Recretos

This is a very cool topic (no pun intended), so I decided to share a few words and graphics. First of, the N Atlantic SSTs on 12.10.13 00z. Taken by NOAA/NESDIS geo-polar orbiter with 5km horizontal resolution. The Atlantic is still running quite warm, with positive anomalies, slightly variying from year to year, but keeping the general trend of the positive phase of the Atlantic Multi-decadal Oscillation (AMO). Just to the SW, the gulf stream looking warm and healthy, connecting with the positive AMO. Now for the global image. The first things that stand out are of course the warm Atlantic, cold PDO off the W USA coast, and the cold biased neutral ENSO regions. The main two players, PDO and AMO are generally staying in their respective phases over the last years, with ENSO having more colder biases than warmer, due to the cold PDO above. Too bad the USA government is in shut-down, because I could make some nice reanalysis about it, but oh well I shall work with what ever I have at the moment. Now compared with the CFSv2 forecast. CFS was quite "on par" with the reality, which was kinda expected from the latest model runs. The PDO could be colder, and there is the warmer main ENSO region. As we all know by now, CFS is also forecasting a weak to moderate El Nino well into summer, while the general consensus of the climate models is more in favour of a neutral phase, with a slight possibility for an El Nino by summer. But that is too late in the forecast to have any significance just yet. The CFS kinda rushes the ENSO into the warmer phase, while I am a bit sceptical about such a fast warmup, but of course it is possible. When looking at CFS geo-potential anomaly forecasts, to me it looks like it is trending more towards an overall positive AAM (Atmospheric Angular Momentum) feeling, semi-explaining the forecasted El Nino, since ENSO is basically much more sensitive to AAM than vice-versa (a huge simplification ). What is interesting, is the UK-Met forecast. Their ENSO forecast keeps it under as a weak La Nina, while the 500z anomaly forecast look more El Nino-ish. I guess they are aiming for at least one verified forecast. Now to go away from the future and into history. Last year around this time, the PDO was more defined, north Atlantic locally warmer (while this year we are generally warmer in the Atlantic) and the warm biased neutral ENSO going soon into a cold-neutral transition. (Note: Extreme anomalies at the poles are not realistic. That is just unreliable data because of the ice cover. ) This would be all for now. As usually, I tried to write in a user-friendly style, so I hope that most (if not all) of the text is understandable. Best regards.

It is for 13/14, becasue it is a 3 month lead from september 2013. That is just a text error.

Yes. They have two options: Monthly and annual subscription. You can check the pricing and other information at their site.

Well, to make it clear: I am just presenting this, not defending it. I agree with you. I posted it because I think that since the analog forecasting is quite popular, some might find this automated system interesting. For any further discussions about this "model" (which isn't really a model at all), please contact Mr. den Dool over email. I will finish off with a fairly interesting current SSt setup. Regards

Well, for all the lovers of analog forecasting, here is a special treat for you. Using the automated constructed SST analog method. The original description: " The construction is a linear combination of past observed anomaly patterns in the predictor fields such that the combination is as close as desired to the initial state (or 'base'). We use as our predictor (the analogue selection criterion) the leading EOFs of the global SST field for consecutive 3-month periods during the year prior to forecast time. Data extending from 1955 to the present are used for a priori skill evaluation under cross-validation (CV)." To interpret the process above : Check current SST/ocean conditions over a certain period - check back in time time find a similar setup - create a composite of the years that have the best match. Of course there are other processes involved, but just to make it easier to understand the very basic idea. Now last year, this system was quite good, at least in my opinion. These images are self explanatory. What this system is trying to do this year, is to force in another neg. NAO episode with an average/final feel similar to the last winter. But we all know what the "game" is on a local scale, under these huge 3-month means. Best regards.

Hello everyone! Yes I am alive and kicking, and ashamed by my lack of activity here. But as we head slowly into the winter season, my activity will slowly increase again, just like the mid-stratosphere temperatures. And about the notorious GWO composites: I made around 70% of the work in around 3 weeks. Then I had some other stuff to do, and postponed the whole project, thinking I have a good system to extract the data and I can finish it at any time. Well, I got a new computer in between (saved all the data of course ), but the first problem was the new Excel version. And the biggest problem was, that I last worked on the project somewhere in March. So when I looked into the data last week, I had no idea how to finish it. I basically forgot my data extraction system. It was a huge mix of different functions which worked great, as long as I knew how to actually use it all together. But do not be sad, for I will finish the project. I always finish something that I decide to finish. It will only take a bit longer, because I will basically have to create a new system. But first I will have to interpret the finished data, because it is still in data form, and needs to be properly coded into files that will be uploaded to NOAA servers and plotted. Just have some faith in me, if there is any left at all for me. Ok, since this topic is about snow coverage, this is my post where I posted some correlations I made, between Oct-Nov snow and stratosphere temp. and U-wind. http://forum.netweather.tv/topic/74587-stratosphere-temperature-watch-20122013/?p=2446929 And some accumulated snowfall forecasts: Best regards.

