OCTOBER PATTERN INDEX (OPI) MONITORING WINTER SEASON 2014-2015

[shaky]

Very good 12z so far!!opi could be even more negative on this run!!heights look more robust!!

[karyo]

Beautiful 12z, the low pressure is even more scattered then in the 06z and higher geos above the arctic.

Yes, the gfs is great! Once again, the UKMO has the least blocked set up.

[Mucka]

A little more info from Guido Guidi after I explain what I was trying to understand.

 

I was interested in how the forecast OPI affected the correlated OPI as I thought the +1.5 swing (-3.3 to -1.8 from the 19th to 20th was quite large given we had 19 days correlated data and only 10 forecast days.

Since we are not given the OPI for the 10 forecast days and only the OPI for the consolidated days + forecast days I wondered if there was an obvious equation to back engineer the OPI for the forecast value only but of course we need to know the OPI without the forecast data to do that.

This is what confuses me. Even the consolidated data contains forecast data so that the value for the 19th for example contains forecast data out to the 29th but the figure will never change for the 19th no matter how accurate or inaccurate those forecast charts prove to be?

If the daily figure is a monthly running mean then that also means that even the consolidated data is actually previous unmodified forecast data + that actual days data?

 

I guess I still don't understand it. 

Can anyone tell me what the OPI would be if at the 25th we had an OPI of -2 and the last 6 days were forecast to have an OPI of +2?

would the +2 be divided by 25 and added?

Could we not have actual daily values superimposed on the bar chart?

I seem to have real blind spot on this so I doubt I phrased my questions very well for Guido and apologise to him but here is his response.

 

 the OPI index is a mean value, so, of course, when you add new members to your calculation they have influence on the mean depending on their absolute value. Jumps from one day to another are entirely due to forecast charts, expecially when the number of consolidated charts is lower than the forecasted charts. But this is not the case anymore since we are almost done with October. 

Daily stands for daily calculation, not for daily values and, once again, since the conceptual model is the (still unknown as well as for the SAI index) causation of the high correlation of the October developement with the winter developement, taking it as a daily (calculated only with a daily run) value would be maybe challenging but useless

 

I also asked as to whether they considered trends (negative first half of month.positive second half etc) in their final forecast or whether it was just based on the final OPI to which his response was as follows;

 

 my answer must be generic, since what we actually need is exactly something giving us a clue about the development of the winter not only the 'mean' AO index, and we know that this could be in the data, but is still matter of research.

 

So I will leave poor Guido and his team in peace now but have of course invited them to the forum with any updates or information.

Edited October 22, 2014 by Mucka

[bluearmy]

that appears to say that the number of the OPI today is the running figure for todays date + forecast data. its not about working out the mean of all the days figures thus far. todays -1.85 is the actual OPi for october, assuming that the actual charts for the remainder of the month are as per the 06z gfs run. that does mean a sudden flip in the modelling could throw the number much less negative but according to guido's message, they dont consider that likely based on historic data.

Edited October 22, 2014 by bluearmy

[feb1991blizzard]

that appears to say that the number of the OPI today is the running figure for todays date + forecast data. its not about working out the mean of all the days figures thus far. todays -2.75 is the actual OPi for october, assuming that the actual charts for the remainder of the month are as per the 06z gfs run. that does mean a sudden flip in the modelling could throw the number much less negative but according to guido's message, they dont consider that likely based on historic data.

 

So really then, the only figure relevant is the latest one, apart from dissecting the start of the month for the purpose of further research of course.

[Barry12]

-1.89 on gfs 12z, nearly an increase of +1

Edited October 22, 2014 by Barry95

[Continental Arctic Front]

OPI back to -1.89. Strange, in my opinion the 12z was better then the 06z.

[Barry12]

OPI back to -1.89. Strange, in my opinion the 12z was better then the 06z.

 

I know

[shaky]

OPI back to -1.89. Strange, in my opinion the 12z was better then the 06z.

exactly!!I thought worse come to worse it would equal the 06z run!!!hmm maybe we missing something?am sure steve murr will explain later!

[SMU]

HI Bluearmy-

 

Guido Wrote-

 

The OPI index is calculated on a daily basis, but becomes a significant tool only if thought as a mean value for the month of October.

The main objective of the calculation (and therefore of the page where is shown) is to collect as many charts as possible for the month to start giving a suggestion of what is going on since the beginning of the period. That means that the reliability of the value as it builds up depends strictly on the reliability of the forecast at the beginning of the months and starts to be really reliable only when the calculation has inside a greater number of consolidated than forecasted charts.

To be clear, think that for a given day x and a given run, the calculation will be based on the whole consolidated (past) days from October 1^st to day x plus the forecast x+10 days charts. Since the figure is less reliable until 2/3 of the month are gone, the bars on the graph are gray. When the value starts to be stable (remaining 1/3 of the month, the bars are blue or red depending on the sign of the index.

Obviously, we could calculate the index only with consolidated data, and of course this is what we do at the end of the month, but that would mean that we could calculate it (and start discussing on it as on this forum) only when the number of charts (days) considered starts being sufficient to get a stable result, leaving us without chances to start evaluating it before the end of the month. Additionally, at this stage, the value is stable enough to be probably pretty close to what we could have leaving the forecast charts out of the calculation.

Hope this helps,

Guido

 

Guido says its the consolidated charts + the 10 days forecast

 

So it must be the Running Mean + the forecast = OPI.

 

So for example if it was the 5th & the OPI was -1, & the net forecast for the 10 days was -2, then the OPI would be -1.66.

 

However heres the rub or shall I say poor presentation peice.

 

When you go on the OPI link to the graph below these CANNOT be the running mean values because every day the graph is the plotted as of the final OPI updates from the consolidated value + the forecasts- usually from the 12z run.

 

So the graph at the bottom cant be used for ACTUAL running mean for us becuase the data on it ISNT the actual running mean, its the running mean of the OPI Actuals & forecast.

 

Based on that then we may as well only use the first 21 days as a guide, with particular resistence to the first 10 days as it has ZERO data, dates 10-20 have 10 days Actuals & 10 days forecast so will be better

Days 21-31 will be the BEST data as they have 21 days of data & only ten days of forecast...

 

 

REMEMBER THEN ALL THE VALUES ON THE GRAPH ACCORDING TO GUIDO ARENT ACTUALS. They are Actuals + 10 days forecast ( they may be CLOSE to actuals though)

IGNORE ANY POST WITH RUNNING MEAN & GO WITH THE OPI FIGURE AT THE TOP FROM TODAY AS ITS THE CLOSEST THING TO ACTUAL RESULTS WITHOUT HAVING ANY ACCURATE DATA!!!!!!

 

There should be a graph below with actual recorded values.!! for running mean purposes, then you can see the model trending in the 10 day timelines!

 

 

S

Edited October 22, 2014 by Steve Murr

[easton888]

The OPI figure now (-1.89) cannot be the running mean total because it fluctuates so wildly, even around the -1 to -3 regions.  Surely some forecast data must have caused it to jump? Weird though, as you say the 12z was a pretty good run!

[Barry12]

Maybe better blocking, but the angle of blocking wasn't good, we need steve murr!

[neige57]

I for one OPI index is stable now and oxcille between -1.5 and -2.5 so the final figure should be within this range at the end of the month which is well below

Edited October 22, 2014 by neige57

[johnholmes]

why oh why use more than 1 output at such long time scales-it really is close to a waste of time and effort using multiple output times at time scales beyond even maybe 144h and certainly 240h. Simply look at how often EC shows differences in its 2 runs and how often at 240h is GFS is even similar on its 4 runs let alone identical. Close in by all means use all the data.

 

I suggested at the beginning of this thread that using a forecast to predict what is another forecast to then predict something else was just going a bit too far.

Edited October 22, 2014 by johnholmes

[Gael_Force]

I don't think it is the mean of the daily numbers, it's the mean of the NH pattern that is being calculated along with the next projected 10 days. The numbers just reflect the calculations - at least, that's how I read the email contents where they emphasise charts rather than figures.

[chionomaniac]

HI Bluearmy-

 

Guido Wrote-

 

The OPI index is calculated on a daily basis, but becomes a significant tool only if thought as a mean value for the month of October.

The main objective of the calculation (and therefore of the page where is shown) is to collect as many charts as possible for the month to start giving a suggestion of what is going on since the beginning of the period. That means that the reliability of the value as it builds up depends strictly on the reliability of the forecast at the beginning of the months and starts to be really reliable only when the calculation has inside a greater number of consolidated than forecasted charts.

To be clear, think that for a given day x and a given run, the calculation will be based on the whole consolidated (past) days from October 1^st to day x plus the forecast x+10 days charts. Since the figure is less reliable until 2/3 of the month are gone, the bars on the graph are gray. When the value starts to be stable (remaining 1/3 of the month, the bars are blue or red depending on the sign of the index.

Obviously, we could calculate the index only with consolidated data, and of course this is what we do at the end of the month, but that would mean that we could calculate it (and start discussing on it as on this forum) only when the number of charts (days) considered starts being sufficient to get a stable result, leaving us without chances to start evaluating it before the end of the month. Additionally, at this stage, the value is stable enough to be probably pretty close to what we could have leaving the forecast charts out of the calculation.

Hope this helps,

Guido

 

Guido says its the consolidated charts + the 10 days forecast

 

So it must be the Running Mean + the forecast = OPI.

 

So for example if it was the 5th & the OPI was -1, & the net forecast for the 10 days was -2, then the OPI would be -1.66.

 

However heres the rub or shall I say poor presentation peice.

 

When you go on the OPI link to the graph below these CANNOT be the running mean values because every day the graph is the plotted as of the final OPI updates from the consolidated value + the forecasts- usually from the 12z run.

 

So the graph at the bottom cant be used for ACTUAL running mean for us becuase the data on it ISNT the actual running mean, its the running mean of the OPI Actuals & forecast.

 

Based on that then we may as well only use the first 21 days as a guide, with particular resistence to the first 10 days as it has ZERO data, dates 10-20 have 10 days Actuals & 10 days forecast so will be better

Days 21-31 will be the BEST data as they have 21 days of data & only ten days of forecast...

 

 

REMEMBER THEN ALL THE VALUES ON THE GRAPH ACCORDING TO GUIDO ARENT ACTUALS. They are Actuals + 10 days forecast ( they may be CLOSE to actuals though)

IGNORE ANY POST WITH RUNNING MEAN & GO WITH THE OPI FIGURE AT THE TOP FROM TODAY AS ITS THE CLOSEST THING TO ACTUAL RESULTS WITHOUT HAVING ANY ACCURATE DATA!!!!!!

 

There should be a graph below with actual recorded values.!! for running mean purposes, then you can see the model trending in the 10 day timelines!

 

 

S

I think I said that about 10 days ago Steve.......

[Weather-history]

why oh why use more than 1 output at such long time scales-it really is close to a waste of time and effort using multiple output times at time scales beyond even maybe 144h and certainly 240h. Simply look at how often EC shows differences in its 2 runs and how often at 240h is GFS is even similar on its 4 runs let alone identical. Close in by all means use all the data.

I suggested at the beginning of this thread that using a forecast to predict what is another forecast to then predict something else was just going a bit too far.

It's interesting in terms of research but IMO far too many people on this site are getting too hung up on all this, reacting with every nuance. I post on TheWeatherOutlook forum and Ukweatherworld as well as here and it hasn't drawn anywhere near the interest it has on this site. 40 pages already this thread?!

God knows what the mood will be if next year it returns a large +ve value, judging by some of the posts reacting to values that are not as negative etc. Winter over posts galore and it will be at the start of November!

Edited October 22, 2014 by Weather-history

[radiohead]

The OPI figure now (-1.89) cannot be the running mean total because it fluctuates so wildly, even around the -1 to -3 regions.  Surely some forecast data must have caused it to jump? Weird though, as you say the 12z was a pretty good run!

 

It fluctuates because it's still using 9 days of GFS output in addition to the actual data from the previous days this month. In the coming days it will use less and less GFS output so there will be less deviations until eventually by the end of the month we get the final figure. The daily figures are only really useful for giving an idea of what the final number will be as you get closer to the end of the month. The final number is the only important one.

[lorenzo]

6z and 12z for the shredder , gfs missing key satellite data.

[SMU]

I think I said that about 10 days ago Steve.......

You did - but you didnt enlarge the font & put it in bold !!! Lol

Plus it was slightly uncertain at that point because i wasnt checking the actuals were the same as the forecast the day before-

Edited October 22, 2014 by DiagonalRedLine
(Added in the quote from the person he was responding to)

[Yarmy]

From the very first post of this thread:

 

"The software analyzes the previous consolidated daily charts plus 10 forecast charts as given by the GFS model. For example on the 15th of October, the software calculates the OPI index based on the 15 consolidated daily charts (1-15 October) adding 10 forecast charts as given by the GFS model (16-25 October)."

 

It's always seemed pretty clear to me. It's just the nature of the calculation itself that is unknown (yet).

[chionomaniac]

6z and 12z for the shredder , gfs missing key satellite data.

I think it has been missing for days (20/10) from what I have been reading and is linked to the loss of satelite sea ice and snow data.

[Mucka]

that appears to say that the number of the OPI today is the running figure for todays date + forecast data. its not about working out the mean of all the days figures thus far. todays -1.85 is the actual OPi for october, assuming that the actual charts for the remainder of the month are as per the 06z gfs run. that does mean a sudden flip in the modelling could throw the number much less negative but according to guido's message, they dont consider that likely based on historic data.

 

Not as I understand it. The daily value is the mean of the consolidated data up to that date + up to 10 forecast days.

[joggs]

Its starting to resemble the mod output thread in here lol.

Winters over now its gone up 1!!!

[Barry12]

make of this as you will