Introduction

Worldwide, biodiversity is declining and tropical forest areas are shrinking, mainly as a result of land-use changes and shifts in social–ecological systems (Curtis et al. 2018). This is mainly driven by the extraction of natural resources, such as minerals, fuel, and timber (Bebbington et al. 2018; Downey et al. 2010; Hosonuma et al. 2012). Understanding changes in land use therefore requires the analysis of trends in extractive activities.

Frameworks have been developed to portray such changes, for example, the multi-level perspective, describing transitions to new socio-technical systems (Geels 2002), and the Forest-Transition Theory that represents the transition between “a shrinking to an expanding forest area” (Mather 1992). However, such frameworks display limitations when it comes to exploring trends in land use: the multi-level perspective specifically considers technological transitions and innovations, focusing on describing the path from “niche” to “regime” more than an established (land-use) system; the Forest-Transition Theory shows (a) a place-based bias—the model was created based on trends observed in the Northern hemisphere; (b) a bias in time—the model uses a single timescale, thereby hiding overlapping processes; and (c) a bias in quality—such as the differences in forest type before and after regrowth (Angelsen 2007; Barbier et al. 2010; Perz 2007; Rudel et al. 2010). Social–ecological realities proper to the Global South, such as colonization, risk being rendered invisible by poorly suited frameworks of analysis (Lorenzen et al. 2020; Perz 2007), causing an oversimplification of the reality, for example the assumption that farming is always the main driver of deforestation (Barbier et al. 2010). Meanwhile, studies point to the importance of considering historical development and multi-level dynamics to understand long-term change in a given landscape (Coral et al. 2021).

The concept of resilience has been broadly applied to analyze changes in social–ecological systems (Allen et al. 2019). While some definitions of resilience have been criticized for failing to recognize the dynamic and multi-scale nature of social–ecological systems (Allen et al. 2019; Garmestani et al. 2020; Jozaei et al. 2022), in this paper we draw from ecological resilience, which recognizes the ability of a system to exist in different states (Allen et al. 2019). Ecological resilience is defined as the ability of a system to absorb shocks, so that it maintains its structure and function; it is the degree to which a social–ecological system can self-organize and adapt (Walker et al. 2004; Gunderson and Holling 2002). The resilience of social–ecological systems can be represented through the adaptive cycle and panarchy, two heuristics that describe change at multiple time and space scales (Gunderson and Holling 2002; Allen et al. 2019). Panarchy puts in evidence the adaptive and transformative capacity of a given system (Jozaei et al. 2022), thus providing a lens through which to assess resilience. Early concerns that panarchy was “highly abstract” and had “little propensity for empirical application” (Walker 2008) have recently been met with a growing body of literature demonstrating the utility of the panarchical approach, for example to study land-use change and to engage policy-makers in improving ecosystem management (Garmestani et al. 2020). Panarchy was also used to provide new perspectives on historical changes in complex social–ecological systems of Latin America and Europe (Antoni et al. 2019; Escamilla Nacher et al. 2021; Jiménez et al. 2020; Winkel et al. 2016), to indicate early-warning signs for regime shifts (Winkel et al. 2016) and to provide guidance for the design of future management interventions (Antoni et al. 2019; Escamilla Nacher et al. 2021; Jiménez et al. 2020). Gunderson et al. (2022) documented the advances made in panarchy scholarship since its conceptualization and the practical ways it can influence natural resource management.

In this study, we apply the adaptive cycle and panarchy (Gunderson and Holling 2002) to a historical analysis of extractive activities in the Peruvian Amazon. Our study area, the department of Madre de Dios, specifically challenges the assumption that agriculture is always the main driver of deforestation and forest degradation, since in this context goldmining plays a more prominent role. Madre de Dios also offers an ideal case to study adaptive cycles and panarchy, due to the rapid changes in land use that have occurred since early colonization. In just over a century, the region went from isolation and inhabitation by Indigenous groups, to becoming a highly intervened and interconnected area of the Amazon showing some of the highest deforestation rates country-wide (Rosales 2019). While the environmental impact of extractive activities such as gold mining or Brazil nut (Bertholletia excelsa) harvesting has been widely studied in Madre de Dios (e.g., Caballero Espejo et al. 2018; Garrish et al. 2014), few studies have looked at historical trends, and important sectors of the economy, for example farming, are still understudied. We examine the degree to which the history of extractivism follows the continuous and multi-scale cycles of adaptive change, and the implications of panarchy for land-use management interventions in the Amazon. We hypothesize that past extraction cycles resemble the adaptive cycle and that, throughout time, these are becoming more numerous as a result of the diversification of livelihood strategies (Callo-Concha et al. 2022). Based on our findings, we discuss possible future scenarios for regional development and how to prevent unfortunate ones.

Theoretical framework

Complex systems are increasingly analyzed under the social–ecological systems (SES) unit of analysis. The concept is based on the perspective that all human systems function in reciprocal connection with their environment: social and ecological processes are part of a same, larger, integrated system (Callo-Concha 2014; Ostrom 2009).

