Hand Analysis of Weather Charts: A Lost Art and Science?

I’m fortunate to work at an unnamed national severe weather forecasting center that serves as one of the last bastions of frequent manual map analysis in the nation. Twice a day, the mandatory level (925, 850, 700, 500 and 250 mb) upper air charts appear from printers on 11″ x 17″ paper for analysis in support of the core forecast mission. Surface maps are printed for analysis many times per day as forecasters take the time and interest to do so. When these analyses are done in a careful, detailed way, this is a very, very good thing; for it forces the mind to slow down and truly immerse in the observations, connecting them into a carefully constructed, multivariate conceptual model of that snapshot of our atmosphere.

Excellent analysis is time-consuming, and requires a good deal of both skill and practice. Unfortunately, especially in local offices, hand analysis is vanishing, partly because of increasing volumetric workload (and resulting loss of time) and partly from ignorance and laziness. I often feel that time pressure myself, and fear that the same fate may befall the national center(s) if we’re not diligent and careful. The slow creep of forecast automation — and of the related phenomenon of “meteorological cancer” (first noted in this 31 year old paper by Snellman) — is on the move in some insidious ways. As I’ve pointed out here before, procedure is taking over forecasters’ time at the expense of meteorology.

Sometimes it seems like those of us who bother to do detailed hand analysis are becoming an endangered species. It utterly baffles me why so many put so little value in it, and why the quality of many hand analyses is so poor — with lines on the wrong side of data, closed highs or lows unlabeled, missed boundaries, unlabeled lines, time continuity jumps (e.g., a cold front oscillating backward from one map to the next) and other fundamental errors. Were I still a synoptic meteorology teaching assistant, I would assign grades of “D” or lower to 90% of the drawn maps I have seen, anywhere, in the last several years. I don’t understand this inattention to excellence in hand analysis. Is it lack of exposure or training, time pressure, disinterest, inexperience or all of the above?

Acquiring the deepest possible grasp on the current state of atmosphere is absolutely essential to forming the conceptual models needed for consistently good forecasts, and also, for good forecasts of rare and extreme events that are most poorly handled by objective analyses and guidance. Throughout all the technological advances, this basic truth remains unaltered.

I engage both hand-drawn maps and objective analyses of all kinds on a daily basis, in considerable detail, and can assure you that the latter is a piss-poor substitute for the former, when excellence in hand analysis is made top priority. Now please understand that I wouldn’t violate the Golden Rule by holding anyone else to a greater standard than myself; nor do I claim to be a more skilled analyst. The principle of (and need for) analytic excellence absolutely applies to me as much as anyone. I can (and do) go back and find inexcusable and shameful mistakes in my own analyses.

A missed max or min here or there, or a boundary not drawn under hurried circumstances, may not seem like much, but how does one know what “minor” feature is really unimportant until the event is over?

Superficial skimming of objective, computer-drawn analyses — which often miss or misplace small but critical features — is not the same as truly diving into the data, taking the time to thoroughly draw for and interpret it. If it comes to a choice between more hand analysis or more model output, I’m choosing the hand analysis. An informal (not yet published) experiment called Project Phoenix, managed by Pat McCarthy of Environment Canada’s Prairie and Arctic Storm Prediction Centre, showed over several seasons that forecasters who looked at observational information only (including hand analyzed charts, as well as satellite, radar and other observed data) performed better at forecasting basic variables during day-1, and often into day-2, than those on regular, operational shifts looking at numerical models. Why would this be?

I don’t advocate eschewing model guidance any more than I would dropping hand analyses. In the practical sense of operational forecasting, it would be dumb to ignore the most pertinent nuggets of prognostic information. I make intensive use of models in day-to-day forecasting, including an increasingly heavy reliance on clues provided by short-range ensemble forecasts (SREF). But when pressed for time, I’ll sacrifice a deterministic model or two for more insight into the current state of things. After all, how can we consistently predict the future atmosphere without deepest possible understanding of the present atmosphere? Those who claim to have such understanding by ignoring manual diagnostics and looking at objective analyses instead have deluded themselves, amidst the intoxicating abundance of quick-n-ready digital diagnostics — the choice drug of forecasting, so to speak. These forecasters are dooming themselves to a fate they probably deserve — automation of their jobs — but in the process, increasing that risk for the rest of us as well. Just remember: garbage in, garbage out. Without corroboration from reality, how does one know the computer drawn map is accurate?

Failing to do good hand analysis also is quite selfish. The subjective analyses are not just for the forecaster doing them; they are for the entire forecast team and shift successors, and are important to continuity of understanding of an evolving situation. Even if a forecaster doesn’t “get much” from his own analysis (which is a surefire sign of a scientific and conceptual deficiency on his/her part), it has meaning to others, and as such, should be done and done thoroughly. Hence…

Understanding the current state of the atmosphere isn’t a matter of personal choice; it’s a matter of professional responsibility.

The solution for the troubled forecaster is simple: Get out those colored pencils and start drawing! To help get going, here are links to online, national upper air maps, ideal for hand analysis and optimized for 11 x 17 printer paper (but useable at smaller sizes):

12Z Plots with VAD and Profiler Winds (PDF format): 925 mb | 850 mb | 700 mb | 500 mb | 250 mb

00Z Plots with VAD and Profiler Winds (PDF format): 925 mb | 850 mb | 700 mb | 500 mb | 250 mb

The first step in making the best forecast that the science permits requires a thorough high quality diagnosis. Successful short range forecasts are more the result of good diagnosis than of prognosis.

– Len Snellman, 9 October 1991