Mon, Aug 15, 2022: 4:30 PM-4:45 PM
520E
Background/Question/MethodsEcologists are making increasing use of joint species-distribution models to predict how changes in climate or habitat suitability are expected to lead to shifts in species distributions. Recent years have also seen tremendous progress in the development of statistical models that predict interactions between co-occurring species, with a particular focus on capturing the role of trait matching (e.g., between predators and prey or plants and pollinators). The time therefore appears ripe for some sort of unification of these two approaches to enable predictions of both (i) when and to where a species' distribution is expected to shift and (ii) with which other species we expect it to interact after doing so. With this in mind, we will describe our own efforts at combining these two ideas together to study interactions between host trees and climbing plants in a secondary lowland tropical forest in Singapore.
Results/ConclusionsOur analyses indicate that there are numerous barriers to unambiguously linking joint species-distribution models and predictive models of species interactions. These obstacles are pervasive across many common ecological contexts, they strongly impact analysis capabilities, and perhaps most importantly they hinder the biological utility of many parameters that emerge from statistical inference. Chief among these include: the information loss that occurs when modelling interaction presence-absence rather than interaction frequency; the limits created by lacking intraspecific variation in species trait values; the ambiguities that arise without independent estimates of background species densities; and the complexity of linking statistical interaction models to appropriate functional-response equivalents. Though some of these can be overcome with additional and improved data collection, others will require substantial reconsideration of how ecologists interrogate and confront statistical and mathematical models with data.
Results/ConclusionsOur analyses indicate that there are numerous barriers to unambiguously linking joint species-distribution models and predictive models of species interactions. These obstacles are pervasive across many common ecological contexts, they strongly impact analysis capabilities, and perhaps most importantly they hinder the biological utility of many parameters that emerge from statistical inference. Chief among these include: the information loss that occurs when modelling interaction presence-absence rather than interaction frequency; the limits created by lacking intraspecific variation in species trait values; the ambiguities that arise without independent estimates of background species densities; and the complexity of linking statistical interaction models to appropriate functional-response equivalents. Though some of these can be overcome with additional and improved data collection, others will require substantial reconsideration of how ecologists interrogate and confront statistical and mathematical models with data.