the impact of atmospheric aerosol from biomass burning on amazon ...

3 downloads 0 Views 5MB Size Report
Key words: Aerosol; Amazon; (A)ATSR; Biomass burn- ing. 1. INTRODUCTION .... ties is based on a physical model of light scattering, and requires no a-priori ...
THE IMPACT OF ATMOSPHERIC AEROSOL FROM BIOMASS BURNING ON AMAZON DRY-SEASON DROUGHT Suzanne L. Bevan1 , Peter R. J. North1 , William M. F. Grey2 , Sietse O. Los1 , and Stephen E. Plummer3 1

School of the Environment and Society, Swansea University, Singleton Park, Swansea, SA2 8PP, UK, Email: [email protected] 2 Met Office, FitzRoy Road, Exeter, Devon, EX1 3PB, UK 3 International Geosphere Biosphere Programme-European Space Agency Joint Projects Office, European Space Agency, European Space Research Institute, 00044 Frascati, Italy

ABSTRACT It is increasingly apparent that the future of the Amazon rainforest is under threat from both climate change and agricultural practices. Here we use a 13-year time series of (A)ATSR derived aerosol optical depth (AOD) measurements to examine the role of aerosols in the interaction between deforestation, biomass burning and drought over the Amazon. The seasonal cycle of AOD shows peaks in September and March correlated with local and more remote biomass burning, respectively. A decreasing trend in dry-season AOD between 1995 and 2000 and a subsequent increase from 2000 to 2004 can be explained by deforestation practices driven by economic forces. Throughout the time series dry-season AODs are inversely correlated with dry-season precipitation suggesting a positive feedback between aerosols and drought that may contribute to enhanced drought under climate change. Key words: Aerosol; Amazon; (A)ATSR; Biomass burning.

1.

INTRODUCTION

The Amazon rainforest plays a major role in regulating the Earth’s climate via the exchange of water, momentum and carbon between biosphere and atmosphere. However, the Amazon rainforest is under threat from both deforestation and the effects of climate change [1, 2]. Uncertainties in predicting the future of the Amazon rainforest include anthropogenic influences such as deforestation and fire [3, 4, 5]. The neglect of fire disturbance in dynamic vegetation models is likely to overestimate the robustness of tropical forests to warming and drying [1]. Forest fires increase atmospheric aerosol concentrations and these can have both regional and global impacts on the solar heating of the surface and atmosphere, and on the hydrological cycle [6, 7].

_________________________________________ Proc. of the '2nd MERIS / (A)ATSR User Workshop', Frascati, Italy 22–26 September 2008 (ESA SP-666, November 2008)

In the Amazon deforestation and biomass burning can trigger a positive feedback cycle of increased fire disturbance and local drought conditions, amplifying drought linked to both anthropogenic global climate change and natural climate variability [8]. Two causes of drought amplification are: 1) a reduction in moisture recirculation by a reduction in evapotranspiration and 2) rain suppression caused by elevated atmospheric aerosol concentrations. 25 to 35% of the moisture for Amazon precipitation is contributed by regional recycling [9] and modelled deforestation leads to a significant reduction in Amazon summer rainfall [7]. The aerosol effect is less clear cut. During the dry season, enhanced atmospheric aerosol concentrations are a direct manifestation of biomass burning — remotely sensed aerosol optical depths (AODs) averaged over a wide area in the Amazon are strongly correlated with fire activity [10]. It is these fire-associated aerosols that have the potential to affect precipitation processes. During the rainy season a positive feedback process connects warmprocess precipitation wash-out of aerosol, cleaner air, fewer condensation nuclei and faster droplet coalescence [11]. In contrast, during the dry season, biomass-burning enhanced aerosol concentrations can delay the onset of precipitation to higher elevations and suppress low-level washout processes [11]. Countering this positive feedback hypothesis is the possibility that although warm-rain processes are suppressed, biomass burning may enhance convection and intensify cold-rain processes leading to a local increase in precipitation [12]. A number of global-scale studies report a positive correlation between cloud fraction and AOD [13, 14, 15], however, over the Amazon region a negative correlation has also been observed [16] and modelled [17]. Evidence of the effect of aerosols on precipitation is much sparser [18]. In this study we generate a time series of AODs over the Amazon region from Along Track Scanning Radiometer (ATSR) top of atmosphere (TOA) radiance measurements. We analyse the seasonal signal and multi-year trends in conjunction with a regional firecount product and measured precipitation. The aim is to investigate the possibility of a synergistic effect between de-

forestation, biomass burning and climate-change induced drought over the region.

