The Loess Plateau is a vast, windblown dust deposit in north-central China covering roughly 640,000 square kilometers, an area about the size of France. For millions of years, fine silt carried by winds from Central Asian deserts and mountain ranges accumulated here layer by layer, creating deposits that in places exceed 300 meters thick. This landscape has been both a cradle of Chinese civilization and the site of what researchers have long called the most severe soil erosion anywhere on Earth, with sediment losses in some areas reaching 15,000 tonnes per square kilometer per year. Over the past two decades, a massive government-led revegetation campaign has dramatically changed the plateau’s appearance and ecology, but the story is far more complicated than a simple greening success narrative.
Where the Dust Came From
The Loess Plateau did not form from rock weathering in place. Its material was generated hundreds or even thousands of kilometers away, ground into fine silt by glaciers, frost, and tectonic forces in distant mountain ranges, then blown southeast by prevailing winds and deposited across what is now Shaanxi, Gansu, Shanxi, and parts of surrounding provinces. Research tracing the mineral signatures of loess grains has identified the stony deserts (gobi) of southern Mongolia and the adjoining sand deserts of northern China as the primary holding and staging areas for this dust. But those deserts were more like warehouses than factories. The actual silt production happened in the Gobi Altay, Hangayn, and Qilian mountain ranges, where glacial grinding, frost weathering, salt weathering, and river action broke rock into particles fine enough for wind to carry.
The picture gets more nuanced when you look at different parts of the plateau. Zircon dating and mineral analysis of loess samples from multiple sites shows that most of the plateau’s material has a close affinity to Yellow River sediments and northern Tibetan Plateau sources. But at the northeastern margin, near a site called Jingbian, the dust fingerprint shifts: it resembles material from northern deserts and the ancient bedrock of the North China Craton.
A 700,000-Year Climate Diary
One of the most scientifically valuable features of the Loess Plateau is that it records climate history in extraordinary detail. During cold, dry glacial periods, strong winter monsoon winds carried more dust onto the plateau, depositing thick layers of pale, unweathered loess. During warm, wet interglacial periods, summer monsoon rains increased, vegetation grew, and the loess surface weathered into darker, more organic-rich paleosols (ancient soils). The result is a layer cake of alternating loess and paleosol that researchers can read like tree rings.
At one site in the northwestern plateau, a 180-meter-thick sequence records the last seven glacial-interglacial cycles. Scientists use the magnetic properties of the sediment as a proxy for summer monsoon strength, and grain size as a proxy for winter monsoon intensity, reconstructing monsoon behavior stretching back hundreds of thousands of years. More recent work has pushed these reconstructions further by using chemical signatures preserved in the soil, specifically branched membrane lipids produced by bacteria, to estimate how much rain actually fell in past millennia. One such record spanning 430,000 years revealed something unexpected: monsoon intensity and actual rainfall did not always move in lockstep. During transitions between glacial and interglacial conditions, the chemical proxies that track monsoon strength could become decoupled from actual precipitation, meaning a “stronger” monsoon did not always translate to more rain on the ground.
Why Loess Erodes So Catastrophically
Loess is inherently fragile. It consists of loosely packed silt grains held together by weak bonds between particles, and its structure is riddled with vertical pores and channels. When water enters, the consequences are dramatic. Laboratory studies of undisturbed loess samples show that water infiltration washes out fine particles, increases porosity by about 20 percent on average, and transforms the internal pore structure from mostly disconnected to more than 90 percent connected. That network of connected pores creates preferential flow paths, essentially internal plumbing that channels water deep into the deposit and sets the stage for sudden, large-scale collapses and landslides.
On the surface, gully erosion is the dominant form of land degradation. The plateau is carved by an extraordinary density of gullies, some tens of meters deep, that advance headward with each rainstorm. Research using controlled scour tests has shown that soil loss increases with both the rate of gully retreat and the initial height of the gully head, and that sandy loess soils erode faster than silty ones. On unpaved roads, which are common across the rural plateau, gully formation follows a threshold relationship involving road gradient and runoff volume, with extreme rainstorms producing total gully volumes roughly twelve times larger than regular storms.
At the plateau scale, erosion rates in the most severely affected areas have been measured at up to 15,000 tonnes per square kilometer per year, with roughly 91,200 square kilometers subject to rates above 8,000 tonnes.
Centuries of Human-Driven Degradation
The erosion crisis on the Loess Plateau is not purely geological; it is largely a human creation. The region was the heartland of early Chinese civilization, well covered by grasses and trees at least through the Western Han dynasty, which ended in 8 AD. Over the following centuries, long-term vegetation destruction driven mainly by human land clearing, and partly by climatic shifts, stripped the landscape bare. The plateau was transformed from relatively flat, vegetated terrain into a labyrinth of barren hills and deep gullies, and from one of China’s most culturally and economically advanced regions into one of its poorest.
