Desertification, as defined by UNCCD:
[…] the degradation of land in arid, semi-arid, and dry sub-humid areas. It is a gradual process of soil productivity loss and the thinning out of the vegetative cover because of human activities and climatic variations such as prolonged droughts and floods […]
United Nations Convention to Combat Desertification (UNCCD) & the Land Productivity Dynamics (LPD)
By taking ‘desertification’ to be the loss of soil productivity alongside the thinning out of vegetative cover due to human activities and climatic variations such as prolonged droughts and floods within arid, semi-arid, and dry sub-humid territories, a complex planetary crisis had been transformed into measurable metrics. This is necessary for the legibility of this process to state and extra-state entities as it further enables similarly measurable actions to be taken. Thus, it is far from simply a scientific achievement — it is an act of administrative translation.
This legibility is a prerequisite for institutional action; more consistent and comparable reporting helps “policymakers track progress and take action where it is most needed”, on the basis that global efforts to combat desertification, restore degraded land and soil, and strive to achieve a land degradation neutral require first the harmonisation of methods and data across countries to strengthen the global knowledge base. As UNCCD put rather succinctly in their latest release (2025) of GPG for calculating the proportion of degraded land over total land area (Indicator 15.3.1), this indicator “not only provides accurate and comparable global data, but also serves as a reliable tool for measuring progress towards land degradation neutrality”.
This update also addressed the challenges of difference in definitions, methodologies, and datasets that limited the comparability of reported results globally. It establishes common nomenclature and definitions, providing harmonised guidance for measuring and comparing the three sub-indicators while clarifying data sources, processing workflows, and integration methods. The other key advancement was the improvement of Land Productivity (LPD) metrics, simplifying the statistical methods for detecting degradation with clearer interpretation of severity and confidence levels while refining the One-Out-All-Out (1OAO) approach, hence preventing false positives or negatives in cases like land restoration activities temporarily lowering productivity readings.
However, this drive for ‘comparable reporting’ necessitates a reduction of reality.
To make a landscape actionable for policymakers in Riyadh or Beijing, the nuanced, shifting sensitivities of dryland ecosystems, with all the complex cycles and feedback-loops they are part of, must be flattened into the standardised nomenclature of the Indicator 15.3.1, derived mainly from the following three sub-indicators:
1. Trends in land cover – assessing land cover change over time;
2. Trends in land productivity – measuring variations in vegetation Net Primary Productivity (NPP);
3. Trends in carbon stocks – estimating changes in soil organic carbon (SOC) stocks.
2. Trends in land productivity – measuring variations in vegetation Net Primary Productivity (NPP);
3. Trends in carbon stocks – estimating changes in soil organic carbon (SOC) stocks.
For instance, to run the JRC LPD tool, the input dataset required consists primarily of a time series of raster data representing vegetation productivity, the assessment’s ecological and analytical context enhanced by optional supplementary layers. Thus the core input is a multi-band raster, in which each band corresponds to a specific year’s productivity value — typically derived from satellite-based indices such as NDVI — which often represent annual integrals or seasonal summaries such as SumNDVI, assumed to reflect total vegetation productivity for each pixel in each year.
Although these metrics provide a global baseline, they produce a ‘binary’ logic for the desert. Through the lens of satellite-derived indices, specifically those for vegetation, a pixel can be categorised as either ‘degraded’ or stable’. Even with refined models allowing for a spectrum from ‘improving’ to ‘degrading’, the fundamental logic removes rooms for nuances, for it is beneficial to policy-makers to leave no room for ambiguity. This reinforces the logic of a border between the ‘green’ of productive lands and the ‘gold’ of desertified lands. This logic then reinforces the top-down approach towards land management, with individual and local actions erased by the smoothing effects of data resampling or reduced to rounding errors.
Furthermore, administrative reliance on course-resolution sensors (e.g. MODIS at 1km, or various Land Cover products at around 300m and 500m) ensures that the most sub-pixel realities, where ecologically significant changes occur, remains invisible. The mathematical smoothing to resolve resolution jumps creates a blurring of reality, reducing the nuances to rounding errors.
WAD & the Convergence of Evidence
While the UNCCD provides regulatory baselines, pushing LPD as the primary instrument to monitor degradation and efforts to combat it in pursuit of Land Degradation Neutrality (LDN), World Atlas of Desertification (WAD) — a JRC publication commissioned by the European Commission in support of UN goals — is standardising at a global level ‘desertification’ itself.
Going back to the definition of ‘desertification’, UNCCD tracks desertification and intervention efforts through the LPD, which works on the datasets providing information on change in productivity and vegetative cover. Meanwhile, WAD employs the ‘convergence of evidence’ methodology to move beyond isolated indices, using both biophysical markers (e.g. aridity, tree loss) and socio-economic markers (e.g. population density, livestock, infrastructure) to note where they appear as ‘stressors’ to the environment, mapping out zones of concerns where they intersect. Thus, despite acknowledging that desertification does not occur in a vacuum and therefore broadening beyond UNCCD’s narrow focus on productivity, WAD still fell prey to the need of legibility, similar to that required of UNCCD, to make the information produced actionable on by state actors and policymakers.
Although it makes ‘intangible’ drivers of degradation visible, recognising a road to be a potential contributor alongside the lack of rain, what is omitted is often just as important as what was selected. Among its full body of work in the WAD itself, plenty of other information had also been studied and mapped, only to be stripped out of the final, official convergence of evidence. This has a simplified set of indices for informing the general population on ‘desertification’ and by extension, how to ‘combat desertification’. It is through this act of selecting where the influence of anthropocentric, global lens with the tendency towards believing in territorially localised causes and effects and therefore solutions shows through, despite studies showing very much planetary and global involvements in many issues.
One instance is that of tracing virtual waters. Other than a detailed study of surface water and groundwater alongside evapotranspiration and others, WAD looked into the phenomena of virtual water, the water embedded in the production of commodities. However, these layers with messier, planetary and global involvements were removed from the final map. The finalised, executive summary map then goes on to inform both the general public, the stakeholders, and policymakers.
This omission first aligns with the general anxiety that desertification is the consequence of ‘too many people’. Combining census data with soil productivities, agricultural activities, and subsequently ‘desertification’ creates a spatial correlation allowing for the blame for the failure of the land to fall on the individual, which also happens to be a convenient simplification, since inclusion of global flows like that of virtual water breaks the ‘territorial’ logic of desertification. This creates a blind spot overlooking supra-national economic flows and industrial extractions, or even resources drawn by populations outside these areas in fact driving soil exhaustion or resource depletion. But the inclusion of flows would mean having to consider, for instance, the connection between depleted groundwater in the middle east and dinner tables in Europe, complicating matters — including these global metabolisms make it hard to blame the ‘local’ livestocks and populations. WAD and UNCCD sets out to simplify the extremely complex phenomena in an attempt to make it legible for policy-driven actions to be taken; this complication is explicitly and fundamentally against what they set out to do.
Thus desertification is reduced to a set of indices, its global and planetary dimensions removed, such that global metabolic flows are transformed into local technical problems put forward within tidy and actionable maps produced for state-level actors and individuals. The freezing of stressors into a map suggesting the solution to be a counter-pressure of a similar scale, ignorant of global pressures and planetary linkages, can then present top-down stabilisation of these variables as the solution, reinforcing the idea that the crisis can be ‘engineered’ away by targeting these individual factors to change the colours of these pixels, ignoring global cycles driving them in the first place and blind to the temporality of the others.
Three North Shelter-Forest Programme (TNSFP) & the Price of ‘Success’
Established in 1978, the Three-North Shelter Forest Programme (TNSFP) in China is perhaps the world’s most ambitious action taken in response to this logic of legibility. Covering 775 counties across thirteen provinces of three norths — Xibei, Huabei, and Dongbei — and the Xinjiang Production and Construction Corps (XPCC)’s area for a total of 4.486 million square kilometres (4,486,000 km2), which is about 46.7% of China’s area, and spanning 73 years from 1978 to 2050, it is split into three large stages with eight phases total, and currently (2026) it is in the sixth phase. In the sixth phase, there are 68 critical projects, with three main ‘battles’ in the core areas encompassing 35 projects.
Official numbers are staggering, for it had accumulated after forty years and five phases 480 million mu of afforested area (320,000 km2), fixed 1.28 billion mu (853,333 km2) of degraded grassland and 500 million mu (333,333 km2) of sandy (desertified) land, with its forest cover raising from the 5.05% of 1977 to the current 13.84%. By all measures from UNCCD and WAD, the TNSFP must be a monumental success; it had changed the colours of the pixels viewed from the satellite from gold to green, after all. However, as the project is governed through this lens, it had fallen into a series of devastating and often invisible failures.
Green Deserts
Veiled by the upward trend in pixel greening is the creation of ‘green deserts’: vast, silent stretches of biological infrastructure satisfying the spectral requirements of a satellite sensor while supporting zero biodiversity. By prioritising canopy cover over ecological complexity, the TNSFP produces standing grids of monocultural poplars and pines, offering no habitat for local fauna and actively suppressing native understory vegetation. These biological infrastructure look healthy even in 10m pixel resolutions, yet they are simply aesthetic placeholders that remain profoundly fragile, for they lack the species diversity required to survive climatic shifts and conceal the absence of indicators like subterranean fungal networks, which points to the health of the soil crust and the water sources available.
This ‘tree-centric’ view of restoration was subtly directed by the limitations of the technology used to monitor it. Because trees are more legible at the 10m to 1km resolutions used for monitoring purposes across large swaths of land than the sparse native shrubs or grasses, the TNSFP devaulted to massive, monocultural plantations of fast-growing species. They were then propogated across the vast, disparate terrains with little regard for local specificity, effectively replicating a singular ‘biological’ solution at a continental scale. This lack of ecological complexity left the landscape without robustness, leaving the Green Wall vulnerable.
