Population growth and deforestation in Amazonas, Brazil, from 1985 to 2020

Amazonas state represents 37% of the Amazônia biome in Brazil. Although Amazonas remains 98% forested, its contribution to annual biome deforestation increased substantially in the past ten years. Herein, the connections between population and deforestation in Amazonas are investigated from 1985 through 2020. Anthropogenic landcover fraction and population density varied spatially and temporally across the 62 municipalities of the state. The temporal variability had specific geographic patterns, and three microregions were identified. Economic development along the southern border, arising from agricultural activities in the pattern of classical deforestation in Amazônia, was characterized by large increases in anthropogenic landcover but only small changes in population. Economic development along the Amazon River, characterized by large increases in population and anthropogenic landcover, represented urbanization and the growth of industry and agriculture. Economic development along the western border, based on trade and commerce with Peru and Colombia, corresponded to increases in population without large increases in anthropogenic landcover. The three microregions were quantitatively characterized by different slopes between anthropogenic landcover fraction and population density. The connections between deforestation and population varied by a factor of 50 × among the different microregions, suggesting important considerations for the future forest preservation in Amazonas. That time is now given the increasing importance of this region, which twice approached 20% of the total annual deforestation in Amazônia over the past decade.


Introduction
The Amazônia biome consists mainly of tropical forests throughout the drainage basin of the Amazon River (Pires & Prance, 1985;Eva et al., 2005).The biome also incorporates a diverse array of smaller ecosystems of deciduous forests, flooded forests, savannas, and more.Humans have been actively transforming Amazônia long into the past on centennial and millennial scales (Clement et al., 2015).Beginning in the 1950s, the process greatly accelerated (Rodrigues et al., 2009;Garrett et al., 2021).It was driven by business opportunities afforded by mechanized technology and global markets (Hecht & Rajão, 2020).The overall extent of biome deforestation, mainly for agricultural activities, now exceeds 15% (Albert et al., 2023).Implications for the carbon cycle and other biogeochemical cycles relevant to climate forcing are significant yet complex (Gatti et al., 2014;Aragão et al., 2018;Covey et al., 2021).The landcover changes have caused biodiversity loss (Bustamante et al., 2016) and degraded the lands of indigenous peoples (Barretto Filho et al., 2021).Through effects on evapotranspiration and land albedo, the changes have also altered regional temperature and rainfall patterns (Marengo et al., 2018).These changes in recent decades beg the question of possible future trends for Amazônia (Lovejoy & Nobre, 2018;Albert et al., 2023).
In this context, several overlapping theoretical approaches have achieved varying levels of insight with respect to understanding the coupling between landcover changes and demography dynamics (Bremner et al., 2010;Caviglia-Harris et al., 2012).These approaches include consideration of the household life cycle (Walker & Homma, 1996), the livelihoods and capabilities of the family unit across decadal timescales (Serageldin & Steer, 1994), and multiphasic responses to household evolution (Bilsborrow, 1987;Davis & Lopez-Carr, 2014).Given the multiple overlapping dynamics of these approaches, a demo-livelihoods theoretical framework combined these separate frameworks into an overarching form (Barbieri, 2023).The interacting factors are complex.For example, the relation between cause and effect of migration and land use can evolve across phases, such as migration as a cause of deforestation in pioneer stages (i.e., in-migration and population size as predicting factors) and loss of agricultural productivity as a cause of migration in later stages (i.e., out-migration and population size as lagging factors) (Gori Maia & Schons, 2020).Partially for these reasons, there is a debate about the balance between deforestation as endogenous, at least in part, with population, rather than fully dependent on it (Sydenstricker-Neto, 2012).A simple linear relationship between these poles of possibility is not accurate, and a multidimensional approach provides more insights.There are bidirectional relationships and feedbacks between the environment and in/out-migration (VanWey et al., 2011).A case study of one microregion of municipalities in southwestern Amazônia highlights the involved complexities (Caviglia-Harris et al., 2012).Thus, scaling up from microregions to states within Amazônia and further to Amazônia as a whole is challenging.
In this regard, the history of change in Amazônia is geographically heterogeneous (Garrett et al., 2021).The government of Brazil under President Kubitschek, elected in 1956, initiated a large-scale, decades-long program to occupy Amazônia (Homma, 2013).Along the southern portion of Amazônia (e.g., the states of Rondônia and Mato Grosso), there was greater infrastructure, and the land topography was more amenable to agribusiness.Moreover, the research activities of the Brazilian government focused on soil improvements (e.g., mass liming) and agriculture (e.g., new crop varieties) in this region.In the northern regions (e.g., Amazonas and Roraima), there were also attempts at resettlement and agriculture in the 1960s and 1970s (Teixeira et al., 1985).At the same time, there was a parallel attempt at an industrial zone in the northern regions, specifically Manaus, Amazonas (Brito & Maciel, 2019).The urban economic activities there proved attractive to settlers, who found farming non-productive in this region, and in large part they abandoned the rural lands.For these reasons, largescale agricultural activities did not develop in this region.Likewise, Roraima had other economic opportunities than farming, especially related to mining (Diniz & Gonçalves Lacerda, 2014).
In consequence to these and other factors, Amazonas stands out as undergoing much less deforestation than its neighboring Brazilian states (Table 1).Figure 1 shows the landcover for the state of Amazonas, Brazil, and its environs in 1985 compared to 2020.Amazonas has a land area of 1.56 million km 2 , which is similar in size to the country of Mongolia, and it constitutes 37% of the total area of Amazônia.Amazonas thus represents a vast forest reserve for preservation or loss in the coming decades.Therefore, both because of its large relative area and its current low relative extent of deforestation, Amazonas represents a key aspect of possible future change for Amazônia.
Table 1 Comparison of the anthropogenic landcover of Amazonas, Brazil, to its five neighboring states for the Amazônia biome Anthropogenic landcover is listed for 1985 and 2020 for each state, including the absolute change during this period.The relative share of land area with respect to the contribution to the total Amazônia biome lying within the six states is listed.The relative share of anthropogenic landcover associated with each state in 2020 for this group of six states is also listed.The study herein analyzes the links between historical population growth and anthropogenic landcover in Amazonas, including the recent acceleration of the latter, to provide context to how policy and governance choices related to each might influence deforestation in the coming decades (Geist & Lambin, 2002).The growth of anthropogenic landcover is synonymous with deforestation.Given that anthropogenic landcover change is by its nature tied to human activities (Davidson et al., 2012), the study approach focuses on population size (Bilsborrow et al., 2004;Carr et al., 2005;Bremner et al., 2010;Reis & Guzmán, 2015;de Toledo et al., 2017).Population growth can be a precursor to anthropogenic landcover change, thus serving as a leading indicator, or can accompany anthropogenic landcover change, thus serving as a lagging indicator (Gori Maia & Schons, 2020;Cowan et al., 2021).
The population of Amazonas increased from 1.1 million in 1975, to 2.2 million in 1991, to 4.3 million in 2021, thus doubling every two decades (IBGE, 2023a).Of this state population, the main municipality of Manaus had a population of 0.39, 1.0, and 2.3 million during these same years, corresponding to 35%, 45%, and 53% of the state population across this period of 46 years and indicating a trend in urbanization.The national government has afforded Manaus preferential tax treatment as a development strategy for Amazônia since the 1960s, transforming Manaus into a national hub for manufacturing (Brito & Maciel, 2019).Waves of Japanese, Korean, and, most recently, Chinese companies have established operations.Deforestation in Amazonas has occurred dominantly outside of Manaus, and changes in the state population ex-Manaus should thus likewise be considered.For Manaus excluded, the population of Amazonas increased from 0.7 million in 1975 to 1.2 million in 1991 to 2.0 million in 2021.This study explores how population and landcover changes are coupled among the municipalities of Amazonas.The results are considered in the context of possible scenarios for ongoing changes in anthropogenic landcover in Amazonas over the coming decades.

Landcover
The state of Amazonas consists of 62 contiguous municipalities (IBGE, 2023b).The shape files for these 62 polygons were obtained from the Brazilian Institute of Geography and Statistics (IBGE) (Section S1).Landcover data were obtained from the MapBiomas project for 36 years from 1985 through 2020 (Collection 6; Fig. 1 Landcover for Amazonas state, Brazil, and its environs in 1985 and 2020.The orange polygon outlines the borders of Amazonas.The neighboring regions include the Brazilian states of Acre, Rondônia, Mato Grosso, Pará, and Roraima (orange line) and the countries of Venezuela (red), Colombia (purple), and Peru (magenta).The images were prepared from the MapBiomas dataset for Brazil (Sect."Landcover").The green regions represent forest cover, the blue regions show water bodies, the yellow regions indicate farming and pasture regions primarily tied to deforestation, and the red regions are urban areas.The rectangle of the image represents {+ 3.0°, − 74.7°} for latitude-longitude of the upper-left corner and {-11.0°,− 54.5°} for the lower-right corner ▸ 27 Page 6 of 18 Section S1) (Souza et al., 2020).The MapBiomas project analyzed annual imagery of the Landsat satellite at a spatial resolution of 30 m × 30 m across Amazônia from 1985 to 2020.Based on the analysis, the MapBiomas project assigned a landcover type for each pixel.For Amazonas, the pixel count was 1.75 × 10 9 .Natural landcover types included forest, savanna, mangrove, wooded restinga, wetland, grassland, salt flat, rocky outcrop, and sand.Anthropogenic landcover types designated pasture, agriculture, forest plantation, urban, and mining.Other landcover types indicated surface waters and aquaculture.The methodology, including an accuracy analysis, is available (MapBiomas, 2022).The IBGE geometry files were rasterized on the MapBiomas grid to develop mask files for the 62 municipalities of Amazonas.

Population
Population data were obtained from the IBGE database (IBGE, 2023a).Census data were typically collected every ten years in Brazil, although there were exceptions, such as recent delays associated with the COVID pandemic.The database consisted of a mix of actual census data for the years of decadal collection and data modeled by the IBGE for the in-between years.For the state of Amazonas, the accuracy of the annual population data had significant uncertainty.For instance, for some municipalities the reported population increased during an initial period but then decreased in a subsequent period, and these changes were unrealistically coincident exactly with the census years (Section S2).Because of the suspect population count, in some cases the population was determined by a judicial order based on claims brought by interested parties.

Data analysis
As stated, the IBGE population dataset had possible year-to-year inaccuracies, especially for Amazonas, during the study period.Even so, the dataset still maintained accurate secular trends given the decadal grounding in data collection.For this reason, the population data were fit herein by linear functions for each municipality (Table S1).This approach captured the secular population trends of the municipalities across the study period.The population difference between the start and the end of the study period was computed for each municipality, and these difference values were employed throughout this study (vide infra).Likewise, differences in anthropogenic landcover fraction across the study period were calculated for each municipality and used in the analysis.Anthropogenic landcover fraction was calculated as the total area of the pixels identified as anthropogenic types in the MapBiomas dataset divided by the area of the municipality.For Amazonas, the dominant anthropogenic type was pasture (97%), which was followed by urban (2%) and other types (e.g., mining).Geographic patterns within the individual datasets of population difference and anthropogenic landcover difference, as well as cross trends between them, were established by plotting the values in false color within the framework of the municipality map of Amazonas state, thereby identifying microregions of different trends.

Variability and patterns in changes in landcover and population
Among the 62 municipalities of Amazonas, there was significant variability in the change of anthropogenic landcover fraction across the 36-year study period.At one extreme (top line, Fig. 2A), the anthropogenic fraction changed from 0.02 to 0.17 for Careiro da Várzea, representing an increase of 0.15 across the study period (Figure S2).The increase was also greater than 0.10 for Urucurituba (second line).Overall, nine municipalities had increases of 0.05 or more.By comparison, at the other extreme, the increases were less than 0.02 for 42 municipalities and less than 0.05 for another 11.For Manaus, which is the dominant population center of the state, the initial anthropogenic fraction was 0.04, and the increase was 0.01.There was no net reforestation for any municipality across the study period.
