Isolating the effect of biomass burning aerosol emissions on 20th century hydroclimate in South America and Southeast Asia

Biomass burning is a significant source of aerosol emissions in some regions and has a considerable impact on regional climate. Earth system model simulations indicate that increased biomass burning aerosol emissions contributed to statistically significant decreases in tropical precipitation over the 20th century. In this study, we use the Community Earth System Model version 1 Large Ensemble (CESM1-LENS) experiment to evaluate the mechanisms by which biomass burning aerosol contributed to decreased tropical precipitation, with a focus on South America and Southeast Asia. We analyze the all-but-one forcing simulations in which biomass burning aerosol emissions are held constant while other forcings (e.g., greenhouse gas concentrations) vary throughout the 20th century. This allows us to isolate the influence of biomass burning aerosol on processes that contribute to decreasing precipitation, including cloud microphysics, the radiative effects of absorbing aerosol particles, and alterations in regional circulation. We also show that the 20th century reduction in precipitation identified in the CESM1-LENS historical and biomass burning experiments is consistent across Coupled Model Intercomparison Project Phase 5 models with interactive aerosol schemes and the CESM2 single-forcing experiment. Our results demonstrate that higher concentrations of biomass burning aerosol increases the quantity of cloud condensation nuclei and cloud droplets, limiting cloud droplet size and precipitation formation. Additionally, absorbing aerosols (e.g., black carbon) contribute to a warmer cloud layer, which promotes cloud evaporation, increases atmospheric stability, and alters regional circulation patterns. Corresponding convectively coupled circulation responses, particularly over the tropical Andes, contribute to further reducing the flow of moisture and moisture convergence over tropical land. These results elucidate the processes that affect the water cycle in regions prone to biomass burning and inform our understanding of how future changes in aerosol emissions may impact tropical precipitation over the 21st century.


Introduction
Aerosol and greenhouse gas (GHG) emissions are the two largest radiative forcing agents [1] and can significantly alter precipitation regimes [2].However, they often have competing effects, which reduces the magnitude of overall trends and makes it difficult to disentangle their specific impacts on climate [3].For example, increasing GHG concentrations increase surface temperature through the greenhouse effect, while increased aerosol concentrations cause cooling by reflecting incoming solar radiation [3].The GHG-induced warming increases the intensity and amount of precipitation, whereas increases in aerosol particle concentrations can suppress regional precipitation due to interactions with cloud microphysics and radiation [1].Despite these opposing effects, Earth system model simulations suggest that 20th century precipitation changes were largely aerosol-dominated, with decreases over tropical land [4,5].In contrast, 21st century precipitation projections are GHG-dominated and suggest wetter tropical conditions in the future [5,6].This transition from aerosol-to GHG-dominated change results from both a projected increase in GHG emissions and a reduction in future anthropogenic aerosol emissions [1,3].The changes in large-scale forcing expected over the 21st century highlight the need to better quantify past climate responses to aerosol forcing to improve our understanding and ability to project climate responses to changes in aerosol emissions.
Cloud and aerosol processes occur at scales that are many orders of magnitude smaller than the gridspacing of Earth system models, which makes it challenging to represent these processes and constrain the effects of aerosol on climate [7].The importance of such processes for precipitation is evident when comparing models that include advanced aerosol schemes linked to cloud processes, such as twomoment cloud microphysics [8], to those with more simplistic aerosol schemes [4].For example, models from the Coupled Model Intercomparison Project Phase 5 (CMIP5) with explicit treatment of aerosolcloud interactions simulate large 20th century reductions in precipitation over tropical land, while CMIP5 models with simpler aerosol schemes (e.g.only represent the direct radiative effects of aerosol) have less pronounced and inconsistent changes [4].When considering aerosol effects on climate, it can be useful to differentiate between anthropogenic (e.g.industrial activity) and biomass burning (e.g.forest fires) aerosol, whose emissions can have different spatial and seasonal distributions and different factors driving future reduction and mitigation [1].The impacts of these two sources of aerosols, both of which contribute to primary organic matter (POM) and black carbon (BC) particle concentrations, may be similar for a given region, but the spatial and temporal distributions of their impacts can be very different [5,9,10].While anthropogenic aerosol concentrations are expected to decrease going forward, changes in climate are likely to increase fire risk in the tropics, and emissions from increased fire occurrence could be impactful in regional hydrological cycles [11].
