from this study, while the black line represents the MODIS MCD64A1 burned area product. For each year, cumulative burned area (in million hectares, Mha) is plotted against individual burned patch sizes (in hectares, ha, logarithmic scale). The x-axis represents burned patch size on a logarithmic scale, allowing for a clear visualization of both small and large patches, which vary over several orders of magnitude. Across all years, this study consistently maps a larger cumulative burned area and detects more burned patches across a wide range of sizes compared to MCD64A1. The differences are especially pronounced for large patches in major fire years (2019 and 2023), but are also evident for small, burned patches, particularly in low-fire years (2020–2022), indicating that MCD64A1 underestimates both small and large fires in Indonesia.

The statistical results in Table 2 support these findings. We applied the Kruskal-Wallis test to compare scar size frequency distributions across datasets at four patch size thresholds (25, 100, 1,000, and 5,000 ha). At fine resolutions (25 and 100 ha), p-values are extremely low (< 10 ⁻ ²⁴), indicating significant differences in the size distributions among Sentinel-2, MCD64A1, and official maps. This confirms that the products diverge most strongly at small patch sizes, where MCD64A1 consistently underrepresents burned areas. In contrast, at larger size thresholds (1,000 and 5,000 ha), the p-values exceed 0.3 and 0.6, respectively, showing no significant differences.

Table 2. Tests statistics with respect to differences in burned-area scar size frequency distributions for Sentinel, MODIS, and official maps.

Scar size (ha)	Kruskal-Wallis test (p-values)
25	2.9537e-291
100	3.8813e-25
1000	0.32109
5000	0.65824

Open in a new tab

Discussion

In Indonesia, fire is both a farming tool and a natural feature of savanna ecosystems, but it can also become a destructive force that degrades humid forests, swamps, and agricultural lands. Government agencies already monitor fire activity every month [23], with some success, but current systems rely heavily on visual interpretation of Landsat, and on FIRMS hotspots, which makes it difficult to capture the complete extent of fire-affected areas.

This study presents the first operational monthly burned-area processing chain designed to detect fires in unprecedented detail for Indonesia. The system combines high-frequency Sentinel-2 imagery (every 3–5 days) with FIRMS daily fire hotspots to produce 20-m resolution monthly burned-area maps. The processing chain runs in Google Earth Engine and delivers timely, monthly information that can support emergency response, forensic investigations, law enforcement, and conservation planning across Indonesia.

From 2019 to 2024, our processing chain mapped 5.62 million hectares of cumulative burned area, equivalent in size to the country of Croatia. Most fires occurred in Kalimantan, Nusa Tenggara, and Sumatra. Six of Indonesia’s 38 provinces — Nusa Tenggara Timur, Kalimantan Tengah, Papua Selatan, Sumatra Selatan, Jawa Timur, and Kalimantan Selatan — accounted for 68% of the total burned area. These regions reflect three main fire regimes, i) large wildfires on degraded previously deforested lands, including peatlands in Kalimantan Tengah, Sumatra Selatan, and Kalimantan Selatan, ii) frequent savanna wildfires in Papua Selatan and NTT islands such as Sumba, and iii) frequent maintenance agricultural fires in Jawa Timur. Among these fire regions, savanna and maintenance fires pose comparatively lower environmental concerns.

We found that severe fire seasons (2019 and 2023) coincided with positive phases of both the Indian Ocean Dipole (IOD) and the El Niño–Southern Oscillation (ENSO), the latter measured by the Oceanic Niño Index (ONI). This pattern aligns with previous findings [7], which showed that the combined influence of ENSO and IOD intensify dry spells and fire activity across Sumatra and Kalimantan. Fire activity accelerated from July and peaked in September–October when ONI ≥ ~+0.5 °C and IOD ≥ +1.5 °C. In contrast, 2020–2022 saw minimal burning under neutral or negative phases of both indices, while 2024 showed moderate fire activity without strong anomalies. These findings reaffirm the importance of tracking both indices (ONI and IOD) for early warning. We note that this analysis is descriptive and intended to illustrate temporal co-variation between climatic indices and monthly burned-area fluctuations, rather than to quantify causality. Future work will use statistical models and incorporate additional climatic variables such as air temperature, rainfall, vapor pressure deficit, or climatic water deficit. These variables are known drivers of fire occurrence [42] that can determine the magnitude and causality of climatic influences.

The Indonesian government has reduced forest fires in recent years [43], and our analysis confirms this trend. The last two fire seasons in which both the El Niño–Southern Oscillation (ENSO) and the Indian Ocean Dipole (IOD) were in positive phases (2015 and 2023) produced comparable drought across Indonesia. The 2019 fire season, although not linked to El Niño, was driven by an exceptionally strong positive IOD, the most intense since the 1960s, which also caused severe rainfall deficits in Kalimantan and Sumatra [44,45]. Despite the extreme dry conditions, we found that in both 2019 and 2023, 122,164 hectares of primary humid forest burned, equivalent to 2.2% of the total burned area. This contrasts sharply with previous extreme fire years: in 1997/98, an estimated 5 million hectares of forest burned [12], while in 2015/16, the figure was 450,000 hectares [46]. These findings suggest that recent fire prevention and response efforts have a tangible impact, particularly in humid forests during severe drought years.

Our monthly Sentinel-2 approach consistently detected more burned area compared to the widely used MCD64A1 product. This discrepancy is especially marked in major fire years like 2019 and 2023, when our processing chain detected larger, burned patches (>1,000 ha). Even in low-fire years such as 2021 and 2022, our approach identifies more burned area in small patches (<100 ha), which highlight the inability of MCD64A1 to detect small fires compared to our results. This underestimation is expected because the 500-m MCD64A1 product cannot reliably detect smaller fires [17]. Similar conclusions were reached by over Sub-Sahara Africa [20] when comparing the MCD64A1 product with ‘Small Fire Dataset’ derived from Sentinel-2 [47]. Our burn-size analysis showed that Sentinel-2 results follow a scale-free distribution [48,49], which is common of wildfire regimes worldwide [50]. In contrast, MCD64A1 showed an S-shaped curve that indicated bias at both small and large scales. Our results further demonstrates that coarse-resolution products underestimate both the smallest and largest burn scars. Accurate detection of both ends of this size spectrum is important for characterizing fire regimes (wildfire, agricultural fire), their scale and severity. Despite this, MCD64A1 remains a valuable globally operational product that provides consistent long-term burned area records across the world. Understanding how it compares with finer-resolution datasets such as Sentinel-2 is useful for bias correction and better interpretation of global fire trends.

Validation confirmed the reliability of our processing chain, with a user’s accuracy for the “burned” class of 98.7% and a commission error of only 1.3% (CI: 0.4%–2.3%). The producer’s accuracy of 65.3% and omission error of 34.7% (CI: 27.3%–42.1%) reflect our conservative approach that minimizes false positives. We prioritised a low commission error rate (i.e., high user accuracy) to address sensitivities [51]. This makes the maps suitable for legal and enforcement use. The omission of burn scars commonly occur in low-intensity fires beneath forest canopies [52], short-lived fires in savannas grassland that re-green within a week, or burn scars that are obscured by persistent cloud or haze. Although our approach produces conservative burned area estimates, it ensures that every detected scar can be used with high confidence, particularly when overlaid with contextual layers such as land zoning plans. For instance, the overlay of our burned area maps with concession boundaries allows authorities to produce objective fire evidence every month. This can help to act quickly, impose fines, and hold landowners accountable.

The proportion of burned area occurring on peatlands in our analysis (16.1%) is lower than that reported by the MCD64A1 burned-area product (27.4%) and lower than our own previous estimates for year 2019 [30]. This is because we used an updated national peatland map from the Indonesian Ministry of Agriculture, which has a smaller spatial extent than earlier peatland datasets [53]. These differences show that estimates of peatland area burning are sensitive to how peatland extent is mapped. They emphasize the importance of continued improvements in national peatland maps given that peat fires reduction is central to addressing concerns about regional air quality and achieving Indonesia’s ambitious national carbon-emissions reduction target.

This study presents the first operational method that produces monthly 20-m resolution burned-area maps for Indonesia. The workflow is largely automated and fully replicable, and provides high-confidence data suitable for consistent national monitoring. To support wider adoption, the burned-area maps and their monthly updates have been integrated into the Nusantara Atlas platform (www.nusantara-atlas.org), an open-access platform for monitoring deforestation in Southeast Asia. This supports month-by-month burned-area estimates across a broad array of land units, including oil palm, pulpwood and mining concessions, administrative units (down to the municipality level), and protected areas, for use by government agencies, NGOs, and the public. The operational execution, automation, and integration of the processing chain with the Nusantara Atlas platform are described in Supplementary Methods. Long-term availability of the data will depend on future cloud-processing and data-hosting costs under Google’s pricing policy. Overall, the system provides a robust and replicable framework for operational burned-area monitoring in Indonesia and can be adapted to other tropical regions facing similar challenges.