2.

METHOD AND DATA

The 13-year time series of AODs over the Amazon region (30◦ –70◦ W, 0◦ –15◦ S) at a resolution of 10 × 10 km was generated using ATSR-2 and AATSR TOA radiance measurements. The AODs are validated via a comparison with shorter time series of in-situ AERosol RObotic NETwork (AERONET) measurements and MODIS retrievals. A correlation analysis is performed between monthly composites of AODs, ATSR World Fire Atlas data, and monthly precipitation anomalies and a regional trend analysis of AOD is carried out.

2.1.

Aerosol optical depths from (A)ATSR

Aerosol optical depth or thickness is an integration of the column concentration of aerosol and describes how solar radiation is attenuated by the aerosol [18]; it is therefore a dimensionless parameter that has a wavelength dependence. A number of methods exist to determine AOD over land, from a variety of remote sensing instruments [18, 19]. Here we exploit a 13-year time series of TOA radiance measurements collected by the ATSR series of instruments, and an algorithm developed by [20], implemented as described by [21]. Some validation of the algorithm has already been performed [21, 19]; additional validation for the Amazon region is presented here. Aerosol optical depths are retrieved at 550 nm from ATSR-2 (1995–2002, on ERS-2) and AATSR (2002– 2007, on ENVISAT) TOA radiances. The multi-angle, multi-channel observing capability of (A)ATSR combined with a physical model of the spectral change with view angle of surface reflectance allows estimates to be made of the AOD, if assumptions are made regarding other aerosol optical properties such as phase function and single-scattering albedo. Simultaneously, the surface bidirectional reflectance is retrieved at both viewing angles and all four optical wavebands, using a look-up table (LUT) parameterised by the 6S model [22, 23]. The method used to retrieve atmospheric aerosol properties is based on a physical model of light scattering, and requires no a-priori information about the land surface spectral properties [24]. Instead, the dual-view capability of the sensor is used to separate the contribution of surface from atmospheric scattering, allowing AOD to be estimated over a wide range of surfaces, with results relatively unaffected by changes in surface brightness. In addition to retrieving AOD, the method is also able to identify the best aerosol model from five candidate models. (A)ATSR radiances are retrieved at a spatial resolution of 1 × 1 km but groups of 10 × 10 pixels are averaged before

processing in order to reduce noise and minimise errors in coregistration between the forward and nadir images. At this latitude repeat coverage time for the A(ATSR) instruments is about every 6 days, however, in order to generate as complete an image as possible, i.e. to minimise the number of missing pixels due to cloud, the results are spatially and temporally composited. The resulting images are georeferenced, reprojected monthly mean composites with a spatial resolution of 0.1◦ .

2.2.

