Holism and Climate Change

Climate change is usually presented as a problem of magnitude. How much warming? How much sea-level rise? How much carbon remains in the budget? How severe will the next drought, flood or fire become? These are necessary questions, but they quietly preserve one enormous assumption: that the system in which these changes are occurring remains sufficiently unchanged for its future to be understood mainly by measuring how far its variables move.

That assumption becomes increasingly dangerous as change accumulates. A warmer ocean does not simply contain more heat. It exchanges energy with a different atmosphere, encounters diminished ice, participates in altered circulation and rainfall, and affects ecosystems already carrying the consequences of previous change. The next event occurs in conditions partly produced by the events before it. Causes persist through their consequences, feedbacks arrive after delays, relations that once restored familiar conditions weaken, and processes separated by our models continue changing one another. Eventually the important question is no longer simply how far the climate has moved from some previous state. It is whether the relations that made that state reproducible still exist.

This is where climate change becomes a problem of the whole. Not because everything affects everything, and not because understanding climate requires an impossible view of the entire planet at once. The problem is more precise. The atmosphere, ocean, ice, soil, life and human activity do not merely sit beside one another as components of a larger machine. Their persistent relations help constitute the systems we distinguish, while those relations can themselves become systems entering further relations. Change enough of that structure and the planet does not merely move to another state within the same climatic system. The system changes the conditions from which its subsequent states can be produced.

Climate change is not a problem occurring inside a stable planetary system. It is a change in the conditions through which planetary stability is reproduced.

This is not an argument for replacing climate science with systems language. The physical mechanisms of anthropogenic warming are well established. Human activity has altered atmospheric composition and the planetary energy balance, warming atmosphere, ocean and land while changing the cryosphere, hydrological cycle and biosphere. The problem begins when phenomena necessarily separated for measurement and modelling are mistaken for processes that are similarly separate in the world.

They are not. The ocean absorbs and transports heat while exchanging energy, moisture and momentum with the atmosphere. Ice loss changes subsequent absorption of solar energy. Vegetation participates in exchanges of carbon, water and energy. Atmospheric circulation changes rainfall and ocean conditions, which alter vegetation, ice, soils and subsequent circulation. Each process changes conditions encountered by others.

The climate system is relational before anyone draws the arrows.

A useful holism begins there. A whole is not produced by adding enough parts together, and understanding one does not require knowing everything at once. The problem is to identify which differences remain consequential to one another, which relations persist across changes of scale and time, and which cannot be removed without changing the behaviour being explained. The distinctions we call parts remain real and measurable, but their identities need not be complete prior to their relations. Persistent relations can participate in producing the identities, boundaries and possible behaviour of the systems we subsequently distinguish as parts.

A Climate Is Something That Recurs

The first consequence is that climate cannot be understood simply as a collection of states. It exists through recurrence.

No two days are identical. No two summers reproduce one another exactly. No El Niño event simply repeats the previous one. Yet enough relational structure persists beneath this variation for climate to be distinguishable from weather at all. Seasonal cycles recur. Temperature and rainfall occupy characteristic distributions. Circulation patterns form, move, weaken and return. Ocean and atmosphere repeatedly enter recognisable configurations without reconstructing precisely the same world.

Recurrence is not repetition. It is persistence through difference.

An event does not acquire its climatic significance from magnitude alone. A temperature extreme means something different according to its distribution, season, humidity, surrounding ocean temperature, antecedent soil moisture and the systems exposed to it. The same rainfall deficit arriving in already dry soil is not dynamically equivalent to one following a wet period. What happens depends partly upon the conditions into which it happens.

Those conditions have temporal structure. Climate contains processes operating across hours, seasons, years, decades and centuries. Some reinforce one another, some interfere, some respond rapidly, while others retain consequences long after the initiating event has passed. Frequency matters, but so do amplitude, relative phase, coupling and delay. Apparently irregular behaviour can contain persistent relational structure, and that structure can itself change.

