Saturday, August 17, 2019

Preliminary 2019-2020 Winter Forecast

Hello everyone, and welcome to my Preliminary 2019-2020 Winter Forecast. Within the Word document I composed it in, this publication consumed almost 55 pages and about 50 figures, clocking in at almost 20,000 words. I believe it is the most detailed / most well-researched article I have written on this blog, and I'm very pleased to have you read it.

Certain constraints to my publication mean some of the data below are marginally out of date by approximately two to three weeks, as the forecast was prepared in late July. Where I am able to, I have made efforts to update the information, but on the whole the forecast is not affected.

For those wishing to skip the 'meat' of this forecast, feel free to scroll to the end for the forecast graphic.

Please enjoy the outlook.



1. Sea Surface Temperatures
a. El Nino – Southern Oscillation (ENSO)

Seasonal forecasting is almost always based off of the state of the El Nino – Southern Oscillation, or ENSO for short. This forecast is no different. We must first understand the concept of ENSO, however, and why we care about it.

The ENSO phenomenon, in a nutshell, is a primary driver of seasonal (and, through other shorter-term oscillations, weekly or even daily) weather patterns by way of sea surface temperature (SST) anomalies in the waters across the Equatorial Pacific. When these sea surface temperatures are above normal by a magnitude of at least 0.5 degrees Celsius, we call it an 'El Nino' event. When these anomalies are below-normal by at least 0.5 degrees Celsius, we call it a 'La Nina' event. If anomalies are positioned within the -0.5 / +0.5 degree range, the environment is called ‘ENSO-neutral’. While we monitor the entire Equatorial Pacific to analyze the ENSO phenomenon, there are four primary "zones" through which to observe, a graphic of which is shown here. They are:

·       • Nino 1+2. This is a small slice of the Pacific located between the Equator and the 10º South latitude line, extending from the far western tip of Peru to the 90º West longitude line.

·       • Nino 3. This is a larger slice of the Equatorial Pacific which spans from 5º North to 5º South latitude lines, and from 90º West to 150º West longitude lines.

·       • Nino 4. This is also a larger slice, and also extends between 5ºN and 5ºS on the latitude markers. For Nino 4, however, the space is spread by longitude from 150º West to about 160º East, crossing the dateline in the process.

·       • Nino 3.4. This is the critical area to watch, and is typically viewed as the primary space with which to assess the state of the ENSO phenomenon. Spatially, it extends from 5ºN-5ºS latitudinally, and 120º West to 165º West longitudinally.

Figure 1: The four ENSO monitoring regions (CPC)

Why do we break this space up into four different pieces rather than just average out the sea surface temperature anomalies and call it a day? A number of scientists with far more knowledge and research than I have come to determine that there can be more than one type of El Nino - where typically El Nino events bring warmer than normal waters to the eastern Pacific, an "El Nino Modoki" event brings warm waters to the western Pacific, and cooler waters to the eastern Pacific. This is not a trivial difference, but for our purposes here, we won't dive into that topic. Indeed, the probability of an El Nino Modoki event is quite low at this time, and looks to remain quite low through the winter months. For now, the key is understanding there are four different regions in which we monitor the ENSO phenomenon, with the Nino 3.4 region broadly being of most importance.

Figure 2: SST anomalies by ENSO region (CPC)
The Climate Prediction Center, the United States government agency most directly involved with seasonal forecasting, declared an ‘El Nino Advisory’ earlier this year to signal that an El Nino was ongoing. The Advisory remains in place as of this writing, and as such we remain in an El Nino state.
Let’s now view SST anomalies by each region described above.

In the Nino 1+2 region (bottom panel), sea surface temperature anomalies (SSTAs) have turned decidedly below-normal within the last month or so. This comes after a period from October 2018 through mid-March of this year, where anomalies were firmly in positive territory. From the end of March to early June, however, the trend shifted to mixed anomalies hovering around zero, and has now moved to comfortably negative anomalies. This is not a good sign for the sustainability of the ongoing El Nino, and stands as a caution flag for the other three regions that the El Nino is on a rather-shaky foundation.

The picture given by the Nino 3 region isn’t any more reassuring, with SSTAs falling into weakly-positive territory as of the most recent data reading. Indeed, the latest value sets a fresh 10-month low for sea surface temperature anomalies in that region, indicating that at the very least the ongoing El Nino has very much come off the boil.

Moving to the key Nino 3.4 region, some cooling has also been observed, but SSTAs are still firmly in positive territory and are only marginally below the +0.5 degree C threshold that typically defines an El Nino. After peaking above +1.0 degrees C at the end of 2018, anomalies in the Nino 3.4 region have gradually tapered off, and with an eroding subsurface warm pool (to be discussed later), this trend is expected to continue, placing the future of the El Nino in doubt.

]In the aggregate, the Nino 4 region has seen little change since November 2018, with latest anomalies only a bit below +1.0 degrees C. Still, the erosion of warm waters below the surface means some weakening is also possible here in coming months.

Figure 3: Observed SSTAs (CPC)

Putting all the pieces together into one yields a graphic like the one above, showing weekly SST anomalies over the Equatorial Pacific. As of July 17th, the picture was a very muddy one as far as surface anomalies go, with broad slightly-above-normal SSTs evident in the western ENSO monitoring regions, but some below-normal sea surface temperatures now appearing from the eastern portion of Nino region 3.4 all the way to the coast of Ecuador.

This is a troubling development for the state of the El Nino. Indeed, it seems as if the previously-stable El Nino has given way to a disintegrating El Nino, or perhaps even an El Nino that is about to fade out altogether. There’s still quite a bit more to discuss in this publication, but even just a cursory analysis of the state of the El Nino prompts quite a bit of alarm.

The concern over the stability of the El Nino is not assuaged when looking at water temperature anomalies below the surface, from the top of the water all the way down to 450 meters below. The top panel of the graphic below shows such anomalies, with raw temperatures displayed on the bottom panel. Looking at the anomaly panel shows why the El Nino seems to have been dissipating. To understand why, we need to learn a little bit about Equatorial Kelvin Waves.

Figure 4: Water temperatures in terms of anomalies (top panel) and in raw values (bottom panel); (CPC)
While the phrase ‘Equatorial Kelvin Waves’ seems rather daunting, it can be explained pretty simply for our purposes. From time to time, these ‘waves’ form in the far western Equatorial Pacific as surface winds, which usually blow from east to west over these waters, literally pile up water in the western Equatorial Pacific. Granted, this accumulation of water is not on a significant scale, but it is detectable on a scale of centimeters. Periodically, these surface winds weaken, which allow that piled-up warm water out in the western Equatorial Pacific to flow back to the east as the ocean tries to even itself out. This ‘wave’ of warmer-than-normal waters making its way eastward is an Equatorial Kelvin Wave, and in particular the movement of warm waters eastward is a downwelling EKW.

In sum, from time to time a ‘wave’ of warmer than normal waters pushes east across the Equatorial Pacific. To be sure, this is not a literal wave like a tsunami, but a slow-moving, expansive wave in its own right. Suppose this scenario were to happen. It leads to that body of warmer than normal waters in the central and eastern Equatorial Pacific now, since it was transported from out west. If that sounds like an El Nino, you’re right: downwelling Equatorial Kelvin Waves are identified as both triggers and enhancements to El Nino events, since they transport a body of warmer than normal waters into the ENSO monitoring regions.


On the flip side, however, the ocean is now out of balance in another respect: suddenly in the central and eastern Pacific, there’s a body of warm water. Again, the ocean tries to get itself into balance, and a body of cooler than normal waters now flows from west to east along the Equatorial Pacific. This is called an upwelling EKW, and commonly transpires in the wake of a downwelling wave. Additionally, just as a downwelling EKW can provoke or intensify an El Nino, an upwelling EKW can provoke or intensify a La Nina.

We have now ascertained that there are two types of Equatorial Kelvin Waves: a ‘downwelling’ wave, which transports warmer than normal waters from the west Pacific to the east; and an ‘upwelling’ wave, which transports cooler than normal waters from the west to the east. This may be one of those topics that is difficult to understand in words but is easier to grasp with visualizations, which are available to show these processes in the chart on the right.
Figure 5: Anomalous depth of the 20 degrees Celsius temperature line between 2 degrees N and 2 degrees S latitude (CPC)

The chart above shows the anomalous depth of the 20 degree Celsius line along the Equatorial Pacific between 2 degrees N and 2 degrees S latitude. Longitude markers are shown on the x-axis, and time is on the y-axis. Warm colors on the chart indicate that the 20 degree Celsius level is deeper down in the waters than normal, usually because the waters are warming (i.e. in an El Nino and/or downwelling EKW). Similarly, cooler colors indicate that the 20 degree C mark is closer to the surface than normal, brought about by a cooling of the subsurface (and surface) water temperatures (i.e. by way of La Nina and/or upwelling EKW). This graphic enables us to visualize the downwelling and upwelling Equatorial Kelvin Waves, which I have marked in light-blue and red, respectively.

Figure 6: Sea level anomalies (CPC)

For example, the development of a downwelling Equatorial Kelvin Wave was seen in fall 2018, highlighted with the solid cyan line, with the 20 degree Celsius temperature line moving further down below the surface as waters warmed with the EKW’s passage. One could argue a weak upwelling EKW then followed between October and December 2018, but then another downwelling EKW pushed through beginning in November 2018 and trailing off in January 2019.
It was here that the El Nino first ran into trouble. Beginning in late November, as the red line shows, a more-formidable upwelling Kelvin Wave traversed the Equatorial Pacific, briefly wiping out the positive water temperature anomalies that the previous two downwelling waves had delivered. A strong downwelling EKW came in response, however, right at the end of 2018 and only moved out of the picture in early April 2018.

The El Nino has never recovered from that point on, as another upwelling Equatorial Kelvin Wave sliced through the Pacific during the spring months of this year. Alarmingly, a downwelling EKW never really followed, with only marginal positive anomalies showing up in early June. Now, beginning in late June, another upwelling wave is on the move, putting to rest any positive anomalies and further endangering the already-fragile El Nino. It remains to be seen if another downwelling wave is on its way – the 20 degree C depth has recently begun to hint at such a development between the longitude lines of 170 E and 150 W – but unless a strong downwelling wave does develop and moves quickly, the floundering El Nino’s outlook into the fall and winter months is bleak.

I made mention of the literal ‘piling up’ of warm waters in the western Equatorial Pacific that signified the initial stages of a downwelling Kelvin Wave, and the Climate Prediction Center allows us to see this in action. On the right, sea level anomalies are shown for the most recent data available. To confirm the presence of a downwelling EKW, we would want to see positive sea level anomalies juxtaposed over the longitudes where the 20 degrees C depth is further below the surface than normal. Looking at the marginally-positive sea level anomalies between 170 W and about 150 E, and seeing the (very) marginally-deeper-than-normal 20 degree C line in the chart from the previous page over the same longitudes, it does appear that a downwelling Equatorial Kelvin Wave is building. This is significant, because it may represent a lifeline for the struggling El Nino, especially if this downwelling EKW is a strong one. Time will tell just how this situation evolves – I believe that if this downwelling EKW is not strong enough, or doesn’t even fully materialize because it’s still only now beginning to build, the El Nino will not survive into the winter.

Now, the El Nino – Southern Oscillation mechanism is not only evident in ocean temperatures. Since it affects weather patterns worldwide, it should not be a surprise that scientists have found the atmosphere to act in particular ways when an El Nino is present, and in particular ways when a La Nina is present.

Figure 7: The Walker Circulation during an El Nino (bottom), La Nina (top) and Neutral-ENSO conditions (middle); University of British Columbia

Above, an image from the University of British Columbia in Canada shows the typical atmospheric set-up during a La Nina on the top graphic, a neutral-ENSO situation in the middle graphic, and in an El Nino event on the bottom graphic. However, it may seem strange that these views of the atmosphere only cover a longitude-by-height view over the Equatorial Pacific, instead of a look at how the global atmosphere reacts to different states of the ENSO phenomenon. Why is that?
This graphic has an important purpose. As I just mentioned, the atmosphere is also a channel by which the ENSO phenomenon shows what state it is in. A primary method of determining the ENSO state is by viewing the Walker Circulation, a pattern over the Equatorial Pacific.

In general, when a La Nina is present, the Walker Circulation will see surface winds moving east-to-west across the entire extent of the Equatorial Pacific, pushing the air up on itself roughly over Darwin, Australia. This motion of air being pushed together is called convergence, and when it occurs at the surface the air has nowhere to go but up, thus creating convection. Thus, the air travels east-to-west across the ocean and is transported to the top of the troposphere by way of convection. When the warm surface air parcel cools at those high altitudes to the point that the air parcel is the same temperature as the air around it, the air parcel stops rising: remember, an air parcel will only rise on its own if it is warmer than the environment around it. However, the air still has to go somewhere: the convective process continues, and there are more air parcels still moving up via convection. Since this original air parcel can no longer rise, and can’t sink back down right where it is, it must spread out. The motion of air spreading out from a point is called divergence. We see this in the La Nina graphic as the green arrow now switching directions and pulling the air parcel from west-to-east, back towards South America and the eastern reaches of the Equatorial Pacific. Briefly, recall at the start how our air parcel was being transported from the east Pacific to the west at the surface. That air has to come from somewhere: it can’t just appear out of nothing! It’s convenient, then, that the air traveling west-to-east high up in the atmosphere starts to slow down when it reaches the eastern Equatorial Pacific, and actually begins to descend back to the surface. This air descending over the eastern Equatorial Pacific (called subsidence) makes for sunny skies and calm weather – after all, convection (rising air) can’t physically occur in an area of subsidence (sinking air). This then completes the Walker Circulation.

You might be thinking ‘Okay, that’s great and all, but why does it matter to finding out if the atmosphere is in an El Nino or La Nina state?’ I can assure you this leads to a revelation – wasting time is only a negative for everyone involved! Take a look at the Neutral-ENSO and La Nina graphics of the Walker Circulation. You’ll notice that while the premise of a circulation remains intact for the other two ENSO states, the location of the convection and subsidence, and even the direction of winds both at the surface and aloft, changes as the ENSO state changes. In a Neutral-ENSO scenario, for example, the circulation remains intact as in a La Nina, but now the area of surface convergence has moved offshore of Australia, somewhere between Darwin and Tahiti.

Perhaps the largest change is in an El Nino Walker Circulation. Initially things are the same as a La Nina, with surface winds going from east-to-west from the eastern Equatorial Pacific. However, rather than surface convergence occurring over Darwin, it now occurs over Tahiti. That’s a pretty big change – what happened? Let’s go back to the concept of divergence. Remember how I said that air high up in the troposphere, transported there from the surface by convection, had to spread out because it could no longer rise and had no room to immediately sink? It spreads out in all directions, not just to the east like the La Nina panel might lead you to believe. So, the air from the convection over Tahiti spreads out both to the east and the west. When it spreads out west, it then sinks somewhere over Darwin, and then starts traveling along the surface again – but this time, from west-to-east. Now we’ve got surface winds in the eastern Pacific going to the west, and surface winds in the western Pacific going to the east. That makes a lot of air pile up all in one place. You can probably guess what happens then – a lot of convection right over Tahiti, where the surface winds collide and create surface convergence.

Figure 8: Surface vector winds averaged from July 15th through July 21st over the Tropical Pacific (ESRL)

Now that we’ve identified how the Walker Circulation should appear in each of the three stages of the ENSO phenomenon, let’s take a look at recent atmospheric conditions to see if the atmosphere is acting as if the ENSO phenomenon is in a different state than an El Nino, or if it is affirming the El Nino. We will use 200-millibar winds to judge wind direction high up in the troposphere, surface winds to judge the surface winds, and outgoing longwave radiation (OLR) to examine the presence of convection. I will explain the concept of OLR when we come upon it.

We begin with a seven-day average of surface winds, as shown in Figure 8 above. The purpose of using a seven-day average as opposed to a monthly average is to make sure any sudden change in wind pattern that may have occurred in only a few days out of the month won’t be glossed over, just to give an example. Analyzing Figure 8 shows a pretty remarkable testimonial to the presence of a La Nina, instead of an El Nino. Indeed, surface wind vectors are seen pointing from the east to the west along almost the entirety of the Equatorial Pacific, with the vector arrows converging around the 160 degree E line of longitude. While this isn’t the same as convergence over the same longitude as Darwin, Australia, it is remarkably close, and the firmly-westward winds at the surface imply the presence of a La Nina… at least at the surface. Figure 9 looks at vector winds at the 200 millibar level, about the level of the atmosphere where the jet stream resides here in the U.S..

Figure 9: 200 millibar wind vectors averaged from July 15th through July 21st (ESRL)

The interpretation of 200-millibar winds is more mixed than the surface. Whereas surface winds showed a rather-convincing portrayal of La Nina conditions, upper-level winds actually seem more aligned with an El Nino, particularly given the westward winds along the Equatorial Pacific from about 160 degrees West longitude into Papua New Guinea. This is a feature most commonly seen in El Nino events, typically produced by convection over Tahiti creating divergence aloft. However, we do not see a corresponding band of eastward upper-level winds to the east of that 160 E line of longitude, which would have really cemented the case for El Nino conditions. Instead, there are a couple of areas of circulation along the Equator, which does not help our interpretation of how ENSO is affecting the atmosphere.

Figure 10: Outgoing Longwave Radiation anomalies from July 15th through the 21st (ESRL)

To try and find more-concrete evidence, Figure 10 takes a look at outgoing longwave radiation (OLR). While the actual definition and mechanics behind it are slightly more complex, for our purposes we can simplify OLR to know that negative values indicate the presence of enhanced convection, while positive values indicate the presence of suppressed convection. Please note the (admittedly) strange color scale, where negative (positive) OLR values are shaded in warm (cool) colors.

Figure 10 seems to be the tiebreaker between surface winds (which favored a La Nina) and upper-level winds (which seem to lean in favor of El Nino conditions), in favor of an El Nino. The July 15th – July 21st OLR averages show enhanced convection broadly in the vicinity of Tahiti, with subsidence north of Darwin near Papua New Guinea as well as just northwest of Ecuador. The fact that these areas of rising and sinking motion are not along the same line of latitude makes the idea that this seems to portray El Nino conditions a little more fragile, but we will press ahead with it anyway. With convection over/near Tahiti and suppressed convection in the general area of Darwin and Peru/Ecuador, the atmosphere seems reflective of a weak El Nino, which goes along with what sea surface temperatures indicate.

Up to this point, through sea surface temperature anomalies, subsurface water temperature anomalies and a look at the atmosphere, we have determined there is a weak El Nino in place. However, the official source for what state the ENSO phenomenon is in is the National Oceanic & Atmospheric Administration (NOAA). As of July 11th, NOAA’s Climate Prediction Center (CPC) branch determined that an El Nino Advisory was warranted, indicating that an El Nino is currently ongoing. But, given how fragile the El Nino is looking, what does the forecast hold?

Figure 11: Multi-model ensemble forecast of SSTAs in the Nino 3.4 region through January 2020 (CPC).

Shown above is a combination of climate models’ forecasts for the sea surface temperature anomaly in Nino region 3.4 from July 2019 through January 2020. As a brief reminder, El Nino’s are technically in place when the anomaly is at or above +0.5 degrees C, while La Nina’s exhibit anomalies at or below -0.5 degrees C. Neutral-ENSO conditions are defined when anomalies are within that +/- 0.5 degree range.

Within the graphic as shown in Figure 11, these models agree on the Nino 3.4 region continuing to cool down, albeit to varying degrees, through the fall months. By September, all models included have the Nino 3.4 region in Neutral-ENSO territory, with one model even predicting a slight La Nina in that month. The trend then shifts back upward heading into the late fall months, but a recovery in the El Nino currently seems unlikely, with the average of these models still staying in neutral-ENSO territory.

This solution is anticipated by the NOAA as well, with their most recent El Nino Advisory cautioning that a decline to neutral-ENSO conditions was expected in coming months, and was likely to persist into the fall and winter months.

If we should expect to be in an ENSO-neutral situation this winter, what can we expect? Figures 12 and 13 below provide some context.
Figure 12: DJF Temperature Anomalies in Neutral-ENSO situations (ESRL)

Figure 13: DJF Precipitation Anomalies in Neutral-ENSO situations (ESRL)
Figure 12 above shows temperature anomalies in the December-January-February period when the ENSO phenomenon was in its neutral state. I identified neutral-ENSO winters by using the Climate Prediction Center’s Oceanic Nino Index (ONI), and included almost all neutral-ENSO winters since 1950. Figure 12 shows a pretty one-sided view, with exclusively below-normal temperature anomalies evident in neutral-ENSO winters, maximized in the northern Plains and upper Midwest, with near-average temperatures in the Southwest and far Southeast. It seems as though a neutral-ENSO winter entails cold air in Canada being rather prone to ejecting south to lower latitudes in the United States, and could be something to watch closely.

Figure 13 analyzes precipitation anomalies over the same three-month window, using the same years as in Figure 12. Here, the pattern is oriented in a pretty interesting middle ground between what one typically sees in a La Nina vs an El Nino. Whereas a La Nina brings wetter than normal conditions to the Midwest and Ohio Valley regions, and an El Nino brings dry weather to those same areas with wet weather along the East Coast, Neutral-ENSO winters seem to strike a compromise, placing modest below-normal precipitation anomalies in portions of the Midwest and Plains while keeping wetter anomalies in the eastern portions of the Ohio Valley towards the Gulf Coast. From all this, it seems like a Neutral-ENSO winter may very well bring about colder than normal conditions for much of the country, while a snowier than normal winter could be in store for the Appalachia / Ohio Valley region.

What does the broader atmospheric pattern look like in neutral-ENSO conditions? I’m glad you asked!

Figure 14: DJF 500-millibar geopotential height anomalies in Neutral ENSO situations (ESRL)
Figure 14, above, shows 500-millibar geopotential height anomalies over the Northern Hemisphere in the same years as outlined in Figures 12 and 13, as well as over the same three-month window. Recall that negative values / cool shadings in this graphic represent negative anomalies (therefore colder and stormier weather), while positive values / warm shadings represent positive anomalies (therefore warmer and calmer weather). Per the figure, neutral-ENSO winters have historically featured high pressure ridges in the Bering Sea, as well as one positioned squarely over Greenland. A ridge positioned over the Bering Sea is identified as the negative phase of the West Pacific Oscillation (WPO), which results in cooler than normal weather over the eastern two-thirds of the United States. Similarly, a ridge over Greenland signals the negative phase of the North Atlantic Oscillation (NAO), which tends to buckle the jet stream south and provide colder weather for the Central and (especially) East U.S. That jet stream buckling also gives credibility to the above-normal precipitation anomalies in Appalachia / the Ohio Valley, as the -NAO is what tends to be the key factor to ignite big storms in the Eastern U.S., including Nor’easters. 

What is seen preventing those Nor’easters and pushing the above-normal precipitation track further inland as opposed to right along the East Coast is the modest ridge just offshore the Southeast, which acts as a diverting mechanism for storms traversing the Southern Plains, forcing them northeast into the Ohio Valley instead of due east towards Georgia and Alabama. Also of note in this graphic is how the -WPO ridge and -NAO ridge are strong enough to force the tropospheric polar vortex to lower latitudes. We will discuss this more later on, but it is critical to know that there are essentially two ‘versions’ of the polar vortex: the stratospheric polar vortex, and the tropospheric polar vortex. Of course, when placed on a 3D scale they are one in the same, but given that we look at slices of the atmosphere rather than 3D graphics, it is better for us to think of two polar vortices as opposed to one. Again, this will be elaborated on later in the Stratosphere section.

In Figure 14, it seems as though a Neutral-ENSO state will encourage the tropospheric polar vortex to split up and be forced to lower latitudes as a result of the two aforementioned ridges. Figure 15, below, illustrates what happens to the stratospheric polar vortex, at about the 50-millibar level. At the stratospheric level, it appears as though an ENSO-neutral situation provokes strong ridging over the Arctic Circle, forcing the polar vortex to lower latitudes at a weakened strength. As a consequence, it becomes far easier for below-normal temperature anomalies to flow to lower latitudes over the course of the winter. It also becomes far easier to achieve a sudden stratospheric warming (SSW), which will be discussed in detail in the Stratosphere section. For now, however, understanding that a neutral-ENSO state appears to materially disrupt both the tropospheric and stratospheric polar vortices is a key takeaway.

Figure 15: DJF 50-millibar geopotential height anomalies in Neutral-ENSO situations (ESRL)
In summary, we are currently in a weak El Nino environment, as per both sea surface temperatures and signals from the atmospheric pattern. However, forecast models anticipate the El Nino will weaken into the fall, and neutral-ENSO conditions appear probable for this winter season. In the event of a neutral-ENSO winter, it appears cooler than normal temperatures are more likely than warmer temperatures, with the coldest anomalies most likely to be maximized in the upper Midwest and northern Plains. Enhanced snowfall could hit the Ohio Valley and Appalachia regions. In addition, a neutral-ENSO winter may set up a pattern very favorable for intense cold across wide swaths of the country (mainly in the eastern two-thirds) as disruptions to the tropospheric and stratospheric polar vortices are seen as more likely.



     b. Pacific Decadal Oscillation (PDO)

There are a number of ocean-based oscillations that are highly relevant to seasonal forecasting, and one could argue that the Pacific Decadal Oscillation (PDO) is the second-most-important for North America, after ENSO. As a result, before we dive in to see what the PDO is doing, we need to understand what the PDO actually is.

Figure 16: Screenshot from the University of Washington's PDO page showing SST anomalies (shaded) and surface winds (arrows) in a positive PDO (left panel) and negative PDO (right panel) (UW)
Figure 16, displayed on the right, provides a look at the Pacific Ocean during the positive phase of the PDO (left panel) and the negative phase of the PDO (right panel). Note that, unlike ENSO, there is no neutral phase. In a positive PDO, sea surface temperatures along the coast in the Gulf of Alaska down to the coast of California are above normal, with positive SSTAs extending into the Equatorial Pacific. In contrast, below-normal SSTAs exist from the well-offshore reaches of the Gulf of Alaska westward to Japan. Aside from sea surface temperatures, surface winds in a positive PDO are generally eastward from Japan into the Gulf of Alaska, with northward vectors along the coastline of North America.

You might draw out that those warmer than normal anomalies in the Equatorial Pacific during a -PDO kind of look like an El Nino. That’s not a mistake: according to the Earth System Research Laboratory (ESRL), the PDO and Nino 3.4 region’s SST anomalies have a correlation of 0.42.
In a negative PDO, the script is flipped, with below-normal SST anomalies seen along the western coast of North America into the Equatorial Pacific, and warmer than normal waters deep in the Gulf of Alaska out towards Japan. With regards to surface winds, westward flow is seen from the Gulf of Alaska towards East Asia, while surface winds are southbound just offshore western Canada and western U.S. We’ll begin with recent SST anomalies in Figure 17.

Figure 17: Recent SST Anomalies over the entire globe. This section will focus on the Northern Pacific. (NOAA)

As shown above, while there are a few areas of significant anomalies across the northern Pacific, there isn’t really any textbook pattern that quickly tells us what phase the PDO is in. On one hand, below-normal SSTAs west of Baja California that try and curve southwestward near the -130 degree line of longitude would typically signal a -PDO. However, such below-normal anomalies don’t extend into the waters just west of Canada and into the shoreside Gulf of Alaska region. On the other hand, well below normal sea surface temperature anomalies northeast of Japan are typical of a +PDO, but they don’t extend far enough east to really indicate to me that it’s a “real” positive PDO. Instead, it just seems like a really messy amalgamation of SST anomalies.

Although this is messy, that doesn’t mean these anomalies aren’t worth looking at in their own right – quite to the contrary, in fact. In the Regions of Interest section of this Sea Surface Temperatures portion, we will go over the northeast Pacific and Gulf of Alaska and investigate why the placement of certain anomalies can prove quite pivotal to how the winter turns out, even without accounting for the state of the Pacific Decadal Oscillation. That will come later on, however – for now, the focus is on deciphering what the PDO is up to.

We did note that surface winds over the north-central Pacific and along the western coast of North America can be reliable in determining the PDO phase. To try and get a sense for it, Figure 18 shows surface vector wind anomalies from June 1st through July 21st in the northern Pacific. You will notice this is a wider timeframe than the images we viewed as part of the ENSO section. This is deliberate: unlike over the Equatorial Pacific, the northern Pacific regularly experiences significant low pressure systems that trek across and around the Aleutian Islands, easily distorting a 7-day view of surface winds if a storm system happens to be moving through at that particular time. To try and assuage these distortive effects, the timeframe is expanded to just under two months.

Figure 18: Surface wind vector anomalies over the north Pacific from June 21st through July 21st (ESRL)

Viewing Figure 18 above, we are able to finally see something that’s more concrete in pointing towards a specific PDO phase. With surface winds anomalously pushing to the west from the Gulf of Alaska almost all the way to Japan (though admittedly not nearly as strong when they get west of the Aleutian Islands), and winds forcing their way from north to south just offshore the western coast of North America, surface winds seem representative of a negative PDO. To be sure, we can’t yet confirm a negative PDO, given the conflicting SST anomalies above, but it appears the atmosphere is leaning in the direction of a -PDO state.

One could argue the SSTs can make the case for a -PDO stronger when viewing the negative SSTAs west of Baja California and positive SSTAs in a horseshoe-style pattern in the Gulf of Alaska, but that pool of deep negative anomalies east of Japan is too significant for me to be comfortable calling this a -PDO event.

Figure 19: Pacific Decadal Oscillation since late 2009 (NOAA/NCDC)

The deciphering process can be made somewhat easier by the presence of the Pacific Decadal Oscillation index, calculated at NOAA. A nearly-decade-long view of the PDO index is shown in Figure 19 on the right. After undergoing a prolonged and strong negative spell from early 2010 through early 2014, the PDO then turned positive for the following two years. Since about late 2016, however, the PDO index has been mixed, with a tendency to lean negative in the last year or so. On a month-by-month level, the PDO was seen modestly positive in April and May of this year, after having been negative since February 2018, and clocked in at -0.00 in June. That’s not a typo, and tells us what we’ve already ascertained: SST anomalies aren’t clear enough to determine which phase the oscillation is in, and what strength that would be (note that the PDO index is based solely off of SSTs and not surface winds).
The last item to look at and try to determine the PDO phase is the tendency of sea surface temperature anomalies over the last year, as seen in Figure 20.

Figure 20: June 2019 SST anomalies (top) and change from June 2018 (bottom), (CPC)


Figure 20 is a two-panel graphic, showing sea surface temperature anomalies from June 2019 on top and the change in SST anomalies from June 2018 to June 2019 on the bottom. Since we have more-recent SST anomaly data available in Figure 17, we will only pay attention to the bottom panel. When glancing over the SST anomaly change from a year earlier, it’s quickly apparent that the trend has been towards a negative PDO. That does not mean things have transitioned into a -PDO; as we found out earlier, SST anomalies are too messy to really determine a firm phase for right now. Instead, the trend has been for SST anomalies to go away from a comfortably-positive PDO state into at least a non-existent state, as evidenced by the -0.00 figure in the PDO index.

In other words, while we don’t appear to be in a negative PDO with certainty, the trend in sea surface temperatures has certainly been in that direction, and surface winds appear to be of the same opinion. Further, this isn’t just a year-long trend: it has continued in the short-term, with SSTs cooling further west of Baja California, cooling south of the Aleutian Islands, and warming in the rest of the Gulf of Alaska, all symptoms of a negative PDO event if the trends continue.

Given that the PDO technically is not strong enough to have a specific phase (via the PDO index), and given how all these negative PDO signals are only drawn from trends rather than actual observed data (i.e. prominent negative SSTAs as opposed to anomalies that are only trending from warm to average), I will go forward in this PDO section and explain what each PDO phase could result in, while not favoring any specific phase for this winter. Instead, I will go over what each phase entails, and in the coming weeks and months I will continue reviewing data to eventually make a determination as to what the PDO is doing. The goal of this format is so that when this winter’s PDO phase is eventually determined, you will have already learned beforehand what each PDO phase is capable of, and thus there should be far less confusion as to how that phase will affect the country.
I am hard-pressed to imagine that this winter’s PDO phase is not determined by the time my 2019-2020 Official Winter Forecast is prepared in October, and the intention is to effectively use the PDO in the forecast then instead of laying out the different possibilities, as I will have to do here.

The graphics below depict the correlations between the PDO and temperature (Figure 21) as well as between the PDO and precipitation (Figure 22) during the December-January-February window. Positive readings in these images indicate that the PDO is positively correlated with either temperature anomalies or precipitation anomalies in that given area, depending on the figure being viewed, while negative readings imply negative correlations.

Figure 21: DJF Correlation between PDO and Temperature (ESRL)
Figure 21 above depicts the correlation between the PDO index and temperatures in the DJF period, as described. One of the most noticeable features is the high positive correlation observed along the western coast of the United States, exceeding 0.50. This suggests a pretty reliable signal that a positive PDO event will provoke warmer than normal temperatures in the West during the winter months, while a negative PDO event will entice cooler than normal temperatures. An opposite picture is seen in the eastern half of the country, with negative correlation values deeper than -0.4 seen along the Gulf Coast. In other words, if the PDO is positive this winter, cooler than normal temperatures seem more plausible in the Southeast, and vice-versa. In the Plains and Midwest up into the Northeast, though, correlation values are situated around 0.00, suggesting that no matter how the PDO turns out this winter, it will be difficult to ascertain temperature anomalies based solely on this oscillation.

Figure 22: DJF Correlation between the PDO and precipitation (ESRL)

Turning to precipitation, Figure 22 provides a look at the correlation between the PDO index and precipitation anomalies, once again during the DJF time period.

In general, precipitation anomalies are most likely to be negatively correlated to the PDO, with correlation values deeper than -0.5 in portions of Kentucky and southern Indiana. As such, if a negative PDO were to evolve for this winter, a snowier than normal cold season could be in store for the Ohio Valley into the Midwest. In contrast, notable positive correlation values are evident in the Southeast, as well as in parts of the southern Plains. For those regions, a positive PDO could present an opportunity for above-normal precipitation this winter.

In summary, the recently-positive PDO has weakened to a level too weak to accurately discern at this time. While the PDO phase should be identifiable in time for the 2019-2020 Official Winter Forecast, it seems most prudent for now to lay out the different impacts each phase of the PDO can have, and then determine the phase itself in the next outlook. This is in the best interest of maintaining the forecast’s integrity and not putting the cart before the horse, so to say.



         c. Atlantic Multi-Decadal Oscillation (AMO)

Rounding out the top three seasonal ocean-based oscillations most relevant to North American weather is the Atlantic Multi-Decadal Oscillation, or AMO. As the name implies, this is a very long-range oscillation, and changes phases on a scale of decades (note the plural), as opposed to years in the case of the PDO, or even months in the case of ENSO. Note that this change in phases refers to a change in the overarching state – it is not uncommon for the AMO to turn negative for a short period even though it may be in the middle of a multi-decade positive state, for example.

