I’m lucky I have a stable office chair. When my collaborator Xugao Wang send me new data on species aggregation of tree species in 21 large forest inventory plots I hurried to plot the new results by plotting aggregation versus abundance. What suddenly appeared before my eyes made me almost fell off my chair…. The intriguing results of this analysis suggest that species aggregation changes systematically along latitude, this is now figure 2 in our paper published this week in Nature, please have a look!

But let’s start from the beginning

We have been working on the quantification of spatial structures in forest communities for almost two decades or so. Why do we think that spatial structures are important? Well, take for example conspecific aggregation of species. It is a key spatial structure because it is closely related to ecological processes such as negative conspecific density dependence, dispersal limitation, mycorrhizal associations, and habitat association. Conspecific aggregation means that individuals of a species are closer to each other than would be expected in a random distribution, or in other words, trees have on average more neighbors of the same species than expected. There are numerous studies quantifying species aggregation in forests, but there has been no assessment of whether aggregation patterns change systematically as one moves along the latitudinal gradient from tropical to subtropical to temperate forest, and how such aggregation patterns would affect species coexistence has remained elusive even almost five decades after Steve Hubbell’s seminal 1979 study.

We were lucky to collaborate with PIs of the Forest Global Earth Observatory network (ForestGEO) of permanent forest dynamics megaplots, which allowed us to analyze aggregation and other spatial patterns for many species in many forest plots across latitude. One of us (Xugao Wang) is PI of the temperate Changbaishan plot, and we have  one ForestGEO plot  in Germany, the Traunstein plot, where Anderas Huth is one of the PIs. In these datasets, every tree with a diameter at breast high only somewhat larger than a pencil (!!!) is mapped, measured, and identified to species every five years on plots of typically 16-50 hectares. Such a plot, especially in a tropical forest, easily comprises more than 100,000 trees to be mapped and measured, an undertaking that often takes more than one year. Our graph below visualizes data of the 50-ha plot of tropical forest at Barro Colorado Island, the oldest plot in the network, which was established in the 1980ies. We show about 90,000 trees with a diameter at breast  height (dbh) larger than 2.5cm, where the diameter of the circles is proportional to dbh. Only the 29 most common species have been colored, all other individuals are represented by black circles.

We now used these data sets to quantify the aggregation of large trees (dhb ³ 10cm) for all species that comprised more than 50 large trees. Following earlier work, we investigated whether the species in a given forest plot exhibited a relationship between aggregation and abundance, and we quantified the strength of such a relationship by the exponent of a power law that we fitted to the data. Now, plotting the exponent of the power law against latitude revealed the surprising emerging latitudinal pattern that almost made me fall off my chair. We found that tropical forests show a large range of aggregation among species, but basically no relationship with abundance. However, the further we moved into temperate latitudes, the stronger the relationship became: in temperate forests, aggregation increases strongly if abundance decreases.

What does this mean?

Such spatial relationships are somewhat difficult to grasp, but we can better interpret this key result if we explain it in terms of the number of neighbors rather than aggregation. Conspecific aggregation is basically the number of conspecific neighbors within a given distance from the “typical” individual of the species, divided by number of neighbors expected from a random distribution. Note that the number of neighbors expected from a random distribution is proportional to species abundance. Now, we found that there is no correlation between aggregation and abundance in tropical forests, and as a consequence, the number of conspecific neighbors (which is proportional to aggregation multiplied by abundance) is smaller for less abundant species, a pattern that can lead to a rare species advantage under negative conspecific density dependence. However, in temperate forests, where aggregation is negatively correlated with abundance, trees of rare and abundant species have more or less the same number of conspecific neighbors. This is really a surprising pattern, but which processes cause them and what consequences does this have for the coexistence of tree species?

In our new study we found evidence that animal seed dispersal and mycorrhiza were involved in the formation of the mysterious latitudinal pattern in aggregation, and that they led to contrasting spatial structures in temperate versus tropical forests, which each promote coexistence. Let us take you now through the journey behind our study. All started with the idea that spatial patterns might stabilize multispecies communities such as forests. To formalize this idea, we developed in our previous 2021 study, published in Nature Ecology and Evolution, an upscaling method that allowed us to incorporate measures of spatial aggregation into larger-scale models of the Lotka-Volterra type. To test our theoretical results we used an individual-based and spatially-explicit implementation where a dispersal kernel placed new recruits close to their parents. However, to our deep disappointment, the model did not show the expected coexistence, but was quite unstable. So what happened? Looking closer we found that species aggregation in our simulations was not constant, as assumed by our theory, but depended on species abundance, as described above for temperate forests!

