by Judith Curry
In Atlanta right now, we have about 2″ of snow, overlain by freezing rain, a classical “winter mix.” The whole city is pretty much closed down (including Georgia Tech). This is a fairly wimpy storm relative to what I used to encounter in Boulder or Chicago, but dealing with weather is relative to what you have adapted to. Madhav Khandekar from India posts at Pielke Sr.:
For last two weeks or about most of north and central India are witnessing cold wintry weather; some places in Kashmir and the Himalayan foothills have low temperatures at -5C to -20C! This is cold for India, since most houses are not insulated, not heated (except some small room heaters in north India) . . .
So what’s going on with the weather? Do holiday blizzards provide more signs of global warming, as per this article in Time? Or is this just “weather roulette,” what happened to come up this year in the chaotic coupled ocean atmosphere system? Well regardless of what greenhouse warming is doing to our climate or whether we have recently seen a climate shift, there are roulette-like elements to how the weather systems actually play out in a given season and location.
How predictable is winter weather? Based upon my experience of providing weather forecasts for a client in the energy sector using the ECMWF forecast system (which I regard as the best forecast system currently available), there is pretty good predictability out to one week for specific weather systems. There is also some longer term predictability in the system associated with ENSO, the AO/NAO, PNA, and MJO.
This winter we have been in La Nina conditions, spiced by predominantly negative AO/NAO (which reached extreme negative values in mid December) and a PNA index that was negative from mid Nov to mid Dec, and is now hovering around neutral. The combination of these indices have a strong control on the winter weather patterns in the northern hemisphere.
Seasonal predictability of winter weather
The basis for predictability on subseasonal (weeks) to seasonal (months) time scales is the ability to predict certain teleconnection modes that influence regional and global circulation patterns and hence storms and regional variations in surface meteorological conditions. The particular teleconnection modes of greatest relevance on these timescales are the Madden Julian Oscillation (MJO), El Niño Southern Oscillation (ENSO), and the North Atlantic Oscillation (NAO)/Arctic Oscillation (AO). While these are viewed as independent teleconnection modes, there are relationships between these modes, whereby the MJO influences ENSO and ENSO modulates the MJO, with resulting storm track shift and NAO/AO response.
Predictability on these time scales of surface weather and extreme events is greatest for regions that are influenced by the MJO, ENSO, PNA, and the NAO/AO. Further, predictability in a particular region varies seasonally. The summertime MJO has a dominant control on the active and break period of the Asian monsoon, and also hurricane activity in the Pacific, Indian Ocean, and North Atlantic. Winter weather in the northern hemisphere is dominated by the NAO/AO and PNA patterns.
So the bottom line is that starting around Dec 1, you can get a pretty good picture of what the winter weather will look like for the next two months. So why have there been such spectacular failures in some recent seasonal forecasts? In particular the UK Met Office has been taking alot of heat over its prediction in 2009 of a “barbecue summer” and this year for not making public its forecast of a cold winter. There are two parts to the forecasting issue: how good the models are, and how to extract useful information from ensembles.
In terms of how good the models are, again in my experience, the modeling system that is currently far superior is ECMWF. There is good news on the seasonal forecast model front. Over the course of the next several months, substantially new and improved seasonal forecast products will be released by NOAA CPC (CFSv2; available Jan 18) and ECMWF System-4 (available sometime in the next six months.) The NOAA CFSv2 represents a substantial change to all aspects of the forecast system, including all components of the model, the data assimilation system, and the ensemble configuration. The ECMWF System-4 includes changes to model parameterizations and the ocean model and an increase in resolution.
Statistical forecast methods
In addition to the global models, there are statistical forecast methods that have been developed for specific regions. Most recently, Judah Cohen has received a lot of publicity for his statistical forecast for winter weather over the U.S.
One of the challenges with statistical forecast methods is that the statistics are non stationary. In particular, in the face of a climate shift (which has arguably occurred sometime over the past decade), the statistical relationships will undoubtedly be different. The various teleconnection indices (e.g. ENSO, NAO/AO) have multidecadal variations associated with the PDO and AMO. The analogue for our current situation (warm NAO, cool PDO) was last seen in the 1950’s, where we might find some analogues to our current weather patterns.