The Amazon rain forest experiences the combined pressures from human-made deforestation and progressing climate change, causing severe and potentially disruptive perturbations of the ecosystem's integrity and stability. To intensify research on critical aspects of Amazonian biosphere–atmosphere exchange, the Amazon Tall Tower Observatory (ATTO) has been established in the central Amazon Basin. Here we present a multi-year analysis of backward trajectories to derive an effective footprint region of the observatory, which spans large parts of the particularly vulnerable eastern basin. Further, we characterize geospatial properties of the footprint regions, such as climatic conditions, distribution of ecoregions, land cover categories, deforestation dynamics, agricultural expansion, fire regimes, infrastructural development, protected areas, and future deforestation scenarios. This study is meant to be a resource and reference work, helping to embed the ATTO observations into the larger context of human-caused transformations of Amazonia. We conclude that the chances to observe an unperturbed rain forest–atmosphere exchange at the ATTO site will likely decrease in the future, whereas the atmospheric signals from human-made and climate-change-related forest perturbations will increase in frequency and intensity.
The Earth is increasingly shaped by human activities (Crutzen, 2002; Song et al., 2018). Concerning the atmosphere, global climate change and air quality impacts on human health are two of the most important recent consequences (e.g., Stocker et al., 2013; Lelieveld et al., 2015; Cheng et al., 2016; Reinmuth-Selzle et al., 2017). The Amazon rain forest and its atmosphere are particularly vulnerable since they are experiencing the combined pressures from human-made deforestation and progressing climate change (Lenton et al., 2008; Malhi et al., 2008). Davidson et al. (2012) presented comprehensive perspectives on the ecological and atmospheric “transition” of the Amazon biome due to continuous land use change and a cascade of related perturbations and feedbacks. Particularly, the hydrological cycle with its large amounts of recycled water and energy represents an Achilles' heel in the ecosystem's integrity and stability (e.g., Andreae et al., 2004; Rosenfeld et al., 2008; Hilker et al., 2014; Machado et al., 2018). The Amazon is defined by a pronounced continental gradient in climatic conditions, socioeconomic activities, and land use change. Climatically, the northwestern part is characterized by high precipitation rates with comparatively weak seasonal amplitudes, whereas the southeastern part experiences a much stronger seasonality, associated with dry season drought stress for the vegetation (e.g., Malhi et al., 2008, 2009). Socioeconomically, the northwestern part is protected by its remoteness and, therefore, still mostly unperturbed, whereas the southeast is heavily influenced by infrastructure development, logging, and agro-industrial expansion (e.g., Soares-Filho et al., 2006; Nepstad et al., 2008; Silva et al., 2013). The regional and global consequences of the Amazon's transition process for the Earth's climate system, water resources, biodiversity, and human health are still widely unknown.
To address the mechanisms and consequences of the anthropogenic perturbation of the Earth atmosphere, a sound understanding of the starting point – the background state – of this transition process is required. However, regions of definable background state conditions are becoming increasingly rare worldwide (Andreae, 2007; Hamilton et al., 2014). To some extent, the Amazon Basin still represents one of the last continental exceptions and, thus, a unique outdoor laboratory for atmospheric science. Certain – although short – episodes in its clean wet season still open a window into the preindustrial and unpolluted past, while the dry season is influenced by heavy pollution from numerous deforestation and land management fires (Martin et al., 2010b; Andreae et al., 2015; Pöhlker et al., 2018). The Amazon Tall Tower Observatory (ATTO) has been established in the Amazon Basin for two main reasons: first, it supports a better understanding of key processes in biosphere–atmosphere exchange and, therefore, helps to assess the global relevance of the Amazon's ecosystem services. For this task, the frequent occurrence of very clean episodes provides crucially important baseline data to approximate the era before globally pervasive anthropogenic pollution. Referring to Andreae et al. (2015) it is “urgent to obtain baseline data now, to document the present […] conditions before upcoming changes, especially in the eastern part of the basin, will forever change the face of Amazonia”. Second, the ATTO research documents the progressing change in the Amazon and, thus, provides essential knowledge to try to avoid irreversible damage to this unique ecosystem. The extent and complexity of meteorological, trace gas, aerosol, and ecological studies at the ATTO site are steadily increasing, promising more and more insights into the manifold facets of biogeochemical and hydrological cycles in this unique ecosystem (e.g., Nölscher et al., 2016; Rizzolo et al., 2016; Wang et al., 2016a, b; Chor et al., 2017; Oliveira et al., 2018; Yáñez-Serrano et al., 2018).
Detailed knowledge on the spatial and temporal variability of the site's footprint region and, thus, the effectively probed land cover mosaic is a prerequisite to embed atmospheric observations at ATTO into a broader Amazonian context. In the course of our recent studies, the analysis of backward trajectories (BTs) and geographic information system (GIS) data helped substantially to explore air mass history and the variability of atmospheric composition (e.g., Moran-Zuloaga et al., 2018; Pöhlker et al., 2018; Saturno et al., 2018a, b). Along these lines, the current study presents a systematic BT and GIS data analysis, providing a robust characterization of spatiotemporal patterns in the advection of air masses towards ATTO, relevant hydrological regimes, and current land use patterns and future trends in the ATTO footprint region. We envision that this work may serve as a helpful resource and look-up reference for the interpretation of current and future observations in the region. Furthermore, we conclude with a discussion on anticipated future developments within the ATTO footprint region in response to progressing climate and land use change, which are influences of crucial importance for future ATTO research.
The remote ATTO site (position of 325 m tall tower, operational since 2017:
2.1459
The systematic backward trajectory (BT) analysis is based on the Hybrid
Single-Particle Lagrangian Integrated Trajectory model (HYSPLIT, NOAA-ARL)
with meteorological input data from the Global Data Assimilation System
(GDAS1, 1
As a basic data set, 14 d BTs were calculated every 1 h for
different starting heights at ATTO for a multi-year period, namely 1 January 2008 until 30 June 2016. This time period from January 2008 to June 2016
was chosen for two main reasons: (i) it includes 8.5 years of
BT data, which is a comparatively broad statistical basis to retrieve
characteristic BT clusters for the ATTO region. (ii) It includes the
AMAZE-08 campaign at the ZF2 site (February and March 2008; Martin et al., 2010a),
which brought together a comprehensive set of instrumentation and represents
the starting point of continuous atmospheric observations in the region,
operational since 2008 at the ZF2 site and also since 2011 at the ATTO site.
The 14 d HYSPLIT backward trajectory (BT) ensembles (starting
height 1000 m a.g.l.) with color coding representing BT transport times:
Figure 1 shows two exemplary 14 d BT ensembles for selected wet and dry
season months. In the context of the present study, we focused on 3 and
9 d BTs, which are shortened versions of the initially calculated 14 d
files. The 3 d BTs cover the air mass movements over the South American
continent and, therefore, represent the continental sources that are
presumably most relevant for the ATTO site. Figure 1 illustrates that 3 d
BTs in the wet and dry seasons typically span from ATTO to the Brazilian
coastal regions and, in some cases, even onto the Atlantic Ocean. The 9 d
BTs cover the entire Atlantic region. On these timescales, they provide (at
least some) information on potential source regions of long-range transport
(LRT) aerosol from the African continent. However, the relative error of the
BT model scales with the absolute length of the BTs. The uncertainty of the
location of the trajectory center of gravity of the contributing air masses
is estimated to be 15 %–30 % of the travel distance
(
The 14 d BTs were calculated for starting heights of 80, 200, 1000, 2000, and 4000 m a.g.l. (above ground level). The overall BT variability of all starting heights has been compared and summarized in Sect. 3.1 and 3.2. However, only the BTs started at 1000 m a.g.l. were used systematically throughout this study and particularly for the GIS analysis in Sect. 3.3. As a result of the rather simple terrain upwind and because of the boundary layer (BL) height variability, this arrival height is representative for sampling heights at ATTO. A comparison of BT ensembles at 200 m vs. 1000 m a.g.l. gave similar results, indicating that the analysis was not very sensitive to varying start heights within the lower 1000 m of the troposphere (for further details see Sects. 2.4 and 3.2). Ensembles of BTs were converted into maps of relative trajectory densities by IGOR Pro routines as illustrated in Fig. S1 in the Supplement. Slightly modified, this procedure was also used to obtain averaged trajectory heights. The BT density maps are called “air mass residence time maps” throughout this study.
Precipitation was available through GDAS (HYSPLIT model) for every hourly
data point of the individual BTs. Based on these data, cumulative
precipitation,
Trajectory models have been constantly improved, while gridded meteorological data became more sophisticated (Gebhart et al., 2005). In addition to the HYSPLIT model, the FLEXible PARTicle dispersion (FLEXPART) model is frequently used in the atmospheric sciences. The FLEXPART model is a Lagrangian transport and dispersion model to simulate long-range and mesoscale transport, diffusion, and dry and wet deposition. A detailed description can be found elsewhere (Stohl et al., 1998, 2005; Stohl and Thomson, 1999).
The present study is mostly based on HYSPLIT BT data. The HYSPLIT BT ensembles combine the trajectories representing the center of gravity (“lines”). We conducted a comparison with FLEXPART for two periods, March 2014 and September 2014. The FLEXPART model accounts for lateral mixing (or dispersion in forward mode), provides distributions of residence times of air masses, and, hence, reflects the relative importance of potential source areas. The same starting heights and time frames were applied in both models. The purpose of this comparison was to validate the results of line trajectories (HYSPLIT model) compared to backward modeling accounting for dispersion. The HYSPLIT BT ensembles for the case study periods comprise 9 d BTs, started every 1 h at ATTO at altitudes of 200 and 1000 m a.g.l. during the months March and September 2014. Similarly, the FLEXPART BTs were started at ATTO at altitudes of 200 and 1000 m a.g.l. with a backward integration time of 9 d, spanning the months March and September 2014.
Based on the entire set of 74 496 3 d BTs (1 January 2008–30 June 2016) we
conducted a
An analysis of the total within-cluster sum of squares as a function of
In this study, we have chosen
The footprint concept has been introduced for atmospheric measurement sites to quantify the distribution and extent of biosphere–atmosphere exchange in their surroundings, which contributes to the variability of observed trace compound concentrations (Gloor et al., 2001). Schmid (2002) formally defined the footprint (sometimes also called airshed or effective upwind fetch) of a measurement as “the transfer function between the measured value and a set of forcings on the surface–atmosphere interface” as a general description of various footprint modeling frameworks. Footprints have been mostly used in the context of long-term trace gas observations (e.g., greenhouse gases) (e.g., Thompson et al., 2009; Winderlich et al., 2010). The size of a footprint largely depends on the height of the measurement, and tall tower sites are known to cover large footprint regions, which makes them particularly valuable for representative regional monitoring (Gloor et al., 2001). The geographic distribution of a footprint is defined by the distribution of relevant sources and the predominant wind directions.
The modeling of specific footprints for the ATTO observations based on
specific source–receptor relationships (i.e., for specific compounds and
observational conditions) is beyond the scope of this work and will
presumably be the subject of dedicated future studies. Here, we use the term
footprint in a more simplistic and general sense as the area on the South
American continent that is covered by the air mass residence time maps,
which are based on multiple years of 3 d HYSPLIT BTs. In order to
discriminate our approach from footprint modeling attempts according to
Schmid (2002), we use the term “BT footprint” in this study. The choice of 3 d
for this analysis is justified, for example, by a study by Lammel et al. (2003),
reporting the characteristic formation times of secondary
aerosols of about 48–72 h as well as the fact that coarse-mode particles
“were derived from emissions
Filters applied to the multi-year BT ensemble. Height filter crops segments along individual BTs if values exceed the height threshold (e.g., height of 1000 m; named “H1000”). En route convection filter crops segments along individual BTs if sun flux values are below threshold (named “Cer”). Convection filter at ATTO removed entire BTs from analysis if sun flux upon arrival (i.e., in the ATTO pixel) is below threshold (named “Catto”).
The sensitivity of the geographic extent of the BT footprint region towards different BL heights and convective mixing inside the BL has been tested by applying dedicated filters to the BT data. We calculated different versions of the BT footprint region on the basis of multi-year BT ensembles with settings as specified in Table 1. Three types of filters were applied, which introduce increasingly strict conditions into the approach:
Based on these differently filtered BT ensembles defined in Table 1,
corresponding footprints were calculated statistically as outlined below.
The original 1 h trajectory points were interpolated to minute-wise
steps and then counted within a 0.1
With the same statistical approach, the BT cluster footprints of the 15
clusters from the
We conducted a detailed analysis of the land cover type and/or change in the
ATTO BT footprint region using a number of different GIS data sets processed
using the QGIS software package (“Las Palmas” version 2.18.2, QGIS
development team). The QGIS software (formally known as Quantum GIS) is
freely available under
The classification of land cover types as well as the quantification of land cover changes and dynamics is a complex task for various reasons: GIS data sets are often based on different spacecraft instruments and data acquisition methods (e.g., satellite grid resolution, spectral retrievals and sensitivity, image processing, and categorization), which could restrict their comparability. Artifacts such as cloud and/or smoke cover as well as terrain-related shadow effects have to be considered. GIS layers typically differ with respect to their acquisition time frames, which mostly represent states in the past (e.g., GlobCover 2009 representing land cover in the year 2009). Moreover, land cover is subject to dynamic seasonal and phenological changes (e.g., in agricultural lands), which is not covered by all GIS data sets (e.g., Ju and Roy, 2008; Jin et al., 2023; Tyukavina et al., 2017). Thus, all corresponding GIS maps are subject to limitations and uncertainties. In this work, mostly well-established GIS products have been used, which have been documented and discussed in previous studies. In Sect. S1.1, we also point out major uncertainties and restrictions with respect to the comparability of different GIS maps; however, for more detailed information we refer the reader to the referenced studies.
