China at an Ecohydrological Crossroads, Part III: New Perspectives

How ecosystem recovery dynamics can change long-term projections of China's water cycle

In Part I: Basic Notions, we discussed difficulties associated with empirical quantification of the major terms in the water budget, especially its invisible components: atmospheric moisture transport, evaporation, and transpiration. In Part II: Common Assumptions, we discussed how the reported decline in water availability, aka runoff, associated with re-greening was not an actually observed decline, but instead concerned a hypothetical variable calculated under the assumption that added vegetation did not influence atmospheric circulation. In this concluding post of the China at an Ecohydrological Crossroads mini-series, we consider what actually happened to China’s water cycle during the decades of re-greening and how the available data can be interpreted when ecosystem processes are taken into account.

Disclaimer: Water availability (WA), water yield (Y) and runoff (R) are used interchangeably in this mini-series, with their very subtle differences discussed in Part I: Basic Notions. Evapotranspiration is denoted as either ET or E.

Increasing Precipitation and Water Availability, Decreasing Transpiration?

Figure 5a from An et al. (2025), showing ERA5 reanalysis trends in precipitation P, evapotranspiration ET, and water availability WA = P - ET during the two decades of re-greening.

We can see that water availability increased, while evapotranspiration decreased, during the studied period. This is the opposite of what a reader might take from the abstract of An et al. (2025), where land use and land cover changes (LUCC) are said to have increased evapotranspiration and decreased water availability:

…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). …

Readers who do not examine how these results were obtained, and media reports that repeat the headline conclusion, may therefore miss the central point.

The direct ERA5 pattern is different: over the last two decades, water availability in China has been slightly increasing. Precipitation has been increasing as well, while evapotranspiration has been decreasing. How should we interpret this? And how can evapotranspiration decline over a re-greening area?

There are two possibilities.

The first is that the evapotranspiration trend itself is uncertain. As discussed in Parts I and II, evapotranspiration is not measured directly over large areas. It has to be inferred from flux-tower data, satellite information, land-surface models, and multiple parameterizations.

The most robust estimates usually come from watershed-scale mass balance: we measure precipitation, we measure streamflow, and we estimate evapotranspiration as the difference between them. But this method is not local. It depends on river-basin measurements, which remain uncertain in many regions. In China, consistent national-scale streamflow data have only recently become available (Wang et al., 2025).

This means that an evapotranspiration trend estimated as the difference between precipitation and runoff trends accumulates the uncertainties of both. It cannot be more robust than the underlying precipitation and runoff data.

It may therefore be that actual evapotranspiration in China has been increasing, together with rainfall and runoff. This would be similar to the wet-season pattern inferred for the Loess Plateau by Tian et al. (2022), using a different reanalysis product, MERRA-2.

A summary of discussed trends, Table 1 from https://arxiv.org/abs/2604.09510

The second possibility, which we discuss below, is related to ecosystem processes.

Ecosystem Recovery, Atmospheric Dynamics, and the Water Cycle

When an ecosystem begins to recover from a major disturbance, plant biomass increases. Transpiration increases as well, adding moisture to the atmosphere. If plants have enough water to sustain transpiration, then as vegetation recovers, the atmosphere above and around it can become moister. This is indeed what has been reported for China: re-greening has been associated with an increase in atmospheric moisture content (Jiang et al. 2013).

At the initial stage of recovery, we can therefore expect a positive correlation between transpiration and atmospheric moisture. Atmospheric moisture, in turn, affects the probability of precipitation, but in a non-linear way. When moisture content is low, and the atmosphere is still dry and far from the dew point, adding more moisture may have little effect on precipitation. This is the dry regime: atmospheric moisture and transpiration increase markedly, while precipitation changes more slowly, if at all.

In this regime, shown in the picture below to the left of the red dot, water availability, or runoff, indeed decreases as transpiration increases:

ΔY = ΔP − ΔE < 0, because ΔP < ΔE.

Conceptual relationship among atmospheric moisture content, stage of ecological succession and changes in hydrological variables—precipitation P, evapotranspiration E and water yield Y=P-E —relative to their initial values. The red dot marks the transitional point at which the water-yield trend becomes positive. Fig. 2 from https://arxiv.org/abs/2604.09510

If the ecosystem is allowed to recover further, the atmosphere may eventually become moist enough for atmospheric dynamics to change, condensation to intensify, and precipitation to increase rapidly. At the same time, transpiration does not increase as readily in an already moist atmosphere. This is the wet regime: water availability begins to rise as evapotranspiration grows slowly and precipitation grows rapidly:

ΔY = ΔP − ΔE > 0, because now ΔP > ΔE.

If we recall the three-stage scheme discussed in Part II, in the section “Re-Greening and Moisture Convergence,” then early succession in the figure above corresponds to panels (a, d), intermediate succession to panels (b, e), and mature ecosystem to panels (c, f).

If a region alternates between dry and wet regimes during the year, recovering vegetation could affect the water budget in different ways in different seasons. During the dry season, better soil cover could reduce direct evaporation from previously bare soil, thus increasing water availability. During the wet season, vegetation could increase transpiration, when additional moisture input to the atmosphere is more likely to enhance moisture convergence and precipitation, thus raising water yield, as in the Loess Plateau example shown in the table above.

