Helping travelers identify ticket flexibility

We’ve added an optional airline selector to the inputs widget: if the user specifies an airline, we’ll check for applicable waivers along their itinerary. While airlines have meaningfully improved general rebooking flexibility by eliminating change fees for most tickets, a fare difference can still inhibit rebooking in the face of interruption. These waivers unlock an additional level of flexibility to navigate yourself around potential disruption.

We’re manually collecting waivers to minimize investment in an unproven feature, but with signs of traction will automate this process1. Until then, there’s likely to be latency in adding them to our database (or even some missed waivers).

From a former SITREP author with ❤️

We also now allow users to specify a single airport, rather than origin and destination (and layover). We think this is especially well-suited for operators, who we hope will use this to save time while writing briefs and SITREPs.

Accordingly, we’ve encouraged GPT to output longer responses when summarizing just one airport and will likely explore further prompt differentiation (like permitting it to use more technical language) in the future.

Latest version is available at close-in.aerology.ai—blue skies and tailwinds!

But as the above-linked post suggested, we’ll mostly be using this space to update readers on what we’re building. So what’ve we been hacking on this last week? We think it’s something that’s especially relevant as travelers start to look ahead to their Thanksgiving return flights. Give us the airports you’ll be traveling from, to and via; and we’ll distill our much-loved forecast discussions from the National Weather Service using OpenAI’s GPT-4 Turbo.

Some readers might draw parallels to earlier iterations of this newsletter, when we were writing mostly about weather. While those posts could be quite practical, we struggled to provide ample coverage relative to all the airports a traveler might transit (and by the time we had finished writing, some information could already be stale). This distillation delivers much of the utilitarian value that our outlook posts provided, with the benefit of encoding the most recent information and a geographical focus tailored to your travels. You could almost consider it an on-call Tim.

We’ve experimented some with how we prompt the model, but—here especially—we welcome feedback: a broad cross-section of outputs will help us more scientifically tweak how we guide the model in its task. We’ve asked it to overweight the contents of the forecast discussion’s aviation section, avoid jargon or highly technical concepts, and highlight uncertainty. If you’re seeing summaries that are not consistent with this prompt engineering, please pass it along! Finally, we’ve initially designed this product with the mobile experience prioritized, but as we iterate on this new tool alongside our months-ahead forecasts, we’re likely to deliver some desktop improvements.

A beta version is available at close-in.aerology.ai; no credentials are required. Smooth travels!

Tim has also strayed further from Excel in the last few months, tempted more and more by chatGPT and Jupyter notebooks. So he spun up a quick Prophet model (and of course drew from the posterior predictive distribution to keep it probabilistic). Survey says? Perhaps most impressively, it gives the Sunday after Thanksgiving nearly a 1 in 5 chance of reaching 3 million screened. It also gives that Sunday about a 50/50 chance of setting a new all-time record (it even thinks the Wednesday before has about a 1 in 8 chance to reach 2.885 screened).

But while we’re liable to spend a paragraph or two on goings-on around the industry, for the foreseeable future, this newsletter will be about what we’re building. To that end, you can expect us back in your inbox a bit more frequently—probably just under once per week. What’s this first update hold? Let’s get to it.

Giving mobile users some love

Our first version was designed with desktop users in mind1; users could access it on a mobile device, but the ported-over desktop layout was cramped (enough so that usability was impaired, especially on smaller phones oriented in portrait).

With this update, mobile users will see more elements distributed vertically in the search section. And once they click get chances, they should benefit from relatively wider prediction visualizations [such that mobile users should horizontally scroll within the results section].

We expect to continue optimizing the at-booking layout for desktop then remold a serviceable mobile experience (though Google Analytics could prove this assumption wrong). But looking ahead, as we develop a next-day/day-of product, we expect to optimize that layout for mobile users.

Multi-airline itineraries are ‘Coming soon’ no more

Version 1.0.0 included a multi-airline switch, but—if we’re in the trust tree—it was really just there to provide some visual balance. Users would discover it was disabled, with a ‘Coming soon’ tooltip at least teasing its future availability. Version 1.1.0 leaves less to the imagination in this regard, with users able to create an itinerary that features different airlines for leg 1 and 2.

That leaves our Stops selector as the only decorative element, though if users were to think the layout’s visual order guides our task prioritization… we wouldn’t discourage them.

Heads up about that itinerary you’re considering!

Last but not least for this update, we’ve built the plumbing for alerting that’s conditional to the itinerary a user is evaluating. For the moment, we only have two alerts loaded: (1) when you’ve selected a date that is likely beyond airlines’ most recent major load and (2) if you’re transiting SFO when 28L is liable to be closed next year.

We’d love suggestions on other events that you think travelers should be advised of before hitting book—particularly ones our model might not “see.” SFO’s 28L closure is a good example in this respect, as it hasn’t learned the impact of these construction-induced capacity reductions.

As always, general feedback appreciated! Latest version is running at at-booking.aerology.ai.

Because that’s the thing: most OTA’s and aggregators prognosticate about the direction of prices (even if you ignore it). And increasingly, price forecasts are presented alongside emissions estimates. So you’re guided towards the most sustainable option with the most economical timing. But what about the most reliable option? Frequent fliers can recognize a red flag in a routing, though a behavioral economist would wince at the most well-tuned mental model of the national airspace system. (To say nothing of the infrequent flier.)

Layovers and it’s-gonna-be-tight arrival times can be a gamble; we think travelers should be better informed about the odds. To that end, we’ve developed an algorithm that answers how likely is an itinerary to get you there when advertised. More technically, we built a density network. It’s an AI architecture that taps the power of neural networks to estimate the range of possible outcomes—and the likelihoods within that range.

In our case, we’re aiming it at the distribution of delays, a problem where uncertainty abounds even with a prediction horizon of a couple hours. Is your aircraft going to require an unplanned tire change? Is your pilot going to sleep through their alarm? Your catering truck? Could get in a fender bender. Passengers? Pesky things are one airport beer too many away from having to be deplaned. Plus a ripple effect. And that’s when you have a reasonably good idea what the weather is doing and where your plane is coming from!

At a months-ahead horizon—when weather forecasts are more coarse and less skillful—it’s all the more important to frame flight outcomes in terms of chances, not single estimates. Within a practical range of over-blocked, tailwind-aided 60 minutes early to might-as-well be cancelled 24 hours late, no exact delay minute is all that likely. While it might not feel like it, arrival delays gravitate towards negative values (read: early arrival); but even within this relative peak, the single most likely outcome (10 minutes early) only happens about 2.5% of the time.

One approach to dealing with this uncertainty is binning several delay minutes together (see below), rather than treating each delay minute as its own unique possibility (like the above chart does). It makes for an easier prediction task: you’d be trying tease apart fewer outcomes with higher likelihoods. If you bin arrival delays into less than 30 minutes, between 30 and 60 minutes, between 60 and 120 minutes and greater that 120 minutes—like Amadeus does—the bin with the lowest frequency (120+ minutes) is more probable than our most probable single value (-10 minutes exactly).

But easier for the forecaster doesn’t mean more usable for the traveler. The bins mask some material differences. Maybe you can absorb as much as 90 minutes delay, but beyond that you’ll miss the start of the meeting. Or even more consequentially, maybe your 65-minute layover is doable with a 30 minute delay but hopeless with a 55 minute delay.

Our density network is well suited for capturing this uncertainty with much higher resolution, i.e. distinguishing between the 30 and 55 minute delay. Let’s see what that looks like.

Get chances

Consider somebody flying from Syracuse (SYR) to Nashville (BNA) on the Tuesday before Thanksgiving. They’ve found two options with similar travel times that allow for a close-to-full day of work.

They’re a frequent enough flyer to vaguely know EWR’s reputation, but 95 minutes seems generous? And they’re seriously averse to getting in the door after midnight—so the 40-minute earlier arrival should provide a better chance to be in the Uber by 11:45 p.m., right? They switch from their Expedia tab to Aerology’s forecast, input the details for the via EWR option and hit “Get chances.”

Under-the-hood, we’ll estimate the distribution of arrival delays for the SYR-EWR flight and the distribution of departure delays for EWR-BNA, then randomly sample from those distributions. That first sample might “draw” an arrival delay of -6 minutes and a departure delay of 2 minutes; given the scheduled layover of 95 minutes, that first simulation nets out to 103 minutes of available connecting time. Then we do it 9,999 more times and visualize 100 representative outcomes. By default, simulations that provide less than 30 minutes to connect are considered missed and, in the case of our BNA-bound traveler, we forecast they’ll be left with an insufficient layover duration about 9% of the time1.

And while a missed connection is a good bet to screw up your schedule, there’s likely some amount of delay to your originally scheduled arrival that you can’t afford. For this reason, we’ll also sample from our estimated distribution of arrival delays to your second leg (EWR-BNA, in this setup). Here, by default, we’ll assume you can only absorb about an hour of delay; for our BNA-bound, the default is about right with respect to getting in the door by midnight.

Because we don’t need to net out the layover or a second delay for this part of our forecast, you can also use it as a workaround to one our design decisions. To maximize applicability, we’ve geared this initial version towards one stop itineraries; it may be a bit hack-y, but if you want to glean disruption chances for a non-stop itinerary, you can treat the second leg as your non-stop (i.e. use the layover airport as your origin). You’ll need to invent a first leg and, depending on the fictional first leg, the misconnection and headline chances should only be referenced for a laugh. But because our forecast of the second leg’s arrival delay is independent from the first leg, the too late forecast is representative of your non-stop’s chances.

That last sentence is an important distinction. The too late forecast does not ask what are the chances I arrive by this time, given that I make my connection. It provides the chances that the flight arrives by that time—whether you’re on it or not. It’s also why the headline number (12% for our SYR-EWR-BNA option) may not equal the sum of misconnects and too late: there’s some too-late outcomes that could have still resulted in a misconnect. The headline asks what are the chances that I misconnect or arrive too late; it could also be framed as what are the chances I make my connection and arrive on time-ish.

To contextualize that 12% disruption probability, it’s about the same as parents having three children of the same sex. But we think our forecasts are best employed in bunches. It provides better context than thinking through the boy/girl mix of your friends and family. So our BNA-bound traveler opens another tab (… our design road map takes note), inputs the details for the via-ATL option, adjusts the too-late threshold to 11:30 p.m., and finds something like an 8.1% chance to mis-connect or arrive too late. Despite a scheduled arrival time that’s 40 minutes earlier, the over-EWR option is nearly 50% more likely to leave our hypothetical traveler short of their door by midnight.

🤨 And why should I trust you guys?

It’s a fair question. A single prediction can be hard to judge on its own, especially when it’s of the percentages variety. For that reason, we took great care to create a test set numbering more than 4.5 million flights that we can use to evaluate our models’ performance. Because we’ve approached this problem probabilistically (i.e. we’re not making a single estimate but rather estimating likelihoods within a range of possible outcomes), one of our most important metrics is the degree to which forecasts are well-calibrated. If they are well-calibrated, it means an outcome actually occurs at roughly the same frequency as we predicted.

For each flight in our test set, we can segment our predicted distribution. For example, calculate the minute where we’d say there’s a 35% chance that actual delay will be less than or equal to that threshold—then we should find, with a well-calibrated forecast, that actual delay falls into this segment about 35% of the time. In the case of the predicted segment where we’d say there’s a 35% chance for delay less than or equal to that minute, 34.5% of actual values fell within the segment.

We tend to be a bit over-confident at lower probabilities, which is to say these events actually happen less often than we forecast. For the segment where we’d predict 15% of delays will fall in, just 5% actually do. Oppositely, we exhibit some under-confidence for probabilities higher than 50%. But, overall, we think our forecasts performed quite well—certainly well enough to underpin an MVP. It’s an open beta, so no credentials required and feedback welcome!

The primary culprit was a piece of energy seemingly meandering around the upper atmosphere. Its slow-moving nature combined with ample moisture to antagonize New York City airports from Saturday to Tuesday; with associated convection focused across northern New Jersey, already-disadvantaged Newark (EWR) was especially snarled. This proximity no doubt contributed to relatively more challenging conditions at the airfield, where it produced 109 mentions of thunder in EWR METARs. (Across the river, “only” 18 mentions at JFK and 12 at LGA.) But more critically, EWR relies on the airspace above northwestern NJ to a much greater extent for ingress and egress than JFK or LGA.

This unsettled weather pattern has likewise upset the tenuous relationship between the DOT-led FAA and airlines. Given that United’s New York City franchise leans heavily on EWR, CEO Scott Kirby has seemingly volunteered to command this particular skirmish for airlines. In a nominally-internal email, Kirby told employees that the FAA had failed them; calculatedly, we’d bet, he would soften his tone2 towards Administration leadership a few paragraphs later. We agree with Scott’s tonal pivot: the controller shortage is years in the making and we don’t fault current FAA employees (especially not the one that can pull the plug on our SWIM subscription). But underneath the headline-grabbing language, he started to pull at some interesting threads. 

What’s a thunderstorm worth in terms of capacity?

One should expect thunderstorms to reduce airport capacity by some amount—and there’s a lot of history to glean how much weather “should” pressure capacity. However Scott would suggest that FAA staffing constraints depressed capacity below weather-attributable depths. For their part, the FAA hasn’t refuted this insinuation, though notably only a pair of Sunday evening operations plans included a mention of ZNY staffing. So let’s concede staffing contributed to even-lower rates on Sunday; there were still eye-watering ground delay programs on Saturday, Monday and Tuesday. If staffing dragged on capacity for the duration of this 4-day event, that seems like a more defensible hill for Scott to occupy. 

We’ve cross-posted this with our friends at AirlineGeeks. While we’re not trying to sell anything, it feels like we should make the boilerplate disclosure: we may earn some advertising revenue if you click on the link. (So don’t not click on the link.)

Absent additional mentions of ZNY staffing, we think our modeling can help to detangle non-weather influences. To wit, we prefer to aim our forecasts at the future, but occasionally they can help to contextualize past performance. And importantly, since our last product update, we’ve evolved from point-estimates to probabilistic predictions. So what does that look like?

Let’s consider the first hour of the Tuesday GDP as an example (i.e. 13:00L, or 1 p.m. ET, on the 27th). In the final forecast we’d make for the hour, we’d predicted capacity would be 35.3 aircraft per hour: this is our best estimate of what capacity “should” have been when controlling for weather (and runway status—all of which were open during the times in question). But historical capacity data is noisy and, even when we make our final forecast at t-1h, there’s still some weather uncertainty.

For that reason, we want to focus not on our point-estimate of capacity, but the range of possible capacity outcomes. To this end, we’ve visualized something like a tropical storm’s cone of uncertainty for each hour. We’d expect approximately 50% of actual values to fall within the blue bar, while the remaining 50% of values would be split evenly above and below the bar. For that 1 p.m. hour on Tuesday, when accounting for weather and runway status, we put chances at approximately 50% that the actual value would fall between 30 and 39. (While not visualized, our point-estimate would generally fall towards the middle of this bar.) We’ve also charted the actual, recorded arrival rate with a red dot (36, in the case of 1p Tue).

You can think of each hour as the roll of a 4-sided die, where:

• rolling a one yields low-end, below-our-uncertainty-bar capacity for that hour,

• rolling a two or three results in expected, within-our-bar capacity,

• and rolling a four returns a high-end, above-our-bar arrival rate.

