The Integration of Ecological Forecasts Into Environmental Policy

Date: July 14, 2021

Post by: Güray Hatipoğlu1, 2Kira Sullivan-Wiley, 3Jaime Ashander

1Middle East Technical University, Earth System Science, Ph.D. Candidate, 2University of Notre Dame, Biological Sciences, Postdoctoral Associate, 3Resources for the Future; SESYNC Affiliate Scholar in Socio-Environmental Systems Modeling

In this series, we are asking: How might ideas from the social sciences improve ecological forecasting? What new opportunities and questions does the emerging interdisciplinary field of ecological forecasting raise for the social sciences? Using social science to understand how and when ecological forecasts are included in environmental policy can improve the ways in which ecological forecasters frame, communicate, and adapt their work.

This installment engages with the relationship between ecological forecasts and environmental policy. The central aim of environmental regulations is to maintain critical services for local populations, including clean air, clean water, and biodiversity. The central aim of ecological forecasting is to predict future conditions under a particular set of circumstances, including policy and regulatory scenarios.

Because environmental regulation and ecological forecasting are both concerned with future environmental conditions, one might expect to find ecological forecasts included and/or integrated into environmental regulations. In this piece we ask, is this the case? 

We identified several operational forecasts that are either currently used or mandated for future use by national or international regulations and treaties. These include a decision support system in the Baltic Sea, a harmful algal bloom monitoring system in the Republic of Korea, and a system for pest forecasting in Latvia. In this blog post, we trace each system from inception to operation to give insight into how ecological forecasts can be integrated into legislation and action.

An increasing problem in the Baltic Sea is widespread eutrophication. This issue, and the associated reduced fish stocks resulted in the creation of HELCOM, the Helsinki Commission, a multilateral decision making body for all states having a coast on the Baltic Sea. At first, HELCOM relied on close monitoring and measurements taken by individual states to address eutrophication. Over time, policies and binding rules were developed by HELCOM, but with incomplete scientific understanding of the complicated ecosystem dynamics of the Baltic Sea. Ultimately, HELCOM realized the utility and significance of an international independent scientific organization, to maintain a database where each state would report its pollutant loads and other relevant information. Moreover, HELCOM instituted a requirement of modeling ecosystem dynamics to generate both environmentally sound and economically feasible plans. In 1999, it initiated MARE (Marine Research on Eutrophication) to construct a user-friendly decision support system, NEST, for the Baltic Sea that is based on, inter alia, ecological models. NEST’s outputs were considered to be the “best available scientific information” in the 2007 HELCOM Baltic Sea Action Plan. The Plan explicitly incorporates NEST outputs into decisions made to control eutrophication. The NEST is still operational and can be used by anyone freely (scroll to the bottom of this webpage to find details about the different versions of NEST and how to download the tool).

In addition to the direct use of forecasts by government decision makers, there are also instances of the government supporting ecological forecasts intended for the public, akin to daily weather forecasts. The National Institute of Fisheries Science of the Ministry of Oceans and Fisheries of the Republic of Korea actively monitors harmful algal blooms in its country’s coastal areas and issues forecast alarms regarding its spread. This system, Red Tide Control Room, developed after dramatic fish losses owing to the harmful algal blooms, with their forecasts widely broadcasted to everyone that might be affected.

Another example is Latvia’s Plant Protection Center’s (PPC) and its use of RIMPro forecasts to aid in responsible use of pesticides. RIMPro is a forecasting tool initially developed to forecast apple pests (though it has diversified since) and is intended to inform growers’ decisions on the use of pesticides on apple trees. This particular product was chosen by PPC after substantial research, a growing body of knowledge including from the local people, and comparisons with several other available forecasting tools. The Plant Protection Center has been working for more than a hundred years and has historically collaborated directly with governmental bodies on preparing legislation. The institute’s recently separated research arm (Agrihorst) conducts state-sanctioned monitoring and research on pest management and broadcasts RIMPro forecasts, normally a commercial product, on their websites freely for regions of Latvia.

In these examples, ecological forecasts of complicated ecological phenomena — MARE NEST for eutrophication in the Baltic Sea, the Republic of Korea’s Red Tide Control Room for harmful algal blooms, RIMPro for pests in Latvia — are incorporated directly and formally into environmental decisions. These examples indicate that the widespread use of ecological forecasts may depend not only on their predictive power but also on how they are integrated and recognized in formal decision processes. Experience and social science research suggest alignment of stakeholder goals and co-production with stakeholders are two necessary components to achieve this tight integration. Use of this and other knowledge from decision sciences or organizational sciences can help to develop forecasts that are fit for purpose.

