Biodiversity Restoration

Established Biodiversity restoration is the attempt to reverse a trajectory, and almost everything the world does under that heading measures an input rather than an output. The dominant instrument is the protected area, and the dominant metric is how much of the planet has been legally designated as one. Established Designation is a legal act performed by a ministry; protection is a physical outcome delivered by people standing on ground, and the measured gap between the two is not a rounding detail. It is the subject.

Established The single number that reorganises this field is a headcount. Appleton and colleagues surveyed 176 countries and territories covering 64.4 percent of the world's protected land and found roughly 555,000 protected-area personnel, one per 37 km², of whom about 286,000 are rangers, one per 72 km². To manage the planned 30 percent coverage by 2030 they estimate a need of about three million personnel, one per 13 km², including more than 1.5 million rangers, one per 26 km². Established That is roughly 5.4 times the current workforce and 5.2 times the current ranger corps — arithmetic on two sourced figures, not a separate claim. Established And staffing is not a hygiene factor: Gill and colleagues found that marine protected areas with adequate staff capacity had ecological effects 2.9 times greater than those without. Frontier A brief that treats holding biodiversity as a conservation-science problem mislocates the binding link by an entire discipline. It is a problem in public administration and recurrent finance.

Established A note on what this brief will not tell you. The research behind it ran under a retrieval gate that refused every primary conservation agency: Protected Planet, the Convention on Biological Diversity, IUCN, UNEP-WCMC and IPBES were all unreachable. Established So the current global protected-area coverage percentage is not printed here, and the verbatim text of Kunming-Montreal Target 3 is not quoted here; both are declared gaps with a named route to closure, not omissions of convenience. Frontier The argument this brief makes does not require a current percentage; it requires the structure, and the structure is in the peer-reviewed record.

Established Start with the word that disappears. Mouillot and colleagues open their 2024 analysis by stating that more than 17 percent of the world's terrestrial surface and 8 percent of the ocean are “formally designated” as protected, and they attribute that pair to Maxwell and colleagues in 2020. Established Two separate defects are visible in that one sentence. The first is staleness: a 2024 paper in a leading journal is transmitting four-year-old coverage figures by citation rather than re-querying the underlying database, and the figures carry no base year at the point of use. The second is the qualifier: the source says designated, and downstream the word reliably falls off and the number is read as protection delivered. Established This brief quotes the pair only as it appears in the literature — undated at source, second-hand within it — and states plainly that the current, dated figure from the World Database on Protected Areas could not be obtained. Frontier Anyone quoting a coverage percentage as a measure of protection is quoting an input as an output, and the rest of this section is about how large that substitution error is.

Established The paper-park problem has been measured, and the measurement contains a dose-response. Gill and colleagues assembled management data from 433 marine protected areas and fish population data from 218 of them. Established Many fell short of benchmarks for effective and equitable management, with pervasive staff and budget shortfalls, and about 71 percent showed positive effects on fish populations with substantial variability. Established The finding that matters most is the conditional one: MPAs with adequate staff capacity had ecological effects 2.9 times greater than those with inadequate capacity. Frontier A dose-response is what distinguishes a working mechanism from a correlation. It means the instrument is not broken; it means the instrument is unpowered. Established Watson and colleagues stated the structural version a decade earlier — financial support is dwarfed by the benefits protected areas provide, but those returns depend on effective management — and the reporting metric did not change.

Established The best counterfactual test of whether protected areas work returns a mixed and uncomfortable answer. Santangeli and colleagues compared occupancy in protected and unprotected sites across four decades in Finland for 638 species — 150 birds, 23 mammals, 39 plants and 426 phytoplankton. Established Positive responses to protection: 19 percent of birds, 13 percent of mammals, 3 percent of plants, 4 percent of phytoplankton. Benefits increased with protected-area size and were traceable to the timing of protection. Established The authors describe the benefit mechanism explicitly as “slower rates of decline inside protected areas,” and conclude that the network can partly slow declines but alone will not halt the biodiversity crisis. Frontier Read that carefully: the measured effect is decelerative, not restorative, and a network that decelerates loss everywhere still loses monotonically. It changes the slope, not the sign. The taxa that did worst were the 426 phytoplankton species, the least charismatic and least monitored, at 4 percent.

Established Now the workforce, because this is the brief's thesis and it is arithmetic rather than argument. Appleton and colleagues built a 176-country census covering 64.4 percent of the world's protected land: approximately 555,000 personnel at one per 37 km², of which 286,000 rangers at one per 72 km². Established The stated requirement for the planned 30 percent global coverage of protected areas and conservation measures by 2030 is approximately three million personnel at one per 13 km², including more than 1.5 million rangers at one per 26 km². Established Their diagnosis of the last two decades is the sentence to remember: staffing has “merely kept pace with area expansion without improving per-unit-area management,” which is to say the world has been adding hectares at the density of supervision it already had, and that by construction cannot raise effectiveness. Frontier Combine this with Gill's 2.9 times capacity multiplier and the policy conclusion is direct: the marginal hectare designated without a marginal salary attached buys close to nothing, and may buy less than nothing if it dilutes the budget across a larger estate.

Established Where protected areas get placed is not where biodiversity is, and this is now characterised rather than alleged. Mouillot and colleagues modelled the multidimensional socioeconomic and environmental niche of protected areas using fourteen factors, and found that protected areas, particularly the most restrictive categories, over-aggregate where human development and the number of non-governmental organisations are high. Established They separate high-vertebrate-diversity regions into those favourable to protected-area establishment — “potential” gains — and those unfavourable, which they call “unrealistic”; they find that IUCN category I areas cover only a reduced fraction of global social-environmental conditions; and they judge 30 percent marine coverage “hardly achievable” without overcoming socioeconomic barriers. Frontier The generous reading is that protection follows capacity, which is rational; the unforgiving reading is that the global estate is assembled from residual land — ground nobody else wanted — and that a target expressed in area will be met with whatever land is cheapest to designate. Speculative That NGO geography is causal rather than confounded is not established: it would require NGO presence to predict designation after conditioning on development and accessibility, and the partial effects behind that fourteen-factor niche were not obtained here.

