China has transformed millions of hectares of degraded land into forests and grasslands. After 2015, certain ecosystem services fluctuated under the effect of climatic factors and land use, while drought episodes affected some of the gains achieved. The Three-North Shelterbelt program, the largest ecological restoration effort ever attempted, faces limits that its designers had not anticipated: the adaptation of ecosystems to the pace of climate change.
The essentials
- China has planted trees on 32 million hectares in the country’s north since 1978, with real local gains documented by satellite remote sensing through 2013, then fluctuations after 2015.
- On the Loess Plateau, vegetation productivity increased measurably according to remote sensing studies, a concrete result after decades of massive erosion.
- Drought episodes affected certain regions of northern China between 2020 and 2023, causing significant mortality in monospecific plantations and revealing their fragility in the face of climate shocks.
- Kazakhstan is conducting its own steppe restoration experiments, but over still marginal areas—5 to 10% of targeted degraded zones.
- As warming accelerates, large-scale ecosystem restoration faces challenges whose current models and satellite surveys do not yet allow us to determine the scope.
The Loess Plateau, laboratory of a visible transformation
Forty years ago, the Loess Plateau resembled an open wound. Centuries of overgrazing and deforestation had devastated 640,000 km² in northern China, leaving ravines, barren lands, and erosion that carried billions of tons of sediment each year toward the Yellow River. The World Bank financed a vast campaign of terracing, reforestation, and grazing bans beginning in the 1990s. The results were striking.
Indicators of vegetation recovery and erosion reduction were observed according to remote sensing studies. Rural communities that lived in precarity regained lands capable of producing. Studies published in Nature between 2021 and 2024 confirmed these gains by satellite remote sensing, measuring the vegetation index (NDVI) over long series.
These results matter because they respond to genuine skepticism: Chinese plantations sometimes lacked ecological foundation. Approaches combining community management, grazing limitation, and species diversification showed a better trajectory than monocultures. Lands that were simply reforested with monocultures of trees ill-adapted to local climate fared worse. The difference between the two approaches widened after 2015.
Satellite progression stalls
The NDVI index, which measures vegetation cover density from space, showed progression between 2000 and 2013, then stabilization or fluctuations after 2015 depending on the areas.
Several factors explain this plateau. First, the most easily restorable areas were treated first. Remaining lands are more arid, more degraded, more exposed to wind. Planting a tree on the Loess Plateau is difficult but feasible; planting in transition zones toward the Gobi Desert is another undertaking. At some sites, lowering of the water table and prolonged water stress associated with land use changes lead to plantation dieback, rendering them vulnerable to late frosts.
Next, monoculture plantations, which still represent a significant share of reforested areas, lack resilience. A poplar planted in a line is not a forest. It does not create the diversity of microhabitats that would allow other species to establish, nor the complex root systems that stabilize soil in depth. Faced with prolonged water stress, it dries out faster than a mixed forest.
This slowdown does not mean the efforts have failed. It means they have reached a different regime, more demanding, where additional gains become harder to obtain.
Three years of drought test the program
Droughts affected certain parts of northern China in 2020 and 2023, but the 2020-2023 period does not correspond uniformly to a severe sequence for all of the north. Rainfall deficits and droughts affected certain regions and localities in the north and northwest. Temporary rivers disappeared. Wells ran dry.
For the Three-North program, this meant significant tree mortality, protective embankments torn from the ground by wind on soil that had become friable, and a reduction in vegetation indicators in certain areas. Scientific studies document that monospecific plantations in semi-arid zones suffered high mortality rates during these extreme episodes.
Areas where species diversity was associated with community management show greater recovery capacity. The signal is nonetheless clear: the resilience of restoration depends as much on its method as on its scale. Planting 32 million hectares guarantees nothing if the chosen species are ill-adapted to the future climate of these areas.
Climate projections suggest that northern China could experience an increase in the frequency of dry episodes as global temperature rises, though the precise magnitude remains uncertain. The Three-North program was designed in a different climatic world from the one approaching.
Kazakhstan and other regional attempts
China is not alone in attempting large-scale restoration in Asia. Kazakhstan has launched programs to restore steppes degraded by decades of Soviet collectivization, which had plowed millions of hectares of natural grasslands for wheat crops that eventually eroded. Restored areas still represent 5 to 10% of targeted zones according to UNEP estimates.
These Kazakhstani efforts illustrate a different constraint than China’s: lack of financial and institutional means, not absence of will. Steppe restoration is less costly than reforestation, as it often consists simply of stopping plowing and allowing natural vegetation to return. But it demands time, livestock management, and coordination between nomadic herders and authorities that proves difficult in a country with still fragile institutions.