This is the first reason climate change cannot be understood simply by asking whether particular events have become larger or more frequent. The distribution from which events emerge can change, but so can the relations through which that distribution is reproduced. The system producing the events is itself under change.

Relations That Become Systems

The El Niño–Southern Oscillation (ENSO) provides a useful example. The ocean and atmosphere can be distinguished, measured and modelled separately, but ENSO is neither simply an ocean event nor simply an atmospheric event. Its identity belongs to a persistent relation involving both. Remove the relation and there is no remaining object called ENSO hiding in either system.

This matters because relation is not secondary in the account being developed here. Suppose we distinguish systems A and B within a persistent relation. Their relation can itself constitute another system, C. C is not reducible to A or B, while A and B need not be fully specified independently of the relation through which they are distinguished. Once C persists sufficiently to become consequential, relations between A and C, B and C, or larger ensembles containing them can themselves constitute further systems. Relation does not merely connect completed things. It participates in producing distinguishable things while becoming available to subsequent relation itself.

The relation is also doubly oriented. A acts through conditions presented by B while B acts through conditions presented by A, and those transformations are not interchangeable. Each participates in producing conditions encountered by the other. The persistent relation can change both while acquiring a history and dynamics that cannot be located entirely within either considered alone.

This does not terminate at three. In an ensemble of N systems, relations can involve apparent components, relations among components, relations involving those relations and higher-order structures produced through their recurrence. What appears as a relation at one resolution can become a system at another, while that system remains relationally constituted in turn.

This also means that systems are not simply given first and related afterwards. Some distinctions can be made independently for particular purposes, but recurrent relations can stabilise differences through which identifiable systems persist. Identity is not erased by relation. It can be produced and maintained through persistent difference within relation.

Not every conceivable relation is consequential to every behaviour. The relevant question is whether changing or removing a relation changes the behaviour, persistence or accessible transitions under examination. Holism requires discrimination, not indiscriminate connection.

There is no intrinsic final level at which relation must stop becoming available to further relation. This is not a hierarchy requiring us to climb from one level to the next until a final system is reached. It is a recursive property of relational structure: systems can arise through relations and become available to further relations without ceasing to depend upon the relations through which they persist.

For climate change, this matters immediately. Ocean and atmosphere form coupled systems whose behaviour affects rainfall, vegetation, fire, ice, ecosystems and human activity. Those interactions acquire their own histories and can become consequential to still larger configurations. What changes at one apparent level can alter the relations through which another level exists.

Stability Has a History

This changes what we mean by climatic stability. A stable climate has never been still. Energy enters and leaves the planetary system. Air and oceans circulate. Water changes phase and location. Ecosystems grow, die and reorganise. Seasons arrive and depart. What appears stable over longer intervals is produced through continual activity at shorter ones.

Stability is not the absence of change. It is the persistence of relational structure through change.

That distinction matters because a system can continue to look familiar while the relations reproducing its familiar behaviour are weakening. A mean temperature, rainfall total or seasonal pattern can remain within an historically recognisable range while ocean heat, soil moisture, ice, ecological condition or circulation has already changed the conditions under which the next event will occur.

Every recurrence arrives after something else has happened. Oceans retain heat. Soils retain or lose water. Ice loss changes the surface encountered by later radiation. Vegetation altered by drought, heat or fire changes subsequent exchanges of water and carbon. Infrastructure persists beyond the decisions that produced it. The present contains consequences whose initiating events no longer need to be present.

The system carries its history because previous activity has changed the conditions of subsequent activity. The past remains active without remaining present in its original form.

This does not require any reversal of causation. Later events do not reach backwards and alter earlier ones. They can, however, reveal consequences through which an earlier event remains active. What appeared transient can become part of a persistent change once its relations to later events become visible. The meaning of the past can change because more of its consequences have occurred, while the causal sequence continues forward.

