Below is a guest post from Jacob Meisel, a rising junior at Harvard College. He has over seven years of experience analyzing complex weather models, and has over a year of direct market trading and analysis within the natural gas futures market. He serves as the Chief Meteorologist over at Bespoke Weather Services, where he plans to release subscription options for natural gas and energy traders. These products will be available for purchase beginning next week at bespokeweather.com; they contain both weekly weather and natural gas market fundamental and technical analysis along with daily trade and weather model updates. Jacob will also occasionally write for other Bespoke products, updating us on the latest happenings within the energy markets. For any additional questions on future products or services from Bespoke Weather Services, please email Jacob at [email protected]
Millions of New Yorkers woke up this past January 27th and looked outside for the ‘feet’ of snow predicted to fall over the prior night. Instead, they saw only a few inches. The major blizzard that everyone was predicting shifted out just east enough to spare the city from most of the impacts. The forecast bust, when later dissected, occurred due to too heavy a reliance on a certain weather model: the European ECMWF. Known as one of the most accurate weather models in the world, this model consistently showed an all-out blizzard for New York City. In the end, though, the American weather guidance proved correct, shifting the storm just enough further east.
This past week, we were shown exactly why this forecast bust occurred in one of the seemingly most unrelated places: the natural gas futures market. Last week, prices closed at their lowest weekly point since April, below $2.59, as a mixed suite of weather models showed a pattern fairly close to average. The ECMWF European guidance was the warmest, and the GFS American guidance was the coolest. Many looked to the GFS American guidance and saw a lack of cooling demand for much of the US into the end of June, using that as partial justification for the major market decline in the previous two weeks.
Yet over the weekend, the model rapidly transformed. The ECMWF European guidance remained consistent in its warmth, while it was the American GFS model trended almost perfectly in line with it through the end of June, showing significant warmth for much of the Eastern half of the country. The natural gas market proceeded to gap up Sunday evening from the Friday close, logging sizable gains Monday, Tuesday, and Wednesday (circled).
This rapid change in the American weather guidance is part of the reason that meteorologists back in January were so ready to dismiss its solution in favor of the European solution; when the model is wrong, it is very wrong. At Bespoke Weather Services, one major indicator we use to forecast and interpret global weather patterns is the Madden-Julian Oscillation (MJO). Its 8 phases indicate, among other things, where in the Pacific Ocean convection (or storminess) is most active. To the right, we show just how inaccurate the American model suite has been at forecasting this over the past week. The green line indicates the ensemble forecast (a blend of all runs of the American guidance) while the blue line was the most recent GFS run of the MJO guidance at the time. The red line (circled) is what actually ended up happening. Instead of diminishing in influence by moving towards the center circle, the MJO continued to move into phase 3 and looks headed to phase 4, which is one of the warmer phases for the US.
Shown to the lower right is the ECMWF European forecast from Monday (green line). Though not perfect, this clearly had a superior take on how the global pattern would materialize; the successful MJO forecast verification being a microcosm for the superior forecasts overall the ECMWF model has printed out recently. Once the natural gas market realized that the warmer ECMWF scenario was accurate, prices spiked to account for the increased cooling demand now expected later in the month. Over at BWS, we alerted the public that these trends were occurring at the end of the day Friday and then again at the market open Sunday via Twitter and our updates at the end of the day on Friday and then again at market open Sunday. It didn’t take long for prices to take off from there.
So where do we go from here? Prices have given back some gains yesterday and today as the American guidance has trended back significantly cooler, and Canadian and European guidance tipped slightly cooler as well. However, the European ECMWF guidance remains the warmest, and still appears to be forecasting the movement of the MJO the best. Until evidence is presented otherwise, or model output shifts, it may be best to side with that warmer European guidance, something the market may not be doing adequately. Though occasionally that can result in a January-esque forecasting miss, more often that not the guidance will nail the forecast on the head.