Supporting information

S1 Text. Processing pipeline for Nusantara Atlas.

Description of the processing pipeline used to integrate the results into the Nusantara Atlas platform.

(DOCX)

pone.0331831.s001.docx^{(221.8KB, docx)}

Data Availability

The data underlying the results presented in the study are available from Zenodo at: https://zenodo.org/records/19104282.

Funding Statement

The author(s) received no specific funding for this work.

References

1.Gritzo LAL. Mitigating social and economic impact of wildfires. Appl Energy Combust Sci. 2024;20:100285. https://www.sciencedirect.com/science/article/pii/S2666352X24000402?via%3Dihub [Google Scholar]
2.Bowman DMJS, Balch J, Artaxo P, Bond WJ, Cochrane MA, D’Antonio CM, et al. The human dimension of fire regimes on Earth. J Biogeogr. 2011;38(12):2223–36. doi: 10.1111/j.1365-2699.2011.02595.x [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Bowman DMJS, Balch JK, Artaxo P, Bond WJ, Carlson JM, Cochrane MA, et al. Fire in the Earth system. Science. 2009;324(5926):481–4. doi: 10.1126/science.1163886 [DOI] [PubMed] [Google Scholar]
4.Tacconi L. Preventing fires and haze in Southeast Asia. Nature Clim Change. 2016;6(7):640–3. doi: 10.1038/nclimate3008 [DOI] [Google Scholar]
5.Field RD, Van Der Werf GR, Shen SS. Human amplification of drought-induced biomass burning in Indonesia since 1960. Nature Geosci. 2009;2(3):185–8. [Google Scholar]
6.Goldammer JG, Seibert B. The impact of droughts and forest fires on tropical lowland rain forest of East Kalimantan. In: Fire in the tropical biota: Ecosystem processes and global challenges. Springer; 1990. 11–31. [Google Scholar]
7.Nurdiati S, Bukhari F, Julianto MT, Sopaheluwakan A, Aprilia M, Fajar I, et al. The impact of El Niño southern oscillation and Indian Ocean Dipole on the burned area in Indonesia. TAO. 2022;33(1). doi: 10.1007/s44195-022-00016-0 [DOI] [Google Scholar]
8.Wooster M, Perry G, Zoumas A. Fire, drought and El Niño relationships on Borneo (Southeast Asia) in the pre-MODIS era (1980–2000). Biogeosciences. 2012;9(1):317–40. [Google Scholar]
9.Gaveau DLA, Sloan S, Molidena E, Yaen H, Sheil D, Abram NK, et al. Four decades of forest persistence, clearance and logging on Borneo. PLoS One. 2014;9(7):e101654. doi: 10.1371/journal.pone.0101654 [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Tsujino R, Yumoto T, Kitamura S, Djamaluddin I, Darnaedi D. History of forest loss and degradation in Indonesia. Land Use Policy. 2016;57:335–47. doi: 10.1016/j.landusepol.2016.05.034 [DOI] [Google Scholar]
11.Miettinen J, Shi C, Liew SC. Fire Distribution in Peninsular Malaysia, Sumatra and Borneo in 2015 with Special Emphasis on Peatland Fires. Environ Manage. 2017;60(4):747–57. doi: 10.1007/s00267-017-0911-7 [DOI] [PubMed] [Google Scholar]
12.Page SE, Siegert F, Rieley JO, Boehm H-DV, Jaya A, Limin S. The amount of carbon released from peat and forest fires in Indonesia during 1997. Nature. 2002;420(6911):61–5. doi: 10.1038/nature01131 [DOI] [PubMed] [Google Scholar]
13.Gaveau DLA, Pirard R, Salim MA, Tonoto P, Yaen H, Parks SA, et al. Overlapping land claims limit the use of satellites to monitor no‐deforestation commitments and no‐burning compliance. Conserv Letters. 2016;10(2):257–64. doi: 10.1111/conl.12256 [DOI] [Google Scholar]
14.Dennis RA, Mayer J, Applegate G, Chokkalingam U, Colfer CJP, Kurniawan I, et al. Fire, People and Pixels: Linking Social Science and Remote Sensing to Understand Underlying Causes and Impacts of Fires in Indonesia. Hum Ecol. 2005;33(4):465–504. doi: 10.1007/s10745-005-5156-z [DOI] [Google Scholar]
15.Purnomo H, Shantiko B, Sitorus S, Gunawan H, Achdiawan R, Kartodihardjo H, et al. Fire economy and actor network of forest and land fires in Indonesia. Forest Pol Econ. 2017;78:21–31. doi: 10.1016/j.forpol.2017.01.001 [DOI] [Google Scholar]
16.Chuvieco E, Mouillot F, van der Werf GR, San Miguel J, Tanase M, Koutsias N, et al. Historical background and current developments for mapping burned area from satellite Earth observation. Remote Sens Environ. 2019;225:45–64. doi: 10.1016/j.rse.2019.02.013 [DOI] [Google Scholar]
17.Giglio L, Boschetti L, Roy DP, Humber ML, Justice CO. The Collection 6 MODIS burned area mapping algorithm and product. Remote Sens Environ. 2018;217:72–85. doi: 10.1016/j.rse.2018.08.005 [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Lizundia-Loiola J, Franquesa M, Boettcher M, Kirches G, Pettinari ML, Chuvieco E. Operational implementation of the burned area component of the Copernicus Climate Change Service: from MODIS 250 m to OLCI 300 m data. Earth System Science Data Discuss. 2021;1–37. [Google Scholar]
19.Lizundia-Loiola J, Otón G, Ramo R, Chuvieco E. A spatio-temporal active-fire clustering approach for global burned area mapping at 250 m from MODIS data. Remote Sens Environ. 2020;236:111493. doi: 10.1016/j.rse.2019.111493 [DOI] [Google Scholar]
20.Ramo R, Roteta E, Bistinas I, van Wees D, Bastarrika A, Chuvieco E, et al. African burned area and fire carbon emissions are strongly impacted by small fires undetected by coarse resolution satellite data. Proc Natl Acad Sci U S A. 2021;118(9):e2011160118. doi: 10.1073/pnas.2011160118 [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Boschetti L, Roy DP, Justice CO, Humber ML. MODIS–Landsat fusion for large area 30 m burned area mapping. Remote Sens Environ. 2015;161:27–42. doi: 10.1016/j.rse.2015.01.022 [DOI] [Google Scholar]
22.Roy DP, Huang H, Boschetti L, Giglio L, Yan L, Zhang HH, et al. Landsat-8 and Sentinel-2 burned area mapping - A combined sensor multi-temporal change detection approach. Remote Sens Environ. 2019;231:111254. doi: 10.1016/j.rse.2019.111254 [DOI] [Google Scholar]
23.SiPongi. Forest and land fire monitoring system Jakarta, Indonesia. Ministry of Forestry, Government of Indonesia; 2025. https://sipongi.gakkum.kehutanan.go.id/indikasi-luas-kebakaran [Google Scholar]
24.Afira N, Wijayanto AW. Mono-temporal and multi-temporal approaches for burnt area detection using Sentinel-2 satellite imagery (a case study of Rokan Hilir Regency, Indonesia). Ecological Informatics. 2022;69:101677. doi: 10.1016/j.ecoinf.2022.101677 [DOI] [Google Scholar]
25.Arisanty D, Feindhi Ramadhan M, Angriani P, Muhaimin M, Nur Saputra A, Puji Hastuti K, et al. Utilizing sentinel-2 data for mapping burned areas in Banjarbaru Wetlands, South Kalimantan Province. Inter J Forestry Res. 2022;2022:1–12. doi: 10.1155/2022/7936392 [DOI] [Google Scholar]
26.Arjasakusuma S, Kusuma SS, Vetrita Y, Prasasti I, Arief R. Monthly Burned-Area Mapping using Multi-Sensor Integration of Sentinel-1 and Sentinel-2 and machine learning: case Study of 2019’s fire events in South Sumatra Province, Indonesia. Remote Sens Appl Soc Environ. 2022;27:100790. doi: 10.1016/j.rsase.2022.100790 [DOI] [Google Scholar]
27.Indah LA, Lalita L Dyah, Indah FH, Harwahyu R, Fitri SR. Machine learning approaches for burned area identification using Sentinel-2 in Central Kalimantan. J Appl Eng Science. 2020;18(2):207–15. doi: 10.5937/jaes18-25495 [DOI] [Google Scholar]
28.Hawbaker TJ, Vanderhoof MK, Beal Y-J, Takacs JD, Schmidt GL, Falgout JT, et al. Mapping burned areas using dense time-series of Landsat data. Remote Sens Environ. 2017;198:504–22. doi: 10.1016/j.rse.2017.06.027 [DOI] [Google Scholar]
29.Schultz MG, Heil A, Hoelzemann JJ, Spessa A, Thonicke K, Goldammer JG, et al. Global wildland fire emissions from 1960 to 2000. Global Biogeochemical Cycles. 2008;22(2). doi: 10.1029/2007gb003031 [DOI] [Google Scholar]
30.Gaveau DLA, Descals A, Salim MA, Sheil D, Sloan S. Refined burned-area mapping protocol using Sentinel-2 data increases estimate of 2019 Indonesian burning. Earth Syst Sci Data. 2021;13(11):5353–68. doi: 10.5194/essd-13-5353-2021 [DOI] [Google Scholar]
31.Gorelick N, Hancher M, Dixon M, Ilyushchenko S, Thau D, Moore R. Google Earth Engine: planetary-scale geospatial analysis for everyone. Remote Sens Environ. 2017;202:18–27. [Google Scholar]
32.Vetrita Y, Albar I, Santoso I, Prasasti I, Kartika T, Usman AB, et al. Monthly mapping of Indonesia’s burned areas: implementation, history, techniques, and future directions. Inter J Remote Sens. 2024;46(2):636–60. doi: 10.1080/01431161.2024.2421942 [DOI] [Google Scholar]
33.Giglio L, Schroeder W, Justice CO. The collection 6 MODIS active fire detection algorithm and fire products. Remote Sens Environ. 2016;178:31–41. doi: 10.1016/j.rse.2016.02.054 [DOI] [PMC free article] [PubMed] [Google Scholar]
34.NASA. From MODIS to VIIRS: continuing the legacy. 2025. https://www.earthdata.nasa.gov/news/blog/from-modis-viirs-continuing-legacy [Google Scholar]
35.Ministry of Environment and Forestry. Indonesia’s Second Forest Reference Emission Level (FREL) Submission to the UNFCCC. Jakarta, Indonesia: Ministry of Environment and Forestry (MoEF); 2022. [Google Scholar]
36.Liu S, Zheng Y, Dalponte M, Tong X. A novel fire index-based burned area change detection approach using Landsat-8 OLI data. European Journal of Remote Sensing. 2020;53(1):104–12. doi: 10.1080/22797254.2020.1738900 [DOI] [Google Scholar]
37.Fletcher K. Sentinel 2: ESA’s Optical High-Resolution Mission for GMES Operational Services. European Space Agency; 2012. [Google Scholar]
38.Breiman L. Random forests. Machine Learn. 2001;45(1):5–32. doi: 10.1023/a:1010933404324 [DOI] [Google Scholar]
39.Schroeder W, Oliva P, Giglio L, Csiszar IA. The New VIIRS 375 m active fire detection data product: algorithm description and initial assessment. Remote Sens Environ. 2014;143:85–96. doi: 10.1016/j.rse.2013.12.008 [DOI] [Google Scholar]
40.Olofsson P, Foody GM, Herold M, Stehman SV, Woodcock CE, Wulder MA. Good practices for estimating area and assessing accuracy of land change. Remote Sens Environ. 2014;148:42–57. doi: 10.1016/j.rse.2014.02.015 [DOI] [Google Scholar]
41.Gaveau DLA, Locatelli B, Salim MA, Husnayaen, Manurung T, Descals A, et al. Slowing deforestation in Indonesia follows declining oil palm expansion and lower oil prices. PLoS One. 2022;17(3):e0266178. doi: 10.1371/journal.pone.0266178 [DOI] [PMC free article] [PubMed] [Google Scholar]
42.Descals A, Gaveau DLA, Verger A, Sheil D, Naito D, Peñuelas J. Unprecedented fire activity above the Arctic Circle linked to rising temperatures. Science. 2022;378(6619):532–7. doi: 10.1126/science.abn9768 [DOI] [PubMed] [Google Scholar]
43.Sloan S, Tacconi L, Cattau ME. Fire prevention in managed landscapes: recent success and challenges in Indonesia. Mitigation and adaptation strategies for global change. 2021;26:1–30. [Google Scholar]
44.Du Y, Zhang Y, Zhang L, Tozuka T, Ng B, Cai W. Thermocline warming induced extreme indian ocean dipole in 2019. Geophy Res Let. 2020;47(18). doi: 10.1029/2020gl090079 [DOI] [Google Scholar]
45.Zhang L, Han W, Hu Z-Z. Inter-basin and Multi-time Scale Interactions in generating the 2019 Extreme Indian Ocean Dipole. J Clim. 2021;1–39. doi: 10.1175/jcli-d-20-0760.1 [DOI] [Google Scholar]
46.Goldman E, Carter S, Sims M. Fires drove record-breaking tropical forest loss in 2024. World Resources Institute; 2025. https://gfr.wri.org/latest-analysis-deforestation-trends [Google Scholar]
47.Roteta E, Bastarrika A, Padilla M, Storm T, Chuvieco E. Development of a Sentinel-2 burned area algorithm: Generation of a small fire database for sub-Saharan Africa. Remote Sensing of Environ. 2019;222:1–17. doi: 10.1016/j.rse.2018.12.011 [DOI] [Google Scholar]
48.Karsai I, Schmickl T, Kampis G. Forest Fires: fire management and the power law. In: Resilience and stability of ecological and social systems. Springer International Publishing; 2020. 63–77. doi: 10.1007/978-3-030-54560-4_4 [DOI] [Google Scholar]
49.Falk DA, Miller C, McKenzie D, Black AE. Cross-Scale Analysis of Fire Regimes. Ecosystems. 2007;10(5):809–23. doi: 10.1007/s10021-007-9070-7 [DOI] [Google Scholar]
50.Malamud B, Morein G, Turcotte D. Forest fires: an example of self-organized critical behavior. Science. 1998;281(5384):1840–2. doi: 10.1126/science.281.5384.1840 [DOI] [PubMed] [Google Scholar]
51.Rochmyaningsih D. Wildfire researcher deported amid growing rift between Indonesian government and scientists. Science. 2020. doi: 10.1126/science.abb2763 [DOI] [Google Scholar]
52.Van Nieuwstadt MGL, Sheil D. Drought, fire and tree survival in a Borneo rain forest, East Kalimantan, Indonesia. J Ecol. 2004;93(1):191–201. doi: 10.1111/j.1365-2745.2004.00954.x [DOI] [Google Scholar]
53.Anda M, Ritung S, Suryani E, Hikmat M, Yatno E, et al. Revisiting tropical peatlands in Indonesia: Semi-detailed mapping, extent and depth distribution assessment. Geoderma. 2021;402:115235. doi: 10.1016/j.geoderma.2021.115235 [DOI] [Google Scholar]