For any given hour, you have a 25% chance of getting low-end capacity—hardly a remarkable event. Even across 8 rolls (read: hours), you have a better than 1 in 10 chance of getting 4 or more low-end rates. We’d be reluctant to conclude there’s some influence other than weather, runway status and noise at play in the wake of one ugly afternoon/evening. Maybe that’s why Scott qualified his remark that the reduction in rates was almost certainly a reflection of staffing constraints.

Who’s right?

But we can decrease variability by rolling the die more times, i.e. increasing sample size. We’ll set aside Sunday, when there was some self-acknowledgement of staffing constraints, and consider just the Saturday, Monday and Tuesday ground delay programs. Those three days provide a sample of 24 hours during which a GDP was active; a low-end, below-bar arrival rate was recorded 14 out of these 24 hours. If you were to roll our 4-sided die 24 times, what’s the chance that you get 14 or more low-end rates? On the order of 1 in 2,000. At those odds, we’d take a bet there’s some factor—most likely FAA staffing—influencing capacity that our model can’t see. 

Share

Could Scott (and, more recently, Joanna Geraghty from JetBlue) use this math to credibly claim that the FAA has underperformed during this event? Those would be his words, not ours. Regardless of blame-casting, these shades of capacity are material: given the compounding nature of delays (like compounding interest, but decidedly bad), delivering even one or two more aircraft per hour would seriously knock down EDCT delays.

But much like we’ve attempted to separate the hand dealt to the FAA (again, bad… like “storm motion of 5 knots” bad) from how they played it (arguably poorly), the same question should be asked of the airlines. Air traffic delays of 3 hours are an unquestionably weak hand to be dealt; even so, there should be some expectation as to how the airlines execute on those cards, however lousy. Unfortunately, we haven’t reached the mile marker on our data science roadmap that would help to diagnose that. That said, when we’ve erected our model stack, we’d wager that a re-forecast supports a credible counter-claim from Secretary Pete.

And if I don’t care who’s right but I’m flying on Friday?

As for Friday, high pressure remains in control for the New York metro area, though another slow-moving low pressure system lurks across the Great Lakes. This low will swing a warm front through the Ohio Valley, with thunderstorms firing ahead of this boundary. Metro terminals look to remain dry, though convection could sneak into eastern Pennsylvania late day to exert some pressure on capacity. Using that same cone of uncertainty approach—but around air traffic delays—we think the 75th percentile of hourly average delays tops out at 58 minutes (in the 7 p.m. hour).

But for travelers booked on United on Friday, they’d be advised to pay closer attention to FlightAware than the National Weather Service. United is racing to get their operation fully recovered for this key test, though if they’re successful, it will be by an uncomfortably narrow margin. After canceling approximately a quarter of their mainline operation on Tuesday and Wednesday, their Thursday cancelation settled at 18%. As of 11 a.m. ET, there’s 218 cancels on the books for today, though—of all the gambles we’ve made herein—we’d be most confident betting that more cancels are forthcoming. For one, we suspect humpty-dumpty EWR has not quite been put back together; and there’s likely some recovery taking place related to yesterday’s irregular operations at DEN as well as ORD; and DEN figures to get hit again today. It’s just a matter of whether it’s a trickle or tranches. If Friday’s cancel count grows by less than 15% from here, we’d say they held the line. Accordingly, United’s hoisted waivers for EWR, DEN and ORD.

But! We’re not here to prognosticate about what travelers face this weekend. At least not directly. Today’s post is filed under our favorite category: Product launch.

We sent our first such post in November, wherein we shared our airport arrival capacity predictions.

Aerology

Introduce? Unveil? Announce? Anyways, we've got our first deep learning model to share

We’ve been writing for about a year now and when a post veered into more quantitative waters, it was generally underpinned by Tim working in Excel. Don’t get us wrong, Excel is great—communication ✅, trust ✅, physical intimacy ✅, curiosity ✅, conflict resolution ✅. But its suitability for blog posts doesn’t extend to machine learning and ML engineering…

Read more

4 years ago · 7 likes · 2 comments · Tim Donohue

Somewhere between the Excel sex joke and discussion about our forecast error reduction, we wrote in that product launch:

Admittedly there’s an interpretability hurdle for travelers, especially… For now, we’d recommend travelers pair our [capacity] predictions with the FAA’s airport arrival demand chart (AADC), which helpfully encodes hourly arrival demand by way of the stacked columns. It’s an important piece of information—and not yet piped into Aerology—to triangulate capacity, demand and more-interpretable delay.

Capacity is generally measured in arriving aircraft handled per hour; it’s a sneakily technical concept that some operators don’t entirely grasp. We instructed users on how to manually perform that triangulation but knew it was a tall task for travelers. And even if users gleaned some sense of delay from our capacity predictions, any insight was more or less binary: They could determine if unpublished delay existed or not, but how much delay remained obfuscated.

Minutes, not arrival rates

So the next mile marker on our data science roadmap was a pipeline to gauge demand; that was followed by an algorithm to translate that enigmatic demand overage to more-interpretable delay minutes. With a minimally viable pipeline and translation program built, we’re excited to launch our first delay beta. We think it represents a step change in usability.

Ergo, we don’t think its necessary to spend a few paragraphs instructing users on how to combine this product with other information. But we should define what type of delay we’re predicting (and pin a few qualifiers to these forecasts). In service of a definition, let’s review: When arrival demand for an airports’ runway(s) exceeds its capacity, air traffic control will explicitly or effectively assign arrivals a landing slot. It’s these landing slots we’re aiming to predict—after which we can measure delay relative to some planned landing time. For starters, we’re aggregating the resulting delays; later we’ll reproduce how airlines swap slots among their flights.

We want to highlight a few corollaries to this definition. It reflects air traffic dynamics at the arrival airport—not an imbalance in en-route airspace1 or at the departure airport. So we're not trying to assimilate any delay that occurs at a flight's departure gate or during taxi to takeoff. (Yet.) Similarly, these predictions stop short of modeling any taxi-in delay; nor do they simulate how this landing delay can ripple through the network. (For now.)

That’s a lot of things it’s not. But the type of delay it is—let’s call it airport arrival imbalance delay—was responsible for at least one-third of arrival delays at Core 30 airports in 20192. With a lot of delay types to eventually account for, we think this a good place to start.

It’s also analogous to the delays resulting from ground delay programs and ground stops3 that the FAA communicates via their NAS dashboard. We’re frequent visitors to their dashboard, but Google Trends suggests we’re in a pretty small minority. FlightAware, however—which had more site visitors than Geico in January—has done their part to promote awareness of these delays. They repackage FAA data: when we grabbed the FlightAware screenshot, the NAS dashboard coincidentally featured an active BOS ground delay program with 73 minute average delays.

Delay minimization and you

So why go to some lengths to predict what the FAA already shares? Because of the FAA’s wait-and-see approach to administering these delays. On average, ground delay programs are published less than 60 minutes4 before the first effected flight lands; delays from ground stops are effective immediately with no notice. It's a rational, if optimistic, tack from the FAA. By assuming demand won't materialize or the weather forecast verifies favorably, they avoid creating unnecessary delay. But their wait-and-see approach betrays airlines' ability to plan and squanders rebooking options for travelers.

For the moment, Aerology isn’t charged with minimizing delay. We “just” fit known demand5 to our best, unbiased estimate of capacity. Our delay modeling shares our capacity predictions' 16-hour forecast horizon; flights scheduled or estimated to land inside of 16 hours may be delayed beyond the horizon. Like our capacity predictions, we've used Newark (EWR) to prototype when modeling delay—though our infrastructure and modeling readily scales with customers. And we’ll initially re-forecast delay hourly—around the 6th minute of the hour after the capacity model runs—though could refresh more frequently with the latest demand information.