The adaptive cycle (AC), first introduced in the study of ecosystems and later enlarged to SES (Holling 1986), is mainly applied as a heuristic for studying system change (Meuwissen et al. 2019). It describes system dynamics and divides them into four successive phases: exploitation or growth (r), conservation (K), release or collapse (Ω), and reorganization (α) (Fig. 1). Resilience varies throughout the four phases of the AC as follows: it shrinks as the cycle moves from growth (a phase with high uncertainty and potential for growth) to conservation (high certainty and low potential for growth), and then expands again as the cycle moves from collapse to reorganization, letting new opportunities arise in a more flexible and unpredictable state (Gunderson and Holling 2002). Panarchy was later developed to describe change across nested ACs (Gunderson and Holling 2002). Initially created as an antithesis to the rigid, top-down approach of hierarchy, it was designed to be based on integration, transcending boundaries of scale and discipline (Gunderson and Holling 2002). As a part of the resilience approach, AC and panarchy allow the understanding of transformation from one system regime to the next by portraying changes across different spatial and temporal scales (Allen et al. 2014).

Fig. 1
figure 1

The adaptive cycle, adapted from the modified representation of Burkhard et al. (2011) in which the cycle is tilted leftward to better represent social–ecological dynamics

Full size image

Study region

Madre de Dios is one of Peru’s 24 departments (its highest-level districts) and shares borders with Brazil and Bolivia (Fig. 2). It is 85,300 km2 in size and almost entirely covered (> 95%) by forest (Global Forest Watch 2022). Madre de Dios is a hotspot for biodiversity (Myers et al. 2000) and was officially declared the Peruvian “Capital of biodiversity” (Peruvian Law n° 26311). The region is in an early stage of forest transition, meaning a landscape with > 80% of forest, and human population densities around 1 km−2 (Dewi et al. 2017). The climate is tropical with mean annual precipitation of around 2200 mm. Historically colonized by consecutive migration waves, the department remains the least populated nation-wide, with about 150,000 inhabitants (INEI 2017). Today, many migrants, mostly from the highlands, are still driven by the economic prospects offered by gold mining and agriculture. These land uses are also the main drivers of deforestation (Nicolau et al. 2019). Despite large, protected areas (amounting to 44% of the total territory), Madre de Dios is among the Peruvian departments with the highest deforestation rates (0.34% in 2021 alone) (Global Forest Watch 2022). Most recent deforestation hotspots are located along the recently paved Interoceanic Highway—a main driver of economic development (Finer et al. 2017, 2018). The department suffers from weak institutions and high levels of corruption, intensifying the environmental and social challenges (Froese et al. 2021; Mathez-Stiefel et al. 2020).

Fig. 2
figure 2

Map of the current department of Madre de Dios, Peru

Full size image

Materials and methods

For this study, we use the following methods: (1) a literature review, (2) interviews with key informants, and (3) categorization and analysis of the information gathered. Each step is detailed in the following paragraphs.

Literature review on past land use

We conducted an extensive literature review on past and present land uses and resource extractions in Madre de Dios using the Ecosia, Google Web, and Google Scholar databases. We chose these platforms for their inclusion of local, old, and not peer-reviewed material. We used keywords combining the names of natural resources (i.e., quinine, rubber, Brazil nut, timber, gold) with “Madre de Dios” and other relevant terms such as “history” or specific dates, in both English and Spanish. Historic records, mainly consisting of secondary literature, including online and locally published regional scientific productions, were scanned. Peer-reviewed articles were also included to cross with other types of literature. The review also included national statistics and records from local governments, to include quantitative data.

Our search focused on changes in land use in Madre de Dios, including both social and ecological components, such as economic trends, social conflict, demographic dynamics, deforestation, and forest degradation rates that together form the social–ecological system. Once the main extractive activities that drive change were identified, each of them was researched more in depth. This included literature recommended by key informants during and after interviews. The body of literature was considered complete once theoretical saturation was reached, i.e., new documents became redundant with the information already gathered. Based on a detailed inventory of the history of extractive activities, we focused on the most important resources and on the resulting patterns of social, ecological, and environmental change.

Extensive interviews with key informants

In parallel with the literature review, we carried out extensive interviews with key local informants for the main (currently extracted) resources. A list of 17 informants was created based on recommendations from actors in academia and the non-profit sectors, and in two cases through informant’s publications. A selection of 4 informants was then made based on their availability and expertise on the extraction of the main resources still extracted in the region (timber, Brazil nut, and gold). No experts could be found for resources that had not been extracted in significant amounts for more than half a century (quinine, rubber), for which we relied exclusively on literature. For the sake of anonymity, we only name informants by initials: AC is a forestry engineer specializing in timber; MZ is a specialist of Brazil nut; LV is a researcher on gold; TM is a researcher who specializes in the region’s history, to provide insight into how each cycle affected Indigenous populations. The interviews aimed to reflect on and expand the information already gathered in the literature review. Informants were first asked to describe their personal experience and/or source of knowledge about the use of the resource. Then, they were asked to describe the history of its use, the political drivers that impacted it, which actors played an important role in the activity’s development, the impact on changes in land use in the region, and the activity’s more recent development. Finally, the informants were presented with the adaptive cycle and asked to locate the present state of the resource’s use in one of the four phases.

Data analysis

For each natural resource, we identified periods of growth, conservation, release and reorganization, referring to the AC phases. We also identified such changes for the overall social–ecological history of Madre de Dios, by exploring changes in human population and in forest cover. Based on yearly production data of extracted resources, we explored overlaps between ACs of various scales in time (i.e., years, decades, centuries) and space (i.e., local, sub-regional, departmental), and interlinkages between ACs.

Results

In this section, we identify and describe the trends related to important natural resources that have shaped Madre de Dios’ land-use history and influenced its present state: quinine, rubber, Brazil nut, timber, and gold (Fig. 3). We connect the changes associated with each resource to the four phases of the AC. We then zoom out to provide an overview of all ACs and to identify general trends in resource extraction over time and overarching social–ecological change, for example forest cover and human population changes.

Fig. 3
figure 3

Successive and overlapping resource extractions in Madre de Dios, Peruvian Amazon, over the last two centuries, starting from the beginning of the activity’s exponential growth (industrialization in the case of gold, Brazil nut and timber)

Full size image

Extractive activities in Madre de Dios

Quinine extraction

The modern colonization process of Madre de Dios started in the early nineteenth century. In 1800, botanical explorations revealed high densities of quinine trees (Cinchona officinalis), whose medicinal bark was much sought after in Europe (Diaz et al. 2016). The few expeditions aiming to collect bark suffered from violent encounters with Indigenous people, keeping them from having any significant impact on the forest (Moore 2019). In 1860, an English expedition collected seeds to develop high-density plantations in Southern India, which finally succeeded. As a result, the AC of quinine ended in collapse as prices dropped and wild quinine extraction lost its economic value, ending quinine extraction in Madre de Dios (Diaz et al. 2016; García 1982).

Rubber extraction

At the end of the nineteenth century, the collapsed quinine AC gave way to a reorganization phase as colonial states shifted their interest away from quinine and toward rubber (Moore 2019). The growth phase was quick to begin: global prices increased rapidly and, in northern Peru, the industry of rubber extraction grew exponentially (Rumrrill 1984). The activity took a few more decades to expand to Madre de Dios, because the region was isolated from commercial routes and Indigenous populations were still successfully protecting their land. In 1890, Carlos Fitzcarraldo, a rich Peruvian rubber merchant, discovered a shortcut that allowed for a commercial route between the central Peruvian Amazon and Madre de Dios, causing rubber exploitation to expand southward (García 1982) (see “Fitzcarraldo isthmus” in Fig. 2).

Around the same period, the production of rubber latex extracted from overharvested Castilla ulei trees declined, in favor of Hevea brasiliensis and other rubber species that regenerate every year (Morales and Rumrrill 1986). The high abundance of H. brasiliensis in Madre de Dios favored the creation of permanent settlements (Lossio 2001) and a fast growth of the sector. The population grew too, with a first immigration wave in the early 1900s (Morales and Rumrrill 1986). Harvesters were mainly enslaved Indigenous people, and the harsh working conditions, recurring massacres, and the expansion of introduced viruses resulted in the decimation of native populations, eradicating entire ethnic groups and causing a demographic hecatomb (bringing the number down from 40,000 to 2000 Indigenous people) that allowed colonists to occupy the region almost entirely (Huertas and García 2003; Moore 2019). Modern infrastructure expanded: steam-powered boats were introduced and a new commercial route connected the region to the highlands of Puno via the Tambopata River (Fig. 2), allowing for the exportation of a monthly average of 50,000 kg of rubber by land, or over half the department’s production (Morales and Rumrrill 1986; Rumrrill 1984). In the first decade of the twentieth century, the borders among Peru, Brazil, and Bolivia were defined—a result of years of war over land, caused by the political interest fueled by rubber (Huertas and García 2003). This created a need for institutionalism and political representation, leading to the foundation of Puerto Maldonado in 1902, and to the recognition of Madre de Dios as a Peruvian department in 1912 (Dourojeanni 2013; Lewis Ulmer 2014). Meanwhile, economic growth continued to fuel immigration and infrastructure development, for example with the arrival of Japanese migrant workers and the implementation of the first Dominican missions (Ruiz Vela 2012).

After a short conservation phase, the first rubber boom ended around 1912 as a result of highly productive, competing rubber plantations in South-East Asia (Lossio 2001) which the English developed with the stolen seeds of Brazilian H. brasiliensis in 1898. In 1915, those plantations were twice as productive as the entire Amazonian region, and by 1920 they were 8 times more productive (García 1982). In 1915, the rubber AC was already collapsing: Madre de Dios was still exporting up to 800 tons of rubber, but this number was dropping along with international prices (Morales and Rumrrill 1986). In the following years, almost all rubber haciendas went out of business. Regional immigration nearly halted and many settlers left (García 1982). The collapse of the rubber industry is well documented for the department of Loreto, in Peru (Fig. 4). By the time rubber production started to decline in Madre de Dios, the total colonist population was around 5000. As part of the reorganization phase, many of these early settlers reoriented toward farming activities, sugarcane plantation for alcohol export, and wild animal fur and skin commerce (Huertas and García 2003; Morales and Rumrrill 1986). Surviving rubber camps complemented their activities with Brazil nuts harvesting to remain profitable (Lossio 2001). The profits remained low, and this period was one of austerity. Native territories were reclaimed by Indigenous people and some of the original access routes started to degrade, making the region more isolated than previously (Huertas and García 2003; Morales and Rumrrill 1986). The population grew slowly, with the small increase attributed to the arrival of Dominican Christians for evangelization missions (Ruiz Vela 2012). Environmentally, despite a lack of reliable data, it is assumed that rubber harvest had little impact on forest degradation and deforestation rates, and that forests were able to recover naturally (Moore 2019).

