Integrating collapse theories to understand socio-ecological systems resilience

Historically, we find that civilizations are built around water sources and land resources that make agriculture possible. Mesopotamia, Egypt, China, the Indus Valley, Andean South America, and central Mexico are some examples (Lucero 2002). This is not surprising, since it was agriculture that made placed-based societies possible (Ehrlich and Ehrlich 2013). But agriculture was not a panacea (Ponting 1993). We can imagine our ancestors after discovering agriculture asking themselves how to increase water availability and stability to feed the growing population that was a product of the now stable food source, and so on in a vicious cycle (Tainter 1988, Harari 2016). From this example we can easily perceive the resilience-fragility tradeoffs inherent in securing the most basic needs of societies: food and water.

History reflects how, in different cultures, human curiosity, intelligence, and imagination have been enough to overcome these challenges and build different irrigation infrastructures and the rules to govern them, not only to manage the valuable resource, but also to respond to environmental threats (Diamond 2005, Costanza et al 2007, Ehrlich and Ehrlich 2013). However, we have also observed that many civilizations were unable to overcome challenges, and collapsed (Diamond 2005).

The world is facing new environmental challenges that may trigger a collapse, or, more precisely, 'loss of socio-political-economic complexity usually accompanied by a dramatic decline in population size' (Ehrlich and Ehrlich 2013, p 1), of some socio-ecological systems (Young et al 2006). Field et al (2014) predicts that more extreme weather events, like heat waves, droughts, floods, and violent storms, may be much more common in the decades to come due to climate change. Although we have an idea of what climatic events to expect in each region, we know less about how systems can cope with these challenges (Field et al 2014). The Peruvian Piura Basin, an agricultural based society, is highly exposed to environmental perturbations (Angulo 2006). The basin has been exposed to harsh environmental events associated with the El Niño Southern Oscillation (ENSO) for centuries and is now threatened with more frequent and intense flood episodes due to climate change (Bustamante 2010). The Piura basin was also home of an ancient civilization named Moche, which collapsed due to a combination of factors, but strong El Niño events likely played a significant role (Fagan 2009). Given that many societies have experienced droughts, floods, or other environmental perturbation without showing major social disruption (Tainter 1988, Diamond 2005), authors suggests that the resilience of socio-ecological systems (SES) to those disturbances may depend on combinations of specific factors such as political, economic, or ecological characteristics (Motesharrei et al 2014). We perform a systematic literature review to identify prominent collapse theories and extract propositions as key factors observed in SES collapses. We then leverage those propositions to analyze the resilience of SES to environmental disturbances, using as a case study the Peruvian Piura Basin.

A resilient system does not necessarily perform at its maximum potential (Csete and Doyle 2002), but it does remain functional despite internal (e.g. population growth) or external (e.g. droughts) perturbations. To study resilience of a SES, we need to be explicit about specific perturbations, given that systems face tradeoffs between resilience to particular expected disturbances and uncertain fragilities to others (Janssen and Anderies 2013). We analyze the resilience of the Piura Basin to flood events, and by performing a longitudinal analysis of the basin we illustrate how this outcome has been affected by efforts to increase water availability and stability to feed a growing population, among others.

This article is arranged in the following sections: (i) we describe our case study, the environmental perturbation to be analyzed, the framework and methods used; (ii) we highlight prominent collapse theories and their propositions for resilience-fragility trade-off analysis; (iii) we assess fragility-resilience trade-offs in the Piura Basin; (iv) finally, we conclude. We found that the Piura basin is apparently very fragile based on almost all of the predictions of collapse theories, but an important strength is its growing stock of social capital. In small steps, user associations have been collectively working towards solutions for water conservation and public-infrastructure maintenance. There is a long way to go yet, but with the right policies to encourage the strengthening of these associations, the Piura basin could become more resilient to future El Niño events.

2.1. Case study

The Piura River is 280 kilometers long, and the basin surface is 12 216 km² (see figure 1). The basin is divided into two irrigation systems, 'Alto Piura' on its right margin in the highlands, and 'Medio y Bajo Piura' on its left margin on the coast. This study is focused on the Medio y Bajo Piura sub-basin, because this area is more exposed to El Niño flood events than the Alto Piura sub-basin. Water from the Piura River is almost completely used before reaching Medio y Bajo Piura, but the sub-basin receives water from the Poechos Dam (on the Chira River) through the 'Daniel Escobar' canal that was built in the 1970s (GRP et al 2009).

