Technical Comments

Response to Comment on “The Incidence of Fire in Amazonian Forests with Implications for REDD”

See allHide authors and affiliations

Science  17 Dec 2010:
Vol. 330, Issue 6011, pp. 1627
DOI: 10.1126/science.1195063

Abstract

Balch et al. suggest that the increased fire frequency reported in our study is principally due to post-deforestation activities. We present a new analysis demonstrating that for the majority of grid cells with positive fire trends, there is a low likelihood that these trends have resulted exclusively from post-deforestation activities. We therefore confirm that fires pose a growing threat to reducing emissions from deforestation and degradation (REDD) policies.

Balch et al. (1) challenge our conclusion (2) that the widespread increase in fire incidence across Amazonia, which has occurred despite a reduction in deforestation rates, poses an emerging threat to the efficacy of reducing emissions from deforestation and degradation (REDD) policy. We address their objections and present new evidence to support our conclusions.

The criticisms of Balch et al. (1) arise from an inaccurate interpretation of our conclusions. We did not claim that post-deforestation fires reverse emission reductions achieved through REDD but, instead, that there is an increased risk of carbon losses resulting from increased fire frequency, even with the reduction in deforestation rates, because of ongoing enlargement of forest edges, secondary forest cover, and fragmentation, favoring the leakage of accidental fires from farms. Balch et al. appear to suggest that the observed fire trends are the product of post-deforestation activities alone. Their proposition is grounded on two studies (3, 4) that have not evaluated this issue. Our conclusions are based on the fact that active fires data, in addition to post-deforestation fires, capture high-energy fires in forest edges, degraded forests, and forest crown (5). Active fire data are likely to underestimate understory fires and, therefore, our conclusions are rather conservative (2). Agreeing on methods for monitoring, reporting, and verifying emissions (MRV) from these areas within the REDD framework is thus urgently required.

Balch et al. do raise an interesting question about the extent to which active fire detections capture fires other than post-deforestation fires. In response, we have performed a further evaluation of the patterns found in our study (2). Despite our inability to detect fire types, if it is assumed that active fire detections are proportional to the biomass available to burn (fuel loads) in a given location (6), then the temporal trends in fuel loads, after successive deforestation, incomplete combustion, and vegetation regrowth in a grid cell, can provide an estimation of the likelihood of post-deforestation fire trajectories in that grid cell.

To verify whether post-deforestation activities can generate positive trends in fuel loads and, hence, sustain an increase in fire incidence along an 8-year time frame, we conducted 250 simulations (four experiments, Fig. 1) based on 110 deforestation scenarios with negative temporal trends. For each scenario, annual fractional deforestation in a 5 × 5 grid cell (total of 25 subcells) was randomly assigned following a normal distribution. The simulations included five levels of combustion completeness (CC), three levels of fractional area deforested (Af), and two levels of biomass regrowth rate (Rb). CC was calculated as a proportion of the initial biomass, which ranges from 19.5% to 61.5% in Amazonia (7). For each deforestation event, we assumed an average initial biomass (Bi) of 300 t ha−1. The biomass remaining at each time interval t (Br,t) was calculated as Br,t = Br,t-1 – (Bi × CC) + Rb. Cumulative annual fuel load was estimated as the sum of Br,t in each of the 25 subcells. Trend slopes were calculated using normalized values of deforestation, fuel loads, and fire. The likelihood of obtaining positive fuel load (simulations) and fire (data) slopes was crudely estimate as a proportion of all simulation results in each class of deforestation slopes (Fig. 1).

Fig. 1

Scatter plot of trend slopes of fire and fuel load across the range of deforestation. All slopes presented in the figure were calculated using normalized values of deforestation, fuel loads, and fire. Yellow diamonds correspond to the slopes derived from the original fire (AVHRR/NOAA-12) and deforestation (PRODES) data presented in (2). Different colors represent fuel load slopes for three simulations (experiments), based on 10 random deforestation scenarios, using a regrowth rate Rb = 15 t ha−1 year−1 (16), and with varying levels of fractional area deforested (Af). Experiment 1 assumes Af = 20% (black), experiment 2 assumes Af = 40% (green), and experiment 3 assumes Af = 100% (blue). Symbols represent the five levels of CC used in each experiment: CC = 20% (open triangle), CC = 30% (open circle), CC = 40% (open square), CC = 50% (open diamonds), and CC = 60% (asterisks). Experiment 4 is the most likely scenario (red diamonds) for fuel load calculations. It is based on 100 random deforestation scenarios. For experiment 4, fuel loads were calculated assuming CC = 40% [approximate mean value of values reported for Amazonia (7)], Af = 40% (encompass 96% of all values for annual fractional area deforested in Amazonia) and Rb = 8 t ha−1 yr−1 [average net primary production for grasslands in Amazonia (7)]. Lines indicate the likelihood (percentage of all simulation results in each class of deforestation slopes) of obtaining positive fire (original data, blue line) and fuel loads (simulations, red line) slopes in each category of deforestation slopes (0.0 to –0.5, –0.5 to –1.0, –1.0 to –1.5, –1.5 to –2.0, –2.0 to –3.0, –3.0 to –4.0, –4.0 to –5.0, –5.0 to –6.0, –6.0 to –7.0, –7.0 to –8.0, and –8.0 to –9.0). Categories match with the vertical grid lines.

Considering all simulations, 57% of the grid cells with positive fire slope fall in a region with less than 30% likelihood of positive fuel load trend. This value increases to 77% considering the most likely scenario alone (experiment 4). The likelihood of extra biomass burning from secondary forests, adjacent forest edges, and disturbed forests is therefore high for the majority of the grid cells. Conversely, as deforestation slopes approach zero, fires associated exclusively with clearing and management are more likely to occur. Other than emissions from primary forest clearance and maintenance fires, none of the above processes are fully considered in previous basin-wide net carbon emission estimates (711).

Balch et al. highlight the contribution and difficulties in reporting understory fires emissions under REDD. This was also a central point of our discussion (2). Emissions from understory forest fires [0.01 to 0.2 Gt C year−1 (9, 12)] and long-term changes in fire-affected forest carbon pools (13) are likely to be unaccounted for, not only because of the lack of agreement on methods to quantify and monitor these events (14) but also if the MRV processes mirror the guidelines on land use, land-cover change, and forestry used by Annex I countries (15).

There is a high probability that fires other than those from post-deforestation activities have an important contribution to the increased fire incidence in Amazonia, confirming our conclusion that REDD benefits may be undermined by the increased risk of unaccounted fire emissions. We hope that this exchange advances the discussions on tropical forest fires and spurs urgent actions to place REDD as an efficient climate change mitigation policy.

References and Notes

  1. We thank the U.K. Natural Environment Research Council for providing funds to L. E. O. C. Aragão (NE/F015356/2).
View Abstract

Navigate This Article