Toxic algae closures: What kinds of predictions are needed?

When I was growing up we had weather forecasts, and it was great. You could either plan your day or just complain about how wrong the weather forecast was (or a bit of both). Now, there are forecasts for almost everything one could monitor, from pollen forecasts to election forecasts to covid forecasts. There is even something called “predictive policing”–a forecast for potential criminal activity before it happens. Artificial intelligence, networked data, and rapidly advancing science has made it tempting to forecast many things and to deliver the predictions on nice websites and “data dashboards”. Harmful and toxic algae are no exception, and as aquaculture becomes more and more essential to feeding the world, forecasts for harmful blooms are poised to have more and more utility.

It’s easy to see the benefits of new forecasting technologies. Harmful algal bloom (HAB) forecasts can help get ahead of possible effects on human health, wildlife, and water quality. But like with any new technology, it’s important to be thinking about potential unintended consequences. The predictive policing example that I mentioned has been shown to be rife with bias and injustices that are embedded in the algorithms. These biases are sometimes unintentional, merely an accidental reflection (or magnification) of biases that exist in the data, model, or assumptions–but with important real-world consequences (Vera et al. 2019).

Environmental forecasters are catching up to other fields now in terms of understanding unintended consequences of data technologies (Grasso et al. 2019). Other ocean forecasting programs have been caught off guard by how their forecasts have been used or interpreted, even when the forecasts were getting it right. For example, forecasts for protected species have been used by poachers; forecasts for fisheries have jolted markets in unexpected ways; and in some cases, forecasts can even reverse the outcome that they were trying to predict (Pershing et al. 2018, Hobday et al. 2019, Record et al. 2021).

As HAB forecasts start to proliferate, it’s a good time to get ahead of potential pitfalls. At the 2021 International Conference on Harmful Algae meeting, an international group of experts presented draft guidance documentation for developing early warning systems and forecasts to meet stakeholder needs. This effort was supported by the Intergovernmental Oceanographic Commission (IOC) Intergovernmental Panel on Harmful Algal Blooms (along with other United Nations agencies, FAO, IAEA), and it will be used to guide the development of HAB forecasts in places around the world. The documentation outlines the wide range of approaches, from monitoring to machine learning technologies, for predicting the staggering variety of harmful algae and toxins that occur. Among the types of guidance provided, there is a strong emphasis on the involvement of stakeholders and users throughout the development process.

To give a bit of background, there are already quite a few HAB early warning systems currently in use. There is a concentration of such programs in European waters, but they’re used in other parts of the world as well, primarily in wealthier countries. Traditionally, warning systems have been designed based on an ecological forecasting paradigm found in the scientific literature. The approach is typically presented as a cycle that aligns roughly with the scientific method–it begins with hypothesis and model building, along with data analysis, and only includes stakeholders at a later stage when the forecast is being communicated, or in a separate process altogether (Dietze et al. 2018). Using a cycle like this has many benefits, but one drawback is the potential for a mismatch between the forecast design and the stakeholder needs. Indeed, this mismatch is one of the factors that has led to the unintended consequences mentioned earlier.

To help address this problem, the FAO-IAEA-IOC guidance documentation walks through a deliberate process of including stakeholders from the beginning of the forecast development. Rather than starting with a scientific hypothesis or some data mining, the process begins with a series of conversations that include stakeholders–that is, the forecast users. The idea is that having stakeholders engaged from the early design stages can help guide the science so that it benefits the people who need it. Of course, there is always a risk of unintended consequences with the development of any technology, but close collaborations with stakeholders–in this case, members of the aquaculture community–can hopefully reduce that risk.

In the spirit of open collaboration with stakeholders, we have launched a new initiative, the Tandy Center for Ocean Forecasting, at Bigelow Laboratory in Maine. One of our tenets is that the creation of ocean forecasts, like those used by aquaculture, should be as collaborative as possible. Our first HAB project has been an early warning system for paralytic shellfish toxins in coastal Maine. The project has been a collaboration between growers, the Maine Department of Marine Resources, and Bigelow Analytical Services. The trial system is in its second year and can be viewed at the Maine Department of Marine Resources data site (https://mainedmr.shinyapps.io/bph_phyto/).

The toxicity prediction gives a one-week advance notice of the probability of a closure-level measurement for each of a collection of sampling sites all along the coast of Maine. During its first year, the system correctly predicted the times and locations of the two closure events that occured. We recognize, though, that the system is far from perfect, there are still lots of improvements that can be made, and one of the best guides for improvements are the people who could be using the forecast. That’s where you could help us. We are encouraging people to provide us feedback on this effort, or on their hopes for future forecasting programs, including questions, requests, and concerns. Feedback can be sent to our email address at forecast@bigelow.org. Send along an email and join the conversation.

In the long term, we hope to help bridge the gap between science and stakeholders through a co-development process (Norström et al. 2020). In the end, we will all probably still complain about weather (and other) forecasts, but hopefully we can do so with equitable access to the benefits of those forecasting technologies.

This post was written along with Johnathan Evanilla for the National Shellfisheries Association Quarterly Newsletter http://dx.doi.org/10.13140/RG.2.2.20122.52165

References

Dietze, M.C., Fox, A., Beck-Johnson, L.M., Betancourt, J.L., Hooten, M.B., Jarnevich, C.S., Keitt, T.H., Kenney, M.A., Laney, C.M., Larsen, L.G. and Loescher, H.W., 2018. Iterative near-term ecological forecasting: Needs, opportunities, and challenges. Proceedings of the National Academy of Sciences, 115(7), pp.1424-1432.

Grasso, I., Russell, D., Matthews, A., Matthews, J. and Record, N.R., 2020, October. Applying Algorithmic Accountability Frameworks with Domain-specific Codes of Ethics: A Case Study in Ecosystem Forecasting for Shellfish Toxicity in the Gulf of Maine. In Proceedings of the 2020 ACM-IMS on Foundations of Data Science Conference (pp. 83-91).

Hobday, A.J., Hartog, J.R., Manderson, J.P., Mills, K.E., Oliver, M.J., Pershing, A.J. and Siedlecki, S., 2019. Ethical considerations and unanticipated consequences associated with ecological forecasting for marine resources. ICES Journal of Marine Science, 76(5), pp.1244-1256.

Norström, A.V., Cvitanovic, C., Löf, M.F., West, S., Wyborn, C., Balvanera, P., Bednarek, A.T., Bennett, E.M., Biggs, R., de Bremond, A. and Campbell, B.M., 2020. Principles for knowledge co-production in sustainability research. Nature sustainability, 3(3), pp.182-190.

Pershing, A.J., Mills, K.E., Dayton, A.M., Franklin, B.S. and Kennedy, B.T., 2018. Evidence for adaptation from the 2016 marine heatwave in the Northwest Atlantic Ocean. Oceanography, 31(2), pp.152-161.

Record, N.R. and Pershing, A.J., 2021, December. Facing the Forecaster’s Dilemma: Reflexivity in Ocean System Forecasting. In Oceans (Vol. 2, No. 4, pp. 738-751). Multidisciplinary Digital Publishing Institute.

Vera, L.A., Walker, D., Murphy, M., Mansfield, B., Siad, L.M., Ogden, J. and EDGI, 2019. When data justice and environmental justice meet: formulating a response to extractive logic through environmental data justice. Information, Communication & Society, 22(7), pp.1012-1028.