Ramblings of a Graduate Student

Above is a computer generated forecast for Tuesday morning (6 AM CST) 9 February 2010.  This forecast comes from the Global Forecast System (GFS) model and is used by meteorologists around the world to help predict the weather.  It is run 4 times a day (00Z, 06Z, 12Z, 18Z — 6 PM CST, 12 AM CST, 6 AM CST, 12 PM CST) and goes 384 hours into the future.  Now, as the forecast goes farther into the future the less accurate it tends to be.  This is why you will hardly ever see a meteorologists make a 384 hour prediction.  Often times something that is seen in the later parts of a forecast run (200+ hours into the future) will change from run to run.  This means that a major snow storm forecast at hour 288 might not be there on the next run.

If the forecasts are so unreliable that far into the future, then why do the models run that far into the future?  Well, the short answer is that while the exact details may change, often a pattern will emerge and you can get a big picture idea of what may happen in the future.  Take the above forecast image (valid 216 hours into the future).  The green on the map is areas where the model is predicting precipitation and areas to the north of the blue ‘0′ line are places that are forecast to be below freezing.  Based on this forecast, a meteorologist might make a prediction that there is a chance of snow in northeast Oklahoma, SW Missouri, and NW Arkansas.  However, being so far into the future, that is not a good idea.

What a good meteorologist should take from this image (and the entire forecast) is that the model is forecasting cold air to come southward out of Canada into the southern plains sometime next week.  Furthermore, the model is predicting some precipitation mechanism to develop in the southern plains / Gulf Coast around the same time.  Now, will this occur at 12Z (6 AM CST) on 9 February?  It’s too far out into the future to know for sure.   If several model forecasts in a row try to develop a storm on 9 February 2010, then a meteorologist would have more confidence that will happen.  If the model continues to change the date and time of when this might occur, a good forecaster will refrain from making a specific prediction about where and when, but understand that ingredients are coming together for a possible winter storm.

As for what the models are currently forecasting?  Well, I wouldn’t put too much stock in the exact timing of things, however, several models are forecasting cold air to move south from Canada.  At the same time, the models continue to have cyclones (lows) move across northern Mexico or southern Texas into the southeast United States.  The cyclones will provide a source of lift for the atmosphere and help to bring moisture from the Gulf of Mexico northward into the southern United States.  Add in the cold air from Canada, and the models are hinting at the potential for more winter storms in the coming weeks.  It’s far too early to tell where and when and how much.  However, it isn’t too soon to begin paying attention to forecasts for next week and beyond.  A lot can (and probably will) change in the coming days, so stay tuned!

Another non-technical discussion to atone for the last few days!

I have shown this map before (after all the snow the country had the last week of 2009) but in light of the last snow storm, I thought it would be interesting to view again.

Below is the National Snow Analysis for 30 January 2010.  You can clearly see the track of higher accumulations across the southern United States.  It’s going to be a few more days before a significant amount of that snow is going to melt!

After several days of more technical discussions, tonight we’ll tone it down a bit.

The winter storm that has blanked much of the Texas Panhandle, Oklahoma, and western Arkansas with a coat of ice, sleet, and snow is slowly moving east.  The graphic above displays all valid warnings.  Notice all the hot pink in the south?  Those are current winter storm warnings associated with the shortwave trough we’ve discussed this week.  Last night, the hot pink extended west into eastern New Mexico, and tonight it extends to the Atlantic Ocean.  This will certainly be a storm that is talked about for years!

As for where I live?  I received 1″ of glaze ice on the windward side of objects and 0.25″ of glaze ice on the leeward side of objects (freezing rain), 1″ of sleet, and 6-7″ of snow.  A taste of just about every form of winter precipitation!

Last night I posted that it appeared that central Oklahoma (including the Oklahoma City Metro area) would have a good chance at experiencing a dry slot, and that precipitation wouldn’t amount to as much as some of the previous forecasts had indicated.  Examining a regional radar view from 00Z (6 PM CST; displayed above).  It would appear that a substantial dry slot had indeed overtaken the central Oklahoma area.  After all, precipitation had ceased to exist.  In fact, using the warm conveyor belt and dry slot arrows previously used, we could easily construct what would appear to be classic cyclone structure (displayed below).

However, the morning runs of the numerical weather prediction models did not capture such a dry slot in their predictions…at least not one of this size and speed (this thing as moving fast!) nor as early in the storm.

Above is the forecast of the amount of precipitation that would accumulate over a 6-hour period ending 12 hours into the forecast.  It is taken from the Global Forecast System (GFS) numerical weather prediction model simulation that was started at 12Z (6 AM CST) this morning.  In other words, the image above displays how much precipitation the GFS was predicting to fall between 18Z (12 PM CST) and 00Z (6PM CST).  As you can see, the heaviest amounts of precipitation were forecast to fall along the I-35 corridor in Oklahoma and north central Texas.  Certainly not what actually occurred.   In fact, the model didn’t really decrease the precipitation in central Oklahoma until almost 12 hours later (image below).

If we look at the same kind of plots for the North American Model (NAM; 12-hour forecast above, 24-hour forecast below), we find almost the exact same pattens, albeit with different precipitation amounts (possibly related to the different resolutions of the model).

If we think back to the post on dry slots, and examine a 700mb chart, we see that there isn’t a clearly defined dry slot in the GFS 12-hour forecast (above), but there is a hint of one in the 24-hour forecast (below).

I’ve annotated the image above to highlight the possible dry slot, and this can be seen below.

Examining the exact same plots in the NAM model we find, once again, a very similar signal.

Once again, I’ve annotated the image above and it can be seen below.

So, if the models weren’t forecasting such a pronounced “dry slot” then what did the model’s miss? How could they have performed so badly 12-24 hours into the future? While I cannot say for sure, I would like to offer one possibility that is evident when examining a water vapor loop for today. Below, I’ve included a water vapor image taken 3 hours apart. I’ll give you a chance to come up with your own guess before I give mine.

Water Vapor image valid 1145Z on 28 January 2010 (5:45 AM CST on 28 January 2010) .

Water Vapor image valid 1445Z on 28 January 2010 (8:45 AM CST on 28 January 2010).

Water Vapor image valid 1745Z on 28 January 2010 (11:45 AM CST on 28 January 2010).

Water Vapor image valid 2045Z on 28 January 2010 (2:45 PM CST on 28 January 2010).

Water Vapor image valid 2345Z on 28 January 2010 (5:45 PM CST on 28 January 2010).

Water Vapor image valid 0315Z on 29 January 2010 (9:15 PM CST on 28 January 2010).

Any thoughts? This time I’ll add some annotations to the water vapor imagery. The images will be identical to the ones above with the following exceptions:

1. Large “X” is the location of the “main” upper-low / short-wave trough.
2. Small “x” is the location of a considerably smaller upper-low / short-wave trough.
3. As the images advance in time, I leave the previous time step(s) X(s) on the image so you can track it’s movement.

Water Vapor image valid 1145Z on 28 January 2010 (5:45 AM CST on 28 January 2010) .

Water Vapor image valid 1445Z on 28 January 2010 (8:45 AM CST on 28 January 2010) .

Water Vapor image valid 1745Z on 28 January 2010 (11:45 AM CST on 28 January 2010) .

Water Vapor image valid 2045Z on 28 January 2010 (2:45 PM CST on 28 January 2010) .

Water Vapor image valid 2345Z on 28 January 2010 (5:45 PM CST on 28 January 2010) .

Water Vapor image valid 0315Z on 29 January 2010 (9:15 PM CST on 28 January 2010) .

One thing that should have stood out was the the “main” upper-low / short-wave trough did not move much until toward the last few images, and even then it didn’t move quickly. Since this short-wave trough didn’t move much, one could make an argument that the precipitation field shouldn’t have changed all the much either. We certainly wouldn’t expect to see a fast moving dry slot, like what was observed.

The other thing that probably stood out was a subtle short-wave trough / upper-low (and infered upper-high) that rapidly moved from northern Mexico at 12Z (6 AM CST) through western Oklahoma and into Kansas by 03Z (9 PM CST). Without getting into all the physical reasoning and mathematics behind why, meteorologists tend to expect rising motion ahead (or downstream) of a moving short-wave trough / upper-low and sinking motion behind (or upstream) of a moving short-wave trough. I propose that the enhanced precipitation rates that were experienced in western Texas, the development of thunderstorms so early in the day, and the enhanced radar reflectivity on the back-edge of the extensive precipitation shield over much of the southern plains (look at the yellows on the western side of the greens in the radar image at the top of this post) were in response to this aforementioned short-wave trough. However, as the short-wave trough moved through the area, the sinking motion in the wake of the subtle short-wave trough was strong enough to decrease (and temporarily stop) precipitation across western Texas and Oklahoma. Thus, in the wake of the short-wave trough, we had a short-wave ridge that briefly ended precipitation. However, as the short-wave trough and short-wave ridge continued to move away from Oklahoma, precipitation began to redevelop – which is why freezing rain is once again falling in Norman.

Now, if a short-wave trough did move through Oklahoma, and short-wave ridging and sinking motion were occurring, it should be evident on a sounding. Below is a sounding taken at 18Z (12 PM CST) in Norman, OK. Notice how smooth the temperature (red line) and dewpoint (green line) are. This would not indicate sinking motion taking place. (Which is what we would expect since at 18Z Norman was experiencing freezing rain.)

Below is the Norman, OK sounding taken at 00Z (6 PM CST; or six hours later). Again, based on the radar image at the top of this post, Norman was located in what appears to be a dry slot. If we examine the temperature (red line) and dewpoint (green line) in the sounding we see that they are no longer “smooth”. There is a kink in both the temperature and dewpoint at around 500mb (or 6 kilometers). Notice how the dewpoint line becomes relatively far away from the temperature line. This would indicate that something cause the atmosphere to dry out. Now, this drying is located above where we would expect to find a dry slot (remember 700mb?). Based on this drying out and the warming of the temperature at the same level, I suspect this is a subsidence inversion, in other words sinking motion, in response to a short-wave ridge. As I mentioned in the dry slot post, sinking motion tends results in compressional warming…and a drying out of the atmosphere – which is exactly what happened.

If this subtle short-wave trough was not accurately forecast by the models, this would explain why the forecast precipitation and observed precipitation did not match well in time. Now, I have made some hand-waving arguments and over-simplifications, but hopefully you can see how this is a plausible explanation for what has transpired today across Oklahoma. Another explanation combines both the dry-slot and the subtle short-wave trough theory by arguing that today’s dry slot was actually associated with the subtle short-wave trough and not with the main upper-low. Which theory is correct? I’m not sure. All I know is that the model’s did a poor job in forecasting this small, subtle short-wave trough that moved out of Mexico and a lot of precipitation timing forecasts were blown as a result!

Oh, as an aside, as the “main” short-wave trough begins to move into the southern plains, I would expect yet another round of precipitation to develop. Hopefully this round will actually bring central Oklahoma snow instead of ice…

Ah, the night before a forecast big storm. Meteorologists keep checking the data non-stop hoping the latest piece of information is available. When the numerical weather prediction models’ forecast transitions away from the forecaster’s desired forecast, a lot of “wishcasting” takes hold. Meteorologists will often tend to find any piece of information to sustain hope that what they want will ultimately develop. This internal struggle can be a challenge for a meteorologist. I know I struggle with it from time to time.

As residents throughout the southern plains head to bed this evening, most are still unsure what tomorrow’s weather will hold.  This is because forecasters are still struggling to understand the evolution of the winter storm.  It is apparent that somewhere will see heavy snow, somewhere will see sleet and freezing rain, and somewhere will only see rain.  The problem is pinpointing those exact locations is difficult even 24 hours in the future.

A couple of things struck me tonight with respect to the shortwave trough (upper-low) responsible for the difficult forecast…

Above is a satellite image depicting the moisture content of the middle-to-upper atmosphere (brighter colors means more moisture aloft).  This is important because the most efficient way to get moisture aloft is for there to be rising motion in the atmosphere.  Thus, wherever there is the brighter colors, we can infer rising motion.  Rising motion is important because it is a key (but not a sufficient) ingredient  in the development of precipitation.  Thus, with all that rising motion in the southwest United States, one would expect a lot of precipitation in the southern plains as the rising motion moves into the area.  This agrees quite well with the precipitation forecasts from the Hydrometeorological Prediction Center (HPC) shown over the last few days.

Below I’ve identified the center of the upper-low (red L) and what I suspect will be the eventual forecast path based on a combination of observational and model trends.  This is considerably farther north and west than what was forecast several days ago.  As I mentioned last night, a farther north/west track will result in more places being affected by the dry slot – including Oklahoma City and points south and east.  However, before the dry slot affects central and eastern Oklahoma, these places will spend an extended period of time in the warm conveyor belt which will bring warm, moist air.  This warm, moist air will result in heavy precipitation and a warming of the temperature aloft (and possibly at the surface).  This warming aloft will prevent snow crystals from forming in central Oklahoma which leaves sleet, freezing rain, or a cold rain as the resulting precipitation types.

As you can imagine, I’ve been pretty busy the past few days handling all the requests for information and discussion regarding tomorrow’s (possible) winter storm.  I didn’t have time to annotate any more graphics tonight, but I will leave you with something better.  Below is a recording of today’s HWT Map Discussion.  It may be a little too technical, but it can give a better glimpse as to what a forecaster must look at in situations like these.  Please feel free to ask questions and provide feedback regarding anything in the last few days worth of blogs and / or the video below!

27 January 2010 NOAA HWT Map Discussion from Patrick Marsh on Vimeo.

I have had a difficult time trying to write tonight’s blog. Normally I have this problem because there is nothing really “exciting” to write about. Tonight the problem is I don’t know what to write without creating a dissertation! As I hinted at last night, almost every meteorologist in the southern plains is focused on the exact evolution of the southern plains winter storm. There were close to 70 National Oceanic and Atmospheric Administration (NOAA) and University of Oklahoma (OU) meteorologists in attendance at today’s Hazardous Weather Testbed (HWT) map discussion. I’m sure tomorrow we’ll have even more.

The official precipitation forecast from the National Weather Service’s (NWS) Hydrometeorological Prediction Center (HPC) continues to have over 2.5-3″ of liquid equivalent (the amount of rain or water from melted ice/snow) falling across portions of Oklahoma, western Arkansas, and northern Texas.  Generally speaking, almost all of today’s numerical weather prediction (NWP) models are in excellent agreement with this HPC forecast.  Where the models differ – between each other and also with different simulations of the same model – is in the form that this precipitation will fall.  Hopefully by the end of this post someone will have an idea as to what will happen…

The culprit for the soon-to-be winter storm is easily identified in current satellite imagery.  In the image above, the shortwave trough (upper low) that is the batch of white clouds off the southwest coast of California.  Over the next 24 hours the upper low should track south-east into northern Mexico.

Sometime during the 24-36 hour time-frame the upper-low should begin to turn more toward the east and then northeast.  When this turn to the northeast happens is crucial for determining precipitation types and duration across portions of the Texas Panhandle, Oklahoma, and western Arkansas.  One reason why where the turn occurs is important, is that it will have a direct impact in where several important features of the cyclone will become established.

Above is the 700mb (~10,000 feet above ground level) chart from the 00Z (6PM CST) North American Model (NAM).  The 700mb chart is important for winter weather forecasting because a lot of important features are easily identified on it.  Below, I’ve annotated the same chart as displayed above.  The red arrows represent initially warm, moist air (moving in the direction of the arrows), whereas the blue arrow represents initially cold,dry air (also moving in the direction of the arrow).  The red arrow is often referred to as the “Warm Conveyor Belt” and the blue arrow is often referred to as the “Dry Slot”.

In the warm conveyor belt, warm moist air from near the surface flows northward into the developing cyclone.  As it flows northward, it tends to encounter colder, drier air at the surface.  Warm, moist air is less dense then cold, dry air so the warm, moist air flows up and over the cold air.  So, in addition to the warm, moist air flowing northward, it is also flowing upward (from the ground).  As warm, moist air reaches higher altitudes, it encounters lower atmospheric pressure and begins to expand.  This expansion of the warm, moist air causes the temperature of the warm, moist air to cool, condensate, and eventually precipitate.

Depending on the strength of the developing cyclone, the warm conveyor belt will either continue to develop ahead of the (weak) cyclone or a portion of the warm conveyor belt will get wrapped around the backside of the cyclone…typically near and north of the 700mb low.  When the warm conveyor belt wraps around the cyclone, the cyclone takes on the typical “comma” shape often seen on satellite imagery and in text books.  This conveyor of warm, moist air aids in the development of clouds and precipitation on the backside of the low.  Precipitation resulting from this process is often referred to as “wrap-around” precipitation because it has wrapped all the way around the low.  Because this is occurring on the backside of the low, the surface temperatures are often falling as the surface cold front has most likely passed through.  Thus, in winter, snow is often found in the “wrap-around” precipitation.

Equally important in the life-cycle of a cyclone is the “Dry Slot”.  Unlike the warm conveyor belt that starts near the surface, the dry slot originates in the upper portion of the troposphere.  Here, cold, dry air begins to get entrained into the mid-level cyclone.  As mentioned above, cold, dry air is more dense than warm, moist air – which is typically found near the surface.  Thus, the cold, dry air aloft attempts to sink toward the surface.  As it does this, the cold, dry air encounters air at a high pressure and is forced to compress.  This sinking and compressing results in a relative warming and substantial drying.  This drying out of the atmosphere tends to supress precipitation development which also aids in the development of the comma shape often seen.  The dry slot is often found just south of the track of the 700mb low.

So what do the warm conveyor belt and dry slot have to do with when the storm turns north?  Well, the earlier the upper-low makes the northward turn, the farther north the 700mb low will track and a good portion of southern Oklahoma into western Arkansas will be “dry slotted” at the same time the temperature become cold enough to support snow.  Places that are dry slotted, in turn, would receive considerably less snow than places that remain in the warm conveyor belt.