The damage accelerated in the twentieth century. Research reconstructing erosion drivers since 1901 found that extensive deforestation and farming, driven by population growth, intensified erosion between the 1930s and 1970s. Only after that period did policy-driven conservation schemes and deliberate revegetation begin to reverse the trend.
The Grain for Green Campaign
China’s most ambitious attempt to heal the Loess Plateau is the Grain for Green Program (also called “Returning Farmland to Forest and Grassland”), launched in 1999. Under this program, farmers on steep slopes were paid subsidies to stop cultivating and allow their land to revert to forest, shrubland, or grass. The program has been the largest ecological restoration effort in the world by area, and on the Loess Plateau its effects are visible from space. Satellite data shows a strong correlation between cumulative afforestation area and vegetation greenness (measured by NDVI) from 1998 to 2013.
On the ground, the program has been paired with physical erosion-control structures: terraces carved into hillsides and check dams built across gullies. Studies of slope-gully systems show that terracing alone reduced runoff by roughly 46 to 84 percent and sediment yields by about 26 to 82 percent, depending on conditions. Check dams added further reductions of 8 to 33 percent for runoff and 10 to 31 percent for sediment. When terraces and check dams were combined, they produced synergistic effects, meaning the total reduction was greater than the sum of each measure working alone. A separate modeling study found that combining terraces and check dams reduced overall sediment connectivity by about 32 percent, with terraces primarily cutting slope erosion and check dams increasing channel deposition.
The Water Tradeoff
For all its success in reducing erosion, large-scale revegetation has created a serious side effect: it is consuming more water than the plateau’s semi-arid climate can easily supply. Trees and shrubs transpire far more water than the crops and sparse grasses they replaced, and this shows up across multiple lines of evidence.
A meta-analysis of 52 case studies across the plateau found that streamflow decreased in 41 of them, and that the decrease was mostly not explained by changes in precipitation. Land use change accounted for about 64 percent of the streamflow reduction, while climate change accounted for roughly 36 percent. An increasing area of soil and water conservation measures correlated with greater streamflow decline. Across the Yellow River system, sediment yield dropped much faster than streamflow, with the sediment decline rate roughly three times the streamflow decline rate in the mainstream river.
Underground, the picture is equally concerning. Field measurements have shown that artificial forests consume more deep soil moisture than the cultivated crops they replaced, inducing soil desiccation and the formation of dry soil layers. One ecohydrological modeling study estimated that the plateau’s average vegetation cover had already reached 0.48, exceeding the climate-defined equilibrium cover of 0.43. In other words, parts of the plateau have been overplanted relative to what the available rainfall can sustain, and this overplanting is primarily responsible for ongoing soil drying.
This does not mean revegetation was a mistake, but it does mean there are limits. The gains in erosion control and carbon storage come at the cost of reduced water availability, and pushing vegetation cover beyond what the local climate supports risks creating a new set of problems.
Can Greening Create Its Own Rainfall?
One of the more intriguing questions is whether the plateau’s new vegetation might partially compensate for its water consumption by generating additional rainfall. Plants release moisture into the atmosphere through transpiration, and if enough of that moisture recycles into local precipitation, the water budget might partially rebalance.
Recent atmospheric modeling using water vapor tracers suggests this is happening, at least to a degree. One study found that greening promoted growing-season precipitation by about 0.45 millimeters per day, with roughly 15 percent of the increase coming from local evapotranspiration recycling and 85 percent from increased water vapor inflow from outside the plateau. A separate analysis covering 2000 to 2019 found that local precipitation increased on average by 0.16 millimeters per year, while evaporation increased by 5.17 millimeters per year. Local moisture recycling rose from about 6.9 percent in the pre-restoration period to 8.3 percent afterward. The conclusion was that revegetation had a positive but modest effect on local precipitation, and that the observed increase was driven by a combination of local greening and broader atmospheric circulation changes.
The effect is real but small relative to the extra water the vegetation consumes. Greening alone is not going to make the plateau wet enough to support unlimited forest cover. The 85 percent contribution from external moisture inflow also means the precipitation boost depends partly on factors beyond the plateau’s borders.
