Patchy distribution of a Madrean Sky Island squirrel shaped by historical habitat configuration

This study demonstrates that the patchy distribution of the Arizona gray squirrel across the Madrean Sky Islands is primarily shaped by historical habitat size and connectivity during the Last Glacial Maximum rather than current conditions, highlighting a legacy of Pleistocene connectivity and the need for management strategies that account for historical refugia and future climatic shifts.

The Gist

Imagine the mountains of southeastern Arizona and New Mexico as a giant archipelago of "Sky Islands." These are tall, forested peaks rising out of a hot, dry desert sea. Just like real islands in an ocean, these mountains are isolated from one another by the "water" of the desert, which is very hard for forest animals to cross.

This paper is a detective story about the Arizona Gray Squirrel, a squirrel that lives only in these specific mountain forests. The researchers wanted to answer a big question: Why are squirrels living on some of these Sky Islands but missing from others, even when those empty mountains look like perfect squirrel homes today?

Here is the story of their discovery, explained simply:

1. The Mystery: The "Ghost" of the Past

If you look at a map today, you might see a mountain range that has plenty of trees, water, and food. You'd expect squirrels to be there. But often, they aren't.

The researchers asked: Did the squirrels just fail to find these places? Or is there a deeper reason?

They tested a theory called the "Constraint-Based Dynamic Island Biogeography" (C-DIB) model. Think of this model like a time machine. It suggests that where an animal lives today isn't just about the weather now; it's heavily influenced by the landscape from thousands of years ago.

2. The Time Travel: The Ice Age Connection

To solve the mystery, the team looked at four different time periods:

• The Last Glacial Maximum (LGM): About 21,000 years ago, during the Ice Age. It was cooler and wetter.
• The Mid-Holocene: About 6,000 years ago. It was hotter and drier.
• The Present: Today.
• The Future: A projection for the year 2100.

The Big Reveal:
During the Ice Age (LGM), the "desert sea" between the mountains was actually filled with forests. Imagine the desert floor turning into a giant, continuous carpet of trees. This allowed squirrels to walk easily from one mountain to another, like walking across a bridge.

The researchers found that the squirrels living on the mountains today are mostly the descendants of those Ice Age travelers.

• The Winners: Mountains that were big and well-connected during the Ice Age still have squirrels today.
• The Losers: Mountains that were small or isolated even back then are still empty today, even if the climate is perfect for squirrels right now.

The Analogy:
Think of it like a family reunion.

• Scenario A: A family lived in a big house together 20 years ago. Even if the house is now broken up into small apartments, the family members still live there because they never left.
• Scenario B: A different family lived in a tiny, isolated cabin 20 years ago. Even if you build a brand new, perfect mansion next door today, that family won't move in because they never had the chance to meet the neighbors back then. The squirrels are the family that stayed; the empty mountains are the ones they never reached.

3. The "Hysteresis" Effect: Why Can't They Just Move In Now?

You might ask, "If the weather is good now, why don't squirrels just walk over from the nearest mountain?"

The answer is isolation.

• The desert between the mountains is a "no-go zone" for these squirrels. They need trees and water to survive.
• During the Ice Age, the "bridge" (the forest) was there.
• When the climate got hotter and drier (the Mid-Holocene), the bridge disappeared. The forests shrank back up the mountains, leaving the squirrels stranded on their peaks.
• Because the squirrels can't cross the hot desert, they can't recolonize the empty mountains. They are stuck in a "distributional disequilibrium." In simple terms: They are living in a landscape that hasn't caught up with the climate yet.

4. The Future: A Shifting Home

The study also looked at the future. As the climate continues to warm, the "perfect squirrel weather" is moving.

• Upward Shift: The squirrels are being pushed higher up the mountains, closer to the peaks.
• The Danger: If they get pushed too high, they run out of mountain! They hit a "summit trap" where there is nowhere left to go.
• The Hope: Interestingly, the model suggests that in some areas, the climate might actually get wetter again in the future, potentially allowing forests to grow back down the slopes. This could create new "bridges" for the squirrels to cross, if we protect the rivers and trees that connect them.

The Takeaway: What Should We Do?

The paper concludes with a clear message for conservationists:

1. Don't just look at today: When planning how to save these squirrels, we have to remember their history. Some mountains are empty not because they are bad homes, but because the squirrels never got there.
2. Build the bridges: We need to restore riparian corridors (the tree-lined river valleys). These are the only safe "bridges" across the desert. If we protect and plant trees along these rivers, we might help the squirrels cross the gaps they missed thousands of years ago.
3. Protect the big islands: The mountains that were big and connected during the Ice Age are the most important refuges. They are the "safe houses" where the squirrel populations have survived for millennia.

In a nutshell: The Arizona Gray Squirrel's map is a fossilized map of the Ice Age. To save them, we need to understand that their current home is a legacy of the past, and to help them survive the future, we need to rebuild the connections they lost long ago.

Technical Summary

1. Problem Statement

The study addresses the biogeographical puzzle of the Arizona gray squirrel (Sciurus arizonensis), a riparian specialist with a fragmented distribution across the Madrean Sky Islands (isolated mountain ranges in the southwestern US and northern Mexico). Despite the presence of climatically suitable habitat in many unoccupied mountain ranges, the species remains absent from them.

The authors aim to test whether current species distributions are driven primarily by contemporary landscape conditions (current habitat area and connectivity) or by historical legacies (habitat configuration during past climatic periods). Specifically, they evaluate the Constraint-Based Dynamic Island Biogeography (C-DIB) model, which posits that historical connectivity and habitat size create "hysteresis" (ecological inertia), where populations established during periods of connectivity persist even after isolation, while historically isolated ranges remain unoccupied despite current suitability.

2. Methodology

Study System and Data:

• Target Species: Sciurus arizonensis.
• Study Area: 29 U.S. mountain ranges in the Madrean Sky Islands (Arizona and New Mexico) rising above ~1,500 m.
• Data Source: 187 georeferenced museum specimen records (GBIF), thinned to 157 unique records to reduce spatial autocorrelation. Occurrence was classified as binary (1 = occupied, 0 = unoccupied) based on verified presence.

Species Distribution Modeling (SDM):

• Algorithm: Maxnet (implemented in the Wallace R package).
• Predictors: 19 CHELSA bioclimatic variables were screened; 5 were retained (BIO01, BIO07, BIO12, BIO15, BIO17) based on Variance Inflation Factors (VIF < 6) to minimize multicollinearity.
• Temporal Projections: The model was projected across four time periods:
  1. Last Glacial Maximum (LGM): ~21,000 years ago.
  2. Mid-Holocene (MH): ~6,000 years ago.
  3. Present: 1970–2000.
  4. Future: 2071–2100 (CMIP6 SSP2-4.5 scenario).
• Thresholding: A 10th-percentile training presence threshold (p10) was applied to convert continuous suitability maps into binary habitat layers for area calculations.

Connectivity and Spatial Analysis:

• Resistance Surfaces: Climatic suitability grids were converted to resistance surfaces ($Resistance = 1 - Suitability$).
• Least-Cost Paths (LCP): Using GRASS GIS (r.cost and r.drain), the study calculated cumulative dispersal costs from mainland source forests (Mogollon Rim/Rocky Mountains) to the highest-suitability pixel on each sky island.
• Metrics: For each time period, two predictors were derived for each island:
  1. Habitat Area: Total climatically suitable area within the island boundary.
  2. Dispersal Cost: The minimum accumulated cost to reach the island.

