Uncertainty exists in many things in life and the atmosphere is no exception. No specific weather event is "sure" or "certain" to happen. The final outcome of a weather event is one of an infinite number of possibilities.
A weather forecast can be made using deterministic models and it can also be made using probabilistic data.
First a brief discourse on weather "models". (These models are called "computer guidance". Guidance as such is just that: you can accept it,use it as you deem fit or you can ignore it all together.)
Computer models used in weather forecasting (also called numerical weather prediction or NWP) take an initial state of the atmosphere - observations from the ground and aloft and then ingest this data into powerful computers to predict future atmospheric conditions. These predictions are done using various equations of motion, thermodynamics and physics and calculated in a 3-dimensional array. The main deterministic or operational models used in forecasting are called the GFS, ECMWF (EC), UKMO (UK), GGEM and NAM. The last mentioned model and the first are models of the USA. GGEM comes from the Canadian Met Service, the EC from the European Center for Medium range Weather Forecasting and the UKMO comes from the United Kingdom (Great Britian) and the British Meteorological Office.
After these models finish all the calculations and come up with their expected forecast output of how the atmosphere will change over time, the initial conditions/observations are then used again in ensemble models. The GFS, ECMWF, GGEM and NAM have ensemble models run for them. But ensemble models don't use the SAME EXACT initial conditions, rather the initial conditions are slightly altered for each ensemble member.
By altering these conditions an ensemble can show another possible outcome to a forecast weather event. Again keep in mind what actually happens with respect to a weather event is ONE of an infinite number of possibilities, so by using ensembles an attempt is made to "forecast" other "possible" outcomes to an event. Thus if ensembles are "clustering" around one possible solution then there is a distinct possibility that MAY very well be what happens.
The more that the ensembles cluster around a solution (e.g., a storm may directly hit us, miss us or give us just a "glancing" shot) and if they are closer to the deterministic or operational model then a forecaster's confidence in a weather forecast/event is much HIGHER than if the ensembles show little clustering. When the latter occurs confidence is much lower than average. There are even cases where the ensembles may cluster around a completely different solution than what their deterministic counterpart has!
The thing to know about NWPs and ensembles is that they are sensitive to the initial conditions. Errors in the initial analysis are carried forward in time. The errors in the initial analysis increase as the data is "projected" farther out in time (e.g., days even weeks, forward). This why weather forecasts beyond 4 days can have varying degrees of uncertainty or "hedging" attached to them. This is why as a meteorologist we can assign both a possible probability of occurrence to an event and a confidence level such as low, moderate, or high (in both short term and long term forecasts).
In addition NWPs can have difficulty during times of "pattern changes": let's say the current pattern is one of a high amplitude flow, that is, the jet stream resembles a giant "sine" wave with large and amplified high and low pressure areas. Now let's say the pattern is forecast to change to a "zonal" flow; that is the jet stream takes on a more west to east orientation with the highs and lows having much less amplitude to them. (Even a zonal flow pattern can wreak havoc on the NWPs with timing the arrival and departure of weather systems (lows and highs), especially the farther out one goes with the forecast.
The state of the science (sometimes) does not permit more from a forecaster saying "We just have to keep an eye things" about a particular weather event many days ahead (sometimes even for the next day's weather - I'm thinking especially about severe weather during the late Spring and Summer). Frustrating yes! This is why meteorologists should convey this confidence (or lack of) factor to the end user. Perhaps a more probabilistic forecast would better serve the end user.
Because of the potential uncertainty involved in longer range forecast (let's say over 7 days, especially for days 5 through 7 of a 7 day forecast) the end user, the viewer, should look at the latter part of a 7 day forecast more as a probability of occurrence versus an absolute or deterministic forecast.
The state of forecasting IS actually getting better but I really doubt there will be a totally perfect forecast made anytime soon day in day out. I remember a sage meteorologist saying to me way back in my college days, "The atmosphere is capable of doing whatever it wants, when it wants. We can out forecast it at times and at other times we can't. Just accept being humbled and learn from what went wrong this time, so that there won't be a next time."
No comments:
Post a Comment