Established Representation is short even where area is not. Maxwell and colleagues found persistent shortfalls in ecological representation and management effectiveness, and that expansion since 2010 had limited success in increasing coverage across ecoregions, 12,056 threatened species, Key Biodiversity Areas, wilderness areas and ecosystem services. Speculative The same paper carries the widely repeated statement that one-third of global protected land is under intense human pressure; that figure is attributed in-text to a reference not opened for this brief, so it is a claim in the literature here rather than a verified number. Established Di Marco and colleagues add that extinction risk in wilderness communities is on average less than half that in non-wilderness communities and that high-priority wilderness remains poorly protected, so the two estates cannot be added together. Frontier Maron, Simmonds and Watson put the general objection in print in 2018 — targets are needed for the retention of nature, not just its formal protection — and the headline metric did not move.

Established An area target can be met while ecological function is absent, and one study measured both in the same system. Endo and colleagues modelled demographic connectivity of reef fish across Brazilian marine protected areas between 10°N and 23°S. Established The Brazilian exclusive economic zone had recently achieved the quantitative requirement of Aichi Biodiversity Target 11, and its areas were nonetheless highly isolated by extremely low demographic connectivity: self-recruitment of 0.69 percent at Trindade in summer, 0.54 percent at Costa dos Corais in summer, 5.93 percent there in winter, and recruitment close to zero at most sites between 2008 and 2015. Frontier This is the cleanest published case of target achieved and function not delivered, measured in the same water in the same years, and it implies something about the unit of analysis: if self-recruitment is under one percent, the object being conserved is not the node, it is the flux between nodes.

Established Restoration, the other half of the subject, has one strong global optimisation and very weak cost data. Strassburg and colleagues found that restoring 15 percent of converted lands in priority areas could avoid 60 percent of expected extinctions while sequestering 299 gigatonnes of CO₂, which they characterise as 30 percent of the total atmospheric CO₂ increase, or 14 percent of total emissions, since the Industrial Revolution. Established Cost effectiveness increases up to 13-fold when spatial allocation is optimised, and multiple biomes must be included to obtain biodiversity and climate benefits together. Two qualifiers carry the whole magnitude: it is 15 percent of converted land, and it is in priority areas. Frontier The 13-fold allocation swing is what makes the second one load-bearing, because it says the same hectare count buys thirteen times more or less depending on where it lands.

Established The only hard restoration cost figure available here comes with a correction attached, and the uncorrected version is in the paper's title. Szalkiewicz, Jusik and Grygoruk analysed 119 river restoration projects implemented across 19 European countries between 1989 and 2016. Established The title figure is 310,000 euros per hectare of restored river, and reported unit costs ranged from 99,000 euros per hectare for projects in 2000 and earlier to 353,000 for 2006 to 2010, and amortised over a 40-year asset life the authors compute an annual ecosystem service value of 7,757 euros per hectare per year. Established When four statistical outliers were excluded, the average dropped to 195,000 euros per hectare — the headline is 59 percent above the corrected value. Frontier That is a 3.6-fold spread within a single intervention class on a single continent. Frontier A global per-hectare restoration cost aggregated across biomes, when one biome's own range is 3.6-fold and the reported mean moves 59 percent on an outlier decision, is not a usable planning number, and costs for other biomes were not obtainable for this brief.

Established Measured restoration outcomes exist, and they measure land cover rather than survival. Barasa, Gudoyi and Pule surveyed households adjacent to restored and unrestored wetland and forest ecosystems in Uganda — 315 valid questionnaires from 1,092 households contacted, across 12 wetland and 15 forest villages — alongside high-resolution imagery classified at 91 to 93 percent accuracy. Established Between 2023 and 2025 papyrus cover at Rwambu wetland rose above 45 percent while adjacent cropland fell 17.7 percent, and tree cover at Mutufu rose above 50 percent from a 23.63 percent baseline; ecosystem-service recovery was significantly greater in restored wetlands than forests. The paper contains no cost data and no vegetation survival rates, and says so. Frontier That absence is not a defect of this one paper; it is the field's standard shape, and a hectares-restored figure with no year-ten survival term is a planting statistic rather than a restoration statistic.

Frontier One variable appears twice in this record, and that is the finding. Ban and colleagues, synthesising 118 studies of the human well-being outcomes of marine protected areas, found that enforcement quality predicted favourable outcomes for people as well as for fish. Frontier Capacity explains the ecological result in Gill, the null result in Santangeli, the deceleration rather than recovery everywhere, and the social result too. What is missing from the world's biodiversity estate is not designation and not science. It is staffed, funded, continuous presence.

Established Three sensing modes have become cheap enough in a decade to change what a protected area could in principle report about itself. Environmental DNA metabarcoding recovers community lists from water, soil and air samples without catching or seeing anything; autonomous acoustic recorders run unattended for months and are read by neural-network classifiers rather than by ear; and airborne imaging spectroscopy resolves plant functional traits across whole landscapes at metre scale. Established The BioSCape campaign flew that combination over South Africa’s Greater Cape Floristic Region in late 2023, pairing imaging spectroscopy, thermal and LiDAR instruments with coordinated field survey; it is cited here for its design and its existence, because the campaign literature was not read for this brief and no result is reported from it. Frontier What has not changed is the thing this brief is about. Established A sensor is a capital purchase with an announceable installation date; a reference library, a classifier and a servicing round are recurrent costs that produce nothing announceable — the same asymmetry that leaves roughly 555,000 people standing in a network sized for three million.

Established An eDNA detection is a match against a library, and the library is the instrument. A sequence read can only be named if a reference barcode for that taxon exists in the database being searched; where it does not, the read is returned as an unassigned cluster, and the organism is present in the sample and absent from the result. Established Library completeness is deeply uneven by taxon and by region: vertebrates in northern temperate systems are well covered, invertebrates and tropical microbiota much less so. Frontier The operational consequence is a coverage bias running in the same direction as every other bias in this subject — toward the places and organisms already studied — and it is invisible in the output, because an unassigned cluster looks like noise rather than like a gap. Speculative No source obtained for this brief gives a per-region completeness figure for any major barcode reference database, and a network that published its own library completeness beside its species list would be reporting something closer to an error bar than anything the field currently prints.