Fig. 4
figure 4

The yearly production of rubber (in kg) for the departments of Loreto (North of Peru) and Madre de Dios, showing a collapse after a period of exponential growth. Source: Flores Marin (1977), Maurtua (1911), Morales and Rumrrill (1986) and Pando (2013)

Full size image

A second, shorter, rubber AC took place a few decades later. During the Second World War, Allied powers feared losing their rubber plantations to Japan. The 1941 treaty signed between Peru and USA reactivated rubber extraction in Madre de Dios by nationalizing some of the last surviving rubber camps (Huertas and García 2003). This renewed interest was short-lived and never reached the production volumes of the first boom. Nevertheless, this provided a solution to the region’s isolation and boosted the commercial development of the Brazil nut and timber industries (see below), by opening the first aerial route and constructing a new terrestrial route from Puerto Maldonado to Cusco (García 1982; Morales and Rumrrill 1986). Improved access drove an exponential population growth (Lossio 2001) and the creation of infrastructure such as the first secondary school (Ruiz Vela 2012). After WWII, government-led export ceased, ending the commerce of rubber in Madre de Dios (García 1982).

Brazil nut extraction

After the collapse of the first rubber AC, the extraction of Brazil nut, the seed of the Bertholletia excelsa tree, became an important economic activity (García 1982). In 1940, the first nut sheller was introduced in the region, allowing for the upscaling of processing (Huertas and García 2003), and nut production entered an industrial phase in the 1950s. Global demand attracted large companies who mechanized the peeling and processing to make the nuts export ready (Morales and Rumrrill 1986). The Brazil nut AC entered the growth phase in the 1960s, when production became exponential, boosted by the completion of the new road (García 1982). In the late 1960s, the first Brazil nut export companies were established in Puerto Maldonado, and an official airport was built, leading to further increase in immigration rates, also motivated by the 1968 agrarian reform (Huertas and García 2003). Around that time, yearly Brazil nut production stabilized around 2000 tons before increasing again in following decades (Lossio 2001), favored by ever-improving access conditions (MZ, personal communication July 2021). Overall, total production volumes continued to grow steadily throughout the twentieth century, even though historic data are only available for exported Brazil nuts (Fig. 5). This steady growth suggests that the AC was in the conservation phase. Around 2008, the price of motorcycles dropped significantly, allowing many smallholders to buy one and accelerate the harvest, which until then was only done by foot. The process became better organized and more processing plants opened in the region. Prices started to peak, compensating for unstable and often decreasing production per tree (MZ, personal communication July 2021). In parallel, forest degradation was increasing too. Studies have found that Brazil nut harvesting exerts pressure on the ecosystem because the activity is correlated with increased wildlife hunting, timber exploitation, and even deforestation due to subsequent farming activities (Escobal and Aldana 2003). Nevertheless, the activity received considerable support from environmental conservation programs, which see it as a more sustainable activity than logging or gold mining.

Fig. 5
figure 5

Yearly production of Brazil nuts (in tons of seeds) for the department of Madre de Dios. Sources: FAO (2021) (for exported seeds), Regional Department of Forestry and Wild Fauna (GRFF 2021)

Full size image

Today, the activity is reaching a point of inflection, suggesting a transition to a phase of collapse or release. The solid line in Fig. 5 (representing regional government data) suggests a region-wide drop in production since 2011. Production in the National Reserve of Tambopata (an iconic protected area of the department) has also been in steady decline since the early 2000s (− 88% tons of seeds between 2004 and 2019). Several trends could be causing this drop: the decline in productivity of Brazil nut forests, where large trees are dying and few seedlings are left for natural regeneration; deforestation causing the loss of natural pollinators and the loss of natural habitats for parrots, which therefore feed on young Brazil nut fruits; and climate change, such as the fall in production around 2016 linked to droughts (MZ, personal communication July 2021; Jansen et al. 2021; Pastana et al. 2021; Thomas et al. 2021). Aside from the ecological explanations, a drop in productivity could also be due to higher labor cost or alternative labor opportunities (e.g., gold mining). Recently, the COVID-19 crisis caused a sharp drop in global prices and serious losses for harvesters. Some Brazil nut producer associations are already anticipating a future collapse of the activity and have started to work on Brazil nut tree restoration in agroforestry systems (MZ, personal communication July 2021).

Timber extraction

Similar to Brazil nut, timber extraction was practiced for local use for decades, before becoming a commercial activity as a result of road access and industrialization (Morales and Rumrrill 1986). The timber AC growth phase started with the construction of the road in the 1960s (see above) which generated a first timber extraction “fever”, supported by the arrival of small sawmills and the targeting of valuable species, such as mahogany (Swietenia macrophylla) and Spanish cedar (Cedrela odorata) (Dourojeanni 2013; SPDA 2016). Environmental concerns over uncontrolled exploitation, which had already started arising in the 1950s, motivated public institutions to order forest surveys, which then served as a basis for the creation of the department’s first protected area in 1973 (Dourojeanni 2013; Huertas and García 2003). Nevertheless, timber companies simply relocated and continued to grow their activities in non-protected areas (Moore 2019). Furthermore, the sector benefited from political and institutional support throughout the 1970s (Morales and Rumrrill 1986): the 1975 Forestry Law (Decree Law 21147) provided the first legal basis for commercial timber extraction, and the 1978 Native Communities Law (Decree Law 22175) facilitated arrangements between companies and communities (TM, personal communication May 2021). All this led to a second timber extraction “fever” (characteristic of the AC growth phase), which fueled further infrastructure development and attracted investors from other regions (SPDA 2016). In the 1980s, the first industrial sawmills were built, and entrepreneurs surpassed the legal limits of extraction areas (TM, personal communication May 2021). Around the turn of the century, mahogany wood prices skyrocketed, generating an extractive boom that left the species locally extinct in many areas (TM, personal communication May 2021). The conservation phase may already have been reached in 2000, when a modification in the Forestry law changed the timber extraction landscape: a system of forestry concessions was established, consisting of a 40-year lease of land by the state (Moore 2019). Official figures show a sharp increase in (formal) timber extraction around that time (Fig. 6). Initially believed to help fight illegality, the law created a wave of corruption and social issues instead. Indeed, large-scale industrial timber concessions were authorized, benefiting large-scale foreign actors, pushing small loggers to illegality and threatening native people’s land rights (Moore 2019; Ruiz Vela 2012; TM, personal communication May 2021). In addition, improvements in sawmill technology allowed for the overextraction of hardwood species and new roads allowed for the logging of previously inaccessible forests. The law was changed again in 2015 and, despite an intention to be more inclusive of local stakeholders, it forced local and native communities to form unequal partnerships with timber companies (Moore 2019). This legal change and the decline in high-value timber species led many to log illegally or to abandon this activity (TM, personal communication July 2021). Indeed, public data, despite discrepancies between regional and national data sets, all show a sharp drop in legally produced timber around 2015 (Fig. 6).