Increased biomass burning over the 20th century was driven by human activities, such as deforestation [12], and resulted in a significant increase in biomass burning emissions which yields an opportunity to evaluate the impacts of aerosol on regional hydrological cycles.Past studies utilized models and observations to assess the impact of biomass burning on clouds and precipitation, and they revealed that biomass burning aerosols reduce precipitation through aerosol-cloud and aerosol-radiation interactions [10,[13][14][15].Aerosol-cloud and aerosol-radiation interactions are complex and are not mutually exclusive, which makes establishing dominant mechanisms challenging.Specifically, aerosol-cloud interactions affect both precipitation efficacy and indirect effects on radiative transfer, which are both associated with the role of aerosol particles as cloud condensation nuclei (CCN) in the formation of cloud droplets.The indirect effect of aerosol results from higher concentrations of CCN, which cause an increase in cloud droplet number and decrease in droplet size [16,17].Smaller cloud droplets backscatter more solar radiation [16] and impede the formation of precipitation, lengthening cloud lifetimes [17], both of which can increase albedo.Aerosol-radiation interactions include the direct effect of aerosol on clearsky radiative fluxes, which is characterized by surface temperature cooling due to reflection or absorption of radiation by aerosol [18].The semi-direct effect is an extension of the direct effect, which pertains specifically to absorbing aerosols that can heat the aerosol layer and contribute to cloud evaporation and reductions in precipitation [19].The direct, semidirect, and indirect effects of aerosol can also affect precipitation by increasing atmospheric stability due to surface cooling and/or atmospheric heating, thus inhibiting convection [10,15,20].
Another less-studied mechanism by which increased biomass burning aerosols may alter clouds and precipitation is through changes in circulation and regional moisture transport.Previous work has yielded mixed results that can depend on the region, season, and prevailing circulation patterns; some studies have found low level winds are reduced due to aerosol, while other research suggests that they are strengthened [9,10,21].A better understanding of the impact of long-term moisture transport changes, in addition to the roles of the direct, semidirect, and indirect effects, is important for explaining long-term precipitation changes due to biomass burning aerosols in South America and Southeast Asia (SEA).In particular, large-scale moisture transport is essential to the inland hydrological cycle over South America.Previous research focuses on the effect of aerosol on clouds, precipitation, and moisture transport on small temporal scales, and more research on the longterm effect of biomass burning aerosol on hydroclimate is needed.
Progress in understanding aerosol influences on precipitation has also been limited by the influence of internal variability on aerosol and cloud distributions, which can contribute to the opposing results discussed above.In analyses of short timescales, even several years, internal variability can impact the pathways of aerosol transport and alignment with clouds [22].Recent community efforts have produced large ensembles of simulations (e.g.Kay et al [23]), which provide new opportunities to better account for the influences of internal variability.In this study, we aim to isolate the mechanisms by which biomass burning aerosol affects tropical precipitation, particularly in South America and SEA, during the late 20th century relative to the preindustrial period.We utilize single-forcing experiments from the Community Earth System Model version 1 (CESM1) Large Ensemble experiment to evaluate the effect of biomass burning aerosols on precipitation and identify the long-term effects of biomass burning on regional hydroclimate while also accounting for internal variability.These experiments allow us to determine how biomass burning aerosol perturbs regional hydrological cycles independently from anthropogenic aerosol and GHG forcing.

