Professor of Earth & Environment Michael Dietze is the lead author of “Near-term Ecological Forecasting for Climate Change Action,” published in early November 2024, in Nature Climate Change. His work aims to increase ecological forecasting accuracy using ecoinformatics. His ongoing project, PEcAn (the Predictive Ecosystem Analyzer) has worked on the development of iterative data assimilation algorithms, similar to those used in numerical weather forecasting, and their application to terrestrial carbon cycle “reanalyses” and forecasts.
Arts X Sciences asked Dietze more about his research and PEcAn in a recent Q&A.
How and when were you introduced to ecological forecasting and climate related quantitative data analysis?
As an undergraduate I switched majors from Engineering to Ecology and was drawn to the quantitative side of the field from the get go. That said, there were a few key moments that stand out in my career that led me to the field of ecological forecasting. First was a graduate course on computational statistics (Maximum Likelihood and Bayes) that fundamentally changed how I thought about science from a very mechanistic perspective to a very probabilistic one, where probability distributions represent our imperfect understanding of the world. Next was the publication of the Science paper “Ecological Forecasts: An Emerging Imperative” in 2001, which was a seminal paper defining the idea of ecological forecasting and really sparked my interest in the field. Also critical was being introduced to the iterative data assimilation algorithms used in other fields, such as weather forecasting, which occurred for me at a 2007 NSF workshop and really sparked my imagination around the potential applications of these approaches to ecological systems.
What has kept you in this field? What questions drive you to continue research and teaching?
At the top of this list are the global climate and biodiversity crises. Like many ecologists, my research and teaching are motivated directly by the need to address these challenges. For me, ecological forecasting has a critical role to play in how we address these challenges and their impacts, and it is also the place where I feel I can best put my own particular skill set to best use. Beyond this, I’m also excited about the research questions in the individual topical areas where we’ve been developing forecasts, as well as the overarching questions in forecasting about understanding the patterns of predictability across ecological systems: How accurately, and how far into the future, we can we predict different aspects of nature; why are some ecosystems more predictable than others; and how do the patterns in predictability across systems relate to biological constraints (e.g., evolution, physiology, ecological interactions) and abiotic drivers.
Finally, the other thing that has kept me in the field is the community. In 2018 I helped launch, and have since directed, the Ecological Forecasting Initiative, a community of practice aimed at building and growing the forecasting research community. Since then we have engaged thousands of academic, agency,NGO, and industry scientists and end users through a broad mix of international chapters, conferences, workshops, working groups, journal articles, social media, webinars, YouTube videos, standards, and policy briefs. We have developed multiple forecast training opportunities including: a book, >25 short courses and/workshops, courses at twelve universities, and online materials. We have collaborated closely with professional societies, organizing >30 conference sessions, launching an Ecological Forecasting award and co-organizing events with the Ecological Society of America, and helping the American Meteorological Society and National Ecological Observatory Network (NEON) launch Ecological Forecasting working groups. We organized a series of federal roundtables to promote interagency coordination and helped USGS and NASA develop white papers on forecasting. Through our Research Coordination Network we launched an ongoing NEON Ecological Forecasting Challenge, through which >200 teams (including >12 university courses) submitted over 2.5 million individual forecasts.
How will an increased accuracy in predictions around ecological systems aid in the effort to mitigate climate change?
Fundamentally, all decisions are about what we think will happen in the future, either under the status quo or different decision alternatives. Forecasts represent our best scientific estimate of these possible different futures and their associated uncertainties. Most ecological forecasts are focused on how we adapt to climate impacts and mitigate the impacts of climate change on natural and managed systems. A subset are focused on nature-based climate solutions, and thus represent efforts to mitigate climate change itself. For example, my lab has done work with the Environmental Defense Fund around how forecasts of the terrestrial carbon cycle inform the voluntary carbon markets (e.g., through modified agricultural practices) and I’ve served as an independent third-party reviewer for a number of commercial climate mitigation projects under the Verra and Climate Action Reserve markets.
How does the PEcAn project under the Ecological Forecasting Lab use information about the carbon cycle to inform climate change efforts?
PEcAn (the Predictive Ecosystem Analyzer) is a long running community project focused on terrestrial ecosystem models and data, which I’ve led since its inception in 2009. In recent years our work in PEcAn has focused primarily on the development of iterative data assimilation algorithms, similar to those used in numerical weather forecasting, and their application to terrestrial carbon cycle “reanalyses” and forecasts. A reanalysis is essentially a forecast run in the past but constrained by multiple different types of data in order to generate out best harmonized estimates of a process (in this case full carbon cycle budgets). As part of NASA’s Carbon Monitoring System we’ve expanded our initial prototypes, which analysed individual forests, to all of North America constrained by multiple forms of remote sensing (lidar, microwave, and multiple different optical satellites). Other members of the PEcAn team at the Finnish Meteorological Institute have also used PEcAn as the basis for Finland’s cropland carbon accounting system, and we’re currently working with the California Air Resources Board to develop an analogous open-source system for state of California’s croplands and wetlands. We also just got a grant to develop similar tools for Department of Defense installations. In addition to informing carbon inventories at national and subnational levels, which are often part of mandated reporting requirements, these systems will increasingly provide forecasts under different management scenarios. On area the lab is particularly interested in is the idea of forecast-guided assisted recovery, whereby forecasts are used to aid and optimize restoration efforts after natural disasters, such as wildfires, hurricanes, and pest and pathogen outbreaks, all of which are becoming more frequent, severe, and widespread under climate change.
Interview by Kelly Broder (COM’27)