Desert Greening

Established Desert greening means deliberately raising and holding vegetation cover in arid and semi-arid land. Three different projects travel under the one name and should be kept apart. (a) Restoring degraded drylands that still receive enough rain to carry vegetation — difficult, partially demonstrated, and the subject of almost all the measured evidence. (b) Greening genuinely hyper-arid desert, where the water is not present and must be imported from a coast or an aquifer. (c) Continental surface-property engineering intended to change the circulation that makes the desert in the first place. Nearly every inflated claim in this field comes from reporting (a) as though it were (b), or from citing the modelling of (c) as though it were a plan.

Established The flagship is Africa's Great Green Wall, launched in 2007 across the Sahel–Sahara. Even its headline dimensions are not stably reported: Bruckmann and colleagues (2022) give the original specification as a 15 km-wide strip along a 7,000 km belt from Dakar to Djibouti, while Deng, Hao and Qu (2024) describe the initiative as spanning 8,000 km across 11 key and 7 associated countries, and neither reconciles the two. This brief uses 7,000 km for the original plan and the eleven-plus-seven footprint for the measured zone, and flags the gap rather than averaging it: a programme whose basic parameters drift 14% between sources is exactly where pledged quantities get reported as delivered ones.

Frontier The most informative single fact about the field is that the programme's own framing has moved from a wall of trees to “tree-based development programs” engaging stakeholders from government to local level (Mbow 2017). Established That is not a rebranding. It is a concession that the intervention which the public still pictures — nurseries, seedlings, a line of trees across a continent — is not the intervention that works.

Established This brief is distinct from Terraforming, which is about another planet; from Future Agriculture, which is about yield rather than cover; and from Water Infrastructure Megaprojects and Continental Irrigation Systems, which move water to the desert rather than holding what already falls on it. Where the carbon claims land is Carbon Capture at Scale; where the consent problem lands is Geoengineering Governance.

Established The Great Green Wall's audited record is the most important set of numbers in this subject, and it is not flattering. Bruckmann, Chotte, Duponnois, Loireau and Sultan (2022), in Land, report the UNCCD's 2020 progress document: about 20 million hectares of a planned 100 million had been restored — 20% — 350,000 jobs had been created, and 300 million tonnes of CO2 had been sequestered, which the report gives as 30% of the planned carbon target. The report, in the reviewers' words, “showed a mixed picture concerning the meeting of the initial targets.” Established That audit falls at roughly the two-thirds point of a 2007–2030 programme window. Established Note also that 20% and 30% are shortfalls against different denominators; they are usually collapsed into a single “behind schedule” sentence, and the carbon figure is the one that would appear in any climate-mitigation accounting of the programme.

Established At the 2021 One Planet Summit, 19 billion US dollars was pledged for 2020–2025, alongside creation of the Great Green Wall Accelerator. Frontier Disbursement against that pledge is not reported by the source that reports the pledge, which states explicitly that it does not give the Accelerator's funding amount or disbursement rates. Established The retrieval pass behind this brief could not reach UNCCD, FAO or any multilateral primary source, so the audit above arrives second-hand, through a peer-reviewed paper reporting a UNCCD document rather than the document itself — which is the ordinary condition of public claims about this programme.

Established The satellite record is better than the ground record, and it is modest. Deng, Hao and Qu (2024), in Remote Sensing, analysed 2001–2023 against a 2001–2020 climatology across the Great Green Wall zone. Regional mean NDVI rose +0.00060 per year (p = 0.000): statistically significant, and small. The desert boundary “moved northward in 79.24% of the regions along the longitude direction, while the other 20.76% of regions moved southward” — retreat across roughly four-fifths of the transect, advance across one-fifth. Established Over the same period and domain, precipitation fell significantly, -0.00028 mm per hour per year (p = 0.002), and soil moisture trended down but not significantly (p = 0.382).

Frontier Vegetation rising while rainfall falls is the most interesting single result in the topic, and it cuts both ways. It is what makes an anthropogenic attribution arguable rather than assumed, because greening over a drying period cannot be a simple wet-cycle artefact. Frontier But it is equally consistent with CO2 fertilisation, which requires no programme at all. Established And it means greening is occurring against a drying trend, which is a persistence question rather than a success story. Progress narratives quote the NDVI number and omit the precipitation number; restoring the second sentence is the cheapest available improvement to Great Green Wall reporting.

Established The regional average conceals countries moving in opposite directions. Ethiopia shows NDVI +0.00074 per year (p = 0.029) alongside precipitation +0.00045 per year (p = 0.017) — greening and wetting together, so Ethiopian greening is at least partly climate and cannot be booked to the programme. Established Nigeria, a core Great Green Wall country, shows NDVI -0.00028 per year (p = 0.014): statistically significant browning.

Frontier And then the counterfactual evidence points the other way for the same country. Wang, Scott and Dallimer (2026) built ensemble species distribution models for avian species richness, compared a 2007–2015 baseline against 2016–2024, matched treated and untreated locations 1:1 on pre-treatment environmental covariates by propensity score, and estimated the average treatment effect on the treated by difference-in-differences. Frontier Nigeria: ATT = +7.45, p < 0.001 — the best of the three countries, with scenario decomposition attributing the effect to vegetation change rather than climate, that is, to the restoration itself. Frontier Ethiopia: ATT = -2.48, p = 0.13 — no significant difference. Frontier Senegal: ATT = -4.17, p < 0.001, with climate-driven change dominating and vegetation effects described as constrained. The authors' own conclusion is the right one: “large-scale restoration does not uniformly deliver biodiversity co-benefits.”

