The dream of the 90s is soon to no longer be alive in Pittsburgh. Or at least not at the Pittsburgh International Airport. This afternoon, the Allegheny County Airport Authority, which manages the airport, announced a $1.1 billion "Terminal Modernization Project". The major project will reconfigure the airport, which is currently split into separate "Landside Terminal" (check-in, security, baggage, and rental car facilities) and "Airside Terminal" (gates and most AirMall shops) buildings. Currently, post-security, you take an automated tram from one building to the other.

The new plan eliminates the separate Landside Terminal building entirely. All facilities currently in the Landside Terminal building will be moved to a large, new addition to be directly built onto a reconfigured Airside Terminal.

About that "dream of the 90s" reference? The current Pittsburgh International Airport opened in 1992, replacing "Greater Pittsburgh International Airport" (or "Greater Pitt" as it was sometimes locally referred to), located on the northeastern corner of the current airport property. It was built with a capacity of 100 gates, a size justified by its role as a hub airport for USAir (renamed US Airways in 1997). Several design decisions, beyond just the large gate count, were made because of its role as a primary hub. One of these include an interesting situation for international travelers where you would have to pick up your baggage the pass through customs, then re-deposit and have your bags re-screened. If your travel terminated in Pittsburgh, your bags would then be handled a second time by your airline to deliver them to the baggage claim in the Landside Terminal building.

After 9/11, US Airways faced some major financial difficulties after after unsuccessful negotiations, pulled out of PIT as a primary hub in 2004. Many gates went disused in the years following, eventually prompting the Airport Authority to erect walls to shut off sections at the end of some terminals that were completely unused. The "E" terminal, used for commuter aircraft and (somewhat ironically) attached to the Landside Terminal Building, was completely demolished in 2011, expect for a small area currently used as a second security checkpoint.

Another feature of the Pittsburgh airport lost after 9/11 was the novelty of being able to shop at the AirMall even if you weren't flying. This was most useful as a way for family or friends dropping someone off or picking them up to pass some time at the airport. However, post-9/11 security required all persons passing through security to have a valid boarding pass, eliminating this amenity to non-travelers.

The "modernized" airport should be a great upgrade to a growing, vibrant region. In addition to surely upgraded lighting and signage (which I'm actually almost most excited for, to be completely honest), the airport will be "right-sized" to a capacity of 51 gates, 11 more than the 40 currently in use. I'm assuming the large new entry will enable the possibility of more than one and/or larger security screening areas. PIT always seemed like a bit of an outlier compared to most mid-size to large airports by really only providing one security line. The elimination of the people-mover and reduced number of gates will hopefully make it faster and easier for people to get to their gate as well.

A modern entrance to the airport will also be a welcome facelift as Pittsburgh continues to re-invent itself in the 21st century. Naturally, some noise was made that this plan was announced just days after the region was cited as a potential spot for Amazon's new HQ2 headquarters. While I'm not entirely sure the two events are directly correlated, having plans to modernize the city's international airport surely won't hurt Pittsburgh's bid.

.@PITairport unveils $1.1B new terminal project. Leadership comments, "We could name the new terminal 'Amazon'." https://t.co/rRMiStLrGh pic.twitter.com/hCiGAhXZna

— Justin Meyer (@JustinMeyerKC) September 12, 2017

There are only two small things I'll miss with this new airport. First, it's not clear in the renderings released so far if the new entry building will keep something of the airfoil-shaped roof currently architected into the airport design. It's a neat element that would separate it from just the flat rooflines and glass commonly used in other airport.

Secondly, with no more people mover, who will welcome us to the City of Pittsburgh? (Currently Mayor Bill Peduto, and I believe someone else, provide voiceovers during the 90 second-long ride between terminals). Thankfully, the Airport Authority has confirmed Franco Harris and George Washington are staying put, and maybe in 2023 we'll still have a T-Rex to greet you with some Pittsburgh Pride.

Facts sourced from Wikipedia, NextPittsburgh, and PITTransformed.com

New airport renders courtesy of Allegheny County Airport Authority, Landside Terminal photo courtesy Wikipedia, T-Rex photo courtesy Boring Pittsburgh

The press, and likely your social media feeds, are abuzz over President Trump's decision today to leave the Paris Agreement, signed by 195 countries in December 2015. His decision will result in the United States joining only Syria and Nicaragua as countries that are not in agreement with this plan. Syria, mind you, is currently trapped in a civil war, and Nicaragua did not sign the agreement because it did not go far enough (the agreement is non-binding, meaning there are not penalties for failing to follow the agreement).

