Let's talk ENSO

[weatherwiz]

Just now, 40/70 Benchmark said:

OMFG, I kept forgetting to select the second data set after checking the group 2 box...my god.

What sucks too is (and I don't know why this is) but after a certain amount of time, when you hit the back button going from the plot to the selection screen, it will remove the dataset 2 (even though the box is checked). So you have to uncheck the box and re-check. Not sure if this is a bug or glitch but it's annoying  

 

[40/70 Benchmark]

2 minutes ago, weatherwiz said:

What sucks too is (and I don't know why this is) but after a certain amount of time, when you hit the back button going from the plot to the selection screen, it will remove the dataset 2 (even though the box is checked). So you have to uncheck the box and re-check. Not sure if this is a bug or glitch but it's annoying  

 

That must be what happened to me.....I had the window open and went back after a few days.

[weatherwiz]

11 minutes ago, 40/70 Benchmark said:

That must be what happened to me.....I had the window open and went back after a few days.

Yup...it seems to happen after like a few minutes. It will look like this (just happened to me too)

[40/70 Benchmark]

19 minutes ago, weatherwiz said:

Yup...it seems to happen after like a few minutes. It will look like this (just happened to me too)

Great tool, but that is an obnoxious bug. Which climo period are you going to be running your composites against? I can't stand using the latest ones, even though that is the protocol.....I don't feel as though it provides an accurate portrayal because 2/3 of the composite ends up smeared in blue (it looks like winter-mode Kev made them)....especially in a climate in which cold underperforms at least excuse imaginable. I would rather my visual presentations depict a canopy of higher heights.

[weatherwiz]

27 minutes ago, 40/70 Benchmark said:

Great tool, but that is an obnoxious bug. Which climo period are you going to be running your composites against? I can't stand using the latest ones, even though that is the protocol.....I don't feel as though it provides an accurate portrayal because 2/3 of the composite ends up smeared in blue (it looks like winter-mode Kev made them)....especially in a climate in which cold underperforms at least excuse imaginable. I would rather my visual presentations depict a canopy of higher heights.

For temperatures (at least) I'm using these climate periods. The year range in parenthesis will be the years to compare to that climo period.

1871-1900 (1901-1910)

1881-1910 (1911-1920)

1891-1920 (1921-1930)

1901-1930 (1931-1940)

1911-1940 (1941-1950)

1921-1950 (1951-1960)

1931-1960 (1961-1970)

1941-1960 (1971-1980)

1951-1980 (1981-1990)

1961-1990 (1991-2000)

1971-2000 (2001-2010)

1981-2010 (2011-2020)

1991-2020 (2021-2030)

Since NCEP/NCAR reanalysis I only dates back to 1948 for years from 1904-1980 I ran against the 20th Century Reanalysis V3 and years from 1981-present I ran against the NCEP/NCAR reanalysis. The reason I started from 1981 for NCEP/NCAR reanalysis is given the dataset does not begin until 1948 the starting climo period would be 1951-1980. 

What I do want to check though is to make sure there are no big discrepancies. When I originally started I just did 20th century Reanalysis (which goes through 2015) but I noticed that yielded much different results for 2010-2011 then the NCEP/NCAR reanalysis did. 

I am unsure if I want to use this breakdown for precipitation, outgoing longwave radiation, wind anomalies, etc but I am thinking of doing so and compare periods...this way we could gauge whether or not we're seeing different response over time due to warming. 

[40/70 Benchmark]

32 minutes ago, weatherwiz said:

For temperatures (at least) I'm using these climate periods. The year range in parenthesis will be the years to compare to that climo period.

1871-1900 (1901-1910)

1881-1910 (1911-1920)

1891-1920 (1921-1930)

1901-1930 (1931-1940)

1911-1940 (1941-1950)

1921-1950 (1951-1960)

1931-1960 (1961-1970)

1941-1960 (1971-1980)

1951-1980 (1981-1990)

1961-1990 (1991-2000)

1971-2000 (2001-2010)

1981-2010 (2011-2020)

1991-2020 (2021-2030)

Since NCEP/NCAR reanalysis I only dates back to 1948 for years from 1904-1980 I ran against the 20th Century Reanalysis V3 and years from 1981-present I ran against the NCEP/NCAR reanalysis. The reason I started from 1981 for NCEP/NCAR reanalysis is given the dataset does not begin until 1948 the starting climo period would be 1951-1980. 

What I do want to check though is to make sure there are no big discrepancies. When I originally started I just did 20th century Reanalysis (which goes through 2015) but I noticed that yielded much different results for 2010-2011 then the NCEP/NCAR reanalysis did. 