Established So Nigeria is simultaneously the worst NDVI performer and the best counterfactual biodiversity performer among Great Green Wall countries, in two peer-reviewed studies published two years apart using different instruments. Frontier This brief declares the tie rather than resolving it. Four explanations are available and none has been tested: greenness and biodiversity decouple; Nigeria's no-intervention trajectory was a steeper decline than what was observed; the intervention footprint and the national aggregate do not overlap; or one result does not replicate. Frontier The resolving experiment — greenness and counterfactual biodiversity co-measured on the same plots — is cheap, and nobody appears to have run it. Established Meanwhile the contradiction is the strongest available evidence that NDVI is a poor proxy for restoration success, which is a problem for every satellite-based progress claim about this programme, including the ones two paragraphs above.

Established The Senegal line deserves its own paragraph, because of how it was described. An ATT of -4.17 at p < 0.001 is a large, highly significant negative effect of intervention on modelled avian species richness relative to matched controls. The abstract describes Senegal as having “exhibited limited benefits.” Established “Limited benefits” describes a small positive effect. The same abstract correctly describes Ethiopia's non-significant negative estimate as no significant difference; the Senegalese one is not correctly described. Frontier Report the three numbers, never the three adjectives. Frontier The honest caveat is that this is a conference abstract rather than a final paper and the wording may be corrected on publication — but the direction of the euphemism is itself the finding, and it is this field's characteristic direction.

Frontier Attribution to the programme is not established anywhere it has been tested carefully. Karwowski (2024) examined Burkina Faso's foundations using satellite imagery from 1990 to 2020: massive vegetation loss 1990–2002, recovery 2002–2007, renewed strong growth 2014–2020 — and concluded that correlation with the Great Green Wall cannot be confirmed without additional precipitation and observational data. Established Note the timing: the strong recovery began before the programme launched in 2007, and any time series baselined at 2007 will hand a pre-existing trend to the intervention.

Established What planted trees actually do is measured, but not where it matters. Ferreira, Cordeiro and Sampaio (2022) followed four tropical seasonal species over 68 months: average survival 85%, with Astronium at 94% by direct seeding against 73% by seedling planting, and direct seeding twice as cost-effective. Established Zahawi and Holl (2009) found Erythrina stakes above 90% survival but most other species between 40% and 70%, with stake biomass 7 to 50 times seedling biomass. Established Holl, Zahawi and Cole (2011) found planting design made little difference while species choice and nitrogen fixation did; Frontier Schweizer and Brancalion (2019) found performance rising with phylogenetic distance from the overstory, so conspecific enrichment does badly. Established Every one of these comes from a tropical, comparatively wet system. Frontier Dryland survival is expected to be substantially worse, and this brief has no dryland survival curve to give. That is the largest measurement gap in the subject, and every Sahelian carbon claim sits on top of it.

Frontier What measurably worked in the Sahel was not planting. Vieitez-García and Roca (2024) report that in the Mauritanian Great Green Wall the interventions with demonstrable soil results were fencing plots, preventing the passage of livestock, and stabilising sand dunes, with consequent improvements in soil organic carbon, pH, structure, colour and texture. Established Those are exclusion interventions, not silvicultural ones. Established The same paper names the structural obstacle plainly: “In the Sahel, land ownership is significantly influenced by the complexity of hierarchical, ethnic, and tribal aspects as well as communal, religious, and state rights. This complicates the development of large-scale ecological projects.”

Frontier Farmer-managed natural regeneration is the intervention most practitioners name as having the best cost-effectiveness evidence in the Sahel, and this brief states no figure for it. Protecting and pruning existing rootstock and volunteer woody regrowth on farmland, rather than raising and out-planting nursery stock, is widely argued to dominate planting on cost per surviving stem and on area covered. Speculative The retrieval pass behind this brief obtained no primary source on it, so the honest treatment is to name the claim, name the missing measurement — cost per surviving stem, and hectares under regeneration against hectares planted — and refuse to publish a number. Frontier What can be said with sources is that everything adjacent points the same way: the highest survival in the retrieved literature came from stakes and direct seeding rather than nursery seedlings; the measurable Mauritanian result was livestock exclusion; and the programme's own framing shifted to tree-based development. Speculative If the claim survives auditing, the Great Green Wall's central design decision was wrong for two decades in a way parts of the field already suspected.

Frontier Planting has measured negative effects on some native fauna. Licata, Thiaw, Ndiaye, Andreone and Ficetola (2025), in Restoration Ecology, surveyed eight reforestation plots in northern Senegal: woody plant density inside plots was significantly higher than in the surroundings, and negatively related to amphibian abundance, with predicted decreases of 74% for Tomopterna milletihorsini and 29.8% for Sclerophrys pentoni within plots. Their framing is worth quoting: tree planting “is increasingly called into question, as it may lead to abrupt habitat changes, with unpredictable effects on faunal communities.” Established Planting trees on land that was not forest is a habitat conversion, not a null operation. Frontier One study, eight plots, one country, carrying a great deal of weight — but the principle it illustrates has a canonical citation nine years older, which section 14 takes up.