Using scripts written in Python 2.7, an analysis weighted by air mass
residence time of several GIS data sets was conducted to assess the relative
significance of land cover categories, forest cover, and loss as well as
fire events for the ATTO observations. Each data set was weighted using
rasterized BT density maps. The rasterized BT density maps were calculated
with cell sizes of 0.09 the air mass residence time-weighted part of each land cover category
and forest cover in all 15 BT cluster footprints. the air mass residence time-weighted forest loss for each BT footprint
in relation to the forest cover in 2000, per year. the air mass residence time-weighted fire counts for each BT
footprint, per year and land cover category.
The precipitation data used in this study are based on the Precipitation
Estimation from Remotely Sensed Information using Artificial Neural Networks
for Climate Data Record (PERSIANN-CDR) data product (Ashouri et al., 2015),
which has been obtained from Google Earth Engine via
The following anomaly indices for the Pacific and Atlantic sea surface
temperatures (SST), which were obtained from
Anomalies in precipitation for the aforementioned regions as well as for the BT frequency of occurrence have been calculated as the relative differences of the monthly averaged values to a multi-year monthly mean. For PERSIANN-CDR precipitation anomalies, the reference time frame spans from January 1983 to December 2016. For BT frequency of occurrence anomalies, the reference time frame spans from January 2008 to June 2016.
The results and discussion section of this paper consists of two main
parts:
Sect. 3.1 and 3.2 summarize the large-scale geographic patters and
seasonal variability of the ATTO BT ensembles as well as their links to
precipitation regimes and selected teleconnections. Sect. 3.3 defines a BT-based footprint region of the ATTO site and relates
it to the current state and anticipated future change of the covered land
use mosaic. Sect. 3.3.1: climatic conditions, biomes, ecoregions, and the “last-of-the-wild”; Sect. 3.3.2: land cover; Sect. 3.3.3: deforestation and agro-industrial expansion; Sect. 3.3.4: fires; Sect. 3.3.5: infrastructure, cities, traffic, and mining; Sect. 3.3.6: protected areas; Sect. 3.3.7: deforestation and climate change scenarios.
In particular, Sect. 3.3 is meant to be a resource and look-up reference
summarizing ATTO-relevant land cover information subdivided into the
following categories:
All seven subsubsections, Sect. 3.3.1 to 3.3.7, begin with a concise literature
synthesis section and then relate the discussion to its specific relevance
for the ATTO research. Due to its length, the entire Sect. 3.3 has been
structured and written in a way that facilitates nonlinear reading.
All BT and GIS results obtained for the ATTO site also generally apply to
the nearby ZF2 site, due to the fact that both sites are located close
enough to each other (straight-line distance 144 km) to be influenced by
similar (large-scale) circulation patterns (Fig. S6). During wet and dry
season conditions, the air masses first pass ATTO before reaching the ZF2
region after
HYSPLIT backward trajectory (BT) ensembles showing the large-scale
trade wind circulation in the Atlantic region and the pronounced seasonal
oscillation between northern and southern hemispheric influence at ATTO by
means of air mass residence time maps (
The annual north–south oscillation of the intertropical convergence zone (ITCZ) defines the large-scale trade wind circulation patterns in the Atlantic region, which govern the atmospheric seasonality in the central Amazon (compare Martin et al., 2010b; Andreae et al., 2012, 2015; Moran-Zuloaga et al., 2018). Here we conducted a multi-year HYSPLIT BT analysis (for five different starting heights: 80, 200, 1000, 2000, and 4000 m) to visualize the large-scale trends in the ITCZ-related air mass advection towards ATTO with respect to BT geographic patterns and transport altitudes (Figs. 2 and S7). For the lower starting heights (i.e., 80, 200, and 1000 m), the overall circulation pattern is predominantly defined by the seasonal ITCZ oscillation and shows two comparatively narrow paths: a northeasterly path during the wet season (February–May) and a southeasterly path during the dry season (August–November). The center of the northeasterly path spans straight from ATTO to the Cabo Verde Islands and the northwest African coast (area of Mauritania and Western Sahara). The air mass transport from the western African coast towards ATTO takes approximately 6–7 d (Fig. 1). The center of the southeasterly trajectory track represents a curved circulation pattern, which is directed eastwards over the mouth of the Amazon River and then curves towards the southeast along the Brazilian coast. For increasing start heights, the separation into distinct northeast and southeast paths becomes more and more smeared out. While the northeast and southeast paths are still somewhat resolved for the 2000 m case, the separation mostly disappears for 4000 m. The observation of the tightest BT bundles close to the ground and their divergence with increasing altitude results from the Hadley cell circulation. Its low-level trade winds feed boundary layer air into the deep-convective ITCZ belt as confluence between northeast and southeast trades (Talbot et al., 1990; Shpund et al., 2011; Dudley et al., 2012; Makowski Giannoni et al., 2016).
Transport within the northeast and southeast BT paths generally occurs at comparatively low altitudes for all start heights (Figs. 2 and S7). This is clearest for the northeasterly circulation during the wet season with its low-level trades. The average transport height of the southeasterly trades tends to be somewhat higher than the northeasterly trades, which can be seen for all starting height cases. For example, at a BT starting height at 200 m (Fig. 2b), the northeast trades are mostly located below 300 m over the Atlantic Ocean, whereas the southeast trades range mostly up to 800 m. At a BT starting height of 1000 m (Fig. 2d), the northeast trades are mostly below 800 m, whereas the southeast trades are mostly below 1400 m. A side aspect in the context of the low-level air mass circulation is that the HYSPLIT BT results (for starting heights 80 and 200 m) indicate a topography effect over the northeast basin. On their way to ATTO, the air masses tend to rather closely follow the Amazon River valley (Fig. S8), underlining the relative importance of trace gas and aerosol sources along the river as discussed in more detail in Sect. 3.3. A potential major uncertainty to keep in mind is the influence of local breeze circulations at the large rivers in the region that could significantly alter BTs in lower atmospheric layers. However, in a previous study we found that HYSPLIT performed quite well in reproducing the river breeze circulation over the Amazon River near Manaus (Trebs et al., 2012)
Comparison of FLEXPART (blue) and HYSPLIT (red) back trajectory
results for wet season (1–31 March 2014, BT start height 1000 m a.g.l.,
The HYSPLIT BTs were found to be a useful tool in the context of this study to analyze temporal and spatial trends of atmospheric circulation patterns. However, it has to be kept in mind that the individual BTs represent a simplified picture by providing center of gravity lines of the transported air parcels, not accounting for dispersion. To assess the relative importance of dispersion for the large-scale circulation patterns, we compared the HYSPLIT BTs with corresponding FLEXPART results (accounting for dispersion) for wet and dry season periods. Selected results are shown in Fig. 3 and illustrate that dispersion yields substantial differences between the HYSPLIT and FLEXPART outputs. Generally, the FLEXPART BTs cover a much larger area than the HYSPLIT BT ensembles, which is particularly obvious over the Atlantic region. The comparison further illustrates that deviations become larger the further the BTs reach into the past, as lateral mixing (i.e., dispersion in forward mode) becomes more significant. With respect to air mass transport over the northeast basin (i.e., from ATTO to the Brazilian coast), however, the main BT paths of both model outputs appear to be relatively similar.