Established Genetic rescue is the one intervention in this subject with a clean, decades-long demonstration, and it is small. Eight female pumas were moved from Texas into the Florida panther population in 1995; the admixed population subsequently showed higher heterozygosity, sharply reduced incidence of the kinked tails, cryptorchidism and cardiac defects that had accumulated under inbreeding, and roughly threefold growth in numbers. Established The Isle Royale wolf translocations of 2018 and 2019, following an inbreeding collapse in a closed island population, are the second case usually cited. Frontier Both are rescues of populations already at the bottom, which is why they are legible: the counterfactual was visible, the defect load was visible, and the response was fast. Speculative Neither case establishes that the same move works pre-emptively on a large, still-connected population, which is what assisted gene flow for climate adaptation proposes to do.

Established Three interventions are conflated under one heading, and they carry different risks, different consents and different law. Assisted gene flow moves genotypes within a species’ existing range to shift the local adaptive optimum; assisted migration, or assisted colonisation, moves a species beyond that range; assisted evolution selects or conditions organisms for a trait before release. Established Only the first has a large operational precedent, and it is not a conservation one: commercial forestry has moved tree seed along climate gradients under provincial and state seed-transfer rules for decades, and Canadian federal guidance treats assisted migration as an available forest-adaptation option rather than as a proposal. Frontier That precedent is narrower than it looks, because it rests on a planting industry that already owned the nursery, the seed zone and the recurrent budget. Speculative For a species with no commercial planting chain the identical intervention has no institution to sit in, and no source obtained for this brief describes one.

Frontier Rewilding is the most active frontier and the least consolidated, and the evidence base for it is thinner than its public profile suggests. A framework literature exists that treats trophic rewilding as an operational programme rather than a sentiment — Pedersen and colleagues' TRAAIL framework for advancing trophic rewilding in anthropogenically impacted landscapes, and Hoeks and colleagues' modelling of shifts in ecosystem equilibria following trophic rewilding. Speculative Both were verified as bibliographic records and neither text was read, so what can be said is that the papers exist, on those subjects, by those authors — no finding and no direction of effect is reported here from either. Frontier The equilibrium framing is the scientifically interesting one, because it converts rewilding from a species list into a state transition with predicted endpoints, which is falsifiable in a way a species list is not.

Frontier The core premise of trophic rewilding — that introduced large herbivores can substitute for extinct megafauna and deliver their ecological function — has a live empirical challenge and a live empirical defence, both in preprint. Speculative Bescond-Michel, Bacher and Vimercati report, by title, that harms of introduced large herbivores outweigh their benefits, with both greater on islands and at higher trophic levels. Dvorsky and colleagues report, by title, that rewilding with large herbivores increased plant species richness in dry grasslands. Speculative Rech and colleagues test habitat preferences and complementary niches of large herbivores inside an operating Danish rewilding area — a direct field test of the complementarity premise. Handwave None of these three texts was obtained; they are named as the shape of the disagreement, not as evidence in it. Frontier The binding scientific unknown behind all of them is whether a surrogate taxon delivers the function of the taxon it stands in for, or only the silhouette, and that has no settled answer for any surrogate anywhere.

Frontier Rewilding also has a counterfactual problem that one recent paper exists specifically to fix. Ceci's 2026 preprint establishes pre-intervention baselines — mammal density, occupancy and activity patterns — at two Italian candidate sites. Frontier A paper whose contribution is measuring the site before anything is done implies that this is usually absent, which is the standing methodological complaint against the field: without a baseline and a matched control, a rewilding project can report change but cannot report effect.

Handwave Oostvaardersplassen is the case everyone reaches for, and this brief will not reproduce it. The Dutch reserve, its herbivore population dynamics, its winter mortality, the culling decision and the successive independent review committees are the standard reference point for whether rewilding is ecological restoration or an aesthetic project with animals in it. Established No Oostvaardersplassen source was retrievable for this brief: two targeted bibliographic queries were rate-limited and a third returned unrelated European rewilding chapters. Here the contested numbers are the argument — mortality rates, carrying-capacity assumptions and what the reviews concluded are exactly what the two sides dispute — so writing them from memory would be worse than leaving the hole, and the hole is declared with its retrieval route.

Frontier Half-Earth has an independent Earth-system route to roughly its own number, arrived at without invoking its founder. Rockström and colleagues place the safe Earth-system boundary for intact natural ecosystem area at 50 to 60 percent of global land, with the just boundary at the upper end of that range, and record it as already exceeded; they add a functional requirement of 20 to 25 percent seminatural habitat per square kilometre in human-modified land. Frontier That convergence is worth noticing, and it is also fragile: the same research programme, with overlapping authorship, uses a different control variable for biosphere integrity elsewhere, with van Vuuren and colleagues operationalising it as Mean Species Abundance at 90 percent. Intact-area-fraction and mean-species-abundance are not the same quantity, do not respond to the same interventions, and cannot be converted into one another. Speculative Montoya, Donohue and Pimm published a critique titled Planetary Boundaries for Biodiversity: Implausible Science, Pernicious Policies; its record was verified and its arguments were not obtained, so it is cited as evidence that a serious disciplinary objection exists and not as a summary of what that objection says. Frontier A target of half the planet resting on a control variable with at least two operational definitions has a stable numerator and an unstable denominator.

Frontier The other-effective-area-based-conservation-measure question is the live institutional frontier. Mouillot and colleagues propose these measures, and privately protected areas, precisely for the regions where conventional protected areas are unrealistic. Frontier That is either the escape hatch that makes the 2030 target reachable or the accounting device that makes it meaningless, and which one depends entirely on whether a measurable standard distinguishes such an area from an unmanaged one, applied after the fact. Speculative Nothing retrieved for this brief describes such a standard in operation, though a named preprint literature on avoiding paper parks through socio-economic strategy was record-verified, which at least establishes that the failure mode has an organised research response.

Established The false-positive problem in eDNA is not a laboratory hygiene problem, and treating it as one is the commonest error in the applied literature. Contamination, tag jumping between multiplexed samples and amplification bias all produce reads that are real sequencing output and wrong ecology; so does a true detection whose source was upstream, dead, or arrived on a bird. Established The long dispute over silver and bighead carp eDNA in the Great Lakes basin turned on exactly that distinction: detections ran ahead of physical capture for years, and the argument was never about whether the assay worked but about what a positive meant. Frontier The methodological answer the field has converged on is to stop treating a positive as a detection and start treating it as an observation with a detection probability, estimated from replicated samples inside an occupancy model. Speculative Whether managers will accept a probabilistic detection as a trigger for action is a separate question, and no jurisdiction obtained for this brief publishes a written rule mapping an eDNA result to a management decision.

Frontier Integration across sensing modes is where the observatory idea is either real or decorative, and it is not an additive problem. Each mode carries its own detection function, spatial support and temporal grain: eDNA integrates over a catchment and a few days, an acoustic recorder over a few hundred metres and a season, a spectrometer over a pixel and an overflight. Established Stacking them without a joint observation model does not average the biases out; it transfers the strongest sampling bias into the fused product and dresses it as agreement. Frontier The statistical machinery for doing it properly exists — integrated distribution models that give each data stream its own likelihood over a shared latent state — and the binding constraint is not the mathematics but the metadata, because such a model needs each stream’s effort, protocol and software version recorded in a form it can condition on. Established This is why the Essential Biodiversity Variables framework puts a defined variable layer between raw observation and published indicator instead of letting indicators be computed from whatever was collected. Handwave Whether an observatory built this way could ever deliver the year-ten survival series restoration reporting lacks is undemonstrated, and the obstacle is again the funded decade of attention rather than the sensor.

Frontier Assisted evolution is furthest along in coral, and coral is the least representative system available. The programme — selective breeding for thermal tolerance, assisted gene flow between warmer and cooler reefs, and manipulation of the symbiont and microbial community — exists as a funded research line with published framework papers, and the reef is a rare case where the organism can be reared, crossed and outplanted at scale. Speculative Whether any of it changes reef-scale outcomes under the heat exposure actually arriving is untested, and the mismatch between hectares outplanted and the area bleaching in a single event is the objection that has not been answered. Frontier The transferable result is methodological rather than ecological: the field moved outbreeding depression from a default prohibition to a bounded risk with a published decision procedure, which is what made deliberate crossing thinkable at all. Established That reframing is now the standard justification for genetic rescue, and it is why the intervention is argued about on ecological grounds rather than genetic ones.

Speculative Reversibility is the property this family of interventions does not have, and the brief’s governance problem changes shape because of it. A protected area can be degazetted and a fence can be taken down; an introduced allele in a wild breeding population cannot be recalled, and a species established beyond its range is, in the statutory language most jurisdictions actually use, an introduction. Frontier The collision is direct: the statutes written to prevent deliberate release of non-native organisms are the statutes under which assisted colonisation would have to be authorised, and this brief could obtain no instrument distinguishing the two by anything other than intent. Established Consent compounds it, because the receiving landscape is usually somebody’s, and the recognition problem this brief already names for protected-area expansion applies with more force to a move that changes what lives there. Speculative Monitoring is the quiet failure mode: a genetic rescue evaluated at year three and never again is an unaudited intervention, and the year-ten survival convention this brief asks of restoration is precisely what would make assisted adaptation falsifiable. Adjacent material sits in de-extinction technologies and genetic engineering.

Established The workback runs from an endpoint nobody has reached: a protected-area network whose interior species trajectories are flat or positive rather than merely less negative. Today's measured value for comparison is Finland's 19, 13, 3 and 4 percent of birds, mammals, plants and phytoplankton responding positively, where responding positively mostly means declining more slowly.

Established L1 — the reported unit changes from area to effectiveness. The headline indicator would have to become effectiveness-weighted rather than square-kilometre-weighted. Established This binds institutionally, not scientifically, and it is first because everything else is downstream of what gets counted. Established L2 — management effectiveness is measured everywhere rather than sampled. Gill needed a bespoke database of 433 sites to say anything at all; Appleton needed a 176-country survey. Frontier A recurring, mandatory, comparable assessment is cheap relative to what it measures and binds on institutional will plus a small recurrent budget.

Established L3 — the workforce exists. This is the binding link and it is not ecological. From roughly 555,000 personnel and 286,000 rangers to roughly three million and more than 1.5 million: about 2.4 million additional trained people, employed on recurrent salaries, concentrated in tropical low-income states, within about five years. Frontier Everything upstream of L3 is cheap; nothing downstream of it works without it. Speculative The chain also contains an untested assumption: that ranger effectiveness does not saturate with density. Gill's 2.9 times is a comparison between adequate and inadequate capacity, not a dose curve, so the shape of the response above adequacy is unknown.

Established L4 — recurrent finance is separated from designation finance. Designation is capital expenditure incurred once; management is operating expenditure incurred forever. Established Watson and colleagues state that financial support is dwarfed by the benefits but that those returns depend on effective management, which is the qualitative form of this constraint. Established The quantitative form — the split of global conservation finance between designation and recurrent management — could not be obtained for this brief and is a declared gap. Frontier It is also the single most decision-relevant missing number in the subject.

Frontier L5 — placement decouples from the socioeconomic niche, which binds on tenure, consent and sovereignty rather than on mapping. Frontier L6 — connectivity becomes a design constraint rather than an outcome, which binds on a genuine scientific unknown: dispersal fields are unresolved for most taxa, and Endo's sub-one-percent self-recruitment is what happens when connectivity is left to chance. Frontier L7 — restoration supplies the net-positive term the network lacks, which binds first on a measurement convention and only then on money, because until hectares are counted at year ten rather than at planting the link is unauditable. Established Overall the chain binds at L3. Frontier The scientific unknowns determine how well the network performs; the workforce determines whether it performs at all.

Frontier L2 acquires a sensing corollary that does not relax it. Measuring management effectiveness everywhere rather than sampling it is now technically cheaper than when Gill needed a bespoke database of 433 sites, because an acoustic recorder and a water sample cost a fraction of a dive team. Established What has not become cheaper is curation: a classifier trained on one region’s soundscape loses precision on another’s and has to be re-estimated locally, and a confidence score from such a model is not a probability until somebody validates it against annotated audio from that site. Speculative A network that bought sensors and not curation would manufacture the appearance of continental monitoring from an instrument whose error rate nobody had measured, which is the paper-park failure repeated one layer up.