Aside from the mass die-offs resulting from these trees exhausting a land not suited to support such large numbers of them of its nutrients (e.g. Phosphorous), there was also the catastrophic consequence of a lack of biodiversity: a single blight could decimate thousands of hectares of forests, as demonstrated in the Minqin disaster. Lacking the natural buffers provided by species diversity, the Asian Longhorned Bettle outbreak in the 1990s and early 2000s thus resulted in nearly a billion trees either killed, or culled to halt the infestation, setting the program back 20 to 30 years - a systemic collapse of an ecosystem built to cater to the metric.
Furthermore, preparing the landscape for these plantations often required destruction of what kept it stable. This can mean clearing existing biocrusts and ground covers essential for soil stability and moisture retention in these arid zones to make room for tree saplings, or breaking the regoliths to plant the saplings and releasing the silt beneath. This ironically accelerates erosion in the short term, desertifying the land further in pursuit of an optical success of a ‘green’ pixel replacing a ‘gold’ one.
Diversions & Displacements
In the endorheic basins of the Taklmakan, life is dictated by the vast, shifting deltas of the Alluvial Fan, where glacial melt from the Himalayas and Kunlun Mountains meet the desert floor. These inherently mobile systems rely on wandering rivers which shift seasonally and annually to feed the ephemeral ecosystems and indigenous communities who had mirrored its mobility for centuries. However, the TNSFP’s drive for a fixed vegetative cover creates in place of this fluidity a rigid plumbing infrastructure. By diverting the water with concrete pipes and irrigation canals to feed the shelterbelts, the state had effectively severed the metabolic link between the mountains and the basin’s sinks.
A consequence of this hydrological diversion is the expansion of saline and sodic dead zones. These appear at the terminal ends of vanished rivers and in the vicinity of the shelterbelts.At the former sites, salt-crusted desiccated basins remain. Near shelterbelt sites, however, a more complex redistribution of salt occurs.
Excessive irrigation raising the mineral-dense groundwater table triggers the capillary action pulling these minerals to the surface, where the water evaporates instantly due to the extreme aridity. This leaves behind the salt and over time, creates the sodic seal, a rock-hard crust suffocating the roots of the trees greening the landscapes.
Simultaneously, the irrigation creates a groundwater mound travelling laterally beneath the dunes and resurfacing eventually in low-lying depressions further from the tree lines, creating hypersaline ‘white pools’ otherwise known as Solonchaks. While the timber grids might be periodically flushed with fresh water to let the roots breath, this merely displaces the minerals into surrounding deserts. The chemical scarring of the land thus results from the toxic redistribution satisfying the immediate needs of the shelterbelts while poisoning the broader territory.
This destruction remains largely invisible to the official gaze. As wandering rivers and native ecologies are largely nomadic and seasonally inconsistent, they show up sparingly in large-scale datasets. This leads the state actors to misread these areas as barren or unproductive voids. This perceived emptiness invites the installation of the biological infrastructures, filling in the voids with static, irrigated forests that are consistently legible both spatially and temporally. This results in a green pixel that suffocates alluvial fans, preventing it from supporting the ephemeral systems. Furthermore, as the official gaze is calibrated to measure land productivity through the greenness of the pixel, the emergence of white pixels is brushed off as a byproduct, since this secondary failure does not register against the primary optical measure of success.
Soil Moisture Loss & Groundwater Debt
In the pursuit of the green pixel, water is treated as an infinite resource rather than a finite one. Beneath the visible salinisation of the lands lies a more hidden failure: the desiccation of bio-crust and the extraction of terrestrial water stores.
To make room for the biological infrastructure, drought-resistant ground cover is either cleared or ploughed. This disrupts the land’s primary mechanism for retaining moisture, accelerating the depletion of the nutrients and moisture within the biocrust to support trees that typically demand more than a dryland can provide. Should the plantation fails, the result is an accelerated erosion more severe than it had been; even during the successful phase of the project, the cleared land is sandier than it would have been before.
Furthermore, the ‘tree-centric’ logic driven by the vegetation cover which would show up on satellites leads TNSFP to trade a horizontal ground-cover that reduces dust which could be picked up by passing winds for a vertical shelterbelt. Intended to break the wind, these trees are nonetheless subjected to more dust-carrying winds than the shorter ground covers, which could induce photo inhibition, particularly in plants not suited to extra sandy conditions. This creates an added dimension of stress for the vegetation, which is lost when the ecological relationships with its lands are simplified into the belief that sufficient greening can eventually resolve the hydrological debt. For the logic overlooks both the stress the environment exerts on the trees, and the one the trees exert on the hydrological reserves.
Data regarding terrestrial water storage in the Three-North region over the last two decades uncovered that in anomalies hotspots where the water reserves were decreasing, vegetation exerted a more pronounced influence than climatic factors, although precipitation governs water increases in some other areas. Nonetheless, neither the hydrological debt of plummeting water tables nor the depletion of soil nutrients and moisture are legible in the administrative gaze, for in the logic of the standard metrics a green pixel is considered ‘fixed’. Thus state-level and extra-state entities remain blind to the reality that their greening projects are not the solution to desertification and instead catalyses it in the longer term.