Reversal → recovery Third-party evidence

Mu Us Sandy Land

China · Ordos plateau · approx. 38.5° N, 109.0° E

Aridity class
Semi-arid
P/PET range
P/PET 0.25–0.45
Area covered
3,710,000 ha
Source: 32,100–42,200 km² · claimed extent; converts to 3,210,000–4,220,000 ha
Duration
≈ 43 years (tipping point ≈1986)
Main intervention
Afforestation, sand fixation
Approach
Technique-based
Evidence strength
Third-party evidence
Last assessment
Volet A methodology, 2026

Assessment summary

Groundwater moved through three phases — stable, then depleted, then recovering. Attributing the full decline to restoration overstates the effect by an estimated 251% once coal extraction and irrigation are subtracted; net consumption attributable to trees is closer to 11.7 mm per year. The greening itself holds. The water-table reserve remains a research question, not a refutation of it.

Visual context

Site imagery — Mu Us Sandy Land
Illustrative placeholder. Location: Ordos plateau, China. Source: to be supplied at implementation.
Local map — project extent
Geographic extent of the documented intervention.
Before / after — cover change
Illustrative before/after pairing. Real imagery and dates to be sourced at implementation.

Overview

Mu Us is a semi-arid sandy land on the southern edge of the Ordos plateau, in northern China. Since the late 1970s, the Chinese state has run a large afforestation and sand-fixation programme here as part of the wider Three-Norths Shelterbelt system. Vegetation cover has increased substantially since a documented turning point around 1986.

The programme is often mentioned alongside the neighbouring Loess Plateau as an example of dryland regreening. Deserts.eco treats the two as related but distinct cases: the greening is well documented, but how much of the change in local water tables can be credited to tree planting — rather than to coal mining or irrigation nearby — has been a genuinely contested question in the scientific literature, and one this project's evidence trail helps to unpack.

Context

Climate
Semi-arid continental, P/PET 0.25–0.45. Regional rainfall trend is itself debated between reanalysis products.
Hydrology
Shallow aquifers overlying the sandy substrate; active coal mining in parts of the area depresses the water table independently of vegetation.
Soil
Aeolian sandy substrate, low organic matter, historically mobile dunes.
Topography
Low-relief sandy land at the transition between the Ordos plateau and the Loess Plateau.
Initial degradation
Historically mobile sand dunes with sparse, unstable cover prior to the programme.
Local communities
Not separately documented at project scale in the sources reviewed; governance is state-led rather than community-initiated.
Governance
State-run, subsidised, top-down (Three-Norths programme architecture).

Intervention

What was done: Large-scale afforestation and mechanical/vegetative sand fixation across the sandy land, converting mobile dunes toward stabilised, vegetated cover.

Who carried it out: Chinese state forestry administration, as part of the national Three-Norths Shelterbelt programme.

When it started: Programme activity traceable from the late 1970s, with a documented cover turning point around 1986; the assessed window spans roughly 43 years.

How it was financed: Public subsidy.

Approach type: Technique-based — a defined set of afforestation and fixation techniques applied at scale, rather than a site-specific spatial design.

Estimated cost: Not reported — see Economics and investment below.

Trajectory

Groundwater in the assessed sector moved through three distinct phases rather than a single trend:

1987–1998

Stability — water table broadly stable.

2003–2010

Depletion — water table declines.

2011–2020

Recovery — water table rises again.

Textual alternative: the trajectory is not a straight line of improvement or decline. Reading it as continuous decline, or as continuous recovery, would both misrepresent it.

Once coal extraction and irrigation are subtracted from the total, the decline attributable to restoration is estimated at less than a third of the commonly cited figure — around 11.7 mm/year net consumption. In actively mined sectors, the water table sits below −70 m for reasons unrelated to tree cover.

LMDC assessment

Evidence is organised from descriptive documentation (N1) through to climate-drift context (N4). Higher levels demand more measurement and are, for this project as for most, less completely filled in.