Frontier The workback acquires a genomic rung that sits below L2 rather than beside it. Genetic diversity within populations is now cheap to measure — reference genomes for non-model species are routine, and heterozygosity and runs of homozygosity are standard outputs — and it is still not routinely reported as a management indicator. Established This brief could not obtain the text of the 2030 framework and so quotes nothing it says about genetic diversity; what can be said is that the measurement is available and the reporting is not, which is the same shape as the effectiveness gap one level up. Speculative A network reporting heterozygosity trends for its flagship populations beside its hectares would be running the cheapest available test of whether protection preserves anything beyond presence.

Frontier This brief waits on results in three other fields and on one thing that is not a result at all. The first dependency is a single operational definition of biosphere integrity, which belongs to the planetary-boundaries programme rather than to conservation biology and is treated in Planetary Stewardship. Frontier Half-Earth arguments lean on that boundary, and it currently carries at least two control variables in papers with overlapping authorship. Speculative Until one definition wins, an area target justified by an Earth-system threshold is justified by a quantity that has not been fixed.

Frontier The second is dispersal and connectivity science. Designing a network as a metapopulation rather than an estate requires larval and propagule dispersal fields resolved operationally for the taxa that matter, at the spatial and seasonal resolution Endo's Brazilian result shows to be decisive. Speculative That is a decade of oceanography and movement ecology, and no current global indicator consumes its output.

Established The third is survival monitoring, which is a dependency on a convention rather than on a discovery. Neither restoration paper obtained for this brief reports vegetation survival rates. Frontier Year-five and year-ten survival by biome and by planting method is not scientifically hard; it is simply not required by any funder, which is why it is not collected.

Frontier The fourth is not a result but a disclosure: the composition of conservation finance. Without the split between one-off designation spending and recurrent management spending, the central policy claim in this brief — that the world is buying hectares instead of salaries — can be argued from structure but not audited from accounts. Established That disclosure exists inside institutions that were unreachable for this brief. Frontier It is worth naming that this is a dependency on transparency rather than on research, and that it is therefore the cheapest of the four to close and the least likely to be closed by anyone whose funding depends on the current framing.

Established Six measurements would settle most of what is contested here, and none of them requires an instrument that does not exist.

Frontier One: run the Santangeli design continentally. A counterfactual, multi-taxon, matched-site occupancy comparison over four decades, replicated across several biogeographic regions rather than one country, would establish whether the Finnish result — a small minority of taxa benefiting, mostly through slower decline — is a property of protected areas or a property of Finland. It is the highest-value single study in the subject and it runs mostly on archives.

Frontier Two: convert Gill's capacity comparison into a dose curve. Gill established that adequate staffing multiplies ecological effect by 2.9 relative to inadequate staffing. Speculative What nobody has published is the shape of the function above adequacy, which determines whether the three-million target is a floor or an overshoot, and therefore whether L3's cost is right.

Frontier Three: the dilution test. Regress per-unit-area management effectiveness against area added, by country, over the expansion period. Speculative Appleton's finding that staffing merely kept pace with expansion predicts a null or negative coefficient; if the coefficient is negative, designation without staffing actively harms the existing estate, which would be the most consequential result in the field and would use data that already exists.

Established Four: a year-ten survival cohort. Register restoration sites at planting, with biome, method, cost per hectare and a control, and report survival at five and ten years. This is the measurement that would make every national restoration pledge auditable, and its absence is why none of them currently is.

Frontier Five: a connectivity threshold trial. Endo measured self-recruitment at 0.54 to 0.69 percent in summer in a network that met its area target. Speculative Modelling minimum exchange thresholds between nodes for a set of indicator taxa, and then testing designation of a corridor against a matched no-corridor control, would convert connectivity from a diagnosis into a design rule.

Frontier Six: an ex-post standard for other conservation measures. Apply a candidate effectiveness standard retrospectively to areas already reported under that heading and report what fraction passes. Speculative If the fraction is high the mechanism is real; if it is low the 2030 target has been met on paper by reclassification.

Established The engineering requirement in this subject is unglamorous and it is mostly payroll infrastructure. Employing 2.4 million additional people on continuous salaries in mostly low-income tropical states requires recruitment pipelines, training institutions, identity and payroll systems that pay reliably in remote locations, equipment supply chains, medical and insurance provision, and a career structure that retains people past their third year. Frontier None of this is technically difficult and all of it is institutionally hard, which is the general shape of the whole subject.

Frontier Monitoring is where the actual instrumentation sits, and it is close to ready. Barasa and colleagues classified restoration outcomes from high-resolution commercial satellite imagery at 91 to 93 percent accuracy, which is adequate for land-cover change at the scale of a few hectares. Frontier Land cover is not biodiversity, so imagery has to be paired with in-situ methods — passive acoustic monitoring, environmental DNA, camera traps, standardised point counts — whose costs have fallen enough to make continuous multi-taxon monitoring plausible for the first time. Speculative The engineering gap is not the sensor. It is a common data schema and an obligation to report into it, without which every assessment stays bespoke, as Gill's and Appleton's both had to be.

Frontier Restoration engineering has a cost-accounting problem before it has a technical one. A 3.6-fold cost spread within European river restoration, and a 59 percent movement in the reported mean from an outlier decision, mean that engineering estimates for restoration at scale currently rest on a distribution nobody has characterised. Speculative The fix is a reporting standard: costs disaggregated by biome and intervention, with the outlier treatment stated and survival included in the denominator.

Frontier There is one genuinely hard piece of engineering in the subject and it is hydrological and ecological rather than digital. River restoration at the unit costs in the European dataset involves earthworks, structure removal, floodplain reconnection and long-lived maintenance obligations, and the ecosystem service value the same authors compute is amortised over a 40-year asset life. Speculative That framing is the useful one: a restored river is an asset with a maintenance schedule, not a completed project, and no standard exists for the maintenance liability of a restored ecosystem — which is how restoration comes to be financed as capital works and operated as though it were not works at all.