Thanks weatherjunkie. To be honest, I was really short on time recently, so for the past two weeks, I haven't even opened the project files, let alone edit them. Next week I will have some more spare time, and I will try to make as much progress as I can.

Since it looks like we are going to go (more like speed through) into MJO 2, 3 and 4, I am going to add the phase 2 and 3 composites I made just for this occasion. Now in these composites, there is actually quite a difference when compared to the composites we usually look at, mainly in the strength of the anomalies. The main difference is of course because of the filtering. I cant wait to start the MJO project. I wrote the conditions on the composite. No lags applied. @Glacier Point: So I guess that the Feb. phases 2 and 3 actually are supportive for a negative NAO after all. And here are some forecast anomalies to compare. The Canadian ENS, and the NCEP analog super ensembles. And GEFS to finish off. Any professional opinions? Best regards.

Wow GP, it seems you kinda have it all figured out. For the sake of curiosity, I took the MJO data, and quickly extracted the phase 4 data, and ran it through the same filters as the GWO (Amp>1, neutral ENSO). And I must say that it is not really what I expected, when looking at the existing composite at http://raleighwx.americanwx.com. I found out, that his composites have the configuration: All AMP and all ENSO phases. That kinda "filtering" is quite easy and I could make plots like that in a day or two. Not to take anything away from those composites or his work. I think they proved very useful. It is just that I think we need MJO composites which will be ENSO and Amplitude filtered. Here is the filtered phase 4. In total 22 days in the composite. No lags applied. And the comparison. There are some similar features tho. +7 days lag applied. Edit: And once I extract and filter the MJO data, it should also prove useful for stratosphere forecasts, in combination with the GWO, since I will have the coded datafiles, and will be able to make many kinds of plots for the usage of stratosphere forecasting, so we wont be limited to just 500mb anomalies.

A very good point. It will be quite interesting to see these composites in action. If I quote: "assuming a phase 4-5-7-8 type orbit.". This was the main reason why I wanted to do MJO rather than GWO. MJO is being actively forecasted, so we can kinda anticipate the phases, while with GWO it is a bit more complicated. EDIT: I can provide the composites and all kinds of plots, but calculating correlations is a bit out of my league.

As some of you may know, I am doing the GWO project, making composites for each phase, for each month, for each ENSO phase, and for all amps and amps>1. I will finish the ENSO filtering in about a week or so. So I can start plotting in about 10 days from now. And then the next project will be updated MJO composites. This is the GWO wheel, with marked phases. Here are February, neutral ENSO, GWO composites, as requested by GP. Best regards.

Well, I think that is what we are going to find out, once we put the composites into practical use. I am already doing the ENSO filtering, and February is now 100% completed, and ready for coding the data files for upload. And here is the finished Phase 8 (0.5) for February, in ENSO neutral phase, and with AMP>1 P.s.: The MEI dataset goes till December 2012, so that is why I also decided to use the dataset for GWO Jan 1970-Dec 2012. So no 2013 dates included. EDIT: Just for fun, I did the same thing as yesterday. Taking the Phase 8 (0.5) which is now corrected with Neutral ENSO and AMP>1, and I plotted 10mb anomaly, with a 12-day negative lag applied. So before the onset of the phase. And I added the anomaly of this year for a similar period, around 12 days before the onset of the phase 8. Of course the anomaly of this year is much greater in the reanalysis, because it contains much less days than the GWO phase composite. Now I am not really trying to make a specific point here. Only trying to say that it should be quite interesting to experiment with all this data.

Ok so no extra composites with amp below 1. I have already completed the basic extraction for all phases for all months. Here is a composite of Phase 8 in February, all amps and all ENSO phases. In total 113 days in the composite. Maybe the signal would be even better, if amp>1 would be applied and only ENSO neutral years. EDIT: This is again phase 8 for February, but I quickly applied the AMP>1 filter. In total 77 days in the composite. A bit better match, but still with all ENSO phases. Maybe an even better match would come put if I would apply the ENSO filtering, but that would take some time, and as I have just find out, that will be the most time consuming part. EDIT2: Just for the sake of curiosity and to make the post less offtopic, here is the 10mb anomaly plot, with a negative 11 day lag applied. So 11 days before the H500 anomaly above. The 10mb mean is basically useless, because only the anomalies show the actual "tendencies" at this height, with so many days in the mix. For a comparison, I added the ECM analyse.

