1,565 georeferenced records, 12 species, 28 co-occurrence pairs. Most apparent associations are sampling artifacts. Range-shift analysis finds nothing — the study area is too small by ~10×.
54% of records (838/1,565) are humpback whale sightings from December–March. Any species with winter records will appear to "co-occur" with humpbacks regardless of ecological reality. The data supports: (1) distinguishing plausible pairs from sampling artifacts; (2) qualitative habitat groupings; (3) a range-shift test — which finds nothing, as expected, because the study area is too small by ~10×.
Co-occurrence counts are extremely sensitive to when and how records were collected.
Co-occurrence window: 25 km, 24 hours. Lift = observed ÷ expected under independence. Lift 1.0 = random; 5× = five times more often than chance. Lift >5 with low n is a screening metric, not a statistical test.
| # | Species A | Species B | n co-obs | Expected | Lift | Interpretation |
|---|
All 28 unique pairs shown (dataset contains 12 species; many theoretically possible pairs have zero co-observations). Lift values > 5 flagged green; 2–5 flagged amber.
This pair has 0 co-observations in the dataset — the two species' GBIF records do not overlap in space and time within 25 km / 24 hours. Spotted dolphins are recorded primarily in September–November offshore; humpbacks peak December–February nearshore. The apparent seasonal segregation in the GBIF record likely reflects survey-effort patterns rather than avoidance, but the co-occurrence signal is null here.
n = 1 co-observation, lift = 0.29 — a weak negative association. Both species use the Guerrero coastal corridor, but their GBIF records rarely overlap within the 25 km / 24 h window. Olive Ridley nesting records concentrate September–November at beach patrol sites; Green turtle satellite-tracking records span March–April offshore. No evidence of co-nesting in spatial proximity. The corridor is wide (100 km bounding box); species are not segregated, but the dataset does not support a co-nesting claim.
Six qualitative clusters from species metadata and habitat context. Not output of a clustering algorithm — n too small for unsupervised methods.
Climate-driven range shifts in marine and coastal species are typically detectable across hundreds to thousands of kilometers and over multiple decades. The study area here is roughly 25–100 km across — about ten times too small for a meaningful signal. Any apparent geographic drift in where species are recorded is overwhelmingly more likely to reflect changes in where observers went across years than genuine distributional change. These results are exploratory only.
| Species | n obs | n years | Trend (°lat/decade) | Direction | R² | p (raw) | p (Bonferroni) |
|---|
Olive Ridley sea turtle (*Lepidochelys olivacea*) shows an apparent northward drift of about +0.23 degrees latitude per decade (roughly 25 km northward), with an uncorrected p-value of 0.0017 (probably real on its own terms). But it does not survive the multiple-comparison correction for testing 8 species simultaneously (corrected p=0.013, above the required 0.00625). Even setting statistics aside: the pattern most plausibly reflects a shift in where observers collected records. Early records (2000–2006) come from established turtle patrol programs near Ixtapa. Later records (2018–2025) include iNaturalist reports spread across the full study area. That geographic spread alone would produce an apparent northward drift with no turtles actually moving.
Conclusion: treat as null. A genuine range-shift study for Olive Ridley on the Guerrero coast would require beach-patrol data standardized by patrol effort — not opportunistic wildlife sightings submitted to GBIF.
Across 1,565 observations of 12 species (2000–2025), the analysis found 28 pairwise co-occurrence associations. The strongest apparent signal — humpback whales appearing alongside lagoon species like crocodiles and iguanas — is an observer artifact: both groups are recorded in December–March, the winter survey peak, not because the animals actually share habitat. The highest lift score (43.8× for Olive Ridley sea turtle and pantropical spotted dolphin) comes from nine co-observations on a single research cruise in September 2003. That is a sampling artifact, not an ecological association. The ecologically coherent signals are: (1) American crocodile and green iguana co-occurring at the lagoon edge (lift 4.4×, 112 co-observations — plausible shared habitat); (2) Leatherback and Olive Ridley turtles sharing the coastal corridor (lift 21× but only 13 co-observations — treat with caution). Six habitat groupings were identified by combining species metadata and habitat context. Range-shift analysis found no species with a geographic trend that survived correction for multiple simultaneous tests. The study area (~25–100 km across) is too small by roughly a factor of ten for climate-signal detection.
GBIF 2000–2025, hasCoordinate=true. Input: _whale_climatology.js (889 records), _turtle_climatology.js (256), _lagoon_climatology.js (420). Co-occurrence: 25 km haversine, 24 h window; lift = observed ÷ expected under independence (point estimate, no CI). Range shift: annual centroid latitude ~ year, OLS, Bonferroni α=0.00625 over 8 tests. Script: scripts/analyze_co_occurrence.py → functions/api/_findings_co_occurrence.js.
Artifact generated: —