Mitigating climate change effects on plants through assisted migration with the application of a modeling approach: A lesson from forest trees

Background/Question/Methods

Changing climate may elicit various population and species responses, including migration and the acquisition of novel adaptations.  In some cases, human intervention may be needed to facilitate population and species survival, especially for those with habitat boundaries limited by physical or environmental barriers or for species with a slower migration pace.  Loblolly pine (Pinus taeda L.) is a forest tree species with an extant broad and largely continuous natural range.  It can serve for modeling migration, however, as it was likely constricted to two southern refugia during the most recent glaciation and has since been expanding northward.  Previous work showed a correlation between height growth and the minimum winter temperature.  Moreover, individuals planted somewhat north of their site of origin grew taller than the locals.  This may indicate a disjunction between the current climate and the distribution of local adaptations, possibly lagging behind historically warming climate.  Despite relatively high heritability of height in loblolly pine, modeling growth has been challenging because of confounding effects of within-population variation.  We used multinomial logit regression to model growth response to selected climate variables.  We utilized historical data from the Geographic Seed Source Study, established and maintained by Western Gulf Forest Tree Improvement Program.

Results/Conclusions

We chose planted tree volume at age 15 as the response variable, and mean minimum temperature of the coldest month, growing season precipitation, and their respective variation measures as explanatory variables.  The planted tree volume was categorized into five performance categories following its normal distribution.  In general, the analysis supported the earlier findings regarding loblolly pine growth, suggesting that the families with the optimal adaptations for a given planting site are located southward.  The impact of precipitation seemed to be more complex, possibly due to less regular precipitation pattern in the west.  The developed model, termed Categorical Universal Response Function (CURF), showed good performance according to the AUC score, and mostly agreed with the current consensus on loblolly pine seed movement observed in managed forests.

By assigning performance categories, the proposed approach provides an alternative way to account for the unexplained portion of variation in more commonly used regression methods.  The major potential application of this approach is to provide support for assisted migration decisions.  The method is flexible and readily implementable in decision support systems.  Moreover, as illustrated by the analysis of loblolly pine performance, the method can be generalized to any species with categorical response variables.