The many paths to ecological network inference: reconciling machine learning, empirical data, and ecological knowledge

点击翻译按钮获取中文摘要

Ecology is rapidly becoming a science tasked with delivering actionable predictions to a wide variety of stakeholders. In complex adaptive systems, like ecological networks, this is a difficult taksk, both for lack of data, and for lack of robust predictive movels. Being able to formulate these predictions requires to simultaneously increase our understanding of the dynamics of complex ecological systems, and to translate this understanding into predictive models that can deliver data-driven predictions. Technical and methodological innovations from the field of machine learning and artificial intelligence are promising in this regard, as far as their successes in other disciplines reveals. In this proposal, I suggest three research questions, all aiming at predicting the structure of ecological networks, that will result in a systematic exploration of the potential of machine learning approaches to formulate predictions on biodiversity. First, I will develop models to predict, and then forecast, the structure of food webs over space, and investigate to which environmental drivers this structure responds. Second, I will develop a suite of Essential Biodiversity Variables for the structure of ecological networks, thereby examining which aspects of network structure have the highest information content. Finally, I will assess the potential to reconcile data from current and past sampling techniques, to evaluate whether hindcasting can be supported by better data integration. All of these projects will result in the development of free and open source software that will be re-usable by other initiatives, in addition to increasing the amount of existing open data; finally, they provide HQPs with a superb opportunity to acquire marketable skills in data science, in addition to furthering their ecological expertise.