In a completely unexpected turn of events, however, the city of Pittsburgh ended up benefiting from generally positive PR in the wake of this announcement. Although I haven't lived in the Pittsburgh area since graduating college in 2014, I still consider myself a proud Yinzer and am always interested in following the new economy of the Steel City as well as how it is portrayed on the national and world stages. Thus, I thought it was necessary to bring some attention to this interesting development.

In President Trump's speech announcing the U.S. would be leaving the Paris Agreement, he stated "I was elected to represent the citizens of Pittsburgh, not Paris." Besides the fact that the only thing Paris has to do with the agreement is that negations and drafting of the document took place there, the quote rather immediately got the attention of current Pittsburgh Mayor Bill Peduto (D). He was quick to go on a mini-tweetstorm explaining that "Pittsburgh", the city, fully supports the Paris Agreement and will continue to follow the guidelines set forth in it:

The United States joins Syria, Nicaragua & Russia in deciding not to participate with world's Paris Agreement. It's now up to cities to lead

— bill peduto (@billpeduto) June 1, 2017

Fact: Hillary Clinton received 80% of the vote in Pittsburgh. Pittsburgh stands with the world & will follow Paris Agreement @HillaryClinton https://t.co/cibJyT7MAK

— bill peduto (@billpeduto) June 1, 2017

As the Mayor of Pittsburgh, I can assure you that we will follow the guidelines of the Paris Agreement for our people, our economy & future. https://t.co/3znXGTcd8C

— bill peduto (@billpeduto) June 1, 2017

His words were, as you'd expect, picked up and mentioned in many articles covering the story. However, other stories also emerged covering how Pittsburgh was a rather, if not very, poor choice of city when talking about leaving an agreement designed to have a positive impact on the environment. Pittsburgh has reinvented itself over the past 20-30 years and has a thriving service economy based in the healthcare, education, and tech sectors. Long gone are the days of smog so thick it literally killed people, as was the case in 1948. Today, the city is known for its beautiful views, great food scene, and vibrant culture and most Pittsburghers are proud of what the city has become.

This was noted in a Lifehacker article "What Pittsburghers Know About the Environment That Trump Still Needs to Learn" posted this evening. The local Pittsburgh Post-Gazette also captured a range of responses. Others made reference to the 13,000 clean-energy jobs in the Pittsburgh metro as of 2015. This Washington Post reporter summed things up quite well:

If we’re still talking Pittsburgh, I’d just add that the people who think it’s still a struggling “steel city” have not been there.

— Dave Weigel (@daveweigel) June 1, 2017

Support for leaving the agreement fell largely along party lines, but an indirect result was definitely a lot of coverage of the economy and current state of of the City of Pittsburgh. And I've yet to see someone cast the city as still stuck in a grey cloud of smog, literally or otherwise.

Article photo by Robpinion - Own work, CC BY-SA 3.0, Source

Winter 2017 is certainly picking up in intensity across New England. After a strong storm brought 12-18" of snow and 4+ hours of blizzard conditions to much of Massachusetts, another storm is is expected to drop 1-2 more feet of snow from northeast Mass to much of the Maine coast Sunday through Monday night. Then there's a chance at yet another storm later in the week.

The National Weather Service posted Winter Storm Watches for MA/NH/ME Friday afternoon, then on Saturday afternoon about 4:30 PM, upgraded the watches to Winter Storm Warnings (as expected) and issued Blizzard Watches along the immediate coastline.

Winter storm warnings are now in effect for all in pink. Blizzard watches from Maine to Nantucket. https://t.co/GIDW3xJABI #nh1news pic.twitter.com/TsBsw5g8Sx

— Ryan Breton (@RyanBretonWX) February 11, 2017

In 2017, there are many ways to get notified of these warnings, from the traditional local TV broadcast, to apps by major weather providers, to variable message signs along highways. Each of these can contain varying amounts of information. Obviously a highway sign conveys much less information than a human speaking with graphics behind them.

