This tutorial introduces participants to key concepts in ecological + forecasting and provides hands-on materials for submitting forecasts + to the National Ecological Observatory Network (NEON) Forecasting + Challenge (hereafter, Challenge), hosted by the Ecological Forecasting + Initiative Research Coordination Network. The tutorial has been + developed and used with >300 participants and provides the + ecological understanding, workflows, and tools to enable ecologists + with minimal forecasting experience to participate in the Challenge + via a hands-on R-based tutorial. This tutorial introduces participants + to a near-term, iterative forecasting workflow that includes obtaining + observations from NEON, developing a simple forecasting model, + generating a forecast, and submitting the forecast to the Challenge, + as well as evaluating forecast performance once new observations + become available. The overarching aim of this tutorial is to lower the + barrier to ecological forecasting and empower participants to develop + their own ecological forecasts.
+Ecological forecasting is an emerging field that aims to improve
+ natural resource management and ecological understanding by providing
+ future predictions of the state of ecosystems
+ (
The NEON Forecasting Challenge aims to create a community of
+ practice that builds capacity for ecological forecasting by leveraging
+ recently-available NEON data
+ (
This 90-minute tutorial was initially developed for a workshop at + the 2022 Global Lakes Ecological Observatory Network (GLEON) + All-Hands’ conference, which had participants with a range of + forecasting and coding experience. The GLEON workshop was given on the + first day of the five-day conference. This timing enabled forecasts + submitted at the workshop to be evaluated throughout the conference, + allowing participants to see near-real time forecast performance, and + for a “winner” to be declared on the final day. The tutorial has since + been taught in nine workshop/classroom settings (Table 1).
+Table 1 Implementation of the tutorial by the authors across a + range of settings. The participants in these workshops covered a + wide range of forecasting and coding experience.
+| GLEON* All-Hands conference | +50 | +In-person | +Aquatics | +
| InventWater PhD training programme | +15 | +Synchronous on-line | +Aquatics | +
| GLEON All-Hands’ Virtual conference | +70 | +Asynchronous on-line (pre-recorded) | +Aquatics | +
| AEMON-J**/DSOS*** Hacking Limnology | +70 | +Synchronous on-line | +Aquatics | +
Global Change Ecology Lab, +University of Edinburgh |
+ 10 | +Synchronous on-line | +Terrestrial | +
| NEON Technical Working Group on Ecological + Forecasting | +10 | +Synchronous on-line | +Terrestrial | +
Ecological Society of America +conference |
+ 50 | +In-person | +Terrestrial | +
| Graduate environmental data science + class | +40 | +In-person | +Terrestrial/Aquatics | +
| Ecological Forecasting Initiative + conference | +20 | +In-person | +Aquatics | +
*Global Lake Ecological Observatory Network; **Aquatic Ecosystem MOdeling Network - Junior; ***Data Science Open Science
+The audience for this tutorial includes individuals who: 1) want to + participate in the Challenge but are not sure how to start; 2) want a + ‘hands-on’ way to learn about ecological forecasting ; and/or 3) are + involved in the broader forecasting enterprise (e.g., researchers + collecting data used for forecasting) and want to submit forecasts + themselves. We encourage users to modify the tutorial as needed, as + all materials are open-source.
+The overarching objectives of the tutorial are:
+-
+
Build an understanding of foundational ecological forecasting + concepts;
+Apply forecasting concepts to submit a simple forecast to the + Challenge; and
+Learn about additional forecasting resources.
+These objectives can be adapted depending on the context of the + tutorial. If the participant/instructor’s goals are geared towards + understanding forecasting concepts then the emphasis of the + presentation and hands-on workshop can be modified accordingly.
+The R-based tutorial is in a public GitHub repository (Olsson et
+ al.
+ (
The tutorial is designed as a 90-minute, standalone session that
+ includes pre-tutorial materials, an introductory presentation (20-30
+ minutes), a guided demonstration of forecast code (30-40 minutes),
+ and discussion (20-30 minutes). If more time is available, the
+ tutorial has additional content that includes more advanced topics
+ (Figure 1). Details on each of these sections is detailed in the
+ workshop’s README.md at the workshop GitHub repository (Olsson et
+ al.
+ (
-
+
Introductory presentation: introduces the participants to + forecasting concepts, the Challenge, NEON data, and tools that + will be used in the R coding portion of the tutorial. This + presentation can be tailored to the audience based on their + familiarity with forecasting concepts and NEON data (Figure + 1).
+Coding walk through: participants walk through a pre-written
+ Rmarkdown forecast workflow script. This code is written
+ primarily using
Open time for discussion: the remaining time can be used for + multiple purposes (detailed in the R markdown and README) + depending on the interests of the participants (e.g. debug code + issues, modify forecast models or form teams to submit + additional forecasts to the Challenge).
+Optional extension: to extend the tutorial beyond 90 minutes
+ (Figure 1), we provide additional materials that show
+ participants how to automate forecast submission (directory
+
The primary tutorial focuses on the Aquatics theme (specifically, + water temperature) of the Challenge as an example of how to generate a + forecast, though the tools and workflows are applicable to all + Challenge themes. For example, we adapted the materials to the + Terrestrial theme (see Table 1) based on the interests of the + participants and have further modified the materials for other + forecasting challenges.
+Several lessons learned have emerged from earlier implementations + of the tutorial (Table 1). First, engagement in the Challenge + post-tutorial is best when there is an opportunity for follow-up + discussion, troubleshooting, and continuation of team collaboration + beyond 90 minutes. Second, we found that providing installation + instructions and preparatory material in advance promoted best + engagement during synchornous and in-person workshops. Third, the + introductory presentation can be adapted to meet the needs and + experience level of the participants (Table 1). Finally, the tutorial + requires a stable and relatively fast internet connection. We found + that slow internet speeds limited access to downloading weather + forecasts used to generate ecological forecasts. This issue can be + addressed by either having the rendered Rmarkdown document available + so that individuals can follow along even if connectivity becomes an + issue or providing access to remote computational environments.
+Figure 1 Potential workshop/course structures using this
+ tutorial. The original 90-minute workshop setup is shown as the
+ “regular tutorial” which can be expanded and modified according to
+ the time available, the anticipated skills and background of the
+ participants and the goals of participation. The alternate modes of
+ delivery were delivered from administering the tutorial to audiences
+ of mixed coding and forecasting experience shown in Table 1.
+
This tutorial was supported by the National Science Foundation + through grants 1926050, 1926388, 1933016, and 2209866. We thank the + initial design teams, contributors and participants in the EFI-RCN + Challenge, and the many tutorial participants for their enthusiasm, + interest, and feedback that helped us iteratively improve the + materials (like our forecasts!).
+