Monday, January 6, 2014

Extreme cold lake effect and what to expect on Lake Ontario, part 2

 Communities east of Lake Ontario are going to be literally buried over the next few days as an extremely cold airmass will spread eastward from the upper midwest states.  The front has passed and the snow band has begun.  These airmasses have visited the lakes before and I've experienced two similar ones in the past, the Jan 20, 1985 and the Jan 19, 1994.  Each of these events brings a little historical context as to what people will experience in the next few days and it also brings up unusual phenomena that the OWLES project may be able to document.

At 500 mb, this cold wave is manifesting itself with very low geopotential heights across the western Lakes.  Any time I see heights falling below 500 dm as far south as Lake Michigan, I'm looking at an unusually deep midlevel trough.  Now the media has spread the term 'Polar Vortex' like wildfire as if it's something right out of the movie 'Day After Tomorrow'.  Hopefully there's no panic ensuing should the mythical  'eye' of the vortex passes overhead.  Again the only unusual aspect of this 'Polar Vortex' is that a lobe of it extends pretty far to the south.  But as with so many times this happened before, we'll survive.

Perhaps more impressive is that the 850 mb temperatures have fallen to -30 C over the upper midwest and as I wrote in the last entry, these sub -30 C temps will advect eastward to Pennsylvania.  Again, as far as the big cold waves are concerned, this is not unprecedented, only unusual to the extent that they occur every several years or so.

At the surface in mid afternoon, I focus on the eastern lakes after the cold front passed to the east and the temperatures are in a free-fall.  Sub zero (F) temperatures have swept into Lake Michigan and are being moderated to the low single digits.  But further south, the sub zero air swept unimpeded across newly fallen snow in Indiana and into Ohio, sweeping up into northwest Pennsylvania.  Temperatures have fallen rapidly across the lakes and the lake effect machine turned on across Lakes Huron, Erie and Ontario.  While surface temperatures along Lake Ontario's snow band were typical for the beginning of lake effect, the temperatures approaching the western end of Lake Erie were extremely cold for mid afternoon.  Surely these cold temperatures will create quite a different type of snow compared to that of Lake Ontario.  But there are other considerations that will make the lake effect snow quite unique for extremely cold snow.

To see what I mean by unique, I go back to basics of snow crystal formation.  Crystal shape is highly dependent on the temperature in which it grows and the level of supersaturation that occurs (RH>100% but not by much).   Here, in a typical environment of snow, the type of snow crystal is heavily dependent on temperature based on this diagram from Ukichiro Nakaya, and replicated many times since (see this review article from Libbrecht 2005).  The fluffy dentrite-dominated snow preferably forms from -10 to -20 C under typical supersaturations, and can be most often observed naturally with modest ascending air associated with orographic snow in the intermountain west, some lake effect snow, and on the cold side of extratropical cyclones.  It is at these temperatures that the precipitation production efficiency reaches its peak in saturated ascending air.  The two consequences of this are that precipitation rates reach a maximum (all other considerations being under control), and the snow becomes less dense.  Clusters of dendrites (snow flakes) reach their fluffiest potential when the -10 to -20 C layer is firmly embedded within a cloud.

That's not going to happen once the eastern lakes are firmly embedded in the coldest of the arctic airmass.  According to the forecast sounding early Tuesday morning, places like Watertown, NY will exhibit a surface temperature of around -14 to -17 C and the temperature within the lake effect cloud will only be colder. When viewing Nikaya's diagram above, ice crystals forming in temperatures from -20 to -30 will yield more simple ice crystals that may find a more difficult time in joining together to create snow flakes.  Or even if they do, they would likely be denser.  Thus the snow will be denser.  In addition, small supersaturations may yield less efficient precipitation production processes may yield smaller amounts of precipitation.  What the result will be is likely less precipitation of denser snow. This super cold lake effect snow from early tomorrow morning till Wednesday should be predominantly a dense assortment of plates and columns.