One solution I'm a big fan of are the weather alerts Google puts out. The alerts are well integrated into Android, the most popular smartphone operating system, and the webpage they link to is an excellent, concise tool for weather communication. When your current location is under a warning issued by the NWS, a card stating so will appear in the Android notification tray and on Google Now. A similar alert appears when you're browsing an area in Google Maps that is currently warned.

When you tap that alert on your phone or click "More Info" in Maps, you are brought to a page containing collection of information related to this Warning. This includes the original warning text, a map with the affected area outlined (all alerts are outlined in orange, not some other color scheme depending on what the alert is), preparation tips, the definition of what this particular alert is, and even tweets and news stories Google thinks is related to the alert.

The Winter Storm Warning Alert page currently in effect for coastal Massachusetts

Most weather apps simply provide the original warning text, and while the warning text often contain some calls to action, putting this information in context seems much more useful and more likely to elicit a response. Studies have previously shown than people often like to confirm weather information from more than one source or individual before reacting. If they were to come to this page and see a tweet from a local police department or a story from the local news station, that may be the additional trigger they need to take action. While I don't think you need to do much convincing for New Englanders to take a Winter Storm Warning seriously, these alert pages also exist for other events like Tornado Warnings, and could definitely be even more useful when minutes or even seconds count.

In short, these alert notifications and pages are a very effective means of weather communication. Most broadcast mets would likely agree that providing context and reinforcing call to actions is an important part of severe weather communication. Google Weather Alerts do so in a logically modern way.

In my group at The Weather Company, speed is of the utmost importance. We serve traders in the energy markets, so getting our customers weather data faster enables them to in turn make decisions faster than their competitors on the trade floor.

One of our major products is plotting weather model data on maps from all the major global weather models. One of these is the Global Forecasting System (or GFS) model, run by the National Centers for Environmental Prediction (NCEP) within the United States government. NCEP offers GFS model data for download in a handful of different spatial resolutions. As a result of the speed requirements in our products, our site offers graphics produced with the lowest (that is, coarsest) resolution data since that data will be available and process significantly faster than the full resolution data. We also produce maps made with this full resolution data, though, so that clients can see the finer details of this weather model output.

This background sets the stage for a situation I was tasked to explain earlier today. Hurricane Matthew is currently churning in the middle of the Caribbean Sea and is expected to make impact as a major hurricane to Jamaica and Cuba in a few days. I was asked about a situation where our graphics seemed to be showing two different centers to the Hurricane as it approached southeastern Cuba depending on the resolution used:

These graphics shade 1000-500mb Thickness values and contour Mean Sea Level Pressure (specifically at Forecast Hour 84 of Friday's 1200z run). You can see the lowest pressure of Hurricane Matthew is much more centered among the concentric isobars surrounding and is located just northwest of the eastern shore of Jamaica in the high resolution data, while the lowest pressure is much more "off center" and almost the entire length of Jamaica (about 200km) to the west in the low-resolution data.

My first thought when analyzing why there might be a difference in the plots was that we were plotting Mean Sea Level Pressure. MSLP is a derived field that corrects for terrain when measuring atmospheric pressure. Air pressure decreases as height increases (due to gravity), so the surface pressure at the top of a 6000 foot mountain will be significantly lower than the surface pressure along the ocean shore. By reducing surface pressure to MSLP, meteorologists can compare pressure readings anywhere on earth.

As it turns out, the reduction of surface pressure to MSLP was significant in the difference between the graphics, but indirectly, as the source grid (raw surface pressure) had significant differences between resolutions to begin with.

Below are the surface pressure fields (in hPa) from the 0.25° GRIB data (first image) and the 2.5° GRIB file (second image)

Many more contours (isobars) are present on the 0.25° image. On that image, you can see concentric circles of surface pressure isobars located over the water between Jamaica and Cuba that represents the center of Hurricane Matthew. However you also see even lower pressure in a tight gradient right along the southern coast of Cuba, and also towards the center of Jamaica. Why is that? Those "low pressure" minimums are due to topography.

As seen in this Google Maps topographic map, there are steep mountains right along the coast, the same spots where the lowest surface pressures were modeled. (Many of the peaks along that mountain ridge are 4000+ feet in elevation, or >1200 meters.) At a high resolution, where grid points are only 0.25° longitude (~28km) across, there's plenty of grid points to represent both this change in elevation and the tropical system offshore. On a 2.5 degree grid (where each grid point is ~275km wide and 275km tall), there are many fewer grid points that cover the same geographic area. As we see here, because of how tall the terrain is in southern Cuba, it actually has a dominating effect on the surface pressure field, so you end up with a stretched minimum isobar that essentially encircles the mountain range, while the tropical cyclone is essentially reduced to nothing.

