Air that has crossed extensive tropical forest produces roughly twice the rainfall of air that has crossed sparse vegetation. That satellite finding is widely cited. The argument is about why.
Most earth system models attribute the difference to evapotranspiration and moisture recycling, and they represent those processes reasonably well. A smaller, contested body of work argues that forests also do something no standard model currently represents: generate the pressure gradients that pull ocean moisture inland in the first place. If that is right, every restoration scenario you run is missing a mechanism. If it is wrong, the most convincing place to show it is inside a full earth system model.
That absence is a choice made during earth system model development, not a finding. The Biotic Regulation Principle starts from the premise that living systems help maintain the conditions they depend on, including through forest-driven moisture transport. A mechanism that is never coded can never be tested, and one that is never tested can never be ruled out.
I am putting this to modellers rather than to policymakers for a simple reason: policy can only act on what your models are able to represent. If you build land surface schemes, hydrology modules, dynamical cores or coupling infrastructure, or you run the ensembles and review the parameterisations, this is your territory. You do not need to accept the principle to find this useful. You need only be curious about what your model structurally cannot say.
In Brief
- Earth system models represent forest effects on rainfall through evapotranspiration and moisture recycling, and that physics runs in both directions.
- It makes three predictions that ordinary recycling does not, and two of them can be checked against a control run you already have.
- They do not represent the proposed biotic pump, in which condensation over forest draws ocean moisture inland. The hypothesis is contested.
The Baseline Problem: One Mechanism Is Missing in Both Directions
Modern land surface schemes are a long way from the bucket models of the 1960s. Multi-layer soil hydrology, prognostic vegetation and explicit canopy water budgets are standard in many configurations. Moisture recycling is represented too, and it matters. Van der Ent and colleagues estimated that around 40 per cent of precipitation over land originates from land evaporation. The Río de la Plata basin alone depends on Amazon evaporation for about 70 per cent of its water resources.
CMIP6’s Land Use Model Intercomparison Project ran an idealised global deforestation experiment across multiple models, and the tropical drying it produces is a genuine mechanistic result of decades of earth system model development. That physics runs in reverse, so models are not limited to degradation. But Spracklen and colleagues noted that the size of the modelled deforestation effect is model and resolution dependent. Their own observation-based estimate for an idealised Amazon deforestation scenario was a 12 to 21 per cent fall in seasonal rainfall.
The biotic pump hypothesis proposes an additional mechanism: condensation over continuously evaporating forest lowers pressure and draws ocean air inland. Makarieva and Gorshkov’s original analysis used International Geosphere-Biosphere Programme transects to argue that over non-forested land, precipitation decays exponentially within a few hundred kilometres of the coast, while over intact forest it does not. In most current configurations that mechanism is absent, and absent symmetrically. A deforestation run cannot lose it, and a restoration run cannot gain it.
If that gap exists in your model, we would like to test it properly rather than argue about it. Get in touch today.
Reframing Earth System Model Development Around Regulation, Not Only Response
The Biotic Regulation Principle treats the biosphere as part of the control system, and that reframing changes what earth system model development is for. In a response-led architecture, biology sits largely downstream of physics: circulation delivers moisture, plants return a share of it, and the feedback to the atmosphere runs mainly through surface energy and water fluxes. In a regulation architecture, biology also helps set the pressure field that drives moisture transport in the first place.
The debate should be stated plainly. The biotic pump has been challenged on its physical foundations: Meesters, Dolman and Bruijnzeel argued in a published comment that the evaporative force it relies on is not supported by basic physical principles. Its proponents have since developed the theory and responded to critics, but the disagreement has not been settled to the satisfaction of the wider field.
My own position is simple. I think this mechanism deserves a test rather than a verdict, and I would rather see it falsified in a well-run experiment than defended in a press release. A mechanism that is debated but never implemented stays debated indefinitely. A model that cannot represent a proposed process has not rejected it. It has simply never been asked.
Three Predictions Earth System Model Development Can Test
If I were in your seat, this is the section I would read first. A hypothesis earns a place in the code when it predicts something the existing physics does not. The biotic pump literature offers at least three such predictions, each with a diagnostic you can run.
| What to look at | Usual expectation | Biotic pump prediction | Diagnostic you can run |
| Rainfall with distance inland | Recycling tops up air over forest and slows the decline, but ocean moisture import is set by the large-scale circulation. | Over intact forest, precipitation holds roughly constant inland. Over non-forest, it decays exponentially within a few hundred kilometres. | Plot precipitation against distance from the coast along forest and non-forest transects in your control run, and compare with observations. |
| Moisture convergence response to extra transpiration | More evapotranspiration is broadly expected to reduce atmospheric moisture convergence. | The sign depends on humidity. In a wet atmosphere, extra transpiration increases moisture import and runoff. | Perturb transpiration across a humidity gradient and check whether the convergence response changes sign. |
| Surface pressure on rainy versus rainless days | Rainy days coincide with lower surface pressure, as air ascends. | At central Amazon forest stations, rainy days averaged 0.5 to 1 hPa higher pressure. At non-forest stations in Brazil, 1 to 2 hPa lower. | Composite surface pressure on rainy versus rainless days over forest and non-forest cells, by season, in your control run. |
The second prediction comes from Makarieva and colleagues’ 2023 paper in Global Change Biology. It argues that in a sufficiently wet atmosphere extra transpiration enhances moisture import and water yield, while in drier conditions it does the opposite. They point to reforestation in China as consistent with this, with water yield falling in some drier reforested areas and rising in wetter ones. The third comes from their analysis of long-term Brazilian station records, published as a preprint, which they interpret as a signature of forest-generated low pressure systems.
None of these is decisive on its own, and the observational analyses behind them have their own critics. But notice what they give you. The first and third need no new code at all, because they can be checked against your existing control run this month. A model that reproduces those patterns without the mechanism would be an important result in its own right. A model that cannot reproduce them has found the place to start.
Feedback Loop Modelling: Where Regulation Would Enter the Equations
Feedback loop modelling for biotic regulation happens in the coupling layer rather than in a new plant functional type. Three intervention points matter most.
- The moisture source term. Evaporation from continuous vegetated surfaces would need to influence low-level pressure gradients, not only latent heat flux totals.
- Horizontal moisture convergence. Condensation-driven pressure lowering needs representation in the dynamical core, or in a parameterised proxy that can be switched on and off for attribution.
- Memory. The seasonal persistence of soil, root and canopy water stores decides whether any such signal can carry across a dry season, so those timescales need checking against observations.
Each of these is testable in isolation. That is the point. Isolated switches give you a clean attribution chain instead of an unfalsifiable narrative about a living planet.
Step by Step: Testing a Restoration Scenario in Your Model
Beyond the diagnostics above, a full test needs one deliberate cycle of earth system model development, structured so the result is attributable either way.
- Audit your climate modelling assumptions in writing. List every place where vegetation state affects the atmosphere, and every place where it cannot. The gaps in that list are your experimental design.
- Run a diagnostic baseline. Force your current configuration with a prescribed large-increase vegetation scenario at continental scale, and record the precipitation response. Separate how much comes from evapotranspiration and recycling from how much comes from albedo and roughness. While you are there, close the basin water budget: over long periods, moisture convergence into a basin should match river discharge out of it.
- Implement one mechanism, not three. Add a single biotic regulation pathway behind a runtime switch, most usefully the link between surface evaporation and low-level pressure gradient.
- Compare against observational constraint. Test the resulting moisture transport against inland rainfall transects, trajectory-based rainfall observations of the kind Spracklen and colleagues used, and the three diagnostics above.
- Quantify the divergence. Report where the regulation-enabled run separates from your control, by region, season and lead time. Divergence is the useful output, not agreement.
- Publish the negative results too. If the mechanism fails against observations in your configuration, that finding is more valuable to this field than another untested assumption.
What to Watch Out For
Most of the risk in this kind of earth system model development is methodological rather than physical.
- Treating CMIP comparability as a reason not to experiment. Protocol runs and regenerative system simulation can live in separate branches of the same model.
- Tuning the new pathway to reproduce the present-day climate. A degraded baseline will penalise any mechanism that only operates at high vegetation continuity.
- Assuming resolution alone settles it. Finer grids resolve more of the circulation, but whether a condensation-driven pressure effect emerges also depends on how your dynamical core treats the mass removed by precipitation.
- Conflating restoration with afforestation accounting. Carbon stock is not the variable of interest here. Moisture transport is.
- Letting unfamiliarity stand in for a result. Scepticism about new physics is healthy, and the best answer to it is a test, whichever way it comes out.
Quick Checklist
- Write down the five assumptions that determine whether your model can represent biosphere recovery beyond albedo, roughness and evapotranspiration.
- Plot precipitation against distance inland along forest and non-forest transects in your existing control run.
- Composite surface pressure on rainy versus rainless days over forest and non-forest cells.
- Check how your dynamical core handles the mass sink from precipitation, and whether it could accept a condensation-driven pressure gradient term.
- Check the memory timescales in your soil and canopy water stores against seasonal signals.
- Register your interest in a shared validation protocol with other modelling groups.
FAQ
Do earth system models include the biotic pump?
Not in standard configurations, as far as we are aware. Models represent vegetation’s effects on rainfall through albedo, roughness, evapotranspiration and moisture recycling, in both directions, and CMIP6 land use experiments quantify those effects. The proposed biotic pump, in which condensation over forest actively draws ocean moisture inland, is not a standard process. The gap is one mechanism, not one direction.
What is the Biotic Regulation Principle in plain terms?
It is the proposition that living systems actively help maintain the environmental conditions they require, rather than only adapting to whatever physics delivers. In hydrological terms, it means continuous natural forest helps generate and sustain the atmospheric flows that bring rainfall inland. It is contested, but it is a physical claim, and therefore testable in a model.
How do I test the biotic pump in my model without rewriting the dynamical core?
Start with the diagnostics that need no new code: the rainfall-distance profile and the rainy versus rainless day pressure composite, both run on your existing control. Then add a parameterised proxy behind a runtime switch, linking surface evaporation over continuous vegetation to low-level pressure tendency. Core modification is a later step, justified only if the proxy shows a meaningful divergence from your control run.
Is the biotic pump accepted science?
Not yet. Makarieva and Gorshkov set it out in 2007, it has been challenged on its physical foundations, and it remains debated. Observational work such as Spracklen and colleagues’ shows a strong link between forest exposure and downwind rainfall, but does not by itself settle which mechanism is responsible. The theory’s proponents have begun setting out how global climate models could help assess circulation changes from restoration, and independent groups with no stake in the outcome are what the question now needs.
Conclusion
Every model carries assumptions about the future, set by the physics available when its land surface scheme was written. The question here is not whether your model is wrong. It is whether earth system model development has ever given this one proposed mechanism a fair hearing inside the equations.
You can leave that question open and keep producing projections that cannot, by construction, include it. Or you can start this month with two diagnostics on a control run you already have, then spend one development cycle finding out what your model says when the mechanism is physically representable. That is my whole ask. It carries scientific risk, and it might falsify the mechanism entirely. That is precisely why it is worth doing, and why this field needs modellers rather than advocates.
Bring your configuration, your scepticism and your diagnostics and let’s discuss today.
Sources
Makarieva, A. M. and Gorshkov, V. G. (2007) Biotic pump of atmospheric moisture as driver of the hydrological cycle on land. Hydrology and Earth System Sciences, 11, 1013-1033.
Meesters, A. G. C. A., Dolman, A. J. and Bruijnzeel, L. A. (2009) Comment on ‘Biotic pump of atmospheric moisture as driver of the hydrological cycle on land’. Hydrology and Earth System Sciences, 13, 1299.
van der Ent, R. J., Savenije, H. H. G., Schaefli, B. and Steele-Dunne, S. C. (2010) Origin and fate of atmospheric moisture over continents. Water Resources Research, 46, W09525.
Spracklen, D. V., Arnold, S. R. and Taylor, C. M. (2012) Observations of increased tropical rainfall preceded by air passage over forests. Nature, 489, 282-285.
Makarieva, A. M. et al. (2013) Why does air passage over forest yield more rain? Examining the coupling between rainfall, pressure and atmospheric moisture content. arXiv preprint 1301.3083.
Boysen, L. R. et al. (2020) Global climate response to idealized deforestation in CMIP6 models. Biogeosciences, 17, 5615-5638.
Makarieva, A. M. et al. (2023) The role of ecosystem transpiration in creating alternate moisture regimes by influencing atmospheric moisture convergence. Global Change Biology, 29, 2536-2556.
Makarieva, A. M. et al. (2025) Assessing changes in atmospheric circulation due to ecohydrological restoration: how can global climate models help? Frontiers in Environmental Science, 13, 1516747.




