If all the data were believed to be good and the analysis tried to tightly fit all the data, which of these drawings
most closely resembles how the analysis 700 hPa temperatures would probably look? (color scale same for all)
Choose one choice, then click Done. (To undo a selection, click the choice again.)
Discussion
Choice a) is what the analysis actually looks like. It did treat all the data as good, but it
smeared out the gradient instead of tightly fitting all the observations. We will discuss why shortly.
Choice b) is what a best human interpretation would look like, using a conceptual model to fit the
temperatures where there isn't any temperature data. Two pages later, we will discuss work underway to help the
analysis look more like this than like choice a).
Choice c) is what that tight-fitting analysis would look like. The cold air under the rainband was only
sampled in a limited area near DFW airport. The limited sampling poses real problems for detecting mesoscale
features and for including them in the model. Surface observations did not pick up this feature, nor would
aircraft flying over at cruising level. Analyzing the rain-cooled air only in the circle where it was observed rather than over
the elongated band over which it probably extended could produce a bizarre forecast you probably wouldn't find
useful. As the models move toward finer resolution, providing adequate observations to take advantage of the
improving resolution is becoming ever more difficult.
Observations may not only have insufficient coverage to capture a mesoscale
feature, but the data may be systematically biased. For instance, profiler
winds may be (properly) flagged as bad in moderate to heavy rain and satellite
IR radiances for temperature and moisture information are not available
in cloudy areas. In this case, aircraft approached the airport from the
north, providing many observations of the rain-cooled air and few observations
of the warm air at and south of the airport.