Zoom In Zoom Out Reset image size

The Piura River's flow is normally low, but in El Niño events the river grows to a point that it becomes dangerous. For example, in the station of the river that is called 'Sanchez Cerro,' river flow in a normal year is no more than 140 m³ s⁻¹ at its highest, but in 1983 it increased to 3200 m³ s⁻¹, and in 1998 to almost 4500 m³ s⁻¹, damaging irrigation and road infrastructure (GRP et al 2009).

Agriculture is an important activity in the basin, involving about one-third of the population (INEI 2016). 48 534 ha are used for agriculture, of which 84% is irrigated. There are 75 176 farmers with small parcels averaging 0.65 ha. By 2016, the main crops in the Medio y Bajo sub-basin were rice (67%), corn (23%), and cotton (7%), all of which have a safe market, with their growers having access to credit and technical assistance.

2.2. ENSO effects on the Piura Basin

2.3. Framework

For internal validity, and to facilitate logical reasoning (Miles and Huberman 1994, Yin 1994) we use the Robustness Framework proposed by Anderies et al (2004) as a guide for data collection and analysis. The Robustness Framework was created to systematically develop a theory of SES robustness; a concept related to resilience that also addresses the question of conscious institutional designs (Anderies et al 2004, Janssen and Anderies 2013).

Unlike other frameworks for resilience (e.g. Leach et al 2010, Garmestani and Harm 2013), or SES (Ostrom 2005, 2007) research, the Robustness Framework enables the analysis of system relationships (links 1–6 shown in figure 2) among key SES components (A–D) to foresee potential perturbations (links 7 and 8) and the system outcomes they might produce (Anderies 2016). Hence, we use the Robustness Framework to analyze how in a specific context (Piura Basin), the characteristics of the resource system (A; watershed), their main users (B; farmers), public infrastructure (D; reservoir, canals, rules, laws), public infrastructure providers (C; regional and national government, water association), and the relationship among them, contribute to the resilience of the basin to flood events. The framework was also used to organize identified propositions from collapse theories reviewed. Components (A–D) and Links (1–8) were used to code selected research works and to identify and classify proposition as explained in the next subsection.

Zoom In Zoom Out Reset image size

2.4. Literature review

For building validity, we follow the analysis of propositions (Miles and Huberman 1994, Yin 2003) from theories or hypotheses proposed in previous research of collapse of societies. We reviewed articles and books (referred in 4) that, by applying case study comparisons, proposed explanations for the collapse of societies beyond exogenous environmental or other perturbations. Even though the complexity of SES makes it impossible to find key recipes of collapse or resilience (Weiss and Bradley 2001, Costanza et al 2007 , Butzer 2012, Middelton 2017), we consider each of these propositions as elements that can shed light on the design of resilient systems. Following the systematic literature review method suggested by Gough et al (2012), we developed a search protocol with inclusion criteria for the selection of studies to be reviewed. A flowchart of studies selection and selection criteria are detailed in figure 3. To locate references, we used Google Scholar search engine by entering as key words: 'society', 'civilization' and 'collapse'. A total of 48 500 references matched with the search criteria, which after filtering by relevance and title screening 250 studies were preselected. 38 references where chosen after abstract screening, and after a full text review 36 references were selected for its inclusions in the analysis.⁵ Figure 4 shows how the propositions extracted for the literature review were organized with the guidance of the Robustness Framework, along with the references of authors that supports each of them.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

2.5. Data collection and analysis

Societal collapse is a fascinating topic that has been of interest to many scholars (Butzer 2012). As a result, it is possible to find diverse propositions as determinants of socio-ecological systems (SES) collapse (see figure 4), that recognize the combination of environmental or other external perturbation as stressors, and weak institutional arrangement as determinants of collapse (Butzer 2012). Tainter (1988), for example, suggests that as societies mature, institutions and hard infrastructure become so complex and rigid that they become extremely vulnerable to shocks. When a civilization reaches that point, it becomes unable to withstand a disturbance of any type (environmental, wars, social, etc), and the population either disappears or migrates out of the system. Brunk (2002) agrees with Tainter's theory, and argues that in societies that are more interconnected, cascade effects of collapses may occur.