I am unsure if I want to use this breakdown for precipitation, outgoing longwave radiation, wind anomalies, etc but I am thinking of doing so and compare periods...this way we could gauge whether or not we're seeing different response over time due to warming. 

I only use analogs going back to 1950....just using an 1951-2010 catch all.

[weatherwiz]

7 minutes ago, 40/70 Benchmark said:

I only use analogs going back to 1950....just using an 1951-2010 catch all.

It will be great to compare results and see if there is much of a difference say comparing 1976 to a 1941-1970 climo vs. 1951-2010 climo. 

[40/70 Benchmark]

3 minutes ago, weatherwiz said:

It will be great to compare results and see if there is much of a difference say comparing 1976 to a 1941-1970 climo vs. 1951-2010 climo. 

The result will be your composite will have overall slightly lower heights, and mine higher.

[40/70 Benchmark]

52 minutes ago, weatherwiz said:

For temperatures (at least) I'm using these climate periods. The year range in parenthesis will be the years to compare to that climo period.

1871-1900 (1901-1910)

1881-1910 (1911-1920)

1891-1920 (1921-1930)

1901-1930 (1931-1940)

1911-1940 (1941-1950)

1921-1950 (1951-1960)

1931-1960 (1961-1970)

1941-1960 (1971-1980)

1951-1980 (1981-1990)

1961-1990 (1991-2000)

1971-2000 (2001-2010)

1981-2010 (2011-2020)

1991-2020 (2021-2030)

Since NCEP/NCAR reanalysis I only dates back to 1948 for years from 1904-1980 I ran against the 20th Century Reanalysis V3 and years from 1981-present I ran against the NCEP/NCAR reanalysis. The reason I started from 1981 for NCEP/NCAR reanalysis is given the dataset does not begin until 1948 the starting climo period would be 1951-1980. 

What I do want to check though is to make sure there are no big discrepancies. When I originally started I just did 20th century Reanalysis (which goes through 2015) but I noticed that yielded much different results for 2010-2011 then the NCEP/NCAR reanalysis did. 

I am unsure if I want to use this breakdown for precipitation, outgoing longwave radiation, wind anomalies, etc but I am thinking of doing so and compare periods...this way we could gauge whether or not we're seeing different response over time due to warming. 

Paul I mean more for a composite, when you have a plethora of seasons from various decades...you have to choose one climo set, which is partly why I opt for the most all-encompassing set.

[weatherwiz]

31 minutes ago, 40/70 Benchmark said:

Paul I mean more for a composite, when you have a plethora of seasons from various decades...you have to choose one climo set, which is partly why I opt for the most all-encompassing set.

Ahh gotcha, I see what you mean. You're asking what period would I choose if I were to throw all Nina winters to get an averaged temperature output. This is a great question, I think in this case I would choose exactly what you did and go with 1951-2010 period. But since I'm also doing events prior to 1950 I would consider something like 1895-2000. 

[40/70 Benchmark]

Just now, weatherwiz said:

Ahh gotcha, I see what you mean. You're asking what period would I choose if I were to throw all Nina winters to get an averaged temperature output. This is a great question, I think in this case I would choose exactly what you did and go with 1951-2010 period. But since I'm also doing events prior to 1950 I would consider something like 1895-2000. 

I like what you did for individual seasons, though...I may adopt that.

[weatherwiz]

2 minutes ago, 40/70 Benchmark said:

I like what you did for individual seasons, though...I may adopt that.

Something that always didn't sit well with me when doing a singular composite is the uncertainty of whether the composite is being influenced by seasons which were more on the anomalous side. Now there is nothing you can really do about this, though I'm sure someone with an impeccable background in mathematics and programming could develop something. 

For example, I've been working on this tornado project (which I started like 12-years ago) which I used for senior thesis, and then finally made some significant progress with during this past spring. But when calculating averages I was always worried about an average being skewed due to outliers. So let's say I wanted to calculate the spring tornado average for the whole U.S. Naturally you would just add up the number of spring tornadoes from 1950-present and divide by the number of years. But what I did was find outliers using statistics and removed any outlier years in the calculating of the average. 

You have some Nina's that were quite cold and other's which were quite warm and at the end of the day how do these extremes influence the composite? It's with this that I think composites can be a bit misleading but that doesn't go to say there is value in them but I think the best value overall is when you're doing a composite of something that is extremely similar with little deviation (for example, creating a composite of the top 10 -NAO December's).

[40/70 Benchmark]

13 minutes ago, weatherwiz said:

Something that always didn't sit well with me when doing a singular composite is the uncertainty of whether the composite is being influenced by seasons which were more on the anomalous side. Now there is nothing you can really do about this, though I'm sure someone with an impeccable background in mathematics and programming could develop something. 

For example, I've been working on this tornado project (which I started like 12-years ago) which I used for senior thesis, and then finally made some significant progress with during this past spring. But when calculating averages I was always worried about an average being skewed due to outliers. So let's say I wanted to calculate the spring tornado average for the whole U.S. Naturally you would just add up the number of spring tornadoes from 1950-present and divide by the number of years. But what I did was find outliers using statistics and removed any outlier years in the calculating of the average. 

You have some Nina's that were quite cold and other's which were quite warm and at the end of the day how do these extremes influence the composite? It's with this that I think composites can be a bit misleading but that doesn't go to say there is value in them but I think the best value overall is when you're doing a composite of something that is extremely similar with little deviation (for example, creating a composite of the top 10 -NAO December's).

This is what I try to address by binning the ENSO events by structure...you can't explain everything because there are so many extemporaneous factors, but it does illuminate the issue somewhat.

 

Given that a strong la nina is not anticipated this season, here is our running composite of successive la nina events, which remains largely unchanged.

 

 

 

However, the composite is heavily skewed by the very anomalously mild 2011-2012, as composite is much cooler if this season is subtracted.

 

 

 

An examination of cool ENSO events by intensity confirms that the outlier of 2011-2012 heavily skews the dataset of weak to moderate successive la nina events warm.

 

Cool Neutral:

 

 

Weak La Nina: 

 

Moderate La Nina: 

Strong La Nina: 

 

Clearly there are significant differences with respect to the pattern across the northern hemisphere that are at least partially governed by the intensity of the cold ENSO event. Thus it is important to focus not necessarily on whether or not the la nina event is successive, or as we will come to learn not even solely on the intensity itself, but rather precisely what structural nuances are attributable to these differences in order to properly diagnose and forecast a la nina season. Indeed, the primary difference as it pertains to North America is with respect to the Aleutian Low and its interaction with the polar fields.

The stark contrast between the 2011-2012 season, which peaked with a marginally moderate ONI at -1.1, and the robust la nina of 2010-2011 with a peak of -1.6 ONI demonstrates this disparity quite well. What is also clear is that intensity, while undoubtedly a contributing factor, is not the only determinant, otherwise the strong la nina season of 2010-2011 would have been much milder for the eastern US than the rather modest la nina event of the 2011-2012 season.

 

 

 

 

And for those wondering, yes, the 2010-2011 la nina was the first event, and the 2011-2012 episode was the second consecutive. However, the remaining "mild second year la nina in the east" crowd should also ponder why the strong, second year la nina event of 1955-1956, which peaked with an ONI -1.7, evolved in much the same manner as the 2010-2011 event.

 

 

 

Thus far it has been illustrated that while stronger la nina events tend to be milder for the eastern US due to a flat Aleutian ridge and dearth of high latitude blocking, there are some stronger events such as 1955-1956 and 2010-2011 that seem to have a more poleward Aleutian ridge, similar to the weaker la nina composite. Likewise, there are also more modest la nina events, such as 2011-2012, that behave similarly to the stronger ONI composite in that the Aleutian ridge is very flat and the eastern US is very mild. This tends to correlate to a very active Pacific jet that inundates most of N America, save for Alaska, with mild Pacific air masses.

[40/70 Benchmark]

One thing worth noting is not all East-based Nina's are frigid, as there is variability due to the relatively weak

 ocean-atmosphere coupling. This is because many east-based events, as is the case with modoki el nino events, are 

weaker. You can have warmth, especially if the event is weaker, and one such case is 2005-2006.

Likewise, there were some weak modoki events that were fairly cold across the eastern US, such as 2000-20001.

This is further proof that there does indeed exist a structural mechanism beyond strength that dictates the overall character of la nina events and one of the focal points of this writing is to elucidate that point.

There are some "hybrid" events that share both east-based and modoki traits. You can have a Nina that is a mix of east-based versus central-based, or "basin-wide", such as both this season and last.

[weatherwiz]

Woah...that is amazing stuff! It's explained so beautifully and simply too. 

[40/70 Benchmark]

21 minutes ago, weatherwiz said:

Woah...that is amazing stuff! It's explained so beautifully and simply too. 

It goes on the show how the placement of the SST anomalies within the ENSO region and the resultant forcing schemes at least partially explains the vastly different outcomes in terms of sensible weather. 