As expected, when you then apply the reduction to mean sea level pressure, the high resolution MSLP data (first image below) neatly accounts for the terrain and forms concentric isobars only around Matthew. Meanwhile, while there is some correction to account for elevation when the low-resolution MSLP field is calculated, it is still significantly stretched and the lowest mean sea level pressure value is found around the single center of lowest surface pressure, which is not aligned with the center of Hurricane Matthew at all.

(In these plots, colored contours are again surface pressure, while the black contours are MSLP, both in hPa)

Hopefully this article may help you to understand the difference between surface pressure and Mean Sea Level Pressure, and some of the pitfalls that may arise when viewing or visualizing model data. It was an interesting and deeply analytical problem that had me snugly wearing my "meteorologist hat" this afternoon.

Quick Links: Project Website | Source Code on Github

I recently published an interactive map that shows the current US Drought Monitor levels and the 5-day National Weather Service Quantified Precipitation Forecast (QPF). As stated on the site, "by overlaying these two datasets, you can analyze where drought conditions may improve or worsen over the next 5 days."

Background

As of Sept. 8, 2016, all of Massachusetts was under some level of drought or dryness

There were two primary reasons that led to the creation of this site. The first is that I had been reading about the Mapbox GL JS framework and wanted some sort of project I could build using it. Secondly, living in eastern Massachusetts, I was well aware of the worsening drought that has gripped the state for effectively the past 2 years. While attending a Maptime Boston meetup back in July where there was a discussion about intersecting and overlaying different datasets, I came up with the idea to create this product.

Technical Details

The US National Weather Service provides a significant cache of its products in common geographic formats. They provide their QPF formats in both Shapefile/KML (vector) and GRIB2 (raster) formats, so it's relatively easy to use, no matter what your application. The Drought Monitor data, hosted by the National Drought Mitigation Center is also available in a variety of formats, including Shapefile.

Mapbox GL JS requires vector data in GeoJSON (as it's effectively the standard for transmitting geographic data over the internet today). I found a Node module shp2json which can easily convert shapefiles to GeoJSON. Since the Mapbox documentation states that URLs are preferable to passing the framework a GeoJSON-compatible object directly, I decided it would be best to serve the data as endpoints in an Express.js app.

I wanted the site to always display the latest data available, however since I was going to host it on Heroku, that presented a small challenge, since I couldn't just set a cronjob to go fetch the new data and save it somewhere. Instead, I decided to determine if new data should be available whenever a request was made for it. It was easy to determine this based on timestamps (defined in the dates.js module). Since it would take a short but noticeable amount of time to download and convert the data to GeoJSON, I also decided to cache the datasets in a Redis database. Redis does have support for complex and geographic datatypes, but for the purpose of this project, just storing the GeoJSON as one string seemed sufficient. Because I'm running on the free Heroku Redis tier, I was sure to purge any old data from the datastore when inserting new data.

Working with Streams

This project was a great opportunity for me to learn more about Streams in Node.js. As with the event-driven model Node uses (and perhaps because it's built on it?), streams can be a bit challenging to understand at first. But they are certainly very powerful when used correctly!

For example, perhaps the code snippet I'm most product of in this project fetches a drought monitor data .zip file and converts it to a GeoJSON stream (in this project's case, the HTTP response):

6 lines, excluding whitespace. No need to store and extract the zip file first. Neat stuff.

Certainly, I'm no master of streams (or Node in general), and one current inefficiency in my code is that I download and convert the data twice - once if there is a cache miss (the data is not in Redis), and then a second time to get the data as a string and store it into Redis. A good future enhancement would be to split the GeoJSON output stream to both return as the HTTP response and save into Redis simultaneously.

Mapbox GL JS Client

I appreciated the Mapbox GL API's design to separate layer sources and styling, and also how easy it was to implement custom styling. Both datasets I used have distinct color tables that are used almost universally wherever the data is displayed. Thus I wanted to stick to these color schemes on my site. It was super easy to make a color table array and use that to both style my map and populate my legend! I also appreciated it's responsive design; it made adding my query panel (see below) that works just as well on a smartphone as a desktop a simple addition.