I really hope that once the special thread is created, all these posts are going to be moved there, because they are offtopic. But until then, I hope people dont mind if we share a word or two in here. @GP: I understand now. How about AMP below 1? Also required I assume? Edit: About amps: Are talking above 1, or above & equal to 1? and the same for below.

Well, the hardest part is extracting the data. Once I collect and code the data, I can plot what ever the reanalysis interface has to offer. So the plan is: 8 phases for each month, separated by Neutral, warm or cold ENSO phase. That means i have to create 8x12x3 custom coded data files for AMP≥1. The same number goes for a date set with the criteria AMP≤1. And shall you want all AMPs, it would be again the same number. So if I separate it by AMP criteria, that would mean 864 data files. Now you said the plots would be H500 mean, H500 anom, and T850 anom for N HEM. That would mean in total of 2592 graphics. the size of one is around 35KB, so that would mean around 90MB of graphical data. Now my plan is: 8 phases for each month, divided into 3 ENSO phases and divided into 2 AMP criteria, above or below 1. I dont think making a composite of all amps even has a meaning. I am already working with the data, and it is really confusing. edit: I will also have all the collected data saved, so I can do custom plots or "on demand" plots if it may be required.

I have finally perfected my data extraction system, which will speed up the process. Took me around 6 hours, and another hour to figure out how to convert the extracted data directly into a code form suitable for NOAA reanalysis system. I am not so familiar with excel and data editing, so that is the reason it took so long. So I am now "officially" launching my GWO project. You have to confirm if this is the most optimal configuration, so I adjust my data extraction to this configuration. And by magnitudes, you mean the AMP (amplitude) column in the GWO text data? And I have selected the data time span to be 1970-2013. I will not include the 58-69 period, because I kinda doubt the 100% accuracy. And about MEI, the data is in dual month format, basically like a 2 month average. So I am just going to take above 0.5 as El Nino and below -0.5 as La Nina. I am also going to use the ONI index to really determine the ENSO phases. One heck of an apple to bite in, but it should be done with some patience. I apologise for being offtopic again. I really hope a new thread can be arranged for this matter. Best regards.

Before you go on with the models, here are some skill scores for the near past. Basically a "percentage" like factor: 0.900 is 90% accuracy, 0.850 is like 85% accuracy, and so on. Below that, is the RMS (Root mean square) error. Basically the lower the better. How good were the model with forecasting N. Hem. 500mb pattern, 5 days in advance. First, this is the success of individual GFS runs. The 12E and 00E, are experimental runs. Next to it, are comparisons of specific models, GFS, ECM, UKMO, MSC (CMC-GEM) and NGP-NOGAPS. The third graph represents the comparison between GFS, ECMWF, and their control runs. In average, the ECMWF control run had better performance than the GFS operational. Next up are ensembles. First the NCEP. Black line is the operational, red line is the control run, and the green line is the ensemble mean. This is for day 8. And GEFS skill score "die-off" with time. It is normal for the ensembles to die-off, because as the ensemble deviation gets stronger, it is not picking up specific features that good anymore. And ECMWF ensembles. First at day 8. Not bad I would say, and quite better than the GEFS. And the die-off. ECMWF control, and ENS mean. At day 10, ECM ensemble mean has a skill of 0.57, while GEFS mean has a skill of 0.49. So you will know where the models and the ensembles stand. Best regards. Edit: Snowking was faster with some skill scores.

I know this is the "classic" model discussion thread, but anyway, here are some additional weekly forecasts to help you with the trends, directly from Asia. They will probably get lost in the massive crowd of posts anyway. JMA weeklies, latest run 31/1. For easier orientation: Atlantic is on top, under the ridge on the week 1. T2m anomaly. And the Korean (KMA) weekly (10-day) ensembles. They have proved to be quite good in the past. Best Regards.

