Instructor Notes

Why ecological forecasting?

Why do we need ecological forecasting?

• Manage environmental change
• Test how well we understand ecosystems

What are some of things we need to be able to make forecasts?

• Models
• Data

Forecastability?

What influences our ability to forecast and how?

• How well the factors driving the response are understood
• How much data is available
• Whether the forecast influences the thing being forecast
  • Are there examples in ecology where a forecast could influence the thing being forecast?

Forecasting, planning, and goals

What kinds of time-scales of forecasts do we want/need in ecology?

How might these different “forecasting horizons” influence our approaches to forecasting?

Determining what to forecast

How should we think about what to forecast in ecology?

• Talk to those who need or are impacted by the forecast
• Consider whether we have the data/models we need

Terminology

What is forecasting?

• Forecasting - probabalistic statement about the future using data assimilation.

The Luo et al. paper distinguishes between forecasting, prediction, projection & prognosis. What are the core differences?

• Prediction - quantitative estimate of an unknown state. Can be about the future. Not necessarily accurate.
• Projection - Possible future state resulting from a choice of future scenarios. Typically quantitative.
• Prognosis - more subjective.
• Note that these definitions vary across and even within disciplines
• Definition of forecasting is much more narrow than most

Are these terms consistently used the two readings? In your experience more broadly?

• Not really. Different fields, disciplines, and authors use them in different ways.

Uncertainty

What is the distinction between point estimates and uncertainty?

• Point estimates are the expected value or mean prediction
• Uncertainty is the variability around that mean value, that describes the reasonable range of predictions.

Do most ecological forecasts include uncertainty estimates?

• No

Why does estimating uncertainty matter?

• Need to know how much to trust forecasts
• What the realistic range of outcomes could be

Forecasting data and methods

What are the general types of models that can be used for forecasting and how do they differ?

• Data-driven vs. Process modeling
  • What is a good example of cross-sectional forecasting in ecology?
  • What is a good example of time-series forecasting?
• Within data-driven “Time series” vs. “Explanatory”

What are the different ways in which model parameters are determined?

• Fixed
• Fitted
• How does this relate to process vs data-driven modeling?

When might these different approaches be most appropriate in ecology?

What are the different ways in which time can be incorporated into these models?

• Explicitly
• Implicitly (driving variables that change)
• Not included

Basic steps

How do the “basic steps” of forecasting relate to ecological forecasting?

• How well do you think our field does at each of these steps?
• What can we do to improve on our current approaches?

Where do you think we are?

• How good do you think we currently are at forecasting in ecology?
• What does the history of weather forecasting tell you about how that might change?