logo
ResearchBunny Logo
Near-complete loss of fire-resistant primary tropical forest cover in Sumatra and Kalimantan

Environmental Studies and Forestry

Near-complete loss of fire-resistant primary tropical forest cover in Sumatra and Kalimantan

T. Nikonovas, A. Spessa, et al.

Discover how deforestation in Indonesia has heightened fire risks, especially in peatlands, causing significant carbon emissions. This pivotal research by Tadas Nikonovas, Allan Spessa, Stefan H. Doerr, Gareth D. Clay, and Symon Mezbahuddin reveals crucial insights into fire resistance in primary forests and the urgent need for conservation efforts.

00:00
00:00
~3 min • Beginner • English
Introduction
Indonesia has experienced among the highest rates of commodity-driven primary forest loss over the past five decades, especially on the islands of Sumatra and Kalimantan, which host most of the country’s carbon-rich peatlands. Intact, closed-canopy tropical forests regulate microclimate, maintain understorey humidity, and buffer drought effects, historically keeping fire rare and preventing carbon release from peat. Deforestation and degradation alter surface energy balance, increase temperatures, dry microclimates, and, in peatlands, drainage lowers water tables and increases peat ignitability. Edge effects from fragmentation extend elevated fire risk kilometers into remaining forests. Anthropogenic ignitions combined with climate variability (El Niño and Indian Ocean Dipole phases) have produced recurrent severe fire episodes (e.g., 1997–1998 and 2015), converting peatlands from carbon sinks to sources. Despite recognition that deforestation raises fire risk, a region-wide assessment of how remaining primary forests resist fire, especially under drought, has been lacking. This study aims to quantify loss and fragmentation of primary forests and analyze associated changes in fire regimes across peatland and non-peatland areas of Sumatra and Kalimantan from 2001–2019 to identify thresholds relevant to forest function as fire barriers and peat carbon protection.
Literature Review
Prior work documents extensive primary forest loss in Indonesia and Southeast Asia and highlights peatlands’ immense long-term carbon stores and historical role as carbon sinks. Studies show intact tropical forests maintain cooler, wetter microclimates and are generally fire-resistant, whereas deforestation and land-use change (e.g., logging, oil palm and pulp plantations) increase local air temperatures, reduce shading, dry near-surface peat, and elevate fire risk. Fragmentation creates edge effects—including higher tree mortality, reduced reproduction, lower biomass, increased temperature and wind, and canopy desiccation—that extend kilometers into forests, further elevating fire susceptibility. In peatlands, drainage can lower water tables up to ~2 km into adjacent peat swamp forests, increasing ignitability. Anthropogenic ignitions are widespread, and regional fire activity is modulated by interannual climate variability (El Niño, IOD), with co-occurring strong positive phases producing the most severe droughts and fires. Recurrent burning and drainage have shifted many peatlands from sinks to sources of carbon. However, a comprehensive, spatially explicit assessment of the fire-resistance of remaining primary forests and the extent to which fragmentation compromises this resistance at regional scale has been missing.
Methodology
Study area: Sumatra and Kalimantan (total ~101.5 Mha), stratified into peatlands (~11.4 Mha) and non-peatlands. Primary forest extent and loss: Primary forest for year 2000 derived from Landsat-based primary forest cover dataset (pixels ~25 m; forest defined as trees ≥5 m height and >30% canopy cover; primary forest = old-growth natural forest blocks ≥5 ha without recent deforestation). Reported overall agreement 90.2% (Kappa 80%) against Indonesia’s 2000 primary forest map. Annual tree cover loss for 2001–2018 from Hansen et al. (v1.6), where loss is stand-replacing disturbance; producer’s accuracy for tropical regions ~83.1%. Loss pixels were intersected with 2000 primary forest to attribute annual primary forest loss. Annual primary forest extent at the beginning of each year computed by subtracting cumulative loss from year-2000 extent. Aggregation and classification: Primary forest pixels aggregated to 0.01° (~1 km) grid to compute primary forest percent cover per grid cell annually. Grid cells classified into four categories: undisturbed forest (>99% cover), partially deforested (50–<99%), mostly deforested (1–<50%), and fully deforested (<1%). The 99% threshold accommodates minor canopy changes and possible misclassification (up to 16 of 1600 25 m pixels) and yields ~20% larger undisturbed extent versus a strict 100% threshold with negligible effect on fire rates; lower thresholds inflated apparent fires in undisturbed forests. Distance to forest edge: For grid cells with >99% cover, distance to edge was computed by treating all cells with <99% cover as non-forest. Cells adjacent (8-connected) to non-forest considered within 1 km of edge; next ring as 1–2 km; interior >2 km. Fire data and filtering: Active fire detections from MODIS Collection 6 MCD14ML (1 km pixel at nadir; up to ~10 km2 at scan edge). Excluded low-confidence detections (<30%). Agglomerated individual detections within 3 km and 48 hours; retained only fire events observed on ≥2 days to focus on persistent burning. This reduced detections by 12% (peatlands) and 45% (non-peatlands), indicating more small, short-lived fires in non-peatlands. Fire-affected area defined as any 1 km grid cell with ≥1 filtered detection in a year. Peatland delineation: High-resolution peat distribution and carbon content maps (Wetlands International/WHC) rasterized to 1 km; cells whose center fell within peat polygons classified as peatland (resulting total peat extent ~10% smaller than source polygons). Climate context: Monthly precipitation anomalies from ECMWF ERA5 (0.25°), computed as departures from 2002–2019 climatological monthly means averaged over land grid cells in Sumatra and Kalimantan. Analyses: Quantified temporal changes in primary forest extent, fragmentation (% cover categories), and fire-affected proportions by category for peatlands and non-peatlands. For areas >99% forested in 2002, tracked transitions across categories and fire incidence, stratified by distance to forest edge (0–1 km, 1–2 km, >2 km).