[weatherwiz]

Just now, 40/70 Benchmark said:

It goes on the show how the placement of the SST anomalies within the ENSO region at least partially explains the vastly different outcomes in terms of sensible weather. 

Certainly does. And it paves the path for additional exploring and factoring in teleconnections (PDO, PNA, Arctic, QBO, etc). 

[40/70 Benchmark]

22 minutes ago, weatherwiz said:

Certainly does. And it paves the path for additional exploring and factoring in teleconnections (PDO, PNA, Arctic, QBO, etc). 

My mentality is to take things a bit further each successive season; building off of the conceptual framework and supporting data (composites) from prior works. I've done that over the past several years if you scrolle through the outlooks.

[It's Always Sunny]

On 9/14/2022 at 12:26 PM, 40/70 Benchmark said:

One thing worth noting is not all East-based Nina's are frigid, as there is variability due to the relatively weak

 ocean-atmosphere coupling. This is because many east-based events, as is the case with modoki el nino events, are 

weaker. You can have warmth, especially if the event is weaker, and one such case is 2005-2006.

 

Likewise, there were some weak modoki events that were fairly cold across the eastern US, such as 2000-20001.

 

This is further proof that there does indeed exist a structural mechanism beyond strength that dictates the overall character of la nina events and one of the focal points of this writing is to elucidate that point.

There are some "hybrid" events that share both east-based and modoki traits. You can have a Nina that is a mix of east-based versus central-based, or "basin-wide", such as both this season and last.

Nice posts. I've run into this observation as well while compositing everything by ENSO type, strength aside. I think(?) we can all agree that ENSO is obviously just one piece to the puzzle, and while it can hold considerable weight in "setting the baseline", that may not always be the case with other external factors in play. I do think that some ENSO phases are more reliable than others, so what I mean by that is I would put more weight in an EP La Nina playing out a certain way than say a CP La Nina as an example. Same would apply to strength.

[It's Always Sunny]

Added ENSO composites to my site based upon strength. 1991-2020 climo really does a number on them, especially washing out what you'd expect to be warm anomalies if you check them out. As an example using my ENSO composite for All La Nina, I'd be inclined to say during La Nina, southern/eastern CONUS is generally average temperature wise.

[40/70 Benchmark]

2 hours ago, It's Always Sunny said:

Nice posts. I've run into this observation as well while compositing everything by ENSO type, strength aside. I think(?) we can all agree that ENSO is obviously just one piece to the puzzle, and while it can hold considerable weight in "setting the baseline", that may not always be the case with other external factors in play. I do think that some ENSO phases are more reliable than others, so what I mean by that is I would put more weight in an EP La Nina playing out a certain way than say a CP La Nina as an example. Same would apply to strength.

Yes. Central based can take on characteristics of both east and west based events.

[40/70 Benchmark]

3 minutes ago, It's Always Sunny said:

Added ENSO composites to my site based upon strength. 1991-2020 climo really does a number on them, especially washing out what you'd expect to be warm anomalies if you check them out. As an example using my ENSO composite for All La Nina, I'd be inclined to say during La Nina, southern/eastern CONUS is generally average temperature wise.

Yea, this is why I use 1951-2010.

[It's Always Sunny]

4 minutes ago, 40/70 Benchmark said:

Yea, this is why I use 1951-2010.

It's a  mental hang-up for me to commit to 1951-2010 because temps are increasing so I almost feel more obligated to speak to the "new normal" if that makes sense. I love the 1951-2010 look because that's what I "know", however I continually remind myself that it likely isn't the case anymore, and according to new climo that's what shows b/c the atmos is warming and those departures that we're used to seeing aren't likely what they used to be back in the 50s-70s b/c we aren't living in that era anymore. That's why I like the approach I sent in that link last week where climo is essentially centered around the years of interest so you get a sample of both however the NOAA/NCEI site doesn't give you that flexibility. I mean honest question though would you feel inclined to say that during La Nina events the southern US should expect around average wintertime temps given what we're "used to" now? Curious to hear your input.

[40/70 Benchmark]

3 minutes ago, It's Always Sunny said:

It's a  mental hang-up for me to commit to 1951-2010 because temps are increasing so I almost feel more obligated to speak to the "new normal" if that makes sense. I love the 1951-2010 look because that's what I "know", however I continually remind myself that it likely isn't the case anymore, and according to new climo that's what shows b/c the atmos is warming and those departures that we're used to seeing aren't likely what they used to be back in the 50s-70s b/c we aren't living in that era anymore. That's why I like the approach I sent in that link last week where climo is essentially centered around the years of interest so you get a sample of both however the NOAA/NCEI site doesn't give you that flexibility. I mean honest question though would you feel inclined to say that during La Nina events the southern US should expect around average wintertime temps given what we're "used to" now? Curious to hear your input.

Well, like I said to Paul...I have no issue with comparing an INDIVIDUAL season to the appropriate climo period...for instance, 1978 to the 1951-1980 period, etc....or 2018 to 1981-2010. However, for a composite of seasons from a myriad of different decades, I prefer the expansive 1951-2010 set.

No, I don't expect normal temps for the south as a baseline la nina composite, however, perhaps a certain individual season will turn out that way....which is why we look at a plethora of other variables. 

[40/70 Benchmark]

On 9/14/2022 at 10:46 AM, 40/70 Benchmark said:

Great tool, but that is an obnoxious bug. Which climo period are you going to be running your composites against? I can't stand using the latest ones, even though that is the protocol.....I don't feel as though it provides an accurate portrayal because 2/3 of the composite ends up smeared in blue (it looks like winter-mode Kev made them)....especially in a climate in which cold underperforms at least excuse imaginable. I would rather my visual presentations depict a canopy of higher heights.

 

11 minutes ago, It's Always Sunny said:

It's a  mental hang-up for me to commit to 1951-2010 because temps are increasing so I almost feel more obligated to speak to the "new normal" if that makes sense. I love the 1951-2010 look because that's what I "know", however I continually remind myself that it likely isn't the case anymore, and according to new climo that's what shows b/c the atmos is warming and those departures that we're used to seeing aren't likely what they used to be back in the 50s-70s b/c we aren't living in that era anymore. That's why I like the approach I sent in that link last week where climo is essentially centered around the years of interest so you get a sample of both however the NOAA/NCEI site doesn't give you that flexibility. I mean honest question though would you feel inclined to say that during La Nina events the southern US should expect around average wintertime temps given what we're "used to" now? Curious to hear your input.

 

[It's Always Sunny]

I ended up caving and used 1951-2010 for the general ENSO composites and without a fixed scale due to the wide variability. As for the individual years I want to mull it over some more and decide how I want to approach that. 

[weatherwiz]

After going all previous corrections (which I have edited all the GIFs in my initial posts) I was able to get going with OLR Anomalies. I do like what Ray uses in velocity potential too and will do stuff with that. Just to be consistent with what I did for temps I used specific climo periods. I also used a constant legend. There were some seasons in which max anomalies were 50 but a range of -30 to 30 I thought was fine.  I'll do breakdowns to based on structure breakdown over the upcoming days. 

All La Nina Winter (DJF) Outgoing Longwave Radiation Anomalies:

Weak La Nina Winter (DJF) Outgoing Longwave Radiation Anomalies:

Moderate La Nina Winter (DJF) Outgoing Longwave Radiation Anomalies:

Strong/Super-Strong La Nina Winter (DJF) Outgoing Longwave Radiation Anomalies:

[It's Always Sunny]

Going to begin re-doing my temp/precip composites for each ENSO type (EP, CP, Basinwide, etc.) using the centered 30-year base periods. I'm trying to decide if JRA-55 or ERA-5 is best but leaning towards JRA-55 since it goes back further. The dataset for ERA-5 is only from 1979-2018 so it limits me considerably. Based on what I've read so far, NCEP is not as good as these two.

[weatherwiz]

57 minutes ago, It's Always Sunny said:

Going to begin re-doing my temp/precip composites for each ENSO type (EP, CP, Basinwide, etc.) using the centered 30-year base periods. I'm trying to decide if JRA-55 or ERA-5 is best but leaning towards JRA-55 since it goes back further. The dataset for ERA-5 is only from 1979-2018 so it limits me considerably. Based on what I've read so far, NCEP is not as good as these two.

This is a great point to keep in mind too regarding all these datasets. Results can vary significantly (even for a singular year) depending on the dataset you choose. 

[It's Always Sunny]

12 minutes ago, weatherwiz said:

This is a great point to keep in mind too regarding all these datasets. Results can vary significantly (even for a singular year) depending on the dataset you choose. 

Yeah I'm seeing MASSIVE differences using what I'm just going to call the ONI Method. The plots correlate much better with the 500 gph plots I have. I'm using the site you are using (thanks for that, btw!). Unfortunately I don't think JRA-55 or CERA-20 gives me a precipitation option so I may just have to use NCEP for those; just about to dig into that. All in all for my new composites I've used a combo of JRA-55, CERA-20 & NCEP R1.