Skip to content
Ecological Intelligence

Ecological evidence.
Independent perspective.
Explore Wild Lands ↗

Ecological Intelligence / Subject

remote sensing

Peer-reviewed

Water-use efficiency trends differ within Iran's drylands

Evidence
Peer-reviewed
Geography
Iran: Zagros, Alborz and other modeled zones
System
Dryland ecosystems / evapotranspiration

WHAT THE EVIDENCE SAYS

A 2001–2025 remote-sensing study mapped valid ecosystem water-use efficiency estimates over about 44% of Iran. Around 23.75% of that valid area showed significant trends, mostly increases; NDVI was an influential predictor in the machine-learning model.

WHY IT MATTERS / OUR READ

A higher productivity-to-water-loss ratio does not by itself mean more ecological water is available or that drought risk has declined.

CAVEATS / UNCERTAINTY

SHAP contributions are model attributions, not experimentally isolated causes; large portions of the country were not covered by valid estimates.

SOURCE

Scientific Reports — Kazemzadeh (2026)
View original evidence ↗

Remote sensing suggests increasing carbon assimilation relative to water loss in some mapped ecosystems.

Read the Wild Lands assessment →
Peer-reviewed Global

Tropical forests' canopy-water buffer could weaken as atmospheric drying intensifies

Evidence
Peer-reviewed
Geography
Pantropical forest regions
System
Tropical forest carbon–water interactions

WHAT THE EVIDENCE SAYS

Researchers integrated satellite, flux-site and climate-model products. Historical analysis for 2001–2015 suggested canopy water moderated ecosystem water-use-efficiency declines by about 18% under atmospheric aridity and 9% under soil drought; late-century simulations projected much weaker atmospheric buffering.

WHY IT MATTERS / OUR READ

Tree-cover persistence may mask diminishing physiological capacity, so ecological resilience assessments should connect canopy function and climate exposure.

CAVEATS / UNCERTAINTY

The late-century decline—including a projected 88% reduction in the modeled buffering function—is scenario- and model-dependent, not measured future dieback or a forest-by-forest forecast.

SOURCE

Water Resources Research — primary remote-sensing and scenario study
View original evidence ↗

Observation-based modeling found water held in canopies moderated past declines in photosynthetic water-use efficiency, but that benefit declined in projections.

Read the Wild Lands assessment →
Europe

Central Europe’s growing season is lengthening fastest in some managed land covers

Evidence
Peer-reviewed remote-sensing time-series study
Geography
Central Europe
System
Agricultural land / forests / grasslands

WHAT THE EVIDENCE SAYS

Two-band MODIS vegetation-index records for 2000–2022 showed earlier growing-season starts, later ends and longer seasons across arable land, broad-leaved forest, coniferous forest and grassland. Median start-of-season trends ranged roughly from 9.3 to 17.5 days earlier per decade and season length increased by about 13.3 to 25 days per decade, with arable land showing especially strong changes and elevation modifying the pattern.

WHY IT MATTERS / OUR READ

Phenology is partly climate signal and partly land-management signal. Comparing land-cover classes and elevation bands helps separate where a longer remote-sensing growing season may reflect altered cropping or vegetation behavior rather than a uniform biological response.

CAVEATS / UNCERTAINTY

The metrics are satellite land-surface phenology, not direct species-level phenology. The 2000–2022 trends are observational, and climate and agricultural management can contribute simultaneously.

SOURCE

Regional Environmental Change — primary study
View original evidence ↗

MODIS land-surface phenology from 2000–2022 found earlier starts and later ends across arable land, forests and grassland, with the strongest shifts often in arable areas and lower elevations.

Read the Wild Lands assessment →
Peer-reviewed North America

For Southern California spotted owls, wildfire—not fuels treatment—drove most mapped old-forest habitat loss

Evidence
Peer-reviewed
Geography
Southern California, USA
System
Old forest / California spotted owl habitat

WHAT THE EVIDENCE SAYS

Researchers combined 1986–2022 California spotted-owl detections, a statewide fuels-management database and remotely sensed predictors in random-forest habitat models. Nesting/roosting habitat declined 54% since 1986, with 92% of remaining habitat concentrated in San Bernardino National Forest. Of attributable post-2006 nesting/roosting habitat losses, wildfire accounted for about 80%, drought/other disturbance 14% and fuels management 6%. Potential territories meeting both nesting/roosting and prey-acquisition habitat thresholds declined 73%.

