Please check out the preprint of my upcoming shared first-author paper with Peter Hajek. We studied the death of thousands of young trees in a tree diversity experiment during the extreme summer drought of 2018. A thread.

#whatkillstrees #TreeDivNet https://www.biorxiv.org/content/10.1101/2020.12.17.423038v1 1/

The 2018 drought and the following dry winter caused a high death toll for the over 20,000 trees of the German branch of the #IDENT tree diversity experiment in Freiburg. 2/

Our analysis is based on the subset 5 x 5 inner trees of each plot that were alive after the 2017 census, resulting in 9,435 trees in total. 34% of these trees (!) died after the drought event.

(aerial photos: @KyleRKovach)
3/

The controlled setting of the #IDENT experiment allowed us to simultaneously look at the 3 mechanisms of drought-induced tree mortality proposed by the seminal hydraulic framework by @McDowell_Trees (2008) - hydraulic failure, carbon starvation and biotic agent demographics. 4/

We used a hierarchical Bayesian model in @mcmc_stan to describe the individual probability of dying of each tree based on species-specific hydraulic traits, NSC depletion, bark beetle infestation, tree size, environmental variables and interactions with neighbor trees. 5/

Our model described 51.8% of the variance in the individual mortality data (for clarity, this figure shows species-wise aggregates on the plot level instead of the binary raw data). 6/

The observed mortality rates differed considerably between species, with average observed mortality rates ranging from 0.6% (Acer saccharum) to 79.5% (Larix laricina).

Plot: observed (circles) and predicted species average mortality rates (posterior mean ± 95% HDI). 7/

Our model indicated that species with narrower hydraulic safety margins and species with increases in the sugar fraction of total NSC content over the vegetation period (which indicates starch-to-sugar conversion, possibly for osmotic adjustment) were much more likely to die. 8/

Taller trees in general were less likely to die, while trees where the presence of bark beetles was recorded in 2018 died in much higher numbers. The effect of infested neighbor trees and of the #IDENT fertilization treatment were negligible for the majority of species. 9/

The height effect varied considerably between spp., but was credibly different from zero on the 95% level for all spp. except Quercus rubra. We suppose that the higher survival of larger trees resulted from more extensive root networks and better access to deep soil water. 10/

The regular planting grid of the #IDENT trial allowed us to estimate how the probability of dying of a tree was affected by the number, size and distance of its direct neighbor trees in a species-specific fashion. 11/

To achieve this, we developed a model for directed neighborhood interactions based on a modified Hegyi competition index.
To our surprise, we found a higher survival of central trees in the presence of *more* and *larger* neighbors of most potential neighbor species. 12/

When looking at all possible neighbor pairs (rows: focal trees; cols: neighbor trees), it's clear that there's a large range of potential interactions & that it's no zero-sum game: increased survival for one species doesn't always translate to a disadvantage for its neighbor. 13/

Interestingly, there were no neighbor pairings that were detrimental for both spp, while a large number of pairings were one-sided or mutually beneficial.

(l. pairs with the large avg. diff., m. pairs with smallest sums, and r. pairs with largest sums of neigh. effects) 14/

Neighbor effects on survival depended strongly on the species identity of the involved trees & were not always positive. A positive biodiversity effect on survival would result only if interspecific interactions on average were more beneficial than intraspecific interactions. 15/

Whether or not biodiversity increases the resilience of a community against drought therefore may be driven by the studied species sample rather than being a general pattern. 16/

Summing up, a species' mortality risk was both associated with hydraulic traits and with changes in carbon metabolism indicative of starch-to-sugar-conversion for osmotic adjustment, which is in line with the assumed close mechanistic link between these two processes. 17/

On the individual level, the mortality risk is modulated by pest infestation, tree-size, and directed and species-specific interactions with neighboring trees. 18/

Thus, our results indicate that to be able to fully understand #whatkillstrees, it is necessary to adapt a holistic perspective on drought induced mortality instead of focusing on single processes and/or risk factors. 19/

Of course, we are grateful to our other coauthors @charlesnock, @cappenberg, Tobias Gebauer, Arthur Gessler, @KyleRKovach , Christian Messier, Alain Paquette, Matthias Saurer, Michael Scherer-Lorenzen, Laura Rose and Bernhard Schuldt for this great collaboration. 20/