The change in population density also had significant variability across the 62 municipalities (Fig. 2B).For Manaus, constituting more than half of the total state population, the population density changed from 89 to 196 persons km −2 , representing a net increase of 107 km −2 (Figure S3).The next largest increase was 12 km −2 for Iranduba, for which the population density changed from 11 to 23 km −2 .Overall, ten municipalities had increases of 5 km −2 or more, thirteen of 1 to 5 km −2 , and thirty-nine of 1 km −2 or less.
The variability in landcover and population had specific geographic patterns to it.Figure 3A shows a map of the change in anthropogenic landcover between 1985 and 2020.Each polygon represents one of the 62 municipalities constituting Amazonas state.The coloring corresponds to an increase in anthropogenic landcover fraction of 0.02 or less as white (42), between 0.02 and 0.05 as orange (11), and 0.05 or more as red (9).The numbers in parentheses indicate how many municipalities out of 62 were included in a grouping.Figure 3B likewise shows the change in population density between 1992 and 2021.The coloring corresponds to an increase in population density of 1 km −2 or less as white (39), between 1 and 5 km −2 as orange (13), and 5 km −2 or more as red (10).
The color clustering in Fig. 3A shows that there were two geographic groupings for changes in anthropogenic landcover fraction (Table S2).In the first cluster in the northeastern portion of the state, anthropogenic landcover increased in the Manaus microregion along the Amazon River.The twelve municipalities of this microregion had an area-weighted increase in anthropogenic fraction of 0.05.This microregion represented an increase of 5100 km 2 in anthropogenic landcover during the study period.In the second cluster, the geographic microregion occurred along the southern border.These municipalities are adjacent to the states of Acre, Rondônia, and Mato Grosso, as well as the southern portion of Pará.These eight municipalities along the southern border had a fractional increase in anthropogenic landcover of 0.04, representing an additional 11,000 km 2 of anthropogenic landcover.For comparison, the 42 municipalities outside the two microregions had a change of 4100 km 2 , corresponding to a fractional increase of 0.003.This increase was an order of magnitude less than that of the two identified microregions.
For changes in population density (Fig. 3B), two geographic groups were again apparent in the color clustering.As a reference point, outside the microregions the population density increased by 0.3 km −2 during the study period.The 17 municipalities constituting the Manaus microregion in the northeast had an area-weighted increase in population density of 14.8 km −2 .The total increase in population was 1.68 million, representing 74% of the total increase in state population (+ 2.27 million).The growth was dominated by the single municipality of Manaus (+ 1.28 million).The analysis for the 16 municipalities excluding Manaus corresponded to increases of 3.9 km −2 , 0.39 million, and 40% of the residual change in the state population (i.e., total change less the change in Manaus).The second microregion, centered on the municipality of Tabatinga, was along the western border adjacent to Colombia and Peru.For the six Brazilian municipalities in this microregion, the population density increased by 2.4 km −2 , corresponding to 0.13 million people and 13% of the residual change in the state population.For comparison to Fig. 3A, the southern municipalities identified therein as centers of deforestation had corresponding changes in population density of + 0.4 km −2 , representing a population increase of 0.14 million and 11% of the residual change in the state.

Correlations between changes in landcover and population
The relationship between anthropogenic landcover fraction and population density is explored in Fig. 4 for the municipalities constituting the microregions of Fig. 3.The municipalities of the Manaus microregion are represented by red lines, those of the southern border microregion by green lines, and those of the western Tabatinga border microregion by blue lines.The lines represent data sets for the period of 1992 through 2020.For all municipalities in these microregions, the population density increased each year from 1992 through 2020 (Table S1).The anthropogenic landcover fraction ranged from 0.0 to 0.17, and the population density varied from < 0.1 up to 23 km −2 .
Several patterns and differences among municipalities are apparent in the data set in Fig. 4.This variety is well organized geographically by the microregions.For the Manaus microregion (red), both the change in population density and the change in anthropogenic landcover were large (Castilhos, 2016).The metropolis of Manaus was itself not in this microregion because the change in anthropogenic landcover fraction was low.Economic development in this region represented continued regional urbanization and growth of agriculture along the trade routes of the Amazon River as well as localization within the sphere of influence of the Manaus industrial center and free trade zone.By comparison, for the southern border microregion (green), there was a large increase in anthropogenic landcover but only small changes in population density.This pattern corresponded to economic development through frontier settlement, deforestation, and agriculture (Browder et al., 2008;Andersen et al., 2009).This pattern was a continuation from the nearby Fig. 4 Relationship between anthropogenic landcover fraction and population density for the municipalities constituting the Manaus microregion (red), the southern border microregion (green), and the western border microregion (blue) of Amazonas, Brazil, from 1992 through 2020.Each black dot represents a data pair of fraction and density for a single year of a single municipality.Each line represents a fit through the data points from 1992 through 2020 for a single municipality.The set of data points associated with an individual line is visually resolved in Figure S4 states of Acre, Rondônia, and Mato Grosso, all well-recognized centers of large agribusiness and deforestation for over half a century (Davis et al., 2014;Mueller & Mueller, 2016).For the western border microregion (blue), the population density increased, and deforestation was relatively less important.Tabatinga as the locus of this microregion grew across the study period as a center of international trade and commerce.It is the primary link among Colombia, Peru, and Brazil, and furthermore it has preferential tax treatment (Euzébio, 2013).For these different models of economic development among the three microregions, the median slopes were as follows: 0.002 in units of change in the fraction of anthropogenic landcover change per change in population density (km −2 ) for the set of municipalities in the western border microregion, 0.006 km −2 in the Manaus microregion, and 0.10 km −2 for the southern border microregion.The sensitivity thus varied by a factor of 50 × among the microregions due to their differing pathways of economic development.
Of the 62 municipalities, 31 are represented by the three microregions in the lines of Fig. 4. For the other 31 municipalities outside of the microregions, population change and deforestation were both small, meaning below the threshold values characteristic of the three microregions.Across the study time period, the economic activity in these other municipalities was mostly traditional, including small-scale fishing and agriculture.Some extractive development also occurred, such as the Urucu-Coari pipeline for natural gas and regional, largely illegal, artisanal gold mining (Barbosa et al., 2003;Isper Jr and Picanço, 2020).

Implications
The importance of Amazonas to the future of deforestation within Amazônia is captured in Fig. 5.For Amazônia (blue), the anthropogenic landcover fraction increased from 0.05 to 0.15 from 1985 through 2020 (Fig. 5A).The fraction in Amazonas, however, increased by much less, specifically from 0.005 to 0.02 (gray, inset).Correspondingly, the anthropogenic fraction of Amazônia outside of Amazonas increased from 0.07 to 0.23 (purple).The green and red points represent the separate contributions by Amazonas and the remainder of Amazônia to the total anthropogenic fraction of Amazônia.One implication apparent in Fig. 5A is that Amazonas contributed only a small portion of the overall change in anthropogenic landcover fraction.As such, the large land area of Amazonas effectively diluted the overall reported deforestation from a higher fraction of 0.23 in regions directly deforested to 0.15 for the biome as a whole.Increasing frequency of deforestation in Amazonas in recent and future years would remove this dilution factor.
The historically small contribution by Amazonas and the recent change in that status are investigated more closely in Fig. 5B.The green stem heights represent the annual contribution by Amazonas to the total increase in anthropogenic landcover of Amazônia (blue stem heights).In the early years, deforestation in Amazonas made a minor contribution (i.e., an annual median of 2.5%) to the total deforestation of Amazônia.However, even as the growth of the anthropogenic landcover slowed after 2005 for the biome as a whole, deforestation in Amazonas continued at its typical 36-year rate.Moreover, it increased in some years, such as in 2015 and 2019.As 1 3 27 Page 12 of 18 Fig. 5 (A) Anthropogenic landcover fraction and (B) change in anthropogenic landcover fraction from 1985 through 2020 for Amazônia in its entirety (blue) and when excluding Amazonas (purple).The separate contributions by Amazonas (green) and the remainder of Amazônia (red) to the total anthropogenic landcover fraction of Amazônia are also shown.The inset shows the plot for the anthropogenic landcover fraction of Amazonas.The sequence of lighter and dark coloring on the data points indicates election cycles in Brazil.Labeling shows when different presidents assumed office in Brazil.The uncertainty on the ordinate of panel B is ± 0.0003 based upon propagation of errors for the landcover assignments in the MapBiomas analysis.Data source: MapBiomas, Collection 6 (Souza et al., 2020) a result, the relative contribution of deforestation in Amazonas to the total anthropogenic change in Amazônia changed from a minor component before 2005 to an important component in the past ten years, reaching 20% of the total deforestation in two of those years (2015 and 2019).