Other Central Asian countries are conducting similar experiments, often within the framework of commitments made at desertification summits under UNEP auspices. Results are heterogeneous, and independent data remain rare. Ecological restoration, when it succeeds, takes one to two decades before producing stable ecosystem gains, a temporality poorly compatible with political cycles that fund it.
Achievements and limits of restoration
Despite the plateau and losses linked to drought, the Three-North program has produced lasting effects that independent assessments confirm. The Loess Plateau exports significantly less sediment to the Yellow River than it did thirty years ago. Agricultural zones bordering restored areas have regained productivity. Sandstorms that regularly reached Beijing in the 1990s have become less frequent, even if they have not disappeared.
These gains are real. And they show that large-scale restoration can alter local water balance, reduce erosion, and offer measurable ecosystem services. Researchers who documented these results in Nature note that the most solid benefits appear where restoration was accompanied by land use reform: less grazing, less plowing, compensation for rural communities for ecosystem management.
The durability of restoration remains exposed to warming that models have not yet calculated with precision for these zones. Trees planted in the 2000s grew in a climate slightly different from that of 2025. Those planted today will grow in the climate of 2035 or 2050. The adaptation of plant species to the future climate of each zone has become as important as the number of hectares reforested.
Between 2035 and 2050: the tipping point that models have not yet settled
At what level of warming does large-scale ecosystem restoration cease to produce net gains: current climate and ecological models present significant uncertainties for the arid and semi-arid zones of Central Asia. They sketch several possible trajectories depending on emissions scenarios and management adjustments.
In the first, warming remains contained below 2 °C on a global scale, and restored zones could benefit from adaptive management to changes in precipitation regimes. Adaptation of planted species to local water conditions becomes an increasingly important issue. Restoration could continue to produce ecosystem services, with variations during dry episodes. This is the scenario in which restoration investments could produce lasting effects.
In the second trajectory, warming exceeds 2.5 to 3 °C above preindustrial levels, and certain semi-arid zones in the north, notably in the Northwest according to some scenarios, could experience more droughts or water deficits, without this implying a uniform increase in rainfall deficits. Beyond a certain hydrological threshold, which 2016 studies identify as a possibility without dating it precisely, plantations could consume more groundwater than ecosystems reconstitute it, worsening regional water deficit.
This second scenario does not invalidate restoration as a tool. It changes its nature. In a warmer and drier climate, the priority would no longer be to plant trees, but to restore herbaceous and shrubby ecosystems that demand less water, sequester less carbon but stabilize soils with greater efficiency per liter of water consumed.
Signals to watch are already identifiable. Monitoring water tables in these provinces is an important indicator of water sustainability, but it must be completed by local data on plantations, withdrawals, and precipitation. Mortality rates during droughts can test the relative resistance of plantations only with comparable monitoring of species, ages, densities, and site conditions. NDVI can track vegetation cover evolution, but assessing stability also requires measurements of survival, biodiversity, soil moisture, and water resources.
Researchers working on these questions identify an important need: field data on plantation mortality, water table levels, and species diversity. Satellite NDVI series are valuable, but they do not always allow distinguishing a living forest from one in water stress before the impacts are already visible on the ground.
Method, not scale, will determine what comes next
Experience with the Three-North program shows that scale alone is not enough. Approaches integrating species diversity adapted to local climate, association of rural communities with management, and limitation of previous degrading uses produce better results.
These lessons are being integrated into new phases of the Three-North program, which increasingly favor species adapted to local water regimes and restored grasslands rather than monospecific forests. This is an adjustment pointing in the right direction, but one that takes time to produce visible effects.
China must continue restoring; the alternatives are less favorable. The issue concerns the capacity of institutions and budgets to adapt as quickly as climate requires. A program launched in 1978 to address twentieth-century desertification must now address twenty-first-century desertification, which obeys different rules. The adaptive capacity of a program of this magnitude, faced with an accelerating climate signal, constitutes the central test of the coming decade.
Sources
- FAO Land and Water, Land Use Data
- UNEP Desertification Atlas, United Nations Environment Programme, reports on desertification in Central Asia
- MERICS China Programs Evaluation 2025, Mercator Institute for China Studies, evaluation of Three-North restoration programs
- Rhodium Group China Program, evaluation of the Three-North Shelterbelt program, 2024-2025
- Nature Scientific Reports, series of studies 2023-2024 on NDVI evolution in northern China and plantation resilience
- Nature Climate Change, 2024 studies on hydrological viability thresholds for plantations in semi-arid zones