WHY IT MATTERS / OUR READ

The result complicates the idea that fuels treatment is the dominant driver of old-forest habitat loss in this system. It supports evaluating treatment effects against the larger counterfactual risk from severe wildfire while still protecting the habitat features owls require.

CAVEATS / UNCERTAINTY

Attribution comes from remote-sensing and statistical habitat models rather than a randomized treatment experiment. Habitat suitability is not the same as owl demography, treatment intensity and placement matter, and the percentages are specific to this Southern California landscape and study period.

SOURCE

Forest Ecology and Management / USDA Forest Service — primary study
View original evidence ↗

USDA Forest Service Treesearch
View supporting source ↗

Thirty-seven years of habitat mapping attributed 80% of post-2006 nesting/roosting habitat decline to wildfire, compared with 6% to fuels management, in the southern range of the California spotted owl.

Read the Wild Lands assessment →
Peer-reviewed North America

A LiDAR ‘canopy variety’ map could help target riparian forest restoration—but it is still a screening tool

Evidence
Peer-reviewed
Geography
Four riparian watersheds, Nova Scotia, Canada
System
Riparian forest

WHAT THE EVIDENCE SAYS

Researchers developed a Canopy Variety Hotspot metric from LiDAR canopy-height models collected in 2013 and 2024 and used Hurricane Fiona as a disturbance reference. Smaller neighborhood windows were most sensitive to proportional canopy change, and multi-temporal comparisons mapped clusters that persisted, appeared or disappeared over the 11-year interval. A proof-of-concept workflow then paired the map with field inventories and other metrics to illustrate how structurally persistent sites might inform climate-adaptive riparian planting design.

WHY IT MATTERS / OUR READ

The useful idea is not that one raster proves storm resistance. It is that managers can convert canopy heterogeneity into a repeatable spatial screening layer, compare it through time and then decide where field verification or restoration design deserves attention. That is a practical bridge between remote sensing and site-level planning.

CAVEATS / UNCERTAINTY

The authors present CVH as a proof-of-concept screening tool. Its output is sensitive to neighborhood and distance parameters, and structural persistence after one major storm does not prove causal windthrow resistance or future restoration performance. Field validation remains necessary before using hotspot classes as prescriptions.

SOURCE

Forest Ecology and Management — primary study
View original evidence ↗

Across four Nova Scotia watersheds, a new hotspot metric tracked canopy-height heterogeneity before and after Hurricane Fiona and highlighted structurally persistent clusters for follow-up.

Read the Wild Lands assessment →
Peer-reviewed North America

Tallgrass-prairie productivity models worked well—until management got intensive

Evidence
Peer-reviewed
Geography
Central Oklahoma, USA
System
Native tallgrass prairie working lands

WHAT THE EVIDENCE SAYS

Researchers compared eddy-covariance-derived gross primary production with two remote-sensing/light-use-efficiency models across six native tallgrass prairie pastures of roughly 32–64 ha. The sites differed in burn interval, grazing intensity and haying. Enhanced Vegetation Index was the strongest predictor of measured GPP, and modeled GPP tracked measured dynamics well overall (R² about 0.72–0.85). But annual cumulative GPP was overestimated by roughly 15–30% in intensively grazed and hayed pastures, while a lightly grazed pasture showed only 0–2% overestimation.

WHY IT MATTERS / OUR READ

For managers using satellite productivity or carbon products, management intensity is part of the calibration problem. A model can perform well across a region yet still bias the sites where grazing, haying or species composition most strongly alter green biomass. Ground-truthing the management extremes may be more useful than assuming a single regional correction.

CAVEATS / UNCERTAINTY

The six pastures are co-located in one central Oklahoma research landscape, so management, soils and weather do not span the full tallgrass-prairie domain. GPP is carbon uptake, not net ecosystem carbon storage, forage quality or biodiversity. Management treatments and C3:C4 composition are not fully randomized experimental factors.

SOURCE

Science of the Total Environment / USDA-ARS — primary study
View original evidence ↗

Across six Oklahoma prairie pastures, remote-sensing models tracked eddy-covariance GPP well overall but overpredicted annual production by 15–30% under intensive grazing and haying.

Read the Wild Lands assessment →
Peer-reviewed Global

Negative vegetation extremes are covering more land—even though event frequency is not rising as clearly

Evidence
Peer-reviewed
Geography
Global terrestrial ecosystems
System
Terrestrial vegetation

WHAT THE EVIDENCE SAYS

An event-based analysis of global leaf-area-index data from 2001–2024 found extreme negative vegetation-growth events showed a significant increase in affected area of about 394,471 km² per year (p=0.04), reaching 39% of global land area in 2024. Event frequency also trended upward by 0.07% per year, but that trend was not statistically significant (p=0.13). Drought alone and drought compounded with heat or wildfire coincided with roughly 58.5–59.8% of events. High-emissions projections suggest substantially wider future exposure.

WHY IT MATTERS / OUR READ

The split between area and frequency matters: ecological stress can become more spatially extensive without a clean increase in how often events are counted. Tracking only event counts would miss that expansion.

CAVEATS / UNCERTAINTY

Leaf-area-index anomalies are remotely sensed vegetation responses, not direct measures of mortality, biodiversity loss or carbon flux. Coincidence with drought-related compound events does not by itself establish event-level causation, and future projections depend on scenario and model assumptions.

SOURCE

Global Change Biology — primary study
View original evidence ↗

Global LAI data show the affected area of extreme negative vegetation events increased significantly from 2001–2024, while the frequency trend was weaker and not statistically significant.

Read the Wild Lands assessment →
Peer-reviewed

Antarctic ocean colour is partly a community-size problem masquerading as a sensor problem

Evidence
Peer-reviewed
Geography
Antarctic waters south of 60°S, with global comparison data
System
Southern Ocean phytoplankton and ocean colour

WHAT THE EVIDENCE SAYS

A Geophysical Research Letters study combined extensive in situ observations with an optical modelling framework and found that Antarctic phytoplankton size structure differs fundamentally from tropical–subtropical oceans. Larger-cell dominance at higher chlorophyll concentrations shifts both algal and co-varying non-algal optical properties, allowing the model to reproduce much of the Southern Ocean’s anomalous colour–chlorophyll relationship without invoking unrelated regional corrections.

WHY IT MATTERS / OUR READ

Satellite records are only as ecological as the biology embedded in the retrieval. If community size structure changes the optical signal, long-term ocean-colour trends can be misread unless algorithms represent that regional biological structure. The result links monitoring error directly to community ecology.

CAVEATS / UNCERTAINTY

The framework is strongest for open-ocean Antarctic waters under Case-1 optical assumptions. Coastal, ice-affected or CDOM-rich waters may behave differently, and the modelling is a mechanistic interpretation rather than a new satellite inversion product already validated across the entire Southern Ocean.

SOURCE

Geophysical Research Letters — primary study
View original evidence ↗

In situ observations and optical modelling indicate that Antarctica’s unusual phytoplankton size structure can reproduce persistent regional biases in satellite chlorophyll relationships.

Read the Wild Lands assessment →
Commentary Global

Satellite soil moisture is becoming a practical regional-to-global monitoring layer

Evidence
Commentary
Geography
Global
System
Soil moisture / terrestrial hydrology

WHAT THE EVIDENCE SAYS

A Tools of the Trade article outlines how satellite-derived products can be used to estimate and compare surface soil-moisture dynamics over regional to global scales, extending observation beyond sparse in-situ networks.

WHY IT MATTERS / OUR READ

The management value is scale: soil moisture links rainfall to plant stress, runoff and fire or drought conditions, so spatially consistent observations can help place local measurements in a broader landscape context.

CAVEATS / UNCERTAINTY

This is a technical overview rather than a new validation experiment. Satellite products differ in sensing depth, spatial and temporal resolution, vegetation sensitivity and retrieval error, so they should not be treated as interchangeable with field measurements.

SOURCE

Nature Reviews Earth & Environment — Tools of the Trade
View original evidence ↗

A Tools of the Trade note surveys how remote-sensing products can track surface soil-moisture dynamics beyond the footprint of field sensors.

