The green movement's incrementalism is not just ecologically inadequate. It is scientifically wrong.

Bogucki argues that small, scattered environmental projects fail on method as well as scale. They treat complex problems as if they were complicated ones.¹ A complicated problem has many parts, but it can be defined, designed for and checked against the design. A hospital, a bridge and a moon rocket are his examples.¹ A complex system such as a climate or an economy behaves differently, he says. Cause and effect show up only in hindsight, and the parts adapt to each intervention.¹ He counts climate destabilisation, biodiversity collapse, soil degradation, ocean acidification and freshwater depletion as complex problems.¹ Such systems have emergent behaviour: it comes from how the parts interact, not from the parts. So, he says, many correct small actions will not add up to the outcome wanted.¹

The feature he dwells on is the threshold. A system can look stable over a wide range of conditions, then shift suddenly and for good once a line is crossed.¹ His example is the Amazon, which he says makes about half of its own rainfall. Clearing forest cuts that rainfall and stresses the forest that remains.¹ The loop ends in savannification, the forest turning to degraded grassland, and past the threshold the fixes that once worked no longer do.¹ Thomas Lovejoy and Carlos Nobre give the same half in a 2018 editorial. They trace it to Eneas Salati's measurements of oxygen isotopes in rainwater across the basin.⁴ They add that the first model of the question put the tipping point at about 40 percent deforestation.⁴

For where to act, he turns to Donella Meadows and her leverage points. These are places in a system where a small shift produces large changes.² He says she published twelve of them in 1999, ranked from weakest to strongest.¹ Her list starts with constants and parameters and ends with the power to transcend paradigms.² Bogucki reads most environmental programmes as sitting at the weak end. Planting trees, cutting emissions and protecting hectares adjust stocks and flows.¹ The strong end, as he puts it, is the information, rules, goals and paradigms that govern a system.¹ He calls a mandatory charge on ecological damage in every transaction high leverage. A voluntary biodiversity credit scheme he calls low.¹ Meadows herself puts taxes and subsidies among the parameters at the bottom of her list, though pollution taxes that internalise costs appear higher, as a way of strengthening negative feedback loops. Incentives she puts under the rules of the system, higher still.² Her book Thinking in Systems covers the same ground at length.

He also draws on Horst Rittel and Melvin Webber's 1973 paper, which separated tame problems from wicked ones.¹ In his summary, a wicked problem has no definitive formulation, no stopping rule and no test of correctness. Every attempted solution has consequences that become part of the problem.¹ The paper itself gives at least ten properties of wicked problems.³ Its tame examples are solving an equation, analysing an unknown compound and a chess mate in five moves.³ Rittel and Webber wrote about public policy, such as where to put a freeway or how to set a tax rate.³ Bogucki applies the term to climate change, biodiversity collapse and the food system.¹

He ends with four properties he says the current approach lacks.¹ The first is to change the architecture of a system, its rules and incentives, not its flows.¹ The second is to run many different experiments at once. System-level feedback then decides which to grow and which to drop. Scattered projects, he says, lack that feedback.¹ The third is to monitor how a system behaves: its feedback loops, its distance from a threshold, its direction. Single numbers such as hectares protected are not enough.¹ The fourth is to aim for resilience, keeping diversity and redundancy, over efficiency.¹ He notes that Jay Forrester's system dynamics dates to the 1950s. The Limits to Growth used it in 1972.¹

Two of those four properties describe what a model of a place can do. Tracking feedback loops and the distance to a threshold, not one indicator at a time, is the work of a running simulation of a watershed or a forest. Comparing many local experiments needs one shared model of the whole system to set their results side by side.