A lager group of scholars (Postan 1966, Culbert 1973, 1988, Abel 1980, Catton 1982, Ladurie 1987, Ponting 1993, Kammen et al 1994, Wood 1998, Holling 2001, Redman 2004, 2005, Wright 2004, Ehrlich and Ehrlich 2013, Motesharrei et al 2014) argues that a major cause of collapse, is the 'overshoot effect,' in which case a large population demands more than the available resources in a system. As a result, people either migrate to another system, or perish. The overshoot effect is also considered in the 'release' phase of adaptive-cycle explanations (Holling 2001) from resilience theory. Combined with the overshoot effect, Pezzey and Anderies (2003) propose that culturally defined subsistence needs affect the process of collapse due to the high levels of what Motesharrai et al (2014) referred to as 'depletion per capita'. They argue that it is not only the population-resources ratio, but also, and most importantly, population consumption needs perception (and thus their consumption) that determines the 'release' or 'overshoot' point.

Another prominent hypothesis suggests that economic stratification of society into elites and masses, has a key role in societal collapses (Khaldun 1958, Brenner 1976, Timmerman 1981, Goldstone 1991, Parsons 1991, Ponting 1993, Diamond 2005, Turchin 2005, 2006, Turchin and Nefedov 2009, Motesharrei et al 2014). Authors stress the relevance of societies' perceptions of equity, arguing that class conflicts can create internal disturbances that weaken the system as a whole. Others suggest that the centralization of power in few decision-making actors has an effect in the speed and quality of reaction to perturbations, making it suboptimal for adaptation (Yoffee and Cowgill 1991, Lucero 2002, Butzer 2012)

Janssen et al (2003) proposed that a sunk-cost effect should be also considered as a factor that may contribute to SES collapses. They claim that some societies that collapse may have been attached to the efforts of achieving high returns after completing large investments, even when experiencing the opposite. As a result, they may have been unsuccessful to overcome a perturbation by failing to switch strategies. Diamond (2005) after studying 15 systems that collapsed and some others that did not, found some commonalities in the cases that collapsed, and brought to the conversation additional potential causes of collapse, that are also supported by other scholars: (i) selfish elites, (ii) moral decay (iii) physical constraints to adapt, and (iv) limited anticipation capacity.

Selfish elites refer to rulers that were benefiting most from the system opposed change at the expense of society and who do not favor ecological conservation either (Butzer 2012). The moral decay proposition refers to society´s attachment to values and activities that were detrimental to the ecological environment (Timmerman 1981, Butzer 2012, Ehrlich and Ehrlich 2013). Soil degradation, deforestation, or groundwater depletion, are examples of, consequences of productive activities that undermine the importance of the ecological conditions. This proposition argues that a stressed or degraded environment potentiates the effects of perturbations (e.g. slope failures, lower productivity that pressures the environment), unleashing catastrophic consequences.

Butzer (2012) supports the physical constraints on people adaptation to new circumstances proposition and highlights how arid lands in the Near East have been more vulnerable than societies in southern Europe. Arid lands tend to turn over to desert after a temporary abandonment making reconstitution very difficult, contrary to non-arid regions. Hence, less social effort and investment is needed for reestablishment. Finally, failure to anticipate or perceive the problem (Johnson 2016) refers to the human capacity to cope with uncertainty whereas another proposed factor, also related to uncertainty, refers to the ecological characteristic 'highly unpredictable' (Ponting 1993, Ehrlich and Ehrlich 2013), as in the case of El Niño events.

The Piura Basin has attracted multiple settlements and supported societies dating from 9000 B.C. (Huertas 1996). The Moche society is one of the most well-known civilizations that settled in the Piura Basin for approximately 700 years, (100–300 AD to 500–800 AD). The Moche were agriculturally based, with a well-developed and large network of irrigation canals and reservoirs (around 816 km) to divert and store river water to supply their crops in desert areas. Archeologists (Larco et al 1945 , Butters and Castillo 2008) suggest that the Moche society was much wealthier than other societies of the same period because of their irrigation capacity. Their main agricultural products were corn, peanuts, cotton, fruits, and, in the highland areas, different types of potatoes (Velásquez 2015), which they traded with other societies (Butters and Castillo 2008), especially in the highlands to the east.

