Trom

Surely it's time for Shannon entropy to be defined. Personally I hate this term when it is a measure of uncertainty or volatility. On the model threads it's used to indicate high uncertainty but really it's just a measure of uncertainty and can be high or low. Entropy just means informational content - and following on from this the likelihood of a predicted event occurring beyond that you get into logarithms pretty quickly. A shannon is just the unit measure of an event occurring 1 shannon = 50% given the forecast of future events. The only important thing to take away is high Shannon entropy = high uncertainty (volatility of potential future outcomes).

I think you may have understood me (or me you). I meant you could look at an individual member and compare it to the mean to see how many standard deviations is was from the mean to see if it was an outlier. I appreciate that the mean is an average but surely the further an a particular member lies from the mean the less likely it is to occur or am I miss interpreting what I'm looking at?

Surely the mean gives an indication of how many std deviations a particular ensemble lies from it and as a result reliability or how much of an outlier you are looking at?

There's nothing in the reliable timeframe to suggest that is there, unless you cherry pick model runs? In fact consensus between models and runs seems very low to me right now at that timeframe. Still, as said, the westerly atlantic pattern will stop at some point. Just that we've all been looking for that since December!

Not sure I agree with you re the Netweather forecast support. Saying it will be a battle of West vs. East is a given for the UK. Trying to pick when the underdog will break through or be knocked back is the key. So anyone picking up on the westerly dominance from December until now, whilst backing the big dog, deserves credit. Normally that easterly pup has a few nibbles at our heals but this year the westerly influence is one big Canadian dog. Still, watching those low pressure fronts spiralling across the Atlantic has been compelling for those that find extreme weather in all forms interesting.

Well GFS 6z really gives no respite in the rain for the next two weeks in the South.

The model watching has been extremely educational for me this winter. I know, very frustrating for cold lovers, but none the less an unparalleled period of weather. Conversely I've learnt a lot about the synoptics we need for prolonged cold from this period. Especially the battle a Scandi high has against the default Atlantic position and exactly where it needs to be for that easterly progression. The models have been far from boring even if they aren't bringing everyone the weather that they want.

Well having got back from a business trip to my head office in Wisconsin which got as warm as -9C in the day and down to -28C at night our current weather is feeling balmy. I love a bit of snow as much as the next person on here but that was a little too cold for me. Currently reading 0.6C in Streatham (SW London).

Whilst I don't want to pee on anyones fire I would point out how quickly the models flipped to this output from a largely westerly dominated set up yesterday. Also the fact that the Meto and models were completely wrong with Fridays forecast as recently as yesterday. I guess what I'm saying is there are a lot posters saying "nailed on cold" which can make us all look like fools if the models flip again. I guess that's life if you live on the margins like we do in the UK.

Lets face it the models and the Meto were forecasting heavy rain for the South East all today, as late as yesterday, and now it looks to be dry day.

Yes basic stats says we need standard deviation and mean to describe a distribution, well a normal one at any rate. Are the ensembles normally distributed or do they show skew from time to time. Also is there any measure of deviation from the mean reported?

With the amount of variability we are seeing in the short range I'm not sure looking a a specific models output in 9 days time is really meaningful. Totally agree that what we need to be seeing is less variability between models and runs before anything has any credibility at this timescale.

I'm not sure the scatter point is correct. I think that would be captured by standard deviation. I think this is looking at variability in the variables from one model run to the next so time is a factor.

Uncertainty in the variables oops looks like I was beaten to it.

Hello I was wondering if anyone could help me with the answer to a post I made last night. I think in the excitement of the models changing so dramatically it rather got missed. That and the fact I was working through the night so made it at 2am! Also please feel free to redirect me where to post queries like these as I’m not sure this thread is where they should be. Also how to quote multiple replies rather than having to copy and paste would be a useful pointer too! Thanks for helping Trom, on 08 Jan 2014 - 01:38, said: Yes thanks - I understand friction is lower closer to the surface, but I've read that the effect is 10 - 15 degrees deviation in direction between boundary airflows and those at surface level coming in off the ocean, and around 40 - 50 degrees from the land. I'm lacking a relative scale, I understand it puts the centre of the low further forward but am not sure to spot this from the model output. Is the best method simply looking at model output at different levels of the troposphere? I guess I'm looking for output that would show wind spiralling towards the low pressure centre, or out from a high pressure centre, at the surface layer, compared to a bigger swing right at higher boundary levels if the Coriolis effect dominated the pressure gradient at this height? The impact would appear significant to me but I lack the experience or confidence in how all the variables interact to work out whether I should be obsessing about this. It just seems a big deviation, and as a economics lecturer and financial modeller it stood out. As said I'm very new to this but I am enjoying watching mathematics versus mother nature. Almost as interesting is the range of posts on this board and the topic of behavioural finance. Some might call it human nature but then I'd be out of a job. Best call it cognitive dissonance and confuse people with terminology! On a side note I think I'm an fan of extremes hot, cold, windy or whatever. Watching the models of these brutal lows steaming in off the atlantic and then seeing the effect has been a buzz. Hey I got stuck on a steam train visiting Santa with a four year old and one less than two, due to a fallen tree. When you've survived three hours of that and the blitz spirit being rekindled by railway enthusiasts singing Rudolph the red nosed reindeer you know its been an amazing period of weather. I've learnt a lot seeing the pressure gradients and temperature differentials between Canada/US and the warmth of temperate air from the South fuelling the Atlantic to unbelievable levels. Not cold for the UK but amazing to watch and a rare event. john w, on 07 Jan 2014 - 23:49, said: Hey, Generally, friction becomes more dominant at the surface whereas the pressure gradient force and coriolis effect tend to balance each other out aloft. Now of course you get rising and falling air etc etc, so air will always converge towards the low centre and diverge from the high centre, but crossing of isobars is more pronounced at the surface. Make sense? Trom, on 07 Jan 2014 - 23:15, said: Hello As a newbie I have a question regarding air circulation around high and low pressure areas and how to interpret them on the various models. I'm not sure if this is the right place to post but the site is fairly daunting with the sheer volume of message boards so apologies if I'm in the wrong place (please feel free to point me in the right direction). Anyway the question is when looking at airflows hitting a low pressure area on one of the models output pages: how can you infer whether the Coriolis effect will dominate or the pressure gradient force? Also I'm struggling with how to interpret the impact of friction on the Coriolis effect and whether I really need to consider it if the airflow is a westerly, given it's coming of the ocean? I'm sure these questions betray my own ignorance but an answer would help me in interpreting model output even if it's "You are missing the big issue and focusing on a small/irrlevant one muppet!"