Intercomparison with AERONET and MODIS

Retrieved AOD measurements from (A)ATSR are compared with in-situ AOD measurements and with other satellite retrievals of AOD. The in-situ data consist of sun-photometer measurements from three AERONET stations. The other satellite-retrieved AODs are from MODIS on board the Terra satellite [25]. The time series of (A)ATSR AOD data points used for the comparison consist of the means of all retrievals per orbit that lie within 0.15◦ of the AERONET site, this may consist of up to 9 observations per AERONET data point. The three AERONET stations selected for the intercomparison are Alta Floresta (56.10◦ W, 9.87◦ S), Abracos Hill (62.36◦ W, 10.76◦ S) and Rio Branco (67.87◦ W, 9.96◦ S). Note that Alta Floresta was one of the stations included in the initial AATSR AOD validation exercise by [21]. AERONET sun-photometer measurements are not made at a wavelength of 550 nm, therefore, in order to make a direct comparison with (A)ATSR measurements it was necessary to interpolate the 440 nm and 670 nm AERONET AOD estimates. The optical depth parame˚ ter β, and the Angstr¨ om exponent, α, are calculated by solving τλ = βλα , where τλ is the optical depth at wavelength λ, using the AODs at 440 nm and 670 nm. The AOD at 550 nm is then calculated using these β and α values. Intercomparisons with Terra MODIS AOD retrievals were also made at the locations of the three AERONET sites, for the years 2002 to 2006 inclusive. The MODIS retrievals have a spatial resolution of 10 × 10 km, therefore, any pixel which overlaps the AERONET site is included in the comparison.

2.3.

Precipitation data

Global Precipitation Climatology Project (GPCP) precipitation data at 2.5◦ spatial resolution were used to investigate the relation between precipitation and AOD. The data are a combined product of gauge analyses and satellite estimates based on microwave, infrared, and sounder instruments [26]. Monthly data were oversampled to produce a grid of 0.1◦ resolution to match that of the remotely sensed AOD, and monthly anomalies were calculated as departures from the 1995–2007 13-year mean.

Figure 1. Time series of AERONET (green), MODIS (blue) and AATSR (red) 550 nm aerosol optical depths at Alta Floresta.

2.4.

Fire data

Monthly fire counts in the region were accumulated from ATSR World Fire Atlas data. These data exist as dates and locations of night-time hot pixels detected by ATSR2 (1995–2002) and AATSR (2003–). Pixel resolution is 1 × 1 km and a hot pixel is defined as having a 3.7 µm brightness temperature of greater than 312 K (algorithm 1) or 308 K (algorithm 2). Data are available for November and December 1995, and then from July 1996 onwards. For this study algorithm 1 data were accumulated as a total firecount per month, within the region (30◦ –70◦ W, 10◦ N–15◦ S). Limitations to the data include the detection threshold, omissions due to cloud cover, and fires persisting beyond the revisit time (6 days or less at this latitude) being counted more than once [27].

3.

RESULTS AND DISCUSSION

Fig. 1 shows time series plots of all the (A)ATSR 550 nm AOD measurements, all recorded AERONET measurements, and all available MODIS retrievals, at the AERONET site Alta Floresta. The Pearson’s correlation coefficients [28] between (A)ATSR and AERONET AODs at 550 nm, for any AERONET observation made within 1 hour of the satellite overpass time, are given in Tab. 1 for all three AERONET sites. All correlations are significant at the 95% confidence level. For these coincident observations the (A)ATSR algorithm has a tendency to underestimate the AOD in comparison with the AERONET observations. Possible contributing factors toward discrepancies between the two methods of AOD measurement are discussed in [21] and include: small differences in aquisition times, cloud contamination, inappropriate aerosol model, and land-surface heterogeneity. [21] also suggest that the algorithm for retrieving AOD performs better for low optical depths. For the three sites considered here, the root mean square (r.m.s.) difference over all sites combined falls from 0.270 to 0.222 if the comparison is limited to optical depths below 1.0. The low correlation coefficient between (A)ATSR and Rio Branco AERONET is caused by a single (A)ATSR point in May where there is a spike in AOD not present

in the AERONET measurements. This point may be due to cloud contamination and excluding it from the intercomparison results in a correlation coefficient of 0.603 and a r.m.s difference of 0.158. In spite of the fairly high r.m.s. differences the high correlation coefficients suggest that the (A)ATSR retrievals of AOD will be valid for trend analyses. A similar comparison with MODIS data makes use of retrievals from 2002–2006 inclusive. In this analysis any MODIS AODs retrieved on the same day as, and therefore also within 1 hour of, (A)ATSR measurements are included. The Pearson’s correlation coefficients between (A)ATSR and MODIS AODs are high and significant but the r.m.s. difference between the two means of retrieval is 0.436 over all sites combined. While MODIS is reported to measure AOD over land to an accuracy of 0.05 ± 0.2τ [18], in a 2-year validation study the global average difference between MODIS and AERONET 550 nm AOD measurements was +41%, with MODIS overestimating in comparison with AERONET; for South America this difference was +21% [25]. The relatively large difference between MODIS and (A)ATSR exhibited here is consistent with a known MODIS tendency to overestimate AOD combined with an (A)ATSR underestimate, a discrepancy which increases as AOD increases.

