Why ecological science must rediscover intellectual humility, question its own foundations, and learn when to leave nature alone

Part I: The Shattered Egg

1. The Illusion of Understanding

Humanity has never possessed more scientific knowledge about the natural world. We can measure photosynthesis, quantify nutrient cycling, identify genetic differences between populations, monitor forests from space, and build computer models with thousands of interacting biological and environmental variables. We can simulate landscapes over centuries, estimate future species distributions, and design interventions meant to make ecosystems more productive, more resilient, or better adapted to conditions we expect to arrive. Yet an uncomfortable question remains largely unanswered: does our growing ability to measure and manipulate nature mean that we are coming to understand it?

The distinction matters more than it appears. Three gaps separate measurement from understanding:

• Parts and wholes. Understanding the behavior of individual components is not the same as understanding the system to which they belong.
• Mathematics and relevance. Describing relationships mathematically does not establish that we have captured the relationships that matter most.
• Consistency and reality. An internally consistent representation of an ecosystem does not demonstrate that the representation behaves as the living ecosystem would.

Modern ecological science nevertheless operates, more and more, as though these gaps can be closed with more data, more computing power, more sophisticated analysis, and more funding. The underlying assumption is familiar. Identify enough components, measure enough processes, and describe enough interactions, and we will eventually understand how the whole works. Understanding it, we will be able to:

• predict its behavior;
• optimize its performance;
• correct its deficiencies;
• steer its development toward conditions we consider desirable.

That assumption is now embedded in environmental research, in resource management, and increasingly in corporate sustainability.

This essay asks what happens if the problem is not that our models are insufficiently sophisticated, but that our conception of what it means to understand nature is incomplete. It is not an argument against science. It is an argument for a more demanding science: one willing to examine the philosophical assumptions beneath its methods, to acknowledge what lies beyond its reach, and to recognize that the ability to intervene in a living system does not establish the wisdom of doing so. It is also an argument that universities have a responsibility to entertain ideas that challenge the foundations of their most successful research programs. The purpose of science is not to perpetuate a method of inquiry. It is to deepen our understanding of reality, including by discovering that our preferred methods are inadequate.

The argument proceeds in five parts (Table 1).

Table 1. Structure of the argument

2. All the King's Horses: Humpty Dumpty as a Theory of Knowledge

An old English nursery rhyme, in print since the early nineteenth century, says more about ecological science than most methods sections (1):

Humpty Dumpty sat on a wall, Humpty Dumpty had a great fall; All the king's horses and all the king's men Couldn't put Humpty together again.

The rhyme is usually read as a joke about fragility. Read carefully, it is a compact theory of knowledge with three layers, and each layer corresponds to one of the arguments in this essay (Table 2).

Table 2. The three layers of the Humpty Dumpty problem

2.1 The intellectual layer

Consider what an egg is before it falls. It can be broken into pieces, and those pieces can be examined in extraordinary detail: chemical composition, mechanical properties, molecular organization. We could build elaborate models of how the fragments interact under different conditions. But no amount of knowledge about the fragments guarantees that we can reconstruct the whole, because the integrity of the original depended on an organization that does not survive the fall and cannot be restored by placing the parts side by side.

This is what reductionist science does to an ecosystem, not physically but intellectually. A forest is not a collection of trees, soil particles, microorganisms, insects, fungi, nutrients, water, and weather. Its functioning emerges from relationships among these, across scales of space and time, and from the particular history through which it developed. Long before any scientist arrived:

• seeds dispersed and organisms colonized;
• trees competed and cooperated;
• fires and insects reorganized composition;
• fungi and soil organisms built and redistributed fertility;
• evolutionary and ecological processes interacted over generations.

The forest is not the execution of a design. It is a historically contingent, self-organizing living system. To study it, we isolate variables, draw boundaries, impose categories, and build representations suitable for experiment or mathematics. Those steps are necessary for much good science, but they transform the object of inquiry. What we investigate is no longer the living forest. It is a representation of selected aspects of it.

2.2 The physical layer

Industrial activity breaks real eggs. Forests are clearcut, drained, roaded, and converted, and science is then asked how to restore them. The usual answer is to reassemble selected components: replant the commercially important trees, re-establish a target density, re-create a measured structure. But regrowing trees is not the same as reconstructing a forest, any more than gluing shell fragments together restores an egg. What was lost was not a list of parts. It was an organization that took centuries to develop and whose full content we never inventoried.

2.3 The institutional layer

This layer is the most important. The rhyme does not say that Humpty could not be repaired. It says that all the king's horses and all the king's men could not repair him. The king's horses and men are the institutional machinery of universities, governments, industries, and funding agencies. They are well resourced, well organized, and sincere. They keep allocating effort to reassembling the pieces (more data, more modules, more calibration, more restoration prescriptions) without asking three prior questions:

1. Is reconstruction possible at all?
2. Do we understand what constituted the whole?
3. Why did Humpty fall in the first place?

The third question is the one the machinery is least equipped to ask, because the answer usually points back at the activities the machinery exists to sustain.

The three layers are linked. The intellectual habit of decomposing nature makes the physical decomposition of nature seem repairable, and the institutional machinery converts that confidence into budgets, careers, and policy. Reductionist science has been enormously successful at revealing mechanisms, and nothing here dismisses that achievement. The difficulty begins when a useful method becomes an exclusive philosophy of understanding: when we become experts in the intricate relationships among the pieces we chose to study while remaining profoundly uncertain about the whole from which they were separated.

3. When the Model Becomes the Forest

3.1 Why simulation appeals

Computer simulation has become one of the most influential instruments of environmental research, and its appeal is understandable. Many ecological processes unfold over decades or centuries, across landscapes too large for experiment. On a screen, researchers can harvest, replant, burn, warm, and dry a forest, compare management strategies, and test assumptions that could never ethically be tested on real landscapes. Used carefully, models are valuable. They can:

• expose the implications of hypotheses;
• reveal inconsistencies in what we think we know;
• organize scattered knowledge into a coherent structure;
• show which observations or experiments would be most informative.

3.2 Every model is a set of choices

Every ecological model is built from choices. Researchers decide:

• which processes to include, and which to leave out;
• how to represent each process mathematically;
• which relationships to assume;
• what spatial and temporal resolution to use;
• which data to calibrate against;
• which outputs will count as success.

These choices may reflect deep expertise, but they remain decisions about how reality will be represented. Three well-established findings from the philosophy and practice of modeling show why this matters (Table 3).

Table 3. Three limits on what models can establish

3.3 How a modeling program studies itself

Consider how a forest-simulation research program typically grows:

1. Decomposition. The program represents photosynthesis, respiration, nutrient cycling, mortality, disturbance, and climatic response as separate mathematical relationships, each calibrated from its own literature.
2. Credibility. The assembled model reproduces observed growth reasonably well, which earns it credibility and funding.
3. Elaboration. Graduate students refine individual modules: a better soil-carbon routine, a more detailed treatment of light competition, a new disturbance agent. Each refinement adds parameters and interactions.
4. Self-study. The model's behavior becomes harder to interpret, and the next generation of students is trained to diagnose it: why it overshoots here, oscillates there, responds nonlinearly to this input.
5. Authority. Publications accumulate, agencies adopt the model, and its authors become recognized authorities on ecological complexity.

What has been discovered about the forest? Some things, certainly. But a substantial part of the complexity the program spends its energy on is complexity it created: the consequence of decomposing the system into separately represented processes and then reassembling them. Researchers can become experts in the mathematical complexity of their representation without demonstrating a corresponding understanding of the living system. The complexity of a representation is not necessarily the complexity of the reality it represents.