Depending on the relative magnitudes of these seasonal effects, annual mean water availability could increase even if annual mean evapotranspiration declines. This could happen, for example, if reduced dry-season evaporation from bare soil outweighs a smaller wet-season increase in transpiration.

This is the second possible explanation for declining evapotranspiration in a re-greening region. The downward trend may reflect a reduction of uncontrolled evaporation during the dry season, due to better soil cover, combined with a smaller increase in transpiration during the wet season, when most moisture convergence and precipitation occur.

This is the essence of ecological succession as a water-cycle process: reducing unnecessary moisture losses while making more effective moisture “investments” through transpiration, at the time when these investments are most likely to support atmospheric moisture import and precipitation.

This dual hydrological behavior during ecological restoration parallels the ecological concept of a landscape trap (Lindenmayer et al. 2022). Mature, relatively undisturbed forest ecosystems tend to be hydrologically stable and resistant to droughts and fires (Xiao et al. 2023; Wolf et al. 2023). Logging or other disturbances can shift such systems toward early successional states that are less hydrologically competent. If disturbance exceeds a threshold, the system may enter a trajectory of drying and burning from which recovery becomes increasingly difficult. The inability of early successional systems to stabilize a moist regime is consistent with the characteristics of the dry hydrological stage shown in the above figure.

In China, contrasting assessments of the relationship between water yield and ecological restoration may reflect the complexity of these ecological processes. After two to three decades of re-greening, large areas remain in relatively early successional stages, often dominated by naturally recovering grasslands rather than planted tree stands (Yu et al. 2023).

In such conditions, analyses confined to moisture recycling alone may suggest that further ecological restoration threatens local streamflow. However, if restoration is viewed as a dynamic progression between hydrological regimes, short-term reductions in water availability may be followed by a trend reversal, with water availability beginning to grow.

If we halt the self-enhancement of vegetation at the dry-regime stage, out of concern about declining water availability, we may miss the larger outcome: full ecosystem recovery and a reinvigorated water cycle.

Conversely, ecosystem destruction should lead to a decline of the water cycle. Recent global forest-change assessments indicate that the dominant hotspots of forest loss and disturbance during 2000–2020 were tropical forests in the Amazon Basin, the Congo Basin, and Southeast Asia, together with extensive tree-cover loss and disturbance in boreal regions of Canada and Russia. In contrast, several temperate and subtropical regions, notably parts of Asia, have exhibited net forest and vegetation gain over the same period, driven in particular by large-scale re-greening in China and India.

Consistent with this pattern, global maps of water-yield trends for 2000–2020 highlight tropical regions, Canada, and Siberia as areas with a tendency toward declining water yield, while India and China stand out as among the few large regions where water yield has increased, at least at regional scales.

Fig. 1B from Zhang et al. 2023. Global trends in water availability in 2001-2020.

Summary

In this mini-series, I wanted to describe some caveats that a general reader may miss in the scientific literature on the vegetation-water nexus. But I also wanted to present an ecosystem perspective on water-cycle recovery, and to highlight the complexity of these processes.

The caveats. If you are a water-cycle practitioner, ecosystem restorer, or forest protector looking to the scientific literature for guidance, it is useful to pause whenever you encounter a statement such as: “[deforestation, afforestation, replanting] caused [runoff, precipitation, cloudiness, evapotranspiration] to increase or decrease.” The first question is simple: did the variable itself actually increase or decrease in the studied region over the studied period?

If the observed trend has the opposite sign, or no clear trend at all, then the reported “vegetation impact” is not a directly observed change. It is a reconstructed effect, obtained by separating vegetation change from the broader climatic background using a particular set of assumptions.

This does not automatically make the result wrong. But it means that the conclusion depends on the method. The assumptions used to isolate the vegetation effect may be appropriate for the region studied, or they may not. They may also miss important processes, such as vegetation effects on atmospheric circulation. If the reported vegetation impact is reconstructed rather than directly observed, the next step is to examine the method: how was vegetation change separated from climate variability, and which effects of vegetation were included or excluded?

Ecosystem perspective. Water is cycled on land through complex and interrelated processes. We still know relatively little about many of them, especially over long timescales and at large spatial scales. This is why the best strategy is to make nature our ally and follow the natural process of ecological succession as closely as possible, allowing the water cycle to recover through ecosystem self-restoration. Where recovery is active and human-assisted, the same principle applies: mimicking natural processes is often reported to be the most productive way forward (Andrade et al., 2020; Pasini et al., 2021).

The high hydrological competence of minimally disturbed ecosystems, together with the slow nature of ecological succession, provides a strong rationale for preserving natural vegetation wherever it remains. Within China, there is increasing recognition that tree cover constitutes a foundation of environmental stability. A consistent extension of this principle is to support the protection of natural forests beyond national boundaries, particularly in Eurasia and the tropics, where deforestation exerts a disproportionate influence on atmospheric circulation and climate stability. In this context, the concept of Ecological Civilization advanced by China can be understood not only as a national development framework, but as a basis for coordinated global action to preserve one of the Earth’s most effective climate-regulating systems: natural forests.

Photo by Alexei Aleinikov

Related reading: