Lima’s Endangered Fog-Oasis Tree Faces a Pincer of Climate Change and Sprawling City Growth

On the coastal hillsides of Lima, one of the world’s largest desert-built metropolises, a rare tree known locally as mito survives on moisture it harvests directly from winter fog. Vasconcellea candicans, a Critically Endangered species found nowhere else, anchors the seasonal plant communities called lomas—fog oases sustained by dense banks of coastal mist below 1,000 meters, generated where the cold Humboldt Current meets the stable subtropical anticyclone. A new study published in Ecology and Evolution has, for the first time, combined ensemble species distribution modelling with spatially explicit urban growth projections to quantify how much of this tree’s habitat is being squeezed between a shifting climate and an expanding city. The answer is stark: the area of high-quality habitat swallowed by urban land is projected to nearly quadruple by 2070, and most of it lies outside any protected boundary.

The research team, led by Rebeca M. Cruz-Zegarra of the Universidad Peruana de Ciencias Aplicadas, focused on the department of Lima, a 35,892-square-kilometer region harbouring the highest density of lomas in Peru. Peru hosts 67 known fog oases covering up to 70,000 hectares at peak fog incidence, and five around Lima—Ancón, Carabayllo 1, Carabayllo 2, Amancaes, and Villa María—have been designated as the Regional Conservation Area Sistema de Lomas de Lima and classified as fragile ecosystems by the national forest service. Yet these ecosystems face the highest anthropogenic stress of any comparable system in the country, driven by informal urbanisation, land trafficking, invasive species, and recurrent fires. Because V. candicans acts as an umbrella species—protecting its habitat also protects the plants and animals sharing its fog-fed microhabitat—and serves as a bioindicator of ecosystem health through its atmospheric water interception, the team treated it as a practical conservation focal point for the wider lomas community.

To map where the tree could live now and in the future, the researchers assembled 101 validated occurrence records, most drawn from citizen-science observations on iNaturalist, spatially thinned to one point per 250-meter grid cell to reduce sampling bias. They started with 31 environmental variables spanning 19 bioclimatic layers from WorldClim, three topographic layers from the Shuttle Radar Topography Mission, and nine soil variables from SoilGrids, then pruned collinear predictors using variance inflation factors until 12 remained. Future climates came from three structurally independent general circulation models—ACCESS-CM2, IPSL-CM6A-LR, and MIROC6—run under two Shared Socioeconomic Pathways: the moderate SSP2-4.5 and the high-forcing SSP5-8.5, projected to 2050 and 2070.

Rather than relying on a single algorithm, the team built an ensemble in the biomod2 platform, evaluating nine modelling approaches and retaining the five that exceeded strict performance thresholds of AUC above 0.90 and true skill statistic above 0.70: Random Forest, MARS, gradient boosting, generalised linear models, and MaxEnt. Random Forest achieved the highest discrimination (AUC 0.968), and all five converged on a consistent hierarchy of environmental drivers. Temperature annual range emerged as the dominant predictor at 27.2% mean importance, followed by soil silt content at 25.4%, slope, total nitrogen, and cation exchange capacity. This combination is ecologically telling: the narrow thermal envelope imposed by the fog-generating ocean-atmosphere system governs the balance between moisture interception and evaporative demand, while the fine-textured, organically active soils reflect fog-driven pedogenic processes on the hillsides.

Under current conditions, the ensemble identified 8,500 square kilometers of suitable habitat across Lima—24.28% of the department—but only 2,891 square kilometers, or 8.26%, qualified as high suitability. That suitable habitat is strikingly concentrated: 69.30% of officially designated fragile ecosystems and 45.39% of fog oases were classified as suitable, and fragile ecosystems held 46.85% of their area in the high-suitability class. Within the protected area framework, Regional Conservation Areas contained 15.15% high-suitability habitat while National Protected Areas contained just 1.18%, revealing a pronounced mismatch between the formal protection network and the species’ distributional core.