The society collapsed after being affected by El Niño, but that climatic disaster was not the sole cause. The Moche would have experienced hundreds of El Niño events over their history. Thus, other factors must have been involved that led to the end of Moche civilization (Diamond 2005, Butters and Castillo 2008). One main hypothesis about the factors that could have contributed to the collapse of the Moche, relates to the centralized and very hierarchical political system, with a caste of religious and military leaders dominating farmers (Bawden 1995). Lucero (2002) analyzed the role of water control of the Maya civilization, a well-studied example that is similar to the Moche in this sense. She proposed that societies that have a centralized governance structure over a main resource for the community, as is the case for many irrigation systems, tend to become trapped in a downward spiral of social crisis when rulers lose control of the main resource as a consequence of climatic changes. Lucero (2002) explains that the crisis starts with the collapse of power of rulers as they lose credibility and their capacity to collect tribute, which at the same time potentiates the disruption of the hard (e.g. the reservoir) and soft (e.g. rule enforcement) public infrastructures, which ends up decreasing the population's wealth. The decrease in wealth causes internal conflicts, population migration, or population loss due to decreasing health.

As opposed to the Moche, the current society settled in the Piura Basin has a more decentralized and less hierarchical governance structure for water management, since farmers are entitled to participate in local water governance. To be eligible for irrigation, farmers must be members of the non-governmental and non-profit National Irrigation Association, Junta Nacional de Usuarios de los Distritos de Riego del Perú (JNUDRP), which is subdivided by valleys, and to be registered in the Local Water Authority of the Region (ALA). For the Bajo y Medio Piura, there are three irrigation associations: Bajo y Medio Piura, Sechura, and Huancabamba. Farmers elect association leaders, and although the participation in elections is low (less than 50% of attendance), farmers feel that they are well represented. This may be a result of a well-articulated network of sub-associations.

The three main associations are divided into users' commissions, which are subdivided into canal committees, which at the same time have 10 delegates that are elected by and represent 200 farmers. The main role of the association is to operate and maintain the minor public irrigation infrastructure (secondary canals and drainage systems), distribute water, collect and manage fees for water use, determine water tariffs, cut water services to non-compliant farmers, represent water users in meetings with other associations and governmental authorities (e.g. National Water National Authority, Regional Government of Piura, Agriculture and Environmental Ministries, and major infrastructure operators called 'Proyecto Especial Chira Piura'), and to generate activities for the economic, social, and institutional development of agriculture in the area (Gallo and Oft 2011).

This multi-level organization that involves 75 176 members, is time efficient, enables fluent coordination, effective monitoring, and increases efficiency in communicating concerns or claims and even in solving conflicts in different instances. Farmers seem to understand the importance of improving water management and the role of the water fee in its success. The water associations still have to reduce the default rate of water-fee payments, and to agree on a higher price that reflects the real cost (public and social) of water. However, they have made an improvement from 2014 to 2015: fees collected increased from S/4 million (around USD 1.2 million) to S/6 million (around USD 1.8 millions). Different actors that have responsibilities as managers (e.g. National Local Water Authority, Mayor Operation Managers), revealed their satisfaction with the progress that the Bajo y Medio Piura Water Users Association has achieved. They have improved by themselves the minor canals that they are in charge of maintaining and have collected their own funds for emergency events. This growing social capital, enhance the resilience of the system (table 1, P2), especially because infrastructure for irrigation (reservoirs, canals and drainage) is crucial for water collection during flood events.

However, farmers' technical knowledge in the basin is limited, and farmers' agricultural practices and water management capabilities are still developing (CEPLAR 2016) which makes their individual economic development slow. Poor farming practices (e.g. excessive watering, pesticide and fertilizer use, no crop rotation) degrade the soil and, at the same time, lock farmers into low economic returns on land and labor. It is more difficult for farmers with low profits to finance the necessary irrigation infrastructure maintenance and to be better prepared for disastrous events (table 1, P7).

Bajo y Medio Piura shares the irrigation infrastructure of the Poecho Project, including the reservoir (see figure 1), with the Chira Basin. This interdependence makes the transaction cost of coordinating to collect enough resources for infrastructure maintenance even higher, especially because the Chira Basin has less-developed collective action than Bajo y Medio Piura (table 1, P12). The Bajo y Medio sub basin's river flow, infiltration, groundwater recharge, sedimentation, and contamination also depend on farmers and other actors in the upper basin (Ostovar 2019). The 1997–98 El Niño, and the 2017 Coastal Niño evidenced how the high deforestation levels (around 80% of the forest in the basin with a majority in the upper basin) reduced the natural rain catchment capacity and facilitated extensive rains to run strong downhill and washed loose sediment downstream into the irrigation infrastructure (Ostovar 2019, Cultivalu 2019).