Map of northeast Amazon Basin with 15 clusters from systematic
The
The first group includes three NE trajectory clusters (i.e., NE1, NE2, and
NE3), which intersect the coastline in the region of French Guiana and then
pass over forest areas towards ATTO. All three trajectory clusters follow
roughly the same geographic track; however, they represent different wind
speed regimes: the longest – and therefore on average fastest – cluster
NE3 spans a distance of
The second group includes four ENE BT clusters (i.e., ENE1, ENE2, ENE3, and
ENE4), which intersect the Atlantic coast north of the Amazon River delta
(over the Brazilian state of Amapá). These clusters also represent
different wind velocities, with cluster ENE1 being, on average, the slowest
(
The third group includes four E BT clusters (i.e., E1, E2, E3, and E4),
which follow the Amazon River valley. The BT clusters meet the Atlantic
Ocean in the area of the Amazon River delta. In this group, cluster E1
represents the slowest (
The fourth group includes four inland BT clusters in the ESE and southwesterly (SW) directions. The east-southeasterly clusters ESE1, ESE2, and ESE3 cross the states of Pará and Maranhão and (on average) do not reach the Atlantic Ocean during the analyzed 3 d period. Cluster SW1 points from ATTO in the direction of the city of Manaus.
For comparison, Fig. S10b shows the clustering results for the 200 m BT ensemble with a partitioning into 15 clusters that is comparable to Fig. 4. Generally, the clusters for the 200 m starting height are shorter, which can be explained by a higher surface shear and friction as well as topography effects. Overall, the comparison of Figs. 4 and S10b underlines that the observed trends do not vary substantially within the chosen starting height range.
Absolute numbers of individual backward trajectories (BTs) and
their frequency of occurrence, resolved by main directions of BT advections
(i.e., NE, ENE, E, and ESE) as well as for all 15 BT clusters from
The absolute numbers of individual BTs in the 15 clusters as well as their
frequency of occurrence,
Seasonality in precipitation regimes and ATTO-relevant backward
trajectory advection.
The clustering further provides time-resolved information on the frequency
of occurrence and, thus, seasonality of the individual BT clusters, as shown
in Fig. 5 along with the seasonal cycles of selected precipitation products.
In terms of rainfall, the Amazon region shows heterogeneous patterns with
different precipitation regimes. Figure 5a compares the characteristic
seasonality in precipitation rates
For further comparison, we added the cumulative precipitation along the BT
tracks,
In relation to the precipitation regimes, the pronounced seasonality in BT
frequency of occurrence is summarized for the main wind directions in Fig. 5b
as well as for all BT clusters in Fig. 5c. In terms of main wind
directions, the following aspects are worth noting. The ATTO site receives
rather stable advection from the Northern Hemisphere (
Beyond the separation into main wind directions, the representation of
Multi-year variability and anomalies in BT advection in relation to
Atlantic sea surface temperatures.
Figure 6 shows the inter-annual variation and anomalies in BT advection. It
is well known that atmospheric circulation, moisture transport, and
precipitation patterns over Amazonia are linked through teleconnections to
the variability of the tropical Pacific and Atlantic sea surface
temperatures (SSTs) (e.g., Good et al., 2008; Fernandes et al., 2015;
Tyaquiçã et al., 2017). In particular, the tropical Atlantic
meridional gradient has a direct influence on the position of the ITCZ and
trade wind patterns towards and over the basin (e.g., Chiang et al., 2002).
Specifically, a warming of the tropical north Atlantic (TNA) relative to the
tropical south Atlantic (TSA) is associated with a northwards shift of the
ITCZ and corresponding weakening of the northeasterly trades, whereas an
anomalously warm TSA relative to the TNA tends to cause a southwards shift
of the ITCZ and a weakening of the southeasterly trades (e.g., Cox et al.,
2008; Espinoza et al., 2014; Marengo and Espinoza, 2016; Erfanian et al.,
2017). Figure 6 confirms these trends by comparing the anomalies in BT
frequency (i.e.,
Closely linked to their influence on atmospheric circulation in the basin,
teleconnections to Pacific and Atlantic SSTs play a crucial role in the
occurrence of droughts and floods in Amazonia (e.g., Fu et al., 2001; Zeng
et al., 2008; Fernandes et al., 2015; Marengo and Espinoza, 2016). The
Pacific SST variability, which is represented by the Oceanic Niño Index
(ONI), plays a central role in the El Niño–Southern Oscillation (ENSO)
and has a pronounced influence on the Amazonian hydrological cycle (e.g.,
Asner et al., 2000; Ronchail et al., 2002). Periods with a high ONI indicate
El Niño conditions and are typically associated with dry or even drought
years in the central Amazon (e.g., Lewis et al., 2011; Marengo et al.,
2011). A negative ONI indicates La Niña conditions, which are typically
associated with rain-rich years. The Atlantic SST further modulates the
hydrological conditions and can intensify ENSO-related anomalies or even
cause hydrological extremes itself (i.e., anomalously high TNA can cause
droughts, whereas anomalously high TSA can cause floods) (Zeng et al., 2008;
Lewis et al., 2011; Marengo and Espinoza, 2016). The ATTO-relevant long-term
rainfall anomalies for the ROI
Overview map of the Amazon Basin – here represented by the
watershed region of the Amazon River and its tributaries – showing the
location of the ATTO site and the geographic extent of its footprint. The
footprint is represented by (i) an air mass residence time map based on the
entire BT ensemble (color code), (ii) contour lines representing the largest
1 %, 5 %, 10 %, 25 %, and 50 % of air mass residence times,
and (iii) 15 cluster center lines from BT cluster analysis (blue dashed
lines; see Fig. 4). For comparison, versions of footprint region based on
filtered BT ensembles are shown in Figs. 8 and S5. Green areas represent
forest cover map (status 2000) according to Hansen et al. (2013). Red areas represent total forest loss
from 2000 to 2014 according to Hansen et al. (2013). The green thick line represents Amazon
Basin watershed region. The red rectangular shape highlights the region of
interest ROI
The BT analysis defines the areas in the northeastern Amazon Basin that can
be regarded as the ATTO site BT footprint region. Its land cover status and
anticipated future land cover change are the subject of the analysis in the
subsequent sections. As a general overview, Fig. 7 shows the geographic
extent of the Amazon Basin in combination with the air mass residence time
map. The distribution of air mass residence times shows steeply decreasing
values with increasing distance from the ATTO site. The upper 1 % of air
mass residence times cover an area of
The forest cover and forest loss map in Fig. 7 illustrates the pronounced
northwest-to-southeast gradient, with the northwest being mostly unperturbed
and the southeast being subject to intense, large-scale deforestation and
land use change (Davidson et al., 2012). Within this gradient, the forest
loss data emphasize the geographic extent of the so-called arc of
deforestation at the southern and southeastern margins of the Amazon forest,
which has been an active frontier of total forest loss (Morton et al.,
2006). The arc of deforestation spans from southern Pará and
Maranhão in the southeast over Mato Grosso and Rondônia in the south
to Acre in the southwest of the basin. The majority of forest
clearance has been concentrated here over the last decades, mainly driven by
agricultural expansion (Malhi et al., 2008). The following sections will
zoom into the ROI
The footprint shown in Fig. 7 takes the entire BT ensemble into account and serves as the base case footprint throughout the subsequent land cover analysis. It is worth noting, however, that mostly those segments of the individual BTs, which were in convective contact to the ground through BL mixing, matter most to identify regions of particular relevance for the ATTO observations. The same is true for those BTs that arrive at ATTO during convective hours (i.e., excluding conditions with decoupled layers during nighttime) and, thus, may introduce emissions from sources in the footprint region into the ATTO BL through vertical mixing. To estimate the impact of such a “diurnally intermittent mixing” on the ATTO BT footprint, we conducted a sensitivity analysis by applying a sequence of filters to the base case BT ensemble (for details see Sect. 2.5). Overall, the filtering does not substantially alter the geographic extent of the footprint's easterly core regions. Only minor variations in the outer parts of the 25 % contour lines were observed (Fig. S5). This underlines that the base case BT footprint is generally suited to identify regions and land cover types that are of relevance for the ATTO research.