We’re ultimately trying to identify the still-lurking downside. To that end, we encourage you to focus on the ‘Predicted Unpublished Delay’ series; for flights scheduled to land in a given hour, this answers how many minutes of delay, on average, arrivals can still expect to absorb. We’ve also included average published delays to contextualize the situation—are we predicting 15 additional minutes of delay on top of 44 or 4 minutes of already-posted delay?

Delay, disruption and denominators

If you're comparing our delay predictions to the averages provided by the FAA—whether directly via their dashboard or piped to places like FlightAware—you’ll find they diverge in cases where the FAA geographically partitions flights or airlines implement cancellations. In the case of the former, the Command Center may exclude trans- or mid-continental flights from delay assignment; because we spread predicted delay across all arrivals scheduled to land in an hour, our per flight amount could be lower.

In the case of the latter, that slinking downside is somewhat amorphous. In response to FAA-assigned slots, airlines will occasionally cancel flights to mitigate the ensuing delays6. This, too, is on our data science roadmap; however, until we’re reproducing this behavior, we’re liable to over-predict unpublished delay at longer lead hours. In the interim, as airlines work through their schedule reductions, we’ll re-forecast delays: Absent the cancelled demand, our unpublished delay predictions should drift down to reflect the amount of delay allocable to still-operating flights. While we re-forecast with the latest demand, the FAA continues to broadcast the delay associated with their original advisory.

And when the Command Center originally runs a ground delay program, average delays can be eye-watering. But after the fact—with the cancellation dust having settled—average hourly delays exceeded 60 minutes just 2% of the time in EWR7. So, until we're predicting cancellations, we'd bet on a schedule reduction when unpublished delay approaches 60 (or even 40) minutes.

In this way, you can think of our predictions as unpublished disruption—cancellations might reduce average delays, but they still leave planes, crews and passengers out of place.

Like our capacity forecasts, we’re launching a decidedly beta version of delay predictions. If you’re curious about what you’ve read and can put up with some bugginess, reach out to tim@aerology.ai: We’ll get you set up beta access!

Though February betrayed some weakness, yesterday’s TSA throughput (2.50 million travelers) compared very favorably to President Day’s Friday in 2019 (2.51 million screened—just a 0.1% difference).

So we figured we dust off our template to take a peek at this weekend’s return travel. Go travel was hampered by ground delay programs at EWR, LGA and LAS (plus a DCA ground stop as well as some ski and Mexico AFP’s), but hardly impassable. Thankfully, Sunday and Monday look to be relatively tranquil across the national airspace system.

Sunday, February 19

Because it doesn’t “see” the holiday, our Holt-Winters model isn’t particularly useful in forecasting traveler volumes this weekend. But we can still glean some patterns from last year (and 2020): We’d estimate between 2.20 and 2.31 million travelers screened.

Offshore high pressure continues to steer moisture towards the Pacific Northwest and Northern Rockies. While not part of our template, we’ll note that probabilities for snow top out at 1% for Seattle (SEA), per the short-range ensemble forecast.

ORD has 127 flights scheduled to arrive in the 6 pm hour1 against an optimal arrival rate of 114; this has been the case since mid-January and EDCT’s have only been required for 5.2% of scheduled arrivals in this hour (in line with the rest of the day2). Somewhat similarly, DCA has 33 scheduled arrivals in the 9 am and 9 pm hours and the Runway 19 operation tops out at a 32 rate; with winds forecast around or above 10 knots and veering from south to west through the period, we’d expect a 19 operation to prevail, though for any delays to be generally imperceptible.

Monday, February 20

We estimate TSA will screen between 2.47 and 2.49 million travelers, which would make for a busier day than the '22-'23 holiday high-water mark (2.44 million from Dec 163—though there was certainly a depressionary effect from Southwest’s meltdown).

Low pressure tracking across eastern Canada will send a cold front across the mid-Atlantic Monday evening: there could be some spotty showers ahead of the frontal passage, though moisture is limited.

Up next?

We should have a product update to write about in the next couple weeks! We’ll also try to write a couple outlooks for Spring Break travel.

Arctic air has already surged into the Plains behind a cold front destined for the Texas Gulf Coast. As the front sinks southeastward, an area of low pressure will form along this boundary then rapidly strengthen as it translates eastward. The center of the low looks to track across the Great Lakes before setting minimum pressure records on the Canadian side of the border.

This system will antagonize air travel in several ways; unfortunately its varied impacts happen to be spread across some of the busiest holiday travel days.

Snowfall

If there’s some relatively good news to be gleaned from this narrative, it’s that no hub faces a paralyzing amount of snow. We’ll have to watch Chicago in this respect, but the forecast discussion indicates 2-5” is generally expected (with lake enhancement focused on northwest Indiana). Even so, while the snow totals aren’t particularly impressive, the bulk of any amount could fall in a relatively short period Thursday afternoon into evening.

Elsewhere, small and medium sized airports on the Great Lakes (e.g. BUF, CLE, GRR) could see locally heavy totals.

Extreme Cold

The most anomalous impact from this event figures to be record-setting cold, with wind chill advisories spanning from the Canadian border to Texas Gulf Coast. The High Plains could see wind chill values as low as minus 70. Practically, this depth of cold can result in inoperable conditions in “just” light snow—hopefully eroding any sense of security the Snowfall section might have instilled. The efficacy of anti-icing fluid is materially reduced when temperatures are below 0°F (and especially below -13°F): Given any intensity of snowfall, it may be impossible to complete the deicing process and taxi to takeoff position before requiring another spray.

Even in the absence of frozen precipitation, typical arrival and departure sequences are drawn-out as agents exposed to the cold (e.g. baggage handlers, fuelers) will need to be rotated inside frequently.

Winds

The very tight pressure gradient—owing to high pressure building behind the front and rapidly deepening low pressure along the front—will promote gusts to 50 mph across the central/northern Plains as well as Great Lakes. Just a few inches of snow can create blizzard conditions from blowing snow, so don’t seek solitude in snowfall headlines.

Ahead of the system, strong southerly winds and rain—at times heavy—hardly preclude Northeast airports from this mess. (As winds veer to 220-250°, east-southeasterly winds, for which NYC is particularly ill-suited, are mercifully forecast to occur pre-dawn on Friday.) We’re also curious about coastal flooding impacts to JFK and BOS associated with Friday morning’s high-tide.

Thunderstorms (for good measure)

As if all of that weren’t enough, a line of low-topped convection may form in the vicinity of Baltimore-Washington Friday morning along with the frontal passage.

So how have airlines responded to the above? For starters, they’ve pretty well blanketed the country in waivers.

Your fare difference is (probably) waived

United seems to have led the pack in terms of timing, breadth and communication. They were out with their first waiver on Monday and follow-on waivers eventually covered 142 stations (more than two thirds of their domestic portfolio); they’ve also gone as far as to tweet about the availability of these waivers.

For their part, Delta has not been quite as liberal in their application, extending coverage no further south than STL. Nonetheless, a still-impressive 64 airports are covered by Delta waivers. American has proven stingiest, with 57 airports included in two waivers (and, notably, no coverage for the Intermountain West). In all cases, the rebooking window is already open.

Your original flight is (probably) disrupted

Airlines are prone to pull down regional schedules with a pretty broad brush; oppositely, they’ll hang onto international operations (where reaccommodation can take days) and hub-to-hub flights (that serve as an outlet) as long as possible. But with multiple hubs impacted, travelers might not have any more success connecting via hub 2 than hub 1 (and airlines can’t readily loan employees between hubs).

Share

It remains to be seen whether airlines’ schedule reductions will slash demand to levels below severely-impaired airport capacities. Regardless, between cancellations, air traffic delays, some suspensions of operations and network effects, we wouldn’t bet on many smooth itineraries. Rebook if you still can!