3.1.

Aerosol optical depth and fire count

A total of 135 monthly composites of AOD for the region (30◦ –70◦ W, 0◦ –15◦ S) were produced using ATSR2 (June 1995–January 2001) and AATSR (July 2002– December 2007) data. Typical spatial distributions of AOD are shown in Fig. 2(a) for July and (b) for September 2005. In July the atmosphere is generally relatively clean with mean aerosol optical depths of the order of 0.3 to 0.4. By September, the biomass-burning season is at its peak and region-mean optical depths are of the order of 0.8 with maxima of up to 2.0. A time series of the monthly 550 nm AOD composites, averaged for the south-west Amazon region ((50◦ –70◦ W, 5◦ –15◦ S), red box in Fig. 2(b)) was produced and is shown by the black line in Fig. 3. The annual signal of biomass burning can clearly be seen throughout the

Table 1. Results of the comparison between (A)ATSR 550 nm AODs and all coincident AERONET observations, and all coincident MODIS retrievals, for three AERONET sites. The correlation coefficients are all significant at the 95% level.

Comparison with AERONET Number of points Correlation r.m.s. difference A(ATSR) mean AERONET mean Comparison with MODIS Number of points Correlation r.m.s. difference A(ATSR) mean MODIS mean

Rio Branco

Abracos Hill

Alta Floresta

All sites

34 0.358 0.239 0.213 0.218

56 0.834 0.274 0.267 0.404

65 0.837 0.282 0.224 0.336

155 0.779 0.270 0.237 0.335

58 0.894 0.219 0.210 0.314

64 0.794 0.547 0.298 0.588

54 0.767 0.463 0.218 0.362

176 0.817 0.436 0.244 0.428

Figure 2. Monthly composites of aerosol optical depth for (a) July 2005 and (b) September 2005. Shown in (a) are the AERONET sites of Abracos Hill (AH), Alta Floresta (AF), and Rio Branco (RB). White areas indicate no AOD value retrieved. The red box in (b) delineates the region for the time series means.

13-year period, with AODs peaking in September and a persistent secondary peak in March of most years. The firecount product follows the same cycle with a peak in September and a smaller secondary peak in March. Over the 13-year period we find a statistically significant positive correlation of 0.64 between the total number of satellite-observed fires and region-mean AOD. The distribution of fire hot pixels (see Figs. 4(a) and (b)) shows that the September peak coincides with fires on the south and eastern borders of the region of rainforest land cover, the peak in March coincides with fires much further to the north, in Venezuela. Climatological 1979–1998 monthly mean wind fields at 850 mb were obtained from the National Center for Environmental Prediction/National Center for Atmospheric Research. The September wind field is consistent with smoke from the fires on the eastern edge of the rainforest being transported across the region. The wind field for March also suggests that the elevated AODs observed over the Amazon region in March may have their source further to the north.

3.2.

Precipitation

Also shown in Fig. 3 is the regional mean of the monthly precipitation anomaly. Clearly shown are the persistent negative anomalies from September 1997 through to September 1998, an El Ni˜no year. Pearson’s correlation coefficient between the regional mean AOD for September and precipitation for September is −0.594, this is signficant at the 95% level. On this region-mean basis there are no significant correlations between September’s AOD and the precipitation anomaly for any other month, i.e. there is no evidence for any lagged correlation between September’s AOD and earlier precipitation patterns. Having identified a region-mean negative correlation between September’s AOD and precipitation, Fig. 5(a) shows the distribution of this correlation. Fig. 5(b) shows the probability values (p-values) of the null hypothesis of no correlation. It can be seen that the regions where there is a negative correlation generally have low p-values, in other words, the negative correlations are significant. The inverse correlation between dry-season precipitation

Figure 3. Mean monthly AOD and precipitation anomaly for the region 50◦ –70◦ W, 5◦ –15◦ S. The total monthly firecount is plotted for the region 30◦ –70◦ W, 10◦ N–15◦ S.