So, based on the images above, what does the NAM forecast?

Southern Oklahoma and western Arkansas dry slot…

With this said, the warm conveyor belt is producing so much precipitation ahead of the dry slot that Oklahoma and Arkansas should still see over 2.5-3″ of liquid equivalent…before the dry slot overtakes them.  For Oklahoma, surface temperature would support much of this falling as freezing rain and sleet (to the tune of over 0.5″ of ice and 2-4″ of sleet!) while west-central Arkansas would see mainly rain (with a brief opportunity for some ice toward the end of the precipitation).  As for the places remaining in the warm conveyor belt?  This run of the NAM predicts over a foot of snow in a wide area from Tulsa westward to Edmond, Enid, Woodward, etc.

Remember, this is just one run of one computer model.  The forecasts continue to change as we learn more about the approaching short-wave trough (upper low).  The above scenario should be taken as 1 possibility out of many others.  In fact, another model run at the same time as the NAM predicts much more sleet across the aforementioned area, and does have as pronounced of a dry slot.

Only time will tell with this storm.  Please refer to your local National Weather Service office for more details for your specific area.

The upcoming winter storm (identified on yesterday’s post) is certainly what everyone is talking about in Norman.  In fact, at today’s Hazardous Weather Testbed (HWT) Map Discussion there were over 40 people in attendance – all focused on what would unfold later this week.  Tonight I thought I would take a look at what the Hydrometeorological Prediction Center (HPC) is currently forecasting.

Above is the current 5-day HPC precipitation forecast.  Almost all of the precipitation across the southern plains is associated with the shortwave trough forecast to affect the area Wednesday night into Friday morning, aka the potential Winter Storm.  Ultimately, the big question is what will the precipitation type of that amount be.  As a rule of thumb (and a starting point), 1″ of liquid is approximately 10″ of snow – however, a lot of things can affect that ratio.

Using the 10:1 ratio above, if all of the precipitation were to fall as snow, a good portion of the southern plains would be buried under almost 2 feet of snow.  The good news for Oklahomans and Arkansans is that all of this precipitation is not forecast to fall as snow.  So what then are meteorologists currently forecasting?  The next few images should help break that down as best as meteorologists can currently forecast.  It should be pointed out that the forecasts below are valid from 00Z Thursday (6PM CST Wednesday) until 00Z Friday (6PM CST Thursday), whereas the precipitation forecast above is valid through 00Z Sunday (6PM CST Saturday).  Thus, additional precipitation will fall after 00Z Friday (6PM CST Thursday) and most of that will fall in the form of some wintry precipitation.

Above is the probability of an area receiving at least 4″ of snow.  As you can see, HPC forecasters have greater than 70% confidence that a large portion of northwest Oklahoma, northern Texas Panhandle, and southern Kansas will receive at least 4″ of snow.

Above is the probability of an area receiving at least 8″ of snow.  Again, HPC forecasts have fairly high confidence that a good potion of the northern Texas Panhandle and northwest Oklahoma will receive over 8″ of snow.

Above is the probability of an area receiving at least 12″ of snow.   It isn’t too often you see confidence high enough to warrant a 70% area for at least a foot of snow this far into the future for a southern plains winter storm.  However, a large swath of 40% probabilities is nothing to sneeze at!  Remember, more snow will fall in parts of that outlooked area after the end of that time period!

And last, but not least, what about areas to the south and east of the heavy snow?  What does all that precipitation fall as?  Unfortunately, it appears that there is enough confidence to warrent a 40% probability that some areas will receive at least 0.25 inches of freezing rain.  This is the criterion for what is known as an Ice Storm.  Couple this much freezing rain with the expected strong winds, some places will most likely see power outages.

Editor’s note:  While I was writing this data from the latest (00Z, 6PM CST) run of the North American Model (NAM) began to come in.  This run is considerably slower and stronger than previous runs.  It holds off most of the heaviest winter precipitation until after 00Z Thursday (6PM Wednesday).  In fact, between 1.5 and 2 inches of liquid equivalent is forecast in the cold sector of the storm Thursday night.  The model forecasts this to fall as heavy, wet snow along the I44 corridor in Oklahoma.  This would easily result in over 12″ of snow in this portion of Oklahoma – again – if this model run turns out to be true.  A lot can (and will) change over the next few days.  Don’t get caught up in the details of a single run, just know that there is the potential for a major winter storm late this week!