Carbon Gains in the Soil
Beyond erosion control, revegetation has been rebuilding the plateau’s depleted soil carbon stocks. Decades of farming had stripped organic matter from loess soils, and the return of permanent vegetation is slowly reversing that loss. A combined remote-sensing and modeling study estimated that revegetated areas on the Loess Plateau had accumulated a total of 0.8 petagrams (billions of tonnes) of soil organic carbon in the top meter by 2022. Under a business-as-usual scenario, the model projects that soil carbon will reach a steady state after about 150 years of continued revegetation, with a potential total storage of 2.9 petagrams.
The type of vegetation matters. Research in the loess hilly area found that converting cropland to shrubland or undisturbed wild grassland was better for soil carbon accumulation than tree plantations, and that most of the carbon gain occurred in the surface soil layers. This aligns with the broader finding that grassland and shrubland may be more appropriate than forests for much of the plateau’s drier territory.
Landslides and Liquefaction
The same porous, collapsible structure that makes loess vulnerable to erosion also makes it prone to catastrophic landslides, and human activity is a major trigger. On the Loess Tableland in Shaanxi province, most landslides are of the “sliding-flow” type, meaning the soil mass does not just slide downhill but liquefies and flows, covering long distances. Field investigations found that the apparent friction of these irrigation-triggered sliding-flow landslides ranged from 0.01 to 0.25, with a mean of 0.159, values lower than those seen in landslides triggered by the 2008 Wenchuan earthquake. Laboratory tests confirmed that saturated loess is highly liquefiable: pore pressures spiked to about 92 percent of the applied normal stress, and effective stress dropped to near zero. The loose structure and large pores of loess provide the physical framework for this static liquefaction.
Irrigation is particularly dangerous because it delivers water slowly and continuously into the subsurface, dissolving the suction forces between grains that keep dry loess stable. Experimental work has confirmed that irrigation causes dissipation of soil suction in near-surface layers, reducing shear resistance and triggering shallow slides. For small slopes, the loss of suction alone can cause failure, though for larger landslides additional factors come into play.
Building Cities on Loess
Despite the geological hazards, rapid urbanization has pushed some Chinese cities to literally reshape the Loess Plateau’s topography. In one of the most dramatic examples, satellite monitoring documented a “mountain excavation and city construction” project that reshaped terrain by up to 80 meters in elevation change and transferred roughly 128 million cubic meters of earth by 2015. Hilltops were shaved off and valleys filled in to create flat land for urban development. The filled areas subsequently subsided at rates up to about 70 millimeters per year during 2014 to 2020, with subsidence proportional to the depth of fill material. Projections suggest the new ground will not stabilize until around 2030. The massive redistribution of weight also generates stress changes in the shallow subsurface, adding yet another hazard in a landscape already prone to landslides.
The Livelihood Question
The Grain for Green Program has been widely celebrated as an environmental success, but its effects on the people who actually live on the plateau are more ambiguous. Research examining livelihood consequences found that at the regional scale, the program actually reduced rural households’ net income and livelihood diversity. The losses stemmed from a substantial decrease in grazing income combined with insufficient government compensation, the failure of economic forest schemes (plantations intended to produce marketable fruit or timber that did not deliver), and ineffective labor training programs. At the household level, the traditional smallholder agro-pastoral system swung from extreme dependence on the natural environment to an extreme prohibition against grassland grazing, which alleviated environmental pressure but lowered overall productivity and undermined the sustainability of rural livelihoods.
The underlying problem, according to the researchers, is that policy still prioritizes ecological goals over economic ones, restricting rural households’ access to land resources without providing viable alternative livelihoods. Any future phase of restoration on the plateau will need to address this imbalance if it expects cooperation from the people who live there.
Matching Vegetation to Water Supply
The emerging scientific consensus is that future restoration on the Loess Plateau needs to be far more selective about what gets planted where. A recent evaluation using multiple environmental factors, including precipitation, temperature, altitude, slope, soil texture, soil depth, and ecological water consumption, mapped out which parts of the plateau are actually suitable for trees, shrubs, and grass. The results suggest that trees should occupy only about 19 percent of the total vegetation area, concentrated in the wetter central and southern plateau. Shrublands are appropriate for roughly 33 percent, mainly across the northern plateau. Grasslands should dominate, covering about 49 percent of the area, mainly in the drier western and northeastern regions.
These proportions are a significant departure from the early years of Grain for Green, when tree planting was heavily emphasized across the board. The modeling also calculated optimal vegetation coverage values for each category, finding that suitable tree areas could support coverage ranging from about 0.14 to 0.87, shrublands from about 0.19 to 0.78, and grasslands from about 0.25 to 1.0, depending on local conditions. The takeaway is that a grassland-dominated plateau with strategically placed forests in the wettest zones would consume less water and likely prove more sustainable than the current trajectory of pushing tree cover into areas that cannot support it long-term.