Figure 23: Long-term view of the AMO (ESRL)

To visualize the extremely-long-term nature of the AMO, Figure 23 shows a 71-year history of the oscillation. Similar to the PDO, the AMO has only two phases: positive and negative. Readings of 0.00 imply the AMO simply does not have a discernable phase; there is no ‘neutral’ phase.
From the beginning of the dataset in 1948 to roughly 1962, the AMO was in its positive phase, with the heavy majority of data points in that span above zero. From about 1962 to about 1995, the AMO was in a comfortably-negative phase, which then reversed into a positive phase that runs from 1995 to present-day. While there were plenty of situations where the AMO briefly spiked into positive territory during its negative phase or dipped into negative territory amidst its positive phase, it is crucial to know what the ‘real’, overarching state of the AMO is. For example, even though the index recently dipped into negative territory, the index remains firmly entrenched in a positive phase over the longer-term. As a result, even though the index recently was negative, we will assume a positive AMO for the coming winter, given that is the ongoing overarching state.

Before moving on to what a positive AMO means during the winter months, I want to briefly address why I believe it is prudent to assume a +AMO, even though recent data is hinting at some weakening in the overall +AMO regime. Figure 24, below, shows a short-term view of the same dataset as in Figure 23, from 2010 to present day.

Figure 24: Short-term view of the AMO, since 2010. Please note that the chart title is incorrect, this is the Atlantic Multi-decadal Oscillation, not Atlantic "Meriodional" Oscillation. (ESRL)

Figure 24 provides a very good look at what I mean when talking about these aberrations of brief negative values during a positive AMO regime or vice-versa. Indeed, since 2010, the AMO turned negative five separate times, though the overwhelming tendency to stay in positive territory shows how the “overarching” and “real” state of the AMO is positive. As the name implies, this is a multi-decadal oscillation, meaning any change in the index should be able to be seen well in advance (i.e. by a year, perhaps more). 

Following the most recent departure into negative territory, at the tail end of 2018, the AMO has been seen moving back firmly into positive territory, now flirting with the highest values in more than a year. This adds credibility to the idea that this coming winter should see a positive AMO. Unfortunately, that credibility is undermined by the incorrect naming of this chart, and the fact that “meridional” is misspelled. Nevertheless, the chart data itself is correct upon examination.

But there’s an issue here – we haven’t gone over how the positive AMO is identified, we have only established that it’s positive. I try to make forecasting a learning experience in addition to a reading experience, and if this publication doesn’t explain how to identify the phase of the AMO, that’s a problem. Refer back to Figure 17 for a moment, showing sea surface temperature anomalies around the globe. More specifically, refer to Figure 25 below, where I have zoomed in on the north Atlantic.


Figure 25: Zoomed screenshot from Figure 17, SSTAs over the North Atlantic (NOAA)
The Atlantic Multidecadal Oscillation comes from the Atlantic, as the name implies, but more specifically we look to the north Atlantic to determine the behavior of the AMO. In a positive AMO, sea surface temperatures are warmer than normal in and around Greenland. Given that we are in a positive AMO state, it is thus no surprise to see widespread and noteworthy positive SSTAs across those waters. If this were a negative AMO environment, one would expect to see below-normal water temperatures around Greenland.

Thus far, this section has illustrated how to determine the AMO phase, discussed what phase of the AMO we are currently in, and pointed out why it’s rather easy to forecast the AMO for a number of months in advance, especially if the oscillation is firmly entrenched in either its positive or negative phase. Now that it’s evident we are in a +AMO regime and will remain in such a regime into the winter, we can see what a positive AMO usually brings for North America in the winter season. For this, I will be using correlation graphics for temperature and precipitation, like what was done in the PDO section. First, though, we need to get a sense for positive AMO winters in general via correlation graphics.

Figure 26: DJF Temperature correlation with the AMO (ESRL)

Figure 26, shown above, depicts the correlation between surface temperatures and the Atlantic Multidecadal Oscillation (AMO) during the December-January-February window. This graphic indicates that while the majority of the country doesn’t see a significant correlation between the AMO and temperatures during the winter months, there are a few areas worth mentioning.

Of note is the maximization of positive correlation values in the state of Maine, where the positive correlation magnitude exceeds +0.40. This implies that in a positive (negative) AMO winter, temperatures in the far Northeast will be warmer (colder) than normal. This positive correlation drops off immediately after exiting Maine, but is a still-notable +0.30 to +0.399 in portions of New England. A modest positive correlation between the AMO and temperatures is evident in the Ohio Valley, Midwest, Great Lakes and Southern Plains regions, while the remainder of the country generally sees a minimal correlation between the AMO and temperatures. Given that we expect a +AMO this winter, the chances for a warmer than normal winter in New England and perhaps into the Midwest/Great Lakes/Plains appear higher.

Figure 27: DJF Precipitation correlation with the AMO (ESRL)

Just above, Figure 27 depicts the correlation between precipitation and the AMO, again during the three winter months. Again, while the majority of correlation values are rather insignificant, there are a few areas to make note of. Primarily of interest here is the positive correlation maximized in the western U.S., especially in northern California. With a +AMO state expected, this would seem to increase the chances of a wetter than normal winter in the West. That in itself would seem to imply a more active than normal storm track for the winter, perhaps a reason why we also see a positive correlation in the Ohio Valley region. This opens the door to the possibility of an active winter storm season, with the primary storm tracks running northeast through the Plains (delivering above-normal precipitation to the Nebraska/South Dakota/Iowa region) and another bowling through the Ohio Valley (bringing wetter than normal anomalies to that region). That possibility would be far higher if the AMO were the only seasonal oscillation with a material effect on North America and if those correlations were stronger, but neither of those is the case. Therefore, the +AMO seems to merely encourage warmer temperatures in the Central and Northeast U.S., but perhaps with wetter than normal conditions in the central Plains and Ohio Valley. Whether this transpires, with numerous other oscillations also in the picture, remains to be seen.

On a final note, it is worth mentioning that the Earth System Research Laboratory (ESRL) has found there is a negative correlation between the PDO and AMO across all months of the year, on the order of -0.28. While this isn’t a significant correlation, it could be of use here as we know the AMO phase but not the PDO phase for the coming winter. If the AMO is presumed in this forecast to be positive for the winter season, this negative correlation would imply that the PDO ought to lean negative, bolstering the case for a negative PDO that was broadly made in the PDO section but not formally proposed as the likely PDO state for the coming winter. On a month-by-month basis, the AMO-PDO correlation for December, January and February is -0.27, -0.25 and -0.32, respectively, good for a three-month average of -0.28 on the dot. In other words, while this adds to the case for a negative PDO this winter, we still have yet to find concrete evidence supporting a formal forecast of a certain PDO phase. That will likely come in the Official winter forecast in late September.

In summary, the Atlantic Multidecadal Oscillation has moved back into positive territory after briefly going negative last winter, continuing the overarching state of a positive AMO environment. The AMO is expected to remain positive for the 2019-2020 winter, with well-above-normal sea surface temperature anomalies evident around Greenland. As a consequence, given positive correlations with temperature across swaths of the Central and Eastern U.S., the possibility of a warmer than normal winter appears to be on the rise. However, the possibility of stormier than normal weather in the Southwest, and snowier than normal conditions in the central Plains and Ohio Valley, also appears to be on the increase should the +AMO persist as expected.




                  d. Regions of Interest
In addition to the three primary oscillations discussed above, there are five areas I monitor that I believe also affect the winter pattern to a non-trivial degree. Figure 28 below, showing weekly SST anomalies across the globe for the period ended July 20th, will serve as the reference graphic for the first three of five regions. These regions, in order of presentation, are: East Asia / West Pacific; Bering Sea / North Pacific; Gulf of Alaska / Northeast Pacific; the northern Atlantic; and the Great Lakes.

Figure 28: SST anomalies for the week ended July 20th (ESRL)

                  i. East Asia / West Pacific
The East Asian / West Pacific region is of high interest to forecasters like myself who enjoy making use of rather-unconventional tools that have, so far, proven to be accurate. One of these tools is the concept that anomalous weather events over the United States can be predicted with pretty high accuracy by viewing those same anomalies over the region of Japan about 6-10 days prior to that event’s occurrence in the United States. While this is not a widely-used forecasting method due to its relative lack of research and relative lack of prominence, I believe it to be important enough to include a brief discussion over the sea surface temperature anomalies in that region.

Looking over Figure 28 and examining the Western Pacific, the picture is rather mixed. While there are pockets of below-normal SSTAs in the waters surrounding southern Japan, as well as due east of the country, there is also a swath of above-normal anomalies just northeast of Japan, with essentially-neutral anomalies in between. Ideally, the predominant presence of above-normal or below-normal sea surface temperature anomalies would allow us to determine how the weather will play out for that area over the coming winter. However, anomalies as weak and spatially-minimized as these don’t afford us that opportunity. At best, an argument could be made that below-normal sea surface temperature anomalies south of Japan favor a stormier-than-normal winter ahead for East Asia, and thereby favor a stormier-than-normal winter in the United States as well. However, that’s a bit of a stretch in my eyes. Instead, it seems most appropriate to acknowledge that there is not yet a clear signal for SST anomalies around Japan, precluding us from determining how the winter there (and by extension, in the U.S.) will play out. Similar to the PDO, I expect the picture to become clearer for the Official winter forecast in a couple of months.


                  ii. Bering Sea / North Pacifc
Another area of interest to this forecast is the waters in and around the Bering Sea, extending into the north-central Pacific in general. This holds special importance in my opinion given how critical the Bering Sea is for the stratosphere during the winter, thus making it a critical factor in the North American winter pattern as well. To understand why this is such an important piece, we need to jump ahead a bit and glean some content from the Stratosphere section of this outlook.
Undoubtedly, if you have any interest in the broad weather pattern during the winter months, you’ve heard of the infamous “polar vortex” at some point or another. The idea that the polar vortex brings severe cold to lower latitudes is correct, but its description has been mangled in the media. Contrary to what some articles will imply, the polar vortex does not “return” or “happen again” – it is a permanent feature in the Arctic Circle during the winter months, and is present every winter. Additionally, the polar vortex is not an “event” – what happens in those cases is that some sort of disturbance in the upper latitudes forces a piece of the tropospheric polar vortex to break off from the main vortex. That smaller piece is then shunted down to lower latitudes, which can (and does) move over northern portions of the United States, bringing intense cold with it. So, if you hear someone talking about “the polar vortex happening again this winter”, make sure you remind yourself that the polar vortex happens *every* winter (indeed, without it, there wouldn’t be a winter as we know it) and what the person should really be asking is if something will happen to make a piece of the polar vortex split off and move to lower latitudes.

Now that I’ve gotten that off my chest, we need to turn to the question of ‘what causes these splits in the polar vortex?’. You might also be wondering how the polar vortex can be present every single winter, even though it only seems to “happen” once every several years in the United States. Both of these are good questions. We’ll use the answer to the second question to also answer the first question.

In the winter, there is a large vortex spanning the entirety of the atmosphere relevant to weather forecasting, from the surface to the upper reaches of the stratosphere. This is the polar vortex, but it acts differently at different levels of the atmosphere. The polar vortex we’re most familiar with is what I refer to as the tropospheric polar vortex. In reality, the atmosphere is three-dimensional, so this is one big vortex, but for our purposes it’s necessary to separate them into two. The ‘other’ one is the stratospheric polar vortex. When the news talks about the polar vortex moving south into our neck of the woods, what they’re really talking about is a piece of the tropospheric polar vortex becoming detached from the main vortex and managing to be moved to lower latitudes over North America. In practice, it can be shunted to any lower latitude, including over Europe, Asia, Siberia, or the oceans. This movement of the polar vortex to lower latitudes happens quite often, but since there’s so much space on this planet and North America is only so large, the stars align only every once in a while for that piece of the tropospheric polar vortex to impact the U.S. What you won’t hear on the news, though, is how that split / disruption in the tropospheric polar vortex happened in the first place. Almost exclusively, especially in the event of a major disruption to the tropospheric polar vortex, this process begins in the stratospheric polar vortex. Usually, that involves a sudden and intense event of warm air rushing into the stratosphere and ravaging the stratospheric polar vortex, which eventually feeds down to the tropospheric polar vortex (usually with a ~2 week lag). Now, we have to identify where those sudden warm air bursts (called Sudden Stratospheric Warmings, or SSWs) come from, and that’s where the Bering Sea comes into play.

For reasons I’m not entirely familiar with, the Bering Sea is an incredibly fertile area for these sudden stratospheric warming events to happen. A ‘typical’ SSW event involves a ridge of high pressure and warm air blossoming over the Bering Sea and then flooding into the Arctic Circle. While this blossoming into the Arctic Circle is not a given – indeed, sometimes the warm air simply dissipates over the Bering Sea and no SSW occurs – it seems to be most likely when that warm air forms over the Bering Sea. I will discuss all of this in the Stratosphere section, so no worries if it seems tough to understand. The key takeaway here is that the Bering Sea is a key area for high pressure to form in the stratosphere, which can then lead to severe cold weather events in North America.
Let’s take a look back at Figure 28 and focus our attention on the Bering Sea. The Bering Sea should draw your attention even if it weren’t the focus of this section: the waters in that area have the highest positive SST anomalies in the Northern Hemisphere, and actually in the world per this chart! This is a significant anomaly, and provides a good look at what the coming winter could entail if these strong positive anomalies persist. The presence of strong positive SST anomalies in the Bering Sea indicate a materially-increased chance of sudden stratospheric warmings this winter, which therefore increase the risk of severe outbreaks of cold air into the United States. This is not a given, of course, but the chance of lobes of the polar vortex protruding into the United States this winter is increased by a non-trivial amount with the presence of such strong positive SSTAs in the Bering Sea.
I will touch on this topic again in the next iteration of this winter forecast, given its importance, and these anomalies will need to be watched to see if they do persist into the fall.



                  iii. Gulf of Alaska / Northeast Pacific
The majority of concerns around the Gulf of Alaska were taken care of in the Pacific Decadal Oscillation section, so this section should be brief. It is necessary to view sea surface temperature anomalies in the Gulf of Alaska in their own right, outside of the PDO, however, warranting a separate section here. Figure 28 shows a swath of well-above-normal SST anomalies in the Gulf of Alaska and northeast Pacific as a whole, an interesting development should it persist into the winter. In general, above-normal water temperatures encourage the development of high pressure systems, while below-normal water temperatures encourage the development of low pressure systems. There’s a bit of ‘the chicken or the egg’ issue as far as if low pressure systems cause negative SSTAs or vice-versa, but what is clear is that negative SST anomalies are more commonly seen in areas with stormier-than-normal activity, while positive SST anomalies are more commonly seen in areas with predominantly-calm weather.

Going along with this, the wide swath of above-normal water temperatures in the Gulf of Alaska seems to favor the development of high pressure systems more often than low pressure systems in that region. To tie it in to the Bering Sea, it seems as though ridges of high pressure could form over the Gulf of Alaska and extend into the Bering Sea, perhaps to the extent that it becomes a blocking ridge and even encourages cross-polar flow to bring significant cold air to North America. All of that contains a lot of contingencies, of course, and situations like cross-polar flow (where a ridge of high pressure pushes so far north that it moves into the Arctic Circle and actually allows air to flow directly from Siberia to North America, instead of being transferred across the Pacific and warming up in the process) can only be forecasted a matter of days in advance. Regardless, the positioning of warmer-than-normal water temperatures across the northeast Pacific and the Bering Sea indicates to me that this is a possibility if those anomalies sustain into the winter season.

Even if this doesn’t materialize, and we keep our attention only on the Gulf of Alaska, the strong positive SSTAs are supportive of a high pressure system over that area. Hawk-eyed weather enthusiasts will recognize the placement of high pressure in the Gulf of Alaska as signaling the negative phase of the East Pacific Oscillation (EPO), which is an atmospheric pattern that favors an influx of colder than normal air to the eastern half of North America when it is negative. Consequentially, should those positive anomalies persist in the Gulf of Alaska (which is subject to what the PDO does, of course), a negative EPO becomes materially more likely, and this increases the chances of a colder than normal winter for the eastern half of the U.S., while also increasing the chances of a warmer than normal winter in Alaska and the western third of North America.



                  iv. Northern Atlantic
As mentioned at the start of this section, Figure 28 would be used for the first three areas of interest. For these last two areas of interest, we will make use of Figure 29, showing a zoomed-in view of SST anomalies that will better serve our analysis for the remainder of this section.

Figure 29: SST Anomalies focused on the Atlantic Ocean and Great Lakes (IRI/Columbia Univ)

Turning our attention to the northern Atlantic, it is necessary to define exactly where we are looking. While the waters near Greenland certainly have some substantial positive SST anomalies, that area is covered by the Atlantic Multidecadal Oscillation, and was already discussed in that section. For this piece, our attention will be given to the waters just east of the East Coast.
In Figure 29, the waters offshore of the Eastern Seaboard are seen as markedly above-normal, particularly maximized offshore of the Mid-Atlantic. The presence of these stout positive sea surface temperature anomalies will alarm winter-weather fans in the East U.S. and excite winter-weather fans in the Central U.S., but for the Northeast it’s more of a ‘glass half full’ situation.

Consider the storm track of the notorious ‘Nor’easter’ system, which is known for dropping feet of snow in locations along the Eastern Seaboard when these systems crawl up the coastline and rapidly intensify thanks to the temperature gradient that the storm usually rides along. Warmer than normal water temperatures offshore the Eastern Seaboard can only intensify that temperature gradient, especially if any Nor’easter is able to pull cold Canadian air down south as it grinds northeast along the shoreline. An increased temperature gradient opens the door for further strengthening, potentially making any Nor’easters that occur even stronger than they would have been without the swath of above-normal water temperatures.

It’s also a cause for concern for those who enjoy Nor’easters, however. As discussed in the prior section for the Gulf of Alaska, warmer than normal SSTs tend to be associated with high pressure systems. Thus, the risk is that Nor’easters could be discouraged from occurring if these warm waters offshore the East provoke high pressure systems and direct any storm systems in the Plains northeast through the Ohio Valley and Midwest, instead of south into the Southeast and eventually northeast into a full-fledged Nor’easter. The absence of strong positive SSTAs near the Southeast U.S. tells me this probability isn’t as high as it could be, but this risk of high pressure systems forming and keeping parts of the East warm throughout portions of the winter is certainly there if waters east of Washington D.C. and Boston remain warmer than average into the winter.



                  v. Great Lakes
The last area of interest in the Sea Surface Temperatures section is the Great Lakes. This is most pertinent for lake-effect snow purposes, but may also dictate temperature trends to some degree for areas immediately downwind of the Lakes.

Referring back to Figure 29, the crude temperature anomalies shown for the Great Lakes show essentially all of the Lakes as being above-normal in sea surface temperatures, an indication that enhanced lake-effect snowfall is possible for areas downwind this winter. A potential exception is Lake Superior, where anomalies are zero to even slightly negative, but anomalies across all of these lakes could change substantially by the time winter actually rolls around. For the time being, though, enhanced snowfall does appear possible for lake-effect snow regions this winter. Temperatures may also average warmer than normal for those areas, especially if the positive anomalies increase.



                  e. Summary
When analyzing sea surface temperatures across the world, it becomes immediately clear that there are many variables affecting the outlook, with the possibility of numerous wrenches being thrown into the forecast and making the outlook miss the mark. However, with the exception of the PDO, it appears the majority of these key variables can be determined at this point. In the Equatorial Pacific, it looks as if the El Nino will struggle to sustain itself into the fall, and ENSO-Neutral conditions are expected for the coming winter. The Pacific Decadal Oscillation is currently too weak to be in any particular phase, and while it appears as though a negative PDO is favored over a positive PDO, there is too much uncertainty to make a definitive statement on that oscillation. The Atlantic Multidecadal Oscillation still appears to be in its overarching positive phase, and as such a positive AMO is expected for this winter. Also of note are well-above-normal water temperature anomalies in the Bering Sea and Gulf of Alaska, both of which seem favorable for a colder than normal winter in much of the country. Indeterminate SST anomalies near Japan preclude any hints of how active the storm track will be for the winter season, but this will become clearer in the Official winter forecast.
In the aggregate, ocean-based oscillations appear to lean in favor of a cooler than normal winter for the majority of the country. It is too early to pinpoint where the strongest below-normal anomalies will land, but I do feel comfortable forecasting a below-normal winter for temperatures at this point in time, based on what the Sea Surface Temperatures section has shown. This may change as we review more material – this is only one section of this outlook, of course!



                  2. Stratosphere
Although the stratosphere does not have a number of influential oscillations like the world’s oceans have, the stratosphere remains an important feature to discuss in any winter outlook. The primary oscillation that involves the stratosphere is the Quasi-Biennial Oscillation (QBO), and that is where we will begin this section.


                  a. Quasi-Biennial Oscillation (QBO)


The Quasi-Biennial Oscillation, or QBO for short, is a rather intimidating name. However, like in the other oscillations reviewed thus far, the concept will become more simple after some explanation. At the start, however, it will likely be one of the most difficult-to-understand parts of this outlook. Let’s dive right in.

Figure 30: Zonal winds on a latitude-height cross-section, as of July 25th (FU-Berlin)































 Figure 30, above, comes from the Free University of Berlin, affectionately annotated as FU-Berlin in reference to the school, not as a slight to the country of Germany. The graphic is pretty complicated at first blush, so let’s dissect it. The x-axis shows latitudes from the Equator to the Arctic Circle, while the y-axis shows the atmosphere by height, in millibars. For reference, the surface is roughly at the 1000-millibar level, the jet stream is at the 200-millibar level, and the stratosphere begins at the 100-millibar level or so.

The graphic itself shows mean zonal winds, in meters per second. Zonal winds are winds that run west-to-east: when the graphic shows positive zonal wind values, it means winds are going west-to-east, while negative zonal wind values indicate winds are running east-to-west. One can confirm this by viewing the large positive swath of zonal winds at the 200-millibar level, positioned around the 40 degree North line of latitude. What might that be? That’s the Northern Hemisphere’s jet stream! The jet stream runs west-to-east, of course, sits at the 200-millibar level of the atmosphere, and generally meanders around the 40 degree N line of latitude.

I have made some annotations on the graphic, most importantly being the positive zonal winds (“westerlies”, from the west) at the Equator from about the 20 millibar level to nearly the 100 millibar level. This is the signal being given from the QBO, which is in its positive phase. When the QBO is in its positive phase, these positive zonal winds / westerlies are strengthened in the stratosphere, as is shown in the red circle. If this were a negative QBO, instead of being positive, those zonal wind values would be negative, indicating winds are going from the east to the west (“easterlies”). We can confirm this by viewing Figure 31 below, which shows a historical time series of zonal winds. Within Figure 31, positive zonal winds are shaded in gray (remember, +QBO) while negative zonal winds (-QBO) are in white. Indeed, in the figure, we see a wide swath of gray has propagated down through the stratosphere as of late, signaling strengthened westerlies and therefore a positive QBO.

Figure 31: Historical time series of zonal winds to depict QBO phase (FU-Berlin)

You’ll notice in looking at Figure 31 that positive QBO “waves” follow a pretty identical pattern. The positive zonal winds start out at the top of the stratosphere, propagate downward rather steadily, and then linger when they reach the area between about 50 millibars and 100 millibars. You’ll also notice how the positive zonal winds seem to quickly be replaced by negative zonal winds (the “easterlies”) at the 10 millibar – 30 millibar layer, a quick change especially when compared to how long the westerlies linger at the lower layers of the stratosphere. That rapid changeover is already beginning to be seen, with the tight gradient between the 10 millibar and 20 millibar layers at the very end of Figure 31. Our assumption for this forecast is that the QBO will remain weakly positive into the winter, after having been firmly positive earlier this year into now. This will become key when we create analogs for our winter forecast later on, but for now we just need to note that the QBO will most likely be weakly positive for the 2019-2020 winter. The assumption that the QBO will keep its positive state into the winter is based on the assumption that the downward propagation of westerlies will be less sudden than that seen in mid-2017, and more akin to that seen in 2008-2009 or essentially any other year in Figure 31.

Now that we have established the QBO is expected to be positive for this winter, we can view correlation graphics like we did for the AMO and PDO to see what a positive QBO might mean as far as temperature, precipitation, and the broader atmospheric flow goes.

Figure 32: DJF Temperature correlation with the QBO (ESRL)

Figure 32, as shown above, depicts the correlation between surface temperatures and the Quasi-Biennial Oscillation phase over the December-January-February period. While viewing the chart, it quickly becomes evident that the QBO does not impose as strong an effect on conditions at the surface as oscillations like ENSO and the PDO do. Indeed, the maximum absolute correlation value is situated in eastern New Mexico, at a value of somewhere between +0.20 and +0.30. The majority of the country exhibits a slight positive correlation of temperatures with the QBO, save for slight negative correlations in southern Florida and the Pacific Northwest.

Taken at face value, this would seem to suggest that the positive QBO will encourage a warmer than normal winter over much of the country, but given how the QBO should only be weakly positive and the correlation itself is only weakly positive, I can’t see the QBO having a particularly-significant effect on temperatures this winter if it does indeed hit the winter in a weakly-positive phase.

Figure 33: DJF precipitation correlation with the QBO (ESRL)

Figure 33, shown just above, shows the correlation between precipitation and the QBO for the same DJF timeframe. While the correlations between precipitation and the QBO are more diverse across the country, they are only diverse in the sign of the correlation, and not the magnitude. Taking the graphic as seen, a positive QBO would appear to most strongly encourage above-normal precipitation near the Front Range of the Rockies, in the Upper Midwest, and in western New York. In contrast, the positive QBO would also act to incite below-normal precipitation along the Gulf Coast and into the southern Plains.

Much like the temperature correlations, however, this would only be the case if those correlations were stronger. Thus, given weak correlations and the expectation of a weakly-positive QBO, I don’t believe the QBO will have a material effect on precipitation either.

That’s not to say we should completely disregard the QBO, however: as I said earlier, wasting time is not something I go out of my way to do. Since the QBO is based in the stratosphere, let’s see if there are any impacts in the stratosphere from a positive QBO, with correlations exceeding 0.30.

Figure 34: DJF 50-millibar correlation with the QBO (ESRL)

Figure 34 above portrays the correlation between the QBO and 50-millibar geopotential height anomalies. Recall that the stratosphere is generally defined as above the 100-millibar mark, so this places us at right about the middle of the stratosphere.

As hoped, the QBO does have correlation values exceeding 0.30 (in absolute terms), seen over the upper latitudes of the Northern Hemisphere at somewhere between -0.30 and -0.40. This means that a positive QBO should cause geopotential heights over the Arctic Circle to be deeper (a.k.a. a stronger stratospheric polar vortex), while a negative QBO encourages geopotential heights over the area to be higher (a weaker stratospheric polar vortex).

A question you might be posing is ‘How can some strong winds at the Equator make the stratospheric polar vortex stronger? It’s thousands of miles away from the Equator!

Consider a basic low-pressure system. At its core, a low-pressure system is exactly what the name implies: an area of low pressure relative to its surroundings. In the Northern Hemisphere, the Coriolis Force means low pressure systems have air that rotates counter-clockwise (think of humid air being pulled from the south ahead of the system and cool air being pulled from the north on the back side of the system)  and high pressure systems have air that rotates clockwise. In the Southern Hemisphere, this is the opposite. Now, imagine you’re up in space, thousands of miles away from Earth, and you’re looking down on the Earth from right above the North Pole. From your view high above the North Pole, the jet stream in the Northern Hemisphere rotates counter-clockwise around the Earth. This is accurate, given the jet stream in the Northern Hemisphere moves from west-to-east.
So, you’re high above the Earth, looking down from your position high above the North Pole, watching the Northern Hemisphere jet stream flow around the Earth in a counter-clockwise fashion. Now, think about what happens if the winds in the stratosphere over the Equator strengthen, as happens in a positive QBO. Recognizing that the Earth’s atmosphere is a fluid and not a 2-dimensional plane, think about what happens if you were to take a garden hose and spray it horizontally underwater. What happens to the water immediately surrounding the garden hose-propelled water? The water immediately around the fast-moving water also speeds up to some degree. The same sort of thing happens if you’re standing on a train station platform and an express train goes by: the train is moving at 50 MPH, for example, and while the air around you speeds up while the train goes by, it’s not going at 50 MPH. It certainly sped up, though (and that’s probably a sign to stand further back from the train tracks).

This can be easily related to a positive QBO event: when the positive zonal winds in the stratosphere get faster due to a +QBO event, the winds at lower levels of the atmosphere also speed up to some degree. In the Northern Hemisphere, that means westerly winds in the upper-level of the atmosphere strengthen, which acts to tighten up the stratospheric polar vortex (stronger westerlies strengthen the wind fields of large vortices) and generally “locks up” cold air in the upper latitudes of the Northern Hemisphere as a result of stronger upper-level winds. This makes sense given that cold air outbreaks here in the lower latitudes occur with a “wavy” jet stream, something that happens less often as westerlies aloft get stronger.

Given this admittedly-rough description, it is no surprise then that Figure 34 shows a negative correlation between geopotential heights at the 50-millibar level and the QBO. A positive QBO leads to those increased westerly winds, which strengthens the polar vortex and thereby decreases geopotential height anomalies. In sum, this positive QBO should act to strengthen the stratospheric polar vortex (and by extension, to an extent, the tropospheric polar vortex) marginally relative to how strong it would be without a +QBO.

But let’s suppose the QBO weakens faster than we expect, and we arrive in the winter season with a QBO stuck at nearly zero. Not a problem – what’s happening right now is already telling us what the coming winter will be like, with respect to the QBO. Take a look at Figure 35 below. In Figure 35, we see another correlation graphic at the 50-millibar level regarding geopotential heights, but now with a twist. This is the correlation between December-January-February 50-millibar geopotential heights and the QBO index during the preceding July-August-September period. In other words, we are seeing the correlation that the July-August-September QBO has to December-January-February 50-millibar geopotential heights.

DJF 500mb Geopotential Height correlation with JAS QBO state (ESRL)

Why should we care? Take a look at the chart: based on this correlation, the 50-millibar geopotential heights over the upper latitudes of the Northern Hemisphere are negatively correlated by a pretty strong margin – stronger than in Figure 34, even – to the state of the QBO five months earlier, in the July-August-September period. Since there is very high confidence in the QBO remaining in its positive phase through September, there seems to be a pretty strong argument that the stratospheric polar vortex will err on the stronger side for the coming winter.

The Quasi-Biennial Oscillation is currently in its positive phase, and is expected to remain in its positive phase – albeit to a weaker magnitude – moving into the winter. This appears likely to strengthen the stratospheric polar vortex, increasing the chances for a broadly warmer than normal winter across the country.



                  b. Recent Temperature Trends
I want to briefly take a look at recent temperature trends in two portions of the stratosphere. While there aren’t any groundbreaking additions to the forecast by examining these trends, per se, I find it beneficial to know the base from which stratospheric temperatures will be starting from heading into the winter, as this could help indicate how strong or weak the stratospheric polar vortex is when the winter kicks off. Figure 36 below starts off at the 70-millibar layer of the stratosphere, near the lower end of that section of the atmosphere.

Figure 36: Recent temperatures at the 70-millibar level, between the 65N and 90N latitude lines (CPC)

The graphic is interesting in that it provides temperature data at the 70-millibar level going all the way back to the start of 2018, allowing us to see how the previous winter affected the stratosphere. For our purposes, however, we need to only focus on temperatures during the spring and summer of 2019. Looking to those data, it appears as though temperatures have recovered to right around the average for this time of year after having meandered near record-lows within the last few months. The same kind of situation unfolded last year, too, with 70-millibar temperatures generally hovering close to the average during the summer months.

When fall rolled around last year, stratospheric temperatures at this level remained around average, only starting to move more towards below-average values moving into November. This is generally the trend for the stratosphere, where temperatures established during the summer relative to the mean hold into the fall. Going by that premise, it doesn’t seem immediately apparent that the stratosphere will be substantially colder than normal to begin winter, which would be a boon to the strength of the polar vortex. Indeed, barring any unexpected event that impacts temperatures, recently-observed temperatures in the lower stratosphere don’t suggest a particularly stronger or weaker polar vortex to begin the winter months.

Moving ahead, Figure 37 (below) depicts recent temperatures at the 1-millibar level of the atmosphere, right at the top of the stratosphere.

Figure 37: Recent temperatures at the 1-millibar level, between the 65N and 90N lines of latitude (CPC)

At the 1-millibar level, it isn’t just the level of oxygen that’s remarkably low. As Figure 37 attests, temperatures at the top of the stratosphere are grinding out fresh strings of record lows on numerous occasions since spring. The same sort of thing transpired in the spring and summer of 2018, and sure enough, temperatures at this level of the atmosphere clung tight to near- or new-record-low temperatures well into the fall.

As the graphic also shows, however, this did not condemn the upper stratosphere to far-below-normal temperatures throughout the winter. In fact, the outcome was quite the opposite, with temperatures spiking to near-record-highs as a stratospheric warming event occurred.

The point of examining recent trends isn’t to try and anticipate how the stratosphere will act during the winter – that’s dependent on the number of stratospheric warming events during the season (a number that can be zero). Instead, this is more relevant for the late fall and early winter period, to try and establish what kind of shape the stratospheric polar vortex will be in as winter begins. Once winter actually gets underway, of course, all attention turns to any and all SSW possibilities.
Based on what Figures 36 and 37 have shown, nothing truly anomalous seems to be looming for the stratosphere as we move towards fall. While temperatures are setting fresh record lows at the top of the stratosphere, this would be far more noteworthy if it were occurring at the lower levels of the stratosphere. This is because the top of the stratosphere is naturally more volatile (one can visually compare the spread between record highs and record lows in the winter to see this difference in volatility), while lower levels of the stratosphere are generally more pertinent to the stratospheric polar vortex as a whole. In other words, if you see a forecaster proclaiming an imminent apocalypse because of a spike in temperatures at the 1-millibar layer, do what I do: ignore it, and wait to see if it shows up lower down in the stratosphere. The stratosphere looks pretty normal heading into the fall months.

                  c. Relations to Other Oscillations

While the Quasi-Biennial Oscillation is the most-relevant (and perhaps only) oscillation that directly involves the stratosphere, there are a number of other oscillations that are not based in the stratosphere, but still exert considerable influence on the behavior of the stratospheric polar vortex during the winter months. The oscillations we will examine relative to the stratosphere are the Atlantic Multidecadal Oscillation (AMO) as well as the Arctic Oscillation (AO).

Figure 38: DJF 50-millibar geopotential height correlation with DJF AMO (ESRL)

Figure 38, attached above, shows the correlation between geopotential heights at the 50-millibar level (~middle of the stratosphere) during the December-January-February period and the AMO during the same three-month period. The idea of this analysis is to see if the AMO, despite being based out of ocean temperature anomalies, also affects conditions in the stratosphere.
The graphic yields an interesting result. Across the upper latitudes of the Northern Hemisphere, positive correlation values are widespread, maximized in northeast Canada at values exceeding +0.40. Compared to some of the earlier correlation values we observed, this is a pretty formidable correlation, indicates that the AMO does seem to wield some influence on the stratosphere. Given that we expect a positive AMO this winter, the implication here is that geopotential height anomalies should also lean higher as a result, indicating a weaker stratospheric polar vortex. Given that the QBO will most likely still be weakly-positive for the winter, this will not be the favored outcome, but still indicates there is room for the stratospheric polar vortex to weaken more than it may in a negative-AMO environment.

The next relationship I’d like to examine involves another correlation, but also with a twist like we used in the QBO portion. In Figure 39, below, I have attached a correlation between 50-millibar geopotential heights during the December-January-February period and the Arctic Oscillation (AO) index from five months prior (July-August-September). It’s worth briefly explaining what the Arctic Oscillation is before delving into the correlation graphic.

The Arctic Oscillation (AO) is an intraseasonal oscillation, in contrast to pretty much all of the oscillations we have analyzed so far. As a result, when using the AO, we look for changes in the index on a basis of days, as opposed to a basis of months with the QBO or years with the PDO, or even decades with the AMO.

The Arctic Oscillation is tracked by observing 1000-millibar geopotential height anomalies over the far upper latitudes of the Northern Hemisphere, above 20 degrees North to be specific. The AO has two phases: a positive phase, and a negative phase. When the AO is said to be positive (+AO), geopotential height anomalies over the upper latitudes are lower than normal. This means the tropospheric polar vortex is stronger than normal, and this stronger vortex “locks up” the colder air at the upper latitudes, keeping it from flowing south. As a result, a positive AO is associated with above-normal temperatures in the United States. When the AO is said to be negative (-AO), geopotential height anomalies over the upper latitudes are higher than normal. This means the tropospheric polar vortex is weaker, and enables colder air masses to flow down to lower latitudes. As a result, a negative AO is commonly associated with below-normal temperatures in the United States.

Figure 39: DJF 50-millibar geopotential height correlation with JAS Arctic Oscillation (AO leads by five months) (ESRL)


Figure 39 shows the correlation between 50-millibar geopotential heights during the December-January-February period and the Arctic Oscillation during the preceding July-August-September (JAS) period, meaning the AO index leads by five months. The figure shows rather-strong negative correlations in the upper latitudes of the Northern Hemisphere, indicating that a positive (negative) Arctic Oscillation during the JAS period tends to bring about negative (positive) geopotential heights in the stratosphere, resulting in a stronger (weaker) stratospheric polar vortex.

Figure 40: Observed Arctic Oscillation (AO) since late March 2019 to July 26th, annotated by the author (CPC)

Now shown immediately above, Figure 40 shows observed Arctic Oscillation index values since late March, giving us a glimpse at how the oscillation has behaved as of late. I’ve highlighted in red the data since July, as the correlation period in Figure 39 spans the July-August-September period. During the month of July, the AO has been overwhelmingly negative, only moving into a positive phase briefly around the middle of the month. This is only just-under-one month of data, to be sure, but if this trend of a negative-AO continues throughout the August and September periods, it would be one of the stronger testaments in favor of a weaker stratospheric polar vortex for the coming winter.

This puts our Stratosphere section in a bit of an awkward situation. On one hand, the Quasi-Biennial Oscillation is generally viewed as the seasonal oscillation most directly affecting conditions in the stratosphere, and it favors a stronger than normal polar vortex for the coming winter, bringing about a tendency towards warmer than normal conditions across most of the country. On the other hand, though, the QBO is expected to be only weakly positive when the winter rolls around, dampening that correlation. Additionally, relationships between the stratosphere and other oscillations in the atmosphere suggest the polar vortex is actually expected to be weaker than normal, based on how conditions so far in July have played out. Where do we go from here?

Figure 41: DJF 50-millibar geopotential height correlation to MJJ Multivariate ENSO Index (ESRL)

There’s one more relationship I want to look at, depicted above in Figure 41. In the figure, a correlation is shown between 50-millibar geopotential heights in the December-January-February period and the Multivariate ENSO Index (MEI) in the preceding May-June-July period, meaning the MEI leads the DJF geopotential heights by seven months. To clarify, the Multivariate ENSO Index (MEI) is an index established to determine the state of the ENSO phenomenon by using more variables than just sea surface temperature anomalies. As such, the MEI is one of the most comprehensive methods used to determine the ENSO phase. As of the most recent data, the MEI was seen just barely in Neutral-ENSO territory, at values of around +0.3 and +0.4.

In data provided by the ESRL – the agency that maintains the MEI – values are provided in the format of a rolling two-month average. This means that the most recent data point is for the May-June period combined, rather than just the month of June. Since we don’t have July data yet, this is to our benefit, because the May-June period covers the remaining two months of the correlation period on the left. For the May-June period, the MEI was clocked at +0.4, indicating conditions are just on the edge of a full-fledged El Nino but aren’t quite established enough. Looking at the correlation graphic on the left, we see that there is a positive correlation between MJJ MEI values and DJF 50-millibar geopotential heights. In other words, a positive (negative) MEI reading for the MJJ period correlates to positive (negative) geopotential heights in the stratosphere, conducive for a weaker (stronger) polar vortex in the following winter. These correlations are rather substantial, too: values along the western portion of North America exceed +0.40, certainly one of the stronger magnitudes of correlation we have seen over the course of this publication. Given the +0.4 MEI value recorded for the May-June period, and the likelihood that July will also contain conditions near or at El Nino status, I expect the MJJ average for the Multivariate ENSO Index to be near or in El Nino territory (at or above +0.5), thus making it more likely that geopotential heights in the stratosphere will be higher, weakening the stratospheric polar vortex and boosting the chances of a cooler than normal winter.

To summarize all of this, it looks as though we will see the Quasi-Biennial Oscillation (QBO) stay in its positive phase into the winter, albeit on a firm weakening trend. The positive QBO should encourage a stronger-than-normal stratospheric polar vortex, which in turn looks to encourage a warmer than normal winter for the majority of the country. However, correlations of the stratospheric polar vortex strength to the Atlantic Multidecadal Oscillation, the summertime state of the Arctic Oscillation and the late spring / early summer state of the ENSO phenomenon all suggest that the stratospheric polar vortex will actually turn out weaker for this winter, in opposition to what the QBO seems to predict. Because these other oscillations have non-trivial correlations to the strength of the wintertime stratospheric polar vortex, and the QBO will likely be in a weak positive phase instead of at its maximum amplitude, I believe it is more prudent to expect the stratospheric polar vortex to be somewhere between average strength and weaker than normal for the coming winter, given how the QBO will have an increasingly-weak influence as the winter moves on. Consequentially, the stratosphere thus seems to favor a cooler than normal winter for much of the United States.






                  4. Seasonal Model Forecasts
In my approach to seasonal forecasting, I like to think of the process as one of constructing pillars, with the forecast itself standing on those pillars. Most of the pillars rely on observed data, such as the various oscillations above, and observed data do form the majority of my forecast. However, I find seasonal forecast models to be of use as another pillar – not for their accuracy, but as another piece to the puzzle. Just as it’s possible to be too broad when creating a seasonal forecast (i.e. relying on certain oscillations that actually have no relation to the winter pattern), it’s equally possible to be too narrow, an example being excluding seasonal forecast models. I will be the first to admit that I don’t particularly like forecast models, due to their inaccuracies, but I’ll also be the first to admit that they are worth looking at, at the very least.

\We’ll look at a handful of forecast models here. In order, and in side-by-side pairs, they will be: the CPC’s official seasonal outlook; the CFS model forecast; the IRI multi-model outlook; the ECMWF’s long-range modeling system; the Japan Meteorological Agency’s long-range modeling system; the Beijing Climate Center’s long-range modeling system; and the CPC’s NMME multi-model system. I will show all images first, in the aforementioned order, and then explain each image and provide a broader discussion afterwards.

Sub-images One (left) and Two (right)
Sub-images Three (top-left), Four (top-right), Five (bottom-left) and Six (bottom-right)
Sub-images Seven (top-left), Eight (top-right), Nine (bottom-left) and Ten (bottom-right)
Sub-images Eleven (top-left), Twelve (top-right), Thirteen (bottom-left), and Fourteen (bottom-right)

Before discussing what these images show, I’d like to identify each one and explain what they show. In sub-images One and Two, the official outlook from the Climate Prediction Center is shown for the December-January-February window, with the probabilistic precipitation forecast on the left and the temperature forecast on the right.

In the graphic showing sub-images Three through Six, the top two images (Three and Four) are from the NCEP’s CFS model (essentially the American seasonal forecasting model), with precipitation anomalies on the left and temperature anomalies on the right, both also valid for the DJF period. The bottom two images that graphic (sub-images Five and Six) display probabilistic forecasts from a multi-model composite created by the IRI establishment at Columbia University, valid in the November-December-January window. The left-hand image shows precipitation probabilities, while the right-hand image depicts the temperature outlook.

Moving on to sub-images Seven through Ten, the seasonal forecasting model from the European Centre for Medium-Range Weather Forecasts (ECMWF) is shown on the top two images (Seven and Eight), with precipitation anomalies on the left and temperature anomalies on the right, each valid for the NDJ timeframe. This is the seasonal forecast model from the same agency that produces the operational ECMWF forecast model, commonly favored during the winter season for reasons that are sometimes far exaggerated and even outright wrong. The bottom two panels (sub-images Nine and Ten) show precipitation (left) and temperature (right) anomalies for the DJF window from the Japan Agency for Marine-Earth Science and Technology’s (JAMSTEC) model, a commonly-used model in seasonal forecasting. Precisely why it’s so commonly used is still a mystery to me. In any event, the output from that model comprises the bottom two images in the four-panel graphic of sub-images Seven through Ten.

Lastly, for sub-images Eleven through Fourteen, the top two graphics (Eleven and Twelve) show precipitation (left) and temperature (right) forecasts during the DJF period from the Beijing Climate Center. That’s right, this is the output from China’s seasonal forecasting model. While I have no information about its accuracy, I find it hard to believe that the computing power of one of the world’s top countries for technology is insufficient to produce a forecast at least worth looking at. To the ‘world’s top countries for technology’ extent, note that South Korea also has a seasonal forecast model, but I have elected to not show it here simply to focus in on the ‘key’ models. If we were to include all the forecast models I have in my bookmarks folder on my laptop, we would still have to go through the Korean model, the Brazilian model, and even the Russian model, among others. Rounding out the last four-panel graphic, and comprising sub-images Thirteen and Fourteen, is another multi-model composite, this one put together by the Climate Prediction Center under the ‘NMME’ moniker, with precipitation and temperature graphics on the left and right, respectively, valid for DJF.

There are a number of points and clarifications I want to make for this amalgamation of forecast models, and we’ll proceed in the same order that they are posted above. Beginning with the Climate Prediction Center’s outlook, I cannot stress enough how ambiguous the agency makes their seasonal outlooks. To that extent, in prior seasonal forecasts, I’ve opted to not show them at all, as historically they align temperature and precipitation patterns almost exactly as what the ENSO phase would dictate. In other scenarios, such as the one here, they’ll simply cover the majority of the country with above-normal temperatures. To be sure, the scientists behind these forecasts are undoubtedly far more intelligent and sharp-minded than myself, but I still fail to see the usefulness behind leaning on the CPC’s outlook. If anything, the precipitation outlook seems to just be a textbook El Nino set-up, which (as we discussed) is rather unlikely heading into the winter. I personally see minimal use in incorporating the CPC’s outlook here, aside from merely showing it as that’s the product which will be shared to the public at large.

Moving on to the CFS model’s outlook immediately below that of the CPC, we find a forecast that seems to resemble a plausible pattern. Examining the precipitation outlook alone, it seems the CFS model supports a ridge of high pressure in the Southwest U.S. allowing the Pacific jet stream to send storm systems into the Pacific Northwest from its namesake ocean. From there, the storm systems would likely ride the jet stream south and eject into the southern Plains, before being forced north by another area of high pressure in the Southeast U.S. (I identify both of these ridges by observing the well-below-normal precipitation anomalies). This results in a storm track through the Ohio Valley and above-normal precipitation anomalies in the same region. The temperature outlook graphic seems to confirm this assessment of the implied atmospheric pattern, though it does seem willing to let that ridge in the Southwest U.S. open up and flow eastward into the Central U.S. at times. This is not too dissimilar from the typical pattern seen in a neutral-ENSO year, like we saw in the ENSO section of this outlook, albeit with below-normal precipitation anomalies maximized in the Southeast and the tract of above-normal anomalies shifted marginally westward. The temperature pattern is almost exactly opposite that of what neutral-ENSO winters typically bring, but given how the CFS model is notorious for being too warm (or even outright hot) in its forecasts, this is not too concerning in my book. All in all, the CFS model seems to present a believable outlook for the coming winter.
We now transition to the IRI multi-model composite, which shows a forecast not too far off from that of the CFS. While there are material differences in the forecast for the Southwest, the idea of enhanced precipitation in or around the Ohio Valley area is intact, with below-normal anomalies also evident in the Southeast. The temperature forecast also aligns remarkably well with the CFS, with above-normal temperature anomalies in the West extending eastward but to a lesser and lesser degree, particularly moving into the Great Lakes region. Perhaps the CFS temperature forecast may be on to something, rather than being too warm.

Moving on, we come across the ECMWF seasonal model outlook. I want to emphasize that just because it’s from the ECMWF agency, it is not necessarily as accurate as the short-range forecasting model of the same name (a.k.a. the “European model”). However, given the success the ECMWF has had with its short-range model, it seems prudent to at least go over what its longer-range counterpart shows. What it shows is a solution structurally different from the CFS and IRI, but in a couple key respects it remains similar. Perhaps the largest difference is the lack of above-normal temperature anomalies in the Southwest, a stark change from the CFS which held the highest positive anomalies in that area. Still, the above-normal anomalies in the High Plains soften with eastward progression as seen in the CFS & IRI, though they quickly ramp up along the Eastern Seaboard thereafter. Perhaps the most similar feature is in the precipitation outlook, where the ECMWF model has a swath of enhanced precipitation just as the CFS and IRI outputs do. In this case, it is shifted more to the southeast, but the presence of the above-normal precipitation anomalies in that same orientation as the other two datasets suggests this may be a credible piece to expect for the upcoming winter.
I don’t find a particular reason to go over the JAMSTEC model output, which closes out this four-panel graphic, as it’s essentially the same forecast as the ECMWF model but with the anomalies on both ends of the scale exaggerated. To that extent, the temperature forecast runs directly opposite of that from the CFS model, but my main takeaway is that above-normal track of precipitation showing up again in the East.

The Chinese long-range model, which comprises the top two graphics of the last four-panel set, shows a forecast that is, in the aggregate, pretty similar to the ECMWF and JAMSTEC models. We again see above-normal temperature anomalies in the northeastern half of the country (if you were to slice the country diagonally from Seattle to Miami) and a tendency towards more-seasonal temperatures in the Southwest. On the precipitation front, hints of above-normal precipitation are still present, but they have been shunted more to the east and drier-than-normal shadings have appeared in the Central U.S. That’s an interesting development, in my opnion, because it means the precipitation outlook from the BCC’s model is now pretty similar to the typical precipitation pattern brought about in a neutral-ENSO winter. The temperature forecast may be diametrically opposite to the typical temperature pattern in a neutral-ENSO winter, but the precipitation forecast is pretty spot on. Needless to say, the massive discrepancy between the alignment of the temperature and precipitation forecasts to their respective composites for neutral-ENSO winters is rather disarming and implores me to not use the Chinese model in building the forecast.

As we come up on the final pair of model forecast images, depicting the NMME outlook, it quickly becomes apparent why the CPC’s official forecast is so warm across the entire country. The NMME, an average of a handful of climate models, shows surface temperature anomalies between 0.5 degrees and 1 degree (I assume in Celsius) covering quite literally almost the entire country during the DJF period. I would like to be very clear in that I do not fault the scientists who put in long hours and a significant amount of intelligence to build the models that went into this averaged output, because it truly is a fantastic resource to have. But at the same time, you don’t have to be a meteorologist to think that it seems pretty much impossible for almost the entire country to experience essentially-identical temperature anomalies during the winter season, the time when the mid-latitude jet stream is at its most amplified and cold air outbreaks are most able to plow southward from Canada. I just don’t see how the NMME forecast is plausible, and will not be using it as part of the forecast.

What can we summarize all of these forecast models into? While there’s quite a bit of variability in the temperature forecasts, one thing kept sticking out again and again across the climate models: the presence of a tract of above-normal precipitation somewhere in the Central U.S. or East U.S., oriented in a southwest-to-northeast fashion and indicative of where the winter’s main storm track could be. That level of consistency is remarkable given how we remain five months out from December, and climate models are notorious for disagreement amongst themselves even a month or two in advance. So, while it’s tough to draw any specific conclusions with regard to temperature, model guidance deems it likely that the winter will be more active than normal with regards to precipitation chances / storm systems.






                  5. Analogs
The creation of analog sets is one of my favorite parts of the winter forecast process, because it’s about as close as you can get to creating your own model for how the atmosphere should play out in advance without having a supercomputer. For the uninitiated, the concept of analog forecasting in a seasonal perspective is using the oscillations we have discussed so far (ENSO, PDO, AMO…), seeing how they are currently situated and seeing how they are expected to be during the winter, and then going back in time and plucking out years that had those oscillations in similar states as they are expected to be. In theory, this should give a more precise idea as to what should happen in the future, as it uses multiple oscillations in finding similar years to create a composite instead of just one oscillation (i.e. comparing +QBO to –QBO). In practice, it’s more of an art than a science, but I believe it’s at least worth going over.

Let’s surmise the key points of the last >40 pages (in the original Word document, at least) of discussion and analysis in a short list.

·       ª The El Nino currently in place is weakening, and the ENSO phenomenon is expected to weaken from an El Nino to a neutral-ENSO environment for the winter.
·      •  The Pacific Decadal Oscillation is too weak to be placed in a specific phase with confidence at this point, and it is difficult to ascertain exactly where it will end up in the winter. Currently, data seem to lean in favor of a negative PDO, but confidence is too low to make this an expectation in the forecast.
·       • The Atlantic Multidecadal Oscillation is expected to remain positive for the winter.
·       • The Quasi-Biennial Oscillation is expected to remain positive, albeit in a weakening trend, for the winter.

Since we don’t yet know how the PDO will play out, and it would be more of a guessing game to assume a –PDO or +PDO in creating analogs, I elected to leave the PDO out of my analog creation process, and instead utilized the ENSO, AMO and QBO components only. The criteria for a winter qualifying as an analog to the forecasted set-up are as follows:

·       ENSO: The sea surface temperature anomaly of the Nino 3.4 region must have been between -0.5 and +0.5 for all three months of the winter season.
·       AMO: The AMO index must have been above 0.0 for all three months of the winter season.
·       QBO: The 30-millbar component of the QBO must have been between 0.0 and +10.0 for all three months of the winter season, and must have been in a declining state (i.e. not the beginning of a new +QBO event).

When applying these criteria, I ended up with 15 analog winters based on the ENSO dataset, 27 analog winters from the AMO dataset, and 10 analog winters from the QBO dataset. I then removed all analog winters that did not show up in at least two of the three datasets. This gave a total of 11 analog winters. There were two winters that appeared in all three datasets, and these were double-weighted in the analog creation process, as you will see in the composite images below.
Based on these criteria and the subsequent filtering and weighting adjustments, I present my analog years for the 2019-2020 winter below.


Figure 42: Analog-implied temperature outlook for the 2019-2020 winter (ESRL)
Figure 43: Analog-implied precipitation outlook for the 2019-2020 winter (ESRL)



I’ll discuss the four images relating to my analogs in order, beginning with the top graphic showing temperature anomalies (Figure 42).

These analog years suggest that the coming winter will be a colder than normal one for almost the entire country, maximized in the central Plains as well as the southern Rockies. I am a bit skeptical of below-normal temperatures covering the entire country, much like I was with above-normal temperatures covering the entire country in the CPC and NMME forecasts, which introduces some uncertainty to this analog set. In any event, the primary takeaway here is that years with a positive AMO, weakening positive QBO and neutral-ENSO environment appear to favor below-normal temperatures for much of the country.

Moving on to the bottom image above (Figure 43), precipitation anomalies during the analog winters are shown. The output is pretty neat: just like we saw in the heavy majority of seasonal forecast models, as well as in the neutral-ENSO winter composite image, there is a tract of above-normal precipitation across the South and extending up into the Eastern U.S. The fact that it shows up in our analog composite as well really adds quite a bit of credibility to the premise of a more active storm pattern this winter, even though the exact location of this above-normal precipitation swath may remain unclear. For the time being, I do feel comfortable in expecting a stormier than normal winter for the eastern two-thirds of the country, with the exact location of this active pattern to be determined but leaning in favor of the Appalachia / Ohio Valley area.

Figure 44: Analog-implied 50-millibar geopotential height anomalies for the 2019-2020 winter (ESRL)
Figure 45: Analog-implied 500-millibar geopotential height anomalies for the 2019-2020 winter (ESRL)
In Figure 44 above (top of the two images), we move up to the stratosphere and examine geopotential height anomalies at the 50-millibar level during the analog winters. The output is dramatically different than what we had expected based on what the QBO is showing, with widespread above-normal geopotential height anomalies across the upper latitudes suggestive of a weaker-than-normal stratospheric polar vortex, likely instigating the below-normal temperature anomalies across much of the country at the surface. It seems that even with a weakening positive QBO, history suggests the stratospheric polar vortex will lean to the weaker side, bolstering chances for a cooler than normal winter across the country here at the surface.

Lastly, Figure 45 shows 500-millibar geopotential height anomalies for our analog winters. The implied atmospheric pattern for the winter is wildly supportive of below-normal temperatures in the United States, with a ridge over the Bering Sea (negative WPO) which extends into the Gulf of Alaska (negative EPO) and a strong high pressure area over Greenland (negative NAO) which extends into the Arctic Circle (negative AO). While the past is not necessarily indicative of the future, the fact that our analog winters show an atmospheric flow so favorable for below-normal temperatures hints at the same sort of result for the coming winter.

To summarize, we constructed analog winters based on the ENSO, AMO and QBO oscillations, with appropriate criteria to capture where they are expected to be during the coming winter. After filtering out those winters which did not appear in at least two of the three databases, and weighting those that showed up across all three databases heavier, our analog set suggests this winter will be a colder-than-normal one for much of the country, with a more active-than-normal storm track laying down above-normal precipitation anomalies somewhere in the Central and East U.S., likely in or around the Appalachia / Ohio Valley nexus.





                  6. Forecast
With all of this data and analysis at our disposal, below is my 2019-2020 Preliminary Winter Forecast graphic.

Figure 46: The Weather Centre's Preliminary 2019-2020 Winter Forecast graphic.

I expect a majority of the country to lean towards cooler weather for the coming winter, based on the analog set and by extension based on some of the key seasonal teleconnections we have gone over. I did not highlight temperature anomalies along the West Coast given a combination of high uncertainty (absence of a definitive PDO state) and my suspicion that a colder Central and East U.S. would primarily transpire with a ridge of high pressure just offshore the west coast, leaving coastal areas average to slightly above-normal. I highlighted an area of higher confidence in below-normal temperatures over portions of the Plains into the Rockies, primarily urged on by the analog set and neutral-ENSO temperature composite graphic. I refrained from highlighting temperature anomalies in the Northeast because I’m not totally convinced a ridge of high pressure would not form immediately offshore if the central part of the country were to be below-normal, but even without any highlighting I do think the Northeast has a better chance at leaning colder than normal for this winter.

In precipitation, I have pretty high confidence in a swath of above-normal precipitation somewhere in the Central or East U.S., in my opinion most likely within the green shaded area. At the same time, the forecast graphic reflects hints of a drier than normal winter in portions of the Central U.S., though this will likely be dependent on how that above-normal swath plays out. I see a predominant upside risk to precipitation anomalies in the Northeast and along the East Coast, particularly if the strong ridge over Greenland plays out as shown in the analog composite images. I once again did not point out any anomalies in the West due to high uncertainty, but think it is more plausible to have an active winter in the Pacific Northwest while the Southwest leans drier. Again, however, this is dependent on the eventual PDO state.

Thank you for reading my Preliminary 2019-2020 Winter Forecast. Please feel free to share the article, link to it, tweet it, etc. My favorite part of making these massive forecasts is having other people take a look and providing their thoughts!

The Official 2019-2020 Winter Forecast will be released on October 12, 2019 at 12:00 PM Central Time.

Andrew