Introduction

Data science skills, such as visualizing, analyzing, and modeling large datasets, are increasingly needed by ecology and environmental science undergraduate students (Farrell and Carey 2018, Auker and Barthelmess 2020, Feng et al. 2020, Cooke et al. 2021). Recent advancements in environmental monitoring technology (e.g., Mcloughlin et al. 2019, Nathan et al. 2022, Dauphin et al. 2023) and the rise of environmental observatory networks (Keller et al. 2008, Weathers et al. 2013, Cleverly et al. 2019) have resulted in a deluge of “big data” in ecology (Hampton et al. 2013, LaDeau et al. 2017, Farley et al. 2018). As a result, analysis of large datasets is now required across a variety of environmental science and ecology careers, necessitating new approaches to training researchers, instructors, and students in data science skills (Hampton et al. 2017, National Academies of Sciences 2018, Feng et al. 2020, Emery et al. 2021).
Currently, a lack of both student and instructor familiarity with data science concepts, methods, and tools presents a major barrier to incorporation of data science into undergraduate life science curricula (Williams et al. 2019, Emery et al. 2021, Naithani et al. 2022, Cuddington et al. 2023). This gap often exists because instructors themselves have not received training in data science skills (Williams et al. 2019, Emery et al. 2021), and students do not have the requisite background skills and confidence to effectively engage in data science training (Williams et al. 2019, Cuddington et al. 2023). Consequently, development of educational materials approachable to both instructors and students is needed to lower the barrier to data science education in ecology and environmental science.
Ecological forecasting is an ideal topic for engaging instructors and students in data science training (Willson et al. 2023). First, ecological forecasting has the potential to guide environmental management decisions (Johnson et al. 2018, Liu et al. 2020, Bodner et al. 2021, Heilman et al. 2022), thereby engaging students in real-world problem-solving. Ecological forecasts, which provide predictions of the future state of ecosystems with uncertainty (Luo et al. 2011, Petchey et al. 2015), are critically needed to help manage natural resources increasingly threatened by climate and land use change (Bradford et al. 2020). Examples of societally important forecasts exist for many ecological systems, including river temperature forecasts to guide reservoir water release decisions and protect fish species (Ouellet-Proulx et al. 2017), temperature-based spring onset forecasts to inform agricultural decision-making (Carrillo et al. 2018), and forecasts of endangered ocean species to avoid bycatch (Hazen et al. 2018).
Second, generating ecological forecasts requires students to step through the scientific method (Moore et al. 2022, Lewis et al. 2023), providing critical skills in developing and testing hypotheses, which are transferable across scientific disciplines. In the iterative forecast cycle, similar to the scientific method, researchers develop hypotheses about how ecosystems function; instantiate hypotheses into a predictive model; use the model to generate forecasts into the future; evaluate forecasts with observations once the future arrives and new data are available; and use evaluation results to iteratively update and improve hypotheses, models, and predictions (Dietze et al. 2018).
Third, ecological forecasting problems are particularly well-suited for active learning, in which students learn by doing rather than passively listening or watching (Bonwell and Eison 1991). Active learning has been shown to enhance student outcomes (Freeman et al. 2014), especially for underrepresented groups (Theobald et al. 2020). Key components of active learning that can be easily embedded within ecological forecasting curricula include: authentic assessments that engage students in real-world, relevant problems similar to what they will encounter in their future careers (Villarroel et al. 2018); scaffolding to help students progressively build more complex skills and problem-solve (Belland 2014); and formative assessments that provide students with specific, actionable guidance on their progress, with opportunities to apply that guidance moving forward (Wiliam 2011).
To effectively use ecological forecasting as a platform for teaching data science in undergraduate classrooms, instructors must have both pedagogical knowledge of active learning and disciplinary knowledge of data science and ecological forecasting (Auerbach and Andrews 2018, Andrews et al. 2019). However, research has demonstrated substantial gaps in instructor knowledge in both active learning (Auerbach and Andrews 2018, Andrews et al. 2019) and data science (Williams et al. 2019, Emery et al. 2021). Given that ecological forecasting is an emerging field (Lewis et al. 2022), and educational resources in ecological forecasting remain rare (Willson et al. 2023), it is unlikely that many instructors have training in this area.
To address gaps in instructor knowledge in the life sciences, multiple models of instructor professional development have been trialed, including short, intensive trainings for teaching assistants (Hughes and Ellefson 2013, Schussler et al. 2015), department-wide training programs for faculty (Owens et al. 2018), and multi-year, multi-institutional programs for postdoctoral researchers (Ebert-May et al. 2011, D’Avanzo et al. 2012, Derting et al. 2016). Outcomes of these professional development activities frequently rely solely on instructor feedback (Ebert-May et al. 2011). However, instructor and student perceptions of the effectiveness of teaching practices in the classroom can differ from each other (Heim and Holt 2018). Consequently, the effectiveness of instructor professional development should be evaluated using multiple methods (e.g., reflection and feedback, observing teaching practices, student assessments; Ebert-May et al. 2011, Heim and Holt 2018) and incorporate input from both students and faculty. Moreover, as lack of time is often cited as a barrier to instructor professional development (Williams et al. 2019), instructional materials associated with active learning activities should include short, accessible definitions and examples of key concepts to provide “just-in-time” (sensu Novak et al. 1999) pedagogical, data science, and ecological forecasting training for instructors as well as students.
To lower the barrier of entry to data science for both undergraduate students and instructors in ecology and environmental science, we developed and assessed a modular curriculum within the Macrosystems EDDIE (Environmental Data-Driven Inquiry and Exploration) program (Carey et al. 2020, Hounshell et al. 2021, Moore et al. 2022, Woelmer et al. 2023a) that uses active learning techniques to teach data science skills in the context of ecological forecasting. While previous educational materials on ecological forecasting have been developed for advanced students, primarily at the graduate level (e.g., Dietze 2017a, Ernest et al. 2023), our curriculum is one of the first that is specifically targeted to undergraduates (Willson et al. 2023). In addition, all materials are designed to be approachable to both instructors and students, as coding experience is not a necessary prerequisite and each module is accompanied by substantial introductory and supporting materials for instructors. Moreover, students work with data from the U.S.’s National Ecological Observatory Network (NEON) to address relevant societal challenges such as predicting freshwater quality impairment.
Here, we present an overview of the Macrosystems EDDIE ecological forecasting curriculum and examples of how it has been implemented in various course contexts. We also analyze student and instructor assessment data to address the following questions: (1) How does completion of Macrosystems EDDIE ecological forecasting modules affect student confidence and understanding of data science and ecological forecasting skills? (2) What are instructor perceptions of module ease of use and efficacy in teaching data science and ecological forecasting concepts? We were specifically focused on student confidence and instructor perceptions, as previous work has shown that two major barriers to integrating data science active learning activities into existing curricula are a lack of instructor training (Williams et al. 2019, Emery et al. 2021) and student confidence (Williams et al. 2019, Cuddington et al. 2023).