But since we last sized-up the regional pilot pool, it was revealed1 that Republic had petitioned for an exemption to the 1500-hour rule, several carriers raised pilot pay rates and the FAA denied Republic’s proposal. So when we woke up to a Monday morning email from Mesa IR about delayed earnings, we figured this would be a good week to circle back on the topic. We’ll approach this post with a slightly different tack: Rather than estimate the depth of the pilot shortage, we’ll take a stab at some financial projections. We think it’ll make for a more interesting analysis insofar as we’re putting ourselves out on a branch. Admittedly, our previous pilot population estimates are tough to verify; our accounting forecast will be proven right or wrong in short order.

What news do we think might be lurking in Mesa’s yet-released earnings? Let’s outline some drivers and assumptions, starting with revenue:

• We’ve backed out $7.1 million from the Contract Revenue line—this represents lease revenue associated with rent GoJet pays to Mesa for 20 CR7s. It leaves $111.8 million as implied CPA2 revenue for the quarter ended Jun ‘22; we’ve taken this amount and divided by flown block hours for that quarter ($1,761 CPA revenue per block hour).

• We’ll hold that $1.7k CPA revenue/block hour constant as forward-looking block varies. We used block amounts from the Dec 5 OAG snapshot and believe selling schedules through March3 are reasonably representative of airlines’ intentions. To account for DHL flying not in OAG and not-completed block, we’ve given OAG block hours a 5.8% haircut (equal to quarter ended Jun ‘22 flown block as percentage of OAG block).

• We’ve carried pass-through and other revenue forward from the quarter ended Jun ‘22.

While unit revenues for CPA flying are flat by design, tumbling block hour capabilities convey decreasing total revenues. Setting aside the DHL/completion factor for the moment, block hours for Jan 2023 are set to be down 39% year-over-year. GoJet lease and pass-through revenue buoy the top-line somewhat, but we still estimate total revenues for the quarter ending Mar 31, 2023 will be down 10% year-over-year.

As for costs:

• Mesa was among the above-mentioned carriers who raised pilot pay rates and their bump was advertised as “industry-leading.” We’ll turn to always-handy Form 41 data, where Piedmont’s $920 pilot and co-pilot expense per air hour was highest among the regionals for the quarter ended Jun ‘22. Mesa was at the bottom of this range, coming in at $407; to jump all its peers calls for a 126% increase in pilot wages.

• Pilot wages fit into the Flight Operations expense line, which also encompasses flight attendant and dispatch personnel (as well as pubs and lodging). Using the same Form 41 data, we estimate pilot wages made up 47% of the line for the quarter ended Jun ‘22; we believe a 126% increase in pilot wages will flow through to a 60% increase in Flight Operations line per block hour by the quarter ended4 Dec 31, 2022 (equal to $1.1k; like CPA revenue/block hour, we’ll hold this constant while varying forward-looking block).

• We assume the remaining operating expense lines5 are fixed and are carried forward at their amount for the quarter ended Jun ‘22.

So what do we have? Mesa presumably raised pilot pay with the expectation that it would attract a greater share of pilots and therefore increase its block hour capabilities. But such a strategy works with a lag and in the interim there appears to be a precarious squeeze occurring. Until block hours can find a bottom, revenues continue to sink; at the same time, variable unit costs have increased owing to the bump in pilot pay and fixed costs are spread across fewer block hours. Given the above assumptions, Mesa swings to a loss of $10.7 million for the delayed-quarter ending Sep 30, 2022. And the projected loss widens to $35 million for the first two quarters of calendar year 2023.

With an estimated $80 million in operating losses on deck, we’re left to wonder if Mesa can make it to the other end of the block-hour-capability-tunnel. We’re a blog6 and this is a back-of-napkin exercise, so we’re going to take the liberty of ignoring non-operating costs (e.g. loss on investments, interest expense, taxes) and cash flows (i.e. cash flows from investing and financing activities). But this back-of-napkin math is something of a red flag: Mesa had $58 million in cash and cash equivalents on hand at the end of Jun ‘22.

There are some green shoots in terms of block hour capabilities. While there’s still time for cuts, Feb ‘23 and Mar ‘23 are both set to notch month-over-month gains in daily block hours; this comes on the heels of month-over-month declines in 9 of the 10 preceding months. But bouncing off a block bottom will only beget narrowing losses initially—by our math, Mesa would need to nearly double block hour output to breakeven. Given all the same assumptions as above, flown block would need to equal 103k hours to achieve an operating profit. To contextualize this, Mesa flew 116k block hours in the quarter ended Dec ‘19, though they did so with 17 more aircraft in-schedule. Absent taking on additional aircraft rent (or finding an interested party for their CRJ-200’s), they’d need to push aircraft utilization to 9.2 hours per day. Not an inconceivable level, but we wouldn’t bet they have the runway to reach it.

Pre-pandemic, it was tough to prognosticate about travel patterns for holidays that are not anchored to a particular weekday; changes to traveler behavior resulting from COVID make it a doubly challenging exercise. So we’re not going to try! Instead we’ll look to airline capacity and commentary.

Tomorrow actually features the most domestic capacity of the period, though Thursday the 22nd edges it out in terms of total capacity (by 2,600 seats—less than one tenth of one percent). American shared that they expect the 22nd to be their busiest day; for their part, United expects January 2 to be their busiest day. We’ve penciled in bookend outlooks covering travel on Dec 22 and Jan 2. As for tomorrow… while not tranquil, low pressure tracking north from the Carolinas will at least deliver rain7 to the coastal Northeast (in contrast to the interior Northeast, where the system will produce up to 2 feet of snow).

7

Valid for 24 hours ending 1am Sat EST.

So as Tim wrote, he dropped hints that his co-founder, Yamaç, was working on a proper ML product. And today we’re really excited to launch a public beta for our first model.

What is it? And what is it not (yet)?

We trained a deep learning model to predict airport arrival capacity, deployed it to the cloud (so that it can infer in real-time) and built a simple frontend that visualizes the results. If you’ve been reading our Substack for a time, you know that we have a complicated relationship with Newark Airport (EWR)—so naturally we chose to start there.

Aerology

The one about EWR

Disruption at Newark Airport (EWR), oft-featured in our air travel outlooks, has been spotlighted after comments from United CEO Scott Kirby during the company’s first quarter earnings release. Kirby (straining to remain calm by his own account) responded to a question about capacity and demand at the airport by saying…

Read more

4 years ago · 4 likes · 5 comments · Tim Donohue

While we’re curious if airports can be featurized—such that the same algorithm outputs predictions for multiple airports—even by itself, our EWR model is trained on a billion data points.

Given how frequently we reference weather and its various models, you’ll likely be unsurprised to learn that ingesting NOAA’s high-resolution rapid refresh (HRRR) is a critical pipeline. There’s tradeoffs (refresh rate, time steps, subscription cost, ensembling), but we think it’s an appropriate solution for an MVP: it allows us to re-forecast capacity once per hour (around the 12th minute of the hour) and provide predictions in hourly time steps out to 15 hours in the future.

For the moment, the model assumes all runways are open, though we’re a couple days away from switching on the pipeline that will read runway status from NOTAMs. We’re also working on the pipeline that ingests actual arrival rates for a window leading up to our prediction. We’ll share on Twitter (provided it’s still around) when these pipelines are up and running.

Follow us on Twitter

As for the model itself? Well that’s sorta the secret sauce, but we will share that it’s based on the same transformer architecture that’s made headlines recently as it relates to generative AI.

If readers aren’t surprised by the initial focus on EWR or the HRRR’s prominent role, they might be surprised to see point estimates of arrival rates. After all, we’ve spilled a fair bit of virtual ink to convey uncertainty… bets, odds, chances, probabilities are oft-used words. So why the single, best estimate of capacity and not a more probabilistic product (e.g. communicating low- and high-end rates)? We’re evolving our model architecture to better quantify the uncertainty around predictions, but Yamaç only has so many Jupyter notebooks. We’re just as excited to iterate as we are to launch and to do the former—with the benefit of feedback—first requires the latter. To that end, we wanted to launch as soon as it contained some value that otherwise doesn’t exist; we think the point estimates satisfy that criteria.

In the meantime, we’ve charted the three most recent initializations to convey some sense of uncertainty. A relatively wider spread in predicted values for the same valid time (e.g. the 22:00 hour above—it’s in UTC) should be interpreted to mean there’s greater uncertainty. Conversely, more tightly grouped initializations (e.g. 17:00 hour) suggest higher confidence on our part.

Cool. What can I do with this?

Admittedly there’s an interpretability hurdle for travelers, especially. For both operators and travelers, we think probabilistic predictions will significantly increase utility. For now, we’d recommend travelers pair our predictions with the FAA’s airport arrival demand chart (AADC), which helpfully encodes hourly arrival demand by way of the stacked columns. It’s an important piece of information—and not yet piped into Aerology—to triangulate capacity, demand and [more-interpretable] delay. Low capacity by itself isn’t necessarily problematic; it’s only when capacity falls below demand that queueing (read: delays) results.

The black horizontal line forecasts the same arrival rate as our dashboard—but reflects the FAA’s outlook. Essentially, we’ve challenged ourselves to forecast rates set by the FAA more accurately than the FAA themselves. Fortunately for the prospects of our burgeoning company, indications are we can do so with a good degree of success. We ultimately want to benchmark using a deployed model, though need to finish the arrival rate pipeline (and collect a sufficiently large sample); in the interim, a trained-but-not-yet-deployed model looks to reduce forecast error by as much as 44%1.

Much of this outperformance likely owes to the FAA's wait-and-see approach to ground delay programs, or GDPs. From where the FAA is sitting, this optimism is largely defensible—it minimizes the chances that they’ll introduce unnecessary delay to the national airspace system. But shifting that decision right—nearer the potential capacity/demand imbalance—costs operators optionality in how they distribute delay (if metering traffic does prove necessary). And loss of operator optionality also flows through to travelers, whose best rebooking alternative might depart before the delay is published.

After sending two or three outlooks per week since August, we think this beta launch is a good opportunity to re-evaluate Substack cadence and content. We want to be metrics-driven and the shorter, more frequent outlooks didn’t really grow subscriber numbers. Now that we have a product we’re iterating on, that figures to provide some previously-inaccessible content. And we might return to the in-depth outlooks for peak travel dates (or author a post on macro trends). So we’re not entirely sure what Substack looks like for us going forward, but we’ll still be writing.

Aerology is, for the moment, unconstrained by this wait-and-see gamble; we’re just making our best estimate of capacity, unbiased by optimism. (And our probabilistic predictions will enable a more scientific, tactical wait-and-see approach… call it hedge-and-see.) So, if the AADC denotes “No GDP”—like the screenshot above—and Aerology’s forecasted rates differ substantially from the FAA’s outlook, we’d bet they’re waiting for some trigger (perhaps demand to exceed some threshold or an updated weather forecast) to publish reduced rates.

Take, for example, the 19:00 GMT hour in the two charts above. Aerology had most recently predicted a rate of 38.8 at 13:00 GMT (which would have been visualized at 13:12); forecasted capacity for the 13Z hour trended generally downward from the previous two initializations (39.2 to 39.6), though not so erratically as to convey high uncertainty. The FAA, for their part, was publishing a 50 arrival rate at the time. Arrival demand was estimated at 40 (less a forecast and more an accounting). If predicted capacity is below both the FAA’s forecasted rate and estimated demand, it’s an indicator that unpublished delay exists.

This can also be the case if a GDP is active and suggests a downward revision to the program may be forthcoming—which would increase delays. If, on the other hand, Aerology’s predicted rates are higher than the FAA’s outlook, we might see an upward revision or cancellation of the program. But delays are sticky. Passengers scatter, agents are reassigned, gate changes are made: It’s the reason the FAA and airlines are inclined to publish a delay only when they’re nearly certain it will occur. Expect the FAA to be quicker with downward revisions than rate increases (or cancellations).

Now the exciting part

We think there will be a step change in traveler interpretability—and therefore utility—when we’re providing probabilities that capacity will be lower than currently-published rates. And another step change when our output is not just capacity, but also delay. However, iterating takes weeks (in the case of probabilities) or months (to reach delay predictions) and we have a product to share today! So comment below or email Tim (tim@aerology.ai) and we’ll get you set up with a username and password; the dashboard is an otherwise publicly accessible website (nothing to download, no VPN required). We intend to leave the beta open for a few weeks, after which point it will be, at least partially, moved behind a paywall.

And it’s just that: A beta. So please expect (and report!) any bugginess. To that end, don’t limit feedback to bug fixes—what would improve the usefulness of the product? As we referenced above, feedback from users is critical to successful iteration. (Perhaps particularly helpful beta users get an extra week of beta access.)

🚀,

Yamaç & Tim

If you have questions about the content of this outlook, the answers might be in this post (we keep adding to it). Otherwise, let us know if something doesn’t make sense (we welcome feedback of all kinds!).

Sunday, November 6

We forecast TSA will screen 2.281 million travelers (± 0.5σ, or a prediction interval of about 38%, is 2.22-2.34 million travelers).

A cold front stalls west of NYC, though shower chances top out around 20% for the metro area—and only briefly1. While we normally wouldn’t concern ourselves with humidity and temperature, the marathon prompts us to take a peek (it also happens to mark one year since our first outlook). To that end, a deep southerly flow will (unfortunately?) promote near-record warmth2, with a forecast high of 71° (and with dewpoints in the low- to mid-60’s, it should feel at least a touch humid too).

While the cold front also stalls west of CLT, rain chances are slower to taper off from west to east. Daytime heating may also support a rumble of thunder, though strong/severe storms are not expected.

Monday, November 7

We forecast TSA will screen 2.119 million travelers (± 0.5σ, or a prediction interval of about 38%, is 2.06-2.18 million travelers).

The cold front that stalls west of NYC receives a nudge Sunday night into Monday morning; as it pushes east across the region, a few showers could reach the metro area during the morning (though like Sunday, coverage is better to the northwest).

A warm front is dragged northward through the DFW-area: cloudy, drizzly conditions ahead of the front will give way only to showery conditions post-frontal passage.

Share

You can check out3 hourly estimates in this workbook.

1

If you have questions about the content of this outlook, the answers might be in this post (we keep adding to it). Otherwise, let us know if something doesn’t make sense (we welcome feedback of all kinds!).

Thursday, November 3

We forecast TSA will screen 2.210 million travelers (± 0.5σ, or a prediction interval of about 38%, is 2.15-2.27 million travelers).

A slowing system should start to produce snow for the central Rockies tonight. As it moves off the mountains late Thursday, at least some snow should mix in with rain. The slowing trend has conveyed somewhat higher snowfall forecasts—albeit with considerable uncertainty—given that more of the precipitation will fall during the cooler evening hours. NWS Boulder puts chances of measurable snow (i.e. at least a tenth of an inch) for DEN at 87%; probabilities for at least one inch of snow are 61%1.

On a flight schedule-related note, ORD looks to be settling into a November schedule with peak scheduled arrival demand of around 106 flights. This is notable because they spent much of October around 117, which is above their optimal rate of 114.

Friday, November 4

We forecast TSA will screen 2.253 million travelers (± 0.5σ, or a prediction interval of about 38%, is 2.19-2.31 million travelers).

The same system that produces heavy mountain snow on Thursday will send a cold front across the Plains on Friday. For the DFW-area, convection will ignite along this front during mid- to late-afternoon. Apparently disorganized clusters of storms may initially develop, though these should quickly merge to form a more-organized line of storms; conditions are fairly supportive of severe weather.