Conclusion

This might be an overly-detailed explanation for a simple one-page map site. But it was an interesting project that exposed me to a number of new tools and ways of thinking and I think the results are quite cool! I showed it off to one of my meteorology friends who quickly started moving the layer sliders and understood the utility of overlaying drought and precipitation forecast data.

This post was originally published on psuchase.org.

Generally, severe weather season in Pennsylvania occurs later in the year than it does out in the Plains and other parts of the country, with the peak of severe weather season usually occurring in mid- to late-June, climatologically speaking. However, if there's one day of the year when memorable damaging weather seems most common in Pennsylvania, that day would be May 31. Here's a sample of four severe weather events that have all occurred, mostly in Western Pennsylvania, on May 31sts.

1889: Johnstown Flood

Today marks the 125th anniversary of one of Pennsylvania's most historic and notable disasters - the Johnstown flood (known locally as the Great Flood of 1889). A large storm had dumped nearly 6 to 10 inches of rain, the previous day according to the US Army Signal Corps (still responsible for a large portion of meteorological observations in the 1880s). The rain overnight was enough to cause severe localized flooding, with the Conemaugh River to already nearly overcome its banks by daybreak. Meanwhile, 14 miles upstream, local residents were concerned by the amount of water that almost overwhelming the South Fork Dam. They worked throughout the morning to try and free the clogged spillway to reduce pressure on the dam, but their best efforts were still futile and less than two hours after they abandoned their efforts, the South Fork Dam collapsed at 3:10 PM. A telegraph had been sent from the town of South Fork to Johnstown twice, trying to warn officials of the impending dam break, but these messages were ignored, as previous warnings had proven to all be false alarms.

The torrent of water, now heading downstream with flow rates approaching that of the Mississippi River, according to recent studies, rushed its way towards Johnstown, destroying houses and businesses in the towns of South Fork, East Conemaugh, and Woodvale. The flood reached Johnstown nearly an hour after the dam broke, causing catastrophic damage and creating a pile of rubble at the Stone Bridge downtown covering 30 acres and 70 feet tall. In all, 2,209 people lost their lives in the Great Flood of 1889 and the flood caused more than $17 million in damage. It was the first major disaster to be served by the newly-formed American Red Cross.

For more, check out the Johnstown Flood Museum (online and/or in person!)

Source: Wikipedia (Photo: ExplorePAHistory.com)

1985: PA's Worst Tornado Outbreak and Only F5 Tornado

The worst tornado outbreak to ever strike eastern Ohio and Pennsylvania occurred on May 31, 1985. A ripe environment in place throughout the day finally gave way to storms by late afternoon. The National Severe Storms Forecast Center (the precursor to today's Storm Prediction Center) issued a tornado watch at 4:25 PM. Within half an hour, a tornado was spotted near the PA/OH border near Erie county, moving into Albion, PA, where nine people lost their lives. At 6:30 PM, a tornado touched down in Portage County, Ohio and quickly strengthened into a powerful twister. By the time it crossed into Pennsylvania near the town of Wheatland, Mercer County, it was nearly a half-mile wide F5 monster.

Tornadoes destroyed buildings and lives on a stretch from nearly Erie to Beaver, PA, and by the time all was said and done, 65 individuals had lost their lives in Pennsylvania alone. Another tornado tore across central Pennsylvania on that day, leveling more than 90,000 trees in the Moshannon/Sproul State Forest. A debris ball was even visible on the primitive WSR-57 radar in State College with so many trees being tossed around in the air as the tornado ran a track parallel to Interstate 80 sixty-nine miles long.

For more information on this outbreak, I highly recommend the book Tornado Watch Number 211 by John G. Fuller.

Sources: Wikipedia | Pennsylvania Highways Feature (Photo source: WCVB slideshow)

1998: A Derecho in Northern PA, Tornadoes Rock Many Other Counties Across the State

On May 31, 1998, another very unstable environment was present across Pennsylvania. With a warm front lifting northward and the jet stream present overhead, deep, strong shear was present, along with high levels of instability. An F3 tornado touched down in Somerset County and stayed on the ground for 15 miles, stretching up to half-a-mile wide at times. The tornado struck the towns of Salisbury and Pocahontas, destroying a furniture factory and house, a significant part of the $4 million total in damages. Tragically, there were 15 injuries and one fatality - a 13-year-old girl killed when a tree struck her car.

Tornado damage in Somerset County, PA

In the eastern part of the state, an F3 tornado caused more damage to many homes in Berks County. Several more weak tornadoes touched down across the state on May 31, 1998, with touchdowns recorded in Elk, Chester, Lackawanna, Luzurne, Lycoming, Montgomery, Pike, and Wyoming counties. Non-tornadic damage was severe as well, with a bow echo flattening trees and causing damage across Pennsylvania's Northern Tier (not far from where the F4 tornado struck in 1985), leading to significant damage in Williamsport, PA as well.

Source (including photo): http://www.angelfire.com/pa/pawx/053198/053198.html

2002: Microburst Causes Fatal Damage at Pittsburgh Amusement Park

A cold front associated with a strong, mature low pressure system pressed through the Pittsburgh region on the evening of May 31, 2002. As the line crossed through Allegheny County, a microburst occurred near West Mifflin, near the amusement park Kennywood. The Whip, an 84-year-old ride at the park, collapsed when the microburst struck, killing one woman and injuring 54 others. This storm was subsequently tornado warned in Allegheny and Westmoreland counties, but no tornadoes were confirmed from this storm. Regardless, a significant amount of wind damage was reported across southwestern PA.

It’s been a long time since we’ve posted some new content to the site, but in honor of Leap Day, let’s leap across the country to do some real basic Skew-T analysis in central California to check out the chance for snowfall later this morning.

Note first that elevation plays a huge role in the type of precipitation for the region we’re taking a look at today. The point that I had my Skew-Ts centered around is located at an elevation of approximately 2100 feet ASL. Go a few hundred feet up in elevation, and it’ll most certainly be cold enough for snow to fall, at least in the layer directly above the surface. Go down several hundred feet and snow is much less likely to occur. As we’ll soon see, there are some very interesting divergences in the models we’re looking at here.

What we’re going to be focusing on primarily in this post is reading part of the Skew-T Log-p diagram to determine (or at least take a stab at) precipitation type. The “Skew-T” diagram is one of the most useful, but rather intimidating tools in meteorology (it’s basically a super-convenient atmospheric calculator). To read more about the Skew-T diagram, click here! Now let’s start analyzing some weather!

All of the model soundings in the diagram are centered around the point linked to above and are for 15z, or 7AM PST.

Skew-T Image	Discussion
	The RUC model is a higher-resolution model that runs every hour. What we see on this diagram from the 05z run is that the surface temperature is solidly above freezing (freezing being the line starting at 0°C on the bottom axis and moving diagonally up to the right). The temperature (and wet-bulb and dewpoint temperatures) all cut across the isotherms (lines of equal temperature) as height above the ground increases. This means the temperature cools off rather quickly as height increases. However, because the surface temperature is well above freezing and there is a fairly tall layer (from the surface to around the 875mb height) above freezing, I would say the most likely surface precipitation type based on this model would be rain.
	The NAM is a short-term forecast model which runs four times a day. This diagram was generated based on data from the 29 Feb 00z run. It is much more interesting in terms of precipitation type for several reasons. First, the surface temperature is almost spot on the freezing mark. The temperature profile is markedly different the first several hundred feet above the surface compared to the RUC output. While this model also is predicting precipitation at 15z, the temperature profile remains ever so slightly below freezing up to about the 860mb level or so (approximately 1300m above sea level) before decreasing much more rapidly with height. I would expect snow showers from a profile like this, and find this scenario to be somewhat believable.
	The final output displayed here is from the GFS model, which is a global forecast model which also runs four times a day. Unfortunately, the latest data I was able to access while writing this article was the 18z run, which is rather outdated (in the world of weather forecasting). On this diagram, the surface temperature is just above freezing, while there is a slight warm pocket noticeable immediately above the surface. This warmer pocket (an inversion, really) contributes to more interesting frozen precipitation types. In this case, because the “warm” layer is so shower, I would be inclined to say that this diagram is indicating mostly snow falling, with some pellets of sleet mixing in, possibly.