Read the Wild Lands assessment →
Peer-reviewed

Forest-volume maps can carry their uncertainty instead of hiding it

Evidence
Peer-reviewed
Geography
Nanning, Guangxi, China
System
Managed forest landscape

WHAT THE EVIDENCE SAYS

Using field plots and UAV-LiDAR metrics, researchers compared forest stock-volume models and found a probabilistic CB-NGBoost approach performed best among the tested models on an independent holdout, with R² of 0.74 and RMSE of 71.24 m³/ha. SHAP analysis indicated that variables driving the central estimate were not identical to those driving predictive uncertainty.

WHY IT MATTERS / OUR READ

For inventories and carbon or harvest decisions, an uncertainty surface can be as useful as the map itself. It helps separate places where the model predicts a value from places where the model has enough support to justify confidence in that value.

CAVEATS / UNCERTAINTY

The field dataset contained 30 original plots and only six original plots in the independent test split. The uncertainty treatment emphasized aleatoric uncertainty, prediction intervals were not perfectly calibrated, and transfer to other forest types or sensors was not tested.

SOURCE

Forest Ecosystems — primary study
View original evidence ↗

A UAV-LiDAR model in southern China paired probabilistic stock-volume estimates with interpretability, showing that predictors of the mean estimate were not the same as predictors of uncertainty.

Read the Wild Lands assessment →
Peer-reviewed

In eastern China’s estuaries, resilience-loss hotspots and conventional risk hotspots did not line up

Evidence
Peer-reviewed
Geography
Liaohe, Yangtze, Minjiang and Pearl River estuaries, China
System
Estuarine wetlands

WHAT THE EVIDENCE SAYS

Using 2000–2024 NDVI time series, lag-1 autocorrelation trends and hotspot analysis, researchers estimated that about 42.26% of the four estuarine study areas showed declining resilience signals. The spatial pattern did not align cleanly with pressure–state–response ecological-risk hotspots; the authors identified areas where resilience-loss signals coincided with relatively low conventional risk scores.

WHY IT MATTERS / OUR READ

A place can still look acceptable on a state-based score while its temporal behavior is becoming less stable. Pairing condition metrics with dynamic indicators may expose places worth closer monitoring before visible degradation becomes the only signal available.

CAVEATS / UNCERTAINTY

Critical-slowing-down indicators are indirect resilience proxies, not observations of an impending regime shift. The analysis assumes alternative-state dynamics and treats tidal flats, salt marshes and mangroves as components of one wetland system; model performance was moderate and causal attribution is limited.

SOURCE

Frontiers in Environmental Science
View original evidence ↗

Remote-sensing indicators flagged declining vegetation resilience in about 42% of the study area, including places that still looked relatively low-risk under a pressure–state–response framework.

Read the Wild Lands assessment →

Ara’s wetland signal has weakened as urban land expanded—but satellite indices are not biodiversity surveys

Evidence
Peer-reviewed remote-sensing time-series study
Geography
Ara, Bihar, India
System
Urban wetland / peri-urban landscape

WHAT THE EVIDENCE SAYS

A Landsat analysis spanning 1995–2025 used multiple surface indices to track water, vegetation and built-up change around Ara. The authors report progressive declines in open-water and vegetation signals as urban land expanded, with strong seasonal differences in some vegetation metrics.

WHY IT MATTERS / OUR READ

Remote sensing can provide a durable early-warning layer where long-term field monitoring is missing. It is particularly useful for showing whether hydrologic and vegetative space is being compressed by urbanization before individual projects are evaluated in isolation.

CAVEATS / UNCERTAINTY

Surface indices are proxies for land and water condition, not field measurements of biodiversity, water chemistry or wetland function. Classification and seasonal effects matter, and the study should not be read as direct evidence of bird or aquatic-community decline.

SOURCE

Discover Geoscience
View original evidence ↗

Three decades of Landsat indices around Ara, India, track declining open-water and vegetation signals alongside urban growth, offering a landscape-change warning without directly measuring species response.

Read the Wild Lands assessment →
Government data Arctic

The 2026 Arctic sea-ice minimum tied for the tenth-lowest satellite-era extent

Evidence
Government data
Geography
Arctic Ocean
System
Arctic sea ice

WHAT THE EVIDENCE SAYS

NASA and NSIDC report that Arctic sea ice reached its 2026 annual minimum on September 12 at about 4.6 million km², tied with 2008, 2010 and 2025 for the tenth-lowest minimum in the continuous satellite record. The years 2007–2026 contain the 20 lowest annual minima since observations began in 1978.