Forecast Spotlights – USA National Phenology Network

January 27, 2021

In this installment of our ongoing “Forecast Spotlights” series, we highlight the USA National Phenology Network (USA NPN) whose mission is to collect, organize, and share phenological data and information to aid decision-making, scientific discovery, and a broader understanding of phenology from a diversity of perspectives.  Responses to the questions below were provided by Theresa Crimmins, USA NPN Director and Research Professor in the School of Natural Resources and the Environment at the University of Arizona. The USA NPN is also an EFI Partner Organization and is a partner on and participating in the design of the EFI RCN NEON Phenology Forecast Challenge which will start accepting forecasts on February 1, 2021.

1. How did you get interested in ecological forecasting?

The business of USA National Phenology Network is to document when seasonal events in plants and animals occur – both historically, in real-time, and in the future. Forecasts are a natural activity for our organization, as advance warning of when various seasonal events will occur is valuable in a wide range of applications including natural resource management, human health, tourism, and agriculture. 

2. What are you trying to forecast?

Our aim is to offer forecasts of management-relevant seasonal phenomena to support effectively timing management activities. So far, this includes forecasts of the start of springtime biological activity, key life cycle events in several insect pests, and green-up in the invasive plant, buffelgrass. In spring 2020, we launched a forecast of winter wheat development to support identification of potential damage following freeze events. In 2021, we plan to add forecasts of flowering and fruiting in additional invasive grasses to support management activities. We continue to engage with various manager and researcher communities to identify other important phenological transitions for which forecasts can improve efficacy of management actions. 

Lilacs, Emerald Ash Borer, and Buffelgrass are examples of organisms USA NPN forecasts.

3. Who are the potential users or stakeholders for the forecasts you create?

Forecast users include land managers, arborists and tree care specialists, invasive species eradication groups, and conservation practitioners as a part of planning and carrying out management activities. In addition, the forecasts – the start of spring forecast in particular – are heavily referenced by the news media as spring is progressing across the country.

4. What are the key lessons you have learned from your forecasts?

Predictions of when an event such as leaf-out or green-up will occur are limited both by the input data used in the forecasts and the models. We have experienced limited forecast performance as a result of both coarse input data and simple models. We hope to address both of these issues incrementally, experimenting with various sources of input data (gridded and station-based) and increasing the sophistication of our approaches. 

In addition, our current forecasts are short-term in nature. In most cases, they are available six days into the future at a specific location, as a result of the short-term availability of the forcing variables,  though some further insight can be gained by looking at phenological activity at lower latitudes or elevations, which generally experience activity earlier than higher latitudes and elevations. However, our stakeholders repeatedly have indicated that predictions of when events will occur on the order of several weeks or months in advance would be of greater value and more aligned with their planning windows. This is an active area of research for us, and we are excited to work with the ecological forecasting community to address this significant challenge.

5. What was the biggest or most unexpected challenge you faced while operationalizing your forecast?

Transforming the forecast output into maps that are clear and concise was a greater challenge than we had anticipated. We discovered the importance of a forecast map being able to stand alone, as our maps are frequently circulated in the news media and social media. We ended up spending much more time than we had planned on refining the map legends, titles, and explanatory text. We wanted to include sufficient detail that a sophisticated user could understand what was behind the forecast, but not so much that a casual user would be confused or alienated. We continue to invest surprising amounts of time in product interpretation, communication, and documentation. More details of this refinement process can be found in this People Refining Forecast Visualization EFI blog post.

Animated image for the forecast of the Emerald Ash Borer from February to August, 2018

6. Is there anything else you want to share about your forecast?

We are actively expanding our suite of forecasts on several fronts, and invite collaboration from the broader community. Our current activities include engaging with end-user communities to identify the most useful species and seasonal events to forecast as well as the best ways to deliver this information. We are also actively working to establish predictive models for phenomena where they do not yet exist. Finally, we are expanding our abilities to operationalize more sophisticated models, moving from simple thermal sum models to mechanistic models. We are excited to continue to grow in all of these directions to offer progressively more meaningful predictions of when seasonal events in key plant and animal species are likely to occur.