Yes thanks - I understand friction is lower closer to the surface, but I've read that the effect is 10 - 15 degrees deviation in direction between boundary airflows and those at surface level coming in off the ocean, and around 40 - 50 degrees from the land. I'm lacking a relative scale, I understand it puts the centre of the low further forward but am not sure to spot this from the model output. Is the best method simply looking at model output at different levels of the troposphere? I guess I'm looking for output that would show wind spiralling towards the low pressure centre, or out from a high pressure centre, at the surface layer, compared to a bigger swing right at higher boundary levels if the Coriolis effect dominated the pressure gradient at this height? The impact would appear significant to me but I lack the experience or confidence in how all the variables interact to work out whether I should be obsessing about this. It just seems a big deviation, and as a economics lecturer and financial modeller it stood out. As said I'm very new to this but I am enjoying watching mathematics versus mother nature. Almost as interesting is the range of posts on this board and the topic of behavioural finance. Some might call it human nature but then I'd be out of a job. Best call it cognitive dissonance and confuse people with terminology! On a side note I think I'm an fan of extremes hot, cold, windy or whatever. Watching the models of these brutal lows steaming in off the atlantic and then seeing the effect has been a buzz. Hey I got stuck on a steam train visiting Santa with a four year old and one less than two, due to a fallen tree. When you've survived three hours of that and the blitz spirit being rekindled by railway enthusiasts singing Rudolph the red nosed reindeer you know its been an amazing period of weather. I've learnt a lot seeing the pressure gradients and temperature differentials between Canada/US and the warmth of temperate air from the South fuelling the Atlantic to unbelievable levels. Not cold for the UK but amazing to watch and a rare event.

Hello As a newbie I have a question regarding air circulation around high and low pressure areas and how to interpret them on the various models. I'm not sure if this is the right place to post but the site is fairly daunting with the sheer volume of message boards so apologies if I'm in the wrong place (please feel free to point me in the right direction). Anyway the question is when looking at airflows hitting a low pressure area on one of the models output pages: how can you infer whether the Coriolis effect will dominate or the pressure gradient force? Also I'm struggling with how to interpret the impact of friction on the Coriolis effect and whether I really need to consider it if the airflow is a westerly, given it's coming of the ocean? I'm sure these questions betray my own ignorance but an answer would help me in interpreting model output even if it's "You are missing the big issue and focusing on a small/irrlevant one muppet!"

I'm trying to get my head around all this too. On a basic level the Stratosphere is where ozone can be found. So heating of the Stratosphere is the result of absorption of of UV radiation from the zone by molecules of ozone. This would explain the seasonal variation in Stratosphere linked to the amount and intensity of UV radiation coming from the sun. This of course doesn't really help with the SSW or explain it.

It is true - it's why I said deviations (meaning difference between observation and mean) rather than 50% of the outcomes. Skew would affect the median, and mode in relation to the mean but not the arithmetic mean itself which is just a mathematical fact. Skew does have the biggest distortion on mean and least on mode with median in the middle. When we say skew has a big effect on mean it's really in comparison to normal distribution. In a normal distribution mean = median = mode. +ve skew mean>median>mode and reverse for negative skew. I do agree that if I had said 50% of the outcomes then I would have been assuming a normal distribution, anyway splitting hairs. Out of interest does anyone know if the ensembles are normally distributed around their mean?

And conversely 50% of the deviations will be less severe than the mean It would be nice to see a +/-1.96 std dev band around mean on the ensembles

Yes but this is a thread about models not gut feel. There lies cognitive dissonance. Human nature but clearly a big factor when we have marginal conditions.