Frontier Restoration is a trade, not a free win, and the evidence base is narrower than the programme. Wang, Dallimer and Scott (2025) assessed ecosystem services across Great Green Wall areas for 2007–2019 — carbon sequestration, soil conservation, sand fixation, water regulation, food provision — and found “synergies between ecosystem services are dominant,” strongest between carbon sequestration and soil conservation, but with trade-offs between those services and food production in certain regions. Established Bruckmann and colleagues separately found the research effort itself lopsided: a predominance of ecological studies and “a concentration of studies in certain geographies of interest, such as northern Senegal.” The evidence base for an eleven-country programme is drawn disproportionately from one country's fenced plots — and a Ferlo insect survey covering more than 500 morphospecies across nine Orders found site effects “much less marked” than season effects, the restoration being too recent to have changed diversity at all.

Established One measured result nobody has explained. Land surface temperature across the zone shows daytime cooling of -0.02052 K per year (p = 0.004) and night-time warming of +0.05618 K per year, the night-time term about 2.7 times the daytime one. Frontier Daily means hide it, and the retrieved literature offers neither mechanism nor heat-stress consequence: a significant satellite result with no follow-up.

Frontier Counterfactual causal inference has arrived in restoration ecology and is producing uncomfortable answers. Matching plus difference-in-differences on species distribution model output is a step-change away from before-and-after greenness, and the first thing it delivered was three countries with three different signs. Frontier Expect the next five years to be a sequence of programmes discovering that their headline improvements do not survive a matched control.

Frontier The albedo-versus-carbon tradeoff is genuinely unresolved in drylands, and is usually asserted with a confidence borrowed from boreal forests. Zhu, Chen and Wu (2024) found albedo in East Asian dryland grasslands driven primarily by vegetation index and vegetation water content in the growing season, and estimated the radiative impact at 0.004 to 0.113 kg CO2-equivalent per square metre per year. Established That range spans a factor of about 28. Frontier A number with a factor-28 spread cannot support a blanket claim either way, which makes “albedo cancels the carbon benefit in drylands” unsafe rather than false.

Frontier And part of the albedo penalty appears to cancel itself. Cerasoli, Yin and Porporato (2020) report that increased forest evapotranspiration “causes an anticipation in the crossing between the ABL and the lifting condensation level” — earlier cloud formation, whose cooling offsets albedo-induced warming. Their title is the claim: land-atmosphere feedbacks reduce the positive albedo forcing of afforestation. Frontier This is a single modelling study and it weakens one of the most-used sceptical arguments about tree planting. It needs replication across dryland hydroclimates before it can carry weight, and it has not had it.

Established Where the albedo objection is strongest is where snow is. At high latitude the term is dominated by snow masking, and the snow-albedo physics is itself sensitive to deposited absorbing aerosol: Pu, Shi and Cui (2020) find black-carbon coating enhances snow albedo reduction by a factor of 1.1 to 1.8 for a non-absorbing shell. Frontier The correct move is to locate the objection where it holds — boreal afforestation — rather than exporting boreal-strength confidence to the Sahel.

Frontier The Sahara wind-and-solar result is the field's most-cited piece of good news, and its principal follow-up is bad news. Li, Kalnay, Motesharrei, Rivas, Kucharski, Kirk-Davidoff, Bach and Zeng (2018), in Science, modelled large-scale wind and solar generation in the Sahara and found it “could cause more local rainfall, particularly in the neighboring Sahel region” through increased surface drag and reduced albedo, with a positive feedback via vegetation expansion. Frontier Lu, Zhang, Miller, Zhang, Berntell and Smith (2021), in Geophysical Research Letters, modelled the same class of intervention and found the regional rainfall and vegetation gains real but accompanied by a “redistribution of precipitation causing Amazon droughts and forest degradation, and global surface temperature rise and sea-ice loss, particularly over the Arctic,” plus northward expansion of Northern Hemisphere deciduous forests and enhanced tropical cyclone activity. Established Citing the first without the second is the commonest error in popular treatments of Saharan greening.

Frontier The companion preprint adds a result with almost no follow-up: the intervention degrades its own industry elsewhere. Global cloud cover is affected, and solar potential in other heavily solar-powered regions is altered by the circulation response. Speculative A Saharan solar megaproject large enough to green the Sahel would change the yield of solar installations on other continents. Nobody has costed that novel category of externality.

Frontier The class of result is not unique to the Sahara. Chen and Xie (2010) found the climatic influence of inter-basin water transfer “positively related to the quantity of water to be transferred”: large surface and water interventions perturb regional climate in proportion to their size. Frontier That implies a self-limiting scale for continental engineering which appears in no classical design methodology.

