Proximal microclimate

Posted on June 28, 2024 by lembrechtsjonas

Spatial resolution is one thing. Temporal resolution an other. The microclimate community has been working hard to improve both, in a continuous search for better microclimate data.

However – and this might be slightly shocking – both are largely missing the point. What we should be aiming for instead, is an improved climate proximity.

The three dimensions of microclimate: spatial and temporal resolution, and proximity.

This ‘climate proximity’ s a new term we introduce in a paper just published in Global Ecology and Biogeography, and it refers to how well climate data represent the actual conditions that an organism is exposed to. This could, but doesn’t have to, relate to the spatial and temporal resolution of your climate data. More importantly, it integrates the important biophysical mechanisms that create the microclimate conditions your study organism is exposed to.

‘Ok, nice theory, Jonas,’ you might say. ‘But can you prove this actually works?’ Oh, yes, we can, and in this new paper, we do so. We compare the accuracy of two macroclimate data sources (ERA5 and WorldClim) and a novel mechanistic microclimate model (microclimf) in predicting soil temperatures (using data from the SoilTemp database. Then, we use ERA5, WorldClim and microclimf to test ecological models in three case studies: temporal (fly phenology), spatial (mosquito thermal suitability) and spatiotemporal (salamander range shifts) ecological responses. In all three cases, one would expect the more proximal microclimate model to do a better job.

The spatial and temporal resolution, and proximity, of the three climate sources used in our study. On the right, you see a list of proximal mechanisms, and how much they are included in the different climate sources

And, oh boy, did that microclimate model live up to our expectations! For predicting soil temperatures, microclimf had 24.9% and 16.4% lower absolute bias than ERA5 and WorldClim, respectively. Even more mindboggling, across the case studies, we find that increasing proximity (from macroclimate to microclimate) yields a 247% (yes, you read that correctly) improvement in performance of ecological models on average! That is compared to a meager 18% and 9% improvements from increasing spatial resolution 20-fold, and temporal resolution 30-fold, respectively.

Emergency rate predictions for ground-dwelling larvae of two crop pest insects in Canada, showing how our microclimate model (microclimf) gets substantially closer at predicting the emergency of the flies.

Temperature predictions (panel a) by ERA5, WorldClim and *microclimf* were similar, yet marginal differences among the three temperature products yielded disparate calculations of growing degree days (panel b). The end result were substantial differences in estimates of insect emergence (panel c): the microclimate model had an average error of 6.57 days, while the next best macroclimate one had already 17.0 days of error.

The paper thus concludes that increasing climate proximity, even if at the sacrifice of finer climate spatiotemporal resolution, may be the way to go to improve ecological predictions. Importantly, that implies we have to use biophysically informed approaches, rather than generic formulations, when quantifying ecoclimatic relationships. Mechanisms first, data second!

Here also, differences in traditional variables like Bio1 were minimal, but for ecologically relevant parameters like fecundity, microclimf generated highly different predictions, with 50% more eggs/female/day than the two other climate sources.

Klinges et al. (2024) Proximal microclimate: Moving beyond spatiotemporal resolution improves ecological predictions. Global Ecology & Biogeography. https://doi.org/10.1111/geb.13884