Methods
To evaluate the hydroclimatic effects of biomass burning aerosol, we utilize output from the Community Earth System Model version 1 (CESM1) Large Ensemble (LENS) experiment [23][24][25].CESM1 is a fully-coupled climate model with the Community Atmosphere Model version 5 (CAM5) as its atmosphere component [26].CAM5 is an advancement from previous versions (e.g.CAM4) partially due to its more complex treatment of aerosol and clouds, which enable an explicit representation of aerosolcloud interactions and indirect effects [27].The twomoment cloud microphysics scheme predicts cloud droplet/ice number concentration based on aerosol concentrations and chemical properties, thus linking aerosol characteristics to clouds and precipitation [8,25,27].CAM5 uses the three-mode version of the Modal Aerosol Model (MAM3), which accounts for coarse, Aitken, and accumulation aerosol size distributions, and includes six aerosol species (seasalt, dust, sulfate, primary organic aerosol, secondary organic aerosol, and BC) [28].The inclusion of aerosol effects significantly impacts resultant atmospheric phenomena such as cloud microphysics and precipitation.For example, CMIP5 models that include prognostic two-moment cloud microphysics schemes, like the one in CAM5, show decreased precipitation over the 20th century that is absent in models with simpler aerosol schemes [4].
CESM1 has previously been shown to reasonably reproduce observed large-scale patterns and trends for many important aspects of the climate system, including atmospheric circulation, cloud microphysics, surface temperature, and precipitation [23,25,27,[29][30][31].For this study, we are primarily interested in the relationship between precipitation and aerosol, and briefly validate the end of 20th century climatological seasonal cycles of CESM1 simulated precipitation and aerosol optical depth (AOD) against observations in supplemental figures S1 and S2.This analysis focuses on four regions: Northwest South America (NWSAM), Southwest South America (SWSAM), Central South America (CSAM), and SEA.For precipitation, CESM1 exhibits a dry bias throughout the year in CSAM and stronger amplitude in SWSAM (i.e.drier dry season and wetter wet season), but the overall seasonal timing in all regions is well represented.For AOD, CESM1 compares well to reanalysis in all regions, except an underestimation of AOD in SEA (see figure 1(e) for regions).These discrepancies are consistent with previous research revealing the presence of a dry bias in South America [32,33] and underestimation of AOD [34,35] in climate models when compared to observations.Additionally, while it would be ideal to further compare simulated changes to observed changes over the 20th century, this is not possible due to the lack of observations in the early 20th century in the tropics [36], so we focus on the model results to evaluate mechanisms driving long-term changes.In the results, we analyze simulated historical changes with consideration of the known biases, but note that general consistency in the timing of biomass burning aerosol and precipitation indicates that CESM1 likely captures important aspects of the mechanisms involved.
The simulations used in this study include both the CESM1-LENS historical (HIST) and singleforcing experiments (CESM1-LENS-SF) [5,23].The LENS experiments allow us to account for internal variability, which is assessed here as the variation across ensemble members.For further information about the structure of the CESM1-LENS project and its utility in isolating trends from internal variability, we refer readers to Kay et al [23], which describes the experimental design of the project, and Deser et al [5], which illustrates how trends due to external forcing are detected across ensemble members.The CESM1-LENS-SF simulations are complementary to the CESM1-LENS historical simulations and can be used to isolate the effect of individual climate forcings on various Earth system phenomena [5].The individual climate forcings considered in CESM1-LENS-SF are biomass burning aerosol (BMB) and anthropogenic aerosol (AER) emissions, and GHGs concentrations.Biomass burning aerosol emissions are based on the observational data compiled by Lamarque et al (2010), which serves as an emissions input for all CMIP5 models [26,35,37].The simulations begin in 1920 and one of these three forcings is held constant at 1920 levels while the rest vary following the historical simulations over the 20th century [5].There are 35 ensemble members for HIST, 20 for AER and GHG, and 15 for BMB; to ensure consistency in our assessment of internal variability with the BMB 15-member ensemble, we use only the first 15 ensemble members for AER, GHG, and HIST [5].The model resolution for all simulations is 0.9 • × 1.25 • .These simulations are unique because they are one of the only ensemble single-forcing experiments to isolate biomass burning aerosol separately from anthropogenic aerosol.
In our analysis, we subtract the simulations that held an individual forcing constant from the historical simulations under present day conditions to isolate the effect of the single forcing on the variable of interest.We compare this to the change from the historical simulation for the last 20 years (1985-2005, hereafter referred to as present day, PD) minus the first 20 years (1920-1940, hereafter referred to as preindustrial, PI) of the simulation.The definition of preindustrial used here is based on the CESM1-LENS-SF experiment design, which holds one forcing fixed at 1920 levels.Although this definition of preindustrial differs from other analyses that use 1850, it captures most of the biomass burning increase that occurs after 1950.Some studies have raised concerns about the assumption of linearly subtracting the excluded simulation from the control [38]; however, Xu et al determined that under low warming, which holds for the 20th century, the linearity assumption is reasonable [39].The statistical significance of the variation across ensemble members is assessed at the 95% confidence level using a two-tailed t-test.To support the CESM1 results, we analyze a subset of CMIP5 models and the single-forcing simulations from the CESM2-LENS experiment, which are further described in the supporting information.

Precipitation trends
CESM1 HIST shows a significant reduction in annual mean precipitation over tropical land in the 20th century (figure 1(b)).This trend is consistent across all ensemble members (figure S4) and indicates that the precipitation trend is robust with respect to internal climate variability.Additionally, the precipitation reduction in CESM1 HIST is consistent with reduced historical precipitation in selected CMIP5class climate models that also included interactive aerosol schemes (figure 1(a); text S1).The results from single-forcing experiments (figures 1(c)-(e)) suggest that the precipitation reduction is driven by a combination of climate forcings, particularly BMB and AER (figures 1(c) and (d)), with an offset by GHG (figure 1(e)) in some regions.In HIST and BMB, the largest precipitation reductions occur over tropical South America and SEA, and the contours of BC + POM burden increase overlap closely with precipitation change (figures 1(b) and (c)).The same regional precipitation trends are also evident in the CESM2-LENS HIST and BMB single-forcing experiments (figure S5), which is noteworthy because CESM2 includes enhanced representation of primary carbon (BC and POM) [40].Overall, the consistency across CESM1, CESM2, and CMIP5 ensembles provides confidence in this pattern of change.
To better understand the mechanisms driving these precipitation changes, we focus on three regions of interest, NWSAM, SWSAM, and SEA (figure 1(e)).These regions have annual mean reductions of −0.61, −0.20, and −0.50 mm d −1 , respectively (table 1), with BMB accounting for over 60% of the precipitation reduction in NWSAM and SWSAM.All regions have a strong seasonal cycle of both precipitation and biomass burning aerosol burden (BC and POM), which tends to peak during the dry season (figure 2).Changes in the seasonal cycle of precipitation from HIST closely correlate with the BMB change due to significant increases in biomass burning since the preindustrial period (figure 2). Figure 2(a) indicates that the precipitation reduction in NWSAM is consistent throughout the year, despite the BC + POM change being smaller than other regions, which suggests that precipitation suppression is a consequence of large-scale regional phenomena.Precipitation changes in SWSAM are likely influenced by both local and non-local processes; precipitation reduction from July to December overlaps with the local biomass burning season, suggesting local aerosol   and the present-day BMB contribution in each region for cloud microphysical and radiation variables.The bold italicized numbers indicate where the change in the specified variable is significant relative to the preindustrial control at the 95% confidence level according to a two-tailed t-test.Regions used for averaging are indicated in figure 1 influences are important, while precipitation reductions from January to May likely result from regional rather than local biomass burning effects.We also consider the CSAM region (figure 1(e)) due to its high biomass burning aerosol burden in HIST and BMB (figure 2(g)).However, the annual mean precipitation change in this region is small (figure 1(b)), despite having a large biomass burning aerosol change, so the aerosol influences in this region are mostly non-local.