WHY IT MATTERS / OUR READ

One relatively less-extreme annual minimum does not reset the baseline. The current record is better read as year-to-year weather variability occurring within a persistently low-ice era.

CAVEATS / UNCERTAINTY

This is a current observational update, not a mechanistic attribution study. Annual minimum extent is one sea-ice metric and does not describe thickness, age, regional distribution or ecological effects on its own.

SOURCE

NASA / NSIDC
View original evidence ↗

EvidenceGovernment dataGeographyArctic OceanSystemArctic sea ice WHAT THE EVIDENCE SAYS NASA and NSIDC report that Arctic sea ice reached its 2026 annual minimum on September 12 at about 4.6 million

Read the Wild Lands assessment →
Peer-reviewed Global

Drylands can green and become less stable at the same time

Evidence
Peer-reviewed
Geography
Global drylands
System
Drylands / rangelands

WHAT THE EVIDENCE SAYS

A 1982–2020 analysis found persistent vegetation greening across global drylands alongside increasing interannual variability over much of the domain. Increasing seasonal maxima and declining minima can coexist in the same long-term record, while dynamic global vegetation models did not reproduce all of the observed variability and divergent trajectories.

WHY IT MATTERS / OUR READ

A positive long-term vegetation trend is not automatically a resilience signal. Managers and analysts should pair average greenness with variability, minima and climate sensitivity when judging whether drylands are becoming more secure or more volatile.

CAVEATS / UNCERTAINTY

Leaf-area index is a remotely sensed vegetation indicator rather than a direct measure of ecosystem function or state transition. The analysis supports increasing variability but does not show that all greening drylands are crossing degradation thresholds.

SOURCE

Nature Climate Change
View original evidence ↗

EvidencePeer-reviewedGeographyGlobal drylandsSystemDrylands / rangelands WHAT THE EVIDENCE SAYS A 1982–2020 analysis found persistent vegetation greening across global drylands alongside increasing interannual variability over much of the domain. Increasing seasonal

Read the Wild Lands assessment →
Peer-reviewed Latin America & Caribbean

In Peru, forest canopy function predicts bird traits that ordinary habitat maps miss

Evidence
Peer-reviewed
Geography
Andean–Amazonian forests, Peru
System
Tropical forest

WHAT THE EVIDENCE SAYS

Researchers linked remotely sensed forest functional composition with trait, taxonomic and vulnerability gradients across more than 1,300 forest-dependent bird species in Peru. Forest functional groups added explanatory power beyond a seven-variable environmental baseline, increasing explained variance by roughly 16–102% on axes where canopy composition was informative.

WHY IT MATTERS / OUR READ

Canopy function can add biologically meaningful information that broad habitat classes and climate/topography alone miss. For landscape assessment, this suggests that remotely sensed vegetation function may help distinguish habitat that looks structurally similar but supports different ecological communities.

CAVEATS / UNCERTAINTY

The analysis is spatial and associational rather than a direct local census or experiment. The strength of relationships varies among response axes, causality is not established, and results from Peruvian forests should not be generalized to all tropical landscapes without validation.

SOURCE

Nature Communications
View original evidence ↗

Remote-sensing-derived forest functional groups added explanatory power for traits and vulnerability across more than 1,300 forest-dependent bird species.

Read the Wild Lands assessment →
Peer-reviewed

Change the map scale and Hangzhou Bay’s ecological trend changes with it

Evidence
Peer-reviewed
Geography
Hangzhou Bay Urban Agglomeration, China
System
Urban / coastal metropolitan landscape

WHAT THE EVIDENCE SAYS

A Google Earth Engine analysis tracked a remote-sensing ecological index from 2000–2024 at 1-, 3-, 5- and 10-km grids and county scale. The regional RSEI increased by about 0.0423 overall, but the share classified as improving rose from 75.34% at 1 km to 84.25% at county scale, demonstrating how aggregation can hide opposing local trajectories. Land-use intensity was the strongest SHAP contributor in the explanatory model.

WHY IT MATTERS / OUR READ

For metropolitan ecological planning, scale is part of the result. County or city averages can be useful for screening, but site and corridor decisions need nested spatial analysis so local degradation is not erased by improvement elsewhere.

CAVEATS / UNCERTAINTY

RSEI combines greenness, wetness, dryness and surface temperature; it is not a direct biodiversity or ecosystem-integrity measure. The study is observational and cannot assign causal effects to land use or policy, and its machine-learning validation was not designed as a strict spatial or temporal holdout test.

