Holling's Landscape Model

§01 Occasion and function

Public systems that have passed a threshold do not recover by removing the original cause. This is the most under-diagnosed property of administrative degradation in the Netherlands. A municipality that gets its budget back after its executing capacity has been hollowed out for years does not get its capacity back. A care chain that has worked with staff shortages for years does not heal through a recruitment campaign. A supervisory system hollowed out by rule fatigue does not recover through a new protocol. The treatment of symptoms repeats itself. The budgeted improvements fail to materialise.

This pattern has been known since 1973 in an unexpected corner: ecology. Holling then showed that ecosystems do not move around one equilibrium, but can tip between multiple stable states, with an asymmetrical return path.¹ Two years later Noy-Meir worked this out for grazing systems: a meadow carries six cows, breaks under seven, and recovers only when grazing pressure is temporarily reduced to one or two cows.² The system has two stable equilibria and an unstable threshold between them, and the transition works only one way of its own accord. What in ecology is now known as resilience theory yields, for the administrative-science eye, a diagnostic instrument no other model in its arsenal offers: the difference between a system that maintains its output and a system that slowly loses the capacity to deliver that output.

This method note introduces the landscape model in its modern form, applicable to public systems. It follows on from an earlier method note about Moore’s Strategic Triangle and about the four triangles arising from Moore and Ofman.³ Where Moore supplies a synchronic analytical compass, at which corner does the tension sit, the landscape model supplies a diachronic reading key, where is the system moving and which thresholds stand in the way. The two models are complementary and belong, in a complete diagnosis, alongside each other.

§02 Origin and formal structure

The landscape model arises from two developments in 1970s ecology. Holling asked why models that represent ecosystems as systems moving towards a unique equilibrium fail when their predictions are tested against large ecosystems under prolonged pressure. His answer was that ecosystems do not move around one equilibrium but occupy a basin of attraction, and are shock-resistant within that basin as long as disturbances remain small relative to its width. Under larger disturbances the system can end up in another basin of attraction, with an entirely different internal dynamic.

Noy-Meir supplied the mathematical formalisation for grazing systems. His model shows that the basins of attraction and the thresholds between them are functions of the external load. Under low load only the high basin exists: lush meadow, productive, resilient. Under rising load, a low basin appears alongside the high one: bare grassland and mud, with an unstable threshold between them. At sufficiently high load the high basin disappears via a so-called saddle-node bifurcation: the stable high equilibrium and the unstable threshold merge and dissolve. Only the low basin remains. The system tips unavoidably towards the low equilibrium.

The modern synthesis has been supplied by Marten Scheffer in Critical Transitions in Nature and Society, which extrapolates the model explicitly to social and administrative systems and supports it empirically with cases from climate, financial markets, ecosystems and societies.⁴ Three core properties mark the model.

Bistability: there are multiple stable equilibria with an unstable threshold between them. The system always sits in one of the stable equilibria and deviates from it only temporarily under disturbance. Threshold: between two stable equilibria lies an unstable equilibrium that can only be crossed under external pressure. The position of that threshold is no fixed quantity but shifts with the external load on the system. Hysteresis: the transition from high to low equilibrium occurs at a different pressure from the transition from low to high. The system remembers its own history. Whoever removes the cause of the collapse does not restore the system, because on the return path the threshold lies far further away than the threshold on the outward path.

§03 The landscape representation

In a two-dimensional representation the model reads best as a landscape with two valleys and a hill between them. The state variable stands on the horizontal axis: vegetation mass, executing capacity, institutional quality. The potential stability stands on the vertical axis. The system behaves as a marble rolling to the nearest valley. Under light disturbance the marble rolls briefly up against a valley wall and slides back. Under heavy disturbance the marble can cross the hill and end up in the other valley.

Three observations are crucial for the administrative-science application.

The landscape itself changes with the external load. As pressure rises, one valley becomes shallower and the hill lower. At a certain moment the shallowing valley merges with the hill and disappears. The marble then falls inescapably into the other valley, even without further disturbance. The system tips under its own weight.

The width and depth of the valley in which the marble sits constitute the resilience of the system. Deep and broad valleys are shock-resistant. Shallow and narrow valleys tip under small disturbances. A system that delivers the output of a deep valley but in fact sits in a shallow one is a disaster waiting for a trigger. Visible output says nothing about resilience. This is the empirical core of what is known in resilience theory as resilience erosion without visible signal: the system works, and simultaneously the system loses the capacity to absorb disturbances.

The return path is structurally unequal to the outward path. Once the marble sits in the low valley, the high valley recovers only when the external load is drastically reduced, far below the threshold at which the original collapse took place. Only then can the marble spontaneously roll back, or only then does the high valley re-emerge. This is hysteresis: the system is caught in its own history. The economic, political and human costs of recovery stand in no proportion to the initial disturbance.

§04 Translation to public systems

The applicability to public organisations is not metaphorical but structural. Three elements make the transfer legitimate.

Public executing organisations are complex adaptive systems in which productive execution rests on a substrate of people, processes, social bonds, institutional knowledge and informal coordination. This substrate has regenerative properties that behave like biomass. It recovers slowly under light load, degrades under high load, and has positive feedbacks: an experienced team attracts new experienced people, and an under-resourced team loses more than it gains.

The load on these systems is largely policy-generable. Decentralisations, reform agendas, availability requirements, coordination obligations and rule-keeping produce a measurable rise in pressure on the substrate. As grazing pressure is a function of the number of grazers, administrative load is a function of the number and nature of regimes that rest on the execution.

The observed empirics fit the model. Public systems show long periods of visibly stable output, followed by apparently sudden collapses that were in retrospect slower than they seemed. The childcare-benefits scandal, the Groningen gas extraction, youth care, Box 3 and the Tax Administration all follow the same temporal pattern: years of invisible erosion of the basin of attraction, followed by a collapse with a trigger that in a healthy configuration would have caused no collapse. Recovery proceeds slowly and at great cost, and usually requires external intervention of an entirely different order from the original disturbance: benefits redress, Groningen damage redress, recovery operations lasting years and consuming multiples of the original policy costs. The recovery-state economy is, in landscape terms, nothing other than hysteresis.⁵

The landscape model thereby provides a reading key for the most dissociated aspect of Dutch public administration: the difference between where the system sits on paper, where it actually sits, and what it would cost to get out again.

§05 Application: the youth-care chain 2015-2025

The Dutch youth-care chain offers an example in which the model becomes visible in its full sharpness. Between 2015 and 2025 a gradual erosion of the basin of attraction unfolds, culminating in the introduction of the Act on Improving the Availability of Youth Care.⁶

In 2015 youth care is decentralised to the municipalities with a budgetary cut of about fifteen per cent. The system presupposes that 342 municipalities each develop the executing capacity to carry the chain. In landscape terms, a high load is imposed on a newly built system in which the high valley has scarcely yet been formed. The basin is wide but shallow.

From around 2019 the budgetary shortfalls become manifest. Open-ended schemes rise, the first structural waiting lists appear, emergency funding is repeatedly needed. The high valley fills up. The hill sinks. The marble still sits safely, but less deep.

Around 2022 the central state, municipalities and providers conclude the Youth Reform Agenda, partly under pressure of cumulative staff shortages and stacked recovery agendas. In landscape terms, the high valley is by then already so shallow that a small additional load suffices to make the threshold merge.

In 2025 the Availability Improvement Act enters into force. The act adds a compulsory regional division, availability requirements and an extra coordination layer to the chain. Administratively presented as the solution to the waiting-list crisis. In landscape terms it is the seventh cow: the load that makes the high valley disappear definitively. The system tips into the low valley of chronic waiting list and crisis care.

The diagnosis according to the landscape model: the act is not the cause of the collapse, but the final push to an already hollowed-out basin of attraction. Removing the act does not restore the chain, because hysteresis means that the high valley returns only at a load far below the original threshold. The chain is caught in its own history. The case-specific point of this example is that the administrative causal attribution (the act solves something, or the act causes something) becomes untenable in both directions. The correct diagnosis is that the act is the tipping moment of a process that began ten years earlier.

§06 Relation to other method instruments

The landscape model does not stand on its own in the Statecraft arsenal. It complements three other core models in a specific way.

Moore’s Strategic Triangle poses three static analytical questions: does the tension sit on public value, on operational capacity, or on political legitimacy? The landscape model adds a temporal dimension and a threshold dimension. A system whose tension sits structurally on operational capacity is, in landscape terms, a system whose high valley is becoming shallower. The Moore diagnosis tells you where the tension sits. The landscape model tells you how far the system still is from the cliff.

The change colours of De Caluwé and Vermaak typify interventions: yellow, blue, red, green, white.⁷ The landscape model adds a condition within which each colour works or does not. When the system still sits comfortably in its high valley, blue interventions (planning, redesign, structural change) work responsibly. When the high valley has almost disappeared, only interventions that lower the load (yellow acts of power to remove regimes, white interventions that create space) can still help. A blue intervention in a tipping system accelerates the collapse. The combination of change colours and landscape model gives the interim manager a matrix on which the colour of intervention and the system condition can be laid over each other.

The Interim Cycle structures time within an assignment: preparation, entry, execution, hand-over, reflection. The landscape model works on a slowness scale that exceeds the interim cycle: it moves across years to decades. An interim assignment can re-centre the marble within a valley, but cannot restore the valley itself when it has disappeared. For the individual interim assignment this means that the first question at entry runs: in which landscape is the marble rolling here, and on that basis what is achievable within the time window of this assignment?

§07 The action perspective

The landscape model implies four rules of action for whoever wishes to maintain or restore public systems.

Read the landscape before designing an intervention. The question is not only where the system sits, in which valley, but also how deep that valley is, how much resilience remains, and how far it is from the top of the hill, how large the margin to tipping. Embedding as primary KPI translates directly in landscape terms: embedding is not maintaining current output but protecting the breadth and depth of the high valley. A trajectory result that maintains output while valley depth declines is no success but deferred failure.

Avoid interventions that further hollow out the high valley under the heading of problem-solving. The Youth-Care Availability Improvement Act is a textbook example. Coordination layers added to help execution often in fact push the valley closer to the hill. The phenomenon has its own logic: political and administrative actors experience the collapse of a valley as a lack of coordination or direction, and respond with more coordination or direction, which raises the load and accelerates the collapse. The 80/20 pattern in which a growing share of ministerial staff time is spent on coordinating execution rather than on execution itself is the structural expression of this.⁸

Recognise hysteresis in the design of recovery interventions. A chain that has tipped past the threshold does not recover by removing the original pressure. The intervention must be heavier than the original disturbance and must reckon with a long period of under-loading before the high valley recovers. This is politically and economically hard to realise because it will cost output during the recovery period, and because the bearer of those costs usually differs from the bearer of the eventual benefits. The recovery path breaks the electoral cycle at both ends.

When hysteresis recovery is politically unattainable, accept that the system sits in its low equilibrium and design around it. Dutch recovery law is in fact a widely branched juridification of precisely this situation: because executing systems sit in their low equilibrium, the judiciary absorbs the damage individually. That is unhappy but coherent. The alternative, pretending the high valley is still there and formulating corresponding policy goals, is the most damaging.

§08 Closing

The landscape model does what no other model in the Statecraft arsenal does. It gives language and mathematics to the difference between a system that maintains its output and a system that slowly loses the capacity to deliver that output. It explains why the stable years are often the most dangerous, for it is in those years that the valley disappears without the visible consequences yet being noticeable. It explains why symptom corrections after the fact rarely work. And it explains why the recovery-state economy in the Netherlands is not an accidental succession of unfortunate files but the predictable consequence of a state apparatus that has, over a long period, exhausted multiple basins of attraction without being willing to read that exhaustion.

For the municipal chief executive, the regional-cooperation programme manager, the senior civil servant and the interim manager the rule is: before designing an intervention, ask where the system stands in its landscape. The visible output says too little. The depth of the valley says everything.

---

Notes

¹ Crawford S. Holling, ‘Resilience and Stability of Ecological Systems’, Annual Review of Ecology and Systematics 4 (1973), 1-23.

² Imanuel Noy-Meir, ‘Stability of Grazing Systems: An Application of Predator-Prey Graphs’, Journal of Ecology 63, no. 2 (1975), 459-481.

³ The Four Triangles of Moore. A diagnostic instrument for the dissociated government and its trajectories, Statecraft Method, May 2026.

⁴ Marten Scheffer, Critical Transitions in Nature and Society (Princeton: Princeton University Press, 2009). For a more accessible overview see also Brian Walker and David Salt, Resilience Thinking: Sustaining Ecosystems and People in a Changing World (Washington: Island Press, 2006).

⁵ For the elaboration of the recovery-state thesis see: Restoration State Netherlands. A diagnosis of the Dutch handling of system damage, Statecraft Position Paper, spring 2026.

⁶ Act on Improving the Availability of Youth Care, Stb. 2025, 283.

⁷ Léon de Caluwé and Hans Vermaak, Leren veranderen. Een handboek voor de veranderkundige (Deventer: Kluwer, third edition 2006).

⁸ The empirical elaboration of the 80/20 pattern in ministerial staffing from 1980 to the present is in preparation as a separate Statecraft publication.

---

Colophon

Statecraft · Method Holling’s Landscape Model House of Viridian OÜ — Tallinn · Lisbon v0.1 · May 2026

Statecraft is an analytical platform on governing, organisational change and interim work in the Dutch public domain. Method notes work out diagnostic instruments used in Statecraft practice.

statecraft.nl · houseofviridian.org