8.1 Transforming Risk Communication with Probability Forecasts for Energy—Weeks to a Century or More

Wednesday, 10 January 2018: 1:30 PM
Room 15 (ACC) (Austin, Texas)
John A. Dutton, ClimBiz Ltd and Prescient Weather Ltd., State College, PA; and R. P. James and J. D. Ross

The power of probability forecasts to transform communication and management of climate risk or opportunity in the energy industry is illustrated with two examples. The first demonstrates how to foresee the statistical consequences of hedging or taking other action to counteract predicted adverse subseasonal or seasonal (S2S) conditions, thus enabling managers to act on forecast with confidence in the expected return and volatility. The second illustrates how to combine international supercomputer climate change simulations with a mathematical model of an electric utility to illustrate alternatives for ensuring financial resilience in the decades ahead as dependence on renewable generation increases and customer demand responds to temperature trends.

Seasonal and Subseasonal Forecasts

The World Climate Service (WCS), a collaborative effort of Prescient Weather and MeteoGroup, provides a calibrated and optimized multi-model combination of the S2S ensemble forecasts of the U.S. National Weather Service (NWS) and the European Centre for Medium-Range Weather Forecasts (ECMWF). The WCS computes and publishes verification and skill information for both retrospective and recent forecasts. With NOAA small business support, WCS developed and is offering S2S probability forecasts for degree days, wind, and solar insolation, in addition to temperature and precipitation.

A compact financial model combines with forecast skill information to estimate the expected return and volatility of hedges or other action taken to mitigate predicted adverse S2S conditions. Three important skill functions are the reliability, which demonstrates whether the forecast calibration is successful, the fraction of adverse forecasts, and the fraction of correct forecasts—all three dependent on the predicted probability.

The critical user question is: What will be the consequences if I act or hedge when predicted probabilities for adverse conditions equal or exceed p? A WCS interactive action and hedge advisor responds to the question by showing how the return and volatility depend on predicted probability and hedge parameters. The results demonstrate that the WCS forecasts are sufficiently skillful to confer advantage to users in hedge transactions at conventional prices.

Climate Change Scenarios

Energy utilities and other climate sensitive activities can attempt to ensure resilience by examining the implications of various climate change scenarios for long-range strategies and capital investment plans. The supercomputer climate change simulations prepared for the 2013 report of the Intergovernmental Panel on Climate Change (IPCC) and archived as part of the Climate Model Intercomparison Project 5 (CMIP5) provide a very large dataset with which to construct scenarios. User-friendly, interactive access to this dataset is provided by a Climate Change Information System for Business and Industry (ClimBiz) being developed by Prescient Weather with Department of Energy small business support.

Illustrating the process that could be used by utilities and other firms, a mathematical model of an electric utility allows for evolution of customer demand in response to temperature trends and for change in the generation fraction and cost of both renewable and fossil sources of energy, with an emphasis on increasing fractions of solar and hydro renewables as the 21st century proceeds.

ClimBiz users can derive time-dependent climate change probability distributions for generation and demand variables from CMIP5 simulations for scenarios depicting mild to severe climate change and use them to create a probabilistic view of generation options and costs. For example, simulations with five fractions each of hydro and solar generation for five 20-year periods in the 21st century will give a total of 125 cases, each to be simulated 5000 times to obtain stable statistics. The results provide options of relatively low generation cost at high volatility or low volatility at moderate cost.

Implications for Transforming Communication and Risk Management

These examples illuminate the critical components of effective risk management with long-range probability forecasts. First, we must have reliable probability forecasts or realistic scenarios for some future time and then we must have a financial or business model that reveals the consequences of various actions and events quantitatively in user terms. For S2S time scales, the forecast skill functions provide probabilities for events in the financial model and thus give a statistical prediction about the consequences of action. For climate change scenarios, using model anomalies relative to a first simulated decade, say, reduces scatter, increases confidence, and allows results to be scaled to present conditions.

In both cases, the plots of financial return or cost versus the volatility take a hyperbolic shape reminiscent of the efficient frontier of portfolio theory. The minimum volatility is midrange to the financial return or cost.

The key to transformational communication and risk management is using the forecast probabilities to describe the probabilities and statistics of the outcomes about which the user is concerned. We must tell the user what the quantitative consequences of acting on the forecasts or scenarios will be.


Prescient Weather Ltd research reported here was supported by the National Oceanic and Atmospheric Agency with Contracts WC133R-11-CN-0147 and WC-133R-16-CN-0103 and by the Department of Energy with award DE-SC0011284.

This talk draws on some concepts and results presented previously, including:

Dutton, John A., Richard P. James, Jeremy D. Ross, 2017. Probability Forecasts for Energy–Weeks to a Century or More, in Weather and Climate Services for the Energy Industry, in press as a Palgrave Pivot book.

Dutton, John A., R. James, J. Ross, P. Knight, K. Mitchell, R. S. Stouffer, 2017. Climate Change Strategies for Electric Utilities, AMS Annual Meeting, 26 January 2017, https://ams.confex.com/ams/97Annual/webprogram/Paper310120.html

Dutton, John A., Richard P. James, Jeremy D. Ross, 2015. Bridging the Gap Between Subseasonal and Seasonal Forecasts and Decisions to Act, AMS Annual Meeting, 7 January 2015, https://ams.confe12.com/ams/95Annual/webprogram/Paper260171.html

Dutton, John A., Richard P. James, Jeremy D. Ross, 2014. A Probabilistic View of Weather, Climate, and the Energy Industry, Weather Matters for Energy, Alberto Troccoli et al., eds., 353-378, Springer.

Dutton, John A., Richard P. James, Jeremy D. Ross, 2013. Calibration and Combination of Dynamical Seasonal Forecasts to Enhance the Value of Predicted Probabilities for Managing Risk, Climate Dynamics 40, 3089-3105.

James, Richard P., Jeremy D. Ross, John A. Dutton, 2014. Skill of a New Two- to -Six Week Forecast System. AMS Annual Meeting, 4 February 2014, https://ams.confe12.com/ams/94Annual/webprogram/Paper233913.html

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner