A cactus of low islands
Cochemiea halei (formerly Mammillaria halei) forms clusters of reclining stems with straight spines and red tubular flowers. Its main sites lie in the Magdalena archipelago off the western coast of Baja California Sur. This small land area includes different substrates, low plains and the elevated ridges of Santa Margarita. The 2020 study by Breslin and coauthors combined a field map of plants with climate models to assess the future suitability of these habitats. [Breslin et al. 2020, pp. 13212–13214]
Eight field visits over four years, made in winter and spring, yielded 1227 georeferenced observations. These are occurrence records used in the analysis, without a complete census of all living plants in the archipelago. A report of a very small population on the adjacent mainland was also included. After highly clustered records were removed, 44 occurrence points at least 1 km apart remained. This thinning reduced excessive weighting of places that had been visited and recorded more often. [Breslin et al. 2020, pp. 13213–13215, 13218]
Ultramafic substrate as part of the habitat
Ultramafic rocks, including serpentinites, produce substrates with a distinctive mineral composition. In the study, they were identified from field observations of soil colour and a geological map. About 60% of C. halei observations came from such places. Other sites occurred on other substrates, so the species' association with ultramafic soil is facultative. The occurrence map alone does not establish the chemical composition of plant tissues. [Breslin et al. 2020, pp. 13213–13215, 13218–13220]
Two substrate categories were included in the model, ultramafic and other substrates. Adding them reduced predictions of suitable habitat in sandy areas between parts of the archipelago where plants had not been observed. This shows the significance of substrate within a similar climate. The study did not experimentally separate the effects of water availability, mineral nutrients and metals. These processes may jointly shape the root environment, but their separate contributions to this result were not determined. [Breslin et al. 2020, pp. 13216–13220]
From occurrence points to a suitability map
Four of the nineteen WorldClim variables were selected to reduce interdependence of information. They were annual temperature range, mean temperature of the warmest quarter, and precipitation of the warmest and coldest quarters. Climate data represented 1970–2000 on a grid of approximately 1 km. Such a cell can include several rock types, exposures and microclimates. The model therefore describes the environment at the scale of the map used. [Breslin et al. 2020, pp. 13214–13215]
The authors compared a generalized linear model, Maxent and boosted regression trees, abbreviated BRT. In the last approach, successive small trees help describe nonlinear relationships between conditions and occurrence. BRT was selected for projections based on an assessment of its ability to transfer results to data not used in fitting. The final dataset reported in the results contained 44 occurrence points and 207 pseudoabsence points representing the comparison background. [Breslin et al. 2020, pp. 13214–13217; Fig. 3–4]
The result depended particularly on temperature
In the selected model, annual temperature range contributed about 40,5% and mean temperature of the warmest quarter about 21%. Together, they accounted for about 62% of the reported relative predictor importance. Substrate contributed about 16%, while the two precipitation variables made up the remainder. Importance indicates a variable's contribution to fitting a particular model. [Breslin et al. 2020, pp. 13216–13218; Fig. 5]
The model curves indicated lower habitat suitability with a greater annual temperature range and a high mean temperature of the warmest quarter. The climate at archipelago sites is strongly influenced by the sea. Future changes in this pattern may shift the climatic space suitable for the species. The curves are not applied to limiting stem temperatures or a pot-culture schedule, because the study did not measure these physiological responses. [Breslin et al. 2020, pp. 13216–13220]
Habitat loss and the emergence of new areas
The projections covered horizons termed 2050 and 2070 in the study and scenarios RCP 2.6, 4.5, 6.0 and 8.5. The RCPs defined different pathways of future radiative forcing in the climate models used. The authors combined results from HadGEM2-ES and CCSM4. This is the set of projections used in the 2020 study, retaining its assumptions and dates. Each map was compared with the baseline area judged suitable by the model. [Breslin et al. 2020, pp. 13214–13217]
| Horizon and RCP | Loss of current area | Newly suitable area | Net loss |
|---|---|---|---|
| 2050 · 2.6 | 46% | 19% | 27% |
| 2050 · 4.5 | 48% | 10% | 38% |
| 2050 · 6.0 | 47% | 14% | 33% |
| 2050 · 8.5 | 52% | 25% | 27% |
| 2070 · 2.6 | 50% | 29% | 21% |
| 2070 · 4.5 | 50% | 5% | 45% |
| 2070 · 6.0 | 50% | 6% | 44% |
| 2070 · 8.5 | 55% | 2% | 53% |
In every version, some areas classified as suitable at baseline lost that classification. The range was 46–55% of the area. At the same time, the model predicted 2 to 29% of newly suitable area outside the baseline area. After both changes were included, net loss was 21–53%. These two ranges answer different questions. The first describes retention of currently suitable places, and the second the balance of the entire projected map. [Breslin et al. 2020, pp. 13216–13218; Table 1]
Why the balance of area is insufficient
Potential expansion was concentrated partly on the higher ridges of Santa Margarita, while losses included warmer lowlands. A newly suitable map cell indicates agreement with the variables used. Establishment of a population there also depends on seed arrival, availability of germination sites and survival of successive stages. For an island cactus, the spatial continuity of new areas and their substrate matter when assessing the possibility of such shifts. [Breslin et al. 2020, pp. 13217–13220; Fig. 6–8]
Loss of part of the existing area may also change distances between remaining habitat patches. Even with a moderate net balance, a local population may lose favourable surroundings. The publication did not convert the suitability map into future numbers of individuals or use it to determine extinction probability. The model indicates the spatial direction of the threat, which can be related to particular sites and their observed demography. [Breslin et al. 2020, pp. 13218–13220]
Projected climate change threatens significant range contraction of Cochemiea halei (Cactaceae), an island endemic, serpentine-adapted plant species at risk of extinction
Breslin P. B., Wojciechowski M. F., Albuquerque F., Projected climate change threatens significant range contraction of Cochemiea halei (Cactaceae), an island endemic, serpentine-adapted plant species at risk of extinction, Ecology and Evolution, 2020, 10(23):13211-13224, doi: 10.1002/ece3.6914
A study of the distribution of Cochemiea halei in the Magdalena archipelago on the western coast of Baja California Sur. Collected field observations were combined with four climate variables and a map of two substrate categories. Models of current habitat suitability were projected onto four RCP scenarios for the 2050 and 2070 horizons. The paper distinguishes the facultative association with ultramafic substrate, projected losses of currently suitable areas and the emergence of potentially suitable new areas.
Plants of the World Online — Royal Botanic Gardens, Kew
Accepted names, synonyms and distribution. A citation beside a taxon leads to its record.
Accessed 2026-10-08.