The Ecological Forecasting Initiative (EFI) is organized around six cross-cutting working groups that aim to develop a broad unification of environmental biology, from core cross-disciplinary theory and decision making through applications of quantitative tools and cyberinfrastructure, which will then be disseminated through education and knowledge transfer.
Is nature predictable? Looking broadly across ecological subdisciplines, are there common patterns to what limits the predictability of different types of problems? Answering many of the deep, overarching questions in ecology requires us to develop new theory about predictability itself. It also requires synthesis across a small but rapidly growing catalog of ecological forecasts. Finally, forecasting itself is inherently synthetic, as making specific, quantitative predictions requires the fusion of our prior data and our current theories (as embedded in the models we use). Prediction provides a means of unification by assessing our ability to extrapolate not just within individual problems but across them; it implies we understand something general about how ecological system works. These are the big-picture problems the EFI Theory & Synthesis working group aims to tackle.
Accelerating the use of probabilistic ecological forecasts requires understanding both how they could be used to make decisions under uncertainty, and how forecasts are actually interpreted and used in practice. It requires understanding how best to communicate forecast uncertainty, as well as how to better learn from model failures. Formalizing expert judgment in key scientific decisions and inference processes are critical to increasing community transparency about decisions and accelerating the process of model learning, reducing bias, and increasing model skill. The Decision Science working group aims to explore these issues from the prospectives of both forecast producers and consumers.
Methods & Tools
The integration of data and models is at the core of ecological forecasting. There is much that can be learned about forecasting methods from other disciplines, from weather forecasters through to economic forecasters. But ecologists also face challenges that outside the mainstream of either of these extremes, such as a high degree of process heterogeneity across many scales and an abundance of semi-mechanistic models, where physical and chemical constraints play an important role but many functional relationships are empirically derived. This working group will advance the statistical methods and tools for forecasting and data assimilation by advancing statistical approaches and best practices for data assimilation, translating these into software tools usable by ecologists, and develop uncertainty estimates on common data.
Regularly rerunning forecastings using the newest data is a core aspect of iterative near-term forecasting. This sets a higher bar for repeatability and presents informatic and computational challenges that go beyond most ecological analyses, particularly for forecasts running closer to real-time. The goal of the cyberinfrastructure working group is to make it easier to implement, archive, and share automated iterative forecasts, so that any research group that can develop a forecasting model can deploy it as an automated system. Efforts of this working group include, but are not limited to, the development of standards and databases for transparent, open, and interoperable archiving and sharing or both forecasts and forecast workflows, and the development of shared community tools for data ingest/interoperability and for forecast workflow automation / continuous integration. We will make the components of our infrastructure available through open source software and open educational resources for using existing tools.
Education & Inclusion
Our focus is to build a diverse community of individuals who are 1) trained on the methods, theory and decision science to create iterative near-term forecasts, or 2) trained to use the forecast predictions in management and policy. This EFI working group will focus on building a community of practice among ecological forecasting educators and developing open, collaborative, and extensible teaching materials at the undergraduate, graduate, and professional levels. We think training in ecological forecasting is important for all ecologists, not just those actively building forecasts. For example, the approaches used to design experiments and collect data can change nontrivially if we want to inform predictions.
The production and use of ecological forecasts often relies on partnerships between many organizations. The Knowledge Transfer working group is focused on fostering collaborations and forecast co-production among academics, government agencies, industry, NGOs, and citizen scientists. We are also interested in promoting the dissemination of ecological forecasts to help society better understand, manage, and conserve ecosystems.