The Skew-T diagram, no matter how useful, is only one tool in the forecaster’s toolbox. As you can see, different models can create very different scenarios for the same location at the same time. That’s part of the challenge of winter weather forecasting! If I were to make a prediction for this setup (difficult considering I am rather unfamiliar with the area), I would go with snow as the most likely precipitation type. It may mix with some rain showers (and possibly sleet, though I find that to be more unlikely). It’s important to note that the Skew-T diagram is by no means likely the easiest, nor the best, way to get a snowfall accumulation prediction. Here, we’re simply examining the profiles to get a sense of what sort of precipitation type is most likely. Perhaps we will tackle the accumulation problem in another blog post. Hopefully for now, you have gained some new appreciation as to one of the most powerful tools in meteorology – the Skew-T log-p diagram!

Thanks to Andy M. for providing today’s question!

For the first time since December 2006, a tornado has touched down in Westmoreland County, PA. This was a very impressive storm which, unfortunately, caused significant damage in Hempfield Township – including the Fort Allen neighborhood which was damaged by the December ’06 twister. This very severe storm began to intensify quickly as it entered western Westmoreland County around 4:25PM on March 23, 2011. The National Weather Service in Pittsburgh observed a tornado vortex signature on Doppler radar at 4:29PM and, correspondingly, issued a tornado warning for central Westmoreland County eight minutes later at 4:37PM.

Some have argued that this was an abnormally long delay between observing the tornado on radar and issuing a warning. Do note a few items in this case, however. First and foremost, the Pittsburgh area does not see tornadic conditions all that often. While I’m sure there have been more (false alarm) tornado warnings issued since 2002, there have only been 7 dates when tornadoes were confirmed in Southwestern Pennsylvania, according to the Tribune-Review. Thus, one could argue that NWS meteorologists in the Pittsburgh office may not be as “experienced” in issuing Tornado Warnings. They also may be a more conservative office, wanting to attempt to minimize the number of false positive tornado warnings. There was also some precedence for severe weather events in Westmoreland County on Wednesday, as the entire region had been under a Tornado Watch since 1:15PM and the county had been severe thunderstorm-warned for a solid 20+ minutes before the tornado warning was issued. Finally, it is important to note that the worst of the damage occurred in Hempfield Township, right where the green arrows are in the below picture (Figure 1). The strongest radar echoes (colors on the radar) were still a few miles west of the area of greatest damage at 4:42PM, and in fact, the preliminary Storm Prediction Center tornado report was not until 4:45PM. This means residents of the Fort Allen area (and surrounding communities near Hempfield High School) had approximately 10 minutes of notice from the Tornado Warning and even more time from the Severe Thunderstorm Warning. Finally, the warnings obviously did their jobs regardless of timing, as no serious injuries were reported.

Figure 1: 4:42 PM Reflectivity and Velocity Radar images (dead image link)

Four minutes later, at 4:46PM, extremely strong rotation was reported on the velocity mode of the Doppler radar. (You can read more about how velocity mode works here, but for now, know that strong values of red and green right beside each other [known as a “couplet”] are very strong indicators of a tornado.) The center of circulation, as represented by the green arrows on the radar image below (Figure 2), appears directly over west-central Hempfield Township area, right near Hempfield Area High School, which sustained significant damage in the storm – including roof damage to the auditorium and damage to the athletic fields. The numbers on the radar image correspond to the figure numbers of the other images listed below and their approximate locations. This should help to frame exactly what the tornado looked like around 4:50PM. Figure 3 really speaks to me about the potential strength of the tornado, as the twister appears to be taking on a classic “stovepipe” shape.

Figure 2: 4:42 PM Velocity mode Radar image with labels

Figure 3: A funnel cloud is seen over Route 30 West in Hempfield Township

Figure 4 (left): Funnel cloud observed from Arona, PA; Figure 5 (right): Funnel cloud observed at the UPS facility in New Stanton, PA

There were a few other very interesting observations from this storm. The first of which was the massive amount of large to “giant” hail that fell during the storm. There are numerous reports of hail at least golfball sized, causing damage to many vehicles. I have seen hundreds of photos of the hail, some of which were actually approaching baseball-sized. Figure 6 below was taken by my brother and shows the interesting cross-section of a hailstone that I’d approximate is about 1.25″ in diameter. For more details on the countless hail reports, check out the Storm Prediction Center Storm Reports page. Finally, it is interesting to note that Mike Smith, founder of WeatherData Services Inc, actually blogged about the hail from this storm as it was occurring, explaining the very strange occurrence of a “hail spike” on radar as the storm moved into Somerset County.