Figure 4. ATSR World Fire Atlas data for (a) September 2005 and (b) March 2005. Insert key shows International Geosphere-Biosphere Programme (IGBP) land cover classifications.

Figure 5. (a) Correlation coefficient between aerosol optical depth and precipitation for September, 1995–2007. (b) Correlation p-value.

and AOD allows for the possibility that elevated atmospheric aerosol concentrations are able to suppress warmrain processes [29] but is in contrast to the positive correlation found by [12]. However, the [12] analysis considered correlation in space only, for the 2000 and 2003 dry seasons, and included a region to the north of the Amazon region where we also find some positive correlations in time. [30] have since shown that in 2003 the lower atmosphere was humid and unstable, and that

the aerosol-cloud relationship differed to that of a drier, more stable year (2002). In 2003 cloud fraction increased and cloud effective radius decreased with AOD which would suppress warm rain processes but might enhance convective rainfall. In contrast, in 2002 cloud fraction decreased with AOD perhaps due to aerosol absorption of solar radiation leading to cloud dissipation [30]. Climate model predictions of reductions in dry-season precipitation for the Amazon combined with an already pre-

Figure 6. Annual trend in aerosol optical depth for (a) 1995–2000 and (b) 2000–2005.

dominantly inverse relationship between AOD and precipitation demonstrated in this study, support a positive feedback hypothesis between biomass burning following deforestation and drought in the Amazon. This delayed transitory positive feedback between deforestation and drought compounds that due to a permanent reduction in transpiration processes following deforestation. At the end of the rainy season, in March, when AODs are slightly elevated, possibly due to fires north of the Amazon region, there is no longer any significant correlation between AOD and precipitation, supporting the theory that an already dry atmosphere is required in order for atmospheric aerosol to suppress precipitation.

3.3.

Regional trend analysis

The regional distribution of annual trend in dry-season AOD, based on the monthly composites, is shown in Fig. 6 for (a) 1995–2000 and (b) 2000–2005. The majority of the region experiences a decreasing trend in AOD for the first period and an increasing trend for the second period. The 2000–2005 increasing trend in AOD confirms findings published by [10] based on MODIS 1 × 1 degree monthly mean 550 nm aerosol optical depth retrievals averaged over August through November each year. We are also confirm the fall in dry-season AOD from 2005 to 2006, with the region-mean decreasing from 0.78 to 0.40. [10] attribute this trend reversal to the activities of a drought and fire alert service established following the disastrous drought of 2005 [31]. For 2007, although Fig. 3 suggests no increase in AOD compared with 2006, an inspection of the geographic distribution of AOD retrievals for September 2007 shows persistent cloud over the regions of expected high aerosol concentrations. The 2007 firecount is at its maximum for the entire series and the AERONET site Alta Floresta reports depths of 2.0 or more (Fig. 1), suggesting that 2007 AODs were, in fact, exceptionally high, and the (A)ATSR retrieval was hampered by cloud. The decline in dry-season AOD from 1995 to 2000 is compatible with the general trend in annual deforestation

rate of the Brazilian Amazon which is dominated by a peak in 1995 relative to earlier years. The 1995 peak was probably due to an increase in the availability of capital and agricultural credit and a peak in land values following Plano Real financial reform in Brazil. From 1995 to 1997 deforestation rates slowed as land values fell by 50% [32]. From 2000 to 2004, increasing global markets for soybeans and beef encouraged deforestation notwithstanding a slow domestic economy in Brazil [32], seen as an increasing trend in dry-season AOD peaking in 2005. From 2004 to 2007 deforestation rates more than halved due to falling soybean prices, a strong Brazilian currency and active government intervention. In spite of this the drought in 2005 still resulted in an exceptionally high number of fires and AOD prior to the decrease in 2006. [2] discuss how a web of economic teleconnections affects deforestation rates, and it seems probable that rising global demands for soybeans and beef, and for biodiesel and ethanol to replace fossil fuels, will drive rates up again. The AODs in 2006 seem likely to be only a temporary reduction.