Forecast Spotlights – Elliott Hazen and Heather Welch

November 23, 2020

In this third installment of our ongoing series, “Forecast Spotlights”, we highlight the EcoCast nowcast and forecasts developed by Elliott Hazen, Heather Welch, and colleagues at the National Oceanic and Atmospheric Administration (NOAA). EcoCast is a fisheries sustainability tool that helps fishers and managers evaluate how to allocate fishing efforts to optimize sustainable harvest of target fish while minimizing bycatch of protected or threatened animals. 

The goal of the Forecast Spotlights blog series is to highlight operational forecasts being conducted by our EFI members, how they got into forecasting, and lessons learned. You can see all the ecological forecast project examples shared on the EFI Projects webpage. If you have an iterative ecological forecast project that you’d like added to this list, you can create a profile for the project using this form.

1. How did you get interested in ecological forecasting?

Elliott: A lot of my interest in ecological forecasting came from my graduate research at Duke University and learning about models forecasting seed dispersal. If such a random-seeming process such as seed dispersal could be modeled successfully it made me wonder what else could be modeled. I ended up using a lot of statistical models in my PhD to deal with the complexities of top predator datasets. In reading about these models, I realized that they could be used not just to understand ecological drivers of animal distribution but also for nowcasting and forecasting distributions moving forward. A paper by Drew Purves titled “Time to model all life on earth” also highlighted the fact that our computing capability has finally caught up to some of the questions that we have been trying to ask ecologically. 

Heather: During my masters at James Cook University I worked with Bob Pressey who was concerned that our global network of MPAs (marine protected areas) was designed to protect static representations of biodiversity, despite common knowledge that many species of management concern have dynamic distributions. We started thinking about how to design management strategies to explicitly accommodate this dynamism. This was a really interesting challenge. At the time there wasn’t much in the literature about how to manage highly mobile species and so there was a lot of room for creativity. It quickly became clear that, in order to manage species that move around, we need to know where species are in real time, or better yet, ahead of time which brought me to ecological forecasting. 

2. What are you trying to forecast?

We try to produce nowcasts and forecasts of the distributions of top predators and human activities to understand and mitigate their interactions, specifically interactions like fisheries bycatch and vessel collisions that put these species at risk. We, specifically Heather Welch in our group, have been predicting fishing behavior to try to identify where and why illegal fishing activities most likely to occur. The models that have predictive skill can be used to direct management and enforcement in the future.

Photo credit: Elliott Hazen

3. Who are the potential users or stakeholders for the forecasts you create?

We target our predictions for use by fishermen and fishery managers in EcoCast, the shipping industry and protected species managers for WhaleWatch, and most recently NOAA’s Office of Law Enforcement and the US Coast Guard for Illegal, Unregulated, and Unreported forecasting efforts. We also usually hope our predictions are interesting if not useful for the broader public.

Real-time predictions for top predators are integrated to produce a surface that indicates areas that are better to fish (blue), and poorer to fish (red) to improve fisheries sustainability. These maps are for the area off the west coast of the US.

4. What are the key lessons you have learned from your forecasts?

  1. We often learn as much from wrong forecasts as right forecasts as it tells us where physical processes may be driving our predictions differently than expected. These outliers can be really useful to understand ecological processes and patterning.
  2. Engaging with stakeholders from the get-go, even before the model is built, is really important to ensure that you’re producing forecasts that are as useful as possible.
  3. Often maintaining and testing ongoing forecasts are as much work if not more work than building the models (Welch et al. 2018). Creating accessible tools collaboratively with stakeholders can ensure the output is as applicable as possible.
  4. Also, Elliott remains interested in comparing the issues and successes in terrestrial vs. marine ecological forecasting systems. While some of the questions being asked are comparable, the processes often change at different scales because of the oceanic medium compared to land and air, which keeps him wondering how the fundamental processes of forecasting may vary across these systems.

5. What was the biggest or most unexpected challenge you faced while operationalizing your forecast?

The biggest challenges in our forecasting process has largely been finding funding and ongoing support for operationalization. Specifically, funding the development of the tools have been manageable but keeping tools working and ensuring predictions remain skillful has been incredibly difficult to fund. This need for ongoing maintenance, often termed research to operations, is a fundamental gap we often face in the forecasting process.

6. Is there anything else you want to share about your forecast?

We have also been moving more and more towards public code libraries to ensure that the lessons we have learned are available to help other forecasting projects get off the ground and also to remain operational. Reproducibility in science has changed as we’ve moved more from bench and field experiments towards modeling efforts. This field, often termed “data carpentry” is going to be growing more and more in the near future to ensure that our coding efforts are done in a publicly available and reproducible manner.

Photo credit: Elliott Hazen