In SEA, the biomass burning and local dry season is historically from August to November, and precipitation decreases from July to December overlap well with the biomass burning season.The large precipitation decrease in July is evident in both HIST and BMB and coincides with the start of the main biomass burning season, so precipitation in SEA is likely impacted by local biomass burning.

Aerosol-cloud and aerosol-radiation interactions
To analyze the mechanisms through which biomass burning aerosol contribute to precipitation reductions over the 20th century, we first examine regional average changes in atmospheric properties associated with cloud microphysics and radiation.To assess the links between cloud microphysical properties and precipitation efficiency, we quantify changes cloud droplet number and cloud droplet radius.
These results indicate that aerosol-cloud interactions, driven by increased aerosol concentrations, contribute significantly to precipitation suppression, though the extent of the contribution varies by region (table 1).Dramatic increases in aerosol burden due to BMB (figure 2, bottom row) significantly increase the number of CCN, leading to higher cloud droplet concentrations and smaller droplet radius compared to the preindustrial period (figure S6; table 1).These changes in cloud microphysical properties due to biomass burning aerosol are significant at the 95% confidence level, and the BMB contribution represents most of the HIST change in all regions (table 1).The largest droplet radius reduction occurs in NWSAM, but there are statistically significant reductions in droplet radius in all regions; BMB accounts for 46.4%, 72.3%, and 68.8% of the historical reduction in droplet radius in SEA, NWSAM, and SWSAM, respectively.In all three regions, BMB is responsible for 57% or more of the increase in cloud droplet number, confirming that increased aerosol concentrations from recent BMB activity have the dominant influence on changes in the microphysical properties of tropical clouds and associated precipitation efficiency over the historical period.We additionally consider whether aerosolradiation interactions contribute to biomass burninginduced precipitation suppression throughout the 20th century.While indirect effects can increase the lifetime of clouds [17], aerosol radiative effects can trigger reduced convection and cloud formation, resulting in overall reductions in time-averaged cloud fraction [10,14,15,19].We evaluate several parameters that pertain to aerosol-radiation interactions, including changes in cloud cover and net surface radiation flux changes.Low cloud fraction is reduced in each region and high cloud fraction is reduced in SEA and NWSAM at 95% confidence in BMB (table 1).The cloud reduction is indicative of the semidirect effect of aerosols (i.e.cloud burn-off) and direct/indirect effects on surface cooling (i.e.reduced downwelling solar radiation), both of which act to stabilize the atmosphere and reduce convection [10,14,15].Atmospheric stability can contribute considerably to precipitation reduction on daily to weekly timescales; studies have demonstrated that reduced radiation at the surface due to high aerosol concentrations limits surface energy available for convection, thus inhibiting convective activity [10,13,15,41].While the CESM1-LENS-SF dataset does not provide output parameters directly related to convective activity (e.g.CAPE), convective suppression is implicated in these results via the net reduction in downwelling shortwave radiation at the surface (table 1) coupled with increased solar heating rate in the lower atmosphere in all regions (figure S7).Based on the literature and our findings, we deduce that reduced surface radiation likely inhibits convection due to reduced energy available for convection, while increased atmospheric heating (i.e.figure S7) enhances stability and prevents development of clouds and precipitation [10,[13][14][15]41].Both aerosol-cloud and aerosol-radiation interactions likely act together to suppress precipitation in the regions of interest, but it is difficult isolate the dominating mechanism in this experiment; however, previous work shows that aerosol-cloud interactions become saturated at an AOD of greater than 0.3, after which aerosol-radiation interactions dominate [14,15].The present-day AOD averages in figure S2 suggest that aerosol concentrations are in a range where both mechanisms of precipitation suppression are important, but one may dominate at different times of year and differ between regions.