Frontier Satellite gravimetry is becoming a vegetation instrument. Velicogna and colleagues (2020) report that terrestrial water storage “provides unique information about deeper water storage limiting grassland growth,” and machine-learning downscaling has since taken it from half-degree to 0.1° over Egypt with correlation up to 0.91. Frontier The storage term that limits dryland vegetation is now partially observable, and restoration planning is not using it to site anything.

Established The water underneath the Sahara is being drawn down, unevenly. Frappart and colleagues (2020) report that over 2003–2016 the Northwestern Sahara Aquifer System lost above 30 cubic kilometres and the Nubian Sandstone Aquifer System around 50 — attributing the first to both climate and abstraction, the second to abstraction alone. Established That is the difference between a rechargeable resource overdrawn and a fossil resource mined. Frontier Greening the Sahara from groundwater means accelerating the second.

Established A discipline on all of the above. Strassberg, Scanlon and Rodell (2007) found, over the 450,000 km² High Plains Aquifer, a correlation of 0.82 between seasonal storage from gravimetry and summed storage components, but only 0.58 against measured groundwater. The satellite measures total mass; partitioning it is a modelled step.

Frontier The largest open item is a study nobody has run. Li et al. (2018) and Lu et al. (2021) are single-model studies reaching opposite overall verdicts on whether continental-scale Saharan surface change is desirable, and no multi-model intercomparison of Saharan surface-property forcing appears to exist. Established Given that the answer governs whether an entire class of intervention is a good idea, that is a striking hole in the literature rather than a technicality.

Established The workback plan for a greened Sahel that persists without maintenance has twelve links, and the binding one is not scientific. They are given here in order; the experiments that close them are in section 6.

Frontier Links 1 to 2, attribution. Establish whether observed greening is programme or climate, by applying matched-control counterfactual designs across the whole Great Green Wall zone rather than three countries; and resolve the Nigeria contradiction by co-measuring greenness and counterfactual biodiversity on the same plots. Established Both are unresolved today, and the second is cheap.

Frontier Links 3 to 5, measurement that is simply not being done. Dryland-specific survival curves at five and ten years by species and establishment method; a test of whether direct seeding's cost-effectiveness advantage transfers from tropical seasonal forest to Sahelian dryland, where seed predation and establishment windows are far harsher; and an honest net radiative accounting of sequestration minus albedo forcing in common units, including the cloud offset. Established None of these requires new science. All three are blocked on nobody funding a decade-long survival cohort.

Frontier Links 6 to 8, the deep unknowns. Establish plant-available deep water storage and its trend from downscaled gravimetry; test whether the vegetation-rainfall feedback can self-sustain, that is, whether a second stable state exists at all; and bound the teleconnection cost with a multi-model exercise over Amazonia and the Arctic. Speculative Link 7 is the core question for the words without maintenance, and link 8 is the one that could make success globally unacceptable.

Established Links 9 to 12, the institutional chain. Secure offtake governance so that grazing and fuelwood removal stay below regrowth; compensate the households bearing the food-production loss; stop converting non-forest habitat that has its own value; and fund the programme to completion against a disclosed disbursement record rather than a pledge. Frontier Link 9 binds, with link 10 as its enabling condition. The direct evidence is in the record: what measurably worked in Mauritania was fencing, livestock exclusion and dune stabilisation, and the obstacle named by the same authors is a land-ownership regime mixing hierarchical, ethnic and tribal claims with communal, religious and state rights. Established Plantings in the Sahel fail overwhelmingly because people and livestock use the trees. That is a tenure and compensation problem wearing a silvicultural costume.

Established The dependency that matters most is a ground-truth network that does not exist. The people trying to measure the programme state the problem themselves: “Because of limited ground measurements for the GGW project, the progress and impacts of the GGW initiative have been a challenging problem to monitor and assess.” Established Every hectares-restored figure, including the 20 million, inherits that sentence.

Frontier Second, model capability for Sahelian rainfall. The vegetation-circulation hypothesis requires a convective parameterisation trustworthy enough to simulate Sahel precipitation, which is among the least reproducible regional signals in the model ensembles. Speculative Until that improves, the Saharan surface-forcing results remain single-model claims about a quantity the models are known to disagree about.

Frontier Third, hydrological observation. Downscaled gravimetry at 0.1°, and groundwater schemes embedded in land-surface models, are the instruments that would let anyone site restoration by plant-available water rather than by administrative boundary. Frontier Both are newly available and neither is in programme use.

Established Fourth, and only for the irrigated variants, the water briefs. Any scheme that greens hyper-arid ground by importing water depends on results argued in Water Infrastructure Megaprojects and Continental Irrigation Systems rather than here. Frontier The warning those briefs carry is worth importing: an empirical assessment of built inter-basin transfer found it “not only fails to improve water consumption…but also further restricts residents' water use due to the high water price caused by this project.”

Frontier Fifth, a dependency on statistical practice rather than on any laboratory. The counterfactual toolkit that produced the only causal estimates in this subject — propensity matching, difference-in-differences, ensemble species distribution models — is borrowed wholesale from programme evaluation in economics and epidemiology. Speculative Its transfer into restoration ecology is recent enough that its assumptions have not been stress-tested on this kind of data: matched controls in a landscape where the treatment was assigned by accessibility and politics rather than at random are exactly where such designs are known to break. Frontier That is a reason to treat the three-country result as the best available evidence and not as settled, and it is a research dependency nobody currently owns.