3.4 From exploring possibilities to claiming authority

When such a model is used to explore distant futures, its outputs tend to travel a predictable route:

1. Simulation results become projections.
2. Projections become expectations.
3. Expectations become management recommendations.
4. Interventions in living landscapes are justified by confidence in a future that has never been observed.

At that point, the model has quietly crossed the boundary between exploring possibilities and claiming authority over reality.

A model may be an excellent instrument for investigating a hypothesis without being reliable evidence that the future will resemble its output.

4. The Future Cannot Validate the Present

Ecological prediction faces a difficulty no amount of computing removes: the outcomes that matter most cannot be observed within the time available for research (Table 4).

Table 4. The mismatch between research time and forest time

There is a simple way to state the problem. The only instrument capable of validating a century-long forest projection is a time machine. We do not have one. So we validate models against the past (historical growth, historical disturbance, historical climate), which is precisely the domain in which the conditions we are trying to predict did not occur. This has three consequences:

• A model calibrated on the twentieth century may reproduce the twentieth century well, yet tell us little about new combinations of temperature, precipitation, pests, pathogens, soils, and species that have never coexisted.
• When several models agree about the future, their agreement may reflect shared assumptions, shared data, and shared theory rather than independent confirmation.
• Consensus among representations is not confirmation by reality.

The researchers doing the most careful work in this field say as much. A Quebec government team running one of the most rigorous assisted-migration field trials in eastern North America reported encouraging five-year results, then cautioned that the findings are short-term, that deep snow had sheltered the seedlings from harsh winters and browsing during the trial, and that longer monitoring is needed to confirm the trends (5). Five years is the time machine we can afford. The trees will live for a century or more.

The appropriate response is not to abandon modeling. It is to treat these limitations as central to the research rather than as caveats appended after the conclusions:

• compare independent modeling frameworks, and investigate why they disagree;
• identify where assumptions produce compensating errors, and which findings survive alternative representations;
• give long-term observation and experiment more weight than simulated agreement;
• distinguish, in every communication, between conditional possibilities and demonstrated outcomes.

A future that can be simulated is not necessarily a future that can be predicted. And a future that can be predicted with some confidence does not thereby become a future we have the wisdom or the authority to engineer.

5. Another Way of Knowing Ecosystems

Criticizing reductionism is easy. The harder question is what a different approach to ecological knowledge would look like. A substantial body of thought answers it, largely developed at the margins of mainstream ecology over the past half century.

5.1 The building blocks

Table 5 summarizes the core concepts.

Table 5. Core concepts of a self-organizing systems view of ecosystems

James Kay and his colleagues at the University of Waterloo combined several of these ideas into a description of ecosystems as self-organizing, holarchic, open (SOHO) systems, and argued that managing them sustainably requires learning to work with their self-organizing processes rather than imposing designs on them (6). Silvio Funtowicz and Jerome Ravetz called the domain where facts are uncertain, values are in dispute, stakes are high, and decisions are urgent post-normal science, and argued that it requires a different relationship between experts and the public, not simply more expertise (11).

5.2 A different purpose for inquiry

The most important consequence of this framework is a change in the purpose of inquiry. If the future of a self-organizing system cannot be reliably predicted and engineered, then science should stop treating prediction and control as its main products (Table 6).

Table 6. Two paradigms of ecological knowledge

This is not a plea for more humility within the prevailing paradigm. It is an argument that a different paradigm already exists and deserves to be explored on its own terms.

5.3 A cautionary tale: the domestication of complexity

Complexity thinking has entered forest science, and what happened next is instructive. Follow the sequence:

1. 2009: the critique. Klaus Puettmann, David Coates, and Christian Messier published A Critique of Silviculture, arguing that traditional silviculture had been built to eliminate complexity in order to maximize production, and that forests should instead be managed as complex adaptive systems (14). It was a serious and influential critique.
2. 2019: the method. Its intellectual descendant, the functional complex network approach, treats a landscape as a network of stands and uses tree functional traits and network properties to decide where diversification, conservation, or reconfiguration will most enhance resilience (15).
3. 2023 to 2028: the program. That approach now underpins DIVERSE, a $7.4 million pan-Canadian research partnership led from the Université du Québec en Outaouais and working across 22 managed-forest landscapes (16, 17).
4. The prescription. DIVERSE's species-selection theme states that its aim is to identify which species and populations will be best adapted to future conditions so they can be favored in forest management, and that this may lead to recommending assisted migration: moving populations within current ranges, extending species ranges, or even introducing species not currently present in Canada (16).

The critique of control became a more sophisticated program of control. The vocabulary of complexity (networks, resilience, functional redundancy) was absorbed into the older ambition of designing the forest of the future.

This is not an accusation of bad faith. The people involved are serious scientists responding sincerely to real threats. It is an illustration of how strong the pull of the intervention paradigm is. Even a framework built on the recognition that ecosystems are self-organizing and unpredictable can be converted, within a decade, into a tool for prescribing what they should become. The alternative paradigm has to be defended not only against reductionism but against its own domestication.

Part II: All the King's Men

6. The Questions Science Is Allowed to Ask

Scientific research does not occur outside society:

• universities depend on funding;
• professors build research programs that must be sustained;
• graduate students need supervision and stipends;
• governments commission studies to serve policy objectives;
• industries finance research that supports their activities.

None of this is improper. But it means that the questions that receive funding are not selected independently of institutional and economic interests.

A research program established to maintain commercial timber production under changing conditions begins with an assumption so basic that it rarely appears in the methods section: that continued timber production is an objective the research should help achieve. The scientific problem then becomes how to sustain it. Researchers compare harvesting systems, planting strategies, genetic improvement, species mixtures, and silvicultural treatments; they build models, run experiments, and publish technically competent work. What never enters the analysis is the possibility that one of the following would be the more defensible recommendation:

• substantially reducing timber extraction;
• leaving large landscapes undisturbed;
• declining to introduce new species, intensify management, or manipulate composition.

6.1 A real example: the harvest that could not change

A recent Quebec study shows how this works without anyone intending it. In 2025, researchers used the landscape model LANDIS-II on a forest management unit in the Mauricie region to compare strategies for making the forest more resilient to climate change and to extreme fire, drought, and insect outbreaks (18). Two of the strategies rest on TRIAD zoning, which divides a forest into a strictly protected zone, an extensive zone managed with ecological forestry, and an intensive zone of high-yield plantations whose purpose is to compensate for the timber not harvested elsewhere (19). Table 7 summarizes the design.

Table 7. Scenarios compared in the TRIAD+ simulation study (Mauricie, Quebec)

Source: Hardy et al. (18). The last two rows are added here to show what the design excluded.

The headline finding was that TRIAD+ offered a good compromise: it harvested the same amount of wood as the other scenarios while increasing protected area by roughly 240 percent and functional diversity by about 15 percent (18).

The authors deserve credit for candor on two points:

• they acknowledged that functional planting was unlikely, on its own, to boost resilience much;
• they could not tell whether their weak result reflected reality or the model's inability to represent the fine-scale interactions through which diversity is supposed to confer resilience (18).

That second admission is the Humpty Dumpty problem in a single sentence: the researchers could not distinguish a finding about the forest from a property of their representation of it.

6.2 What the design could and could not answer

Because timber supply was held constant across every scenario, the study's reach was fixed before it began (Table 8).

Table 8. What a fixed-harvest design can and cannot reveal

Two decisions produced this outcome. The model's architecture decided which processes could matter; the research design decided which management options could win. Both were reasonable within the program's objectives. Together, they made the most consequential question unaskable. An analysis can be conducted flawlessly and still be incapable of answering the question that matters most.