Because of sediment accumulation, the Poecho Reservoir currently has only 50% of its initial capacity (885 mmc). Other irrigation infrastructure, as the major canals are under maintained (water distribution loss is around 15%), and the drainage-system maintenance is almost completely ignored, leaving a significant part of the Valley with salinization problems (CEPLAR, Centro Regional de Planeamiento Estratégico 2016). The physical condition described (natural and engineered) reduces the resilience of the sub basin to flood events (table 1, P10). The government has unofficially announced that it will progressively reduce its financial assistance for major infrastructure maintenance. User associations will therefore need to grow even stronger to ensure public provisioning, in this case through water tariffs. Moreover, self-financed systems are becoming more and more necessary because of the reduction in support from international aid sources. Since Peru has improved its human development in literacy and GDP per capita, it is now a country that is less prioritized for international aid (GRP 2016).

Similar to the Moche's centralized governance system, the current Piura society's disaster prevention policies are highly centralized and main decisions are taken in the national capital, with few local input and low flexibility to adapt to local conditions (French and Mechler 2017). The national government allocates a special budget for disaster prevention to the regional government, however interviewees mentioned that the regional and local governments are more concerned about re-election and public visibility than in the effects of El Niño (table 1, P3). During the 2017 Coastal Niño, the regional government was severely critiqued for spending just 3% of the 2016 budget for prevention programs (French and Mechler 2017). Centralized disasters prevention policies, weak regional governance, and simultaneous floods in different regions of the country combined, limits emergency response (table 1, P8 and P12).

The 2017 Coastal El Niño caused economic losses of around USD 3.1 billion (Leon and Kraul 2017), and affected 1.5 million people (French and Mechler 2017) of which the majority was from Piura (32%), followed by Lambayeque (16%), La Libertad (5%), and Lima (5%), where the capital is located (PAHO 2017). Even though the damage is comparable to the 1982–83 and 1997–98 events (French and Mechler 2017), the rain and flood were not as strong (SENAMHI 2019), revealing a system's resilience decline. Ramírez and Briones (2017) argues that the 2017 Coastal El Niño had disastrous effects partially because it developed rapidly (table 1, P1), there were contradictory forecasts between the national (Peru) and international (US) monitoring agencies that delayed governmental response, and because of an underestimation of impacts, given that some regions (including Piura) were experiencing drought at the time, and because the effects of the previous ENSO (2015–16) were far less than expected (table 1, P4). However, once the danger was identified, the communication channels for evacuation failed, and the majority was unaware of which were the safe areas, or preferred to stay in their residence for fear of losing their belonging to strangers (French and Mechler 2017). The 2017 Coastal Niño also uncovered the lack of proper rainwater drainage systems in urban and agricultural areas (French and Mechler 2017). The effects revealed a fragile social resilience that in a context of growing population (10% average from 2007 to 2017, INEI 2017), can be catastrophic.

Piura is the second most-populated region of the country⁶ (INEI 2017 and has a population density of 52.1 habitants km² (INEI 2017)). The population dynamic in Piura dates back to the 1970s, when to increase the region's capacity to cope with drought seasons, the government built the biggest reservoir in Latin America (Poechos Reservoir). The Peruvian government implemented The Agrarian Reform, that aimed to return the agricultural land owned by hacendados to the people who had worked it for decades and who, during the reform, were organized in cooperatives. To strengthen the reform process, the first water law⁷ was formulated, initiating a series of water policies that, in one way or another, place agricultural activity at the center of the articulation of water law. It was in this context that the Poecho Reservoir and its related infrastructure were built. With the execution of the project, water from the Chira River was diverted to the reservoir, and then released into the Piura River, favoring farmers of the Bajo y Medio Piura.

Thanks to this project, water availability became stable, and agricultural activity started to grow. In only 10 years, agricultural land area grew by 10% (from 84 000 ha in 1976 to 93 000 ha in 1986), and population grew as well (3% per year in the same period). Population density in Piura is problematic for two reasons (table 1, P6): (i) Piura hosts 6% of total population, and the total amount of water available in this region is less than 1% of the total available in the country (CEPLAR 2016), and (ii) Population that migrates from different regions and low-income families settle down in flood high-risk areas due to limited territory available (table 1, 11). Never have so many people lived in the Piura river flood zones exposed to raging torrents during El Niño events.

