Big Data and Predictive Analytics Tools Project Next Month's Weather Risks Today
Click here to view this article in our digital reader
Written by John Plavan
In sectors such as retail and finance, big data analysis has proven transformative – changing how businesses approach everything from product inventory to fleet efficiency to actuarial strategies. Identifying trends and probabilities from patterns within massive volumes of data has led to powerful new techniques for risk reduction and decision making.
In the energy sector, big-data is becoming a game-changer in an unexpected way - by transforming how we forecast the risk of extreme weather events on the horizon. From gas and power futures to grid infrastructure planning, the implications of big-data for the energy sector are huge.
For years, gas and power traders and risk managers have relied on numerical weather-prediction methods from NOAA’s Climate Forecast System and the European Centre for Medium-Range Weather Forecasts (ECMWF) to judge probabilities of weather events for demand forecasting and hedging. While conventional forecasting tools have become significantly more accurate over the years for short and medium-range forecasts, accurate forecasting beyond a seven-day window has remained challenging.
Numerical weather simulation models tend to be volatile in the 5 to 15 day forecast window due to the limits of parameter initialization scope needed in computer simulation modeling. And it is challenging to fully analyze these effects efficiently in an operational setting. For traders to leverage these traditional forecasts to gain a market advantage or to identify mispricing has typically required time-consuming comparisons of historic events to current patterns.
That’s where big-data driven models are coming into play, enabling forecasters to quantify weather risk with greater accuracy over longer periods. Statistical data analysis methods blended with cutting edge new atmospheric science findings can compare numerous weather variable patterns, observationally quantified around the globe, as precursors to events that might happen 30-40 days in the future. For those investing in gas and power or planning a city grid, the ability to see farther down the road, with less effort, is enormously valuable. With big-data driven predictive analytics models, sophisticated algorithms do the heavy lifting and deliver statistical reliability.
The new analytics approach uses more than 60 years of data on global weather pattern relationships in conjunction with current observational data to objectively quantify the risk for future extreme temperature events - with lead times of up to 40 days.
The data is indeed big. Billions of calculations with hundreds of weather patterns compiled over 10,000 days of observations provide a quantitative probabilistic forecast that identifies the relationships between historical weather patterns, current observational data and long-range (subseasonal) extreme temperature events.
For energy producers and traders, this means daily access to the odds of extreme (i.e. market-relevant) heat waves, cold snaps and temperature events that are often missed by other forecasting methods. These additional data points, entirely independent of numerical model simulations and traditional weather forecasting routines, provide traders with a more comprehensive understanding of weather’s impacts on the natural gas and power markets.
Case in point? The winter of 2011-12. By October of 2011, much of the commercial meteorology world predicted that the winter of 2011-12 would be blisteringly cold. But big data analytical models were presenting a different picture, repeatedly indicating elevated “heat” scores in the Eastern U.S ahead of the short range forecasts.
This objective, algorithmic data was one of few indicators that the winter of 2011-12 in the Eastern U.S. featured elevated risk of warmth, not cold at lead times ahead of the market.
For energy analysts and energy brokers who took advantage of this advanced forecast, prior knowledge of the weather trend was money in the bank.
Advanced knowledge of increased risk for extreme temperature events before the market has factored these events into pricing creates an advantaged play opportunity. Traders leveraging big data driven technology are given uniquely derived information with which to identify mispriced markets ahead of the crowd.
The ability to forecast weather event risks beyond two weeks is something meteorologists and market analysts have long sought – and there’s benefit to be had both in profits and public good. By utilizing powerful new predictive analytics tools to project the weather risks, energy traders, producers and utilities can not only gain an edge in the market, but also increase efficiency; avoid unplanned downtime caused by unforeseen adverse weather and reduce resource costs.
John Plavan is the chief executive officer of EarthRisk Technologies, a San Diego based pioneer in research and analytics for projecting extreme weather phenomena and temperature fluctuation risk over long periods of time. www.earthrisktech.com
Read More in Energy Digital's March Issue