But what mechanisms can lead to the observed strong aggregation of less abundant species?

Well, when most recruits are placed close to their parents, locally high densities of adults are generated by the continuous introduction of new recruits into these clumps, with local densities kept in check by density-dependent mortality. This mechanism results in individuals always having approximately the same number of neighbors, regardless of whether the species is abundant or not, which leads to a negative aggregation–abundance relationships given that neighborhood density is proportional to abundance time aggregation. So we learned an important lesson: the placement of recruits with a dispersal kernel caused the instability of our model and led to a strong negative relationship between aggregation and abundance. But is there a way to stabilize the dynamics of our model? In new model simulations we followed the idea of animal seed dispersal: We still placed the recruits in small clumps to generate aggregation, but the clumps were distributed to random locations in our simulation area. Miraculously, the simulations showed stable dynamics, aggregation was largely independent on species abundance, as observed for tropical forests, and our theoretical predictions matched! Indeed, roughly 70 to 80 percent of tree species in the tropics are dispersed by animals, but much less in temperate forests.

But how could seeds and juveniles in temperate forests escape the detrimental Janzen-Connell effect?

This is where mycorrhizae come into play. Temperate forests are usually dominated by so called ectomycorrhizal (EM) tree species, while tropical forests are dominated by arbuscular (AM) mycorrhizal tree species. Interestingly now, the host specific EM fungi provide juveniles physical root protection, allowing them to recruit close to conspecific parent trees, whereas this is not the case with AM fungi. Indeed, several studies using spatial analyses found that EM tree species in temperate forests tend to show higher levels of aggregation than AM tree species. Thus, seeds in temperate forests should have an interest in remaining close to their parents, but seeds in tropical forests should have an interest in moving away from them. If this hypothesis was true, we should observe also a relationship between latitude and the proportion of tree species in a forest that are mostly dispersed by animals and have an AM association. So Xugao Wang gathered additional information with the help of the plot PIs , now about seed dispersal and mycorrhizal association for the 720 tree species used in our analysis. When I plotted the results over latitude, my chair saved me again. The emerging strong relationship we found is now figure 3a in our paper. And, at about the same time when we got our exciting results, a new study was published about the joint evolution of mutualistic interactions, pollination, seed dispersal mutualism, and mycorrhizal symbiosis in trees (Yamawo & Ohno 2024) that supported our hypothesis, based on evolutionary relationships.

To uncover the consequences of the surprising latitudinal pattern in the response of aggregation to abundance, we now expanded our theory to include the power-law relationship into our model equations. This has been a somewhat arduous and lengthy process, as nicely documented in the peer review file that is provided together with our article. The reason for our initial difficulties was our idea that temperate forest should have a disadvantage because of their destabilizing negative aggregation-abundance relationships. But, our data did not really show this and the reviewers scratched their heads. It took us a while to interpret our theoretical results correctly. In short, our individual-based simulations showed a destabilizing effect when we changed the way recruits were distributed, but we assumed in the simulations that all else would be equal. But “all else being equal” did not apply for our data, especially not for the spatial patterns that entered our equations. Eventually we discovered a simple way to express the conditions for species to persist. Overall, it turned out that species in tropical and temperate forests exhibit optimal – but contrasting – spatial structures that each promote coexistence, and animals and mycorrhiza seem to be involved in the maintenance of these contrasting spatial structures.

Our new study is now the starting point for further research. We want to develop a more general theory for understanding the spatial dynamics and stability of species-rich forests. As next steps we will substantially expand our methods and analyses, for example by taking into account the size of the trees, the immigration of species and more detailed species characteristics, as well as by using remote sensing data. This is required to place the spatial coexistence mechanisms in perspective of other mechanisms. The direct link of our theory to spatial patterns gives access to the rich source of data to validate spatial simulations of alternative models against multiple biodiversity patterns. The most exciting part? We were recently awarded an ERC advanced grant that provides us optimal conditions to develop our ideas further.

References

Hubbell, S. P. Tree dispersion, abundance, and diversity in a tropical dry forest. Science 203, 1299–1309 (1979).

Wiegand, T. et al. Consequences of spatial patterns for coexistence in species-rich plant communities. Nat. Ecol. Evol. 5, 965–973 (2021).

Yamawo, A. & Ohno, M. Joint evolution of mutualistic interactions, pollination, seed dispersal mutualism, and mycorrhizal symbiosis in trees. New Phytol. 243, 1586–1599 (2024).