But will that be the case?  One thing to remember is that Nikaya's diagram is the result of crystal growth in a controlled laboratory setting, perhaps in a chamber large enough to document the crystal growth but certainly not too large to control all the parameters researchers desired to change.  How similar will such a setting be compared to a band with a meso-alpha structure similar to that depicted in the forecast below for 0600 EST Tuesday morning?  According to the 4 km NAM (courtesy of NWS BTV), there will be an intense long axis lake effect band with a band-induced convergence zone and a mesoscale inflow jet exceeding 40 kts on its south side.  

The NAM is also quite aggressive in modifying the low-level temperature field due to strong sensible heating as the band accelerated the flow across the lake.  If the NAM is truly correct, then 2m temperatures will stay in the 20's F over the lake which is the reason that substantial amounts of CAPE will exist to aid convection. Nakaya didn't create the chart in moist convective clouds.

Thus if the NAM is correct, then there will be a substantial opportunity to get saturated ascent within the -10 to -20 C layer and that ascent will be strong starting from very shallow layer near the lake.  Will this scenario be the case?  The only way to find out is to sample the very low levels over the lake, or immediately along the shoreline near or in the band.  If we don't see temperatures near 20 F in the band on the shoreline tomorrow morning then the NAM is off the mark and we'd have to find out why.

What if the NAM temperatures wind up being too high?  Would the lake effect be just high density plates and columns of snow falling at a less than optimal rate?  Should the forecasts tone down their snowfall amounts to lower values than the nearly incredible 6" of liquid equivalent forecast by the 4 km NAM over the Tug Hill?  Let's go back and find out from the two previous cases of super cold arctic air that I mentioned earlier.

Back in 1994, a similarly cold arctic outbreak took a similar trajectory around a strong surface low in northern Ontario with a similar pattern. The 500 mb 'Polar Vortex', if you will, took a similar southerly track, the 1000-500 mb thickness was below 500 dm and, if anything, the 850 mb temps were even colder.  At the time I worked in Washington DC and Lake Ontario was just a short 7 hour trip up on Rt 81.  I had to go experience this epic event and I asked my friend Bob Boyd if he wanted to go.  He was interested too.  We departed a couple days before the main event was to begin on the 18th with Oswego as our target.

Like today, the cold front passed to the east of the lake and near surface temperatures fell to near 0 F on the upwind side of the lake.  The band immediately set up in a long axis configuration darkening the skies to the north from the Nine Mile Pt area.  This band stayed over the center of the lake but then at night it shifted south into Oswego.  The band intensified as its depth increased to nearly 4 km and it wasn't too long before we were experiencing blizzard conditions along the south edge of the band.  Not just marginal blizzard conditions.  The band intensified the inflow to 50-60 mph with snow easily 4"/hour.  These conditions persisted for hours overnight, however occasionally the band's convergence axis moved overhead and the winds would quickly drop to near calm.  With surface temperatures nearly 0F, Bob and I witnessed numerous lightning flashes both on the windy south side of the band and within its center.

A view of Lake Ontario on Jan 18, 1994 from near Nine Mile Point facing north.  Thick lake effect snow darkened the sky as a long axis lake effect band developed.  Closer to me, a hole formed in the shore fast ice creating a natural blow hole as waves swept into the narrow constriction.

The sounding from Buffalo that evening was impressive.  It showed that the convective layer could easily have been up to nearly 4 km with lake heating, but starting out with a surface temperature near -20 C!  There was no part of the sounding within the dendrite layer.  Thus the snow that we were getting would likely have formed in temperatures well below the optimal levels to maximize precipitation efficiency.  Yet we were getting absolutely clobbered.

When morning broke, the scene outside our hotel room looked like the picture below.  It was tough to make a good guess of the snowfall overnight but I suspect over 2 feet and less than 3.  What was more interesting though was that every upwind surface was plastered with snow.  This was a scene that I would've expected from a wet snow.  But clearly when the surface temperature was near 0 F, there wouldn't be much liquid water to provide the adhesive properties of this snow.

A view looking east in the morning of Jan 19, 1994 in Oswego, NY after a major super cold lake effect event.
Other views around town showed more examples of how adhesive this snow was.  Icicles on the downwind sides of buildings were coated in snow while most signs facing upwind were plastered with enough snow to form hard faced pyramidal shapes.
Icicles hanging off the east side of the hotel were covered in snow while a parking sign in a nearby Walmart wore a nearly perfect pyramid of snow on its upwind side.
Meanwhile trees and power lines wore mantles of snow as if the temperatures were near freezing.  Even in the face of strong winds on the periphery of the band, the snow stuck fast to most objects.

Snow stuck to trees and power lines throughout the city after the main lake effect snow event.

At near 0 F, I would've expected an easier time of it removing snow from cars.  Not in this case.  The snow stuck more like melted marshmallow.  

My car was frosted by snow from the cold lake effect storm in Oswego, NY.
The adhesive quality of the snow was different than any other kind of snow for which I experienced and the first hypothesis I had for this type of snow was that it was statically charged in the strong electric fields present in the snow band.  Considering the amount of lightning we had overnight, this is the best explanation I had for the nature of the snow.  

Nearly 9 years previous to the 1994 event, I was fortunate to experience another impressive shot of extremely cold lake effect on Jan 20, 1985.  This one was colder than anything today could dish out and the 1994 event.  The 500 mb plot showed a much more exceptional low latitude 'polar vortex' than today's or in 1994.  

Again the Buffalo sounding showed exceptionally cold air starting out nearly -20 C with no sign of a dendrite production layer.  

As in the 1994 event the snow band struck Oswego with a severe dumping with near surface temperatures close in the low single digits F.  The picture below was taken near the height of the lake effect event when we were in the strong inflow south of the band axis.  My friend, Anthony Artusa was surely reveling in this as much as I was.  Later on, we experienced several nearby strikes of cloud to ground lightning and several prominent towers in the area exhibited significant coronal discharge.
Oswego getting pounded by lake effect on Jan 20, 1985.
 The next morning, I saw the same snow adhesiveness as in the 1994 event.  Most objects sticking above ground were coated in snow that was not easy to dislodge.  We found out at this time that the snow was great for making snowballs, at a temperature of 0 F!
A picture of me enjoying the aftermath of the lake effect on the SUNY Oswego campus on Jan 21, 1985.

In both cases, I believe the snow was statically charged in a highly electrified environment within the lake effect band.  Assuming that was really the case, a strong electrical field plays havoc with Nakaya's snow growth chart.  Libbrecht (2005) describes that strong electrical fields can accelerate growth of crystals, typically with strongly accelerated growth of needles under a modest application of an electric field but then significant branching may occur at higher field intensities.  The enhanced growth rates could easily compensate for the slower growth rate of snow crystals at temperatures outside the dendrite production zone of -10 to -20 C.  

Whether or not this idea of electric field-induced crystal growth explained the huge snowfall rates is not something I can prove.  For one thing, I would've had to test whether individual snow flakes had a charge. Second, there's not much understood about snow growth in electric fields, as this article in website explained.  Clearly though, the presence of electric fields means that one who uses the Nakara crystal habit diagram to forecast snowfall rates, or snow density in lake effect events, stands the possibility of being wrong.  

In addition to the uncertainties with electric field induced growth, much of the snowflakes I saw during both past events were heavily rimed.  Many times I witnessed the classic dumps of graupel.  This was not partially melted graupel, obviously, but snow flakes were so rimed that they took on the rounded shapes of graupel, and there were huge amounts of them.   In fact, this may have explained why Oswego only got a 17:1 snow to liquid ratio out of the 1994 event. So this does bring back the idea that liquid water content also played a significant role in determining the snow type in Oswego.  If the lake effect clouds were truly at or below -20 C, I can't imagine too much liquid water within them.  Yes there may be some, and perhaps enough to cause significant riming.  However much of the riming I saw seemed to have come from warmer temperatures.  I suspect there was strong saturated ascent at temperatures warmer than what we saw onshore in Oswego.  But there was no way of knowing.  However, if the near surface temperatures depicted in the 4km NAM forecasters was anywhere close to being true then I can imagine not only an active dendrite production zone but also significant riming.  Certainly the strong buoyancy concentrated close to the lake surface, in combination with saturation, would yield enough flux of liquid water to do the job.