Circulation and precipitation reduction
Biomass burning aerosol can also have a non-local influence on precipitation by perturbing regional moisture transport.In NWSAM, SWSAM, and SEA, changes in vertically integrated moisture transport and moisture convergence coincide with significant precipitation reductions (figure 3).Moisture over northern South America is sourced from the Atlantic Ocean and Caribbean Sea, and it is transported from east to west across the northern part of the continent before it turns south along the slopes of the Andes and travels over Peru and Bolivia via the South American low-level jet (SALLJ) (figure S8).Over northern South America, moisture transport is slowed onto the continent from the Atlantic Ocean and Caribbean Sea, and moisture export from NWSAM to the Pacific Ocean is enhanced in both HIST and BMB.Within the continent, moisture is also sourced from precipitation recycling, which can be reduced due to deforestation [42,43].In addition to deforestation, biomass burning also reduces surface water evaporation in these simulations, further affecting inland moisture recycling (table 1, QFLX).Decreased moisture transport and recycling in SWSAM likely contributes to the decrease in moisture convergence and precipitation over the Andes from ∼10 • S to ∼20 • S evident in both HIST and BMB.Similar changes occur over SEA; moisture transport from the eastern Indian Ocean and Karimata Strait to southern Sumatra and Borneo is reduced relative to the background flow (figure S8).Reduced moisture convergence, such as that depicted by the orange regions in figure 3, leads to enhanced moisture transport downwind away from the region, thus diminishing precipitation in the regions of interest.This pattern of regional moisture transport change is consistent with reduced precipitation efficiency as well as large-scale circulation anomalies associated with atmospheric heating and surface cooling due to high biomass burning aerosol and convectively coupled circulation responses.Increases in low-to mid-level solar heating rate in all regions (figure S7) in both HIST and BMB suggest that absorbing aerosol (e.g.black carbon) contributes to atmospheric warming in the aerosol layer.A warming atmosphere and cooling surface due to absorption of solar radiation reduces low-level convergence, which weakens local convection and alters large-scale pressure gradients and circulation.
Over SEA, the reduction in moisture transport is partially due to a weakening of the land-sea temperature contrast and associated sea breezes, which have been found to reduce precipitation over Borneo during intense biomass burning events [44].The pattern of reduced moisture transport in SWSAM due to increased biomass burning aerosol is consistent with the weakening of the South American Monsoon dryto-wet season transition [45,46].The largest precipitation reduction in SWSAM occurs during the transition to the wet season from September to November (figure 2(b)), which is also the period with the highest biomass burning aerosol burden.This is consistent with previous work which has revealed that reduced surface radiation due to high aerosol loading delays the transition to the wet season and reduces moisture transport and thus precipitation and monsoon onset during September-November [10,46].
In NWSAM, the overall pattern of opposing directions of moisture transport change between the eastern and western sides of the continent suggests a weakening of the convectively coupled circulation in the region [47].There is reduced moisture convergence and transport into NWSAM to the east and increased transport out of NWSAM to the west.This indicates a weakening of the background flow, which normally features high moisture convergence and vigorous convection over Colombia triggered by orographic lifting and low-level jets [48].While it can be difficult to quantify cause and effect associated with convectively coupled circulation responses, it is likely that the dramatic increase in biomass burning aerosols over NWSAM triggers anomalous lower atmosphere heating (figure S7); this stabilizes the atmosphere and, when combined with reduced precipitation efficiency, acts to decrease convective heating in the region.This initial reduction in convection is enhanced by the convectively coupled circulation response, which triggers further reduction of moisture transport from the east.Suppressed convection in NWSAM not only reduces local precipitation but also downstream moisture transport via a weakening of the SALLJ, and thus contributes to lower precipitation over SWSAM as well.

