Moving beyond the panarchy heuristic

Abstract

Panarchy is a heuristic of complex system change rooted in resilience science. The concept has been rapidly assimilated across scientific disciplines due to its potential to envision and address sustainability challenges, such as climate change and regime shifts, that pose significant challenges for humans in the Anthropocene. However, panarchy has been studied almost exclusively via qualitative research. Quantitative approaches are scarce and preliminary but have revealed novel insights that allow for a more nuanced understanding of the heuristic and resilience science more generally. In this roadmap we discuss the potential for future quantitative approaches to panarchy. Transdisciplinary development of quantitative approaches, combined with advances in data accrual, curation and machine learning, may build on current tools. Combined with qualitative research and traditional approaches used in ecology, quantification of panarchy may allow for broad inference of change in complex systems of people and nature and provide critical information for management of social-ecological systems.

1. Simplifying Complex Systems of People and Nature

Systems of people and nature self-organize at multiple scales leading to uncertain dynamics and difficulties in prediction. This uncertainty manifests in accelerating change of social-ecological systems in the Anthropocene and has created an urgency for understanding and managing resilience of complex systems of people and nature. Panarchy, a heuristic or model of nature (Gunderson and Holling, 2002), envisions system change within and across multiple spatiotemporal scales. It provides opportunities to confront and navigate the intensifying sustainability challenges for humanity, such as the increasing risk of regime shifts from local to global scales (Hughes et al., 2013).

The panarchy concept has been rapidly adopted across scientific disciplines, especially in social-ecological research. Panarchy conceptualizes three core aspects of complex systems (Gunderson and Holling, 2002) (Fig. 1): First, systems of people and nature, including ecosystems, are hierarchically organized with multiple scales, processes and structures. These scales compartmentalize structures and processes from narrow spatial extents with fast dynamics to broad spatial expanses with slow change. Second, the dynamic change associated with these structures and processes is controlled by specific sets of key system intrinsic (nutrients, biological interactions) and extrinsic (seasonal cycles, weather phenomena) drivers. These dynamics are portrayed as adaptive cycles which describe recurring periods of system growth, development and succession, decay and collapse, and reorganization and renewal (Fig. 1). Third, all scales in a panarchy are connected; information, including energy and matter all flow across scales. This information flow is reciprocal and can occur from the bottom up (“revolt” feature in the panarchy nomenclature; Fig. 1), and the top down (“remember” feature).

Figure 1:

A roadmap towards moving beyond the panarchy model. The schematic shows how the heuristic can be extended through the development of a quantitative framework, based on current and novel modeling approaches and advances in data collection and processing. Implications for management and resilience science are indicated. The conceptualization is not exhaustive and meant as a basis to inspire future quantitative research.

Together these features (hierarchical organization, dynamic change or adaptive cycling, information flow) comprise the concept of panarchy. A comprehensive overview of the current-state-of-the-art of panarchy can be found in Gunderson et al. (2022). The following example demonstrates the concept.

Lake food webs are comprised of different trophic levels where ecological communities operate dynamically at different spatiotemporal scales. Phytoplankton forms the lowest panarchy scale in the pelagic environment. Phytoplankton dwells in small volumes of water and shows adaptive cycling in the form of fast generational turnover rates. Predatory fish comprise a higher panarchy scale. These fish can have interannual life cycles and exploit more extensive areas of a lake. Information flow occurs in the form of fish predation regulating phytoplankton biomass through trophic cascades (top-down effect) in the food web, and phytoplankton primary production affecting secondary producers through matter and energy flow from the bottom up.

2. Current and future challenges

Panarchy research has almost exclusively been qualitative (Allen et al., 2014). Qualitative research is valuable because it allows accommodation of attitudes, beliefs, and values in analyses which corresponds to a deeper space of processes, phenomena and relationships that cannot be reduced to the operationalization of variables (Maxwell, 2013). However, such analyses have focused on the social side of social-ecological systems, and an outstanding question is whether or not, and to what extent, ecological processes and dynamics fit the core tenets of panarchy.

We provide a roadmap that discusses an extension of current, preliminary approaches to quantify panarchy and critically test the concept (Fig. 1). We acknowledge that the development of a roadmap for quantifying panarchy may take potentially many different routes with different methods likely being developed in parallel as statistical tools and computation become more sophisticated and powerful. The potentially broad spectrum of possibilities for development impedes ascertaining the direction of future expansion of quantitative panarchy. Our roadmap is therefore meant as a teaser to stimulate future research.

Quantifying panarchy holistically ultimately requires spatiotemporal methods that integrally assess cross-scale structures, identify and assess key ecological drivers at each scale, assess information flow and feedbacks, and identify and quantify within scale adaptive cycles. The development of methods that account for these features is likely difficult given data needs and the complexity that the concept embraces. There are currently no methods available that allow for such an integral assessment of panarchy. Understanding of panarchy is therefore based to a great extent on pattern detection rather than combined pattern-process assessment. Current approaches to quantify panarchy are therefore primarily reductionist because aspects of panarchy are assessed individually (Tab. 1). For example, early warning indicators, Fisher information, analysis of variance and an R-modeling (QtAC) framework have been used to quantify the adaptive cycle and aspects of system control that mediate phases of adaptive cycles. Scale-explicit resilience-based methods such as discontinuity analysis, spatial and time-series modeling and fractal analysis have been used to detect scaling structure and analyze dynamic system change both in space and time. Approaches used to infer the flow of information are diverse but most assess flow implicitly, based on pattern detection through, for instance, correlation analysis or the assessment of cross-scale connectivity. Other, more process-oriented methods that assess fluxes (e.g., stochiometric analyses) and dynamic information flow (e.g., multi-state modeling of animal migrations) have so far not been framed around the concept but are potentially suitable.

Table 1:

Representative overview of current quantitative analyses of the core aspects of panarchy and the goal and context in which they have been applied. Examples are supported with references.

Panarchy aspect	Methods	Goal	Application	References
Adaptive cycle (single scale)
	Analysis of variance	Detecting adaptive cycle phases	Assessing shifting systems controls	Angeler et al., 2015a
	QtAC framework	Quantifying the adaptive cycle	Assessing system potential, connectedness and resilience	Schrenk et al., 2022
	Early warning signals	Detecting erosion of resilience	Regime shift detection	Dakos et al., 2015
Scaling patterns (multiple scales)
	Discontinuity analysis	Snapshot or time/space-implicit detection of scale-variant system structure	Regime shift detection Assessing dynamic system structure Spatial regime detection and regime migration	Spanbauer et al., 2016 Garmestani et al., 2009 Roberts et al., 2019
	Time series and spatial analysis based on canonical ordination	Detection of time-explicit dynamic scaling structure and spatial scaling patterns	Regime shift detection Ecosystem vulnerability assessment Detecting “parallel dimensions” in time and space	Spanbauer et al., 2014 Angeler et al., 2015b Angeler and Hur, 2023 Angeler et al., 2015c
	Fisher information	Detection of spatial/temporal transitions	Regime shift detection	Sundstrom et al., 2017 Eason et al., 2016
	Wombling	Spatial scale transitions	Regime shift detection	Roberts et al., 2022
	Fractal dimension analysis	Detection of scale and scale invariance	Detecting shifts of system control and feedbacks	Gunderson, 2008
Information flow
	Correlation analysis	Implicit, patterns-based inference of information flow	Ecosystem restoration and mitigation	Angeler and Hur, 2023
	Network analyses	Assessing cross-scale contributions of individual habitats to overall system connectivity	Landscape planning, management and measurement	Cumming et al., 2022
	Network causal loop diagrams	Information flow as “domino effects”	Cascading regime shifts	Rocha et al., 2019
	Multi-model analysis	Information flow inferred based on connectivity (static)	Teleconnection of spatial regimes	Heino et al., 2020
	Stoichiometric analyses	Assessing fluxes and metabolism	Matter flow in food webs	Welti et al., 2017
	Multi-state modeling	Dynamic information flow	Organism dispersal	Calvert et al., 2009

Despite the preliminary character of quantitative analyses of panarchy, this work has provided useful information for detecting system organization and regime shifts and for assessing the vulnerability and resilience of ecosystems and their management (Tab. 1). In addition, novel insight has emerged that extends the panarchy heuristic and understanding of resilience more generally. This preliminary insight points to an enormous potential for quantitative panarchy research to uncover aspects of resilience that are relevant for basic and applied research. We discuss two aspects that “seed” our roadmap and suggest that further development and application of quantitative approaches to panarchy is warranted.

A first novel aspect of quantifying panarchy was revealed by time-series and spatial modeling that detected the occurrence of orthogonal (statistically independent) patterns in addition to hierarchical organization (Angeler and Hur, 2023). Such patterns are known to comprise differentiated temporal variability within defined cycles of periodicity in ecological dynamics. This orthogonality can be interpreted as “parallel dimensions” existing in ecological data which modeling is able to extract. Parallel dimensions are not trivial and are important for assessing resilience, which is usually carried out by evaluating the distribution of abundances, and the redundance and diversity of ecological functional traits across hierarchical scales (i.e. a “vertical approach”). Assessing such distributions “horizontally” across parallel dimensions would provide a novel and more nuanced way for representing complex ecological dynamics and resilience. Extending the current panarchy heuristic to envision both hierarchical and orthogonal dimensions suggests a path forward for future research (Fig. 1).

A second novel aspect emerging from modeling is that panarchy structure is often not detected; that is, models are often not significant indicating that stochastic processes potentially dominate over determinism that would give rise to the multiscale deterministic dynamics envisioned in panarchy. Not detecting panarchy structure in ecological models, however, does not mean that the heuristic is unsuitable or even, as a theory, does not stand up to the test of quantitative rigor. Ecological models perform as a function of the resolution and extent of spatial data and the length of time-series data. Insignificant models may simply indicate that data might not capture the relevant spatiotemporal scales necessary for detecting system structure. Alternatively, insignificant models may indicate periods of analysis where predominantly stochastic dynamics may be indicative of systems being exposed to disturbances or their sequels. Such periods may cover chaotic transition phases during system collapse, or when a system operates in the re-organization phase of the adaptive cycle after a system has been perturbed or collapsed. Future quantitative research on panarchy is warranted to scrutinize these speculations.

Moving beyond the panarchy heuristic through quantitative analyses not only requires the development of integral modeling tools but also the need to address current data limitations. While ecological long-term monitoring programs increasingly generate data that will be useful for testing panarchy in ecosystems, a combined social-ecological analysis to ascertain how ecosystems influence and are influenced by social factors in a panarchy context is more challenging and an area ripe for future research. This challenge is largely due to the limitations with primary social data; that is data which are aligned with the ecological data for specific study designs. Often when secondary data (social data obtained for other research) are available, they can be spatiotemporally mismatched with ecological data.

Data collection is generally limited by resource availability and therefore requires innovative ways of using secondary data collected for other purposes. There have been recent algorithmic advancements in data engineering and prediction which together with innovative computational infrastructures, have created promising opportunities for novel approaches to capture published data, and impute missing data or using surrogates obtained for different purposes. Such machine learning approaches provide, for example, Generative Adversarial Networks to create artificial datasets, Active Learning methodology for reducing the human workload to label raw data, and different data augmentation techniques to use available but scarce data to generate new data sets for training machine learning models (Sundstrom et al., 2023). The usefulness of artificial intelligence and machine learning technologies has already demonstrated enormous potential for predicting regime shifts in natural systems. For instance, Bury et al. (2021) showed that models producing early warning signals of tipping points were more sensitive than traditional methods. They also showed that their models predicted some qualitative characteristics of the new system that was not possible with generic early warning indicators.

The further development of quantitative approaches combined with improved data accrual and curation will not only provide new opportunities for carrying out basic research on panarchy but also to inform critical ecosystem management needs such as the provision of ecosystem services (Winkler et al., 2022). We suggest that management can capitalize on panarchy through planning and leveraging management at different phases of the adaptive cycle across scales. Specifically, triggering the collapse of an ecosystem regime with limited ecosystem service provisioning is often desirable so that a new regime with improved ecosystem service delivery can hopefully be created. After inducing collapse, management can attempt to guide the system towards a more desirable system state that then can be fostered. Aquatic invasive species management with the invasive rainbow smelt Osmerus mordax exemplifies ecosystem management from a panarchy perspective (Mrnak et al., 2022).

The relevance of quantitative panarchy for management manifests in its potential to determine management efficiency and identify scenarios when current management schemes have become obsolete under fast changing social-ecological conditions. Ongoing climate change demonstrates shifting social-ecological baselines which increasingly challenges ecosystem conservation and the maintenance of current levels of ecosystem service provisioning (e.g., management of coral reef social-ecological systems). Under non-stationary ecological change (i.e., both system dynamics and system bounds are changing), intensifying management efforts are likely needed to force ecosystems towards desirable present-day levels of ecosystem service provisioning in the future.

Ideally, within-scale and cross-scale interventions should act synergistically to optimize management while minimizing costs. However, such conditions are hardly met in nature. There are many social-ecological trade-offs operating at multiple scales of space and time that ultimately affect management efficiency. However, the complexity inherent in systems of people and nature has made the evaluation of trade-offs difficult. Panarchy, which accounts for this complexity, has potential to overcome such limitations. The concept considers that not all scales in a system are manageable and that information flow within and across scales is a matter of degree and efficacy. Furthermore, this efficiency is mediated by the redundancy of management measures taken within and across scales in the panarchy.

The current climate provides an example of the inability to directly manage greenhouse gas concentrations in the atmosphere (i.e. at the highest scale in a climate panarchy). Panarchy envisions the delivery of Earth stewardship (green energy, circular economies, life style and transportation changes, reforestation) at lower scales (individuals) and intermediate scales (societies) to curb greenhouse gas emissions and ultimately stabilize and reduce concentrations in the atmosphere. At the individual and society scales, Earth stewardship is designed to deliberately induce an adaptive cycle collapse phase; that is, Earth stewardship targets the transformation of unsustainable to more sustainable human exploitation of natural resources. Panarchy considers that transformation at the individual and societal scales generates information that percolates up to the global climate scale to foster the current Holocene climate regime and stave of collapse and shift into a Hothouse earth regime at the planetary scale.

This example shows that management based on panarchy can be made efficient by targeting system conservation and collapse independently at different spatiotemporal scales. However, despite increasing awareness and efforts, Earth stewardship has been insufficient, leaving global warming at the planetary scale unabated. Deeply entrenched fossil fuel economies, flawed social-ecological governance, and other factors are major obstacles that offset the flow of information obtained from redundant Earth stewardship measures from the individual to the global scale in the climate panarchy.

Lake restoration research provides another example of the need to implement frequent and costly redundant biological (piscivorous fish stocking, fishing of planktivorous and benthic fish, submerged vegetation replanting) and technological (nutrient precipitation, sediment dredging, water column aeration) management interventions within and across scales to abate the negative effects of eutrophication in lakes. Panarchy suggests that multiple interventions in different phases of the adaptive cycle and information flow across scales in some lakes may not be enough to make a desirable ecological status self-organizing (resilient). For example, management often targets control of frequent toxic algal blooms at the lowest scale in the lake panarchy that is not directly manageable due to the impossibility to fish or filter blooms. Therefore, management aims to induce a collapse phase at the phytoplankton scale through cascading effects from higher trophic levels in the food web. That is, pisicivorous fish are stocked to mitigate depauperate apex predator communities in eutrophic lakes, which corresponds to the management of the reorganization and conservation phase of the adaptive cycle at this scale. This creates a top-down information potential to induce a collapse phase at the scale of planktivorous and benthivorous fish, which is reinforced through active fishing. Ultimately, management of the fish communities is designed to trigger phytoplankton blooms through indirect, cascading effects. However, as is the case with the climate example, lakes frequently remain in a eutrophic state due to other stabilizing feedback effects (e.g., nutrient loading from sediments and non-point runoff from catchments), which suggests that management redundancy and information flow within and across scales in the lake are insufficient.

The climate and lake example make clear that a panarchy must not be regarded as a closed system that exclusively operates according to the intrinsic scale-specific processes and their cross-scale connectivity envisioned by the heuristic. Extrinsic factors, such as catchment-level agricultural processes, influence a lake’s trophic status. From a systemic perspective this suggests that a lake and agricultural/terrestrial system panarchy can entrain each other to form an overarching complex system. Although the increasing complexity of a broader system will challenge the measurement of panarchy even more, quantification will provide a refined picture of the mechanisms that control information flow and feedbacks. The aforementioned parallel dimensions detected by preliminary modeling studies of panarchy highlight one are of research for studying the connectivity of entrained panarchies.

In summary, advances in quantitative panarchy research may assist in identifying management trade-offs and cost-benefit analyses by analyzing management efficiency of redundant management options needed to erode undesirable and foster beneficial ecosystem conditions. Quantitative panarchy approaches based on integral monitoring, modeling and simulations combined with artificial intelligence may provide novel ways for assessing management efficiency in the short and long term (Fig. 1). In turn, this can build the basis for planning scenarios of alternative resource use in the future if current schemes become unsustainable in a fast-changing Anthropocene.

3. Concluding remarks

Panarchy is gaining prominence in resilience and sustainability research to address social-ecological challenges, including ecosystem management and governance, on a rapidly changing planet. We suggest how the currently predominant use of the heuristic in qualitative research can be complemented with quantitative approaches to test the tenets of panarchy objectively. To this end, methodological advances such as the development of integral modeling tools and overcoming data limitations with improved monitoring and machine learning provide a step forward. Using a single method for analysis may be useful to ascertain specific aspects of panarchy; however, applying multiple methods and integral modeling will be needed to infer patterns, processes and feedbacks more broadly. Inference can be strengthened by combining qualitative research with currently available data and future innovative quantitative approaches for resilience-based analyses and methods traditionally used in ecosystem science, such as biodiversity analyses. Quantitative panarchy research needs to be embedded in broader transdisciplinary collaborations across spheres of society to account for environmental factors and social agency across scales (e.g., governance, economic realities) to objectively identify how ecosystems, and systems of people and nature more broadly, influence and are influenced by rapid social and environmental change.