Established The experiments below are ordered by leverage per dollar, and the first four are all cheap.

Frontier One: co-measure greenness and counterfactual biodiversity on the same plots. Nigeria and Senegal, matched controls, NDVI or biomass and species richness from the same footprints in the same years. This resolves the sharpest contradiction in the subject and requires a field season, not a programme.

Frontier Two: a dryland survival cohort. Tagged stems across Sahelian sites, survival reported at five and ten years by species and by establishment method, published as an open register. The benchmark to beat is the 40–70% wet-tropical range; nobody knows the Sahelian number, and every carbon claim about the programme depends on it.

Frontier Three: direct seeding against seedlings at Sahelian scale. Randomised, reporting cost per established stem at five years. The tropical result — twice the cost-effectiveness — is the highest-leverage untested transfer in the field.

Speculative Four: farmer-managed natural regeneration head-to-head against planting. Cost per surviving stem and area under each, on comparable land, with the same measurement horizon. This is the number this brief could not obtain and the number that would most change programme design if it came out the way practitioners expect.

Frontier Five: net radiative accounting at programme scale. Sequestration minus albedo forcing in kilograms of CO2-equivalent per square metre per year, using the dryland grassland method, with the evapotranspiration-driven cloud offset included as a stated term rather than omitted.

Speculative Six: the missing intercomparison. A multi-model exercise on continental-scale Saharan surface-property forcing, with pre-registered diagnostics over Amazonia, the Arctic sea-ice margin, and the solar resource of other continents. One protocol, several models, one question: are the teleconnections robust or model-specific?

Speculative Seven: interactive-vegetation regional modelling for bistability. Does a greened Sahel have a second stable state, or does it relax back once maintenance stops? Frontier Eight: replicate the amphibian result across more countries and plot ages, since one study with eight plots currently carries the entire faunal-cost case. Frontier Nine: measure offtake — grazing and fuelwood removal against measured regrowth, on plots differentiated by tenure instrument. Established Ten: publish a disbursement ledger against the 19 billion dollars pledged for 2020–2025. That one is not an experiment; it is a spreadsheet somebody already has.

Established The engineering problem in dryland restoration is not the tree. The intervention with a measured soil result in the Mauritanian programme was fencing, livestock exclusion and dune stabilisation. The engineering that matters is therefore exclusion infrastructure and its maintenance, at lengths measured in thousands of kilometres and lifetimes measured in decades, in places where the fence has an opportunity cost to somebody with animals.

Frontier Establishment technology is the second lever and it points away from nurseries. Stakes outperformed seedlings on survival and carried 7 to 50 times the biomass; direct seeding beat seedling planting on both survival and cost in the one tropical comparison available. Frontier Scaling either at Sahelian scale is a seed-supply and handling problem — collection, storage, viability, predation — rather than a horticultural one, and no retrieved source costs it.

Established Monitoring is the third, and it is the one the field concedes it lacks. A ground-truth network dense enough to convert satellite trends into audited surviving cover is a civil-engineering and staffing problem with a known cost structure and no funder. Frontier Downscaled gravimetry at 0.1° and sub-basin storage trend fields are now good enough to be operational siting tools; making them operational is software and institutional plumbing rather than science.

Speculative For the irrigated variants, the engineering envelope is set elsewhere and it is unforgiving. Desalinated water carries a levelized cost of roughly 0.5 to 2.0 US dollars per cubic metre, driven, as the review that reports it says in the same sentence, by energy consumption. Established Irrigation water is bought one to two orders of magnitude below that, conveyance and lift from coast to farmland is a separate megaproject usually omitted from quoted costs, and the brine stream runs at about 1.5 litres per litre of product against ecological thresholds of roughly one gram per litre above ambient salinity. Handwave Desalination-irrigated desert agriculture is not impossible; it is an energy-price and conveyance bet dressed up as a membrane story.

Established Water supply. Every hyper-arid greening proposal is a water-supply proposal, and its constraints belong to Water Infrastructure Megaprojects and Continental Irrigation Systems: desalination cost and its energy dependence, brine disposal thresholds, and the finding that a built transfer scheme can raise water prices enough to reduce the water use it was meant to increase.

Frontier Satellite gravimetry and land-surface modelling. Terrestrial water storage as a vegetation-limitation diagnostic, machine-learning downscaling to 0.1°, spatially refined trend fields, and groundwater compartments embedded in river-routing models are all becoming operational at once. This is the adjacent field most likely to change how restoration is sited within a decade.

Frontier Renewable-energy megaprojects. The Saharan wind-and-solar literature is the point where energy buildout and desert greening become the same subject, with the same teleconnections. That makes Climate Engineering and Geoengineering Governance direct neighbours rather than distant ones: an intervention that changes Amazonian rainfall is governed by the same consent problem whether it is intended as energy or as restoration.

Established Carbon accounting and biodiversity policy. Carbon Capture at Scale inherits the survival and albedo problems described here whenever land-based sequestration is credited; Biodiversity Restoration inherits the counterfactual-inference methods and the finding that greenness and biodiversity can move in opposite directions. Frontier Future Agriculture holds the other half of the carbon-versus-food trade-off, and Weather Modification holds the adjacent question of whether regional precipitation can be deliberately altered at all.