Fig. 6
figure 6

Source: National Forestry Service (SERFOR 2021) and Regional Department of Forestry and Wild Fauna (GRFF 2021)

Yearly production of wood (in cubic meters) for the department of Madre de Dios.

Full size image

Today, timber exploitation (responsible for forest degradation) is controlled by large corporations that invest in new logging roads, resulting in land invasion by landless farmers (responsible for deforestation). Insufficient resources available to local authorities have kept chaos and illegality rates high (Urrunaga et al. 2012; Ruiz Vela 2012). As in the case of the Brazil nut, some believe the timber industry as we know it today is close to collapsing. The overharvested species need more time to regenerate than the 40-year lease cycle established in the Forestry Law. When concessions end their first lease round (around 2040), there may be almost no more highly valuable wood left to harvest (AC, personal communication May 2021). A possible future scenario could mirror what has happened in other regions of the Amazon: concessionaries abandon their land, which gets invaded by landless farmers and land speculators. The timber commerce would continue at a much lower pace, with low-value species (AC, personal communication May 2021).

Gold extraction

Ever since the Spanish conquest, gold has been one of the most sought-after resources in the New World. The modern era of gold mining began in the 1930s, after a global recession caused prices to peak (Salo et al. 2016). This first small wave generated a familiar pattern: increased immigration from the highlands, social dispute over land, and political interests of the national state (Huertas and García 2003; Morales and Rumrrill 1986). The gold AC was already in full growth phase in the 1970s, when the activity became of major importance due to both a strong increase in gold value and the discovery of high gold concentrations along rivers (García 1982). In 1972, mining companies were replaced by a public entity: the Peruvian Mining Bank. This institution incentivized miners with small credits and advantageous prices for tools and had exclusive rights to trade gold (Moore 2019). Emblematic of a national government aiming to colonize the Amazon, the Mining Bank granted land to small miners first, then to private companies (García 1982). In the next decade, both the price and production of gold increased almost 20-fold (Fig. 7). Improved regional access allowed for a demographic avalanche, bringing the number of gold miners from 300 in 1975 to 20,000 only 5 years later (Morales and Rumrrill 1986). Some villages became the fastest growing population centers in all of Peru (Huertas and García 2003). In the early 1980s, private companies bypassed the Mining Bank and took over business, launching an unstoppable privatization process. The illegal trade of gold increased and brought mining companies and Indigenous communities into conflict (Moore 2019; Morales and Rumrrill 1986). Mining grew exponentially and boosted development in terms of population growth, economic output (surpassing that of the best rubber years), and employment (Morales and Rumrrill 1986). In the 1990s, neoliberal policies further promoted privatization, mechanization, and year-round extraction (Cortés-McPherson 2019). Social conflicts grew, with Indigenous rights threatened once again (Moore 2019; SPDA 2016). In an attempt to take back control, the national government implemented the first regularization policies in 2002. But the sector, entering the conservation phase of the AC, continued to grow chaotically and informally, driven by road pavement (Cortés-McPherson 2019), exponentially increasing gold prices (Moore 2019), adoption of new technologies (Caballero Espejo et al. 2018), and corrupt political institutions that both supported the activity and hindered the sector’s governance (SPDA 2016). As a result, informal mining exploded, along with poor working conditions and human trafficking (Ruiz Vela 2012; SPDA 2016). The activity that had, until then, mainly concentrated in particular areas, began to spread, with small mines becoming dominant (Caballero Espejo et al. 2018). Social unrest grew and a third cycle of social conflict with Indigenous communities occurred (SPDA 2016). Environmentally, gold mining became the leading cause of deforestation, causing a greater area to be deforested than during the two previous decades (Caballero Espejo et al. 2018) (Fig. 7).