Long time series satellite datasets can suffer from drift and so we must be aware of this possibility. However, the (A)ATSR series of instruments undergo extensive prelaunch calibration and are subject to rigorous onboard calibration ensuring radiometric accuracy and continuity between successive instruments [33]. A 6-month data overlap period between ATSR-2 and AATSR from January to June 2003 was used to find the mean difference between AOD retrievals based on the two instruments. Throughout the 6-month period, and within the region being studied, there were a total of 16,831 pixels in the monthly composites with 1 or more coincident retrievals from both ATSR instruments. The overall mean of the AATSR−ATSR-2 difference was -0.03, the standard deviation of the differences was 0.174. This difference is much smaller than the magnitude of the annual trends identified. Some of the difference may be due to ERS-2 and Envisat orbits being separated by about 30 minutes. Mean September AODs from the three AERONET sites considered in this study also show the trends identified in the (A)ATSR retrievals (Fig. 7).

REFERENCES

Figure 7. September mean 550 nm AOD for the regionmean (A)ATSR retrievals and for three AERONET sites. 4.

CONCLUSIONS

Fire disturbance and associated atmospheric aerosol must be taken into account in modelling the impacts of climate change on the Amazon. Enhanced AOD affects both regional and global energy balance and precipitation patterns. The 13-year A(ATSR) dataset is the longest available time series of remotely sensed AODs for the Amazon region and reveals clear seasonal signals and multiyear trends. The south-west Amazon regional mean of monthly AOD correlates strongly with monthly firecounts with a peak in September associated with Amazonian fires and a secondary peak in March associated with fires in Venezuela. The time series of regional means also shows that prior to the published increasing trend in dry-season AOD of 2000–2005, the region experienced a 5-year decrease. A strong inverse correlation exists between September AOD and monthly precipitation anomalies. The relationship between enhanced atmospheric aerosol concentrations and precipitation is a complex one. The AOD can be driven by land management practices in response to economic forces, as is shown by the correspondence between multi-year trends in deforestation practices, firecount and AOD, but in years of exceptional drought AOD may be governed more by climatic conditions. Our results, showing that dry-season intensity correlates with externally forced AOD trends, suggest that atmospheric aerosols are imposing a positive feedback on drought, not simply rising in response to dry conditions. The prerequisite for this process might be that it can only occur for an atmosphere too dry for convective rainfall to replace the usual low-level warm-rain coalescence [30]. With dry-season rainfall predicted to decrease over the Amazon under future climate change, the opportunity for the positive feedback between deforestation and regional climate change via the aerosol mechanism will increase.

ACKNOWLEDGMENTS The authors would like to thank NERC EODC for providing the (A)ATSR image data, and NASA LAADS for providing the MODIS aerosol products. We would also like to thank principal investigator Brent Holben, and staff of AERONET for their effort in establishing and maintaining the Amazon measurement sites. The ATSR World Fire Atlas data were generated through of ESA’s DUE. This work was funded by NERC CLASSIC.

[1] Peter M. Cox, Phil P. Harris, Chris Huntingford, Richard A. Betts, Matthew Collins, Chris D. Jones, Tim E. Jupp, Jos´e A. Marengo, and Carlos A. Nobre. Increasing risk of Amazonian drought due to decreasing aerosol pollution. Nature, 453, 2008. [2] Daniel C. Nepstad, Claudia M. Stickler, Britaldo Soares-Filho, and Frank Merry. Interactions among Amazon land use, forests and climate: Prospects for a near-term forest tipping point. Phil. Trans. R. Soc. Lond. B, 363, 2008. [3] Peter M. Cox, Richard A. Betts, Chris D. Jones, Steven A. Spall, and Ian J. Totterdell. Acceleration of global warming due to carbon-cycle feedbacks in a coupled climate model. Nature, 408:184–187, 2000. [4] R. A. Betts, P. M. Cox, M. Collins, P. P. Harris, C. Huntingford, and C. D. Jones. The role of ecosystem–atmosphere interactions in simulated Amazonian precipitation decrease and forest dieback under global climate warming. Theor. Appl. Climatol., 78, 2004. [5] J. H. Christensen, B. Hewitson, A. Busuioc, A. Chen, X. Gao, I. Held, R. Jones, R. K. Kolli, W.T. Kwon, R. Laprise, V. Maga˜na Rueda, L. Mearns, C. G. Men´endez, J. R¨ais¨anen, A. Rinke, A. Sarr, and P. Whetton. Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, chapter Regional Climate Projections. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 2007. [6] V. Ramanathan, P. J. Crutzen, J. T. Kiehl, and D. Rosenfeld. Aerosols, climate, and the hydrological cycle. Science, 294:2119–2124, 2001. [7] David Werth and Roni Avissar. The local and global effects of Amazon deforestation. J. Geophys. Res., 107, 2002. [8] Luiz Eduardo O. C. Araga˜o, Yadvinder Malhi, Nicolas Barbier, Andre Lima, Yosio Shimabukuro, Liana Anderson, and Sassan Saatchi. Interactions between rainfall, deforestation and fires during recent years in the Brazilian Amazonia. Philos. Trans. R. Soc. B, 363, 2008. [9] E. A. B. Eltahir and R. L. Bras. Precipitation recycling in the Amazon basin. Quarterly Journal of the Royal Meteorological Society, 120:861–880, 1994. [10] Ilan Koren, Lorraine A. Remer, and Karla Longo. Reversal of trend of biomass burning in the Amazon. Geophys. Res. Lett., 34, 2007. [11] M. O. Andreae, D. Rosenfeld, P. Artaxo, A. A. Costa, G. P. Frank, K. M. Longo, and M. A. F. SilvaDias. Smoking rain clouds over the Amazon. Science, 435, 2004. [12] J. C. Lin, T. Matsui, R. A. Pielke Sr., and C. Kummerow. Effects of biomass-burning-derived

aerosols on precipitation and clouds in the Amazon Basin: A satellite-based empirical study. J. Geophys. Res., 111, 2006. [13] M. Sekiguchi, T. Nakajima, K. Suzuki, K. Kawamoto, A. Higurashi, D. Rosenfeld, I. Sano, and S. Mukai. A study of the direct and indirect effects of aerosols using global satellite data sets of aerosol and cloud parameters. J. Geophys. Res, 108, 2003. [14] Yoram J. Kaufman and Ilan Koren. Smoke and pollution aerosol effect on cloud cover. Science, 313:655–658, 2006. [15] Norman G. Loeb and Gregory L. Schuster. An observational study of the relationship between cloud, aerosol and meteorology in broken low-level cloud conditions. J. Geophys. Res., 113, 2008. [16] Ilan Koren, Yoram J. Kaufman, Lorraine A. Remer, and Jose V. Martins. Measurement of the effect of Amazon smoke on inhibition of cloud formation. Science, 303, 2004. [17] Yan Zhang, Rong Fu, Hongbin Yu, Robert E. Dickinson, Robinson Negron Juarez, Mian Chin, and Hui Wang. A regional climate model study of how biomass burning aerosol impacts land-atmosphere interactions over the amazon. J. Geophys. Res., 113, 2008. [18] Yoram J. Kaufman, Didier Tanr´e, and Olivier Boucher. A satellite view of aerosols in the climate system. Nature, 419:215–223, 2002. [19] A. A. Kokhanovsky, F.-M. Breon, A. Cacciari, E. Carboni, D. Diner, W. Di Nicolantonio, R. G. Grainger, W. M. F. Grey, R. H¨oller, K.-H. Lee, Z. Li, P. R. J. North, A. M. Sayer, G. E. Thomas, and W. von Hoyninge-Huene. Aerosol remote sensing over land: A comparison of satellite retrievals using different algorithms and instruments. Atmos. Res., 85, 2007. [20] Peter R. J. North. Estimation of aerosol opacity and land surface bidirectional reflectance from ATSR-2 dual-angle imagery: Operational method and validation. J. Geophys. Res., 107:doi:10.1029/2000JD000207, 2002. [21] William F. Grey, Peter R. J. North, Sietse O. Los, and Ross M. Mitchell. Aerosol optical depth and land surface reflectance from multiangle AATSR measurements: Global validation and intersensor comparisons. IEEE Trans. Geosci. Remote Sensing, 44, 2006. [22] Eric F. Vermote, Didier Tanr´e, Jean Luc Deuz´e, Maurice Herman, and Jean-Jacques Morcrette. Second Simulation of the Satellite Signal in the Solar Spectrum, 6S: An overview. IEEE Trans. Geosci. Remote Sensing, 35:675–686, 1997. [23] W. M. F. Grey, P. R. J. North, and S. O. Los. Computationally efficient method for retrieving aerosol optical depth from atsr-2 and aatsr data. Applied Optics, 45:2786–2795, 2006.