Overview map of the Amazon Basin – mostly identical to Fig. 7 – with refined air mass residence time map based on filtered BT ensemble (color code). The BT filtering was conducted according to case H1000_Cer_Catto as shown in Fig. S5 and outlined in Sect. 2.5. Contour lines represent the largest 1 %, 5 %, 10 %, 25 %, and 50 % of air mass residence times, based on the filtered BT ensemble.
Beyond the general consistency between the footprint regions with and without the convection filters being applied, some characteristic patterns emerge in the corresponding map in Fig. 8, if diurnally intermittent mixing is taken into account. In fact, this BT refinement suggests that certain footprint regions tend to be more relevant for the ATTO observations than others. Specifically, the geographic locations of the convective segments of those BTs that arrive at ATTO during daytime are more relevant than the non-convective (nighttime) segments of the same BTs. Due to the comparatively constant wind directions and velocities, the resulting dashed lines (convective vs. non-convective) of the large number of BTs result in repeating geographic patterns. Figure 8 suggests that the region directly east of ATTO as well as the easterly valley of the Amazon River are frequently and convectively linked to the overpassing air masses and, thus, most relevant for the ATTO observations.
Map of the ATTO-relevant eastern Amazon Basin (ROI
This section provides a characterization of the BT footprint region from a
climatic and ecological perspective. Figure 9 shows maps of the mean
temperature and annual precipitation in the ROI
Map of the ATTO-relevant eastern Amazon Basin (ROI
Figure 10a presents a geographic biome classification according to Olson et al. (2001), with the following five biomes being included in the
ROI
Most of these ecoregions within the BT footprint belong to the biome
category of tropical and subtropical moist broadleaf forests (Fig. 10b). The
ecoregion with the highest overlap with the ATTO footprint is the
Uatumã-Trombetas moist forest. Moreover, regions with so-called
An additional GIS layer in Fig. 10a visualizes areas with the lowest
anthropogenic influence – so-called last-of-the-wild areas – as an
approximation of the biosphere in a pristine state (Sanderson et al.,
2002). Sanderson et al. (2002) used selected proxies for
contemporary human influence, such as population density, land
transformation, and infrastructure as a basis to define the last-of-the-wild
areas. Note here that the last-of-the-wild map provides only a very general
visualization of biosphere regions in untouched state for the following
reasons: (i) the map represents the status prior to the year 2000, which
means that since then the geographic extent of the last-of-the-wild regions
likely has shrunk substantially, (ii) within regions with human influence
the severity of human impact/pressure on the biosphere can be rather
variable, and (iii) the specification of regions with human influence does
clearly not account of the complexity of human activities and their
interaction with the biosphere. A more detailed discussion along these lines
can be found in Sanderson et al. (2002).
Map of the ATTO-relevant eastern Amazon Basin (ROI
Figure 11 shows the land cover classification according to the GlobCover
2009 data set within the ROI
Quantitative characterization of land cover types (see Fig. 11) in all 15 backward trajectory (BT) cluster footprints (see Fig. S11). The land cover contributions have been weighted with the relative BT density and, thus, represent an ATTO-relevant “land cover mix”. A comprehensive summary on the land cover mix within the ATTO BT footprint can be found in Table S1 in the Supplement.
Summary of GlobCover 2009 categories that account for 99.9 % of land cover variability within the weighted ATTO BT footprint with specification of relative contributions of individual categories. A comprehensive summary of the land cover mix for ATTO BT footprint is available in Table S1.
Beyond the qualitative analysis, we quantified the “land cover mix” within the BT cluster footprints (Fig. 12). For this analysis, the land cover analysis has been weighted by the air mass residence time in the clusters (see Sect. 2.7). Accordingly, regions within the footprint that are located close to the ATTO site were crossed more frequently by BTs and, thus, are weighted more strongly than regions in the periphery of the footprint. As a result, 11 GlobCover 2009 categories account for 99.9 % of the land cover variability within the BT footprint region as summarized in Table 3. The categories 40 (broadleaved evergreen or semi-deciduous forest) and 210 (water bodies, mostly part of the Atlantic Ocean) expectedly dominate the results (i.e., accounting for 87.3 %). Agricultural areas (i.e., categories 14, 20, and 30, accounting for 4.2 %), wetlands (i.e., 160 and 180, accounting for 4.7 %), and shrub- and grasslands (i.e., 110, 130, and 140, accounting for 3.4 %) represent minor fractions of the land cover mix.