Key Findings
- Primary forest extent declined markedly: from ~45 Mha (of 101.5 Mha) in 2001 to ~37 Mha by end-2018 (average annual primary forest loss ~0.43 Mha). - Non-peatlands: primary forest declined from ~39 Mha (2000) to ~33.6 Mha (2019). Peatlands (~11.4 Mha total): primary forest declined from ~6 Mha (53%) in 2000 to ~3.4 Mha (30%) by end-2018. - Deforestation peaked around 2012 and decreased thereafter; 2017–2018 levels were the lowest since 2001–2003, largely due to reduced peatland forest loss, coincident with policy interventions (e.g., peat restoration, moratoria). - Fragmentation increased substantially. In peatlands, area in undisturbed primary forest (>99% cover) decreased from ~32% of peatlands (2000) to ~16% (end-2018). Large increases occurred in partially to fully deforested categories in both peatlands and non-peatlands. - Fire occurrence strongly increased with declining primary forest cover: • Peatlands (annual mean % of grid cells fire-affected): undisturbed 0.9%; partially deforested 6.2%; mostly deforested 11%; fully deforested 8.3%. • Non-peatlands: undisturbed 0.09%; partially deforested 1.3%; mostly deforested 3.3%; fully deforested 2.1%. • In drought years, up to 22% (peatlands) and 8% (non-peatlands) of the mostly/fully deforested categories were fire-affected; maxima in undisturbed forests were 3% (peatlands) and 0.23% (non-peatlands). - Edge proximity governed rare fires inside undisturbed forests: ~94% (peatlands) and ~97% (non-peatlands) of fire-affected undisturbed forest cells were within 1 km of the forest edge; 1–2 km accounted for the remainder; >2 km interior accounted for <1% of total fire-affected area. - Fires did not penetrate undisturbed primary forest deeper than 2 km from the forest edge, effectively decoupling interior forests from regional drought–fire dynamics during 2001–2019. - By end-2018, undisturbed primary forests located >2 km from the edge covered only ~3% of peatlands and ~4.5% of non-peatlands. In peatlands, ~84% of area was under high fire risk (<99% cover) and an additional ~13% under increased risk (undisturbed but within 2 km of edge). Only ~10% of remaining peatland primary forests were in the ‘resilient to fire’ (>2 km interior) group. - Maximum distance from forest edge shrank from 11→8 km (peatlands) and 19→14 km (non-peatlands) over the study period, indicating diminishing interior refugia. - Recently deforested lands contributed substantially to major fire episodes: ~15% of fire-affected area in 2015 and ~17% in 2019 occurred on land deforested since 2002. - Peatlands, while ~11% of the land area, contributed a large share of smoke emissions during extreme fire episodes; overall, peatlands had on average 4.3× higher fire-affected proportions than non-peatlands, underscoring high vulnerability of peat carbon stocks.
Discussion
The analysis demonstrates that intact, closed-canopy primary forests in Sumatra and Kalimantan function as highly effective fire barriers, with interior areas (>2 km from edge) remaining virtually fire-free even during droughts. Conversely, deforestation and fragmentation greatly increase fire susceptibility through microclimate drying, edge effects, and human ignitions, with elevated fire incidence persisting long after initial clearing. Peatlands are particularly vulnerable: drainage and vegetation shifts toward flammable grasses and ferns amplify combustibility and sustain recurrent fires, converting long-standing carbon sinks into carbon sources and driving substantial smoke emissions with severe health and economic impacts. The findings quantify the extent to which loss of contiguous forest cover has expanded high fire-risk areas and degraded the protective function of forests, explaining the growth and persistence of fire activity over time. Policy measures implemented in recent years may be helping reduce peatland forest loss, but the small remaining interior primary forest area indicates that without urgent protection of contiguous forest blocks and peatland restoration, future climate-driven droughts could lead to more widespread fires and associated emissions.
Conclusion
The study provides a region-wide, satellite-based assessment showing that undisturbed primary forests resist fire beyond 2 km from edges, yet such interior forests now occupy only ~3% of peatlands and ~4.5% of non-peatlands in Sumatra and Kalimantan. Extensive loss and fragmentation of primary forest have substantially expanded high fire-risk areas, especially in peatlands, where only ~10% of remaining primary forest is in the ‘resilient to fire’ interior category. The authors conclude that protecting and regenerating remaining contiguous primary forests and restoring peatlands are urgently needed to mitigate increasingly frequent and severe fire events and safeguard carbon stocks, biodiversity, public health, and regional economies.
Limitations
- Reliance on remotely sensed products with known uncertainties: primary forest map (overall agreement 90.2%, Kappa 80%); annual tree cover loss (producer’s accuracy ~83.1% in the tropics); MODIS active fire detections (estimated ~8% false detection rate in South Asia and <10% omission for fires >0.125 km²). - Active fire–based proxy for fire-affected area may miss small or short-lived fires (despite filtering) and is sensitive to view geometry (pixel area varies off-nadir). - Peatland extent rasterization at 1 km resolution yields a total peat area ~10% smaller than the source polygons, potentially under/over-classifying edge cells. - Classification threshold for ‘undisturbed’ (>99% cover) increases the estimated area of undisturbed forests by ~20% versus a strict 100% threshold, though authors report negligible impact on fire occurrence rates for this category. - Annual primary forest extent represents conditions at the beginning of each year; intra-annual dynamics are not captured.
Listen, Learn & Level Up
Over 10,000 hours of research content in 25+ fields, available in 12+ languages.
No more digging through PDFs, just hit play and absorb the world's latest research in your language, on your time.
listen to research audio papers with researchbunny