Fig. 7
figure 7

Registered yearly gold production in Madre de Dios and global gold prices. Sources: Morales and Rumrrill (1986), World Gold Council (2021), BCRP (2021) (gold prices), De La Torre (1987) and MINEM (2021) (gold production)

Full size image

Around 2010, gold production reached a maximum (Fig. 7). The related deforestation also reached a plateau (Caballero Espejo et al. 2018). Successive governments failed to regulate the activity, of which less than half was estimated to be legal (Ruiz Vela 2012). Mining in the buffer zone of a protected area resulted in an emergency decree that declared the regularization of this activity to be of national interest (Cortés-McPherson 2019). However, interventions by the police and military led to social protests from miners and failed to end mining activities, which even increased in 2013, 2015 and 2016 (Fig. 7), driven by skyrocketing gold prices (Cortés-McPherson 2019; SPDA 2016). Yet, since 2017, anti-mining policies have managed to bring gold production down (Fig. 7). In 2019, military interventions on the most affected areas contributed to sustaining the downward trend (SPDA Actualidad Ambiental 2020). Mining-related deforestation dropped by 62% and concentrated in new areas, mainly native community territory (Finer et al. 2021). Today, the grand majority of miners work on a small scale and do not fulfill the regularization’s legal requirements (TM, personal communication May 2021). By using artisanal tools, these small-scale miners can only mine the superficial layer of the soil, as opposed to industrial techniques that access deeper layers (LV, personal communication June 2021). This keeps small miners from accessing the 71 tons of gold total estimated to be available in the region (MINEM 2020) and forces them to move when a given region is depleted. Some miners reflect on the much higher quantities of gold that could be extracted on an average mining day decades ago, compared to the smaller quantities (of more valuable gold) they are able to mine nowadays (LV, personal communication June 2021).

Panarchy as a framework for extractive activities

The colonial history of Madre de Dios, when seen through the lens of extractive activities—as described in the above sections—shows patterns resembling those of the AC, altogether forming the region’s panarchy. Interestingly, it appears that earlier ACs (quinine, rubber) have triggered the start of subsequent cycles, with overlap between reorganization and exploitation (growth) phases of consecutive cycles. In addition, natural resource ACs show a pattern of diversification over time and are part of overarching social–ecological trends (Fig. 3).

An overview of the consecutive adaptive cycles

When organized into the four phases of the adaptive cycle: exploitation, conservation, release, and reorganization, it seems that only the extraction of quinine and rubber have gone through all phases. The short-lived quinine AC, beginning in 1850, followed a short exploitation phase with a negligible conservation phase and a sudden release. As a marker for the start of the human colonization process, this first AC opened the way to the second one, rubber extraction. In this case, a clear phase of exponential growth, followed by a few years of consolidation, ended in an abrupt release. Reorganization was a return to more traditional, subsistence ways of life. Interestingly, both the quinine and rubber ACs entered the release phase after the resource got bio-pirated by British prospectors who developed more productive systems in Southern Asia. In both cases, resource depletion had not yet become a problem, and instead geo-politics and macroeconomics defined the fate of the extractive activity. A second rubber boom, resembling a small AC with an almost negligible conservation phase, can also be perceived as a part of a long regional reorganization phase leading to a reopening to national markets (Table 1).

Table 1 Organization of the history of five primary natural resources extracted over the last two centuries in Madre de Dios, Peruvian Amazon, into the four phases of the adaptive cycle
Full size table

Only decades after those two first ACs were other activities able to take off. Improved infrastructure in the 1960s allowed for the timber—and to a lesser extent Brazil nut—industries to enter the exploitation phase. Brazil nut reached the conservation phase a few decades later, with continuous steady growth, and timber culminated in the early 2000s. The gold AC exploitation phase started in the 1970s, generating a trifurcation in the system’s panarchy. Gold production, dominating all other productive activities in terms of economic importance, started to level off around 2010, entering the conservation phase. Today, all three ACs (Brazil nut, timber, gold) show signs of deceleration and it is unclear if any of them have reached a release phase, but local actors anticipate a reorganization phase. In the case of Brazil nut and timber, harvesters note that the industries are reaching a tipping point. Gold could possibly follow a similar trend, since most miners work on a small scale (Table 1).

Diversification patterns and overarching adaptive cycles

An overview of all trends in resource extraction over time shows a clear pattern of diversification (see Fig. 3). While early cycles (i.e., quinine, rubber) happened successively, the more recent cycles overlap with one another. In parallel, longer-term social–ecological trends (e.g., forest degradation, deforestation, migration waves) have influenced and been influenced by the natural resource ACs. Those trends can be assimilated into ACs as well: the forest ecosystem  is suffering exponential deforestation and is therefore possibly in the early phase of collapse; the pre-Hispanic demographic system has already collapsed (with a close to 95% population decrease) and the post-colonial demographic system is showing decreased growth rates since the 1990s, a characteristic of the conservation phase (INEI 2020). More punctual events, happening on a shorter timescale and smaller space scale, such as the local extinction of mahogany, form part of the natural resource ACs and can also be looked at through the lens of the AC.

Discussion

As demonstrated in “Results”, the historical analysis of social–ecological changes in Madre de Dios uncovers a cyclic pattern that resembles the AC. The panarchy of the region is characterized by a diversification of ACs over the last century and the presence of larger and smaller cycles at multiple scales (i.e., time and space). In this section, we discuss our main finding and draw implications for future scenarios.

Consecutive and overlapping adaptive cycles

The Amazon region, historically considered as a land “empty of people” but rich in resources (Langewiesche 2022), has seen rapid cycles of expansion, stabilization, and decline in succession. Such dynamics, observed in Northern Brazil, led Homma (1992) to theorize that all extractive activities eventually decline, only to be replaced by agricultural activities. This transition can be attributed to technological, demographic, or economic factors (Homma 2012), akin to the “boom-and-bust cycles”, during which trends in supply and demand cause changes in land use (Andreotti et al. 2022). In Madre de Dios, this tendency was noted by Huertas and García (2003), who stressed that international market prices influence the extraction of any given resource more than political and social dynamics. When the economic value of commodities follows cycles of fast growth, stabilization, and decline, it is no surprise that the extraction of (commodified) natural resources follows the same trend. In the middle of the twentieth century, the development of infrastructures in Madre de Dios allowed for a bi- and even trifurcation in the region’s panarchy. Road access generated a multiplication of market opportunities, allowing for several commodities to “boom” around the same time and generate a diversification of land uses. This diversification of land uses, and thereby livelihoods, was also found to be a characteristic trait of the Madre de Dios social–ecological system in a cross-national analysis of the Southwestern Amazon (Callo-Concha et al. 2022).

Relationships among different scales

One of the core principles of panarchy is that ACs exist at various scales in time and space. Larger cycles (e.g., happening across an entire department) tend to be slower, whereas small ones (e.g., happening locally) are faster (Perz and Almeyda 2009). In our analysis, we have identified three main scales of change: large and slow ACs, medium ACs, and small and fast ACs (Table 2). The large, overarching ACs happen on timescales of centuries and cover the entire department (our area of analysis). They influence and are influenced by all other ACs happening at smaller scales, i.e., a sub-regional (or sub-departmental) scale (Gunderson and Holling 2002). For example, early medium ACs (quinine, rubber) had low impacts on the forest ecosystem, but a dramatic one on the human population, which in turn had large subsequent impacts on the forest ecosystem. The rubber cycle, in particular, disrupted the pre-Hispanic social and economic systems of Indigenous groups to the point of collapse. Later, medium ACs (timber, gold) impacted the forest ecosystem greatly, with significant resulting forest degradation and deforestation (Moore 2019). In parallel, these ACs were strongly tied to population dynamics, in this case consecutive colonizing migration waves, by both benefiting from and fueling them. This allowed for the colonial population AC to enter exploitation, with exponential growth between the 1940s and 1990s (INEI 2020), and to then settle in the conservation phase. As a result, only a small proportion of today’s population descended from Indigenous populations, and most extractive activities are carried out by first- or second-generation descendants from neighboring departments. This peculiarity raises the question of how a traditional cultural relationship to the land plays a role in avoiding excessive resource extractivism. One could argue that it is easier for new coming settlers to see a natural resource as a commodity and degrade a landscape that is not part of one’s cultural heritage. This point relates to the findings of higher forest cover in areas populated by Indigenous groups (Sze et al. 2022). Social–ecological systems not only get influenced by lower-scale ACs, but also by each other. For example, researchers increasingly warn about the effects that the crossing of an ecological tipping point can have on social systems (Lovejoy and Nobre 2019). Small and fast ACs happen on timescales of years and cover only small, local areas. They are usually embedded in medium ACs. For example, the timber AC, incomplete to this day, has seen some full ACs happen on smaller time and space scales, with the local extinction of mahogany trees. The second rubber boom can also be understood as a smaller, faster AC embedded in the larger rubber reorganization phase (Table 2).

Table 2 Time and space scales of various adaptive cycles identified in the historical analysis of Madre de Dios, Peruvian Amazon
Full size table

Exploring future land use through the conceptual lens of panarchy

Panarchy exposes the intrinsic dynamism of social–ecological systems that renders change both fundamental and inevitable (Garmestani et al. 2020; Jozaei et al. 2022). The dynamics we describe in this article suggest that natural resource extraction in Madre de Dios is currently in a later conservation phase, characterized by limited potential for growth and low resilience. As extractive activities enter a release phase, a transition toward alternative land uses becomes possible. Activities currently in the early exploitation phase (e.g., farming, eco-tourism) with high potential for growth and resilience could take over from activities currently in the conservation phase and possibly close to a release (e.g., timber, gold). The departmental GDP over the last 15 years shows an increasing share of mercantile and agricultural activities (+ 146%) versus a decline in gold mining (− 49%) (INEI 2021). In this section, we will explore the possibility that farming will become the leading economic activity in Madre de Dios in the coming decades.

Traditionally, farming in Madre de Dios was aimed at subsistence, until the collapse of the rubber industry allowed for the beginning of small-scale commercial agriculture (Morales and Rumrrill 1986). The immigration wave of the 1960s led many Andean peasants to settle along the road and practice rotational (slash-and-burn) farming (Rumrrill 1984), and the agrarian reform in 1969 promoted the creation of cooperatives (Morales and Rumrrill 1986; Rumrrill 1984). Yet, the focus on extractive activities, such as rubber and gold, kept the agricultural sector from growing. Government programs to advance the agriculture frontier only focused on international border areas and most ended in failure due to poor technical support (García 1982; Huertas and García 2003; Morales and Rumrrill 1986). The economic crisis and neoliberal policies in the 1990s mostly left farmers to fend for themselves (Rojas, not published). To this day, the agricultural sector is smaller than other Peruvian departments, both due to low productivity and a small proportion of land dedicated to agriculture (Rojas, not published). However, for the first time the farming sector is gaining the attention of a diversity of actors with diverging interests. Cacao is being promoted as a sustainable land use capable of ending slash-and-burn expansionism and as an alternative to illegal gold mining (Moore 2019). In parallel, the agriculture office of the regional government and private actors (e.g., papaya producers) are pushing for intensification through mechanization and technification. As a result, the farming landscape has changed significantly over the last decade: between 2010 and 2016, the cultivated surface of cacao has increased 80-fold, papaya almost 10-fold, and more traditional rice cultivation has decreased 17-fold (INEI 2017). Even illicit farming activities, such as the plantation of coca (Erythroxylum coca) have started to spread (El Comercio 2021).