Delay Changes the System

History is inseparable from delay. Atmosphere, ocean, vegetation, soils, ice sheets, infrastructure and institutions respond at different rates. Climate is not merely interconnected. It is polyrhythmic.

Delay is not just waiting time between cause and effect. It changes what else is present when an effect arrives. A feedback has duration. During that duration other processes continue, reinforce one another, interfere, overshoot or alter the conditions into which the feedback returns. Timing is part of causal structure.

This means the order of events matters. A second heatwave does not occur in the world that existed before the first if the first changed soil moisture, vegetation, water storage or vulnerability. Heavy rainfall following prolonged drought encounters different soils and ecosystems from the same rainfall following a wet period. Ocean warming arriving alongside altered circulation produces different consequences from the same nominal quantity considered in isolation.

An event can recur while the conditions of its recurrence move underneath it.

This is one reason simple causal diagrams can become misleading when read too literally. An arrow from A to B can identify an important causal relation while concealing the fact that B may respond only after A has changed again, that other processes operate during the interval, and that the returning effect may encounter a system different from the one in which the interaction began. The arrow has a history.

When Possibility Changes

We can now return to the assumption hidden inside the usual language of magnitude. If climate change altered only the values of variables within an otherwise unchanged system, sufficiently accurate measurement could in principle tell us how far through a stable space of possibilities the system had moved. The deeper problem is that the space is not functionally fixed.

Two moments can display similar atmospheric temperatures while differing in ocean heat, soil moisture, ice, vegetation, circulation and the trajectories from which those measurements arose. Similar present measurements need not imply dynamically equivalent states. One can belong to a system capable of readily returning towards a previous regime while the other belongs to a system in which the relations supporting that recovery have weakened.

This matters near transitions. A system does not need to look dramatically different before its capacity to reproduce its previous relational structure has begun to change. Recovery can slow. Variability can change. Relations that previously returned the system towards a familiar regime can weaken while alternative configurations become easier to enter.

A tipping point is not fundamentally a number at which a system suddenly changes its behaviour. A threshold can be an indispensable indicator, but the underlying event is a reorganisation of persistence. One relational configuration becomes less capable of reproducing itself while another becomes more accessible.

What changes first can be not appearance, but possibility.

Possibility here is not a substance, destination or hidden future. It names the transition structure generated by the system as it presently exists. Some subsequent states and trajectories can be reached from that structure and others cannot, while their relative probabilities differ. Change the relations composing the present and the transition structure itself changes. A warmer ocean, lost ice, dry soil, ecological change or altered infrastructure does not merely move the system towards another point within an unchanged catalogue of futures. It changes the conditions from which future states can be produced.

The system does not merely move through a fixed space of possibilities. Its activity changes the conditions governing its subsequent transitions. In this sense, the system reorganises what is possible for itself without determining one inevitable future.

Climate change acts in both the event and the conditions inherited by whatever follows it. State and trajectory cannot be collapsed into one another. A measurement tells us something indispensable about what exists now. A relational description also asks what produced it, what sustains it, which delayed processes remain active, how readily the previous structure can be reproduced, and which subsequent states have become more or less accessible.

We Are Not Outside It

There is another complication. Human beings are not merely observing these changes from outside the system. We participate materially and reflexively in the larger planetary field.

We alter atmospheric composition, land cover, water systems and ecosystems while observing those changes and responding to them. Agriculture responds to rainfall. Energy systems respond to temperature and policy. Insurance responds to losses. Governments and markets respond to models, prices and expectations. Some of those responses return to the physical system through changed emissions, infrastructure and land use.

This does not make atmospheric physics a social construction. It means that a planetary system containing human beings also contains some consequences of human descriptions of that system.

A climate model begins as a representation. Once institutions act through its projections, the model can enter the causal pathway it describes. Observations inform models; models alter expectations; expectations inform decisions; decisions alter material activity; that activity becomes part of the future system being observed. The model does not stand outside the planetary field once actions taken through it become causally consequential within that field.

Language participates in the same limited but consequential sense. Meaning does not alter radiative transfer. It can alter behaviour, and behaviour can alter emissions, land use and vulnerability. Processes separated for analysis can become related again through the consequences of human action.

The attempt to understand the system can become one of the relations through which the system changes.

The Boundary Is Part of the Question

None of this makes decomposition a mistake. Decomposition is how we know as much as we do. The mistake is forgetting what the decomposition removed.

A model can legitimately exclude processes irrelevant to its question. The difficulty is that relevance changes with scale and time. A relation negligible over days can become decisive across decades. A local process can become consequential through recurrence across space. What leaves one model as an externality can return through another pathway as a cause.

A useful holism does not abolish boundaries. It makes them answerable.

Change the scale and see which relations survive. Alter temporal resolution and see what disappears. Couple previously separated models and ask whether their behaviour changes. Remove a feedback, introduce its delay or change a boundary condition and ask whether the phenomenon can still be reproduced. What persists need not be a particular variable or component. It can be a relation, constraint, lag, recurrent pattern or transformation.

A model is holistically adequate to the extent that consequential relational invariants survive its decomposition.

This is not an argument against reduction. It identifies when reduction has removed something necessary to the behaviour being explained.

Nor is there one privileged scale at which the whole finally appears. A forest can be carbon storage, water exchange, habitat, fuel, productive land and culturally significant place. These descriptions are not interchangeable, but neither do they describe unrelated worlds. Changes represented in one can become consequential within another because relations cross their analytical boundaries.

The whole is not a final box around smaller boxes. Something can be a whole for one question and a subsystem for another. A relation among systems can itself become another system, which can enter further relations without ceasing to depend upon those through which it arose. Holism does not mean finding the largest available object. It means refusing to mistake a useful boundary for the end of relation.

What Becomes Measurable

If this were only a philosophical change of vocabulary, it would add little to climate science. Its value depends upon whether differences in relational structure can be observed, compared and tested.

Many can. Temperature, greenhouse-gas concentrations, ocean heat, sea level, ice extent and precipitation remain indispensable. A relational account adds questions about how consequential relations among such quantities change through time. Coupling strength, covariance, lag, persistence, recurrence, spatial coherence, recovery rates, synchronisation, conditional dependence and spectral structure can all be investigated where physically appropriate.

Earth-system modelling, dynamical systems, temporal and higher-order networks, recurrence analysis, causal inference, resilience research and critical-transition analysis already provide substantial machinery for doing so. Different research traditions repeatedly encounter parts of the same structural problem. Temporal ordering changes which paths exist. Higher-order interactions can produce dynamics absent from pairwise descriptions. Recurrence reveals persistence beneath non-identical events. Network histories alter subsequent behaviour. Changing recovery dynamics can precede visible transitions.

The vocabulary differs. The underlying difficulty keeps returning.

This matters because the central claim becomes empirically vulnerable rather than rhetorically convenient. If relational structure matters, then changing the relevant coupling, delay, recurrence, boundary or higher-order interaction should sometimes change explanatory or predictive performance. If it does not, the additional relational machinery has earned no place in that particular explanation. Holism does not receive a free pass merely by calling itself holistic.

From Prediction to Navigation

Climate science has understandably invested enormous effort in prediction: temperatures, rainfall, sea level, extremes, carbon budgets and trajectories under different emissions scenarios. That work remains essential. But prediction is not the whole problem when present activity changes the conditions under which later activity will occur.

Across longer horizons, ecological change, infrastructure, technology, economics, policy and adaptation become part of the trajectory. A projection can alter investment; investment can alter infrastructure; infrastructure can alter emissions and vulnerability. Description and system can become causally coupled without becoming the same thing.

The problem becomes partly navigational. Which relations are becoming brittle? Which slow processes are carrying consequences already produced by earlier forcing? Where are delays producing overshoot? Which interventions preserve possibilities rather than improving one variable while transferring instability elsewhere? Which systems still look familiar while losing the relations that previously allowed them to recover?

Carbon remains central. Anthropogenic greenhouse-gas emissions are the principal driver of contemporary warming. A whole-system view does not weaken that causal fact. It follows the forcing into the relations through which warming propagates, histories accumulate, ecosystems respond, institutions react and subsequent transition probabilities change.

This also changes the meaning of intervention. An intervention is not only an attempt to push a variable in a preferred direction. It enters an already moving relational field, encounters processes operating at different rates and can alter conditions inherited by later interventions. The same action taken at another time, scale or system state need not have the same consequence.

The question is no longer merely how much change will occur. It is what kind of system will be there to receive it.

Holism Without Omniscience

The obvious objection to all of this is that nobody can understand the entire planetary system. That is true and irrelevant. Nobody needs to.

Holism is not omniscience. It is discipline about what our distinctions do.

We isolate the atmosphere to understand radiative transfer, an ocean basin to understand circulation, a forest to understand carbon exchange or an economy to understand investment. Such reductions are powerful. Each also makes a wager that what has been excluded will not materially change the answer being sought.

Sometimes that wager succeeds. Sometimes it succeeds only at a particular scale or interval. Sometimes the excluded relation returns.

This does not imply that a sufficiently enlarged model will eventually contain the final whole. Enlarging a boundary changes the system being described and introduces further relations among what has been joined. The relation between previously separated systems can itself acquire consequential structure. Integration does not abolish relation. It produces further relations capable of becoming systems in their own right.

The whole is neither inaccessible nor completely capturable. We encounter it through partial descriptions whose adequacy can be tested against observation, against one another and against their failures. Reality constrains those descriptions because consequential relations continue operating whether or not a model contains them.

No description completes the system it describes. Change the question, scale or interval and another distinction can become consequential. Extend the boundary and another relation crosses it. Relate previously separated systems and the relation itself becomes available as another system. Explanation does not require a final position from which everything is simultaneously visible.

Climate Change Changes What Comes Next

We began with the familiar questions of magnitude: how much warming, how much sea-level rise, how much carbon, how severe the next extreme. Nothing in this argument makes those questions less important. It changes what we think their answers describe.

A temperature increase is not merely movement along a thermometer. It occurs in an ocean-atmosphere system carrying accumulated heat, beside changing ice, altered water cycles, shifting ecosystems and human systems responding to both observations and expectations. Each change enters conditions encountered by what follows. The quantities matter because the relations through which they acquire consequence matter.

Climate change is not a problem occurring inside a stable planetary system. Anthropogenic forcing is altering the persistent structure of relations through which climatic stability has been reproduced. Energy accumulates differently. Oceans store and transport the consequences. Ice retreats and changes subsequent absorption. Water moves through altered atmospheric and terrestrial conditions. Ecosystems encounter climates carrying the history of earlier change. Human systems observe, interpret and respond, and some responses return materially to the processes being observed.

These are not independent symptoms assembled afterwards under the heading climate change. Nor are they merely components connected inside a larger container. Their recurrent relations help constitute the systems whose behaviour we distinguish, while persistent relations among those systems become systems capable of entering further relations in turn.

Recurrence continues, but it does not recur into the same world. Recurrence carries history. History changes the conditions of recurrence. Those conditions alter what can happen next. What happens next becomes part of the history inherited by what follows.

There is no return to an identical beginning, nor a final state towards which the system is trying to move. Stability is not purpose. Persistence is not intention. What persists does so because existing relations constrain subsequent transitions, making some configurations more accessible than others. Change those relations and the structure of subsequent possibility changes with them.

No central term has to tell ocean, atmosphere, ice, forest, soil, carbon, economy or culture how to compose a planetary whole. No one of them contains that whole, and adding them together does not produce a final object that does. Their relations constitute further systems; those systems enter further relations; and the resulting structure remains open to scales, delays and consequences absent from any particular description.

This is not an infinite regress requiring an unreachable final explanation. It is the structure being described. Relation remains generative because no local term exhausts the field of relations through which it exists.

Nor does relation need to resolve into a completed unity. Relations become systems, those systems enter further relations, and no resulting system exhausts the conditions through which further relation remains possible. The absence of a final centre is not a hole waiting to be filled. It is part of the structure that prevents relational closure.

The consequence is more substantial than saying that climate is complicated. The planet is not simply moving from one climatic state to another through an unchanged space of possibilities. Its present activity is changing the conditions inherited by its future activity. Some relations will persist, some will weaken, some will disappear and others will become consequential in configurations that did not previously exist. What must be understood is not only where the system is going, but how the system that can go anywhere at all is changing as it moves.

Climate change is changing what the climate system is likely to become.

---

Complexity Science: Points of Contact

The argument does not depend upon complexity science having previously assembled precisely the same account. Its relevance is that independent research traditions repeatedly encounter parts of the structure described here. Near-decomposability makes useful decomposition dependent upon interaction and timescale. Temporal-network research demonstrates that temporal ordering affects available dynamical paths. Higher-order network theory shows that collective interactions need not be recoverable from pairs. Recurrence analysis identifies dynamical structure without requiring exact repetition. Network-memory research demonstrates dependence of present relational structure upon its history. Critical-transition research investigates changes in recovery dynamics that can precede an obvious regime shift.

These structures are not philosophical inventions looking for somewhere to land. They have mathematical descriptions, empirical applications and measurable consequences. The points of contact include Simon (1962) on near-decomposability; Kelso (1995) on coordination dynamics; Bar-Yam (2004) on multiscale variety; Marwan et al. (2007) on recurrence; Scheffer et al. (2009) on critical transitions; Holme and Saramäki (2012) on temporal networks; Ghil and Lucarini (2020) on climate variability and change as a complex dynamical problem; Weinberger (2020) on the timescale relativity of causal representation; Battiston et al. (2020) on higher-order interactions; and Williams et al. (2022) on memory in temporal networks.

No one of these works contains the argument made here. Their significance lies in what keeps returning across them: timing, history, scale, coupling, recurrence, conditional boundaries and forms of collective behaviour that disappear when relation is treated as secondary to independently specified things.

Wallace, G. (2026) ‘The Strait of Hormuz: Cascades, Delay and the Architecture of a Crisis’, Daedelus Kite, 30 August.

This essay applies the relational argument to a coupled global system in which energy, logistics, finance, military force, politics, infrastructure and communication operate at different rates while remaining consequential to one another. Cascades expose dependencies that ordinary operation can conceal. No local participant contains the whole, yet changes propagate through relations extending beyond every local description.

Wallace, G. (2020) ‘Global Systems’, Daedelus Kite, 16 November.

Global Systems approaches the same problem from an earlier and more abstract direction. A global system has no final component containing its unity, no privileged boundary beyond which relation ceases and no master description in which every consequential difference becomes simultaneously present. Something remains elsewhere: another scale, timescale, relation, consequence or possible description.

This incompleteness does not prevent the system from being a system. It is one of the conditions under which further relation remains possible. No local term has to complete the whole for relations to continue beyond every provisional boundary, and relations themselves can acquire sufficient persistence to become systems participating in further relations.

There is no climatic object standing behind atmosphere, ocean, ice, soil, life and human activity directing their coordination. Nor does their relation need to resolve into a completed unity. Relations become systems, those systems enter further relations, and no resulting system exhausts the conditions through which further relation remains possible.

The absence of a final centre is not a hole waiting to be filled. It is part of the structure that prevents relational closure. The missing centre is not missing from the explanation. It is part of what the explanation has finally had to explain.

---

Formal Companion

Climate Change and the Logic of the Whole develops the formal companion to this conceptual account, expressing the same relational structure through explicit variables, transition dynamics, delay, probability and testable propositions.