Frontier De-extinction sits adjacent and pulls in an awkward direction. If the binding constraint on holding biodiversity is staffed management of existing habitat, then restoring a lost species is downstream of a problem that has not been solved, and the funding it attracts competes with recurrent salaries for a workforce that is at roughly a fifth of its stated requirement. Speculative The steelman is that a charismatic restoration generates public willingness to pay that ranger salaries never will, making it a finance instrument rather than an ecological one. Handwave That transfer has not been demonstrated. See De-extinction Technologies.

Frontier Agriculture is the actual land-allocation counterparty. Strassburg's finding concerns converted land, and almost all converted land is farmed; the 15 percent that would deliver 60 percent of avoided expected extinctions has to come out of production or out of yield growth. Frontier The land-sparing versus land-sharing question is therefore not adjacent to this subject but continuous with it, and is treated in Future Agriculture.

Frontier Urbanisation is adjacent through both land demand and the placement niche. Protected areas over-aggregate where human development is high, which is where cities are; and the tropical regions where the ranger deficit is largest are the regions urbanising fastest. See Sustainable Megacities. Speculative Engineered and synthetic ecosystems are adjacent in a different way: they test whether ecological function can be assembled rather than protected, which is the same surrogate-function question rewilding asks with wild animals. See Synthetic Ecosystems.

Frontier Two less obvious adjacencies matter more than they look. The first is public employment policy. A workforce expansion of the size implied here is not a conservation programme in any conventional sense; it is a rural employment programme in mostly tropical states, and the delivery expertise for that sits in labour ministries and development banks rather than in environment ministries. Speculative The instrument that has actually placed millions into rural public work at speed is the public employment guarantee, and no retrieved source connects that literature to ranger recruitment at all. Speculative The second is sovereign credit, because the constraint is a permanent salary line rather than a project: whether a low-income state can commit to recurrent expenditure for decades is a question about fiscal space and debt profile, not about biodiversity, and it decides whether the target is deliverable in the countries holding most of the estate.

Established The first institutional requirement is an indicator change, and it is the cheapest thing in this brief. While the reported quantity is designated area, designated area is what ministries will produce, because it is what they are scored on. Frontier An effectiveness-weighted indicator — area multiplied by a measured management score — would change behaviour without changing a single line of treaty text. Established That the retention argument was published in 2018 and the indicator did not change is evidence about how hard this is rather than about whether it is right.

Established The second is a standard for other effective area-based conservation measures, applied after the fact. Without one, a target expressed in area can be met by reclassification, and the peer-reviewed restatements of the target are careful to say “protected areas and conservation measures” where the popular shorthand says national parks. Frontier A chapter by one of the Maxwell co-authors reads the framework's meaning for protected and conserved areas directly; its record was verified for this brief and its text was not read, so what the framework commits parties to is not quoted here. Established The verbatim target text was unreachable and is a declared gap, and no treaty wording appears in this brief that was not read.

Frontier The third is recognition of indigenous and traditional territories, and it is where the honest answer is that this brief's evidence base is thin. The structural argument is strong: protected areas over-aggregate in the socioeconomic niche of high development, the incremental land needed for any expanded target lies largely outside that niche, and tenure, consent and sovereignty are what bind at that link. Speculative A legal literature on the place-people-and-law connections of protected areas exists and its record was verified for this brief, but no quantitative source on indigenous and community-conserved territories — their extent, their measured biodiversity outcomes, or their share of any coverage figure — was obtainable. Handwave That is a gap in this brief, not a judgement about the evidence, and it should be closed before any claim about relative effectiveness is made in either direction.

Frontier The fourth is a finance instrument that funds salaries in perpetuity rather than gazettement once. Designation attracts capital because it produces an announceable event; management produces nothing announceable and needs money every year forever. Speculative Endowments, international transfer payments and biodiversity-linked sovereign instruments are all proposed. Frontier The measurement that would discipline all of them — the current split between designation and recurrent management spending — is the declared gap named in section 4.

Frontier The standards question for observatories is the indicator question arriving early. Common protocols — what volume of water, how many replicates, which marker, which classifier version, which minimum confidence — decide whether two parties’ results can be compared at all, and they are being set now, mostly by consortia rather than by regulators. Established The Essential Biodiversity Variables programme is the most developed attempt to fix a common variable layer above those protocols, precisely so that indicator definitions do not silently absorb whichever method happened to be used. Speculative If protocol standards are set by whoever deploys first, the comparability of the global record will have been decided by procurement rather than by policy, and this brief could obtain no source describing a binding cross-jurisdiction protocol standard for either eDNA or acoustic monitoring.

Established Protected areas produce human outcomes that are genuinely mixed, and the honest summary is a tie. Ban and colleagues found about half of documented well-being outcomes positive and about one-third negative across 118 studies, with social, health and cultural domains under-studied. Frontier A synthesis in which a third of documented outcomes are negative is not an endorsement and not an indictment; it is a finding that the intervention's distributional effects depend on design, and the design variables the same paper identifies — enforcement quality, no-take status, single-zone simplicity — are the variables that cost money.

Frontier The expansion arithmetic contains an unexamined ethical proposition. Any target that increases the protected estate materially must place the increment largely outside the socioeconomic niche protected areas currently occupy, which means on land where people live, hold rights or lack formal title. Speculative No source obtained for this brief addresses how that consent is to be secured, and the historical record of exclusionary conservation is precisely why the question cannot be treated as procedural. Handwave A half-of-the-planet target with no consent mechanism specified is not a conservation proposal; it is an unpriced claim on other people's land.

Frontier Rewilding raises an animal-welfare question that ecology cannot answer. Whether a population left to density-dependent regulation in a fenced reserve is being restored or being neglected is a question about obligation, not about equilibrium, and it is the question the Oostvaardersplassen controversy actually turned on. Established This brief could obtain no source on that case and therefore states no numbers about it. Speculative The generalisable point survives the gap: the moment a reference state is chosen by people, welfare consequences follow from a human decision and cannot be attributed to nature.

Frontier There is also a labour ethic in the thesis itself. Scaling to three million personnel means creating one of the largest new public workforces of the century, mostly in low-income states, in a job with documented physical risk. Speculative Whether that is an employment programme or an extraction of cheap enforcement labour depends on wage levels, tenure and who holds the payroll, none of which appear in the headcount.

