weather_logistics

Hi there. I'm a newbie to NetWeather. I happen to think that statistics are certainly the way forward for seasonal weather forecasting. I've previously done some spread betting for weather events. I just weigh up the risk of losing against the odds. If I placed a bet on a 9:1 event, but my personal skill was 7:1, then over a sufficiently high number of events I would expect to gain 25%. On a single event I take the bet knowing that I have a 87.5% chance of losing, yet I am willing to take the risk. Not surprisingly I didn't make much of a loss, nor a substantial gain. This is because humans often under-estimate risk, whilst computers are unbiased. The process certainly taught me to trust statistics and how to estimate risk. Gambling is highly addictive sport and the only sucussful gambler is one that doesn't have the motive or need to win. It then begs the question. Is it possible to Gamble Aware? I recommend that you read the information on this website www.gambleaware.co.uk. This may teach forecasters to understand how to make predictions. I often hear forecasters say: we were due some cold winters, or cold weather. It was on the cards. Gamble Aware states that: "In games of chance there is no such thing as a win being due." In the sense that somebody could guess, or predict the weather without a model or factual evidence: "It may seem that you, or someone you know, is luckier than other people". There is no evidence that anybody, over a significant number of events, is more lucky than anybody else. It is an interesting point, that the Met Office appeared to have an unlucky time at predicting seasonal weather on numerous occasions. That may suggest that the model they used was not producing successful forecasts. Was the outcome statistically significant, or do they need to start developing different forecasting models for longer range predictions. Could somebody have achieved better results by chance? Weather can also be an obsession or addiction. Human intervention in a seasonal forecast produces bias. I would hazzard a guess that Global warming believers tend to over-estimate seasonal temperatures. It is logical to believe that ski and snow-board lovers, prefer to believe that the winter will be colder and snowier. The mind is often warped from reality into what it wants to believe. I want it to be cold and snowy, because I love the snow. Mild winters are a threat to my snowman, to my skiing practise, to my idealised Christmas of log fires and crisp and even snow. So whether you are a snow lover or cold hater, there has to be some medium in-between that enables unbiased thought. My main seasonal forecasting objective at Weather Logistics UK is to try and push forward the idea that seasonal forecasting can be successful within certain bounds. These bounds can be focussed or refined as time progresses. Statistics are rarely right on one occasion. Variability in weather is high enough on any time-scale to warrant averaging. Small shower clouds can't be predicted the day before, but weather forecasters can say that showers are more likely in one region than another. Models can inform the forecaster how likely the showers are (30%), but they can't tell us for certain that showers will not occur, nor that they will. My model predicted that autumn will be 66% likely to be stormier than average, though I was 1/3 likely to be incorrect. My model predicted that temperatures would be +0.6C above 1961-1990 average during the autumn, they are currently about +0.3C above. Variability between the months is high and is mostly unpredictable and as such my temperature bounds show a large error range. One that I don't really think I can reduce. Recent validation of my autumn forecast tells me that the regional bias in temperature is consistent (reporting a standard deviation of 0.1C), so that the distribution in temperatures appears to be close to those observed. The baseline temperature in the model is actually +0.3C above average, however the model did not predict this. I choose a base temperature to distribute based on the current global temperatures and La Nina - 0.6C was considered to be a good guess. The weather-types and trend in weather was realistic, based on the monthly migration of the jet-stream. Storminess is slightly above average, though not quite as high as expected. What I am happy with is that the results are mostly within my estimated errors. My future work will be trying to reduce these uncertainties in seasonal weather. Weather Logistics UK uses a different forecasting system, based on an empirical temperature baseline and predictions of North Atlantic Blocking patterns. Temperature data was obtained from the National Oceanic and Atmospheric Administration (NOAA). Modelling was undertaken in GNU Data Language. See the validation of version 2 data at : http://www.weatherlogistics.com/Software_Outputsv2.pdf