The coupled planet: how forests, groundwater, rain, & climate shape each other. A complex systems approach

Tipping points, regime shifts, feedback loops, and fitness landscapes

[Aragao 2012]

There is a powerful and surprisingly lucid way to understand the complex, integrated system that is Earth’s ecology, water, and climate. One of the keys to understanding our environmental resilience is that the planet is fundamentally nonlinear. To safeguard our future and manage environmental risk, we must move beyond gradualist assumptions and answer a critical question: What are the specific stabilizing loops that keep us in our current environmental regime, and what are the powerful, invisible drivers that could suddenly tip the system into a fundamentally different state?

The key to this analysis comes from the field of complex systems. This framework reveals that the Earth is not simply a linear machine responding to external forces, but a vast, interconnected network of causal loops. Because of this intricate coupling, the system doesn’t just decline or grow slowly, it can suddenly change. This nonlinearity is evidenced in planetary-scale phenomena where biomes, when pushed too far, abruptly reorganize themselves, shifting into a totally different, often irreversible state, much like liquid water instantaneously freezing into ice. Scientists call these sudden, systemic changes regime shifts or tipping points.

This lens provides the essential language and analytical tool-set for understanding the planet’s stability. Every stable ecological state, whether it be a desert, a rainforest, an ice sheet, is essentially a “mode” held in place by a constant, hidden struggle between two core forces: negative feedback loops that seek to stabilize the system by counteracting disturbances, and positive feedback loops (like a snowball rolling down a hill) that seek to amplify change by pushing the system further in the direction it is already moving. Understanding this struggle, knowing which loops stabilize us, and being wary of the drivers that unleash the amplifying loops, is crucial for climate and environmental policy. A regime shift occurs when the amplifying positive feedback overwhelms the stabilizing negative feedback, pushing the system past a critical threshold.

These regime shifts are a direct consequence of coupling, the non-independent and reciprocal linking of the Earth’s major sub-systems. Complex systems theory shows that the way components lock together determines the number and stability of available modes. Our deepening understanding of the climate system, therefore, requires radical interdisciplinary collaboration, the bringing together of hydrologists, ecologists, atmospheric scientists, and oceanographers. We must move past isolated studies of water or atmosphere to grasp how groundwater, the biosphere (vegetation), the atmosphere, and the oceans all actively couple. These deep, intertwined relationships, where the ocean influences rainfall, where groundwater influences biomes, where biomes and deforestation influences climate, where climate influences oceans, create a single, global network. This integrated approach is the only way to understand how all these things fit together to create planetary regimes. An example of this dramatic principle of nonlinearity and tipping is the fate of the Amazon rainforest.

Consider a forested landscape. Trees pull water from the soil and release it into the air through their leaves. That moisture forms clouds and falls back as rain. The rain sustains the forest, and a healthy forest transpires even more moisture. It becomes a self-reinforcing cycle: the forest creates its own rain, and that rain sustains the forest that created it. This is the “forest-rain mode,” a deeply coupled system in which vegetation, soil moisture, and atmosphere amplify one another.

In some regions, the same land can occupy a very different mode. In a savanna, sparse vegetation and grasses transpire only modest amounts of moisture. With so little water entering the air, the region produces little of its own rainfall. The landscape depends on whatever storms drift in from far away. With less rain, only drought-tolerant plants survive, and those plants contribute even less to atmospheric moisture. The feedback loop runs in the opposite direction, locking the region into a dry state. The difference between rainforest and savanna is not solar radiation or latitude; it is which feedback loops dominate.

Carlos Nobre, one of Brazil’s leading climate scientists, and Thomas Lovejoy have warned for years that the Amazon is approaching a tipping point. The rainforest does not merely receive rainfall; it manufactures it. Up to half of the rain that falls in the basin originates as transpiration from the forest itself. Remove enough forest, and the precipitation-recycling engine weakens. A weakened engine produces less rain, and less rain stresses the remaining forest. Stressed trees transpire less, producing still less rain. Eventually, the system crosses a critical threshold, estimated at around twenty to twenty-five percent deforestation, and the entire basin flips. Forest does not slowly thin; it collapses into a savannized state, even in areas never touched by chainsaws.

We are already at roughly seventeen percent deforestation in the Brazilian Amazon. The tipping point is not so far off distant possibility. And once the system flips, it becomes extremely hard to reverse. A savanna cannot generate the rainfall required to regrow a rainforest. The old feedback loop is gone. Even heroic reforestation efforts struggle because the atmospheric coupling, the invisible engine of moisture recycling, has been broken. The system now remembers its degraded state.

To understand why this irreversibility matters, it helps to zoom out and look at what a regime shift actually is. Any complex system, the climate, a forest, even a human body, stays in a particular mode because negative feedback loops keep it there. Negative feedback is stabilizing: when something changes, the system pushes back.

For instance, think of a thermostat in your home. It maintains room temperature through negative feedback: when the air gets too warm, the heater turns off; when it gets too cold, it turns on. The system stabilizes around a set temperature. Now imagine the thermostat malfunctions or the room is extremely cold. The negative feedback alone cannot maintain balance, and small changes can escalate. The system tips into a new mode where the room is uncomfortably hot or cold.

Another everyday example comes from the “wave” in stadiums. At first, a few people standing and raising their arms is dampened by the surrounding crowd. Negative feedback suppresses the small movement. But if enough people start participating, positive feedback takes over: one section stands because the previous section stood, and the motion propagates around the stadium. The system switches into a new, self-reinforcing mode, a moving wave, that would not exist without the initial excitation to overcome stabilizing forces.

A rainforest behaves this way. If one patch dries temporarily, deeper-rooted trees draw on stored moisture, transpire steadily, and help maintain local rainfall. The system dampens the disturbance. A savanna has its own stabilizing loop: grasses adapted to fire and drought, with low but steady transpiration, prevent total desiccation. Savannahs also have groundwater to draw upon for survival. Each landscape has a “home mode,” held in place by the negative feedbacks that give it resilience.

A regime shift happens when positive feedback overwhelms these stabilizing forces. Positive feedback amplifies change: a little drying leads to less vegetation, which leads to even less moisture recycling, which leads to further drying. Eventually, the stabilizing loop of the forest is too weak to counter the runaway loop pulling the system toward a new state. That is why the transition from forest to savanna is abrupt rather than smooth; the system is not sliding along a gradient but falling into a different basin of attraction, a new self-reinforcing mode. Thinking in terms of modes and feedback loops may sound abstract, but it is one of the most powerful tools we have for understanding how landscapes reorganize themselves and why some changes, once triggered, resist reversal. [Staal et al 2020]

Another idea from complex systems is the idea of fitness landscapes. In the attractor-basins/valleys (which you can see in diagram below) is where the system’s modes will settle into. It requires a perturbation to move it from one of valley to another. When it does the mode or regime shift happens. The idea of hystersis means the system is path dependent. What happens as rain increases depends if you are in the savannah basin or the forest basin.

[The fitness landscape and regime shifts of forest to savannah/grasslands. The system is bistable, meaning it has two modes Diagram from Staal et al 2020 ]

When we add groundwater to this vegetation-rain system this framework adds something extra: the subterranean reservoir, the living canopy above it, and the moisture in the sky form their own web of stabilizing and destabilizing loops, capable of holding a region in a humid, resilient mode or, if the feedbacks turn the wrong way, pushing it toward a drier, more fragile one.

Once we bring groundwater into the picture, the story of modes and feedbacks adds a another systems effect. Groundwater adds a slow, stabilizing reservoir to the system, a deep memory that can carry a forest across months of drought. In a healthy rainforest, this groundwater–tree–atmosphere loop acts as a powerful negative feedback: when the surface soil dries, deep roots tap the aquifer; when trees keep transpiring, they keep moisture recycling alive; when moisture recycling stays intact, rains eventually return; when rains return, groundwater is recharged. Each leg of the loop reinforces the others. The system cushions disturbance. It stays in the forest mode because its stabilizing feedbacks link the underground, the biosphere, and the atmosphere into a single self-regulating circuit.

But these feedbacks have their own tipping point. When groundwater drops below rooting depth, whether by over-extraction, prolonged drought, or cumulative deforestation, trees lose their buffer. They begin behaving like shallow-rooted plants, shutting down transpiration during dry periods. The negative feedback collapses. Suddenly, the system is dominated by positive feedback: less transpiration means less atmospheric moisture; less moisture means less rainfall; less rainfall means weaker groundwater recharge; weaker recharge means groundwater falls even further from reach. The loop amplifies itself, pushing the landscape toward an entirely different mode. What emerges is a reconfigured system in which groundwater, vegetation, and the atmosphere no longer reinforce one another. The reservoirs decouple. The internal rainfall engine falters. The region becomes dependent on whatever moisture external circulation happens to deliver, an identity much closer to savanna than forest.

Groundwater and vegetation interact in other ways too, Studies such as Ulrik Ilstedt and Douglas Sheil [2016] looked at in Burkina Faso have shown that trees create shade and enrich the soil with organic matter, improving its structure and increasing rainwater infiltration. These feedback loops can produce particular “modes” of groundwater storage: for example, in some savannas, intermediate tree cover maximizes groundwater, while too few or too many trees reduce it. Other research suggests different groundwater modes depending on soil type, climate, and vegetation configuration. The system is not linear: vegetation and groundwater reinforce each other, producing emergent states that are stable under certain conditions and prone to tipping under others. In places where groundwater can become saline like coastal areas, and much of Australia, a certain type of bistable mode behavior appears. When the water table is deep, a stable mode exists where healthy vegetation transpires large amounts of water, keeping the water table low. Conversely, if the saline water table rises high enough to intersect the root zone, the salt stresses the vegetation, severely reducing transpiration and creating a secondary, less desirable stable mode. This reduction in water uptake (a positive feedback) allows the water table to persist at a shallow, highly saline level, leading to the self-perpetuation of the dryland salinity regime. [Peterson 2012]

This perspective also sheds light on a long-standing debate in forest hydrology: do forests “use up” groundwater, or do they help sustain it? [Ellison 2012] On the one hand, trees extract water from the soil, which can reduce local groundwater levels, a straightforward demand-side effect. On the other hand, through precipitation recycling, forests generate rainfall that eventually recharges the aquifer, a supply-side effect. Traditional studies often focus on one side or the other, leading to apparently conflicting conclusions. But when we view the system through the lens of coupled modes and feedback loops, the picture becomes clearer. Cutting down trees might temporarily increase groundwater locally, but over longer timescales, reduced transpiration lowers rainfall, weakening recharge and potentially pushing the watershed into a new, drier mode with less total water. In other words, what looks like a linear, short-term gain can translate into a long-term loss once the full coupled system is considered. Mode thinking and complex adaptive systems provide a framework to reconcile these observations, showing that short-term changes in demand or supply are inseparable from the emergent behavior of the watershed as a whole.

Seen through this lens, the Amazon is not just a biome at risk of losing trees; it is a coupled hydro-biological system approaching a phase transition. The forest mode is defined by tight integration: deep groundwater feeding evergreen canopies, canopies maintaining atmospheric humidity, humidity sustaining rainfall, rainfall replenishing groundwater. The savanna mode, in contrast, is characterized by sparser tree cover and grasses that, while able to access groundwater, produce less transpiration overall. Moisture escapes to the atmosphere more slowly and over shorter periods, rainfall is less effectively recycled, and the system’s feedback loops are weaker. The difference between these modes is not only the amount of water available but the structure and strength of the feedbacks themselves: one configuration stabilizes, the other is more fragile. Because regime shifts depend on the architecture of feedback loops, not just averages of rainfall or temperature, the transition between modes can be swift, nonlinear, and difficult to reverse.

Traditionally, climatologists have often described the world’s biomes using the Köppen–Geiger classification, a framework in which temperature and rainfall largely determine vegetation type. In this view, climate dictates the biome: deserts are dry, rainforests are wet, and savannas sit somewhere in between. This is a form of climate determinism, treating ecosystems as passive responses to external climate variables. But the coupled groundwater–tree–rain dynamics we have described suggest a more nuanced picture. Biomes are not merely outcomes of average rainfall and temperature; they are part of a complex, dynamical system in which vegetation, groundwater, and climate co-evolve. Different modes can emerge from these interactions: a forest mode with strong rainfall recycling, and intermediate tree modes with groundwater and rain coupling. In other words, the Earth system itself generates multiple stable states, and which state emerges depends on the internal feedbacks and thresholds, not solely on the imposed climate averages. This perspective moves us beyond simple climate determinism to a framework where the biome–hydrology–atmosphere network actively shapes its own climate.

The logic extends further when we consider that the climate itself has multiple modes, self-reinforcing patterns of circulation that can persist for weeks, months, or even years. Phenomena such as El Niño, the Indian Ocean Dipole, Madden-Julien oscillation, and persistent jet stream blocking are examples of these modes. Each mode represents a stable or quasi-stable configuration of the atmosphere and ocean, much like forest and savanna are stable modes of the land–water system. These modes can shift abruptly, producing periods of intense rainfall, drought, or extreme temperature swings, sometimes called climate whiplash.

Recent research shows that vegetation and land processes are not just passive responses to these modes; they can actively influence them. Transpiration from trees affects regional and even global circulation patterns: it can alter the Hadley cell, which modulates jet stream behavior, potentially influencing blocking events that drive extreme weather. Land surface moisture and evapotranspiration can even affect the development and intensity of El Niño events [Lee 2021]. In other words, trees, soil, and groundwater are part of the chain that links local water use to modes of the entire climate system.

This chain of logic, from groundwater extraction to ecosystem feedbacks, to rainfall patterns, to atmospheric circulation modes, to climate extremes, is long and complex. Scientists are working on each link, but understanding how all the links interact remains a challenge. This is where the framework of complex systems, feedback loops, and regime shifts becomes useful: it provides a way to conceptualize the system’s architecture, highlight which loops stabilize or destabilize it, and identify where thresholds and new modes may emerge. Within this framework, groundwater extraction emerges as not just a local hydrological issue, but a potent climate driver capable of nudging the system into new modes, with consequences that may propagate far beyond the local landscape.

The Madden–Julian Oscillation (MJO), was discovered by observing repeating eastward-propagating waves of enhanced and suppressed tropical rainfall every 30 to 60 days. Scientists noticed that these oscillations could reinforce or dampen rainfall over forests and savannas depending on timing and local conditions. The key insight from the MJO, and from decades of studying atmospheric modes, is that emergent patterns often appear only when data are analyzed across space and time, revealing feedbacks and phase-like behavior that are invisible in short-term or localized observations. This is how we might find groundwater–climate coupling. Just as the MJO could not have been discovered by looking at a single weather station, the multiple modes of vegetation–groundwater–rainfall interaction may only emerge when we integrate long-term hydrological, vegetation, and atmospheric data. Detecting these coupled modes requires careful analysis across seasons, years, and landscapes, looking for correlations, lags, and feedback loops that indicate the system is self-organizing into distinct states.

A significant amount of groundwater can be brought up by trees to then transpire to create rain. Some estimates groundwater creates 10% of rain [Barlage 2021]. That rain can then flow back down into the aquifers. When there is bidirectionality there becomes a feedback loop. Feedback loops might then emerge modes. Currently groundwater-biome-climate coupling where the groundwater can also influence the climate, is an understudied research area with just a smattering of papers.

There is also a broader layer of coupling that also involves the ocean to add to the , atmosphere- land coupling. Traditionally, scientists have viewed the ocean as influencing the land, shifts in currents or the Atlantic Meridional Overturning Circulation (AMOC), for example, can drive changes in atmospheric circulation and rainfall over continents. (AMOC has two basic modes - if we shift out of our current mode it would lead to big climate shifts). However, recent research shows that land changes, such as widespread deforestation, can also affect ocean air [Lague], ocean temperatures and circulation. [Portman 2022].. This two-way interaction opens a whole new world of potential modes, where land/ecosystem, groundwater, ocean, and atmosphere feedbacks can produce emergent behavior that was previously unrecognized.

It is also worth stepping back and thinking about what drives climate and ecosystem changes in the first place. Scientists often speak of “climate drivers,” factors that push the Earth system in one direction or another. Carbon emissions are the most familiar: more greenhouse gases trap heat, alter temperatures, and shift weather patterns. But they are far from the only driver. Water matters too, and in ways that are just as fundamental. When we extract groundwater, we are not only changing aquifer levels; we are triggering a cascade of effects through soil, vegetation, rivers, and the atmosphere. Each loop feeds back into others, reshaping rainfall, surface water, and even the structure of biomes. Groundwater withdrawal may thus be a physical climate driver in its own right: it can flip landscapes from one mode to another, just as deforestation or greenhouse gas accumulation can. We need more research into this area.

There is also a whole other dimenasion. Biodiversity itself can act as a Complex Adaptive System (CAS). Trees, plants, and soil microbes do not just passively transpire water or store nutrients; they respond to local conditions, adapt their strategies, and interact with one another in ways that collectively shape larger-scale patterns. Through mechanisms like tuned transpiration, hydraulic lift, and shading, biodiversity can influence moisture, rainfall, and even regional or global climate in ways we are only beginning to quantify.

Complex Adaptive Systems describes systems composed of interacting agents that respond to each other and to the environment, producing emergent behavior that cannot be predicted from the behavior of individual components alone. So it is to be distinguished from the more basic complex systems where the components don’t make choices. A well-known example comes from economics: in a market, each buyer or seller makes choices based on prices, expectations, and competitors’ actions. The overall behavior of the market, trends, bubbles, or crashes, emerges from these local decisions, not from a single controller. Feedbacks can be stabilizing (negative) or amplifying (positive), and agents constantly adapt their strategies, producing new patterns over time. (See work by Brian Arthur and Doyne Farmer.)

Applying this analogy to forests and watersheds highlights how each tree, patch of soil, and aquifer segment acts as an adaptive agent. Trees make choices, when to transpire, when to conserve water, based on local conditions and interactions with neighboring plants. They can also choose when to bring up groundwater. Together, these choices shape emergent patterns of moisture, rainfall, and even regional climate. Giving trees this adaptive agency, puts a whole different factor into climate models and earth system models that can lead to them exhibiting quite different modes and regimes, and in their ability to regulate its regimes and regime shifts. The system is adaptive, self-organizing, and capable of both resilience and abrupt shifts, just like economic markets. This perspective could help formalize ideas proposed by hydrologists and ecologists: Hubert Savenije (2017, 2024) suggested that trees regulate water in watersheds for its own needs, so that the eco-watershed system is like an organism, while Makarieva and Gorshkov [2000] emphasized the role of living organisms in biotic regulation of the atmosphere. Complex adaptive systems provides a framework for understanding these insights: the coupled land–water–climate system is a network of interacting agents whose choices collectively produce stable modes, regulate the Earth system, and buffer against shocks.

Through this lens, the coupled land–water–climate system becomes not just a passive machine, but an adaptive, self-regulating network capable of buffering change, sustaining life, and guiding human interventions. Only by integrating multiple dynamical mechanisms, groundwater, vegetation, soil, atmosphere, and their feedbacks, can we begin to anticipate how new modes might emerge, and how seemingly local changes could ripple outward to shape global climate. Recognizing these multiple drivers, and understanding how they interact through complex feedbacks in a complex adaptive system, is essential if we hope to anticipate, and perhaps prevent, abrupt shifts in Earth’s coupled water–land–climate system.

Barlage, Michael, Fei Chen, Roy Rasmussen, Zhe Zhang, and Gonzalo Miguez‐Macho. “The importance of scale‐dependent groundwater processes in land‐atmosphere interactions over the central United States.” Geophysical Research Letters 48, no. 5 (2021): e2020GL092171

Ellison, David, Martyn N. Futter, and Kevin Bishop. “On the forest cover–water yield debate: from demand‐to supply‐side thinking.” Global change biology 18, no. 3 (2012): 806-820.

Gorshkov, Victor, Anastassia M. Makarieva, and Vadim V. Gorshkov. Biotic regulation of the environment: Key issues of global change. Springer Science & Business Media, 2000.

Ilstedt, U., Bargués Tobella, A., Bazié, H.R., Bayala, J., Verbeeten, E., Nyberg, G., Sanou, J., Benegas, L., Murdiyarso, D., Laudon, H. and Sheil, D., 2016. Intermediate tree cover can maximize groundwater recharge in the seasonally dry tropics. Scientific reports, 6(1), p.21930.

Lee, Ting-Hui, and Min-Hui Lo. “The role of El Niño in modulating the effects of deforestation in the Maritime Continent.” Environmental Research Letters 16, no. 5 (2021): 054056

Lovejoy, Thomas E., and Carlos Nobre. “Amazon tipping point.” Science advances 4, no. 2 (2018): eaat2340.

Peterson, T. J., A. W. Western, and R. M. Argent. “Analytical methods for ecosystem resilience: A hydrological investigation.” Water Resources Research 48, no. 10 (2012).

Portmann, R., Beyerle, U., Davin, E. et al. Global forestation and deforestation affect remote climate via adjusted atmosphere and ocean circulation. Nat Commun 13, 5569 (2022). https://doi.org/10.1038/s41467-022-33279-9

Savenije, Hubert HG. “The hydrological system as a living organism.” Proceedings of IAHS 385 (2024): 1-4.

Savenije, Hubert HG, and Markus Hrachowitz. “HESS Opinions Catchments as meta-organisms–a new blueprint for hydrological modelling.” Hydrology and Earth System Sciences 21, no. 2 (2017): 1107-1116.

Staal, Arie, Ingo Fetzer, Lan Wang-Erlandsson, Joyce HC Bosmans, Stefan C. Dekker, Egbert H. van Nes, Johan Rockström, and Obbe A. Tuinenburg. “Hysteresis of tropical forests in the 21st century.” Nature communications 11, no. 1 (2020): 4978.