Frontier Programme evaluation methods. The most consequential import into this field in the last three years was not ecological at all: it was matched-control causal inference, arriving from economics and epidemiology and immediately overturning the before-and-after greenness narrative. Speculative The neighbouring subject with the most to gain from watching this happen is any large environmental programme that currently reports its own outputs as its own outcomes, which is most of them.

Established Three institutional conditions govern this subject, and none of them is a research result. The first is land tenure. The Sahelian ownership regime mixes hierarchical, ethnic and tribal claims with communal, religious and state rights, and the paper that names this is the same paper reporting that the interventions which worked were fencing and livestock exclusion. Established Excluding livestock requires the authority to exclude, which is a tenure instrument, and the absence of that authority is why plantings are grazed.

Established The second is the measurement standard. Hectares pledged, hectares under intervention, hectares planted, and hectares restored and surviving are four different quantities, and the headline figure is routinely one of the first three reported as the last. Frontier Changing the governing metric from area treated to surviving cover is an accounting decision, not a scientific one, and the field published the argument for it in 2017.

Established The third is disbursement. Nineteen billion dollars was pledged in 2021 for 2020–2025; the peer-reviewed source reporting the pledge explicitly does not report disbursement rates, and the assessment of global restoration finance is titled around the finding that such finance needs to quadruple by 2030. Frontier A public ledger of money actually moved against that pledge would be the single most informative document anyone could publish about this programme.

Frontier A fourth condition applies only to the continental variants, and it has no existing venue. If Saharan surface change alters Amazonian rainfall and Arctic sea ice, then large-scale desert greening is a global intervention requiring global consent — structurally the same blocker that has kept every continental water-transfer scheme unbuilt. Speculative There is no treaty instrument for a land-cover change with intercontinental precipitation effects, and no forum in which Brazil could object to a Saharan solar buildout on hydrological grounds.

Established And a condition on the evidence itself. The primary sources that hold the audited record for this programme — UNCCD, FAO, and the Accelerator's own tracking — were unreachable during the research behind this brief. That is a statement about one retrieval pass, not about the documents. But it is also the ordinary condition of public discussion of the Great Green Wall, in which secondary reports of a five-year-old progress document stand in for an audit.

Frontier The carbon-versus-food trade-off has a distributional answer that nobody states. Ecosystem-service mapping across the programme found trade-offs between carbon and soil services and food production in some regions. Established Carbon accrues to a global account and food accrues to a household; restoration that improves the first at the expense of the second is a transfer from Sahelian farmers to distant beneficiaries unless compensation flows the other way. Speculative Whether such compensation exists at scale is not established in the retrieved literature, and it is the mechanism most likely to determine whether restored land stays restored.

Established Fencing is an exclusion, and what it excludes belongs to someone. The intervention with the clearest measured soil benefit works by preventing the passage of livestock. Pastoralists are not an externality, and a restoration programme whose effective mechanism is denying access to grazing land is making a political allocation, not applying a technique.

Frontier Habitat conversion is the ethical problem the field named for itself and then ignored. Afforesting savanna is not restoration; measured amphibian declines of 74% and 29.8% inside Senegalese planting plots are the local price. Established The restoration literature's own principle — that afforestation should not replace native grasslands or savanna ecosystems — has been in print since 2017.

Established Reporting euphemism is an ethical matter, not a stylistic one. A large, highly significant negative treatment effect described in an abstract as “limited benefits” changes what a policymaker reading only the prose will conclude. Frontier The remedy is procedural: report effect sizes and p-values in abstracts, and treat a directional adjective without its number as unreported.

Speculative Finally, consent at range. If continental surface engineering dries the Amazon, the people harmed are on another continent and have no standing anywhere. That asymmetry is the ethical core of the exotic end of this subject and it is unaddressed.

Established Drylands are not marginal land; they are where a large fraction of humanity farms and grazes, and where degradation and recovery are both fastest. The measured Great Green Wall record — a small but significant regional greening, a desert margin retreating across four-fifths of a transect, a fifth of the target restored at audit — is neither the triumph nor the fiasco it is alternately reported as. It is a partial, unevenly distributed, poorly monitored improvement bought at a cost nobody has fully accounted.

Frontier The civilizational question is whether dryland productivity is a technology or a governance form. Everything measured in this brief points at the second. The interventions that worked were exclusions; the obstacle named by practitioners is tenure; the metric that governs the field was argued against nine years ago and did not change. Established A civilization that could reliably keep offtake below regrowth on communally held land would not need most of the silviculture.

Speculative The exotic reading is that the Sahara may be a switch. If the vegetation-rainfall feedback is real and strong enough to admit a second stable state, then a sufficiently large surface intervention is not a restoration project but a climate reconfiguration, and someone will eventually want to flip it. Handwave The 2026 reappearance of Saharan hydropower and depression-filling schemes under a Green Sahara framing suggests the appetite outlives the evidence by a wide margin.