Despite the thunderstorm hatching around ORD, NWS Chicago omits mention of thunder from their forecast.

Share

You can check out2 hourly estimates in this workbook.

1

If you have questions about the content of this outlook, the answers might be in this post (we keep adding to it). Otherwise, let us know if something doesn’t make sense (we welcome feedback of all kinds!).

Sunday, October 30

We forecast TSA will screen 2.415 million travelers (± 0.5σ, or a prediction interval of about 38%, is 2.35-2.48 million travelers).

Despite thunderstorm hatching around ATL, the forecast discussion indicates instability is marginal and any isolated thunderstorms should become more showery as they translate east.

Perhaps more notable is the start of the IATA Northern Winter 2022/2023 season. On Sunday, the transition looks to briefly peak EWR’s scheduled arrival demand in the 4 p.m. hour at 46; this relaxes only somewhat and appears to settle in around 44 for the rest of the month. More broadly, the retraction of COVID-related slot usage waivers has stoked NYC scheduled demand to the tune of 5.4%1, including 9.8% at EWR.

Monday, October 31

We forecast TSA will screen 2.229 million travelers (± 0.5σ, or a prediction interval of about 38%, is 2.23-2.35 million travelers).

Again the thunderstorm hatching may be overdone—this time as it relates to CLT, where the forecast discussion suggests precipitation is more likely for the higher terrains of western NC.

Ahead of a weak cold front, some showers are possible for NYC, though most models keep the area dry.

Halloween2 reductions have trimmed scheduled demand across our 10 automated airports by 6.2% versus the following Monday (and 7.8% versus the previous Monday).

Share

You can check out3 hourly estimates in this workbook.

2

We asked Open AI’s DALL-E for “digital art of dieselpunk skeleton sitting in gothic airport terminal.”

If you have questions about the content of this outlook, the answers might be in this post (we keep adding to it). Otherwise, let us know if something doesn’t make sense (we welcome feedback of all kinds!).

Thursday, October 27

We forecast TSA will screen 2.388 million travelers (± 0.5σ, or a prediction interval of about 38%, is 2.33-2.45 million travelers).

A cold front sinking south across the DEN area during Thursday morning will promote showers across the foothills and urban corridor. There’s some question as to whether areas east of I-25 (e.g. DEN) will be cold enough to support snow. For its part, the Short-Range Ensemble Forecast (SREF) keeps probabilities around 20% for the first half of the day; the forecast discussion carries “dusting to trace accumulations” language.

Friday, October 28

We forecast TSA will screen 2.441 million travelers (± 0.5σ, or a prediction interval of about 38%, is 2.38-2.50 million travelers).

Thunderstorms spread from West Texas into the DFW area during early Friday; warmer, moister air will, meanwhile, nose into the area from the southeast. Should this warm sector advance further north than forecast, the development of severe thunderstorms will warrant watching. Regardless, widespread rain looks to last into early Saturday. The closure of DFW’s 13R/31L may compound capacity pressures.

Share

You can check out1 hourly estimates in this workbook.

If you have questions about the content of this outlook, the answers might be in this post (we keep adding to it). Otherwise, let us know if something doesn’t make sense (we welcome feedback of all kinds!).

And we’re back!

After one wedding-induced layoff (not mine—slide into the comments, ladies) and another National Blend of Models-induced layoff, we’re finding our routine again. Thankfully, we missed some relatively quiet days. Core 30 airports1 dodged EDCTs entirely on Thursday and Friday; and only 0.6% of core 30 arrivals were assigned an EDCT on the 16th and 17th, a two-day incidence that’s in the lowest quintile year-to-date.

Perhaps more notable was TSA throughput in our absence, which set a new recovery high-water mark on Sunday, the 16th. 2.495 million travelers went through TSA checkpoints, surpassing the 2.491 million screened on July 1, 2022. It’s an impressive bounce from an Ian-related slump—depending on what weekday mapping you subscribe to, the equivalent Sunday in 2019 was 7-9% off the summer peak.

Let’s get to our regularly scheduled programming.

Sunday, October 23

We forecast TSA will screen 2.492 million travelers (± 0.5σ, or a prediction interval of about 38%, is 2.43-2.55 million travelers).

Models are coming into better agreement around the coastal low developing abeam Georgia; to wit, a weaker and more offshore solution. Nonetheless, a batch of rain should overspread the Mid-Atlantic Sunday, with a later onset (early afternoon) and slightly better chances for NYC.

A strong system headlined by heavy mountain snow may deliver impacts to DEN in the form of gusty west winds.

Monday, October 24

We forecast TSA will screen 2.379 million travelers (± 0.5σ, or a prediction interval of about 38%, is 2.32-2.44 million travelers).

Measurable snow is unlikely for DEN associated with this system, though mixed precipitation could result in some deicing2 delays Monday morning. We’ll take the opportunity to plug our winter operations explainer.

The low associated with this system lifts into the Northern Great Plains on Monday while its cold front pushes through the Southern Plains and Middle Mississippi Valley. Unseasonably warm air around ORD will support the development of showers and possibly a few thunderstorms through the morning. The highest chances for rain looks to be later in the day around DFW, with a few strong to marginally severe storms possible during the evening.

Share

You can check out3 hourly estimates in this workbook.

2

Still using those DALL-E credits. The prompt for this one was, “deicing at denver airport during sunrise, mountains in background, in the style of female indie-pop album cover.”

A Jack Antonoff-produce Taylor Swift album? We couldn’t help ourselves. Unfortunately, DALL-E isn’t similarly inclined and specifying Taylor Swift violates their content policy...

If you have questions about the content of this outlook, the answers might be in this post (we keep adding to it). Otherwise, let us know if something doesn’t make sense (we welcome feedback of all kinds!).

Thursday, October 13

We forecast TSA will screen 2.432 million travelers (± 0.5σ, or a prediction interval of about 38%, is 2.37-2.50 million travelers).

Five days into ORD’s October schedule change and they’ve managed without a late afternoon/early evening ground stop thus far. We’ll need to wait until mid- to late-November for a fuller picture (inclusive of metering, MIT, MINIT delays), but initial indications are disruption resulting from the new 6 p.m. arrival demand peak is largely imperceptible.

Despite the sort order, the focus of tomorrow’s story figures to be NYC, where scattered showers will develop by early afternoon ahead of a strong cold front. A more organized band of moderate to heavy showers is expected along the frontal boundary during evening. A strong onshore flow and evening frontal passage somewhat inhibit—but don’t rule out—thunder. While JFK might be slightly better equipped to handle the gusty south-southeasterly winds, their chances nevertheless feel a bit underdone.

To the south, the ingredients for convection—while still a bit uncoordinated—are more favorable around DCA (and Dulles, for that matter). Here the question seems to be whether severe storms1 develop (not whether storms develop at all).

Friday, October 14

We forecast TSA will screen 2.494 million travelers (± 0.5σ, or a prediction interval of about 38%, is 2.43-2.56 million travelers). Traveler volumes continue their quick recovery from an Ian-induced slump, with the 7-day moving average having returned to early August-levels. And for the first time in a few months, we again get to bet on setting a new high-water mark. No longer belonging to the Sunday after Thanksgiving 2021 (2.45 million screened), July 1 currently owns the recovery record (2.49 million screened). We give Friday about a 50/50 chance to exceed July 1’s checkpoint volume (to Thursday we extend 3 in 10 odds).

More tranquil weather conditions are on-tap for potential record-setting traveler volumes on Friday. Even the rain hatching in the vicinity of ORD owes to chances at the end of the chart’s valid period (i.e. Saturday morning).

Share

You can check out2 hourly estimates in this workbook.

If you have questions about the content of this outlook, the answers might be in this post (we keep adding to it). Otherwise, let us know if something doesn’t make sense (we welcome feedback of all kinds!).