SOURCE

Frontiers in Ecology and Evolution
View original evidence ↗

A 25-year remote-sensing analysis found coarse administrative units made improvement look more widespread while hiding local opposing trends along urban corridors and the bay.

Read the Wild Lands assessment →
Peer-reviewed Latin America & Caribbean

The Caatinga can look stable in the long-term average while local resilience still diverges

Evidence
Peer-reviewed
Geography
Caatinga biome, northeastern Brazil
System
Seasonally dry tropical forest / dryland landscape

WHAT THE EVIDENCE SAYS

A 2004–2024 remote-sensing analysis of Caatinga vegetation found broadly stable long-term net primary productivity and water-use-efficiency trends at biome scale while year-to-year behavior tracked water availability strongly. The study also detected spatially different trajectories among land-governance categories and near boundaries, showing that broad stability can coexist with localized vulnerability or resilience.

WHY IT MATTERS / OUR READ

Biome-wide averages can be reassuring and still be poor management maps. In drylands, surface-water availability, edge context and governance history can separate local trajectories that disappear when the whole region is averaged together.

CAVEATS / UNCERTAINTY

The analysis relies on remotely sensed productivity and water-use-efficiency proxies rather than direct biodiversity or demographic measurements. Governance categories are correlated with landscape history and site conditions, so differences among them should not be read as simple causal effects of tenure or protection status.

SOURCE

Environmental Management
View original evidence ↗

Two decades of remote sensing found little net trend in productivity or water-use efficiency, but strong year-to-year sensitivity to water availability and spatial differences across governance settings.

Read the Wild Lands assessment →
Peer-reviewed North America

In the California Delta, aquatic-weed treatment efficacy depended on habitat, dose and treatment history

Evidence
Peer-reviewed
Geography
Sacramento–San Joaquin Delta, California, USA
System
Tidal freshwater estuary / invasive floating aquatic vegetation

WHAT THE EVIDENCE SAYS

Across 349 sites in 2004–2008 and 333 sites in 2014–2018, higher per-visit herbicide doses generally reduced floating aquatic vegetation more than the same seasonal amount spread across more low-dose visits. Effects varied sharply by habitat; channel treatments were weak or short-lived, while glyphosate effects persisted up to three years in some flooded-island sites.

WHY IT MATTERS / OUR READ

The management signal is spatial and operational: treatment method cannot be separated from flow environment, dose and repeated-use history.

CAVEATS / UNCERTAINTY

This is a retrospective program evaluation, not a randomized experiment, and it evaluates target-plant efficacy rather than full ecological safety, non-target effects or long-term resistance. The historical treatment periods also limit direct transfer to present programs.

SOURCE

Environmental Management
View original evidence ↗

A landscape-scale retrospective found low-dose repeated applications often ineffective, especially in channels, while some slower-water habitats retained stronger control.

Read the Wild Lands assessment →
Peer-reviewed

The Tarim Basin kept greening while groundwater storage fell

Evidence
Peer-reviewed
Geography
Tarim River Basin, Xinjiang, China
System
Arid-basin groundwater, irrigated oases and desert vegetation

WHAT THE EVIDENCE SAYS

A 2003–2024 analysis combined GRACE, GLDAS and MODIS data to compare groundwater-storage anomalies with vegetation condition across the Tarim Basin. Basin-scale groundwater storage declined while vegetation condition generally improved, and the relationship between greening and groundwater varied strongly by location rather than following one basin-wide rule.

WHY IT MATTERS / OUR READ

Greenness is not groundwater security. In heavily managed arid basins, vegetation improvement can coexist with declining subsurface storage, so ecological and agricultural condition metrics need to be read alongside the water account that sustains them.

CAVEATS / UNCERTAINTY

GRACE has coarse spatial resolution and cannot cleanly separate shallow and deep groundwater or every storage component. The analysis is observational, and irrigation, glacier/snow inputs, soil moisture and land-use change can confound spatial relationships.

SOURCE

Water
View original evidence ↗

Satellite and gravity observations from 2003–2024 show improving vegetation condition alongside declining groundwater storage, with the relationship varying sharply across the basin.