One has to also question how in the world did I experience lightning in both cold events if the cloud was too cold for significant charge separation via the interaction of graupel and lighter crystals?  Both of these events would've seriously fallen off the left side of the range of a nomogram proposed by Steiger et al. 2009 namely because the height of the -10 C level was below ground.  They used the -10 C level as one of the important parameters determining the likelihood of lake effect lightning as a proxy for the presence of mixed graupel and snow crystals considered necessary for charge separation.  Their other most significant parameter, convective layer depth, was considered an important contributor to updraft strength.   The one consideration that could not easily be put into the parameter was the lake band morphology.  Just about all lightning events came out of long axis single lake effect bands.  This was not a worry for either of the events I experienced.  But the extreme cold temperatures suggest that the only area with significant graupel production would have to be relatively close to the lake.  And that any loss in graupel production efficiency owing to the cold would have to be made up by intense vertical motions right off the lake surface.  I suspect the extreme horizontal and vertical temperature gradients could have supplied the necessary vertical motion.  Again, if somehow, the NAM is right about the over lake 2m temperatures for this event, then these past events would have had similarly warm over lake temperatures.  I remain doubtful of that the NAM is right.
A nomogram showing the occurrence of lightning (stars) vs its nonoccurence (dots) as a function of the -10 C isotherm height and lake induced equilibrium level based on lake induced CAPE.  This is figure 10 in Steiger et al. (2009)

What does this have to do with this event?  Well, the OWLES project gives an opportunity to answer whether or not there is significant charge splitting from graupel formation in an extremely cold event.  The project can also determine if the NAM is correct in its 2m over lake temperature forecasts and whether that is a significant contributor to vertical velocities.  After all, we need to figure out how strong these vertical velocities can become with such extreme thermal gradients.   And we need more documentation of lightning under extremely cold events.  It's not often that a field project can coincide with a once in 10 year event like this.  Good luck!

p.s.  I heard from Dave Zaff, the SOO in Buffalo, that there was lightning this evening with the Lake Erie snow band.  The 00Z sounding's maximum temperature was -13 C and a convective layer depth was around 2.7 km.   So it has begun.


Libbrecht, K. G., 2005:  The physics of snow crystals, Rep. Prog. Phys, 68, 855-895.  Available online at []

Steiger, Scott M., Robert Hamilton, Jason Keeler, Richard E. Orville, 2009: Lake-Effect Thunderstorms in the Lower Great Lakes. J. Appl. Meteor. Climatol.48, 889–902.

Sunday, January 5, 2014

What is an extreme cold air outbreak going to do over Lake Ontario?

Not for a long time have I seen an opportunity of cold air as extreme as this upcoming outbreak to interact with the eastern Great Lakes.  What will be the impact of such cold temperatures on the lake effect snow machine east of Lake Ontario?  First, let me say that the impact will be huge, not just from the unusually large horizontal and vertical temperature gradients but also the well aligned winds down the long axis of the eastern Lakes and the relatively deep convective layer.  So yes, the lake effect machine will be put into maximum overdrive.  But there are specific impacts and forecasting issues that we may expect to occur based on previous experience, and new ideas that have arisen since some of the last big arctic outbreaks.  I'll start with discussing what I think of applying CAPE to forecasting the intensity of Lake Effect.

From yesterday's NAM forecast, the surface temperatures will be in the single digits above or below zero (F) at the surface by Monday night along the shoreline of both Lake Erie and Ontario.  These temperatures are cold but as you can see they have been modified to some extent by the western lakes.  The unmodified air swirling south and eastward through the Ohio valley will be well below zero F, even for high temperatures.  The only exception to this scenario lies directly over the lakes Superior, Huron and Ontario where 2m temps exceed 15 deg F.  This is a big question if the actual 2m temperatures will be able to remain this high due to sensible and latent heating, or whether the models are having a fantasy?

Meanwhile at 850 mb, sub -30 C air temperatures will likewise swirl around the lakes spreading eastward toward PA on the south side and into eastern Ontario to the north. The lake modification extends to this altitude in the models and I think this is accurate.  By the 700 mb height, just about all lake temperature modification is gone and the NAM forecasts widespread sub -30 C temperatures across the Great Lakes.

The question of CAPE in lake effect

What does the forecast sounding look like from a point over Lake Ontario?  An example appears below where the sounding for early Tuesday morning shows a superadiabatic temperature lapse rate in the lowest 1 km of the atmosphere above the lake and a 2m temp of -10 C.  That's quite a bit warmer than the temperatures at the same level on either side of the lake representing a classic depiction of strong lake induced heating simulated by the NAM.  The result will be a model-based CAPE of 81 j/kg.  Note that the nearest moist adiabatic indicates this CAPE to be from a mixed layer.    As you can see the lowest model layers have the highest lapse rates, exceeding 12 deg/km or almost 3 deg/km greater than an adiabatic lapse rate.  At these cold temperatures, the moist parcel and the dry parcel adiabatic are pretty similar and thus the moisture flux from the lake is not contributing much to the CAPE.  It's mostly the intense sensible heating from below contributing to the CAPE.  Though it's nice to have near saturation from low-levels to help reduce any dry air entrainment into any updrafts.

A CAPE of 81 j/kg may seem pretty tame for summer convection aficionados but consider that all of the CAPE lies below 3 km MSL (only 76 m below lake level).  The highest parcel to environmental temperature excess in the convective layer is roughly 3-4 deg C (1 km LI = -3 to -4).  If a parcel following the yellow curve was unmixed and idealized (e.g., no resistance from air above it, no entrainment), it would reach a vertical velocity of ~12 m/s.  Such a vertical velocity would be likely to be strong enough to separate significant charge should a healthy region of graupel and ice crystals mix in a deep enough layer.  Scott Steiger has a good paper discussing lake effect lightning (Steiger et al. 2009).  I'll return to this later when you see where I'm going.

The question is whether the air parcel should be mixed or not?  The NAM is obviously allowing lake modification to occur, even though the grid resolution of this output is relatively course (~12 km?).  Perhaps the surface parcel should be used to calculate CAPE.  In this case, a much larger value appears and it looks like this below.

Now in the SKEWT the thinner yellow curve from the surface extends beyond the 3 km MSL level and yields a CAPE of 303 j/kg.  Calculating a pure parcel-based vertical velocity yields an impressive 24 m/s!  Now we're talking a vertical velocity akin to summer convection.  What's more impressive, however, is that the 1 km LI is nearly -7 deg C!  If we were to plot the lowest LI found in a convective layer vs CAPE I'm pretty sure a -7 LI would be on the extreme end for that range of CAPE.  But somebody should call me on that assumption.  Needless to say, according to pure parcel theory, the vertical acceleration would be amazingly strong in the lowest km of the atmosphere.

Model-based MUCAPE also depicts values in this range, as can be seen in this forecast made available by the College of Dupage.

We're not done yet, however.  We could apply an empirical technique to modify the 2m land temperature and dew point upwind of Lake Ontario to determine a near surface beginning parcel.  This technique, based on Phillips (1972), would typically warm the 2 m temperature approximately halfway between the upwind surface temperature and the lake temperature after a typical over lake residence time of 90-120 min (winds 30 kts or so).  Assuming the upwind temperature is near -17 C and the lake temperature is near 4 C (see GLERL's lake temp analysis of 4 C) then the modified temperature would be -4 C and the dew point would be ~ -6 C.  Calculate a surface-based CAPE then would yield an incredible 1332 j/kg and 1 km LI of -17!  This would convert to a pure parcel-based vertical velocity of 51 m/s!  I bet even tornado chasers would drool over those numbers in the late spring.

Note that I used the sounding point at Watertown NY which is away from the lake heating, and therefore loses the superadiabatic lapse rates below 1 km.

A vertical velocity of 51 m/s would surely yield a huge precipitation-free cavity surrounded by graupel the size of basketballs and incredible lightning displays.  Well, even a more modest 24 m/s just from the NAM-based SBCAPE would do the same though the graupel would be maybe the size of grapefruits.  Okay maybe we wouldn't see graupel that big because all the graupel would be flushed out the top of the convection and fall out the side leaving a big linear bounded weak echo region (BWER) down the centerline of the band.  While a huge BWER hasn't been observed, smaller ones have been observed by mobile radar during a small field experiment in 2012  (Steiger et al. 2013).  But vertical velocities of 12 m/s would be more than sufficient to loft frozen precipitation, even graupel, out of the updraft.

I have two considerations that cause me to seriously doubt vertical velocities reach those values depicted by the SBCAPE calculations using the 2m NAM temp and dewpoint or using Phillip's regressions of temperature and dewpoint.  For the Phillips equations, I find it hard to believe in the temperatures depicted.  If so then shoreline observations should show similar temperatures, or temperatures that would match Phillips regressions for even a relatively modest 0 C lake temperature assuming that the warmer temperatures offshore would've been overturned before reaching the cooler shelf waters.  But it's not just the temperature I suspect, it's also the concept of applying pure parcel theory to calculate peak vertical velocity in a lake effect band.

Pure parcel theory ignores the impact convection has on its surroundings, and it also ignores the impact of pressure perturbations.  Among other things, the application of parcel theory, the foundation behind using CAPE, depends on the surroundings being completely unaffected by the parcel.  Perhaps parcel theory can be applied on the scale of a cumulus updraft because it's energetics is very small compared to surrounding environment and thus it's impacts can be ignored (still to one's peril).  But when there is a massive heating source residing in the meso-alpha scale (i.e. Lake Ontario) that completely alters the state surrounding any point, the concept of steady base state loses its meaning.  The concept of a parcel also loses its meaning as well.  That's not to say that a lake effect band isn't convection.  The band is releasing energy through convective processes.  But it's not the kind of process that can be approximated by using a primitive parcel theory that forms the foundation of CAPE.  The process is more akin to that of a hurricane where buoyancy is consumed as soon as its produced to provide a meso-alpha scale region of heating from which an organized circulation develops.  In the lake effect example, the circulation develops around a linear axis as opposed to a circular area as in a hurricane.

I believe that using CAPE should be used with even greater caution in a lake effect environment than that of a more typical convective situation.  And using CAPE from the Phillips equations output is nonsense.  There is a great article on the concepts of buoyancy and how the real situation is so much more complicated than can be described by simple parcel theory.  If you're up to it, read Doswell and Markowski (2004).

I suspect that vertical motions for the upcoming event will be observed that lie between the mixed parcel model and the surface-based parcel model.  So that means somewhere above 12 m/s and below 24 m/s.  Isn't it great coincidence that we will actually find out to some extent.  The Ontario Winter Lake Effect Systems (OWLES) project has started its second phase of operations on Jan 4 and will be ready for this event.  They have the Wyoming King Air plane available for direct measurements of vertical velocity ready to provide an answer.  However even with the plane up there, we may miss the most intense portions of the lake effect if the best instability occurs outside their flight times or locations.  But it's certainly a great opportunity for getting a vertical velocity value.  In addition, numerous ground teams will stand ready to collect temperature and dew point data to evaluate how the lake modifies the near surface air.

I'm going to post another entry tomorrow about what folks living east of Lake Ontario may see in this lake effect event based on two previous super cold events I've experienced in the last 30 years.

Doswell, Charles A., Paul M. Markowski, 2004: Is Buoyancy a Relative Quantity?. Mon. Wea. Rev.132, 853–863.

PhillipsD. W., 1972Modification of surface air over Lake Ontario in winter. Mon. Wea. Rev.100662670.

Steiger, Scott M., and Coauthors, 2013: Circulations, Bounded Weak Echo Regions, and Horizontal Vortices Observed within Long-Lake-Axis-Parallel–Lake-Effect Storms by the Doppler on Wheels*. Mon. Wea. Rev.141, 2821–2840.

Steiger, Scott M., Robert Hamilton, Jason Keeler, Richard E. Orville, 2009: Lake-Effect Thunderstorms in the Lower Great Lakes. J. Appl. Meteor. Climatol.48, 889–902.