Science can answer its questions correctly while asking the wrong questions.

This requires neither dishonesty nor conscious bias. Researchers can sincerely believe they are investigating ecological sustainability while working inside a framework that has already decided which economic activities must continue. Once such assumptions are normalized across a discipline, they stop looking like assumptions. They become the background against which questions are formulated.

7. When Universities Reproduce Their Own Limitations

Universities should be among society's most important institutions for challenging established knowledge. Their purpose cannot be limited to producing technically proficient graduates and publications within accepted frameworks; they must also preserve the freedom to discover that prevailing approaches are inadequate. Yet academic institutions contain powerful mechanisms of continuity:

1. An established professor builds a research program, secures funding, develops analytical infrastructure, and attracts graduate students.
2. The students learn the program's methods, assumptions, vocabulary, and questions.
3. Some become professors and secure funding to extend the work they were trained to do.
4. Over a few academic generations, a way of investigating nature becomes an institutional tradition.

Thomas Kuhn called this normal science: puzzle-solving within a paradigm that is not itself in question (20). Continuity supports cumulative progress, but it also produces lock-in:

• the deeper the investment in a method, the harder it is to question its foundations;
• a group whose expertise, infrastructure, and reputation depend on a particular model finds endless opportunities to improve it, and little incentive to ask whether a different framework would serve better;
• a graduate student who improves an established model has a recognizable path to a degree, publications, and a job, while one who asks whether the model's conception of the ecosystem is adequate has no comparable path.

Nobody needs to forbid the question. The structure of academic careers is enough to discourage it.

This is especially troubling when complexity itself becomes a professional specialty, as described in Section 3.3. New parameters create new uncertainties, new modules require new calibration, new computing power permits more ambitious simulations, and each improvement generates further research. But the expansion of a research enterprise is not evidence that its foundational approach is converging on a better understanding of nature (Table 9).

Table 9. What academic success measures, and what it does not

A university should be able to recognize a fundamental criticism of an established paradigm as a scientific contribution, even when it threatens decades of institutional investment. Otherwise it becomes extraordinarily efficient at reproducing the knowledge it already knows how to produce.

Part III: Where the Error Meets the Forest

8. Nature Does Not Have a Production Target

Industrial forestry describes its goals in language that sounds ecologically neutral: maintaining productivity, ensuring resource availability, improving resilience, sustaining ecosystem services. Each term hides a choice:

• Productivity for whom?
• Resilience of what?
• Services to which beneficiaries?
• Compared with which historical condition, and over what period?

A forest can be economically unproductive and ecologically vigorous. The same changes read very differently depending on the lens applied (Table 10).

Table 10. The same forest changes, seen through two lenses

The difference matters because commercial objectives become embedded in definitions of ecological success. If a forest is expected to deliver a predictable flow of wood, anything that interrupts the flow becomes a problem requiring intervention, and any departure from a preferred composition reads as deterioration. But an ecosystem has no obligation to maintain conditions humans find economically convenient. Nature does not run on annual budgets, allowable annual cuts, or supply contracts.

The result is a self-reinforcing loop:

1. Ecological variability interferes with continuity of supply.
2. More management is applied to restore continuity.
3. Management creates new uncertainties and side effects.
4. Those side effects generate demand for more research and more intervention.
5. The problem of controlling nature becomes the justification for more sophisticated control.

Meanwhile, the possibility that the original demand for continuous production is itself the problem receives comparatively little attention.

Perhaps the question is not how to make forests conform more reliably to our economic expectations, but whether those expectations are compatible with the nature of forests.

9. Harvesting Is Not a Wildfire

The most important conceptual bridge between industrial forestry and ecological legitimacy is the claim that harvesting emulates natural disturbance. In the boreal forest, where large fires historically replaced whole stands, the argument runs that clearcutting is ecologically analogous to fire: both reset succession, both create young forest, both belong to the system's natural rhythm.

Quebec built a version of this idea into law. Its 2010 Sustainable Forest Development Act makes ecosystem-based management (reducing the gap between managed and natural landscapes) a central principle of public forest management (21). The principle is sound. The danger lies in how easily it slides from "harvesting should be designed to resemble natural disturbance" into "harvesting is a natural disturbance."

The scientific literature has been clear for a quarter of a century that the two are not interchangeable. A major review in Environmental Reviews compared wildfire and harvesting across Canadian forests and documented differences in their physical, chemical, and biological effects and in their spatial patterns (22). A later industry-commissioned synthesis reached the same conclusion, finding that in the first years after disturbance (23):

• post-fire stands have more standing dead trees and thinner forest floors than logged sites;
• fire produces a stronger pulse of available nutrients and raises soil pH, while harvesting does not;
• plant, bird, mammal, and invertebrate communities respond differently to the two disturbances.

Table 11 summarizes the principal contrasts.

Table 11. Why harvesting is not ecologically equivalent to wildfire

Sources: McRae et al. (22); NCASI (23).

None of this means harvesting cannot be redesigned to reduce these differences; that is precisely what ecosystem-based management attempts. The point is rhetorical as much as ecological. The analogy lends industrial extraction the moral standing of a natural process, and in doing so it shifts the burden of proof. If harvesting is disturbance, the question becomes how to do it well, not whether to do it at all.

9.1 Land sparing: a political promise in ecological form

The same structure appears in the land-sparing argument behind TRIAD zoning: intensify production on a small fraction of the land so that more can be protected elsewhere (19). The logic is coherent and the motive is often conservationist. But the conservation benefit is not a property of the plantation. It depends on conditions that no ecological model can guarantee:

• the protected zone must actually be designated, and stay designated;
• demand for wood must stay constant, rather than growing into the space the intensive zone created;
• intensive zones must remain where they were drawn;
• future governments must honor the bargain struck by present ones.

Land sparing is a political promise expressed as an ecological model, and the model cannot keep the promise.

10. [TBD]

11. The Forests That Organized Themselves

The history of northern forests offers a larger perspective. Much of the land now covered by boreal and temperate forest lay under continental ice twenty thousand years ago. As the ice retreated:

• plants and animals recolonized the emerging landscape;
• tree species dispersed across immense distances;
• communities assembled, dissolved, and reassembled;
• soils formed;
• relationships developed among vegetation, fungi, insects, animals, and water.

No authority selected the species composition, and no model prescribed the succession. Indigenous peoples shaped many of these landscapes through their own stewardship, including fire, but the continental development of the northern forest was not industrial engineering. It was a demonstration, on a continental scale, of the capacity of living systems to organize themselves.

Honesty requires stating what this history does and does not prove (Table 13).

Table 13. What postglacial history shows, and what it does not

The rates matter. Genetic evidence suggests that several eastern North American trees moved northward after the last glaciation more slowly than once believed, on the order of 100 metres a year or less (34). The average velocity at which temperature zones are now shifting across the land surface has been estimated at roughly 0.4 kilometres a year (35). Landscapes today are also fragmented by farms, roads, and cities that did not exist during postglacial migration. These are the strongest arguments for assisted migration, and they deserve to be taken seriously.

But they establish a risk, not a remedy. That trees may lag behind a shifting climate does not, by itself, establish:

• that humans know which trees to move, where, and in what combination;
• that deliberately assembled communities will function better than self-assembled ones;
• that a tree which survives in a new place will become an ecologically appropriate participant in its community, given relationships with soil organisms, mycorrhizal fungi, insects, and other plants that no model represented.

Researchers in this field have themselves warned that assisted migration operates in conditions of "information underload," where ecological complexity, incomplete knowledge, and data deficits make confident prescription difficult (36). The question is not merely whether humans can help trees move. It is whether we have sufficient evidence to justify deliberately rearranging living ecosystems on the strength of uncertain expectations about the future.

12. Moving Trees for a Climate That Has Not Arrived

Assisted migration is no longer hypothetical in Canada. Two provinces show how it is being implemented.

12.1 British Columbia: assisted migration by regulation

British Columbia was an early adopter (37, 38):

• 2008: the province raised elevation limits for seed transfer to anticipate warming;
• 2012: it launched the Climate-Based Seed Transfer (CBST) project;
• 2017 to 2019: CBST became available, first as an alternative and then as an optional standard;
• 1 August 2022: CBST became mandatory for reforestation on Crown land, and the older geography-based rules were removed.

Under CBST, seedlots are matched not to the climate where they grew but to the projected near-future climate of the planting site, with a migration distance added to account for expected warming (37). This is assisted population migration: moving genetic material within a species' range. It is now applied, by regulation, to every Crown-land planting in the province.

12.2 Quebec: assisted migration by experiment, heading toward guidance

Quebec has proceeded on several fronts:

• Seed-orchard redeployment. The provincial forest research directorate, whose seed orchards supply most of the province's reforestation stock, has tested whether southern white spruce seed sources outperform local ones along a climatic gradient of roughly 5.5 to 6 °C, with a view to redeploying orchard seed northward (39, 40).
• The DREAM network. In 2018, a government team led by Patricia Raymond established in the Portneuf wildlife reserve, north of Quebec City, the first site of DREAM (Desired REgeneration through Assisted Migration), an international network of assisted-migration trials (41).
• The Portneuf planting. There they planted more than 9,000 seedlings of nine species from seed sources matched to current, mid-century, and end-of-century climate analogues. Some seed was collected more than 800 kilometres to the south, in Pennsylvania and Ohio, and three species were new to that bioclimatic domain: northern red oak, black cherry, and shagbark hickory (5).
• DIVERSE. The partnership described in Section 5.3 aims to produce guides recommending functional groups of tree species for planting across Canada's bioclimatic regions, explicitly including the possibility of introducing species not currently present in Canada (16).

Table 14. Assisted migration in practice: British Columbia and Quebec

Sources: (5, 16, 37, 38, 39, 41).

12.3 What the early evidence actually shows

The Portneuf results are, on their face, encouraging, but the details deserve attention (5):

• after five years, 84 percent of seedlings survived, and the authors judged the climate-analogue approach to seed sourcing appropriate;
• seedlings sourced more than 500 kilometres to the south generally performed slightly worse than local or moderately moved sources;
• shagbark hickory suffered repeated damage to its leading shoot and the lowest survival, and the authors suggested it may not be a good candidate;
• data loggers recorded winter minimums of −40 °C and late frosts during leaf flush;
• deep snow protected the seedlings during their first winters, and the authors stressed that the results are short-term and need longer monitoring.

Then there was the event the models did not forecast. In spring 2021, an early warm spell advanced white spruce bud flush by about ten days across Quebec, and a hard frost followed at the end of May. In a provincial assisted-migration trial spanning three sites (42):

• researchers assessed 5,376 trees after the cold spell, and virtually all were damaged;
• leader growth was cancelled for between roughly a fifth and three quarters of trees, depending on the site;
• there was no significant difference between moved and local seed sources;
• the authors concluded that seed-transfer models must account for climate instability, not just warming trends, and warned of large-scale maladaptation.

A separate Quebec study found that local conditions mattered more than seed origin for sugar maple frost hardiness at its northern limit (43).

12.4 The counterfactual nobody plans for

The entire logic of assisted migration depends on a projected future: that a site will, within a few decades, have the climate of a place hundreds of kilometres to the south. The practice rarely confronts the obvious questions:

• What if warming is slower than projected, or uneven?
• What if a warmer average coexists with continued extremes, such as polar-vortex outbreaks, late frosts after early springs, and deep cold snaps?
• What happens to southern populations, which break bud earlier and are adapted to milder winters, during the decades before the projected climate arrives?
• What happens to local populations, associated organisms, and forest development if the introduced trees fail, or succeed too well?

A seedling chosen for 2080 must survive 2030 first. And unlike a model run, a planted landscape cannot be reset.

12.5 Not all assisted migration is the same

The distinction between kinds of assisted migration matters (Table 15).

Table 15. Three intensities of assisted migration

Sources: Raymond et al. (5); DIVERSE (16).

Canadian researchers recognized more than a decade ago that disagreement about assisted migration is ultimately ethical, not only technical: a matter of values about naturalness, risk, and human responsibility (44), and of how much uncertainty is acceptable when the consequences are irreversible (45). The forest will carry the consequences of a mistaken introduction long after the funding cycle that produced it has ended, the researchers have retired, and the projections that justified it have been revised. That asymmetry (reversible assumptions, irreversible interventions) is the strongest reason why the burden of justification must rise with the scale and permanence of what is proposed.

13. Intervention Is Not the Default Moral Position

Contemporary environmental management often treats intervention as the mark of responsibility:

• doing something demonstrates concern;
• developing a solution demonstrates expertise;
• a new technique demonstrates progress;
• choosing not to intervene looks passive, negligent, or unambitious.

This framing confuses activity with responsibility. In a complex living system, restraint can be a carefully reasoned response to uncertainty, and the more extensive, persistent, and irreversible an intervention, the stronger the justification it should require. A short-term experiment can demonstrate feasibility. It cannot establish long-term wisdom.

None of this means non-intervention is always better. Ecosystems damaged by pollution, invasive organisms, or past land use may benefit from carefully designed restoration, and some habitats require active stewardship. Doing nothing has consequences too. But that observation does not make intervention the default. It makes honest comparison necessary. A responsible research program should be able to weigh the full range of options without deciding the answer in advance:

• protection;
• reduced extraction;
• active restoration;
• limited intervention;
• non-intervention.

It should also recognize that the right objective may not be to hold an ecosystem in a preferred condition, because an ecosystem's capacity to reorganize may matter more than its ability to reproduce the conditions humans happen to value.

Scientific humility becomes meaningful only when uncertainty is permitted to change what we decide to do.

14. The Problem With a "Healthy" Forest

Environmental institutions describe their aims as forest health, ecological integrity, resilience, and biodiversity. The terms signal good intentions, but they conceal conceptual difficulties. Health is already a complicated idea in medicine, where the absence of diagnosed disease does not guarantee well-being. A forest is harder still: it is not an organism with a unified physiology but a changing community of organisms with different life histories, competitive relationships, and responses to disturbance (Table 16).

Table 16. Unhealthy, or simply changing?

Degradation can sometimes be identified convincingly: severe pollution, soil destruction, the loss of key populations. The difficulty is not that all judgments of ecological condition are meaningless. It is that "health" cannot be made operational without specifying three things:

• what is being valued;
• what is being measured;
• what is being maintained, and for whom.

A forest can be healthy by the standards of timber production while ecologically simplified, and a naturally disturbed forest can be economically undesirable while sustaining processes essential to long-term development. Institutions that invoke forest health should state whose conception of health they are applying. Otherwise an economic preference is presented as an objective property discovered by measurement.

15. Biodiversity: Measuring What We Barely Know

Biodiversity is a legitimate scientific concept, but its breadth creates an unavoidable measurement problem. One widely cited estimate suggests that around 86 percent of the planet's land species and 91 percent of its marine species remain undescribed (46). Many organisms are microscopic, seasonal, or living in habitats scarcely surveyed, and even among known organisms our grasp of ecological relationships in soil communities, microbial interactions, fungal networks, and parasites is thin. New tools such as environmental DNA expand detection, but detecting more organisms is not the same as understanding biodiversity.

15.1 The limits of the inventory

Every inventory is bounded by its sampling methods, taxonomic resolution, geographic coverage, and timing. Each bound shapes the result:

• a forest surveyed in spring yields a different community than the same forest in autumn;
• a one-hectare plot can contradict a landscape survey;
• species counts within plots can rise while differences among habitats collapse into homogeneity;
• succession can change which organisms are present without any unambiguous basis for calling the change an improvement.

15.2 The limits of the index

A diversity index describes a mathematical property of an observed sample. It cannot capture relationships, evolutionary histories, functional roles, or the organisms no one detected, and different ecological realities can produce identical index values. Maximizing species richness is not even a sound conservation objective, since an influx of common or introduced species can coincide with the loss of distinctive native communities.

15.3 What gets counted

Attention gravitates to mammals, birds, and commercially important trees, while nature is full of organisms that are obscure, unpleasant, economically inconvenient, or unknown:

• native insects labelled as pests, despite their ecological roles;
• fungi that govern decomposition and nutrient cycling unseen;
• parasites with no public constituency at all;
• species no inventory can represent, because science has not found them.

15.4 Claims limited to what was measured

This makes comprehensive claims about biodiversity outcomes extraordinarily difficult to substantiate (Table 17).

Table 17. What biodiversity monitoring can and cannot establish

The unknown does not become known because a reporting framework finds it inconvenient, and the impossibility of complete measurement does not weaken the obligation to protect what we cannot see. It strengthens it.

Our ignorance of the living world is itself a reason for caution in disturbing it.

Part IV: The Same Error, Accounted For

16. When Ecological Science Becomes Environmental Accounting

The difficulty of measuring nature becomes most consequential when ecological concepts are translated into corporate targets. Environmental performance systems need standardized boundaries, indicators, baselines, monitoring procedures, and criteria for judging whether commitments have been met. The reasons are understandable:

• governments want comparable information;
• companies want manageable objectives;
• investors want indicators of risk and performance;
• auditors and validators need criteria they can apply consistently.

International frameworks now push strongly in this direction (Table 18).

Table 18. Frameworks translating nature into corporate metrics

Complex ecological realities are thereby translated into measurable activities and pressures: water withdrawn, hectares converted, practices changed, pollution reduced, restoration financed. Such measures can be meaningful. Reducing a well-established pressure can deliver genuine benefits even when the resulting ecological change cannot be measured directly. The problem begins when the language of performance becomes broader than the evidence (Table 19).

Table 19. Activities and conformity are not outcomes

16.1 A thought experiment: the positive company and the dying watershed

The distinction is easy to state and easy to lose, so consider a company that:

1. has degraded a watershed through decades of withdrawals and discharges;
2. adopts better operating practices on site;
3. finances restoration projects in other watersheds;
4. quantifies those benefits with a recognized accounting methodology;
5. has the calculations independently verified and is found to have met its target;
6. is now permitted to make a positive-performance claim.

Meanwhile, the original watershed continues to deteriorate. What, precisely, has been established? Three different things have been verified, and only one of them is an ecological fact (Table 20).

Table 20. What was verified in the thought experiment

Implementing measures intended to reduce harm does not demonstrate that the harm has ceased. Purchasing benefits elsewhere does not compensate for harm that continues here. A measurement system can be internally rigorous and still be an incomplete representation of the reality it claims to assess, just as a financial ledger can be correct to the cent without describing what the transactions did to anyone.

A measurement system can be internally rigorous while remaining an incomplete representation of the reality it claims to assess.

This is the error of Section 3 in a different uniform: a carefully constructed representation acquires authority beyond what its relationship to reality can justify.

17. The Strange Idea of Becoming "Nature Positive"

The clearest example is the spread of claims that organizations can become water positive, biodiversity positive, or nature positive: that they will leave the natural world better off than they found it. Since 2020, several of the world's largest technology companies have pledged to become water positive by 2030, meaning that they will replenish more water than they consume (51, 52). Most measure their replenishment with VWBA, a method published by the World Resources Institute and partners to quantify the benefits of projects such as wetland restoration, irrigation efficiency, and aquifer recharge (50, 51).

17.1 Water is not fungible

VWBA is a serious method, and replenishment projects can do real good. But look at what the balance compares. A cubic metre of groundwater withdrawn in a dry season from a stressed aquifer is not ecologically equivalent to a cubic metre of streamflow restored in a wet season in another watershed. The two volumes may differ in:

• place: they may belong to different hydrological systems;
• people: they may support different communities;
• ecology: they may sustain different ecosystems and processes;
• timing: their availability may matter at entirely different times of year;
• quality: even within one watershed, the condition of the water matters.

A numerical surplus cannot, by itself, establish that the affected aquifers, rivers, wetlands, or people are better off. The calculation can be correct and the claim of positivity unsubstantiated.

17.2 The boundary problem

Then there is the boundary. Microsoft's commitment, for example, applies to its direct operations (52). That is a legitimate and transparent scope. But a company's total water footprint is much larger (Table 21).

Table 21. Inside and outside a typical corporate water boundary

A 2026 analysis of data-center water use noted that upstream, supply-chain water is rarely included in corporate disclosures for lack of reliable data, which makes headline comparisons misleading (53). A company can be water positive inside the boundary its accounting system draws while much larger pressures lie outside it.

This is exactly the problem of the forest model (Table 22).

Table 22. The same boundary problem in science and in accounting

In both cases, the representation can be internally consistent and externally incomplete, and in both cases the party choosing the boundary has an interest in what lies inside it.

17.3 Who owns the water?

Behind the accounting lies a more basic question: who owns water, and who bears the cost when it is taken? Consider bottled water in Ontario. The case of Nestlé Waters Canada's well at Aberfoyle, near Guelph, illustrates the stakes (54, 55):

• Guelph is the largest Canadian city that relies solely on groundwater, and the water taken at Aberfoyle comes from the same aquifer system that feeds the city's supply;
• in 2016, after a drought forced local residents to restrict their own water use, it emerged that the province charged bottlers $3.71 per million litres;
• while the company's expired permit was under review, it was allowed to keep pumping up to 3.6 million litres a day.

Public water became a private product, sold back to the public in plastic, while the community whose aquifer it was bore the scarcity. No volumetric balance can resolve a question of that kind, because it is not a question of quantity. It is a question of:

• who has the right to take;
• who has the duty to restore;
• who decides.

17.4 Biodiversity is not fungible either

The same structural problem runs through biodiversity accounting:

• a restored hectare does not necessarily replace a destroyed hectare of old forest;
• an increase in selected species does not compensate for losses elsewhere;
• an improved landscape indicator does not reconstruct the relationships an intervention disrupted.

Living ecosystems are not collections of interchangeable units, but accounting systems gravitate toward quantities that can be standardized, transferred, aggregated, and netted. That creates the appearance of equivalence where none has been established. Beneficial projects deserve to be described on their own terms, not treated as evidence that ongoing harm has been neutralized.

Paying for a beneficial environmental activity does not, by itself, establish that an organization's overall relationship with nature has become beneficial.

18. The Cost of Turning Nature Into Indicators

Environmental measurement raises a question that is rarely confronted: who pays? Credible monitoring of biological communities requires specialists, repeated field surveys, laboratory and genetic analysis, databases, and long-term quality control, and the burden grows with scale (Table 23).

Table 23. Measurement difficulty rises with scale

Comprehensive knowledge remains unattainable at any price. Standardized reporting nonetheless imposes real costs on farmers, forest owners, and small producers, who are asked to supply data, adopt prescribed methods, and demonstrate conformity with standards written elsewhere, while large corporations finance the platforms, consultants, and verifiers. A new economy of environmental measurement emerges (Table 24).

Table 24. The environmental measurement economy

Some of this improves transparency. But the growth of the infrastructure is not evidence that ecological understanding is growing at the same rate. Two dynamics stand out:

• institutions can become exceptionally good at demonstrating conformity while remaining unable to establish the outcomes that justified the requirements in the first place;
• the limitations of measurement generate demand for more indicators, more methodologies, more technology, and more verification, which feeds the same economy.

We risk building a sophisticated economy for measuring compliance with environmental frameworks rather than an effective means of understanding and protecting nature. The king's horses and men, again.

19. What Our Methods Cannot See

19.1 The invisible becomes expendable

The limitations of reductionist science have practical and ethical consequences. When institutions rely on selected indicators to represent ecological reality, whatever cannot be readily measured receives less weight in decisions, not because it is unimportant but because the method used to describe the system does not reveal it. A forest's merchantable volume can be estimated precisely, its composition sampled, its canopy mapped from space. But many of its relationships are poorly understood, and their importance may become apparent only after they are disrupted:

• an organism looks insignificant until its interactions fail;
• one soil community looks interchangeable with another until a change in vegetation alters nutrient cycling;
• a regenerated stand meets its stocking standard while differing profoundly from the forest it replaced.

What falls outside the measurement framework becomes invisible to the management framework, and what becomes invisible becomes expendable.

19.2 The limits of method are not the limits of reality

This is one reason other ways of knowing deserve serious attention. Indigenous knowledge systems, built on long, place-based relationships with particular landscapes, emphasize relationships, obligations, histories, and meanings that conventional measurement does not capture, and contain detailed, practically tested understanding of ecological processes (56). Other traditions emerge from philosophical or spiritual understandings of humanity's place in the living world.

The deeper issue is philosophical. Modern science is methodologically naturalistic: it investigates what its instruments can detect and its models can represent. That is a powerful discipline. But it is easily confused with a metaphysical claim (Table 25).

Table 25. A rule of method versus a belief about reality

Two principles follow, and both matter:

• The inability of a particular method to investigate a phenomenon does not establish that the phenomenon does not exist.
• Openness to unfamiliar possibilities does not relieve us of the responsibility to test factual claims carefully.

19.3 An example: trees and the moon

Traditional forestry in many cultures held that the moon influences trees: that timber felled at certain lunar phases behaves differently, and that sap moves with lunar rhythms. Mainstream wood science long treated this as folklore. Then the question was tested (Table 26).

Table 26. Testing a traditional claim: tree stems and lunar tides

The question remains open, and the episode is instructive precisely because it is unresolved. A claim that most scientists would once have dismissed as superstition was tested with ordinary instruments, published in a leading journal, disputed, and re-examined. That is how openness and rigor should coexist: neither reflexive dismissal nor uncritical acceptance, but investigation.

A university should be capable of exploring unfamiliar perspectives in exactly this way. The point is not to abandon standards of evidence. It is to recognize that the questions worth asking are not exhausted by the questions an established methodology is equipped to answer, and that a science which cannot imagine its own categories being insufficient has confused the boundary of the discipline with the boundary of reality.

20. The Political Economy of Ecological Certainty

The recurring problems in environmental science and environmental accounting are connected by a larger institutional pattern (Table 27).

Table 27. What each institution depends on, and what it tends to produce

These institutions do not share identical interests, and each produces real public goods. But their incentives converge on a common preference: solutions that allow established economic activities to continue. None of this requires a conspiracy. It requires only institutions responding to the incentives under which they operate. The trouble is what those incentives filter out (Table 28).

Table 28. Alternatives that are filtered out, and why

These alternatives are not refuted. They are simply never funded, never modeled, never audited. Institutional bias need not be a matter of individuals distorting evidence. It can be embedded in the questions institutions are organized to answer.

Ask, then, the question this essay has been circling: cui bono? Who benefits from the belief that forests can be predicted, optimized, redesigned, and reassembled, and that corporate harm can be netted against purchased benefit? The answer is the same set of institutions that fund the research, write the methodologies, and certify the results. That does not make their work false. It makes their framing of the problem something that must be examined, not assumed.

When the continuation of an economic activity is treated as a fixed requirement, science risks becoming an instrument for adapting nature to the economy rather than helping society adapt its demands to the limits of nature.

Part V: Toward a Humbler Science

21. A Different Mandate for Ecological Science

The argument for ecological humility must lead somewhere. A different conception of ecological science would start from the framework set out in Section 5:

• ecosystems are historically developed, self-organizing, holarchic, open systems whose behavior cannot always be understood through the separate study of selected components;
• models are representations built for purposes, not substitutes for the systems they describe;
• some uncertainty is a permanent limit on what research can justify, not a gap awaiting more funding;
• long-term observation, independent comparison of competing theories, and careful study of relationships deserve more weight than simulated agreement.

21.1 Questions to ask of every proposed intervention

Above all, such a science would examine its own assumptions, asking:

1. Why is this ecosystem considered degraded, and who chose the reference condition?
2. Which organisms and processes were included in the assessment, and which were left out?
3. Why is intervention assumed to be necessary, and what alternatives were excluded from the design?
4. What evidence would persuade us that the proposed intervention should not proceed?
5. Could reduced extraction or non-intervention produce acceptable, or preferable, outcomes?

21.2 What would change in practice

These questions should be among the foundations of ecological science, not its afterthoughts. In practice, that means (Table 29):

Table 29. Practical reforms for a humbler ecological science

22. Restraint Is Not Abandonment

An argument for leaving nature alone is easily caricatured as an argument that humans should never touch it. That is not the claim. Humans have shaped forests for millennia, and many landscapes that look untouched were tended by fire, harvest, and ceremony for thousands of years by Indigenous peoples whose stewardship was often sustained precisely because it was embedded in obligations and long observation rather than extraction targets (56).

The position defended here is restraint, and restraint has four distinct components (Table 30).

Table 30. The four components of ecological restraint

The argument is not for universal passivity. It is for restraint grounded in the recognition that ecosystems do not exist merely to satisfy human economic objectives, and that our capacity to act on them far exceeds our capacity to foresee the consequences.

23. Humility Must Have Consequences

Humility is frequently invoked in scientific writing and rarely reflected in practice:

• a researcher acknowledges uncertainty and still recommends confident interventions with unknown long-term effects;
• a corporation acknowledges the limits of biodiversity measurement and still claims to be restoring nature;
• a government acknowledges the complexity of forests and still adopts strategies that assume their future can be directed.

In each case uncertainty is acknowledged rhetorically without constraining the action it should constrain. Genuine humility requires:

• proportionality between evidence and claims;
• a clear separation of what has been observed, what has been inferred, and what has merely been assumed;
• resistance to converting computational sophistication into predictive authority;
• recognition that scientific expertise confers no exclusive authority over the ethical question of what nature should become;
• willingness to recommend stopping, reducing, protecting, observing, or refraining.

This is not a philosophy of paralysis. Decisions must be made under uncertainty, and refusing an intervention has consequences too. But uncertainty cannot justify assuming that action is always preferable. The burden of justification should scale with the magnitude and permanence of what is proposed (Table 31).

Table 31. The burden of justification should scale with the intervention

Good judgment also requires letting science produce conclusions that are inconvenient for established interests. Each of the following would be a success of science, not a failure:

• a study concluding that timber extraction should be substantially reduced;
• a study finding that a given assisted-migration program lacks adequate justification;
• a study showing that a biodiversity indicator cannot support a particular corporate claim;
• a study demonstrating that an expensive model cannot reliably distinguish between management strategies.

Such findings show science functioning as an independent means of inquiry rather than a continuous supplier of technical justifications for intervention.

If scientific humility never changes the actions researchers recommend, the claims institutions permit, or the economic objectives society is willing to reconsider, it remains little more than a reassuring word.

24. Conclusion: Why Did Humpty Fall?

We began with a nursery rhyme. Humpty Dumpty sat on a wall, had a great fall, and all the king's horses and all the king's men could not put him together again.

Modern ecological science has become extraordinarily skilled at studying the pieces. It can measure them, classify them, model their interactions, and build elaborate systems for evaluating selected properties. These achievements are real. But they do not guarantee that the whole has been understood, and they do not guarantee that it can be rebuilt. When our representations become the basis for interventions, policies, and claims of environmental improvement, the gap between the fragments and the whole stops being a philosophical curiosity and becomes a question of responsibility.

The king's horses and men are still at work:

• refining the models;
• redeploying the seed orchards;
• liming the sugar bushes;
• zoning the landscapes;
• verifying the water balances;
• certifying the targets.

Much of their work is careful and sincere. What they rarely do is the one thing the rhyme quietly invites: stop trying to put the egg back together and ask why it fell. In most of the cases examined here, the answer is not a mystery (Table 32).

Table 32. Why Humpty fell: the causes behind the cases

The institutions best equipped to reassemble the pieces are often the ones least able to name the cause of the fall, because the cause is bound up with their own reason for existing.

Humanity does not need to abandon scientific discovery. It needs to distinguish discovery from technical elaboration, understanding from representation, and stewardship from control. It needs:

• universities that reward fundamental questions;
• research institutions that tolerate conclusions unfavorable to their sponsors;
• environmental standards that do not overstate their achievements;
• management systems that recognize restraint as a legitimate outcome.

Most of all, it needs the courage to acknowledge that we may not understand what constitutes a living ecosystem, and that what we can dismantle, measure, model, and manipulate is not necessarily something we know how to reconstruct.

The greatest contribution ecological science could make to the future may not be teaching humanity how to control nature more precisely. It may be helping humanity recognize where its knowledge ends, where its authority becomes uncertain, and where the wiser course is to stop pushing the egg off the wall and leave nature alone.

References

1. Opie, I. & Opie, P. (1997). The Oxford Dictionary of Nursery Rhymes (2nd ed.). Oxford University Press. [Entry: "Humpty Dumpty"; public-domain rhyme, earliest printed versions c. 1797–1810.]
2. Levins, R. (1966). The strategy of model building in population biology. American Scientist, 54(4), 421–431.
3. Oreskes, N., Shrader-Frechette, K. & Belitz, K. (1994). Verification, validation, and confirmation of numerical models in the earth sciences. Science, 263(5147), 641–646.
4. Beven, K. (2006). A manifesto for the equifinality thesis. Journal of Hydrology, 320(1–2), 18–36.
5. Raymond, P., Champagne, E., Dumais, D., Kern, C.C., Périé, C., Munson, A.D., Tremblay, J.-P. & Royo, A.A. (2025). Moving up north: How do translocated seedlings perform in mixed-species plantings at the boreal-temperate interface? Forest Ecology and Management, 597, 123179. https://doi.org/10.1016/j.foreco.2025.123179
6. Kay, J.J., Regier, H.A., Boyle, M. & Francis, G. (1999). An ecosystem approach for sustainability: addressing the challenge of complexity. Futures, 31(7), 721–742.
7. Koestler, A. (1967). The Ghost in the Machine. Hutchinson.
8. Allen, T.F.H. & Starr, T.B. (1982). Hierarchy: Perspectives for Ecological Complexity. University of Chicago Press.
9. Holling, C.S. (1973). Resilience and stability of ecological systems. Annual Review of Ecology and Systematics, 4, 1–23.
10. Scheffer, M., Carpenter, S., Foley, J.A., Folke, C. & Walker, B. (2001). Catastrophic shifts in ecosystems. Nature, 413, 591–596.
11. Funtowicz, S.O. & Ravetz, J.R. (1993). Science for the post-normal age. Futures, 25(7), 739–755.
12. Holling, C.S. (Ed.) (1978). Adaptive Environmental Assessment and Management. John Wiley & Sons.
13. Rosen, R. (1991). Life Itself: A Comprehensive Inquiry into the Nature, Origin, and Fabrication of Life. Columbia University Press.
14. Puettmann, K.J., Coates, K.D. & Messier, C. (2009). A Critique of Silviculture: Managing for Complexity. Island Press.
15. Messier, C., Bauhus, J., Doyon, F., Maure, F., Sousa-Silva, R., Nolet, P., Aquilué, N., Fortin, M.-J. & Puettmann, K. (2019). The functional complex network approach to foster forest resilience to global changes. Forest Ecosystems, 6, 21. https://doi.org/10.1186/s40663-019-0166-2
16. DIVERSE Project (n.d.). Theme 2: Identification and selection of suitable tree species. Université du Québec en Outaouais. https://diverseproject.uqo.ca/theme-2-identification-selection-suitable-tree-species/ (accessed 13 October 2026).
17. Messier, C. (senior author) et al. (2026). DIVERSE: Implementing the Functional Complex Network approach alongside Climate-Smart Forestry in Canada. USDA Forest Service Treesearch record 80877. https://research.fs.usda.gov/treesearch/80877 [full bibliographic details to be confirmed].
18. Hardy, C. et al. (2025). Climate is stronger than you think: Exploring functional planting and TRIAD zoning for increased forest resilience to extreme disturbances. PLOS ONE, 20, e0326627. https://doi.org/10.1371/journal.pone.0326627 [author list to be confirmed].
19. Messier, C., Tittler, R., Kneeshaw, D.D., Gélinas, N., Paquette, A., Berninger, K., Rheault, H., Meek, P. & Beaulieu, N. (2009). TRIAD zoning in Quebec: Experiences and results after 5 years. The Forestry Chronicle, 85(6), 885–896.
20. Kuhn, T.S. (1962). The Structure of Scientific Revolutions. University of Chicago Press.
21. Québec (2010). Loi sur l'aménagement durable du territoire forestier (Sustainable Forest Development Act), RLRQ c. A-18.1. Éditeur officiel du Québec.
22. McRae, D.J., Duchesne, L.C., Freedman, B., Lynham, T.J. & Woodley, S. (2001). Comparisons between wildfire and forest harvesting and their implications in forest management. Environmental Reviews, 9(4), 223–260.
23. NCASI (2006). Similarities and Differences between Harvesting- and Wildfire-Induced Disturbances in Fire-Mediated Canadian Landscapes. Technical Bulletin No. 924. National Council for Air and Stream Improvement.
24. Moore, J.-D., Ouimet, R. & Duchesne, L. (2012). Soil and sugar maple response 15 years after dolomitic lime application. Forest Ecology and Management, 281, 130–139.
25. Moore, J.-D., Ouimet, R. & Roy, M.-E. (2026). Le chaulage, à lui seul, ne garantit pas le succès de la régénération de l'érable à sucre. Avis de recherche forestière no 215. Ministère des Ressources naturelles et des Forêts du Québec.
26. Lemieux, G. & Lapointe, R.A. (1986). Le bois raméal et les mécanismes de fertilité du sol. Université Laval, Québec.
27. Dijkstra, F.A. & Smits, M.M. (2002). Tree species effects on calcium cycling: the role of calcium uptake in deep soils. Ecosystems, 5, 385–398.
28. Association forestière du sud du Québec (AFSQ) (n.d.). Acériculture: principes de base de l'aménagement d'une érablière. https://www.afsq.org/fr/acericulture
29. AFSQ (n.d.). Zoom sur une espèce forestière : le tilleul d'Amérique. https://www.afsq.org/fr/trouver-de-linformation/zoom-sur-une-espece-forestiere-le-tilleul-damerique
30. AFSQ (n.d.). L'érablière à sucre : une source de biodiversité. https://www.afsq.org/fr/trouver-de-linformation/lerabliere-a-sucre-une-source-de-biodiversite
31. van den Berg, A.K. & Perkins, T.D. (2014). A model of the tapping zone. Maple Syrup Digest, 26A(1), 18–27.
32. van den Berg, A.K., Perkins, T.D., Isselhardt, M. & Wilmot, T. (n.d.). Growth rates of sugar maple trees tapped for maple syrup production using high-yield sap collection practices. University of Vermont Proctor Maple Research Center. https://mapleresearch.org/wp-content/uploads/highyieldgrowth.pdf
33. University of Vermont (2013). Abby van den Berg taps the science behind how trees produce sap. https://www.uvm.edu/rsenr/news/abby-van-den-berg-taps-science-behind-how-trees-produce-sap
34. McLachlan, J.S., Clark, J.S. & Manos, P.S. (2005). Molecular indicators of tree migration capacity under rapid climate change. Ecology, 86(8), 2088–2098.
35. Loarie, S.R., Duffy, P.B., Hamilton, H., Asner, G.P., Field, C.B. & Ackerly, D.D. (2009). The velocity of climate change. Nature, 462, 1052–1055.
36. Park, A. & Talbot, C. (2018). Information underload: ecological complexity, incomplete knowledge, and data deficits create challenges for the assisted migration of forest trees. BioScience, 68(4), 251–263.
37. O'Neill, G., Wang, T., Ukrainetz, N., Charleson, L., McAuley, L., Yanchuk, A. & Zedel, S. (2017). A Proposed Climate-Based Seed Transfer System for British Columbia. Technical Report 099. Province of British Columbia, Victoria.
38. Province of British Columbia (n.d.). Chronology of seed transfer and seed planning zone changes. https://www2.gov.bc.ca/gov/content/industry/forestry/managing-our-forest-resources/tree-seed/seed-planning-use/seed-planning-chronology
39. Benomar, L. et al. (2018). La migration assistée comme mesure proactive d'adaptation aux changements climatiques : cas de l'épinette blanche au Québec. 12e Colloque du Centre d'étude de la forêt (abstract). http://cef.uqam.ca/colloque/2018/prg_abstract.php?ID=2982
40. Ministère des Ressources naturelles et des Forêts du Québec (n.d.). Sources de semences. https://mffp.gouv.qc.ca/les-forets/production-semences-plants-forestiers/semences/sources/
41. Royo, A.A., Raymond, P., Kern, C.C., Adams, B.T., Bronson, D., Champagne, E., Dumais, D., Gustafson, E., Marquardt, P.E., McGraw, A.M., Miesel, J.R., Munson, A.D., Périé, C., Tavares Moreira, F.J., Ola, A., Bouchard, M. & Bissonnette, J.-F. (2023). Desired REgeneration through Assisted Migration (DREAM): implementing a research framework for climate-adaptive silviculture. Forest Ecology and Management, 546, 121298.
42. Benomar, L., Bousquet, J., Perron, M., Beaulieu, J. & Lamara, M. (2022). Tree maladaptation under mid-latitude early spring warming and late cold spell: implications for assisted migration. Frontiers in Plant Science, 13, 920852. https://doi.org/10.3389/fpls.2022.920852
43. Mura, C., Charrier, G., Buttò, V., Delagrange, S., Surget-Groba, Y., Raymond, P., Rossi, S. & Deslauriers, A. (2025). Local conditions have greater influence than provenance on sugar maple (Acer saccharum Marsh.) frost hardiness at its northern range limit. Tree Physiology, 45(1), tpae167.
44. Aubin, I., Garbe, C.M., Colombo, S., Drever, C.R., McKenney, D.W., Messier, C., Pedlar, J., Saner, M.A., Venier, L., Wellstead, A.M., Winder, R., Witten, E. & Ste-Marie, C. (2011). Why we disagree about assisted migration: ethical implications of a key debate regarding the future of Canada's forests. The Forestry Chronicle, 87(6), 755–765.
45. McLachlan, J.S., Hellmann, J.J. & Schwartz, M.W. (2007). A framework for debate of assisted migration in an era of climate change. Conservation Biology, 21(2), 297–302.
46. Mora, C., Tittensor, D.P., Adl, S., Simpson, A.G.B. & Worm, B. (2011). How many species are there on Earth and in the ocean? PLoS Biology, 9(8), e1001127.
47. Convention on Biological Diversity (2022). Kunming-Montreal Global Biodiversity Framework. Decision 15/4, Target 15.
48. Taskforce on Nature-related Financial Disclosures (2023). Recommendations of the Taskforce on Nature-related Financial Disclosures. TNFD.
49. Science Based Targets Network (2023). Technical Guidance: Step 3: Measure, Set, Disclose (Freshwater and Land). SBTN.
50. Reig, P., Larson, W., Vionnet, S. & Bayart, J.-B. (2019). Volumetric Water Benefit Accounting (VWBA): A Method for Implementing and Valuing Water Stewardship Activities. Working Paper. World Resources Institute.
51. E&E News (2021). Questions trail big business' 'water positive' pledges. E&E News by POLITICO. https://www.eenews.net/articles/questions-trail-big-business-water-positive-pledges/
52. Microsoft (n.d.). Water replenishment: Our learnings on the journey to water positive. Microsoft Corporation. https://cdn-dynmedia-1.microsoft.com/is/content/microsoftcorp/microsoft/msc/documents/presentations/CSR/Water-replenishment-learnings-on-journey-to-water.pdf
53. Small Bottle, Big Pipe: Quantifying and addressing the impact of data centers on public water systems (2026). arXiv:2603.02705. https://arxiv.org/abs/2603.02705
54. CBC News (2016). Why Nestlé's Aberfoyle well matters so much to Guelph, Ont., residents. https://www.cbc.ca/news/canada/kitchener-waterloo/nestle-guelph-rally-nestle-water-aberfoyle-1.3779649
55. The Canadian Press / The Globe and Mail (2016). Liz Sandals says opposition to Nestlé water-taking based on 'misinformation'. https://www.theglobeandmail.com/news/national/liz-sandals-says-opposition-to-nestle-water-taking-based-on-misinformation/article32107624/
56. Berkes, F. (2012). Sacred Ecology (3rd ed.). Routledge.
57. Zürcher, E., Cantiani, M.-G., Sorbetti-Guerri, F. & Michel, D. (1998). Tree stem diameters fluctuate with tide. Nature, 392, 665–666.
58. Vesala, T., Sevanto, S., Paatero, P., Nikinmaa, E., Perämäki, M., Ala-Nissilä, T., Kääriäinen, J., Virtanen, H., Irvine, J. & Grace, J. (2000). Do tree stems shrink and swell with the tides? Tree Physiology, 20(9), 633–635.
59. Barlow, P.W., Mikulecký, M. & Střeštík, J. (2010). Tree-stem diameter fluctuates with the lunar tides and perhaps with geomagnetic activity. Protoplasma, 247, 25–43.