LMDC assessment for Mu Us Sandy Land
LevelIndicatorCurrent evidenceSource typeTreatment required
N1 Rainfall regime Station data; uneven spatial coverage Third-party To normalise
N1 Hydrography / topography Maps, terrain model, discharge; heterogeneous sources Third-party To normalise
N1 Soil Aeolian sandy substrate, low organic matter Third-party To harmonise
N1 Cost per hectare Not reported — sources describe scale of state investment, not a per-hectare figure — To produce
N1 Governance & local anchoring State-run, subsidised, top-down Third-party To harmonise
N1 Economic model Public subsidy Third-party To harmonise
N2 Vegetation performance Greening documented; indices and thresholds vary study to study Third-party To produce
N3 Groundwater level / stock Sporadic studies; heterogeneous methods Third-party To normalise
N3 Surface energy flux assessment No comparable documentary access — To produce
N3 Dew monitoring No documentary access — To produce
N4 Regional climate drift Rainfall trend contested between reanalyses Active debate To harmonise

N4 here refers only to the documented regional climate trend around the project — not to a claim that the project itself is shifting the regional climate. See the LMDC framework page for that distinction.

Economics and investment

Two distinct metrics are tracked separately and must never be combined into one figure: documented investment (a specific reported amount, tied to a scope and a date) and cost intensity (a cost-per-hectare or cost-per-outcome rate). A project can have one, both, or neither reported — reporting one is not evidence of the other.

Documented investment
Not reported
Cost intensity
Not reported
Economic model
Public subsidy.

The reviewed literature (Kong 2021; Yue et al. 2025) characterises the Three-Norths programme as backed by a scale of continuous state budgetary commitment "that few countries can match" — but reports this qualitatively. No source reviewed for this platform gives a specific investment total, a start/end date for that total, or a comparable cost-per-hectare rate for Mu Us. Reporting an estimate here would imply a precision the underlying documentation does not support.

Editorial interpretation, not a reported fact: The scale and duration of Chinese state afforestation funding is unusual internationally, but "unusual" is a qualitative judgement from the literature, not a number this platform can independently verify.

Evidence gaps

  • Surface energy balance is not documented for this site — cooling versus warming effect of the new cover is unmeasured.
  • Dew and non-rainfall moisture inputs are not systematically monitored.
  • The regional rainfall trend itself is an active scientific debate between reanalysis products, which limits how confidently the site can be placed in a regional climate context.
  • No documented investment or cost-per-hectare figure was found in the reviewed sources for this project.

People and organisations

Chinese state forestry administration
Programme operator and funder — Three-Norths Shelterbelt system.
Independent researchers
Multiple academic teams have measured this site independently; see bibliography below rather than an editorialised list of names.

Deserts.eco does not publish individual contact details without permission.

Bibliography

  • Chen, X. et al. (2024). 65% cover is the sustainable vegetation threshold on the Loess Plateau. Environmental Science & Ecotechnology.
    Supports: historical and sustainable cover-threshold context for the wider region
  • Feng, X. et al. (2016). Revegetation in China's Loess Plateau is approaching sustainable water resource limits. Nature Climate Change.
    Supports: the water-resource-limit reading contested elsewhere in this assessment
  • Tian, L., Zhang, B., Chen, S., Wang, X., Ma, X. & Pan, B. (2022). Large-scale afforestation enhances precipitation by intensifying the atmospheric water cycle over the Chinese Loess Plateau. Journal of Geophysical Research: Atmospheres.
    Supports: the counter-reading that afforestation increases moisture convergence rather than depleting it
  • Wang, X. et al. (2023). Unintended consequences of combating desertification in China. Nature Communications.
    Supports: attribution share of public policy versus other drivers of greening
  • An, Z. et al. (2025). Land cover changes redistribute China's water resources through atmospheric moisture recycling. Earth's Future (AGU).
    Supports: estimated water-availability decline attributed to afforestation (contested)
  • Makarieva, A.M., Nefiodov, A.V. et al. (2026). On the methodology for assessing vegetation impacts on the atmospheric branch of the hydrological cycle. arXiv preprint (physics.ao-ph).
    Supports: methodological critique of the counterfactual used to attribute water decline to afforestation
  • Zhang, Y. et al. (2026). Mining and vegetation greening jointly drive groundwater depletion (Loess). GIScience & Remote Sensing.
    Supports: separating coal-mining drawdown from vegetation-driven water use
  • Shen, Y. et al. (2025). Disentangling ecological restoration's impact on terrestrial water storage. Geophysical Research Letters (AGU).
    Supports: the ≈251% overestimation figure once coal and irrigation are subtracted

See the full bibliography

Methodological note

Assessment scope
Reads existing third-party literature against the LMDC grid; does not include new fieldwork or original satellite processing.
Limitations
Surface energy and dew indicators are unmeasured for this site; regional rainfall attribution is contested in the source literature.
Methodology version
LMDC v0.1 (first iteration)
Reviewer
Deserts.eco editorial team
Date
2026
Revision history
v0.1 — initial publication. No prior revisions.