Skip to main content

Data from: Microgeography, not just latitude, drives climate overlap on mountains from tropical to polar ecosystems

Cite this dataset

Klinges, David H.; Scheffers, Brett R. (2020). Data from: Microgeography, not just latitude, drives climate overlap on mountains from tropical to polar ecosystems [Dataset]. Dryad.


An extension of the climate variability hypothesis is that relatively stable climate, such as that of the tropics, induces distinct thermal bands across elevation that render dispersal over tropical mountains difficult compared to temperate mountains. Yet, ecosystems are not thermally static in space-time, especially at small scales, which might render some mountains greater thermal isolators than others. Here, we provide an extensive investigation of temperature drivers from fine to coarse scales, and demonstrate that the degree of overlap in temperatures at high and low elevations on mountains is driven by more than just absolute mountain height and latitude. We compiled a database of 29 mountains spanning 6 continents to characterize “thermal overlap” by vertically stratified microhabitats, biomes, and owing to seasonal changes in foliage, demonstrating via mixed-effects modeling that micro- and mesogeography more strongly influence thermal overlap than macrogeography. Impressively, an increase of one meter of vertical microhabitat height generates an increase in overlap equivalent to a 5.26° change in latitude. In addition, forested mountains have reduced overlap – 149% lower – relative to non-forested mountains. We provide evidence in support of a climate hypothesis that emphasizes microgeography as a determinant of dispersal, demographics, and behavior, thereby refining classical theory of macroclimate variability as a prominent driver of biogeography.



We compiled published and unpublished temperature data from a combination of author field collection, personal communications, and public data repositories (see data citations) for a total of 29 mountains on six continents representing gradients of vegetation, environmental degradation, and climate (table A1, fig. A1). Temperatures at low and high elevation sites were recorded on each mountain, as was the difference in elevation between low and high sites (yet of note is that not every low site was at the “bottom” of the mountain, nor the high site always at the mountain peak; in addition slope or aspect both varied between sites). Difference in elevation between low and high sites (henceforth “Δ elevation”) was a parameter of interest (see Mixed-Effects Models), therefore a broad range of elevation gradients from 122 to 3080 meters were represented. Time series length varied from 74 days to 64 years; although sampling of 13 of the 29 mountains did not include every day of the year, seasonal coverage did not vary systematically (see “Mixed-Effects Modeling” for how variation in time series length was accounted for). At each elevation band (low and high), at least one of three vertically stratified microhabitats was monitored: soil, surface, and (in forests only) canopy. Of the 29 mountains, 6 sites had all three vertical strata represented, and 14 had at least two vertical strata represented. Soil temperature sensors varied in depth from 2 to 12 cm into topsoil (although most were between 7-10 cm in depth), surface-level sensors were 1-3 meters above the ground, and canopy sensors were typically between 19-25 m high, placed within canopy foliage. When available, temperatures from multiple plots at the same approximate elevation on a mountain– that is, less than 5% of Δ elevation apart– were averaged. All thermal sensors included in our study were shielded from direct radiation (table A1), which is common in ecological research. While shielding reduces irregular sensor performance due to high solar radiation, it also dampens variation depending on shielding methods (Terando et al. 2017). As a result our estimates of microclimate variation may be conservative, but the majority of sites (17 out of 29) followed best practices as recommended by global micrometeorological networks (Beeck et al. 2018; Rebmann et al. 2018). Both microhabitat category (soil, surface, canopy, henceforth “vertical microhabitat”) and depth belowground or height aboveground of sensor (henceforth “microhabitat height”) were used in analyses.

See methods section of manuscript for more details. Additional details will be provided here upon publication.