For the footprints of the individual BT clusters, the land cover
categorization is summarized in Fig. 12. The following trends can be
observed: (i) agricultural lands contribute negligibly to the NE and ENE
clusters (sum of categories 14, 20, and 30:
Based on the land cover characterization, we further conducted an analysis
of the forest phenology within the ROI
Seasonal cycles in normalized difference vegetation index (NDVI)
for the ROI
The results in Fig. 13 show a pronounced NDVI seasonality for all relevant
land cover classes. Specifically, the seasonal cycles follow two contrasting
patterns in relation to rainfall and cloud fraction with the latter one
being an indirect proxy for incoming solar radiation (e.g., Hilker et al.,
2014). The NDVI results for the evergreen moist forest categories (i.e., 40
and 160) show a minimum around February and a maximum around July. Thus, the
seasonality is generally in phase with solar radiation and suggests
sunlight-enhanced growth upon decreasing cloud cover in the dry season, in
combination with a time lag of greening after the rain-rich wet season
(Hilker et al., 2014, and references therein). Presumably, the increasing
drought stress in the dry season is buffered by the comparatively
deep-rooted trees in the moist soils and, therefore, does not
(significantly) affect the NDVI (Nepstad et al., 2008). In contrast, land
cover categories representing agricultural lands (i.e., 14, 20, and 30),
shrub- and grasslands (i.e., 110 and 130), and deciduous forests
(i.e., 50) show a NDVI maximum in May (
Amazonian deforestation and further forest degradation activities (e.g., ecosystem fragmentation, fires, selective logging, illegal mining, overhunting) as a function of biophysical, climatic, socioeconomic, and cultural factors have been addressed by a large number of previous studies (e.g., Nepstad et al., 1999; Laurance et al., 2002; Asner et al., 2005; Malhi et al., 2008; Godar et al., 2012a; Cisneros et al., 2015). Here, we discuss the ATTO-relevant deforestation trends and drivers.
Map of the ATTO-relevant eastern Amazon Basin (ROI
Figure 14a provides an overview of the deforestation patterns within the
ROI For
this calculation, the following numbers have been used: Fearnside (2005)
reported a total deforested area of
The ESE clusters cover the areas with the strongest forest fragmentation and
perturbation within the ROI
The Amazon River and its shores are covered by the E clusters. Here, some deforestation hot spots (i.e., at the northern shore, about halfway between the ATTO region and the river delta) have emerged and could potentially develop into large-scale forest destruction in the future, depending on overall socioeconomic trends and conservation efforts (Fearnside, 2007). Accordingly, these clusters represent a semi-perturbed area of the basin and, thus, an “intermediate state in the Amazon's transition” (Davidson et al., 2012).
The ENE and NE clusters cover areas where deforestation activities are comparatively low (i.e., northern Pará, Amapá, and French Guiana). Therefore, the ENE and NE clusters still represent a mostly unperturbed state of the forest.
The so-called “fish bones” along the major highways represent a typical deforestation pattern, which consist of perpendicular smaller and mostly illegal access roads, penetrating (deeply) into the surrounding forest (Laurance et al., 2009). In Fig. 14, these patterns can be recognized, for example along the highways BR-163 (the so-called “soybean corridor” connecting the international port in Santarém with the soybean production in the southern states) and BR-230 (Trans-Amazonian highway) (Soares et al., 2004). The deforestation is typically associated with a strong fragmentation of the remaining forest areas. The fragmentation creates so-called edge effects, which perturb the humid, dark, and stable microclimate in the forest's canopy and understory with impacts on forest structure, tree mortality, and biodiversity (e.g., Wirth et al., 2007; Broadbent et al., 2008; Dohm et al., 2011; Laurance et al., 2011).
The main socioeconomic drivers of Amazonian deforestation can be grouped
into two categories. The first category comprises subsistence and family
agriculture, including some extractive activities, such as logging and
hunting. These individual actions typically occur on rather small scales,
but they are practiced by a rather large number of smallholders and
colonists, which ultimately sums up to a substantial level of deforestation
pressure (Godar et al., 2012a). The second category comprises deforestation
on larger scales, which has been mostly driven by international economic
interests, market demands, and government policies/subsidies (e.g., Soares
et al., 2014). It is conducted by a comparatively small number of
largeholders The term “largeholder” is frequently used as the
opposite of smallholder in the literature on deforestation (e.g., Godar et
al., 2012a, b; Pacheco, 2012). We have adopted this notation here.
Figure 14b and c zoom into two selected areas within the ROI
Figure 14c zooms into a region in central Pará along the Trans-Amazonian highway (BR-230), which has been a corridor of continuous and active deforestation during the last decades. This region has been selected for three reasons: (i) the characteristic fish bone deforestation patterns are particularly pronounced here, (ii) the region represents a deforestation hot spot that is comparatively close to ATTO (within 10 % contour of highest residence times, Fig. 7) and, thus, presumably has noticeable influence on the atmospheric observations, and (iii) its deforestation dynamics have been well documented previously by Godar et al. (2012a, b). Specifically, the municipalities of Medicilândia and Novo Brazil in Fig. 14c are characterized by contrasting socioeconomic deforestation trends: in Medicilândia, smallholder (family) agriculture has been predominant, which corresponds with a comparatively low degree of deforestation, rather short “fine bones”, small deforested patches that are rather close to the BR-230, as well as declining deforestation rates (Godar et al., 2012a). In contrast, in Novo Brazil, largeholder agriculture with extensive cattle ranching and industrialized soy farming has had a much more pronounced influence, which corresponds to a stronger degree of fragmentation, longer “fish bones”, many large rectangular patches (mostly for cattle ranching with several hundreds of hectares) that are located rather far from the BR-230, and increasing deforestation rates (Godar et al., 2012a; Cisneros et al., 2015).
Quantitative characterization of forest loss trends in cluster BT footprints for main wind directions NE, ENE, E, and ESE, based on Fig. 14. The annual forest loss has been calculated relative to the forest cover in the year 2000. The forest loss data in the cluster BT footprints have been weighted by the air mass residence time. The corresponding trends for all BT clusters are shown in Fig. S16.
Based on the annual forest loss data in Fig. 14, we conducted a quantitative
analysis of the deforestation levels and dynamics within the weighted
footprints of the individual BT clusters (Fig. 15 and Table S2). Figure 15
shows that the overall annual forest loss levels span from The
calculations are based on the annually deforested area of
This section presents the spatiotemporal patterns of fire occurrence within
the ROI
Map of the ATTO-relevant eastern Amazon Basin (ROI
Figure 16 shows a fire map within the ROI
In addition to the agriculture-related fire patterns, Fig. 16 further shows
substantial fire densities in the savanna ecoregions (i.e., high activity in
the Cerrado and modest activity in the Guianan savanna regions), which are
characterized by lower precipitation, a different vegetation type, and a
tendency to higher flammability. In contrast to the fires in moist forested
regions, which are the result of human-made ignition or the consequence of
human-made forest degradation and increased flammability, the low
precipitation levels in the savanna regions makes them more prone to the
occurrence of fires. Accordingly, the savanna vegetation is more adapted to
an (infrequent) occurrence of natural fires as well as more frequent use of
fire by indigenous people since pre-Columbian times (de Carvalho and Mustin,
2017). Parts of both ecoregions, the Cerrado and Guianan savannas, with
their characteristic fire regimes are ATTO-relevant as they are located
within the BT footprint. On a year-to-year basis, our analysis showed
comparable geospatial patterns in fire occurrence within the ROI
Relative fractions of ATTO-relevant fires (weighted with air mass residence times) within the different land cover categories, discriminated by major BT directions: NE, ENE, E, ESE, and SW. Fire analysis is based on the INPE database (see Sect. S1.1). Results shown here are averages of corresponding year-to-year data (2000–2016) as shown in Fig. S18. Fractions of all counted fires in this analysis are provided along with BT frequencies for comparison (see Table 2).