Read the Wild Lands assessment →
Peer-reviewed

Radar satellites are finally following emperor-penguin colonies through the polar night

Evidence
Peer-reviewed
Geography
Atka Bay, Coulman Island and Cape Washington, Antarctica
System
Emperor penguin colonies and sea-ice habitat

WHAT THE EVIDENCE SAYS

A Communications Earth & Environment study combined Sentinel-1 synthetic-aperture radar with optical imagery to reconstruct year-round colony positions at three emperor-penguin colonies from 2017–2024. Radar allowed observation through winter darkness and cloud. Colonies shifted position during winter and sometimes moved synchronously away from the fast-ice edge even when the ice appeared broadly stable.

WHY IT MATTERS / OUR READ

Year-round monitoring changes what counts as habitat use. If colonies move substantially during the winter period that optical surveys largely miss, climate-risk assessment and survey design need to follow the whole seasonal cycle rather than extrapolating from the visible summer window.

CAVEATS / UNCERTAINTY

The study covers three colonies and tracks colony-scale position rather than abundance, breeding success or individual movement. The observed shifts do not by themselves identify the environmental mechanism driving relocation.

SOURCE

Communications Earth & Environment
View original evidence ↗

Eight years of Sentinel-1 tracking across three colonies show that winter colony position can shift even when fast ice looks stable, filling a seasonal blind spot left by optical imagery.

Read the Wild Lands assessment →
Peer-reviewed Africa

Small forest disturbances add up to a Central African carbon loss that coarser maps can miss

Evidence
Peer-reviewed
Geography
Central African dense humid forests, including the Congo Basin
System
Tropical humid forest, secondary forest and disturbed forest mosaics

WHAT THE EVIDENCE SAYS

A Nature Communications study used annual 10-m canopy-height maps built from spaceborne lidar, Sentinel radar and optical imagery to resolve biomass changes from 2019 to 2022. The analysis found that 87% of detected disturbances were smaller than one hectare. These small patches accounted for a substantial share of both carbon loss and regrowth, and the regional budget showed a net aboveground biomass-carbon loss of about 58 ± 9 Tg C per year, with gross losses only partly offset by gains.

WHY IT MATTERS / OUR READ

Deforestation maps can miss the ecological importance of degradation that arrives as many small openings. For carbon accounting and forest management, monitoring needs to resolve the spatial grain at which selective logging, small clearings, fire and regrowth actually change biomass rather than treating only large clearings as meaningful change.

CAVEATS / UNCERTAINTY

The analysis covers aboveground biomass over a short 2019–2022 window and depends on remote-sensing and model calibration; it is not a complete ecosystem carbon budget and does not include all belowground or soil-carbon processes. Attribution of individual disturbance causes is also more uncertain than detection of canopy change itself.

SOURCE

Nature Communications
View original evidence ↗

EGU 2026 · Study abstract
View supporting source ↗

Annual 10-m canopy-height maps for 2019–2022 found that 87% of detected disturbance patches were under one hectare, while dense Central African forests showed a net aboveground biomass-carbon loss.

Read the Wild Lands assessment →
Peer-reviewed Europe

The Alpine peatlands a mapping model misses may be the ones most degraded

Evidence
Peer-reviewed
Geography
European Alps
System
Alpine peatlands and cross-border remote sensing

WHAT THE EVIDENCE SAYS

A Communications Earth & Environment study developed a harmonized multi-sensor workflow for mapping peatlands across the European Alps using 590 ground-reference peatlands, Sentinel-2 optical data, Sentinel-1 radar, Landsat surface temperature and terrain information. The best-performing model was dominated by Sentinel-2 features. Importantly, official mapped peatlands were more likely to be missed where orthophotos showed forest encroachment or agricultural disturbance, while some apparent model false positives may represent unmapped peatlands.

WHY IT MATTERS / OUR READ

Monitoring error is not always random. If degraded peatlands become less visually or spectrally peatland-like, the systems most in need of restoration can disappear from inventories built around intact reference conditions. That makes disagreement between maps and models a useful trigger for targeted field verification rather than just a classification nuisance.

CAVEATS / UNCERTAINTY

Remote sensing cannot by itself diagnose peat depth, hydrologic condition or restoration priority. Official inventories also vary among countries, so some disagreement reflects reference-data limitations as well as ecological degradation. Candidate unmapped peatlands still require field confirmation.