Frontier And the sober reading is that the same result makes the switch unusable. The best available modelling says a greened Sahara comes with Amazonian drought, Arctic sea-ice loss and enhanced tropical cyclones. Speculative A civilization capable of greening a continent would, on current evidence, be choosing which continent to dry in exchange.

These horizons track what could be measured and decided, not what could be planted. The programme deadline is the near-term anchor; everything past it depends on results that are not being produced.

• 10 yr: Established The Great Green Wall's 2030 deadline arrives inside this window, and on the only audited trajectory available — 20% of hectares and 30% of carbon at the 2020 report — it will not be met. Frontier An independent post-2020 audit, the plot-level co-measurement that resolves the Nigeria contradiction, and the first dryland survival curves at five years are all achievable this decade and require no new science. Frontier The multi-model intercomparison of Saharan surface forcing is likewise a protocol-and-funding problem, not a research one. Speculative Expect the counterfactual literature to keep producing heterogeneous and partly negative results, and expect the hectares metric to survive them.
• 25 yr: Frontier This is the window in which the maintenance question is answered empirically: whether areas restored in the 2010s and 2020s still carry cover once fencing, subsidy and project attention lapse. Speculative It is also the window in which restoration finance either quadruples, as the global assessment says it must, or does not, and in which surviving-cover accounting either replaces area accounting or is finally abandoned as unenforceable. Speculative If farmer-managed regeneration is audited and confirmed as dominant, programme design changes here rather than earlier.
• 50 yr: Speculative Saharan renewable buildout could plausibly reach the areal scale at which the modelled circulation response stops being hypothetical, which turns the teleconnection question into a live diplomatic one. Speculative Whether the vegetation-rainfall feedback admits a second stable state should be settled by then, by models with credible Sahelian convection rather than by argument. Handwave A deliberately maintained green belt of continental extent is conceivable in this window; a self-maintaining one is not, on any current evidence.
• 100 / 250+ yr: Handwave A genuinely self-sustaining green Sahel, held by its own rainfall feedback rather than by fences and budgets, is the version of this subject that is currently considered impossible. Handwave It requires a positive feedback strong enough to be bistable, survival past establishment that has never been measured in drylands, offtake governance sustained over centuries, and a global consent regime for an intervention that redistributes rainfall between continents. Speculative Each of those is separately unproven; the interesting thing is that none of them is separately absurd.

• Depends on This brief waits on no result that another brief on this map is producing, which is itself the finding. Its dependencies are a measurement network, a metric, and a tenure regime — none of them a scientific unknown. The nearest genuine scientific dependency is model capability for Sahelian convection, because the vegetation-circulation hypothesis animating the exotic end of the subject is a claim about the least reproducible regional precipitation signal in the ensembles, and no amount of restoration work can settle it. The nearest instrumental dependency is downscaled satellite gravimetry, which has only just reached the resolution at which plant-available deep water storage could be used to site interventions. Everything else this brief needs already exists and is not being applied.
• Requires (not on this map) Three constraints hold this subject, and all three sit outside science. The first is tenure: the intervention with the clearest measured soil benefit in the Sahel works by preventing the passage of livestock, and excluding livestock requires an authority to exclude that the region's mixture of hierarchical, ethnic, tribal, communal, religious and state rights does not reliably confer. The second is the metric: as long as programmes are governed by hectares treated rather than by surviving cover, pledged area will continue to be reported as restored area, and the field has known this since 2017. The third is money that moves: nineteen billion dollars was pledged in 2021 for 2020–2025, the source reporting the pledge does not report disbursement, and no public ledger of what was actually spent could be obtained. Fix all three and the ecology becomes tractable; fix none and better silviculture changes nothing.
• Enables What this subject supplies to its neighbours is mostly a discipline and a warning. To carbon accounting it supplies the rule that a sequestration claim without a survival assumption, a measurement horizon and an albedo term is not auditable. To biodiversity work it supplies the most concrete available demonstration that greenness and biodiversity can move in opposite directions in the same country, and therefore that vegetation indices are not a biodiversity proxy. To energy megaproject planning it supplies the result that continental surface change has intercontinental precipitation consequences, converting a siting decision into a consent problem. And to restoration policy generally it supplies a worked example of a correction published in a field's flagship journal, in the title of the paper, then ignored for nine years while the metric it argued against went on governing the money.
• Adjacent Water Infrastructure Megaprojects and Continental Irrigation Systems for the irrigated variants; Carbon Capture at Scale for the sequestration accounting; Biodiversity Restoration for the counterfactual methods; Climate Engineering and Geoengineering Governance for the teleconnection and consent problems; Future Agriculture for the other half of the carbon-versus-food trade; Weather Modification for the adjacent question of deliberate precipitation change.

Established “The Great Green Wall is restoring 100 million hectares.” One hundred million hectares is the target. The UNCCD's 2020 progress report, as reported in the peer-reviewed literature, recorded about 20 million hectares restored and 300 million tonnes of CO2 sequestered, 30% of the carbon target. Established Hectares pledged, hectares under intervention, hectares planted, and hectares restored and surviving are four distinct quantities and the popular figure is routinely one of the first three reported as the last. Frontier Even the 20 million inherits the field's own stated limitation: “Because of limited ground measurements for the GGW project, the progress and impacts of the GGW initiative have been a challenging problem to monitor and assess.” Pair the target with the audited figure and the date of the audit, every time.