PLoS One. doi: 10.1371/journal.pone.0331831.r001

Decision Letter 0

Bijeesh Veettil

22 Oct 2025

-->PONE-D-25-45724-->-->Nationwide Monthly Burned Area monitoring in Indonesia using Sentinel-2-->-->PLOS ONE

Dear Dr. Gaveau,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please submit your revised manuscript by Dec 06 2025 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:-->

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at . Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols....

We look forward to receiving your revised manuscript.

Kind regards,

Bijeesh Kozhikkodan Veettil

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Please note that PLOS One has specific guidelines on code sharing for submissions in which author-generated code underpins the findings in the manuscript. In these cases, we expect all author-generated code to be made available without restrictions upon publication of the work. Please review our guidelines at https://journals.plos.org/plosone/s/materials-and-software-sharing#loc-sharing-code and ensure that your code is shared in a way that follows best practice and facilitates reproducibility and reuse.

3. When completing the data availability statement of the submission form, you indicated that you will make your data available on acceptance. We strongly recommend all authors decide on a data sharing plan before acceptance, as the process can be lengthy and hold up publication timelines. Please note that, though access restrictions are acceptable now, your entire data will need to be made freely accessible if your manuscript is accepted for publication. This policy applies to all data except where public deposition would breach compliance with the protocol approved by your research ethics board. If you are unable to adhere to our open data policy, please kindly revise your statement to explain your reasoning and we will seek the editor's input on an exemption. Please be assured that, once you have provided your new statement, the assessment of your exemption will not hold up the peer review process.

4. We note that Figures 3, 4, 5, and 8 in your submission contain map/satellite images which may be copyrighted. All PLOS content is published under the Creative Commons Attribution License (CC BY 4.0), which means that the manuscript, images, and Supporting Information files will be freely available online, and any third party is permitted to access, download, copy, distribute, and use these materials in any way, even commercially, with proper attribution. For these reasons, we cannot publish previously copyrighted maps or satellite images created using proprietary data, such as Google software (Google Maps, Street View, and Earth). For more information, see our copyright guidelines: http://journals.plos.org/plosone/s/licenses-and-copyright.

We require you to either (1) present written permission from the copyright holder to publish these figures specifically under the CC BY 4.0 license, or (2) remove the figures from your submission:

1. You may seek permission from the original copyright holder of Figures 3, 4, 5, and 8 to publish the content specifically under the CC BY 4.0 license.

We recommend that you contact the original copyright holder with the Content Permission Form (http://journals.plos.org/plosone/s/file?id=7c09/content-permission-form.pdf) and the following text:

“I request permission for the open-access journal PLOS ONE to publish XXX under the Creative Commons Attribution License (CCAL) CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). Please be aware that this license allows unrestricted use and distribution, even commercially, by third parties. Please reply and provide explicit written permission to publish XXX under a CC BY license and complete the attached form.”

Please upload the completed Content Permission Form or other proof of granted permissions as an "Other" file with your submission.

In the figure caption of the copyrighted figure, please include the following text: “Reprinted from [ref] under a CC BY license, with permission from [name of publisher], original copyright [original copyright year].”

2. If you are unable to obtain permission from the original copyright holder to publish these figures under the CC BY 4.0 license or if the copyright holder’s requirements are incompatible with the CC BY 4.0 license, please either i) remove the figure or ii) supply a replacement figure that complies with the CC BY 4.0 license. Please check copyright information on all replacement figures and update the figure caption with source information. If applicable, please specify in the figure caption text when a figure is similar but not identical to the original image and is therefore for illustrative purposes only.

The following resources for replacing copyrighted map figures may be helpful:

USGS National Map Viewer (public domain): http://viewer.nationalmap.gov/viewer/

The Gateway to Astronaut Photography of Earth (public domain): http://eol.jsc.nasa.gov/sseop/clickmap/

Maps at the CIA (public domain): https://www.cia.gov/library/publications/the-world-factbook/index.html and https://www.cia.gov/library/publications/cia-maps-publications/index.html

NASA Earth Observatory (public domain): http://earthobservatory.nasa.gov/

Landsat: http://landsat.visibleearth.nasa.gov/

USGS EROS (Earth Resources Observatory and Science (EROS) Center) (public domain): http://eros.usgs.gov/#

Natural Earth (public domain): http://www.naturalearthdata.com/

If the reviewer comments include a recommendation to cite specific previously published works, please review and evaluate these publications to determine whether they are relevant and should be cited. There is no requirement to cite these works unless the editor has indicated otherwise.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

-->Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented. -->

Reviewer #1: Partly

Reviewer #2: Yes

**********

-->2. Has the statistical analysis been performed appropriately and rigorously? -->

Reviewer #1: No

Reviewer #2: Yes

**********

-->3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.-->requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.-->

Reviewer #1: No

Reviewer #2: Yes

**********

-->4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.-->

Reviewer #1: Yes

Reviewer #2: Yes

**********

-->5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)-->

Reviewer #1: 1.In remote sensing and computer science, algorithm often implies a novel mathematical or computational procedure.

In this manuscript, the approach is essentially a workflow combining preprocessing, machine learning, and rule-based filtering. The claim of developing a new “algorithm” is overstated, as most components rely on established methods and primarily represent an operational integration of existing techniques. A more accurate framing might be processing chain or workflow/ generic methodology rather than a novel algorithm.

2.If the authors claim this is the first automated nationwide monthly burned area algorithm, reviewers and readers will expect at least a description of how end users can interact with it. Without an interface or demonstration, the system risks being perceived as a research prototype rather than an operational monitoring tool. This is especially critical because the paper emphasizes policy and enforcement applications

3.The link between the ENSO (ONI) and IOD anomalies and fire peaks is described as correlation, but no statistical modeling is provided to quantify strength of influence.

Reviewer #2: The paper is well-written in a consistent manner and is easy to understand. The topic is interesting and very important for conserving forests and vegetation cover in tropical areas. I only have some recommendations, which I have provided below:

The abstract does not mention the methodology.

Check the text for language issues.

Line 67: Provide some references regarding the use of Landsat data in Indonesia.

Line 81: "with advanced machine learning algorithms" – State exactly which methods were used.

Line 91: "To date, no automated burned area detection system exists for Indonesia." I did a brief search and found several papers using AI for burnt area detection. How can the authors claim that their method is the only one published so far? I think there are some other papers published, using AI as an automated way of detecting burned areas in Indonesia.

Line 94: Explain the "FIRMS" dataset, as it might be new to some readers.

What is the difference between this paper and the previous one? (See line 79).

Figure 1 is very simple. The paper is fine without it as it is. Consider adding more detail or removing it, as the text is already sufficient.

The text uses many abbreviations without explaining them fully.

I don't understand why 6.25 hectares was selected as the threshold. This is already a large area (250m x 250m). The authors have stated that not removing fragmented burned areas is a main advantage of this paper. However, many of these areas could be false positives or negatives. What is the impact on the final results if these patches are removed? Is this done only for comparison with other products?

In section 3.2, please explain what happens in the different phases of each oscillation regarding precipitation, temperature, etc., and how these factors affect burned areas.

**********

-->6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our For information about this choice, including consent withdrawal, please see our For information about this choice, including consent withdrawal, please see our For information about this choice, including consent withdrawal, please see our Privacy Policy..-->..-->

Reviewer #1: No

Reviewer #2: Yes:Masoud Jafari ShalamzariMasoud Jafari ShalamzariMasoud Jafari ShalamzariMasoud Jafari Shalamzari

**********

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

To ensure your figures meet our technical requirements, please review our figure guidelines: https://journals.plos.org/plosone/s/figures

You may also use PLOS’s free figure tool, NAAS, to help you prepare publication quality figures: https://journals.plos.org/plosone/s/figures#loc-tools-for-figure-preparation.

NAAS will assess whether your figures meet our technical requirements by comparing each figure against our figure specifications.

PLoS One. 2026 Apr 8;21(4):e0331831. doi: 10.1371/journal.pone.0331831.r002

Author response to Decision Letter 1

29 Dec 2025

Response to Reviewers

Manuscript title: Automated monthly burned area mapping for Indonesia using Sentinel-2 and FIRMS data (2019–2024)

Manuscript ID: PONE-S-25-59671

Authors: D.L.A. Gaveau, A. Descals, M.A. Salim

Journal: PLOS ONE

General Response

We thank the Academic Editor and both reviewers for their constructive and insightful comments. We have revised the manuscript accordingly. Below we address each comment point-by-point. Reviewer comments are reproduced in italics, followed by our responses in plain text.

Reviewer #1

“In remote sensing and computer science, algorithm often implies a novel mathematical or computational procedure. In this manuscript, the approach is essentially a workflow combining preprocessing, machine learning, and rule-burned-area filtering… The claim of developing a new ‘algorithm’ is overstated.”

Response:

We appreciate this clarification. We agree that the term workflow or processing chain more accurately describes our method. We have revised the manuscript throughout to reduce the emphasis on “new algorithm” and instead describe it as an “automated monthly burned-area processing chain” built from established methods — Sentinel-2 pre/post-fire compositing, Random Forest classification, and rule-burned-area filtering — integrated into a single operational workflow.

The novelty lies in (i) its automation at national scale, (ii) its monthly temporal resolution, and (iii) its adaptation to Indonesia’s landscape and policy requirements.

“If the authors claim this is the first automated nationwide monthly burned area algorithm, readers will expect at least a description of how end users can interact with it… Without an interface or demonstration, the system risks being perceived as a research prototype.”

Response:

We agree and have clarified in the revised manuscript how the system will be made accessible to end users. The automated workflow will being integrated into the Nusantara Atlas platform (www.nusantara-atlas.org), which already provides near-real-time deforestation and fire-alert data. We will publish the monthly burned-area layers on Nusantara Atlas. This will ensure operational accessibility for the public, NGOs and relevant agencies. To reflect this, the Abstract now includes:

“Monthly burn-scar updates (post-December 2024 to present) are available for viewing and analysis via Nusantara Atlas (www.nusantara-atlas.org), where the system operates as an open, operational platform for public use..”

The Discussion also now concludes with a paragraph explaining the implementation and access considerations in full detail: “To support wider adoption, we have integrated our burned-area maps into the Nusantara Atlas platform (www.nusantara-atlas.org), which already provides near-real-time deforestation and fire hotspots data. This integration will enable continuous updates of monthly burned-area maps and make them accessible for analysis by government agencies, NGOs, and the public.