This transition in the relative importance of Amazonas in story of deforestation in Amazônia can be considered in the context of governance and leadership at the national level (Hochstetler, 2017;Viola & Gonçalves, 2019;Garrett et al., 2021).The sequence of lighter and dark coloring on the data points of Fig. 5 indicates election cycles in Brazil.Labeling shows when different presidents assumed office.The period of Lula's presidency, which began in 2003, correlates with a step change downward in the rate of deforestation in Amazônia (Fig. 5) (Hecht, 2011;Boucher et al., 2013).From 1986 through 2003, the median increase in anthropogenic landcover fraction was 0.004 y −1 (Fig. 5B).From 2003 to 2019, the median increase was 0.002 y −1 .For Amazonas, however, the statistics and correlation with national governance appear different.The median contribution by Amazonas to the total increase in anthropogenic landcover for Amazônia during the period before Lula's presidency was 9 × 10 -5 y −1 , which can be compared to 2 × 10 -4 y −1 after Lula's incumbency, representing close to a doubling in the more recent period.This result again indicates the growing importance of Amazonas in the deforestation equation in Amazônia.It also highlights that there are differences between Amazonas and Amazônia in the context of national leadership and governance as well as natural capital.
Amazonas is located on the western side of Amazônia and thus receives greater rainfall and has less of a contrast between wet and dry seasons than the eastern and southern regions of Amazônia (Espinoza Villar et al., 2009).As a result, natural fires are absent under most circumstances.Human-initiated fires, however, are common and result in increased anthropogenic landcover (Aragão & Shimabukuro, 2010).More specifically, the anthropogenic landcover of Amazonas increased from 1 to 2% of the state during the study period (Souza et al., 2020;MapBiomas, 2022).By comparison, the anthropogenic landcover within the Amazônia biome of the neighboring states of Acre, Mato Grosso, Pará, Rondônia, and Roraima increased by 4% to 29% (Table 1).Amazonas constitutes 40% of the total area of the Amazônia biome lying within this group of states.
Most studies on deforestation in Amazônia have focused on the geographic region along the southern and eastern edges of Amazônia, including the population dynamics in these regions.Objectively, Amazonas in central Amazônia made minor contributions to deforestation during those past time periods, justifying the literature focus on the edge regions of Amazônia.Nevertheless, in the past ten years, Amazonas has twice been a major contributor to total deforestation (20%).By comparison, relative changes in population size were similar during the past period of low deforestation in Amazonas and the recent period of accelerating deforestation.Specifically, during the past ten years, the population in the northern states of Brazil grew by 17% to 18.9 million, and 26% of this growth occurred in Amazonas (+ 732,000 persons) (IBGE, 2023a).The northern states in the IBGE classification include Amapá, Acre, Amazonas, Rondônia, Roraima, Pará, and Tocantins.In the ten-year period from 1989 to 1999, the population in the northern states grew by 22% to 12.1 million, and 20% of this growth occurred in Amazonas (+ 439,000 persons).Given 27 Page 14 of 18 the non-differentiated growth rate in population across these periods yet the stark difference in deforestation rates, an inference could be that population size does not appear tightly related to deforestation.During both periods, however, most of the population growth was absorbed into the urban microregion of Manaus.For an analysis based on microregions, however, the population is analyzed geographically, and the southern microregion shows strong coupling between population growth and deforestation.This coupling is understood within the demo-livelihoods theoretical framework of phases (Barbieri, 2023), meaning that to this day there is an active and ongoing phase of initial colonization in the southern microregion.By comparison, the Manaus and Tabatinga microregions in central and western Amazonas have large amounts of capital and other economic activities (e.g., industry and commerce), and these microregions are in the late phases of the demo-livelihoods theoretical framework during which population growth is more weakly coupled to deforestation.