Established “Nineteen billion dollars was committed, so the programme is funded.” It was pledged, at the 2021 One Planet Summit, for 2020–2025. The source that reports the pledge explicitly does not report disbursement. Established The global assessment of restoration finance is titled around the finding that such finance needs to quadruple by 2030. Frontier Pledged and disbursed are different quantities and this brief states only the first.

Established The canonical correction to this field already exists, is in the title of the paper, and was ignored. Brancalion and Chazdon (2017), in Restoration Ecology, published “Beyond hectares: four principles to guide reforestation in the context of tropical forest and landscape restoration.” Its principles include that afforestation “should not replace native tropical grasslands or savanna ecosystems” and that “residual carbon stocks should be quantitatively and qualitatively distinguished from newly established carbon stocks.” Established The field's flagship restoration journal published an explicit argument against hectares as the governing metric nine years ago, and hectares remained the governing metric. Established The second principle names an accounting error — counting pre-existing carbon as newly established — structurally identical to counting pledged hectares as restored hectares. Frontier And the first principle now has corroborating measurement in the same journal: planting density inside Senegalese plots predicted amphibian abundance declines of 74% and 29.8%. Established The correction was not missing. It was published, in the right venue, by the right people, and the money kept being counted the old way.

Established “Planting N trees sequesters M tonnes of carbon.” Two omissions make such claims unauditable. Survival: most species show 40–70% survival even in wet tropical restoration, the dryland figure is unmeasured, and carbon claims are typically computed on stems planted. Established Albedo: dryland grassland forcing runs 0.004 to 0.113 kg CO2-equivalent per square metre per year and must be netted off. Frontier A carbon claim citing neither a survival assumption nor an albedo term is not a carbon claim.

Frontier “Albedo cancels the carbon benefit of afforestation.” Over-applied. The objection is strongest where snow masking dominates, and the snow-albedo term is itself sensitive to deposited black carbon. Frontier In drylands the forcing spans a factor of about 28 between studies, and land-atmosphere cloud feedbacks have been shown, in at least one modelling study, to reduce the positive albedo forcing of afforestation through earlier cloud onset. Established Locate the objection where it holds rather than exporting it.

Established “Greening the Sahara would be unambiguously good.” The leading modelling of continental-scale Saharan surface change finds Amazon drought and forest degradation, global surface temperature rise, Arctic sea-ice loss and enhanced tropical cyclone activity, and a companion analysis finds the solar resource of other regions altered. Frontier The optimistic 2018 result and the pessimistic 2021 result are both single-model studies and no intercomparison exists to adjudicate them.

Established “More vegetation means more water.” Across the Great Green Wall zone, NDVI rose while precipitation fell significantly and soil moisture trended down. Established “Vegetation cooling is straightforward.” Daytime land surface temperature fell 0.021 K per year while night-time temperature rose 0.056 K per year — the night-time term is about 2.7 times larger, and daily means conceal it. Frontier “Restoration is a win-win with livelihoods.” The ecosystem-service assessment finds trade-offs with food production in some regions. Established “The evidence base covers eleven countries.” Bibliometric analysis finds studies concentrated in certain geographies, notably northern Senegal, with a predominance of ecological work.

Handwave “Desalination could irrigate the desert.” Desalinated water costs roughly 0.5 to 2.0 US dollars per cubic metre, and the review reporting that band gives the cause in the same sentence: energy. Irrigation water is bought one to two orders of magnitude cheaper, coast-to-farmland conveyance is a separate megaproject left out of quoted costs, and brine runs at about 1.5 litres per litre of product against thresholds near one gram per litre above ambient salinity. Speculative The binding terms are energy price and conveyance, not membrane science.

Handwave “Once green, the Sahel would stay green.” This is the version of the subject currently considered impossible, and it is worth stating why rather than dismissing it. It needs a vegetation-rainfall feedback strong enough to constitute a second stable state, resting on one modelling lineage and no intercomparison; survival past establishment, never measured in drylands; and offtake below regrowth indefinitely, which is governance rather than ecology. Speculative None of the three is absurd; all three are unmeasured, and the third is the one the programme record shows failing most reliably.

Frontier A note on how citations in this field get corrupted. During the research pass behind this brief, a bibliographic query about reverse-osmosis energy returned a result set about organ cryopreservation — a silent substitution in the retrieval layer, with no error and no low-confidence signal. Three other passes in the same session caught their own tools substituting one document for another. Established This matters because dryland restoration is governed by assessment reports assembled through exactly that kind of machine-assisted citation harvesting, and the failure leaves no trace in the finished document. Frontier It is why this brief prefers a peer-reviewed paper reporting the UNCCD audit second-hand, with that status stated, over a confident figure of unknown provenance.

Established And a note on absence. Where this brief says a figure could not be obtained — disbursement against the pledge, dryland survival curves, the cost per surviving stem under farmer-managed regeneration — that is a statement about one retrieval pass, not a claim that the number does not exist. Frontier Each has a named instrument: the UNCCD status report and Accelerator tracking, a decadal survival cohort, and a head-to-head field comparison.