Figure 2. A snapshot over Riau showing the areas burned by the June 2013 fires (red) mapped using LANDSAT 8 imagery acquired on 25 June 2013 (background) with NASA’s fire alerts (yellow dots) detected between 1 and 30 June 2013.

BOGOR, Indonesia (2 July 2013)_CIFOR has made an analysis of new satellite imagery for the area in Riau Province, Sumatra, which appears to have been worst affected by recent fires causing haze problems over Sumatra, Singapore and Malaysia. Although several recent assessments have used NASA’s daily fire alerts to locate the fires, we have additionally used higher-resolution imagery from the recently launched Landsat 8 satellite to map fire scars. The Landsat images were recorded on 25 June 2013.

While our analysis is preliminary, it raises a number of interesting and important questions and insights about the location, context and causes of the fire incidents.

We would like to share these observations and associated hypotheses below, and we encourage your feedback.

KEY OBSERVATIONS
1. There was a distinct peak in NASA’s daily fire alerts during a very short period of time, between 17 and 25 June (Figure 1).

Figure 1. NASA’s daily fire alerts in Sumatra during the month of June 2013, showing a peak in fire activity between 17 and 25 June.

2. We found a strong spatial correspondence between fire alert locations and observed burned areas in the Landsat 8 imagery of 25 June (Figure 2). A total of 100,000 hectares (1,000 km²) of fire scars were mapped within the worst-affected Landsat scene, corresponding to 3% of the total area studied.

Figure 2. A snapshot over Riau showing the areas burned by the June 2013 fires (red) mapped using LANDSAT 8 imagery acquired on 25 June 2013 (background) with NASA’s fire alerts (yellow dots) detected between 1 and 30 June 2013.

3. A very high proportion of the fire scars are on peatland, as opposed to mineral soil (Figure 3).

Figure 3. The 100,000 ha area mapped as burned (red) within the worst-affected LANDSAT scene (black box). NASA’s fire alerts are marked with yellow points. Not all burned areas are indicated due to cloud and haze cover and missing imagery. Most fires are located on peat soils (brown areas).

4. Fire scars are predominantly observed in areas of established plantation land use, both large and small scale. Most June 2013 fires in the studied area have occurred outside natural forests (Figure 4). However, some of the fires appear to have advanced from plantations into adjacent natural forest.

Figure 4. Four LANDSAT 8 snapshots acquired on 25 June 2013 over Riau revealing areas that have been recently burned (dark purple areas). Smoke plumes, indicating active fires, can still be seen in the snapshots to the left. The rectangular features of the burned areas, located in the middle of plantations (grid-like and rectangular patterns), suggest that these fires have been lit as part of the management of existing plantations.

5. Fires occurred both on acacia and oil palm plantation sites with little or no tree cover and on sites that carried standing plantation trees (Figure 5).

Figure 5. (A) A snapshot over an acacia plantation two months before the fires (22 April 2013). The grid-like rectangular pattern reveals an industrial plantation. The green areas reveal mature acacia trees. The brown areas reveal bare soils. The yellow outline shows the area that has burned between 17 and 25 June 2013. (B) The same area shortly after the fires (25 June 2013). The burned area (dark purple with yellow outline) encompasses areas that were previously bare soil and mature acacia trees.

6. Observed fire scars were both inside and outside boundaries of concession areas, as determined from available official maps (Figure 6). Some fire scars outside of these concession areas contain patterns that indicate plantation establishment.

Figure 6. A LANDSAT 8 snapshot acquired on 25 June 2013 revealing existing industrial plantations (grid-like rectangular patterns) outside of the publicly available concession map (red outline: oil palm; blue outline: acacia).

7. Many of the June 2013 fire scars are in areas that were classified as natural forest in 2007 (Figure 7).

Figure 7. Areas that burned in June 2013 (red) and natural forest cover in 2007 (green).

HYPOTHESES
1. Many of the June 2013 fires are part of the processes of plantation establishment and management. The very short period over which fire incidents peaked, the high proportion of fires occurring on peatlands, typical patterns of plantation management in fire areas, and the lack of updated concession maps support this hypothesis.

2. Weather conditions (including wind patterns) exacerbated the haze problem in June 2013 compared with previous fire incidents.

ACKNOWLEDGEMENTS
The very rapid availability of Landsat 8 data allowed us to make a timely assessment of fire scars and correlate the data with known land uses. However, more field data are needed to confirm the precise events and processes that caused the fires.

For more information on the issues discussed in this article, please contact David Gaveau at d.gaveau@cgiar.org or Mohammad Agus Salim at m.salim@cgiar.org.

This research is carried out as part of the CGIAR Research Program on Forests, Trees and Agroforestry.