Frontier Assisted adaptation asks for a consent conservation has no established form for. The subject of the intervention is a population that cannot be consulted and a landscape whose holders often can; the recognition gap this brief names for territorial expansion is the same gap, and moving genes or species into a territory is a heavier ask than drawing a line around it. Speculative A second asymmetry is worth naming: the interventions most likely to be funded are those with a commercial planting or aquaculture chain behind them, so the species that receive adaptive help will correlate with the species that have a market, and no source obtained for this brief tests that correlation.

Frontier A network that actually holds biodiversity would be a permanent civil institution, and civilisations have not built many of those on purpose. The thing being contemplated is a standing global corps of roughly three million people, paid every year indefinitely, doing work whose success is invisible because it consists of things not happening. The nearest analogues are public health surveillance and fire services, both of which took a century to become uncontroversial and both of which are still cut first in a fiscal crisis.

Frontier The alternative future is already the default and it is not collapse. It is a world with a large, well-mapped, formally designated estate in which measured interior trajectories continue to decline slowly, reported as compliance because the indicator is area. Established Santangeli's Finnish result is what that world looks like when someone finally measures it, and Finland is a wealthy, well-governed state with a long-established network. That is the important part: the null was measured where conditions were favourable.

Speculative The upside case is worth stating because it is quantified. Strassburg's optimisation implies that a well-allocated 15 percent of converted land could avoid 60 percent of expected extinctions and sequester 299 gigatonnes of CO₂, which the authors put at 14 percent of total emissions since the Industrial Revolution. Frontier A civilisation that did that would have demonstrated something it has never demonstrated: the ability to spend continuously on an outcome that produces no event, no ribbon and no measurable improvement inside an electoral cycle.

Speculative There is also a quieter implication about knowledge. A staffed network is, incidentally, the largest continuous biological observing system anyone could build — three million people in permanent contact with the places where change happens first. Handwave The scientific return on that, over a century, would plausibly exceed the conservation return, in the way that weather services turned out to matter for reasons unrelated to why they were funded. Speculative That argument has never been made in the funding case, and it may be the only version of the case that survives a fiscal crisis intact.

These horizons track institutional and workforce delivery, not ecological discovery. The science needed to act is largely in hand; what is not in hand is the payroll, the indicator and the consent.

• 10 yr: Established Expect the 2030 coverage target to be reported as approached or met in area terms, because area is what is counted and other conservation measures widen what qualifies. Frontier Expect the effectiveness gap to widen in the same period, because Appleton's mechanism — staffing keeping pace with expansion rather than exceeding it — is the default. Frontier Expect at least one large multi-region counterfactual evaluation on the Santangeli design, because the data are archival and the result is publishable either way. Speculative Expect a serious proposal for an effectiveness-weighted indicator and expect it not to be adopted on this horizon. Frontier Expect eDNA and acoustic monitoring to be written into at least one national protected-area reporting standard on this horizon, and expect the reference-library and classifier-validation costs to be unfunded in it.
• 25 yr: Speculative A recurrent-finance instrument that pays protected-area salaries from an endowment or a transfer mechanism rather than from annual appropriations is plausible on this horizon and is the single most consequential thing that could happen in the subject. Speculative Year-ten survival reporting becoming a condition of restoration finance is plausible and cheap, and would retire most current restoration claims. Frontier Expect the rewilding surrogate-function question to be settled empirically for at least one well-studied system, in one direction or the other. Speculative Expect at least one national assisted-migration authorisation for a non-commercial species to be attempted on this horizon and to be litigated under introduction law rather than under conservation law.
• 50 yr: Speculative A network designed and legally protected as a metapopulation — corridors of water and time rather than polygons of land — is the threshold institutional invention, and no retrieved source proposes the legal instrument that would do it. Speculative A workforce in the low millions is achievable on this horizon if it is treated as an employment programme rather than as a conservation budget line. Handwave Whether interior trajectories can be made non-negative at network scale remains, on the evidence available now, undemonstrated in any system.
• 100 / 250+ yr: Handwave A civilisation that has held its biosphere for a century will have done so with an institution that pays people to be present in places nobody visits, funded through wars and recessions, and it will look in retrospect as ordinary as a fire brigade. Handwave The alternative — a fully designated planet whose designated areas are empty of what they were designated for — requires no new decisions at all, which is precisely why it is the more likely one.

• Depends on This brief depends on results that conservation biology is not the field to produce. It waits on a single operational definition of biosphere integrity from the planetary-boundaries programme, because Half-Earth style area arguments rest on a boundary expressed as intact-ecosystem-area fraction in one paper and as Mean Species Abundance in another by an overlapping author team. It waits on dispersal science capable of specifying minimum exchange thresholds between nodes, since a network whose self-recruitment is under one percent has met an area requirement without meeting a demographic one. It waits on a survival-monitoring convention that reports restored hectares still standing at year ten. And it waits on a disclosure rather than a discovery: the split of conservation finance between one-off designation and recurrent management, unobtainable for this brief and the number that would confirm or refute its central claim.
• Requires (not on this map) Two constraints sit outside anything a research programme can deliver, and the first is the brief's whole thesis. Conservation capital is raised against designation, which produces an announceable event on a finite timetable; management is an operating cost that recurs forever and produces nothing announceable, and the gap between roughly 555,000 personnel today and the roughly three million the 2030 target implies is a recurrent salary problem, not a capital one. No instrument obtained for this brief pays protected-area wages in perpetuity, and the split of existing conservation finance between the two purposes could not be established here. The second constraint is recognition of indigenous and traditional territories. Any material expansion must place its increment outside the high-development socioeconomic niche protected areas currently occupy, which means on land held, used or claimed by people whose tenure is often informal; without recognition of those territories the expansion is either blocked or coercive, and this brief could obtain no quantitative source on the extent or measured outcomes of indigenous and community-conserved territories, which is a hole it declares rather than fills. Two further constraints arrive with the sensing and genomic material and neither is a research result. Curated reference libraries and locally validated classifiers are maintained assets rather than deliverables: a barcode database nobody curates and a soundscape model nobody re-estimates degrade into an instrument with an unmeasured error rate, and no funder obtained for this brief pays for curation as a standing line. And assisted adaptation needs statutory cover for deliberate translocation beyond a native range, because the law that would otherwise apply is invasive-species law, which distinguishes the conservation move from the offence by intent alone.
• Enables Closing the workforce link would enable almost everything downstream: the 2.9-fold capacity effect realised at network rather than site scale; management-effectiveness data that exists everywhere rather than in bespoke research databases; other-measure areas verified against a standard rather than accepted on classification; and restoration audited on survival rather than on planting. A network producing non-negative interior trajectories would also supply something the wider Earth-system programme lacks: an empirically anchored biosphere control variable measured in outcomes rather than in area.
• Adjacent Adjacent to de-extinction, which competes for the same conservation attention and funding while sitting downstream of an unsolved management problem; to agriculture, since restoration's target is converted land and almost all converted land is farmed; to urbanisation, because protected areas over-aggregate in the developed socioeconomic niche and the largest ranger deficits sit in the fastest-urbanising regions; and to synthetic and engineered ecosystems, which pose the same surrogate-function question as trophic rewilding with different components.

Established “Seventeen percent of land is protected.” The source sentence says formally designated, and it is a 2024 paper quoting a 2020 citation without a base year. Established Designation is a legal act and protection is a delivered outcome; Gill's capacity data and Appleton's staffing data show they are different quantities of different magnitude. This brief prints no current coverage percentage because the authoritative source was unreachable, and the enthusiast version of the error — that the world is more than halfway to its 2030 goal — is what treating an input as an output produces.

Established “The 2030 target means 30 percent in national parks.” The peer-reviewed restatements are careful: the planned coverage is of protected areas and conservation measures. Frontier The second half is not a rounding detail. It is the part with no agreed effectiveness standard, and whether it is an escape hatch or a dilution is the live institutional question. Established The verbatim target text was not readable for this brief and is therefore not quoted anywhere in it.

Established “River restoration costs 310,000 euros per hectare.” That is the outlier-inclusive mean of 119 European projects, and it is in the paper's title. Established Excluding four statistical outliers, the same paper reports 195,000 euros per hectare. Frontier The inflated figure is the one that travels, because indexing, abstracting and citation carry titles; if it is used at all it should be given as 195,000 outlier-excluded, or as both with the outlier treatment named, and never as 310,000 unqualified. Frontier This is a clean, fully verified example of a correction that exists inside the source and never reaches the citation.

Established “Restoring 15 percent of land avoids 60 percent of extinctions.” The source says 15 percent of converted lands, in priority areas, and the 60 percent is of expected extinctions under a model. Established Restated as 15 percent of land, the claim silently enlarges from a fraction of already-converted ground in optimally chosen places to a fraction of the entire terrestrial surface anywhere, and the paper's own 13-fold allocation finding is what makes the dropped qualifier load-bearing rather than pedantic.

Established “More rangers would be nice.” More rangers is the arithmetic centre of the target: roughly 5.4 times the current global workforce within about five years. Established Treating it as an implementation footnote rather than as the binding constraint is the most consequential framing error in this subject, and it is committed by almost every summary of 30x30 written for a general audience.

Frontier “Protected areas do not work.” This is the cynic's mirror image of the first error and it is also wrong. Established Santangeli found mixed effects with a real positive signal for birds and mammals and a clear dependence on area; Gill found a 2.9-fold dose-response to staff capacity; Di Marco reports wilderness communities carrying less than half the extinction risk. Frontier The evidence supports “under-resourced protected areas do little,” not “protected areas do nothing,” and dropping the capacity covariate produces the wrong policy conclusion in both directions at once: it tells optimists that designation suffices and pessimists that nothing does.

Established “Brazil's marine network shows that hitting targets works.” The same network that met the quantitative requirement of its area target has measured self-recruitment of 0.54 to 0.69 percent in summer and near-zero recruitment at most sites across 2008 to 2015. Frontier Target achievement and ecological function were measured in the same system, in the same years, and disagreed.

Established “We restored X hectares.” A hectares-restored figure with no year-ten survival term is a planting statistic. Established Neither restoration study obtained for this brief reports survival rates, and one explicitly reports no cost data either. Frontier The enthusiast form is the trillion-tree or national pledge quoted as delivered area, and the corrective is not scepticism but a reporting convention nobody currently requires.

Frontier “Half-Earth is fringe” and “Half-Earth is settled” are both wrong. Frontier An Earth-system boundary analysis independently places safe intact natural ecosystem area at 50 to 60 percent of global land and records it as already exceeded — a serious quantitative route to roughly the same number without invoking its popular advocate. Against it: the observed placement niche says the residual land available for protection is systematically not the land biodiversity needs, the measured conversion efficiency of protection is low, and effectiveness has not risen with area. Speculative For it to work, the 50 percent would have to be ecologically selected rather than residual, effectiveness would have to scale rather than dilute, and tenure over the incremental land would have to be settled without coercion. Handwave That last condition binds and no retrieved source addresses it. Established The Half-Earth primary literature and its principal critiques were not obtainable here, so this passage is built from the Earth-system boundary and the placement literature and labelled accordingly.

Frontier “Rewilding is restoration” is a claim, not a definition. Speculative For it to be restoration there must be a stated reference state that is not merely a chosen historical baseline, pre-intervention baselines actually measured, and surrogate taxa that deliver the function of the taxa they replace rather than the silhouette. Frontier The third is the binding scientific unknown, and a 2026 paper whose contribution is establishing baselines before intervention suggests the second is commonly unmet. Handwave Where those conditions fail, rewilding is a legitimate aesthetic and cultural project that should be argued on those terms rather than smuggled in as ecology.

Established A methodological correction that belongs in this brief because it is a fact about how citations get corrupted. The research pass behind this brief recorded a bibliographic query about protected-area targets returning a result set entirely about organ cryopreservation, in a machine-readable pipeline with no error and no low-confidence signal. Frontier The result was discarded, and the incident is reported here because this subject is governed by exactly the kind of large assessment report that is assembled by machine from bibliographic queries, and silent substitution of one document for another produces confident, well-formatted, wrong citations. Frontier It is also a reason to treat the citation-lag finding at the top of this section as a symptom rather than an anomaly: a 2024 paper carrying a 2020 coverage figure by citation is the same phenomenon at a slower speed.