Pests and diseases can have a disastrous effect on plants and the surrounding ecosystem of which they are a part. While government agencies often try to control and restrict the movement of flora and fauna, international travel and trade facilitates the spread of dangerous pathogens. Is it possible to predict the possibility of plant disease or infestation worldwide?
We provide a tool that can be used to project—in geographic space—the probability of discovering different types of pest-host interactions (e.g., generalist or specialist). To do this, we consider the pest’s host ranges and their phylogenetic constraints, or what would be equivalent to the species’ genealogy. We hypothesize that closely-related species should be more vulnerable to infection if a close relative has suffered a particular disease.
Although this hypothesis seems like a very simple step towards our understanding of biotic interactions, it represents an advantage in the suggestion of actual or potential biological invasions and biotic interactions, but in a spatially explicit context. We demonstrate this methodology using ambrosia beetles and the plants with which they interact; however, this protocol can be used to analyze almost any other set of interacting taxonomic groups.
Original work by Greg Gilbert (University of California, Santa Cruz) and his colleagues yielded nice evidence on how phylogenetic proximity among host plants can lead to a larger or lower probability for sharing diseases. We thought this general framework could be used in tandem with spatially and environmentally-explicit scenarios where these hosts (and associated pathogens) occur (or can potentially distribute if they haven’t reached a particular site but conditions or facilitated dispersal allowed them to establish in other localities). Using this framework, we sought to develop a tool for phytosanitary risk assessment that combines biological information from many sources and puts it in a geographic context.
Phylogenetic signal and probability of host plant genera sharing beetle species. Curves are predicted from logistic regressions. Four relationships are highlighted: a narrow host range and phylogenetically constrained species (Xyleborus xylographus), a wide host range and phylogenetically constrained species (Xyleborus glabratus), a wide host range and phylogenetically dispersed species (Xylosandrus crassiusculus), and all beetle species together.
Our series of algorithms utilize obtaining maps that can advertise locations in which hosts might be more vulnerable; agencies and the general public concerned with predicting risk and location of invasions by different biological agents can currently employ these techniques. One can use this process as a general tool for any group of organisms — not only plants, beetles, and pathogens. With a bit of effort, our algorithms can be used as a general framework to construct more comprehensive and robust tools for sanitary measures, thus uniting other dimensions of biodiversity. This is much like a Lego© puzzle, where one combines pieces to create a nice and interesting toy useful for multiple biological analyses.
Our method provides computer tools that are easy to implement and replicate to evaluate phytosanitary risks using information freely available on the internet. The tools display such information in easy-to-interpret maps.
There are important limitations to successfully projecting the probability of a host plant sharing a pest or pathogen in geographic space. Firstly, this approach requires good information both about the phylogenies of the species of interest and their geographic occurrence. Ideal data would include resolved phylogenies for plants, their pathogens, or other agents interacting with them, with good resolution at the species level. Occurrence information (precise geographic coordinates where species have been observed or collected) is also biased not only by taxonomic group (some groups have been better recorded than others) but also by accessibility (e.g., a road is more accessible than a mountain peak). Relevant to this issue is the problem of detectability, especially when dealing with pathogens or other biological agents that are much more difficult to identify. This implies that not all desired predictions from a phytosanitary perspective can be done with absolute confidence. Use of taxonomic trees instead of phylogenies could solve part of the problem, but we would definitely need to improve our knowledge on species genealogical relationships from robust systematic and phylogenetic methods.
Global host plant genera distribution for where studied ambrosia beetles are most likely to be found. Higher values coincide with higher probability for interaction between plant genera and ambrosia beetles.
Additionally, to ensure that the model predictions are accurate, we need to test them with raw data derived from field experiments and other databases. Although our results match what has been observed in the field for a few species of beetles and the plants with which we know they interact, more work must be done on many realms to improve the availability of biological data useful for achieving valid predictions and respectable measures of uncertainty on the maps that are generated with this protocol.
For more information, see the corresponding paper.