Thanks everybody for the positive feedback. @Matt: Thank you. The data collection might not be as hard as you might think, and as hard as I thought to be. I have just "developed" a certain data extraction system, using Excel. The NOAA GWO data files are very nicely coded, and I was able to convert that into .csv and then into .xls, and separating each value (year, month, day, phase, amp, etc..) into its own column. From there on, I used a simple system of filters, with which I applied a certain "rule" to a certain column (year, month, phase, amp,...). For example: I used filters in the phase column to show me only rows with the phase value being equal to 4.5 for example. And if I apply a filter to the amplitude column, for example "equal to or grater than 1.3", I am only left with date rows that meet the condition of having a phase 4.5 at amp 1.3 or stronger. Then I still have to manually determine the ENSO phase of each year, but this extraction system should still save some time. I just have to "tweak" it a little bit first. After the ENSO filtering, I can already start the coding of the data files that will be uploaded to the NOAA servers as a base to plot reanalysis from. So right now, I am working on the system that would speed up the data extraction, and would of course be useful for the MJO, if the original text data is coded the right way. I have already successfully tested coding a datafile containing dates, and used it as a database for reanalysing. I can also apply custom plot title, which will be useful, so each graphic would contain custom text instead of showing the dates in the composite as per default. I could then apply text such as "GWO phase 4.5 amp>X El Nino", for example. Two test runs: So basically I kinda already have the idea in my head, how I can make this project a reality. I just need some time, and to work out some details first, before I really "dig in". Best regards.

@GP: You gave me quite of a homework there. In my original plans, which go back to autumn 2012, I was deciding between the MJO and GWO climatologies. I basically decided to use the MJO over GWO, for just one particular reason: MJO is something that is being actively forecasted. And knowing the climatology of the phase which is being forecasted, that would be useful to some extent in forecasting. But what attracted me with the GWO, is its actual "connection" with the patterns, and when considering that it goes "hand in hand" with GLAAM, I guess that the climatologies would have more consistency. Now my main dilemma with the GWO as the choice, is in the usefulness of these climatologies. Is the GWO actually forecasted the way MJO is? Or is it forecasted by deriving it from other factors, like AAM tendencies, with combination with GWO? It is clearly not something that would be addressed as "basic" meteorology. The structure you suggested can be accomplished. For the sake of the objective data, I would then use data from around 1970 on. Basically that is when the satellite observations of synoptic scales of weather were getting more "advanced". I have answered Paul in a PM, and I would like to thank him to be open to assist in the matter. Thanks also to Snowking for the kind suggestion. So in February, I will make a few "dummy" runs, to first really determine the best methodology, than to try coding the dates into data files, ready to be uploaded to NOAA FTP servers as custom time-series, and of course, some minor but important details. I am a perfectionist by nature, so I want to do my best, to really deliver something quality, and unique in its own extent. I am already testing the coding of the datafiles and the FTP uploading and practical use. So I think now would be a good time to make a final agreement, if I start the GWO or MJO climatologies project. So far the "big two" have suggested GWO, and since I am 50/50, GWO looks to be the choice for now. I just need to find some nice complied database of the GWO in a text format. The overall project is nothing special, but since it is time consuming, I really want to get around all the details before I actually start collecting data. I think we all know pretty well how important time is these days. Best regards. Edit: Thank you Chionomaniac for being open to cooperate. And a special sub forum wouldn't be a bad idea. But as I said, I first have to determine the "what" and "how". And you posted basically just seconds before me.

@Interitus: Thanks for the suggestion and the link to the archive. . I really have to develop a "fixated" methodology before I start. I was thinking of using phases with amplitude above 0.85. And for the ENSO years separation, the neutral period being MEI from 0.5 to -0.5. Everything below of above, counting as a La Nina or El Nino year, respectfully. I am also thinking of using the ONI index for ENSO phase determination, but it is seasonally averaged. And about the lag. I am planing to use a number of days that will depend on the longevity of the phase. More long lasting the phase, more days I will take out. But of course not all. And the main goal will first be to extract the exact phase days. I can apply lag later when I begin the reanalysis process. @Chiono: The research area would sure need someone knowledgeable. I can do the basic climatologies and everything, but that will have to be "verified" and put into practical use. And about the GWO: I was actually thinking of that, but to be honest, those composites would have far less forecasting usefulness than the MJO, which phases and movement is actually being forecasted. So having climatologies for a certain forcing which is being forecasted, would be more meaningful, at least in my view. I am a bit worried about the deviation about the actual response within each phase, We all know the "system" is more complicated and with more forcings than just the MJO, so the outcome is not always the same. But the current climatologies at http://raleighwx.ame...wx.com/MJO.html have been quite useful as a guidance or a display of tendency, so that gives me some confidence. Just one thing to consider: If I actually manage to make these climatologies, it would probably be quite a few graphics. And I dont really have a way to host these graphics. And just keeping them on my computer is basically nonsense. So I would like to know if there would be a way that these graphics would be hosted here on Netweather, having a page of their own? Of course I have to make the graphics first, but I have to ask in advance. Basically the only really useful climatologies are the ones on the http://raleighwx.ame...wx.com/MJO.html site. So Netweather would basically then be the only site to have updated climatologies, with Europe maps, and different parameters. Just for consideration. Best regards.

Hi everyone. I see the topic is still very active, which is really awesome. I apologise for my lack of posts, but I am really busy with some side projects, which basically takes my time away from all forums. And besides that, I don't really post unless I have something worth posting. Just wanted to say, that during the year, in my free time I have a plan to make fresh MJO climatologies (basically anomaly reanalyses) for different phases. I think I am getting quite good with reanalysis and I have found some MJO archives for the entire 1980-2012 period, from many different sources, which should be enough to make a basic phase climatology. Basically I would do this to kinda test my reanalysis skills. (But as far as it looks, no real skills are needed , mainly time) Now there is something to consider: 1) The before mentioned time lag. I was planing to use only last two days of a certain phase, with the condition that the phase must be "active" for at least 6 consecutive days, so I would use only the really "influential" phases. 2) The different strength of certain phases. Sometimes a certain phase in a certain month in a certain year, is weaker/stronger than in a different time. So putting it all in one mix would give us a certain mean value or "mean anomaly", but the respected deviation would probably be quite high. 3) I had 3 point, but I forgot what the 3rd point was. Will add it later if I remember. So the main difficulty is of course the collection of dates. Once I have all the dates for the phases, the main work is basically done, and I can start playing around. From there on, I can divide it into specific ENSO years, SSW years, QBO years, different combinations, etc,... I could make specific maps for N. Hem, Europe, etc... Different parameters like T2m, H500, U-Wind, etc... I have in plan to make a dates-datafile for each phase in a specific format, so I can upload it to the NOAA FTP servers, and use it as a custom time-series for reanalysing. Basically then the options are almost endless, because I always just use the uploaded file, instead of writing the dates every time into the reanalysis interface. And not to mention, that the daily reanalysis interface supports only 20 custom dates, while I can put 1000 dates into the custom datafile. And from there on, it is basically like a normal reanalysis, but with the MJO phase specific dates. So, this is quite a huge project for me, given the fact that I have in general quite a busy schedule, but I hope I will complete it before the new winter season. I think it could be done in a month, but it would take some serious time dedication. Any suggestions before I start? Best regards.

Yes, all clear about that. I wasnt pointing it out as an "evil omen". Just as a fact.

I was a bit bored, so I plotted some graphics (Hovmoller diagram) about this SSW, from the NCEP/DOE Reanalysis (R-2), which has 4x daily rotating archive. Nothing special. I think it should be interesting for some. These are custom made, so for now only me and Netweather have them, and published. Of course in this "form". There might be graphics out there in a similar style of the same level and parameter. All these graphics represent zonal means. Data courtesy of NCEP. And I plotted the 10mb 60N zonal wind, for a better estimate of the actual SSW timing. And since I was even more bored, I plotted Hovmoller diagram of the 18z GFS. And 2 graphics from the Low resolution GFS 00Z. And some more plotted GFS data (yes I was bored on a Saturday evening ). Basically, the Polar vortex trying its best at a quick recovery in the upper strat. Zonal means. Best regards.

December is over, and it is time I present my January and February analog seasonal forecasts. I made these composites back in October, and first published them on Nov. 1st, on a weather forum here in my country. HERE is the link to that topic, with some extra analogs. I will also make some CFSv2 comparisons with my "forecast composites". And my analogs. The only thing that is not really logical, is that tongue of cold, extending out from E Greenland. But hey, these are composites, so I got the picture that I got. And CFSv2 forecasts one month ago. Precipitation: My analogs: CFS forecast one month ago: Pattern. The funny thing about January, is that the CFS was more easterly with the entire pattern. It has the ridge axis over USA E. coast, while my analog has it over CONUS. But given the patterns so far, it doesn't really look like the monthly reanalysis is going to verify. But there seem to be signals in favour of this analog in the future. February. Kinda grabbing onto something here, but i am not sure if having the CFSv2 backing you up is a good thing or not, given its reputation. CFS forecasts one month ago: Basically my first year of "analog forecasting". Regardless of the final verification, I think it is my obligation to share these analogs here. And keep in mind that these analogs or composites must be interpreted the same way as ensemble means. Best regards.