“The link between ENSO (ONI) and IOD anomalies and fire peaks is described as correlation, but no statistical modeling is provided to quantify strength of influence.”

Response:

We thank the reviewer for this comment. We compared the ENSO (ONI) and IOD anomalies with the burned area monthly as an additional validation exercise to verify that the burned area seasonality aligns with the expected fire season patterns linked with ENSO and IOD and reported in previous studies. Our goal was not to model the statistical strength of influence but to confirm that the temporal variations in burned area are consistent with known large-scale climate drivers. We have clarified this point in the Discussion section. The revised text now explicitly notes that the analysis is descriptive.

We wrote: “We note that this analysis is descriptive and intended to illustrate temporal co-variation between climatic indices and monthly burned-area fluctuations, rather than to quantify causality. Future work will apply statistical modelling approaches incorporating additional climatic variables such as air temperature, rainfall, vapor pressure deficit, or climatic water deficit, known drivers of fire occurrence,to estimate the strength and causality of these climatic influences.”

Reviewer #2

“The abstract does not mention the methodology.”

Response:

We have revised the abstract to include a concise description of the method used to map burned areas:

“Our approach uses a Random Forest model that classifies Sentinel-2 imagery and uses FIRMS fire hotspots to reduce false positives. The resulting 20-m monthly burned-area maps cover the entire country from January 2019 to December 2024.”

“Check the text for language issues.”

Response:

The manuscript has been carefully read to correct minor grammatical and typographical issues, ensuring clear and standard English throughout.

“Line 67: Provide some references regarding the use of Landsat data in Indonesia.”

Response:

We have added the following reference to support the statement about Landsat’s use for burned-area mapping in Indonesia by the Indonesian government:

https://sipongi.gakkum.kehutanan.go.id/indikasi-luas-kebakaran

In this reference, it is written: “The area of forest and land fires is calculated burned-areas on Landsat-8 OLI/TIRS image analysis”

“Line 81: ‘with advanced machine learning algorithms’ – State exactly which methods were used.”

Response:

We have clarified that the workflow uses the Random Forest classifier implemented in Google Earth Engine. The sentence now reads: “a method that combines time-series Sentinel-2 imagery with a machine learning classification model (Random Forest), implemented within the Google Earth Engine.”

“Line 91: ‘To date, no automated burned area detection system exists for Indonesia.’ I found several papers using AI for burnt area detection… How can the authors claim that their method is the only one?”

Response:

We appreciate this correction; and have revised the sentence for accuracy.

We now state: “While previous studies have proposed promising methodologies for monthly burned-area mapping in Indonesia, using Sentinel-2 and other remote sensing inputs (Vetrita et al., 2025; Arjasakusuma et al., 2022), none have yet delivered an automated, nationwide, monthly burned-area mapping system that operates continuously.”

This distinguishes our contribution in terms of spatial coverage, temporal frequency, and operational automation.

“Line 94: Explain the ‘FIRMS’ dataset, as it might be new to some readers.”

Response:

We have added the following text that explains the FIRMS dataset:

We now say in the Introduction: “Our method integrates high-frequency Sentinel-2 time series (every 2–5 days) with FIRMS (Fire Information for Resource Management System) daily fire alerts (Giglio et al., 2016), which are derived from MODIS (Moderate Resolution Imaging Spectroradiometer) sensors onboard NASA’s Terra and Aqua satellites and the VIIRS (Visible Infrared Imaging Radiometer Suite) sensors onboard the Suomi NPP and NOAA-20 satellites (NASA, 2025). These instruments detect daily thermal infrared anomalies associated with active combustion. In this study, the resulting fire hotspot detections serve as both spatial and temporal filters for identifying candidate burned areas detected and delineated using Sentinel-2 imagery.”

“What is the difference between this paper and the previous one (line 79)?”

Response:

We now clarify this distinction in the Introduction section:

“Our earlier study (Gaveau et al., 2021) produced an annual burned-area product suitable for retrospective analysis, whereas the present study develops a monthly, automated system designed for near-real-time monitoring, enabling continuous national updates.”

“Figure 1 is very simple. The paper is fine without it as it is. Consider adding more detail or removing it.”

Response:

We have removed Figure 1.

“The text uses many abbreviations without explaining them fully.”

Response:

We have ensured that all abbreviations (e.g., GEE, FIRMS, NBR, ONI, IOD, etc.) are defined upon first use in the manuscript.

We replace the BA abbreviation with “burned-area”

“I don't understand why 6.25 hectares was selected as the threshold…”

Response:

We thank the reviewer for raising this point.

Our classification system can detect burn scars smaller than 6.25 ha; however, we adopted this threshold to align with the official minimum mapping unit (MMU) used by the Indonesian Ministry of Forestry for national burned-area assessments.

This ensures comparability with official datasets and legal frameworks. Smaller patches are filtered from the published layer to meet this national standard.

The revised manuscript now makes this clear in the Introduction section: “ The workflow generates monthly burn-scar maps at 20-meter resolution, with a minimum mapping unit of 6.25 hectares — aligned with the official minimum burn-scar size recognized by Indonesian authorities”

(MoEF): Indonesia’s Second Forest Reference Emission Level (FREL) Submission to the UNFCCC, Ministry of Environment and Forestry (MoEF), Jakarta, Indonesia, 2022. https://redd.unfccc.int/files/modified_2nd_frl_indonesia_20220529_clean.pdf

“In section 3.2, please explain what happens in the different phases of each oscillation regarding precipitation, temperature, etc., and how these factors affect burned areas.”

Response:

We have added a paragraph in Section 3.2 explaining why positive IOD and El Niño (ONI) phases typically increase fire risk, while negative phases reduce it.

We wrote in Section 3.2: “During positive IOD and El Niño phases, reduced convection and rainfall lead to drier and warmer conditions across much of Indonesia, increasing fuel flammability and extending the dry season. Conversely, La Niña and negative IOD phases enhance rainfall and humidity, reducing fire occurrence.”

This provides the climatic context for the observed burned-area variations. We note in the Discussion section that future work will apply statistical modelling approaches incorporating additional climatic variables such as air temperature, rainfall, vapor pressure deficit, or climatic water deficit, known drivers of fire occurrence (Descals et al., 2022), to estimate the causality of these climatic influences.

________________________________________

Data Availability

Reviewer #1 expressed concern that data were not fully available.

We have clarified in the Data Availability Statement that:

• All input datasets (Sentinel-2 SR, FIRMS, MODIS MCD64A1) are publicly accessible.

• The monthly burned-area outputs will be hosted openly via the Nusantara Atlas upon acceptance. Thus, we will include the link to the dataset in the accepted version.

________________________________________

We believe these revisions address all reviewer concerns and substantially improve the manuscript’s clarity, rigor, and accessibility.

Attachment

Submitted filename: Response to Reviewers.docx

pone.0331831.s002.docx^{(29.4KB, docx)}

PLoS One. doi: 10.1371/journal.pone.0331831.r003

Decision Letter 1

Bijeesh Veettil

22 Jan 2026

-->PONE-D-25-45724R1-->-->Nationwide Monthly Burned Area monitoring in Indonesia using Sentinel-2-->-->PLOS One

Dear Dr. Gaveau,

Please submit your revised manuscript by Mar 08 2026 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:-->

A letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

We look forward to receiving your revised manuscript.

Kind regards,

Bijeesh Kozhikkodan Veettil

Academic Editor

PLOS One

Journal Requirements:

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

-->Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.-->

Reviewer #1: All comments have been addressed

**********

-->2. Is the manuscript technically sound, and do the data support the conclusions?

Reviewer #1: No

**********

-->3. Has the statistical analysis been performed appropriately and rigorously? -->

Reviewer #1: Yes

**********

-->4. Have the authors made all data underlying the findings in their manuscript fully available?

Reviewer #1: No

**********

-->5. Is the manuscript presented in an intelligible fashion and written in standard English?

Reviewer #1: Yes

**********

-->6. Review Comments to the Author

Reviewer #1: The authors provide a detailed narrative of analytical steps using Sentinel-2 compositing, Random Forest classification, post-classification filtering, validation, and dissemination via a web platform. However, these steps are described as analytical procedures rather than as components of an automated processing system.

Specifically:

a. The manuscript does not clearly describe whether the processing chain runs automatically on a fixed schedule.

b. How data ingestion, processing, quality control, and output publishing are orchestrated

c. Whether human intervention is required at any stage

d. No workflow diagram or system architecture figure is presented to illustrate,the interaction between platforms (Google Earth Engine, FIRMS, external climate datasets), and data flow from raw inputs to final products, and The interface between analysis and the Nusantara Atlas website

e. The manuscript suggests that burned-area results are “made available” through the Nusantara Atlas website, but it remains unclear whether the website is dynamically connected to the processing chain, or whether results are manually exported and uploaded.

f. The authors rely on several independent systems (Google Earth Engine, FIRMS, and a web visualization platform), yet the manuscript does not explain how these components are programmatically integrated, and whether a central controller or pipeline manages execution across platforms.