The Piura Basin presents attractive features for agriculture and trade: good weather conditions with different ecological zones that allow for a diversity of crops, forest on the highlands of the basin, sea life on the ocean, significant rivers running from the Andes to the Pacific Ocean, and a central geographic location that is excellent for trade in the region. However, at the same time, the basin presents a fragile ecosystem with major challenges for human settlement. The Moche society was well known for its engineered irrigation system, but was still unable to cope with the severe droughts and floods from El Niño events. In the end, the remaining population decided to migrate to the high areas of the basin. After many decades in the 1970s, policymakers unaware of the dangers of El Niño, attracted and incentivized population growth in a region highly exposed to strong floods, by building hard public infrastructure (e.g. reservoir, canals, roads, bridges) and soft public infrastructure (e.g. water law, agrarian reform). According to Tainter's (1988) theory of collapse, this complexity puts at risk the resilience of the system, and it is now more difficult to mobilize a big population to relocate to another region less exposed to environmental perturbations (table 1, P5 and P9). Moreover, as suggested by Holling and Gunderson (2002), even though the built infrastructure aimed to stabilize food production by reducing water flow variability, it also changed the stability of the landscape, thereby creating more rigid and myopic management, and thus more sensitive to disturbances.

It seems that everyone in Piura is well aware of the fragility of the basin given the impacts of past Niño events. However, many recommendations made after previous El Niño disasters are yet to be implemented. Most of the time the policy focus is on engineered infrastructure and less on improvement of governance infrastructure and social cohesion. Information about engineered infrastructure is available and well understood, but the source of funding for hard infrastructure maintenance is barely considered. From the analysis summarized in table 1, we can see that the most resilient aspect of the system is the way both individual, and networks of farmers' associations are developed. It is, however, still far from being ideal, and the analysis shows that it can make a significant difference if the associations get support to build their capacity and to become more knowledgeable about how authorities can support them, how to get out of poverty traps, how to better manage their resources, and how to prepare better for future threatening events. Since poverty and developing agricultural practices are aspects that are identified as root causes that show fragilities in the system, by underpinning this strength there might be a positive effect on other aspects of the system.

With the guidance of the Robustness Framework and by leveraging propositions from past research on collapse, it was possible to assess the resilience of the Medio y Bajo Piura Sub-basin to flood events. It seems that, in general, the SES of the Bajo y Medio Piura sub-basin is quite fragile to future drought and flooding events. The Peruvian State or International Aid agencies modestly invest in El Niño years in emergency programs (e.g. canals and rivers reinforcement before the event, sensitivity programs for water use, housing, and main infrastructure restoration after the event). Our analysis suggests that without this support, we would likely see the regional Piura system eventually collapse again, repeating the Moche experience. However, the analysis has demonstrated a significant strength in the system: considerable capacity for collective action. This capacity is built on social capital that provides a foundation for generalized capacity that may be mobilized to prevent damage from future Niños and to develop sustainably.

As shown in this study, public infrastructure is an essential feature of functioning societies within SES, and for SES resilience. However, too much attention has been paid to physical infrastructure, with the result that opportunities to strengthen SES have been overlooked. If we pay more attention to soft public infrastructure, we may find some more effective potential solutions for increasing the resilience of the system to floods. In addition to strengthening water-users association as discussed earlier, policies that pay more attention to identifying and reducing the socio-economic drivers of fragility that affect the population and infrastructure should be implemented.

This is the story of the Piura basin SES. There are many other similar cases in Peru, and around the world, that will be exposed to future climate change events. Because this was only one case study, it is not possible to draw any general theoretical conclusions. However, this research provides methodological and theoretical insights that can contribute to theory building for SES resilience, which is an urgent endeavor. By reviewing collapse theories with the assistance of an interdisciplinary framework, we were able to complement and integrate propositions from isolated disciplines or perspectives. Contrary to proposing deterministic causes of collapse (e.g. environmental, war, epidemics), we exemplify how the inclusion of a combination of diverse factors (biophysical, institutional and socio-economical) into the analysis provides a more comprehensive explanation of phenomena. Future research can use the same methodological approach to analyze more cases and refine the theory.

We thank the two anonymous reviewers whose comments and suggestions helped improve and clarify this manuscript. We are grateful to Marco A Janssen, Joshua Abbot, and Jordane Boudesseul for feedback and inspiration. We also acknowledge financial support from the National Science Foundation (Grant GEO-1115054).

The data that support the findings of this study are available from the corresponding author (CR) upon reasonable request.