Author Essays, Interviews, and Excerpts, Books for the News, Mathematics and Physics, Sociology

Excerpt: Masters of Uncertainty

9780226298689

Excerpt:
Masters of Uncertainty: Weather Forecasters and the Quest for Ground Truth 

by Phaedra Daipha

***

PRODUCING THE WEATHER FORECAST

Protecting America’s life and property against the calamities of the weather is a daunting task to manage, above and beyond the formidable meteorological challenges involved. There is a tremendous range of “weather” for the NWS to keep an eye on: land weather, airport weather, marine weather, fire weather, hydrologic weather. To properly protect America’s life and property against the calamities of the weather, therefore, NWS forecast offices are operational around the clock.“The weather never sleeps and neither do we” is the usual stock phrase Neborough forecasters use to enlighten outsiders about their schedule and, by extension, their importance.

The primary responsibility of an NWS forecast office, of course, is to advise and alert about the potential for hazardous weather. Only when hazardous weather warning requirements have been met do forecasters turn their attention to routine products and services. Indeed, during a hazardous weather event, the NWS becomes transformed into what Fine (2007, 40) calls an“activated organization . . . verging on being overwhelmed and understaffed, until routine can again be established.” In this, NWS forecasters readily resemble firefighters, paramedics, and other first responders— primed for hazardous weather, they seem perpetually caught in a lull before the storm. But pushing the analogy any further may be misleading. In contrast to firefighters (cf. Desmond 2007, 81ff.) and other emergency professionals, NWS forecasters experience little or no workload downtime.11 In 2008, five years after the implementation of the IFPS, with the work schedule radically revamped to accommodate for the realities of the new forecasting process and the new forecasting process already an old routine, Neborough forecasters were still struggling to keep up with the weather. To be sure, their workload has been exacerbated by an expanding list of IFPS-driven forecast responsibilities, which, coupled with a series of staff cutbacks due to a shrinking budget, force offices to “do more with less,” to use another favorite NWS catch phrase. What drives this heavy workload in the first place, however, is the fact that we have come to recognize “the weather” not as a finite phenomenon, like a wildfire or a robbery, but as ever present and relevant. As competition among the various weather forecast providers emboldens our appetite for more, faster, better weather information, what counts as the weather further expands in detail and significance. And so, too, does the charge of NWS forecasters. NWS forecast offices more closely resemble a newsroom in this respect—busy during routine operations, verging on understaffed during emergency conditions.

It is this trait of the weather that, in the context of NWS operations, makes weather forecasting an auspicious case study for probing the process of decision making in multiple decision- making regimes. But first, the basics. By way of an introduction to the process of meteorological decision making, the remainder of this chapter goes over the main components of a typical shift at the Neborough office: data analysis; deliberation; and finally, the actual doing of the forecast. In practice, of course, these components are thoroughly intertwined. There is no actual moment when diagnosis ends and prognosis begins. Rather than constituting the means and ends, respectively, of forecasting action, diagnosis and prognosis are in fact“two names for the same reality” (Dewey 1922, 36). Ne-borough forecasters never switch from a diagnostic to a prognostic frame of mind—they just continue making increasingly more consequential decisions as the forecast submission deadline draws nearer. If anything, meteorological prognosis analytically precedes diagnosis, as will become evident in the following pages. This empirical reality of NWS forecasting, while unintelligible from a rational choice perspective, is entirely consistent with pragmatist accounts of the decision-making process. “We do not use the present to control the future,” writes Dewey (1922, 322); “We use the foresight of the future to refine and expand present activity.” The formal distillation and ordering of the NWS forecasting routine into a diagnostic, a deliberative, and a prognostic component denotes therefore the temporal, or processual, unfolding of forecasting action rather than its analytic structure.

TAKING OVER THE HOT SEAT

No shift can start without a briefing by the outgoing forecaster to get the incoming forecaster up to speed with the big weather picture and developing concerns. As in any other work setting whose rhythm is dictated by a shift schedule, weather briefings form an essential, organic part of the forecasting routine as they allow for efficient resource management, minimize duplication of effort, and promote forecast-to-forecast continuity. Weather briefings almost always occur right at the workstation of the outgoing forecaster. Indeed, if the incoming forecaster does not find the outgoing forecaster at his desk as she walks in, she will seek him out in his cubicle and, together, they will walk back to the operations deck to begin the briefing. This is not a mere formality but a testament to the role of screenwork—that is, the processing of information via computer screens— as the organizing principle of meteorological expertise (Daipha 2013). As already noted, if they cannot see it, forecasters cannot think, never mind talk, weather. The departing forecaster relies on the computer screens to make a case for his forecasting decisions, reasoning through the assortment of weather displays he flags as pertinent. For her part, the incoming forecaster relies on the computer screens to keep up with the action, to bring into focus what portends to be the weather forecasting problem of the day. As will become apparent time and time again, screenwork forms the backbone of every aspect of the meteorological decision-making task.

A briefing is in reality two briefings in one, a briefing about forecast concerns and a briefing about technology malfunctions, and forecasters at Neborough will invariably cue “weather wise” and “equipmentwise” to signal the end of the initial pleasantries and the start of the briefing or to segue into the next section. Depending on the weather situation du jour and the familiarity of the incoming forecaster with the current weather system—in other words, depending on whether she has worked that desk the previous day—briefings may last from several seconds to over ten minutes, not infrequently turning into protracted meteorological discussions on model performance and biases or similar past weather events. Throughout, the two forecasters will be poring over the computer screens, the outgoing forecaster guiding the action with the computer mouse. Anything and everything deemed relevant information can be included in the weather briefing: model(s) of choice and reasoning behind it, forecast dilemmas and ultimate decisions, remarkable personal weather observations and puzzling spotter reports, verification concerns, deliberations with neighboring offices, weather features to watch out for, upcoming hazards and how they have been addressed so far, interoffice coordination issues. The conversation is relaxed and collegial, a good-natured back and forth. Yet, even between forecasters who know each other well and have a high regard for each other’s forecasting skill, the exchange is clearly underwritten by a handing off the baton dynamic: the outgoing forecaster is eager to make his forecast stick, especially if he is coming back in a few hours, while the incoming forecaster is intent on not missing a beat but not necessarily committed to the particulars of the existing forecast.

Margaret (short-term desk, day shift): So you guys still thinking some action today?

Dick (short-term desk, midnight shift): Yes. Yes. Todd [at the neighboring office to the southwest] and I were talking it over, and there’s just enough cold air advection aloft, and it’s a sharp enough upper trough, that I was not going to go against tit . . . But, as you can see here, there’s a fair amount of action upstream right now in [adjacent states to the southwest], and the potential is there for all that to work in and just stabilize the air mass out of all severe possibility. So, it’s not something I’m entirely confident in, but with the other factors in place I just showed you . . .

Margaret
: So, it must be all instability aloft, because it feels pretty comfortable out there.

Dick
: Yeah . . . I mean, see here, we are eventually looking at a minus twelve [degrees Celsius] at 500 [millibar atmospheric pressure level]. But it’s going to be this evening before it gets down into our part of the Northeast . . . so, that’s part of the problem, too, that . . . if anything busts the forecast is that the cold pool . . .

Margaret
: Takes forever?

Dick
: . . . takes forever to get in. so that’s something to keep an eye on. . . . Anyway, I did what I could; it’s in your hands now.

“Equipmentwise” the briefing can take up an equal amount of time—hardly surprising given the big science character of NWS forecasting operations. Technology fails forecasters in big and small ways, from the radar going down during a severe storm to the server being slower than usual, and it fails them in multiple ways at once. The log of any given shift during my stay at Neborough contained at least two equipment malfunction entries, with that number doubling during hazardous weather conditions. The weather spares no one, certainly not the people tasked with anticipating its every move.

Finally, a“changing of the guard” of sorts takes place. Throughout the briefing, the outgoing forecaster has remained seated at the desk chair with the incoming forecaster standing or leaning against the desk next to him. The briefing completed, the outgoing forecaster will now stand up and, sometimes with some kind of verbal or nonverbal flourish, he will offer the seat to his relief of the day.

The incoming forecaster is now in charge of the workstation. But not until she has adjusted the computer screens according to the settings stored under her user profile will she have truly taken control of the desk and of the weather. The customized, unique combination of weather display formats, color graphics, and sound alarms effectively transforms the workstation into on’s personal workspace, and Neborough forecasters are quick to switch over to their profile the moment they claim the seat. That is especially so because, stored under a forecaster’s name but accessible by all, is a set of personal “best practices” for looking at weather information: the so- called Procedures, accumulated over one’s career and updated as necessary. To study the wind along the atmospheric column, for example, one forecaster might prefer the 850, the 500, the 250, and the surface millibar height charts, color coded just so, while another might routinely find the 700, the 500, the 200, and the surface charts more insightful. Where the Procedures become truly useful, however, is in their ability to recall elaborate composites of data graphics in a matter of seconds. As a result, the variation among forecaster profiles can appear quite staggering. Yet, despite their seemingly idiosyncratic nature, forecasters’ profiles reflect eminently social decisions, the result of apprenticing at particular meteorology programs and forecast offices, under particular mentors, with particular technologies, and so on (see Daipha 2010). For example, unlike other Neborough forecasters, Biff and Phil have primarily organized their Procedures according to weather scenarios, something they learned to do at X- University, which they attended several years apart. And Margaret and Phil are in the habit of looking at weather data in really busy four-panel displays, something they picked up, as it turns out, while interning in the same forecast office in the Midwest.

To be sure, our forecaster does not, as a rule, have to change the previous weather display settings. Rarely did I witness a Neborough forecaster request that the display settings be changed or explained when hunched over someone else’s workstation— following the weather was straightforward enough. In fact, there is an argument to be made that no profile presets are truly necessary, and that the default settings would more than suffice for the task ahead. Certainly, most forecasters like to claim that relying on their Procedures is only an issue of expediency. In practice, however, temporal constraints and the threat of information overload lead to a near absolute dependency on such preset templates for studying the weather.

2:45 p.m. Phil was working an administrative shift in his cubicle today but has been called on forecast duty because of the potential for severe weather later this afternoon. He logs into one of the vacant workstations, turns to the left graphics screen, goes to File/Procedures/select User ID, selects his user name, clicks on severe Weather Tools from the drop-down menu, and selects to load all six of the included information sources from the new drop-down menu. Repeats the same process for the middle graphics screen, this time clicking on Meso [analysis] stuff and selecting five information sources (mostly guidance products from the storm Prediction Center) out of approximately twenty-five. While waiting for the data to load, he next turns to the right graphics screen but now selects Biff’s user name and clicks on severe_Neborough Radar_Right. He tells me he has been meaning to copy this procedure into his own user profile. He likes how “nice and clean” Biff has set up his radar info for right-moving storms. . . . Within fifteen minutes, in consultation with Tom (short- term desk), Phil is ready to press “send” for the first severe thunderstorm watch of the day. He rubs his hands together in excitement: “And we’re rolling!”

***

To read more about Masters of Uncertainty, click here.