Reply to comment by W. Schroeder et al. on “Reversal of trend of biomass burning in the Amazon”

[1] The fate of the Amazonian rainforest is critically important to the well being of the planet. When studying questions related to the health of the forest and especially to human-induced effects on this fragile ecosystem, we must take special care to insure proper reporting and analysis. The comments of Schroeder et al. [2009] contribute to a welcomed scientific dialogue on important topics affecting the Amazon. [2] Koren et al. [2007] (hereinafter referred to as K2007) analyzed seven biomass burning seasons of aerosol optical depth retrievals from the Moderate resolution Imaging Spectroradiometer (MODIS) satellite sensor and nine seasons of fire count observations from the Advanced Very High Resolution Radiometer (AVHRR) satellite sensor. The data showed a positive trend in both the fire counts and aerosol up until 2006 when a sharp reduction in both variables occurred. Although we acknowledged the fact that interannual variability in meteorology is important, we attributed the sharp reduction in the final year partially to human activities. In particular we suggested that part of the sharp decrease in the amount of fires and smoke in the dry season of 2006 can be attributed to actions done by a tri-national committee and a policy shift implemented by small rural farmers in 2006. [3] Schroeder et al. [2009] (hereinafter referred to as S2009) address two issues in K2007. (1) They question the AVHRR fire count data and the resulting trend shown in K2007. We note that S2009 never question the validity of the data or the trends of the MODIS aerosol optical depth. Only the AVHRR fire count record is criticized. (2) They also question our interpretation that the reduction of the biomass burning in 2006 was in part due to deliberate changes in human activity. [4] We will first address the comments on the fire counts and AOD trends. One of the remarkable points of K2007 is the close agreement between the measured aerosol optical thickness (AOD) and the fire counts. We note that MODIS also produces a fire count product. We found, but did not show in K2007, that the MODIS and AVHRR fire counts were tightly correlated, although offset from one another. Both the AVHRR and MODIS fire counts for the years 2001 through 2005 were correlated with the MODIS AOD (R > 0.95). Having the choice of either satellite’s fire count data, we decided to show only the AVHRR record due to its longer time series. S2009 point out an inconsistency between the AVHRR fire counts in years 1998–1999 and the post2000 record.We cannot rule out this inconsistency. However, even if we limit the analysis to the post-2000 MODIS era we see a clear and consistent increase in both smoke loading and fire counts, followed by a drastic decrease in 2006. The S2009 objections to the 1998 and 1999 data may be valid, but are immaterial to the main points or the conclusions of K2007. Figure 1 shows the time series in AOD for 2000– 2006 and fire counts from MODIS for 2001–2005. The 2000–2005 trends and the correlations between fire counts and AOD are clearly demonstrated. [5] The five years ofMODIS fire counts shown in Figure 1 were all that were available to us during the analysis for K2007, due to MODIS reprocessing at that time. Today we have benefit of additional data. Figure 2 shows scatter plots between the monthly mean AOD over the Amazon forest (Latitude 3N to 16S; Longitude 52W to 72W) versus normalized monthly fire counts from MODIS from 2000 to 2008. The high correlation between the mean AOD and the fire counts is clearly seen in Figure 2. We also note a shift in the relationship between these variables when comparing the dry season (J,J,A) with the transition season (S,O,N). A lesser number of fires per AOD is seen in the wetter months, which can be explained by a higher cloud fraction in the transition season and a larger number of fires obscured by clouds. It is interesting to see that difference in the relationship between smoke and fires in the wetter and dryer parts of the season becomes smaller for the higher AOD cases. This can be explained by the effect of absorbing aerosols on the cloud fraction [Koren et al., 2008]. There is an indisputable relationship between fires and smoke in the Amazon Basin dry season that showed a strong increasing trend in biomass burning activity from 2000 through 2005, then a drastic decrease in both fires and smoke in 2006. [6] We now address the second criticism of S2009. What causes this interannual variability and the trends in biomass burning in the Amazon? First we need to clarify an ambiguity from K2007 that created a concerned response in S2009. There was a severe drought during the 2005 dry season that affected the western Amazon, which is a tri-national region consisting of parts of Brazil, Peru and Bolivia. The policy GEOPHYSICAL RESEARCH LETTERS, VOL. 36, L03807, doi:10.1029/2008GL036063, 2009 Click Here for Full Article