Climate projections diverged dramatically depending on the emissions pathway. Under the low-projection scenario (IPSL-CM6A-LR with SSP2-4.5), high-suitability habitat declined gradually, from a 2.50% loss by the 2050s to a 4.08% loss by the 2070s. Under the high-projection scenario (ACCESS-CM2 with SSP5-8.5), the model instead projected expansion of 28.16% by 2050 and 56.35% by 2070, plausibly reflecting an upward altitudinal shift of suitable conditions. The authors caution, however, that such gains should be read as an upper bound rather than an expectation: high-emissions trajectories also intensify El Niño variability and disrupt the Humboldt Current dynamics sustaining fog formation, processes not fully captured by bioclimatic variables, and ACCESS-CM2 carries a documented energy imbalance that may inflate its signal. Projected climatic gains also do not guarantee colonisation, since dispersal and establishment constraints fall outside the modelled niche.

The urban half of the analysis drew on three decades of Landsat-derived land cover data from MapBiomas Peru. Between 1990 and 2020, Lima’s urban footprint nearly doubled, expanding from 429.0 to 848.5 square kilometers, with the fastest relative growth in the 1990s (3.864% per year) and a progressive deceleration thereafter. Kernel density mapping showed growth hotspots migrating outward from the historic core toward peripheral districts—Carabayllo, Pachacámac, and Ancón—while landscape metrics revealed consolidation rather than classic sprawl: patch numbers fell from 1,615 to 1,051 as mean patch size tripled, and the compactness index rose, indicating coalescing urban fragments that progressively sever connectivity between lomas patches.

Projecting that trajectory forward with a Markov chain and cellular automata model validated at 90.40% correctness and a kappa of 0.864, the team estimated that Lima’s built area will grow from 843 square kilometers in 2020 to 1,286 by 2050 and 1,563 by 2070—an 85.4% cumulative increase. Overlaid on the suitability surfaces, the consequences are severe. Today, 36.12 square kilometers of high-suitability habitat already coincide with urban land, a non-recoverable baseline loss. By 2050 that figure nearly triples to 99.69 square kilometers, and by 2070 it reaches 126.94 square kilometers, a 251% increase concentrated along the southern periurban front through Villa María del Triunfo, Pachacámac, and Lurín, extending into Cañete province. Crucially, most high-suitability habitat at risk lies outside any protected boundary, underscoring the limited buffering capacity of the current conservation framework.

The study’s central message is that single-threat assessments systematically misjudge extinction risk. Climate-driven habitat gains under high emissions may be partly erased by simultaneous urban encroachment into the same newly suitable areas, a direct synergy that only integrated modelling can reveal. Because urbanised land is permanently removed from ecological function, encroachment is the most structurally irreversible of the two threats, and edge effects—altered microclimate, dust, and pollution—likely degrade adjacent habitat beyond the mapped footprint. The authors argue that the 2020–2050 window offers the greatest opportunity for intervention before urban consolidation forecloses it. Their priorities are spatially explicit: extend the Regional Conservation Area network to unprotected high-suitability fragments, fix existing fragile-ecosystem polygons as no-build urban growth boundaries, and embed habitat suitability mapping into Lima’s metropolitan planning. Where formal designation lags, private conservation areas offer a faster route, as demonstrated at Lomas Cerro Campana. The framework itself, the team notes, is directly transferable to other endemic taxa at the urban–desert interface across coastal South America, where rapid urbanisation and climate change are converging as the dominant drivers of biodiversity loss.

Subject of Research: Climate change and urban expansion impacts on the habitat of the endemic Peruvian fog-oasis tree Vasconcellea candicans in Lima

Article Title: Assessing Habitat Suitability and Exposure of the Endemic Vasconcellea candicans Under Climate Change and Urban Growth in Lima, Peru

Article References: Cruz‐Zegarra, R. M., Gutierrez‐Gutierrez, J. R., Rivera‐Fernandez, A. S., Cotrina‐Sanchez, A., & Astete, S. (2026). Assessing Habitat Suitability and Exposure of the Endemic Vasconcellea candicans Under Climate Change and Urban Growth in Lima, Peru. Ecology and Evolution, 16(10), Article e74437. https://doi.org/10.1002/ece3.74437

Image Credits: AI Generated

DOI: 10.1002/ece3.74437

Keywords: Vasconcellea candicans, fog oases, lomas, Lima, species distribution models, urban expansion, climate change, endemic species, conservation, Peru, habitat suitability, biodiversity