Integration of provenance tests and National Forest Inventories to account for intraspecific variation in species distribution models

Background/Question/Methods

One of the biggest challenges for understanding and managing species range shifts in changing climates is that the responses of species strongly vary between populations.

Tree responses to different climates have been addressed through common garden experiments that are regarded as the best tool available for conducting management decisions. However, this type of experiments only cover a portion of the ranges of species, and are not always suitable to answer questions related to climate change.

National Forest Inventories (NFI) on the other hand, offer very valuable opportunities, that when available across climatic gradients, can be complementary to data coming from common gardens to infer climate/vulnerability relationships while accounting for population phenotypic variation in the field. For doing so, an estimation of the phenotypic variation of a given species was estimated with the intraclass correlation measure P_ST, measured on the residuals of a regression between the d.b.h. and the total height across the distribution of Abies alba in France. 

We calculated climate responses to height for A. alba L. nine years old trees measured in four common gardens in France, where provenances coming from all Europe are tested. We used quadratic mixed-models that incorporate annual average temperature and minimum temperature of the coldest month (including interactions) as fixed effects, and site, block and the provenance as random effects.

Results/Conclusions

Results showed that provenances coming from warmer locations would grow faster than those coming from colder ones. Likewise, medium-high values of P_ST calculated for France from NFI showed the existence of phenotypic variation across the range of the species in France confirming Q_ST estimates based on the common garden results.

Phenotypic variation within the range of A. alba was detected by the NFI and common garden data, showing the usefulness of both types of databases to detect the spatial structure of tree growth along large environmental gradients.

In view of these results, spatial predictions for current conditions and for 2080 (representative concentration pathway 4.5) for two populations located at the margins of the distribution have been performed using the random forest algorithm calibrated with data from the NFI as well as from the common gardens. These simulations show the applicability of both databases in assisted migration planning.