Beyond the fire map in Fig. 16, we conducted a quantitative classification of the detected fires by two metrics: the weighted BT cluster footprints and the land cover type in which they were detected. Figure 17 shows the resulting relative fractions of fires per land cover type and grouped into the four main BT directions, NE, ENE, E, and ESE. Note that this analysis has been weighted by BT residence time and, thus, provides an estimate for the ATTO-relevant “mix of fires” (e.g., forest fires vs. savanna fires vs. agricultural management fires). This is relevant for the ATTO observations since the fuel types (e.g., forest vs. agricultural waste) and corresponding combustion modes (i.e., flaming vs. smoldering) typically emit gases and aerosol particles of different composition and properties (e.g., Andreae and Merlet, 2001; Janhäll et al., 2010). For example, breakout understory fires (escaping from ignited land-clearing fires) are typically associated with less efficient smoldering combustion, whereas pasture burning and high-intensity deforestation fires (after clear-cutting and drying of the vegetation) tend to be predominantly flaming combustion (e.g., Tang and Arellano, 2017). Figure 17 shows different fire mixes for the NE, ENE, E, and ESE clusters. The dominant contribution in all cases are fires in rain forest areas (land cover category 40), which account for 60 %–63 % in the NE, ENE, and E directions and for 54 % in the ESE direction. Fires in shrub and grassland categories (i.e., 110, 130, 140) account for a comparable fraction (10 %–16 %) for all directions. Fires associated with agricultural categories (i.e., 14, 20, 30) show a pronounced gradient from NE (11 %) over ENE (13 %) and E (17 %) towards ESE (28 %). This is consistent with the properties of absorbing aerosols measured at ATTO (e.g., the relative fractions of black vs. brown carbon), which have been related to the air mass origin by means of BT directions (Saturno et al., 2018b). Finally, we analyzed the seasonal cycle in fire occurrence for the entire BT footprint, resolved by land cover categories, and found a rather uniform seasonality with its minimum around April and its maximum around September (see Fig. S20). These trends, particularly the onset of the biomass burning season, are consistent with the observed atmospheric composition and variability at ATTO (e.g., Andreae et al., 2015; Pöhlker et al., 2016; Saturno et al., 2018b).
Biomass burning represents the predominant forest perturbation and source of atmospheric pollution in the Amazon Basin. However, several further categories of human-made activities and infrastructure also impact the biosphere–atmosphere exchange to a significant extent. Accordingly, this section addresses the following infrastructure classes: (i) population density and urban centers with their related emissions, (ii) thermoelectric power plants as major fossil fuel burning sources, (iii) major dams and reservoirs with their significant environmental impacts, (iv) major highways as drivers for forest fragmentation and degradation, and (v) (cargo) ship traffic as a further pollution source.
Map of the ATTO-relevant eastern Amazon Basin (ROI
The population density map in Fig. 18 shows that most settlements and cities
are located along the Atlantic coast in the southeast of the ROI
Dams and reservoirs worldwide have a substantial impact on rivers and their
ecology for a variety of reasons (Lehner et al., 2011). An increase in
methane (also carbon dioxide) emissions from reservoirs due to anoxic
microbial decomposition of flooded biomass has been one particular impact on
the biosphere–atmosphere exchange – presumably also with direct relevance
for the ATTO site observations (e.g., Abril et al., 2005; Kemenes et al.,
2007; Fearnside and Pueyo, 2012). Figure 18 shows that several major dams
are located within the ROI
Map of eastern Amazon Basin, including ROI
In the previous sections, we have noted that highways in the Amazon Basin
“have a keystone role in deforestation” and “stimulate the influx of
population and investment” (Fearnside and Graca, 2006). However, it is
still being debated to what extent roads have acted as main deforestation
controls. Figure 18 displays the currently existing road network in the
ROI See, for example,
Ministros assinam contrato que estuda ligação de RR ao Pará pela
BR-210.
Map of the ATTO-relevant eastern Amazon Basin (ROI
Figure 19 displays a map, which zooms out from ROI
Figure 20 presents a map highlighting the ATTO BT footprint in combination
with a GIS layer on mining activities in the basin. Generally, mining –
particularly large-scale pit mines – has caused strong perturbations of the
forest ecosystem in affected areas (e.g., Potapov et al., 2017). Moreover,
large mines are potential sources of industrial air pollution as well as
soil dust suspension (i.e., coarse-mode aerosol particles) (e.g., Huertas et
al., 2012). Both of these factors make mining relevant for the ATTO
research. Figure 20 emphasizes that comparatively large bauxite mines
already exist within the ATTO BT footprint. The largest of those mines,
which is shown in Fig. 14b, is located within the 1 % contour of the
highest air mass residence times according to Fig. 7 and is, thus, a
potentially relevant source for the ATTO observations. It further caused
substantial forest loss rates over the last years ( Annual forest loss of Some information
on the initiative can be found here:
Map of the ATTO-relevant eastern Amazon Basin (ROI
The conservation efforts to protect the Amazonian forests are manifold.
Often, protection is initiated as a response to deforestation frontier
expansion (Nepstad et al., 2006). Worldwide, a large variety of different
types of conserved areas (e.g., in terms of their legal, control, and
habitation status) exists. According to the world database on protected
areas (WDPA) classification, about 14 types are relevant in the ROI
Map of the ATTO-relevant eastern Amazon Basin (ROI
The ATTO site itself is located in a sustainable development reserve (i.e.,
Reserva de Desenvolvimento Sustentável do Uatumã). Such sustainable
development reserves allow a certain level of resource use and extraction,
in contrast to strictly protected areas. Figure 21a further shows that the
ATTO site footprint is to various degrees covered by protected areas.