[24] Peter R. J. North, Stephen A. Briggs, Stephen E. Plummer, and Jeffrey J. Settle. Retrieval of land surface bidirectional reflectance and aerosol opacity from ATSR-2 multiangle imagery. IEEE Trans. Geosci. Remote Sensing, 37:526–537, 1999. [25] L. A. Remer, Y. J. Kaufman, D. Tanre, S. Mattoo, D. A. Chu, J. V. Martins, R. R. Li, C. Ichoku, R. C. Levy, R. G. Kleidman, T. F. Eck, E. Vermote, and B. N. Holben. The MODIS aerosol algorithm, products, and validation. Journal of the Atmospheric Sciences, 62:947–973, 2005. [26] R. F. Adler, G. J. Huffman, A. Chang, R. Ferraro, P. Xie, J. Janowiak, B. Rudolf, U. Schneider, S. Curtis, D. Bolvin, A. Gruber, J. Susskind, and P. Arkin. The Version 2 Global Precipitation Climatology Project (GPCP) monthly precipitation analysis (1979-present). J. Hydrometeorol., 4:1147– 1167, 2003. [27] Olivier Arino, Stephen Plummer, and Diane Defrenne. Fire disturbance: The ten years time series of the ATSR World Fire Atlas. In Proceedings of the MERIS-AATSR workshop 2005, Frascati, Italy, 2005. [28] R Development Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria, 2005. ISBN 3-900051-07-0. [29] Daniel Rosenfeld. TRMM observed first direct evidence of smoke from forest fires inhibiting rainfall. Geophys. Res. Lett., 26:3105–3108, 1999. [30] Hongbin Yu, Rong Fu, Robert E. Dickinson, Yan Zhang, Mingxuan Chen, and Hui Wang. Interannual variability of smoke and warm cloud relationships in the Amazon as inferred from MODIS retrievals. Remote Sensing of Environment, 111:435– 449, 2007. [31] I. F. Brown, W. Schroeder, A. Setzer, M. De Los Rios Maldonado, N. Pantoja, A. Duarte, and J. Marengo. Monitoring fires in southwestern Amazonia rain forests. Eos Trans. AGU, 87:253,259, 2006. [32] Philip M. Fearnside. Deforestation in Brazilian Amazonia: History, rates, and consequences. Conservation Biology, 19:680–688, 2005. [33] D. L. Smith, J. Delderfield, D. Drummond, T. Edwards, C. T. Mutlow, P. D. Read, and G. M. Toplis. Calibration of the AATSR instrument. Adv. Space Res., 28:31–39, 2001.