Title: Fires in the Amazon: An Analysis of NOAA-12 Satellite Data Source: Environmental Defense Fund Status: Distribute freely with accreditation Date: December 1, 1997 ENVIRONMENTAL DEFENSE FUND 1875 Connecticut Ave., NW,10th Fl. Washington, D.C. 20009 Telephone: (202) 387-3500 Facsimile: (202) 234-6049; firstname.lastname@example.org Fires in the Amazon: an analysis of NOAA-12 satellite data, 1996-1997. Stephan Schwartzman December 1, 1997
The number of fires in the Brazilian Amazon between July and November increased over 50% between 1996 and 1997. The NOAA-12 satellite recorded 29,571 fires in the Amazon region on 136 days between July 1, 1996 and November 30, 1996 and 44,734 fires on 118 days between July 1, 1997 and November 22, 1997, an increase of over 50% from 1996 to 1997, even though data are available for fewer days in 1997 than in 1996. The average number of fires per day increased 75%, from 217 in 1996, to 379 in 1997. A previous analysis, based on a more limited sample earlier in the year, had shown a smaller increase.1
The data are generated by the Advanced Very High Resolution Radiometer (AVHRR) on the NOAA-12 weather satellite, which detects thermal anomalies, and passes over the Amazon daily. Fires are mapped and counted by the National Institute for Space Research (INPE) in Brazil (http://condor.dsa.inpe.br/mapas_que).
The largest differences between the two years occurred in November and October, and result from increased economic activity, particularly burning of cattle pasture. The difference also reflects the extended dry season of 1997 caused by El Nino. Normally seasonal rains start in late September or early October in most of the Amazon, curtailing fires. 5/ In 1997, airports were still closing because of thick haze in November. The satellite recorded 2,638 fires in 22 days in November 1997, as opposed to 1,542 in 27 days in November 1996, an increase of 71%, over fewer days. In October 1997, 10,305 fires appear in 28 days, over three times more than the 3,119 counted for 26 days in October 1996.
The actual number of fires in the Amazon in both years is considerably higher than the totals obtained by the NOAA-12 satellite, for two reasons. The NOAA satellites, because of their trajectories and the locations of current receiving stations, cover the northern and western Amazon poorly. In addition, the NOAA-12 satellite passes over the region at night, when the number of fires is lower than during the day. INPE stopped analyzing NOAA-14 images, taken during the day, for the burning season of 1996, arguing that solar reflection on hot days could be confused with fires by the satellite's sensors and inflate the number of fires. While the NOAA-12 images thus under-count fires, comparison of data sets from different years does show changes in the level of burning.
New research from the region strongly suggests that fires themselves are rapidly becoming at least as great a threat to the biological integrity of the Amazon as is deforestation, as well as increasing Brazil's contribution to global CO2 and other greenhouse gas emissions. Fires are set in the Amazon to burn off cleared primary forest, and also to burn old cattle pastures and secondary forest areas. Deforestation per se accounts for only a relatively small part of the fires every year. Some 70% of the fires burn on land already deforested.2
The Woods Hole Research Center and the Institute for Amazonian Environmental Research (IPAM) have shown that selective logging and ground fires-fires that burn largely undetected by the satellites, beneath the forest canopy-are degrading an area approximately equal to the area deforested annually in recent years. Selective logging, as studies by the Institute for Man and Nature in the Amazon (IMAZON) show, contributes to the flammability of the forest through opening up the canopy and leaving combustible material behind.3 Ground fires, often in previously logged areas or areas bordering already deforested lands, in conjunction with dry weather, are making the forest dryer. The increased burning this year means that ground fires, which may cover hundreds or even thousands of square kilometers, also increased, even though they do not appear in the satellite images. Deforestation, according to INPE's last figures (for 1994), was about 15,000 square kilometers a year. This means that a similar area, unrecorded by satellite images, is being degraded through selective logging and ground fires annually.
The Woods Hole, IMAZON and other new findings indicate that CO2 emissions and other global climatic effects of Amazon fires have heretofore been underestimated, by as much as 30%. Recent long-term research on forest fragments in the Amazon shows that up to 36% of biomass is lost in fragments within 100 meters of edges in the first 10-17 years after fragmentation. The authors conclude that decline in biomass in forest fragments could be a significant, and uncounted, source of greenhouse gases such as CO2.4
The Woods Hole Research Center/IPAM research on fires has identified an alarming new trend. Much of the forest of the eastern and southern Amazon, which depends on deep-soil water reserves to stay green in the dry season, is becoming flammable because of logging and drought. Hitherto, virgin forest has prevented the spread of fires because it was too moist to burn. Should large parts of the intact forest dry out enough to burn, as appears to be occurring, much quicker and larger scale destruction of the forest becomes possible, in a vicious circle of drying - larger fires - more drying. The Woods Hole group set an experimental fire in intact closed forest in Par? state for the first time this year. These results show that the rate of deforestation of formerly intact primary forest, as measured by analysis of Landsat images - formerly considered the central indicator of of forest destruction -- is no longer the only significant, or even the most urgent, threat to the forest. Should intact closed forest begin to burn, a previously incremental process (the loss of 0.4%, or 0.5% of the forested area of the Amazon to deforestation yearly, as was the case in the 1980s and 1990s) could become a catastrophic positive feedback loop. Climate models predict a slightly drier climate in tropical areas under global warming.
While increased burning involves hundreds of thousands of acres spread across a continental region, much can be done to address the problem. One half of the area burned in 1994 and 1995 resulted from accidental fires. These fires have substantial costs for small and large farmers alike and benefit no one. Efforts to assist rural Amazonians to prevent accidental fires (through fire breaks or enforcing compensation for fires that damage others' property), and to rely less on the use of fire for agriculture (through mechanization) would make a difference. In addition, passage of the Environmental Crimes Act, currently stalled in the Brazilian House of Representatives, would give the Brazilian environmental agency, IBAMA, statutory authority to enforce the law, including restrictions on burning and deforestation, for the first time since 1989.
Whether or not deforestation rates have increased in the Amazon will only be known with the release of INPE's analysis of Landsat images. INPE has promised to release data for 1995 and 1996 by end of the year. Increased burning, and new research results on the effects of fire, however, unequivocally demonstrate that the rate of deforestation is no longer the only important indicator of threat to the biological integrity of the Amazon forest. Under current conditions of drought stress, fire itself may rapidly become the vector of greater and much quicker destruction than previously imagined possible, with potentially enormous global repercussions.
Summary of Analysis for 1996 July 1740 Aug 10293 Sep 12877 Oct 3119 Nov 1542 Actual Fires Counted 29571 Number of Days in Period 153 Data Days Available 136 Average No. Counted per day 217 Summary of Analysis for 1997 July 2453 Aug 14986 Sep 14352 Oct 10305 Nov 2638 Actual Fires Counted 44734 Number of Days in Period 153 Data Days Available 118 Average No. Counted per day 379
Note: Daily fire totals broken down by state are available on request for July-November for 1996 and 1997.