Statistical Analysis:

• Modeling Approach: Firth's penalized logistic regression (using the logistf package) was used to handle the small sample size (5 presences vs. 24 absences) and prevent separation bias.
• Hypothesis Testing:
  • H1 (Temporal Legacy): Tested period-specific models (LGM, MH, Present) to see if a single historical period best predicts current occupancy.
  • H2 (Structural Drivers): Used Principal Component Analysis (PCA) to create composite variables for "Area across time" and "Cost across time" to test additive effects.
• Evaluation: Models were compared using AICc (corrected Akaike Information Criterion), AUC (Area Under the Curve), Cross-validated AUC, and Brier scores.

3. Key Results

Model Performance:

• The LGM model provided the best fit (AICc = 28.1; AUC = 0.842), marginally outperforming the Present model (AICc = 28.2; AUC = 0.758) and significantly outperforming the Mid-Holocene model.
• The Structural (PCA) model showed moderate support but was weaker than the LGM-specific model.

Predictor Effects:

• LGM Habitat Area: Positively associated with current occupancy ($\beta = 0.80$, OR = 2.23), though not statistically significant at $p < 0.05$ ($p=0.103$).
• LGM Dispersal Cost: Negatively associated with occupancy ($\beta = -1.13$, OR = 0.32), indicating that lower historical dispersal costs (better connectivity) increased the probability of current presence. This was marginally significant ($p=0.063$).
• Consistency: The direction of effects for the Present model mirrored the LGM model but with slightly reduced effect sizes, suggesting a strong historical signal that persists.

Habitat Shifts:

• Elevational Shifts: The mean elevation of suitable habitat shifted non-linearly: rising from 1,643 m (LGM) to 1,792 m (MH), dropping to 1,723 m (Present), and projected to rise to 1,782 m (Future).
• Geographic Shifts: The centroid of suitable habitat moved northwest (LGM to MH), then southwest (MH to Present), and northwest again (Future).
• Connectivity: The LGM period featured extensive, connected low-elevation woodlands, whereas the MH and Present periods show fragmented, isolated patches.

4. Key Contributions

1. Validation of C-DIB in Mammals: The study provides empirical support for the Constraint-Based Dynamic Island Biogeography model in a terrestrial mammal system, demonstrating that historical connectivity (LGM) is a stronger predictor of current distribution than current habitat configuration.
2. Evidence of Distributional Disequilibrium: The findings confirm that many sky islands remain unoccupied by S. arizonensis despite having suitable contemporary climate, due to dispersal limitation and the inability to recolonize historically isolated ranges.
3. Methodological Rigor: The application of Firth's penalized logistic regression effectively addressed the "small sample size" problem common in island biogeography studies, providing stable parameter estimates where traditional logistic regression would fail.
4. Spatiotemporal Dynamics: The study quantifies the non-linear elevational and geographic reconfiguration of the species' niche over the last 21,000 years, highlighting the dynamic nature of sky island habitats.

5. Significance and Implications

• Conservation Strategy: The results argue against a purely "current climate suitability" approach to conservation. Protecting historically connected refugia (large, well-connected LGM habitats) is critical.
• Restoration Priorities: To facilitate range expansion and genetic exchange, conservation efforts must focus on restoring riparian corridors that act as dispersal pathways across desert lowlands, effectively "rewiring" the historical connectivity lost during the Mid-Holocene.
• Climate Change Adaptation: As climate warms, suitable habitat is projected to shift upslope and geographically. The study suggests a risk of a "summit trap" where species are pushed to mountain tops with no further upslope escape. However, future scenarios also suggest potential downslope habitat recovery, provided corridors remain intact.
• Management: Management strategies must be mountain-range-specific, acknowledging that each island has a unique history of isolation and connectivity that dictates its current occupancy status. Assisted migration may be necessary for isolated ranges but requires genomic assessment to avoid disrupting local adaptations.

In conclusion, the patchy distribution of the Arizona gray squirrel is not merely a function of current climate but a legacy of Pleistocene landscape configuration. Effective conservation requires integrating paleoecological history into modern landscape planning to restore functional connectivity in fragmented sky island ecosystems.