One last radar image I’d like to share is a not-so-commonly seen view of radar (see Figure 7): 3D radar! This radar image enables meteorologists to get a vertical view of a storm , somewhat like an MRI a doctor might have performed on your leg to “see inside” it. As stated on the image, this storm was a “Classic Supercell” – another example of a term you never hear in conjunction with the location Greensburg, Pennsylvania! You can see both the mesocyclone complex on the back edge of the storm (as labeled) as well as the hail core directly in front of it. To be honest, I never imagined I’d be looking at a radar image like this over top my hometown!

Figure 6: A cutaway view of hail, approximately 1-1.25" in diameter

Figure 7: A 3D radar view of the supercell over Greensburg, PA at 4:50 PM (dead image link, maybe someday I'll run the volumetric data and restore this)

As noted earlier, this storm did cause significant damage to several dozen homes in central Hempfield Township and also the to the Hempfield Area High School. Many photographs of the damage show roofs partially or completely missing. Some buildings also sustained damage to exterior walls, including the dramatic photo shown above (courtesy of KDKA-TV). According to WTAE.com, this storm has been ranked by the National Weather Service as an EF2 (or “strong” tornado). The tornado was also observed to be about 300 yards at its widest point. Based on damage I have seen, I predict this storm will be rated, at a minimum, a strong EF1 tornado, but more than likely an EF2 tornado. I will update this story with the official results when they are made available.

Wednesday’s storm was easily the worst storm to strike Westmoreland County since at least June 1998 when a significant tornado outbreak occurred across western PA. Although I was watching from the Joel Myers Weather Center at Penn State, I can still say that I have never seen a storm this severe in Westmoreland County. The damage is simply incredible. For one final look, please check out the videos posted below.

Sources

Data for this article came from many readily-available sources including WTAE.com, KDKA-TV, and the Tribune Review, as well as the National Weather Service. The radar images (Figures 1, 2 and 7) come courtesy of Texas Storm Chasers, while the storm damage photos (Figures 3-5) come from WTAE.com’s uLocal member-upload site and Facebook page.

For more details on this storm check out the Tribune-Review article, KDKA article and videos, and WTAE article.

Videos

Update: As of 4:00PM on Thursday, March 24, the National Weather Service in Pittsburgh has confirmed the tornado in Westmoreland County as an EF2, or “strong tornado”. More details on the tornado track may still be forthcoming.

The past few days have been very interesting in terms of weather. A massive storm has moved across the country, bringing anything from blizzard conditions to severe ice accumulation. In Latrobe, one of the results was a strong, but very brief thunderstorm that rumbled through Wednesday morning about 8:40AM. The storm surely looked impressive on radar, especially by February standards!

8:45 AM Radar from Intellicast

The storm was recorded in a special METAR observation. Normally, these automated observations occur around 50 past the hour, as the image below shows. However, if particularly “interesting” (as the system recognizes it) weather occurs, it can also make a special observation. Such was the case here. The circled portion of the report describes the active weather at the time. It reports light rain associated with a thunderstorm, and ice pellets. Some of the other details, not shown on that chart, include “observed” occasional cloud-to-ground lightning. There’s quite a bit of information you can get from this single automated report!*

One other thing apparently with even that chunk of weather observations in Latrobe is that the cold front passed through at that moment. This can be deduced in several ways. First, we see that the temperature peaks at a whopping 50 degrees at 8:41AM (surprising in February for sure). The temperature begins a pretty [rapid] decent after that point. There was also a wind shift, as winds changed slightly from the WNW to the WSW after that observation. Finally, the pressure reached a minimum at the time of that observation and began to rise again afterwards. A pressure minimum is a surefire sign of frontal passage.

This frontal passage is confirmed on a Weather Underground** surface map from 8:32AM. An occluded front passed through the region at that moment, and the combined effects of warmer, moist, and unstable air led to a very interesting (and unusual) February thunderstorm.

* - The full coded METAR read: KLBE 021341Z 29011G16KT 10SM -TSRAPL SCT038 BKN047 OVC075 10/10 A2956 RMK CB MOV ENE OCNL LTGCG

** - ** – Speaking of the Weather Underground, they just launched an overhauled version of their website yesterday. As always, there has been considerable debate about the design, but I tend to like it. Go check it out for yourself!