While other land uses are seemingly reaching a conservation phase in the AC, farming is growing exponentially: between 1994 and 2012, the agricultural area and number of producers have both more than doubled and, between 2005 and 2018, yearly agricultural product exportation went from 1.9 to 18.9 million US$ (BCRP 2022). This transition from extractive economies to farming has been extensively discussed by Homma (2012), who argues that, while extractive activities have historically been key for the development of Amazonian economies, these activities have no future. Limited supplies, growing demands, urbanization, and the emergence of alternative sources of employment are only a few of the factors that make extraction obsolete in the longer term. In addition, the agricultural frontier’s expansion is in direct competition with extractive activities, since extractors often become farmers themselves (Homma 1992). In the future of Madre de Dios, we foresee that commercial agriculture will become the most important activity, both economically and in terms of land use (Fig. 8).

Fig. 8
figure 8

The panarchy of Madre de Dios shows various nested adaptive cycles at different time and space scales. There are two full adaptive cycles in the earlier colonial period (quinine and rubber), multiple overlapping adaptive cycles (gold, timber, Brazil nut), and overarching forest and human population adaptive cycles that influence and are influenced by lower-scale, smaller, faster adaptive cycles. Full lines represent past development, dotted lines represent future, expected development

Full size image

Implications and recommendations

In this paper, panarchy was proposed as a framework for revealing possible trends in the economic activities of the Madre de Dios region. It has allowed for the consideration of a diversity of drivers of change, as well as small and large changes. The interplay between these drivers of change determines the dynamics of social–ecological systems, in this case, the diversification of livelihoods and economic activities.

Further, the panarchy heuristic warns for foreseeable changes, thus pointing to the need to foster systems’ transformative capacity and adaptability (Garmestani et al. 2020; Jozaei et al. 2022). Collapse and regime shift are brought forward by the loss of resilience happening during transition from growth to conservation (Gunderson and Holling 2002). As this is to be avoided from a societal point of view, it has major implications for the design of management interventions in a context of rapid social and environmental change. As we have shown, past economic activities in Madre de Dios have collapsed due to their overreliance on a single resource (i.e., rubber), and current ones (Brazil nut extraction, timber, and gold) are vulnerable to both global market fluctuations and overexploitation. Hence, interventions seeking to influence the social–ecological system need to pay close attention to emerging sectors (e.g., farming) and the diversity (or lack thereof) of resources that are being relied on. For example, current efforts to conserve the rainforest focus on strengthening extractive activities, such as timber and Brazil nut (see Montoya-Zumaeta et al. 2022), making conservation contingent on the continuation of activities that are losing in importance. Conservation initiatives have so far failed to shape land-use dynamics, which is also attributed to a lack of integration with the agricultural sector (Rodriguez-Ward et al. 2018). By ignoring the current decline in extractive activities, conservation interventions fail to envision a transition from an overreliance on few commodities to a more diverse and resilient system.

Our study suggests that the future development of Madre de Dios will likely lead to the dominance of agricultural activities—as has already occurred in the Brazilian Amazon (Callo-Concha et al. 2022; Homma 2012). This implies that the regional farming system, today still understudied and overlooked, should become a key leverage point for both conservation and land-use policies. According to Homma (2012), environmental policies that focus on extractivism are doomed to fail, especially in regions where the agricultural frontier is expanding—in which case policies targeting the agricultural system would be better suited for addressing environmental issues. Management interventions should not seek a return to “normal”, but instead acknowledge the need to either adapt to changes or transform by drawing from past lessons (Jozaei et al. 2022). In Madre de Dios, adapting to change implies that environmental land-use policies should consider the growing prominence of the agricultural sector. To draw from past lessons will mean to help the rapidly expanding agricultural sector avoid the risk of overreliance on and overexploitation of few resources, to maintain resilience and prevent repeating patterns of boom and bust.

One way to buffer the larger social–ecological system against rapid change and to make it more resilient is to allow for multiple smaller ACs to play out at the same time (e.g., multiple crops at the level of farm system) (Kuhmonen and Kuhmonen 2023). For instance, today’s agricultural interventions in Madre de Dios focus on only a few commercial crops, like cacao, causing farmers to specialize, thereby increasing their dependency on markets. Land-use practices that are more diverse and multifunctional, such as agroforestry, have the potential to increase resilience, but their success depends on the development of a diversity of supply chains, as opposed to a focus on a few exportable commodities (Lagneaux et al. 2021). Conveniently, the agricultural system is still benefiting from high potential for growth and flexibility, both characteristics of the AC “growth” phase it is currently in, which is conducive to the development of resilience-building strategies. Such strategies could be applied at various scales for resilience at both the farm level (e.g., though the diversification of land use and crops) and landscape level (e.g., by adapting interventions to the peculiarities of diverse sectors).

We propose that future land-use policies in the Peruvian Amazon should take into account the expanding agricultural sector and investigate ways to enhance its diversify and strengthen its resilience. Adaptive approaches, based on the AC and on understanding past trends and present realities, help to identify scenarios and foresee suitable interventions. We show that the adaptive cycle and panarchy can be useful heuristics for land-use change studies and allow the identification of adaptive approaches for sustainable land use.