Conclusion
Biomass burning aerosol suppresses precipitation over South America and SEA through multiple interconnected processes.Increases in biomass burning aerosol burden over the 20th century led to higher CCN concentrations and yielded smaller cloud droplets, increased solar heating in the aerosol layer, and perturbed regional moisture transport.Aerosolcloud interactions are prevalent in South America and SEA due to the reduction in droplet radius and increase in droplet number resulting from higher aerosol concentrations, while aerosol-radiation interactions contributed to reduced cloud fraction and increased atmospheric stability.Analysis indicates that both aerosol effects contribute to reduced precipitation simultaneously, making it difficult to discern which process dominates [49].Our results are consistent with previous studies which demonstrate that both the microphysical and radiative effects suppress precipitation over the tropics [10,14,15].Changes in moisture transport due to biomass burning are regionally complex and have received less attention in the literature, but our findings suggest that biomass burning aerosol slows regional moisture transport, thus contributing to reduced precipitation.Additionally, these results agree with studies that have assessed changes in circulation and moisture transport in response to biomass burning aerosol [9,21].For example, Zanis et al show that absorbing anthropogenic aerosols reduce the 850 mb wind in NWSAM and SWSAM relative to preindustrial conditions [9].
Our study contributes to a better understanding of the impact of biomass burning aerosol over the tropics on a long timescale.In NWSAM and SWSAM, for example, 61%-62% of the reduction in precipitation in the full historical simulation is attributable to increased biomass burning at the end of the 20th century.These results indicate that the direct, semidirect, and indirect effects of aerosols and reductions in moisture transport all contribute to significant precipitation suppression, which offers insight into the long-term consequences of increased biomass burning.Over the 21st century, tropical precipitation is expected to increase over NWSAM and SEA due to both GHG warming and vegetation responses to rising CO 2 [50], and this trend could be further amplified if expected decreases in aerosol emissions reverse the 20th century drying effects demonstrated here, though projections of biomass burning activity remain uncertain.This study advances awareness of aerosol influences on hydroclimate, thus paving the way for more accurate projections of future climate and constraints on drought and flood risk in regions prone to biomass burning.

Figure 1 .
Figure 1.Spatial distribution of changes in present day (PD,1985-2005) minus preindustrial (PI,1920-1940) precipitation for CMIP5 models with interactive aerosol (see supplemental information for list) (a) and CESM1 HIST (b), and PD HIST minus PD single-forcing for BMB (c), AER (d), and GHG (e).Contour lines indicate increases in BC and POM burden from 1 to 5 mg (blue) relative to the control period, and reductions in burden of −1 to −5 mg (brown).Stippling in (a) indicates where 7/10 models agree on sign of precipitation change, and stippling in (b)-(e) indicates a significant change at the 95% confidence level.The purple squares in the bottom panel indicate the areas over which the data for figure 2 and table 1 are averaged.See supplemental figure S3 for the same figure, but with BC and POM burden magnitude and statistical significance demonstrated.

Figure 2 .
Figure 2. Seasonal cycle of precipitation (a)-(d) and biomass burning aerosol (BC + POM) burden (e)-(h) in Northwest South America (NWSAM), Southwest South America (SWSAM), Central South America (CSAM), and Southeast Asia (SEA).Regions used for averaging are indicated in figure 1(e).Changes (right-hand y-axis) are indicated by dashed lines and represent the present day (PD,1985-2005) minus preindustrial (PI,1920-1940) periods for HIST (green) and PD contribution from BMB (brown).The left-hand axes vary for each panel, but the right-hand axes have the same scale for (a)-(d) and (e)-(h).

3 .
Change in vertically integrated moisture transport (m/skg/kg, arrows) and moisture convergence (mm d , colors) for HIST and BMB.Black arrows show where changes in moisture transport are statistically significant at the 95% confidence level.Arrows depict the change in moisture transport and are scaled to represent values that are 15 times smaller than the background transport depicted in figure S8.