Sunday, October 9

We forecast TSA will screen 2.370 million travelers (± 0.5σ, or a prediction interval of about 38%, is 2.31-2.43 million travelers).

After the Ian-influenced steep decline in 7-day screening volumes, TSA throughput has been quick to bounce back. Following Tuesday, the 7-day moving average sat at 2.01 million (its lowest level since early March); helped in part by yesterday’s 2.46 million screened (which would have been good for the third busiest Friday of the summer), the 7-day moving average has recovered to 2.13 million.

A relatively tranquil weather setup for our automated airports. LGA’s elevated probability owes to a curious October Sunday demand blip in the noon hour (jumps to 41 from a more typical 35 Monday-Friday).

Monday, October 10

We forecast TSA will screen 2.249 million travelers (± 0.5σ, or a prediction interval of about 38%, is 2.19-2.31 million travelers).

On Wednesday, we wrote about a new peak in the 6 p.m. hour that ORD1 is set to wrestle with this month. For Monday, the front depicted that drops down from Canada should have no sensible weather impacts—as a matter of fact, NWS Chicago writes, “splendid fall weather continues.”

With little in the way of weather to write about, we figured we’d spend a few more sentences on ORD’s October schedule. The new arrival peak (of 117 scheduled) exceeding optimal capacity (of 114) is doubly notable, given ORD’s status as a schedule facilitated airport (like EWR). So who might have played a little fast and loose with the FAA’s [quasi-voluntary] runway timings approval? Looks to be American, who re-banked their schedule, effective Thursday. The new schedule is actually 9% smaller week-over-week, with 4 banks (rather than the 6 in previous schedule). But in the re-banking, they effectively shifted the 7 p.m. arrival bank an hour earlier—which now sits on top of an existing United arrival bank. Correspondingly, departure activity in the 7 p.m. hour is also stimulated. We would advise to expect some consistent metering or miles-in-trail for arrivals in the 6 p.m. hour and healthy departure queues for departures in the 7 p.m. hour.

Share

You can check out2 hourly estimates in this workbook.

1

Back to a thumbnail generated by AI, though we’re trying OpenAI’s DALL-E now that it’s out of its private beta. The prompt was “o'hare airport neon light tunnel between concourse b and concourse c, vaporware, cyberpunk”

If you have questions about the content of this outlook, the answers might be in this post (we keep adding to it). Otherwise, let us know if something doesn’t make sense (we welcome feedback of all kinds!).

Thursday, October 6

We forecast TSA will screen 2.166 million travelers (± 0.5σ, or a prediction interval of about 38%, is 2.10-2.23 million travelers).

As an aside, Ian almost succeeded in pushing the 7-day moving average of TSA throughput below 2 million—a (psychologically important?) threshold we’ve remained above since early March. After yesterday, the 7-day moving average sits at 2.01 million. For those invested in such things, we think there’s a better than 4 in 5 chance that more travelers are screened today than last Wednesday (in which case the 7-day moving average would bounce). And estimate there’s only a 6% probability that today’s checkpoint volumes are light enough to drag the 7-day moving average below 2 million.

A strong cold front will quickly sink through ORD during afternoon. Post-frontal showers are expected and a few rumbles of thunder are possible; northerly winds also increase behind the front. Perhaps more notable is an October schedule change1 that sees 117 scheduled arrivals in the 6 p.m. hour… notable because ORD’s optimal rate is 114. We’d manually adjust our overage probability up for ORD tomorrow (the chance of thunderstorm creates a small sample size, which—in this case, too aggressively—de-weights a near-certain hourly probability).

Friday, October 7

We forecast TSA will screen 2.220 million travelers (± 0.5σ, or a prediction interval of about 38%, is 2.16-2.28 million travelers).

The cold front moves through the NYC area on Friday, though moisture is lacking and better chances for showers looks to be to the northwest.

The mixed precipitation hatching continues to inch close to DEN, though the short-range ensemble forecast (SREF) keeps snow and ice probabilities at zero.

Share Aerology

You can check out2 hourly estimates in this workbook.

1

Demand varies somewhat by day of week, with Saturdays “relaxing” to near 104. This is up from approximately 97 in September.

Charts from Monday, October 3 (pre-schedule change) and Thursday, October 6 (post- schedule change) below. Note different lower bound to vertical axis.

If you have questions about the content of this outlook, the answers might be in this post (we keep adding to it). Otherwise, let us know if something doesn’t make sense (we welcome feedback of all kinds!).

Sunday, October 2

We forecast TSA will screen 2.178 million travelers (± 0.5σ, or a prediction interval of about 38%, is 2.12-2.24 million travelers).

The rain shield associated with now-post-tropical Ian will extend far beyond its center (currently moving north-northwestward across central North Carolina). For EWR, the more notably feature may be an increasing easterly flow—there’s enough of a northerly component forecast at the moment to allow usage of runway 4L and 4R, but should direction veer more southerly, we could see (the rare) runway 11 usage.

Out west, low pressure aloft will keep weather unsettled around DEN, with chances for scattered showers and isolated thunder increasing through the day.

Monday, October 3

We forecast TSA will screen 2.096 million travelers (± 0.5σ, or a prediction interval of about 38%, is 2.03-2.16 million travelers).

Remnants of Ian continue to lose organization, eventually absorbing a piece of energy emerging from the northwest and redeveloping as a coastal low pressure system. There’s some uncertainty around how far off the Mid-Atlantic coast this low will sit; guidance is trending drier, but there remains at least a slight chance for showers in the vicinity of NYC and DCA.

Share

You can check out1 hourly estimates in this workbook.

If you have questions about the content of this outlook, the answers might be in this post (we keep adding to it). Otherwise, let us know if something doesn’t make sense (we welcome feedback of all kinds!).

Thursday, September 29

Our Holt-Winters model forecasts TSA will screen 2.287 million travelers (± 0.5σ, or a prediction interval of about 38%, is 2.23-2.435 million travelers), though it doesn’t see cancellations—and the people that don’t travel as a result. With 1,600 flights already cancelled for tomorrow (only two days this past summer had more), we’d expect actual screenings to come in towards the [very] low end of our forecast.

Ian is forecast to make landfall later this morning or early afternoon near Tampa Bay; interaction with a stalled frontal boundary will slow its progression. The center of Ian is expected to move across central Florida through early Thursday before emerging over the Atlantic later tomorrow. By Thursday night, an axis of heavy rain should be just offshore of the Georgia/South Carolina coast.

Though these reopening dates/times are subject to change, as of 10 a.m. Wednesday:

• Key West (EYW) reopens at noon ET Wed

• Sarasota (SRQ) reopens at 8 p.m. ET Thu

• Orlando (MCO) reopens 10:30 a.m. ET Fri

• Fort Myers (RSW) and Tampa Bay (TPA) reopen at noon ET Fri

Elsewhere, weather around DEN turns unsettled as some monsoonal moisture returns to the Southern Rockies (while mixed precip hatching inches closer to the airport, probabilities for frozen precipitation are zero per the SREF).

Friday, September 30

We forecast TSA will screen 2.297 million travelers (± 0.5σ, or a prediction interval of about 38%, is 2.24-2.36 million travelers).

Ian may strengthen slightly before it makes a second landfall somewhere along the GA or SC coast. There’s still some uncertainty as to whether really heavy rain materialize around CLT Friday night; tropical storm-force winds are not expected that far inland. Continued expansion of Ian waivers bears watching: Delta has grown their waiver to include smaller airports in GA (Brunswick, Columbus, Savannah) but not ATL (they’ve also included Hilton Head, SC). American has yet to expand their waiver beyond Florida, though their CLT hub is at least a decent candidate.

Through the operating period on Friday, Ian—or its remnants, more aptly—look to only increase clouds around DCA (though Saturday will be a different story).

Share

You can check out1 hourly estimates in this workbook.