To substantiate the claim of automation, the authors should:

� Explicitly describe the nature of the processing system

� Include a workflow or system architecture diagram

� Clearly distinguish between Analytical methods (classification, filtering, validation), and Operational automation (execution, updating, publishing).Without these additions, the manuscript primarily documents burned-area analysis and result dissemination via a website, rather than a demonstrably automated monitoring system.

**********

-->7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files.). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Reviewer #1: No

**********

To ensure your figures meet our technical requirements, please review our figure guidelines: https://journals.plos.org/plosone/s/figures

You may also use PLOS’s free figure tool, NAAS, to help you prepare publication quality figures: https://journals.plos.org/plosone/s/figures#loc-tools-for-figure-preparation.

NAAS will assess whether your figures meet our technical requirements by comparing each figure against our figure specifications.

PLoS One. 2026 Apr 8;21(4):e0331831. doi: 10.1371/journal.pone.0331831.r004

Author response to Decision Letter 2

4 Mar 2026

Response to Reviewers

Manuscript title: Nationwide Monthly Burned Area Monitoring in Indonesia Using Sentinel-2

Manuscript ID: PONE-D-25-45724R1

Authors: D.L.A. Gaveau, A. Descals, M.A. Salim

Journal: PLOS ONE

General Response

Reviewer #1

“The authors provide a detailed narrative of analytical steps… however these are described as analytical procedures rather than components of an automated processing system.”

Response:

We thank the reviewer for this insightful comment. We agree that the original manuscript did not sufficiently distinguish between the analytical methodology and the operational implementation of the processing chain. In response, we have revised the manuscript and added a dedicated description of the operational system as Supplementary Information.

We clarified the operational and automated nature of the processing chain. We added Supplementary Methods, describing execution, scheduling, orchestration, and system integration. We added Figure S1, illustrating the system architecture and data flow. We revised terminology throughout (e.g., “operational”, “largely automated”) for precision

“The manuscript does not clearly describe whether the processing chain runs automatically on a fixed schedule.”

Response:

We have now explicitly described the execution schedule in Supplementary Methods S1. The system operates on a fixed monthly schedule, automatically initiated on the 5th day of each month, allowing for Sentinel-2 data latency.

“How data ingestion, processing, quality control, and output publishing are orchestrated.”

Response:

Supplementary Methods S1 now describes the orchestration of:

• Automated Sentinel-2 data access within Google Earth Engine

• Automated FIRMS hotspot retrieval and ingestion

• Scripted GEE processing workflows

• Automated post-processing via Python scripts

• Dissemination through Nusantara Atlas

“ Whether human intervention is required at any stage.”

Response:

We have clarified that the workflow is largely automated, with human involvement limited to exception handling, such as system failures. No manual intervention is required during routine monthly execution.

“No workflow diagram or system architecture figure is presented”

Response:

We have added Figure S1, which presents a schematic of:

Data sources (Sentinel-2, FIRMS)

Google Earth Engine processing

Cloud storage

Post-processing

Nusantara Atlas dissemination

“ The manuscript suggests that burned-area results are “made available” through the Nusantara Atlas website, but it remains unclear whether the website is dynamically connected… or manually uploaded.”

Supplementary Methods S1 now clarifies that burned-area outputs are:

Exported automatically from GEE to Google Cloud Storage

Accessed programmatically by Nusantara Atlas services

We revised wording in the manuscript to state that products are made available via Nusantara Atlas, avoiding any implication of manual upload steps.

“The authors rely on several independent systems (Google Earth Engine, FIRMS, and a web visualization platform), yet the manuscript does not explain how components are programmatically integrated”

Supplementary Methods S1 now describes the integration pipeline, including:

Python-based scheduling and orchestration

Automated data exchange between GEE, cloud storage, and post-processing services

Programmatic linkage to Nusantara Atlas

We also clarify that Google Earth Engine serves as the core execution environment, with external scripts coordinating post-processing and dissemination.

“Explicitly describe the nature of the processing system

Response:

We have followed this recommendation by:

Separating analytical methodology (main manuscript)

Operational automation & architecture (Supplementary Methods)

Providing Figure S1 (workflow/system diagram)

We also revised terminology throughout the manuscript for clarity and accuracy.

Attachment

Submitted filename: Second Response to Reviewers.docx

pone.0331831.s003.docx^{(35.1KB, docx)}

PLoS One. doi: 10.1371/journal.pone.0331831.r005

Decision Letter 2

Bijeesh Veettil

9 Mar 2026

Nationwide Monthly Burned Area monitoring in Indonesia using Sentinel-2

PONE-D-25-45724R2

Dear Dr. Gaveau,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice will be generated when your article is formally accepted. Please note, if your institution has a publishing partnership with PLOS and your article meets the relevant criteria, all or part of your publication costs will be covered. Please make sure your user information is up-to-date by logging into Editorial Manager at Editorial Manager® and clicking the ‘Update My Information' link at the top of the page. For questions related to billing, please contact and clicking the ‘Update My Information' link at the top of the page. For questions related to billing, please contact and clicking the ‘Update My Information' link at the top of the page. For questions related to billing, please contact and clicking the ‘Update My Information' link at the top of the page. For questions related to billing, please contact billing support....

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Bijeesh Kozhikkodan Veettil

Academic Editor

PLOS One

Additional Editor Comments (optional):

Reviewers' comments:

PLoS One. doi: 10.1371/journal.pone.0331831.r006

Acceptance letter

Bijeesh Veettil

PONE-D-25-45724R2

PLOS One

Dear Dr. Gaveau,

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS One. Congratulations! Your manuscript is now being handed over to our production team.

At this stage, our production department will prepare your paper for publication. This includes ensuring the following:

* All references, tables, and figures are properly cited

* All relevant supporting information is included in the manuscript submission,

* There are no issues that prevent the paper from being properly typeset

You will receive further instructions from the production team, including instructions on how to review your proof when it is ready. Please keep in mind that we are working through a large volume of accepted articles, so please give us a few days to review your paper and let you know the next and final steps.

Lastly, if your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

You will receive an invoice from PLOS for your publication fee after your manuscript has reached the completed accept phase. If you receive an email requesting payment before acceptance or for any other service, this may be a phishing scheme. Learn how to identify phishing emails and protect your accounts at https://explore.plos.org/phishing.

If we can help with anything else, please email us at customercare@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Bijeesh Kozhikkodan Veettil

Academic Editor

PLOS One

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

S1 Text. Processing pipeline for Nusantara Atlas.

Description of the processing pipeline used to integrate the results into the Nusantara Atlas platform.

(DOCX)

pone.0331831.s001.docx^{(221.8KB, docx)}

Attachment

Submitted filename: Response to Reviewers.docx

pone.0331831.s002.docx^{(29.4KB, docx)}

Attachment

Submitted filename: Second Response to Reviewers.docx

pone.0331831.s003.docx^{(35.1KB, docx)}

Data Availability Statement

The data underlying the results presented in the study are available from Zenodo at: https://zenodo.org/records/19104282.

[pone.0331831.ref001] 1.Gritzo LAL. Mitigating social and economic impact of wildfires. Appl Energy Combust Sci. 2024;20:100285. https://www.sciencedirect.com/science/article/pii/S2666352X24000402?via%3Dihub [Google Scholar]

[pone.0331831.ref002] 2.Bowman DMJS, Balch J, Artaxo P, Bond WJ, Cochrane MA, D’Antonio CM, et al. The human dimension of fire regimes on Earth. J Biogeogr. 2011;38(12):2223–36. doi: 10.1111/j.1365-2699.2011.02595.x [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0331831.ref003] 3.Bowman DMJS, Balch JK, Artaxo P, Bond WJ, Carlson JM, Cochrane MA, et al. Fire in the Earth system. Science. 2009;324(5926):481–4. doi: 10.1126/science.1163886 [DOI] [PubMed] [Google Scholar]

[pone.0331831.ref004] 4.Tacconi L. Preventing fires and haze in Southeast Asia. Nature Clim Change. 2016;6(7):640–3. doi: 10.1038/nclimate3008 [DOI] [Google Scholar]

[pone.0331831.ref005] 5.Field RD, Van Der Werf GR, Shen SS. Human amplification of drought-induced biomass burning in Indonesia since 1960. Nature Geosci. 2009;2(3):185–8. [Google Scholar]

[pone.0331831.ref006] 6.Goldammer JG, Seibert B. The impact of droughts and forest fires on tropical lowland rain forest of East Kalimantan. In: Fire in the tropical biota: Ecosystem processes and global challenges. Springer; 1990. 11–31. [Google Scholar]

[pone.0331831.ref007] 7.Nurdiati S, Bukhari F, Julianto MT, Sopaheluwakan A, Aprilia M, Fajar I, et al. The impact of El Niño southern oscillation and Indian Ocean Dipole on the burned area in Indonesia. TAO. 2022;33(1). doi: 10.1007/s44195-022-00016-0 [DOI] [Google Scholar]

[pone.0331831.ref008] 8.Wooster M, Perry G, Zoumas A. Fire, drought and El Niño relationships on Borneo (Southeast Asia) in the pre-MODIS era (1980–2000). Biogeosciences. 2012;9(1):317–40. [Google Scholar]

[pone.0331831.ref009] 9.Gaveau DLA, Sloan S, Molidena E, Yaen H, Sheil D, Abram NK, et al. Four decades of forest persistence, clearance and logging on Borneo. PLoS One. 2014;9(7):e101654. doi: 10.1371/journal.pone.0101654 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0331831.ref010] 10.Tsujino R, Yumoto T, Kitamura S, Djamaluddin I, Darnaedi D. History of forest loss and degradation in Indonesia. Land Use Policy. 2016;57:335–47. doi: 10.1016/j.landusepol.2016.05.034 [DOI] [Google Scholar]

[pone.0331831.ref011] 11.Miettinen J, Shi C, Liew SC. Fire Distribution in Peninsular Malaysia, Sumatra and Borneo in 2015 with Special Emphasis on Peatland Fires. Environ Manage. 2017;60(4):747–57. doi: 10.1007/s00267-017-0911-7 [DOI] [PubMed] [Google Scholar]

[pone.0331831.ref012] 12.Page SE, Siegert F, Rieley JO, Boehm H-DV, Jaya A, Limin S. The amount of carbon released from peat and forest fires in Indonesia during 1997. Nature. 2002;420(6911):61–5. doi: 10.1038/nature01131 [DOI] [PubMed] [Google Scholar]

[pone.0331831.ref013] 13.Gaveau DLA, Pirard R, Salim MA, Tonoto P, Yaen H, Parks SA, et al. Overlapping land claims limit the use of satellites to monitor no‐deforestation commitments and no‐burning compliance. Conserv Letters. 2016;10(2):257–64. doi: 10.1111/conl.12256 [DOI] [Google Scholar]

[pone.0331831.ref014] 14.Dennis RA, Mayer J, Applegate G, Chokkalingam U, Colfer CJP, Kurniawan I, et al. Fire, People and Pixels: Linking Social Science and Remote Sensing to Understand Underlying Causes and Impacts of Fires in Indonesia. Hum Ecol. 2005;33(4):465–504. doi: 10.1007/s10745-005-5156-z [DOI] [Google Scholar]

[pone.0331831.ref015] 15.Purnomo H, Shantiko B, Sitorus S, Gunawan H, Achdiawan R, Kartodihardjo H, et al. Fire economy and actor network of forest and land fires in Indonesia. Forest Pol Econ. 2017;78:21–31. doi: 10.1016/j.forpol.2017.01.001 [DOI] [Google Scholar]

[pone.0331831.ref016] 16.Chuvieco E, Mouillot F, van der Werf GR, San Miguel J, Tanase M, Koutsias N, et al. Historical background and current developments for mapping burned area from satellite Earth observation. Remote Sens Environ. 2019;225:45–64. doi: 10.1016/j.rse.2019.02.013 [DOI] [Google Scholar]

[pone.0331831.ref017] 17.Giglio L, Boschetti L, Roy DP, Humber ML, Justice CO. The Collection 6 MODIS burned area mapping algorithm and product. Remote Sens Environ. 2018;217:72–85. doi: 10.1016/j.rse.2018.08.005 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0331831.ref018] 18.Lizundia-Loiola J, Franquesa M, Boettcher M, Kirches G, Pettinari ML, Chuvieco E. Operational implementation of the burned area component of the Copernicus Climate Change Service: from MODIS 250 m to OLCI 300 m data. Earth System Science Data Discuss. 2021;1–37. [Google Scholar]

[pone.0331831.ref019] 19.Lizundia-Loiola J, Otón G, Ramo R, Chuvieco E. A spatio-temporal active-fire clustering approach for global burned area mapping at 250 m from MODIS data. Remote Sens Environ. 2020;236:111493. doi: 10.1016/j.rse.2019.111493 [DOI] [Google Scholar]

[pone.0331831.ref020] 20.Ramo R, Roteta E, Bistinas I, van Wees D, Bastarrika A, Chuvieco E, et al. African burned area and fire carbon emissions are strongly impacted by small fires undetected by coarse resolution satellite data. Proc Natl Acad Sci U S A. 2021;118(9):e2011160118. doi: 10.1073/pnas.2011160118 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0331831.ref021] 21.Boschetti L, Roy DP, Justice CO, Humber ML. MODIS–Landsat fusion for large area 30 m burned area mapping. Remote Sens Environ. 2015;161:27–42. doi: 10.1016/j.rse.2015.01.022 [DOI] [Google Scholar]

[pone.0331831.ref022] 22.Roy DP, Huang H, Boschetti L, Giglio L, Yan L, Zhang HH, et al. Landsat-8 and Sentinel-2 burned area mapping - A combined sensor multi-temporal change detection approach. Remote Sens Environ. 2019;231:111254. doi: 10.1016/j.rse.2019.111254 [DOI] [Google Scholar]

[pone.0331831.ref023] 23.SiPongi. Forest and land fire monitoring system Jakarta, Indonesia. Ministry of Forestry, Government of Indonesia; 2025. https://sipongi.gakkum.kehutanan.go.id/indikasi-luas-kebakaran [Google Scholar]

[pone.0331831.ref024] 24.Afira N, Wijayanto AW. Mono-temporal and multi-temporal approaches for burnt area detection using Sentinel-2 satellite imagery (a case study of Rokan Hilir Regency, Indonesia). Ecological Informatics. 2022;69:101677. doi: 10.1016/j.ecoinf.2022.101677 [DOI] [Google Scholar]

[pone.0331831.ref025] 25.Arisanty D, Feindhi Ramadhan M, Angriani P, Muhaimin M, Nur Saputra A, Puji Hastuti K, et al. Utilizing sentinel-2 data for mapping burned areas in Banjarbaru Wetlands, South Kalimantan Province. Inter J Forestry Res. 2022;2022:1–12. doi: 10.1155/2022/7936392 [DOI] [Google Scholar]

[pone.0331831.ref026] 26.Arjasakusuma S, Kusuma SS, Vetrita Y, Prasasti I, Arief R. Monthly Burned-Area Mapping using Multi-Sensor Integration of Sentinel-1 and Sentinel-2 and machine learning: case Study of 2019’s fire events in South Sumatra Province, Indonesia. Remote Sens Appl Soc Environ. 2022;27:100790. doi: 10.1016/j.rsase.2022.100790 [DOI] [Google Scholar]

[pone.0331831.ref027] 27.Indah LA, Lalita L Dyah, Indah FH, Harwahyu R, Fitri SR. Machine learning approaches for burned area identification using Sentinel-2 in Central Kalimantan. J Appl Eng Science. 2020;18(2):207–15. doi: 10.5937/jaes18-25495 [DOI] [Google Scholar]

[pone.0331831.ref028] 28.Hawbaker TJ, Vanderhoof MK, Beal Y-J, Takacs JD, Schmidt GL, Falgout JT, et al. Mapping burned areas using dense time-series of Landsat data. Remote Sens Environ. 2017;198:504–22. doi: 10.1016/j.rse.2017.06.027 [DOI] [Google Scholar]

[pone.0331831.ref029] 29.Schultz MG, Heil A, Hoelzemann JJ, Spessa A, Thonicke K, Goldammer JG, et al. Global wildland fire emissions from 1960 to 2000. Global Biogeochemical Cycles. 2008;22(2). doi: 10.1029/2007gb003031 [DOI] [Google Scholar]

[pone.0331831.ref030] 30.Gaveau DLA, Descals A, Salim MA, Sheil D, Sloan S. Refined burned-area mapping protocol using Sentinel-2 data increases estimate of 2019 Indonesian burning. Earth Syst Sci Data. 2021;13(11):5353–68. doi: 10.5194/essd-13-5353-2021 [DOI] [Google Scholar]

[pone.0331831.ref031] 31.Gorelick N, Hancher M, Dixon M, Ilyushchenko S, Thau D, Moore R. Google Earth Engine: planetary-scale geospatial analysis for everyone. Remote Sens Environ. 2017;202:18–27. [Google Scholar]