SOURCE

Communications Earth & Environment
View original evidence ↗

A cross-border remote-sensing workflow validated against 590 peatlands found that disturbed sites were disproportionately hard to detect, turning classification error itself into a restoration signal.

Read the Wild Lands assessment →
Peer-reviewed Global

Tropical rainforests lost 18% of undisturbed cover since 1990 while edge exposure increased

Evidence
Peer-reviewed
Geography
Pantropical rainforest regions, 1990–2023
System
Tropical rainforests

WHAT THE EVIDENCE SAYS

Annual tropical land-use maps from 1990–2023 indicate an 18.1% loss of undisturbed rainforest cover. Losses were uneven: about 25% in Asia, 14.8% in Africa and 14.5% in the Americas, with especially large regional declines in Southeast Asia, West Africa, the Amazon deforestation arc and Central America. The area of remaining forest within 100 metres of an edge increased by about 6.6%, showing that fragmentation rose alongside outright loss.

WHY IT MATTERS / OUR READ

Forest area and forest configuration are different conservation variables. A landscape can retain mapped rainforest while exposing more of it to edge-related changes in microclimate, disturbance and access. Monitoring only net area can therefore understate how much interior habitat and ecological continuity have changed.

CAVEATS / UNCERTAINTY

The analysis depends on remote-sensing classifications and a fixed 100-m edge-distance metric. Edge exposure is a structural proxy, not a direct measurement of biodiversity loss or microclimatic change at every site.

SOURCE

Scientific Reports · Original evidence
View original evidence ↗

Annual land-use maps show forest loss and fragmentation moving together across the tropics, with especially large declines in Southeast Asia and West Africa.

Read the Wild Lands assessment →
Peer-reviewed Africa

One national park can contain very different fire regimes

Evidence
Peer-reviewed
Geography
Hwange National Park, Zimbabwe
System
Fire-managed savanna protected area

WHAT THE EVIDENCE SAYS

MODIS records from 2002–2022 showed strong spatial differences in fire across Hwange National Park. Roughly 37% of the park burned at least once, but the Robins section had far more recurrent burning than Makona, and total annual burned area did not show a significant park-wide temporal trend. Fire patterns also varied with vegetation, temperature and precipitation.

WHY IT MATTERS / OUR READ

A protected-area average can conceal operationally important fire heterogeneity. Management units may need different monitoring and fire strategies even inside a single legal boundary.

CAVEATS / UNCERTAINTY

Satellite burned-area products capture fire occurrence more readily than ecological severity, ignition cause or population response. The study is observational, and section-level patterns should not be interpreted as proof that one fire regime is optimal.

SOURCE

Journal of Applied Ecology · Original evidence
View original evidence ↗

Two decades of satellite data from Hwange, Zimbabwe, show strong internal variation in burned area and recurrence even where park-wide totals show no significant trend.

Read the Wild Lands assessment →
Peer-reviewed

Fire hotspots and biodiversity hotspots overlap enough in the Zagros to change where monitoring should concentrate

Evidence
Peer-reviewed
Geography
Southern Zagros Mountains, Iran
System
Mountain forest, steppe and rangeland mosaic

WHAT THE EVIDENCE SAYS

Researchers mapped 13,419 fires from 2000–2023, identified statistically significant spatiotemporal fire clusters, and overlaid them with modeled habitat value for 108 vertebrate species. About 60% of burned area fell within high-value ecological zones, including potential ranges of multiple Red Listed species.

WHY IT MATTERS / OUR READ

The analysis turns a generic fire-risk map into a triage tool by asking where repeated fire exposure and modeled biodiversity value coincide. That can focus field verification and prevention where ecological stakes are highest.

CAVEATS / UNCERTAINTY

The study maps exposure and modeled habitat value, not pre- versus post-fire population change or impact severity. Species distributions were modeled with MaxEnt, and the authors explicitly lacked direct before-and-after biodiversity data.

SOURCE

Environmental and Sustainability Indicators · Original evidence
View original evidence ↗

A 2000–2023 remote-sensing analysis layers recurrent wildfire clusters with modeled vertebrate richness to identify conservation priority zones.

Read the Wild Lands assessment →

Keep exploring

Follow the systems. Find the connections.

WildlifeLandWaterOceansClimateCitiesOne HealthGovernanceMethods / Monitoring
Browse the archive →