In conclusion, Amazonas, Brazil, a region still 98% forested, is an important reservoir for protection.The time for that protection is now given its increasing importance over the past ten years to overall deforestation in the Amazônia biome.Its contribution has twice approached 20% of the total deforestation in the past decade.The deforestation in Amazonas occurs primarily as incursions at the edges of the state along border regions with Acre, Rondônia, Mato Grosso, Pará, and Roraima.A reasonable supposition based on linear expectations is that continued incursions deeper into Amazonas might be expected in the coming years.This deforestation can be considered within the coupled deforestation-population frameworks described in the literature for the continuum from family farmers through agribusiness (Caviglia-Harris et al., 2012;Barbieri, 2023).There is also deforestation along the Amazonas River in the metropolitan region of Manaus in central Amazonas.Although Manaus itself is stable with respect to deforestation, its economic activity impacts deforestation in the nearby municipalities.Variability among regions is a common theme in the trajectories of deforestation-population dynamics (VanWey et al., 2011;Caviglia-Harris et al., 2012;Sydenstricker-Neto, 2012), and the microregions identified herein for Amazonas provide three additional contrasting case studies of past and possible future trajectories.The variety of economic models across the three main microregions of Amazonas suggests the possibility of future development without deforestation in Amazonas under a scenario of good governance and economic development (Abramovay, 2018).The converse, however, is also possible.In a widely prognosticated possibility, a point of Amazon dieback might occur, leading through teleconnections to decreased rainfall amounts in regions of food production throughout central and southern South America (Costa et al., 2019).

Data sources
Landcover data were obtained from the MapBiomas Project: Collection 6 of the Annual Series of Land Use and Land Cover Maps of Brazil, https:// mapbi omas.org/ en/ colec oesma pbiom as-1, accessed 2 June 2022.The MapBiomas Project is a multiinstitutional initiative to generate annual land use and landcover maps from automatic classification processes applied to satellite imagery (Souza et al., 2020).The full description of the project can be found at http:// mapbi omas.org.The MapBiomas data are public, open, and free under a Creative Commons CC-BY-SA license.Population data were obtained from the Brazilian Institute of Geography and Statistics (IBGE) (2023a).Data for the geography of Brazilian states and municipalities were also obtained from IBGE (2023b).

Fig. 2
Fig. 2 Minimum and maximum extent of (A) anthropogenic landcover fraction from 1985 to 2020 and (B) population density from 1992 to 2021 for the municipalities of Amazonas, Brazil.There is one horizontal line for each of the 62 municipalities, and the lines are ordered from the highest to the lowest maxima.The lines are separately sorted in each panel, meaning that the top lines in A and B correspond to different municipalities.Selected municipalities discussed in the main text are in color, including Manaus (red), Tabatinga (green), Careiro da Várzea (blue), Urucurituba (purple), and Iranduba (orange).In B, the population density for the municipality of Manaus, shown as the red dashed bar, is divided by 10 for scaling.The analysis for A is based on the MapBiomas dataset for Brazil from 1985 to 2020 (see Sect. "Methodology" of the main text).The smallest and largest values are omitted in the data set of each municipality to eliminate the effects of any single-year outliers.The analysis for B is based on population data obtained from the Brazilian Institute of Geography and Statistics (IBGE) from 1992 to 2021.The underlying data file of the analysis is available (see "Data Availability")

Fig. 3
Fig. 3 Map of the change in (A) anthropogenic landcover fraction between 1985 and 2020 and (B) population density between 1992 and 2021.Coloring is explained in the main text.The dataset used in the coloring corresponds to the interval lengths plotted in A and B of Fig. 2 (see Figures S2-C and S3-C for further detail) (MapBiomas, 2022) calculated based on the summary data available for the MapBiomas product(MapBiomas, 2022).The listed states account for 93% of the land area of the Amazônia biome.Additional states (7%) of the Amazônia biome include Amapá, Mara-