[pone.0331831.ref032] 32.Vetrita Y, Albar I, Santoso I, Prasasti I, Kartika T, Usman AB, et al. Monthly mapping of Indonesia’s burned areas: implementation, history, techniques, and future directions. Inter J Remote Sens. 2024;46(2):636–60. doi: 10.1080/01431161.2024.2421942 [DOI] [Google Scholar]

[pone.0331831.ref033] 33.Giglio L, Schroeder W, Justice CO. The collection 6 MODIS active fire detection algorithm and fire products. Remote Sens Environ. 2016;178:31–41. doi: 10.1016/j.rse.2016.02.054 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0331831.ref034] 34.NASA. From MODIS to VIIRS: continuing the legacy. 2025. https://www.earthdata.nasa.gov/news/blog/from-modis-viirs-continuing-legacy [Google Scholar]

[pone.0331831.ref035] 35.Ministry of Environment and Forestry. Indonesia’s Second Forest Reference Emission Level (FREL) Submission to the UNFCCC. Jakarta, Indonesia: Ministry of Environment and Forestry (MoEF); 2022. [Google Scholar]

[pone.0331831.ref036] 36.Liu S, Zheng Y, Dalponte M, Tong X. A novel fire index-based burned area change detection approach using Landsat-8 OLI data. European Journal of Remote Sensing. 2020;53(1):104–12. doi: 10.1080/22797254.2020.1738900 [DOI] [Google Scholar]

[pone.0331831.ref037] 37.Fletcher K. Sentinel 2: ESA’s Optical High-Resolution Mission for GMES Operational Services. European Space Agency; 2012. [Google Scholar]

[pone.0331831.ref038] 38.Breiman L. Random forests. Machine Learn. 2001;45(1):5–32. doi: 10.1023/a:1010933404324 [DOI] [Google Scholar]

[pone.0331831.ref039] 39.Schroeder W, Oliva P, Giglio L, Csiszar IA. The New VIIRS 375 m active fire detection data product: algorithm description and initial assessment. Remote Sens Environ. 2014;143:85–96. doi: 10.1016/j.rse.2013.12.008 [DOI] [Google Scholar]

[pone.0331831.ref040] 40.Olofsson P, Foody GM, Herold M, Stehman SV, Woodcock CE, Wulder MA. Good practices for estimating area and assessing accuracy of land change. Remote Sens Environ. 2014;148:42–57. doi: 10.1016/j.rse.2014.02.015 [DOI] [Google Scholar]

[pone.0331831.ref041] 41.Gaveau DLA, Locatelli B, Salim MA, Husnayaen, Manurung T, Descals A, et al. Slowing deforestation in Indonesia follows declining oil palm expansion and lower oil prices. PLoS One. 2022;17(3):e0266178. doi: 10.1371/journal.pone.0266178 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0331831.ref042] 42.Descals A, Gaveau DLA, Verger A, Sheil D, Naito D, Peñuelas J. Unprecedented fire activity above the Arctic Circle linked to rising temperatures. Science. 2022;378(6619):532–7. doi: 10.1126/science.abn9768 [DOI] [PubMed] [Google Scholar]

[pone.0331831.ref043] 43.Sloan S, Tacconi L, Cattau ME. Fire prevention in managed landscapes: recent success and challenges in Indonesia. Mitigation and adaptation strategies for global change. 2021;26:1–30. [Google Scholar]

[pone.0331831.ref044] 44.Du Y, Zhang Y, Zhang L, Tozuka T, Ng B, Cai W. Thermocline warming induced extreme indian ocean dipole in 2019. Geophy Res Let. 2020;47(18). doi: 10.1029/2020gl090079 [DOI] [Google Scholar]

[pone.0331831.ref045] 45.Zhang L, Han W, Hu Z-Z. Inter-basin and Multi-time Scale Interactions in generating the 2019 Extreme Indian Ocean Dipole. J Clim. 2021;1–39. doi: 10.1175/jcli-d-20-0760.1 [DOI] [Google Scholar]

[pone.0331831.ref046] 46.Goldman E, Carter S, Sims M. Fires drove record-breaking tropical forest loss in 2024. World Resources Institute; 2025. https://gfr.wri.org/latest-analysis-deforestation-trends [Google Scholar]

[pone.0331831.ref047] 47.Roteta E, Bastarrika A, Padilla M, Storm T, Chuvieco E. Development of a Sentinel-2 burned area algorithm: Generation of a small fire database for sub-Saharan Africa. Remote Sensing of Environ. 2019;222:1–17. doi: 10.1016/j.rse.2018.12.011 [DOI] [Google Scholar]

[pone.0331831.ref048] 48.Karsai I, Schmickl T, Kampis G. Forest Fires: fire management and the power law. In: Resilience and stability of ecological and social systems. Springer International Publishing; 2020. 63–77. doi: 10.1007/978-3-030-54560-4_4 [DOI] [Google Scholar]

[pone.0331831.ref049] 49.Falk DA, Miller C, McKenzie D, Black AE. Cross-Scale Analysis of Fire Regimes. Ecosystems. 2007;10(5):809–23. doi: 10.1007/s10021-007-9070-7 [DOI] [Google Scholar]

[pone.0331831.ref050] 50.Malamud B, Morein G, Turcotte D. Forest fires: an example of self-organized critical behavior. Science. 1998;281(5384):1840–2. doi: 10.1126/science.281.5384.1840 [DOI] [PubMed] [Google Scholar]

[pone.0331831.ref051] 51.Rochmyaningsih D. Wildfire researcher deported amid growing rift between Indonesian government and scientists. Science. 2020. doi: 10.1126/science.abb2763 [DOI] [Google Scholar]

[pone.0331831.ref052] 52.Van Nieuwstadt MGL, Sheil D. Drought, fire and tree survival in a Borneo rain forest, East Kalimantan, Indonesia. J Ecol. 2004;93(1):191–201. doi: 10.1111/j.1365-2745.2004.00954.x [DOI] [Google Scholar]

[pone.0331831.ref053] 53.Anda M, Ritung S, Suryani E, Hikmat M, Yatno E, et al. Revisiting tropical peatlands in Indonesia: Semi-detailed mapping, extent and depth distribution assessment. Geoderma. 2021;402:115235. doi: 10.1016/j.geoderma.2021.115235 [DOI] [Google Scholar]

PERMALINK

Nationwide monthly burned area monitoring in Indonesia using Sentinel-2

David L A Gaveau

Adrià Descals

Mohammad A Salim

Roles

Abstract

Introduction

Methods

Summary of methods

Sentinel-2 pre- and post-fire Sentinel-2 monthly composites

Fig 1. Example of a Sentinel-2 Normalized Burn Ratio (NBR) time series.

Fig 2. Monthly versus annual Sentinel-2 pre- and post-fire composites over an area in Central Kalimantan.

Supervised burned/unburned classification

Fig 3. Location of training and validation points labelled as ‘Burned’ and ‘Unburned’.

Post classification procedure

Masking of small patches.

Morphological operations.

Masking false positives with FIRMS daily fire hotspots.

Masking repeated burning detection.

Fig 4. Persistence of burn scar visibility leading to repeated monthly detection by Sentinel-2 imagery.

Burned-area map validation

Fig 5. Illustrative representation of validation points used for the monthly burned-area layer.

Results

Burned-area detection

Fig 6. Area burned by number of times between 2019 and 2024.

Fig 7. Spatial distribution of the burned area fraction (2019–2024) across Indonesia, mapped at a 1 km resolution.

Fig 8. Total burned area by region (left) and province (right) in Indonesia from 2019 to 2024.

Fig 9. Indonesia-wide cumulative burned area (in million hectares, Mha) by month for years 2019 to 2024.

Fire activity and climatic anomalies

Fig 10. Monthly burned area (top panel) and climate indices (bottom panel) for Indonesia from January 2019 to December 2024.

Accuracy assessment

Fig 11. Confusion matrix of burn-month detection from processing chain versus visual interpretation.

Burn size comparison

Fig 12. Cumulative burned area as a function of burned patch size for each year from 2019 to 2024.

Table 2. Tests statistics with respect to differences in burned-area scar size frequency distributions for Sentinel, MODIS, and official maps.

Discussion

Supporting information

Data Availability

Funding Statement

References

Decision Letter 0

Bijeesh Veettil

Roles

Author response to Decision Letter 1

Decision Letter 1

Bijeesh Veettil

Roles

Author response to Decision Letter 2

Decision Letter 2

Bijeesh Veettil

Roles

Acceptance letter

Bijeesh Veettil

Roles

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases