China at an Ecohydrological Crossroads, Part I: Basic Notions

What We Do and Do Not Know About Re-Greening Impacts on the Water Cycle

What happens to the water cycle when vegetation is added or removed? Are different components of the cycle affected in different ways? And how do these effects scale up? These are complex questions to which no universal answers exist, yet we depend on them critically. We need both water and vegetation to survive.

China, which has added vegetation to its land on a massive scale in recent decades, provides a unique setting in which to study these questions.

Image source: NASA

In October 2025, a very interesting study was published in Earth’s Future, a journal of the American Geophysical Union: An et al. (2025) “Land Cover Changes Redistribute China’s Water Resources Through Atmospheric Moisture Recycling”.

This study concluded that the vegetation added in China increased rainfall and evapotranspiration, but reduced the country’s overall water availability. To quote the abstract:

To achieve sustainable development goals such as mitigating climate change and ensuring food security, China has undergone rapid land use/cover changes (LUCC), including afforestation, grassland restoration, and cropland redistribution, which have substantially transformed the terrestrial surface and affected hydrological conditions and water resources management. However, the hydrological impacts of these changes, particularly through atmospheric moisture recycling processes, remain insufficiently understood. This study quantified the hydrological impacts of LUCC in China from 2001 to 2020 using high-resolution data sets and an atmospheric moisture tracking model. Our findings revealed that LUCC had led to increased evapotranspiration (ET; 1.71 mm/yr) and precipitation (P; 1.24 mm/yr), while decreasing water availability (WA) (P − ET; −0.46 mm/yr). <…>

The study was, unsurprisingly, picked up by many news outlets, including Popular Mechanics, Live Science and Earth.com, and seems to have resonated widely. In late 2025 and early 2026, several colleagues from around the world asked me to comment on it. Since the study is directly relevant to the biotic pump concept — which predicts an increase, not a reduction, in water availability on land — we decided to look into the issue more closely.

An unexpected conclusion, which I would like to share in this mini-series of posts, is that the study’s finding of reduced water availability due to added vegetation was effectively predefined by its methodology. In other words, the assessment was framed in such a way that reduced water availability emerged independently of the data, as a built-in consequence of the underlying assumptions.

This unusual situation is a good illustration of just how complex vegetation–hydrology feedbacks really are and, I hope, offers readers an engaging way into the subject, even if it is somewhat dense. To me, there is no better entry into a complex field than through a controversy. While the textbook route can leave one with passive knowledge and the illusion of understanding, analyzing a live controversy keeps the mind alert and prepares it for the task of creating new knowledge.

So let us follow the path of water in China.

The Water Cycle Components and Terminology

We begin with the fundamentals. To understand the water cycle, it helps to look at it from two complementary perspectives: from the ground, where water is received and stored, and from the atmosphere, which delivers it and takes it away.

From the ground perspective, water arrives as precipitation, P. Once it reaches the land, it can follow three paths. It can return to the atmosphere through evaporation and plant transpiration, together called evapotranspiration, E. Second, it can run off under gravity as either surface flow or subsurface runoff or streamflow, R, eventually making its way to the ocean.

Or it can remain on land, accumulating below ground as soil moisture and groundwater. This third pathway is different from the first two. Evapotranspiration and runoff are flows: they move water from one reservoir to another. Storage change is not a flow. It describes what happens when water stays in place and the local moisture store grows or shrinks.

To make this distinction explicit, we denote the storage term in the water budget as dS/dt, where S is the local moisture store. This notation means the rate at which storage changes with time: a small change in storage, dS, over a small interval of time, dt.

These quantities are tied together by the most basic physical law: conservation of matter. Water does not appear from nowhere, and it does not vanish into nowhere. Whatever falls on land as precipitation must do one of three things: return to the atmosphere, leave as runoff, or remain on land by changing storage. This gives the basic water-budget equation:

P = E + R + dS/dt.

Every drop of precipitation must follow one of these paths. It can rise into the air, flow to the sea, or stay and build the hidden reserves of moisture underground.

Every drop that falls on land must choose its fate: rise, run off, or stay.

Water Yield and Water Availability

Let us now clear up a potential source of confusion. In hydrology, runoff R is sometimes referred to as water yield, as in this old hydrological dictionary:

Screenshot from “General Introduction and Hydrologic Definitions” by W. B. Langbein and K. T. Iseri (1960), Manual of Hydrology: Part 1. General Surface-Water Techniques.

The point of this term is to capture all forms of runoff together: surface runoff, subsurface runoff, and streamflow. The dictionary states that water yield is equal to precipitation minus evapotranspiration, R = P - E. But from the water-budget equation above, we can see that this is true only in a steady state, when local water storage does not change and dS/dt = 0.

Here the difference between flows and storage change becomes especially clear. A steady state means that, over the period considered, the local moisture store remains unchanged. Reservoirs may already be full, with no capacity to hold additional incoming moisture. Or rainfall may fail to infiltrate the soil and instead leave entirely as surface runoff, while the soil itself remains persistently dry. In such cases, storage does not change: dS/dt = 0.

A similar simplification often applies over sufficiently long timescales. Since local storage is finite, it cannot continue to influence the water budget indefinitely. Over long enough periods, storage changes usually become small relative to the fluxes, and one can approximately set dS/dt = 0.

So, in a steady state, the transient storage term drops out, while the flows remain. The budget equation becomes simply

P = E + R.

Only in this case is runoff, or water yield, equal to precipitation minus evapotranspiration.

In the general case, however, not all precipitation that is not returned to the atmosphere becomes runoff. Some of it may remain on land, replenishing soil moisture and groundwater, or offsetting an ongoing depletion of these reserves.

Outside steady state, P - E is therefore better understood not as runoff alone, but as water availability in a broader sense. It is the water left on land after evapotranspiration, to be divided between runoff and storage change. This is the definition adopted by An et al. (2025):

water availability, defined as P - E, includes all runoff plus the rate of change in water storage.

Atmospheric Perspective

Let us now look at the same water budget from the atmosphere. Precipitation, which delivers moisture to the ground, removes it from the air. Evapotranspiration does the reverse: it draws water from the ground and returns it to the atmosphere. There is also a third pathway by which the atmosphere gains moisture locally: transport by the winds.

The winds carry only a small amount of moisture, typically less than 1% of air mass. If air simply passes through horizontally, the net import of moisture, also known as atmospheric moisture convergence, is zero. But if that moist air rises and cools, its water vapor condenses and precipitates. The air leaving the region is then drier than the air entering it, and the atmosphere has delivered moisture to the surface.

In this scheme, we view the water cycle from the perspective of the atmosphere. Moist air enters and leaves the region (light blue arrows), while drier air flows out (white arrows): whatever air enters must, in the end, also leave. Evapotranspiration (curved light-green arrows) adds moisture to the atmosphere, while precipitation (the downward blue arrow) removes it.

The local budget of atmospheric moisture M can be written as

dM/dt = E + Fᵢₙ − Fₒᵤₜ − P.

Here Fᵢₙ is the incoming moisture flux and Fₒᵤₜ is the outgoing moisture flux. Their difference,

C = Fᵢₙ − Fₒᵤₜ,

is called atmospheric moisture convergence.

A key point is that the atmosphere can hold only a limited amount of water. At terrestrial temperatures, water vapor condenses readily, so atmospheric moisture storage is necessarily small. Even over tropical oceans, the total atmospheric moisture content — consisting mostly of water vapor, with only minor contributions from cloud water and precipitation water — is only about 40 mm. That is the depth of water that would form if all atmospheric moisture above a given surface area were condensed and brought to the ground.

This is very little. With typical tropical precipitation of about 4–5 mm/day, the entire atmospheric moisture store turns over in just a few days. So when we average over periods longer than about a week, the storage term dM/dt becomes negligible compared with the fluxes themselves.

In that case, the atmospheric water budget simplifies to

P = E + C,

which means that, from the atmospheric point of view, evapotranspiration and moisture convergence are the sources of moisture, while precipitation is the sink.

From the ground perspective, conversely, precipitation supplies moisture to the ground, while evapotranspiration and runoff remove it:

P = E + R.

Here the link between land and atmosphere becomes clear: whatever moisture the atmosphere brings to a location ultimately leaves it as runoff:

C = P - E = R.

Outside steady state, the balance acquires an additional term. Atmospheric moisture convergence then equals water availability:

the moisture brought in by the atmosphere either leaves as runoff or stays on land by replenishing local storage.

The Big Measurement Problem: The Visible and the Invisible

By now we have assembled quite a list of key variables, all linked to one another and all potentially altered when vegetation is added to the landscape: precipitation, runoff in its various forms, atmospheric moisture convergence, evaporation and transpiration, and local water storage in soil moisture and groundwater. Atmospheric moisture itself is tiny by comparison and, on timescales of several days or more, remains close to steady state.

What is crucial to keep in mind is that all of these quantities are measured in very different ways. The visible parts of the water cycle, such as precipitation and streamflow, are the easiest to measure. In China, one of the earliest references to precipitation measurement appears in Qin Jiushao’s Mathematical Treatise in Nine Sections, published in 1247. According to Strangeways (2010),

The [ancient Chinese] raingauges were conical or barrel-shaped, one being installed at each provincial and district capital. In addition, the book also discusses problems with large snow gauges made from bamboo which were sited in mountain passes and uplands, probably the first ever reference to snow measurement. Qin Jiushao also discusses how point measurements of rainfall were converted to areal averages (Needham 1959). Biswas (1970) suggests that the gauges were necessary because the flooding of rivers and canals has always been a problem in China, but if this was the case the hydrological cycle must have been understood and this seems improbable in the thirteenth century; more likely the purpose was again agricultural.

Below is a modern precipitation gauge in Yakutia. Like its older counterparts, it measures precipitation at a single location.

In contrast, runoff is conventionally measured as the streamflow draining a given area. It therefore represents an integrated, rather than local, property of the drainage basin. For example, the Óbidos gauge station on the Amazon River measures streamflow from a drainage area of nearly five million square kilometers. This is done in a relatively straightforward way, by measuring water level and flow velocity. However, such measurements do not account for groundwater outflow from the basin. That component must be treated separately and can only be included through additional assumptions.

To compare such a runoff estimate with precipitation, we would need multiple rain gauges spread across the basin and capable of capturing the rainfall variability that matters for the area average. In modern times, we can also rely on satellite estimates of precipitation, but these, too, are ultimately calibrated against ground-based measurements.

Soil moisture and deeper groundwater present measurement problems of their own. To mention just one, measuring soil moisture along a deep profile corresponding to the rooting zone of large trees may require drilling or boring holes and installing sensors in the soil. But this very act disturbs the local soil and biota, and may itself alter soil moisture.

A sensor for assessing soil moisture. An electromagnetic pulse is sent along metal rods, and its propagation time depends on the water content of the surrounding soil.

The Invisible

But the real difficulties begin when we turn to the invisible parts of the cycle: atmospheric moisture convergence and evapotranspiration.

To estimate atmospheric moisture convergence, we would need to track, with very high accuracy, the velocity and water vapor content of all air masses entering and leaving a given area. This is already a demanding task. Water vapor is only a minor constituent of the atmosphere, and its distribution is highly uneven in space and time. As a result, even when air motion itself is represented reasonably well, the transport of water vapor may still be estimated poorly.

A natural check is to compare atmospheric estimates of moisture convergence with runoff measured independently at the ground (remember, C = R). This check is not always passed. For example, global climate models may describe large-scale circulation reasonably well, yet when their atmospheric moisture convergence is compared with observed runoff in major river basins, substantial discrepancies remain. For the Amazon, the mismatch can reach about 50%.

Data from Hagemann et al. 2011 showing discrepancies between atmospheric moisture convergence and river runoff in some river basins

And yet atmospheric moisture convergence is not the hardest quantity to estimate. After all, it still depends on variables we know, in principle, how to measure: air motion and water vapor concentration.

Evaporation and transpiration are even more elusive.

The reason is simple. They add water vapor to the atmosphere very slowly compared with how quickly winds mix that vapor over space. Consider a small vegetable garden of 100 square meters. Let cumulative evaporation and transpiration be E = 2 mm/day. This means that the garden contributes 200 kg of water vapor to the atmosphere per day, because a water layer 1 mm deep spread over 1 square meter weighs 1 kg.

Now let the mean wind speed be u = 2 m/s, and let the column water vapor content be M = 30 mm, that is, 30 kg/m². For a square garden of area 100 m², the side length is L = 10 m. Under these conditions, the gross atmospheric moisture flow across one side of the garden in one day is M u L = about 52 million kilograms, or 52 thousand tons of water.

The garden contributes 200 kg of water vapor per day, while the moving atmosphere carries 52 thousand tons across it. The local evaporative signal is thus submerged in a vastly larger moving background.

And this mixing interferes in important ways. Suppose, for example, that we want to infer transpiration intensity from the temperature contrast between a transpiring patch, such as the small fir tree in the picture below, and a nearby dry surface. Under clear and windless conditions, such a contrast can become very large, reaching several tens of degrees Celsius. But in cloudy or windy weather, it is smoothed by atmospheric mixing. How strongly it is smoothed depends not only on wind speed, but also on season, surface roughness, and other factors that are rarely known with sufficient precision.

Local cooling from plant transpiration. With incoming solar radiation of about 1 kW m⁻², dry area on the deforested plot (left) has temperature of 55.3°C. Young transpiring trees (right) lower the surface temperature by almost 30°C. Distance between the two spots is 1 m. Measurements and photo credit Jan Pokorný.

For this reason, local estimates of evaporation and transpiration can be highly uncertain and heavily dependent on parameterization. I often cite the study by Teuling (2018), who pointed out a striking example. When transpiration is estimated from flux towers, that is, from local measurements of water vapor concentrations and turbulent transport (the atmospheric perspective), one may conclude that grasslands transpire more than forests. But when transpiration is estimated at larger scales from the ground-based water budget, as precipitation minus runoff, the conclusion is reversed: forests transpire more than grasslands.

This gives some sense of the conceptual and practical difficulty of measuring this invisible, yet fundamentally important, part of the water cycle.

What We Have Discussed in Part I

• On the ground, precipitation is partitioned between evapotranspiration, runoff, and local moisture storage:

  P = E + R + dS/dt.

• In the atmosphere, precipitation is partitioned between evapotranspiration and atmospheric moisture convergence:
  P = E + C.

• In a steady-state,
  R = C.

• All of these variables can, with varying difficulty, be measured independently. But when we combine them as mass conservation requires, gross mismatches often appear.

• Evaporation and, especially, transpiration are the hardest to measure, because their signal must be extracted against strong atmospheric mixing. In practice, the most reliable estimate comes from the ground-based water budget: from known precipitation and runoff.

In the next part, we will look at how plant impacts on the water cycle can be studied under these conditions — and how built-in assumptions can drastically alter the outcome of the analysis. Please stay tuned. In the meantime, your comments are very welcome.

Also, please note that tomorrow, Saturday, 18 April 2026, from 10:00 a.m. to 12:00 noon ET, I will take part in the online mini-conference “Protect and Restore Ecosystems to Cool the Climate” , organized by our great friends at Biodiversity for a Livable Climate. The event is free and will include a discussion. You are very welcome to register and join.