Specifically, almost the entire northern half of the BT footprint with the
NE and ENE clusters (i.e., in northern Pará and Amapá as well as in
French Guiana) consists of protected lands. In contrast, the southern half,
which overlaps with the arc of deforestation and the ESE clusters (i.e.,
ESE2 and ESE3), comprises only few and rather small conservation areas. The
part of the Amazon River valley that is covered by the E clusters contains
few protected areas, and, therefore, plays an intermediate role in this
overall conservation picture. In Fig. 21a, a comparison of the protected
areas and the patterns in fire occurrence highlights examples for both
successful and less successful conservation efforts. For example, most of
the indigenous areas have been very effective in preventing deforestation
and, thus, do not overlap with the major fire hot spots (see also Ricketts et
al., 2010). Note also the sharp edges between fire hot spots and indigenous
areas in certain cases. In contrast, the formally protected areas along the
highway BR-163 have not prevented the continuous forest fragmentation in
this area. The future effectiveness of the protected areas in the
ROI
In addition to the institutionalized network of protected areas, the
conservation of the Amazon forest can be discussed as a passive de facto
protection due to the remoteness of yet unperturbed regions (Soares-Filho et
al., 2006). Figure 21b visualizes a “remoteness map” of the ROI
The construction of new roads – in particular paved all-weather highways –
has a keystone role in future deforestation as it opens large areas for
colonization and resource exploitation (e.g., Fearnside and Graca, 2006; de
Carvalho and Mustin, 2017). For the ATTO observations, the currently
discussed construction of the Arco Norte with the extensions of the BR-210
and BR-163 would be by all means the most severe impact with a profound
perturbation of the currently still untouched NE segment of the ATTO
footprint. Soares-Filho et al. (2006) conducted a policy-sensitive
simulation of future deforestation scenarios. Within the ROI
In addition to the infrastructural perturbations, climate change tends to further increase the pressure on the Amazon ecosystem. In extreme scenarios, a large-scale rain forest dieback – i.e., a climate-driven substitution of moist forests by semiarid and/or savanna vegetation – due to changing hydrological and seasonal regimes has been predicted, although these predictions still comprise large uncertainties (e.g., Cochrane and Laurance, 2008; Nepstad et al., 2008; Cochrane and Barber, 2009). Furthermore, it has been reported that these effects will likely be most severe in the eastern basin as large parts of its forest are already close to the lower rainfall limit that sustains moist tropical vegetation (Zelazowski et al., 2011). Even minor changes in precipitation patterns could exceed thresholds that irreversibly push the system beyond a tipping point towards seasonal and savanna forests with strong fire feedback cycles (Malhi et al., 2009; Alencar et al., 2015). Accordingly, any changes in dry-season water supply (i.e., precipitation or stored soil moisture) are of critical importance for the rain forest ecosystem (Boisier et al., 2015). Since climate models are known to differ substantially with respect to regional rainfall patterns in Amazonia, future spatially resolved projections are highly uncertain (Cox et al., 2008; Xie et al., 2015). However, a general drying trend towards more seasonal bioclimatic conditions as well as an increase in frequency and severity of droughts is being observed already and will likely intensify, particularly in the eastern basin (Good et al., 2008; Lewis et al., 2011; Silva et al., 2013; Hilker et al., 2014). Accordingly, it appears to be a likely scenario that the ongoing ATTO observations will witness further substantial, maybe disruptive transformations within the eastern basin through the synergistic effects of human-made perturbations and climate change.
This study presents a backward trajectory (BT) and geographic data analysis for the ATTO site in the central Amazon Basin. It extracts spatiotemporal BT patterns in relation to the current state and anticipated future change of the land cover within the ATTO footprint. The overall aim of this work is to provide a robust characterization of the effectively probed land cover mosaic in the footprint region to embed atmospheric observations at ATTO into a broader Amazonian context. Since the footprint regions of the ATTO site and the nearby ZF2 site are similar, the present work may further act as a resource for the interpretation of ZF2 results. We envision that this work may serve as a resource and look-up reference for the interpretation of current and future observations in the ATTO-relevant and vulnerable eastern basin, in which the progressing climate and land use changes are influences of crucial importance.
The multi-year BT analysis (2008–2016) shows the characteristic central
Amazonian air mass advection as the confluence of northeasterly and
southeasterly trade winds, feeding boundary layer air into the ITCZ. In
response to the annual north–south ITCZ migration, the ATTO-relevant BTs
follow a seasonal swing between a NE path during the wet season and a SE
path during the dry season. The northernmost BT advection occurs around
February and the southernmost around July with a north-to-south swing
lasting
The BT-derived ATTO footprint defined here includes a continental area of
In terms of deforestation and agricultural expansion, the ATTO site covers
the full range of deforestation dynamics spanning from major forest
fragmentation and clearing hot spots in the SE (deforestation rates
The following data from this study have been deposited under
The supplement related to this article is available online at:
CP and DW conducted the HYSPLIT backward trajectory analyses. CP, HP, TK, and ERC conducted the GIS analyses. CD conducted the FLEXPART analysis. CP wrote the paper. DMZ, JWK, GL, DP, MLP, NL, JS, and SW supported the backward trajectory and GIS analyses. JB, SC, VRD, BAH, JVL, JM, MP, MS, QW, BW, PA, UP, and MOA contributed to the data analysis and paper writing through fruitful discussions and by providing valuable comments and ideas. All authors contributed to data discussions and paper finalization.
The authors declare that they have no conflict of interest.
This article is part of the special issue “Amazon Tall Tower Observatory (ATTO) Special Issue”. It is not associated with a conference.
This work has been supported by the Max Planck Society (MPG). For the
operation of the ATTO site, we acknowledge the support by the German Federal
Ministry of Education and Research (BMBF contract nos. 01LB1001A and 01LK1602B) and the
Brazilian Ministério da Ciência, Tecnologia e Inovação
(MCTI/FINEP contract 01.11.01248.00) as well as the Amazon State University
(UEA), FAPEAM, LBA/INPA, and SDS/CEUC/RDS-Uatumã. This paper contains
results of research conducted under the Technical/Scientific Cooperation
Agreement between the National Institute for Amazonian Research, the State
University of Amazonas, and the Max-Planck-Gesellschaft e.V.; the opinions
expressed are the entire responsibility of the authors and not of the
participating institutions. Céline Degrendele was funded by the core
facilities of the RECETOX Research Infrastructure, project LM2015051, and by
Actris-CZ RI project CZ.02.1.01/0.0/0.0/16_013/0001315,
funded by the Ministry of Education, Youth and Sports of the Czech Republic
under the activity “Projects of major infrastructures for research,
development and innovations”. We acknowledge the support by the
Instituto Nacional de Pesquisas da Amazônia (INPA). We would like to
thank Reiner Ditz, Jürgen Kesselmeier, Susan Trumbore, Alberto Quesada,
Thomas Disper, Thomas Klimach, Andrew Crozier, Björn Nillius, Uwe
Schulz, Steffen Schmidt, Niro Higuchi, Antonio Ocimar Manzi, Alcides Camargo
Ribeiro, Hermes Braga Xavier, Elton Mendes da Silva, Nagib Alberto de Castro
Souza, Adir Vasconcelos Brandão, Amauri Rodriguês Perreira, Antonio
Huxley Melo Nascimento, Thiago de Lima Xavier, Josué Ferreira de Souza,
Roberta Pereira de Souza, Bruno Takeshi, and Wallace Rabelo Costa for
technical, logistical, and scientific support within the ATTO project.
Moreover, we thank Britaldo Soares-Filho, Florian Wittmann, Henrique
Barbosa, Scot T. Martin, Xuguang Chi, Hang Su, Isabella Hrabě de Angelis,
and Ovid Krüger for scientific support and stimulating discussions. The
authors gratefully acknowledge the NOAA Air Resources Laboratory (ARL) for
the provision of the HYSPLIT transport and dispersion model and/or READY
website (
The article processing charges for this open-access publication were covered by the Max Planck Society.
This paper was edited by Laurens Ganzeveld and reviewed by David Fitzjarrald and Bart Kruijt.