My forecast for Monday and today, posted early Monday morning, called for reasonably high (50 to 70 percent) chances of showers and thunderstorms. Yet little rain has fallen in the immediate D.C. area, generally less than one-quarter inch (although heavy rain fell in extreme northwest Virginia).
Weather can be a game of inches, and that’s been the case over this short stretch. A front has meandered to our north and west, serving as a focus for heavy precipitation. Because the front has not been terribly well-defined, models and humans have had a hard time knowing where the rain would start and stop.
It turns out the dividing line between significant rain and hardly any set up just to the west and north of the District.
We can see below how the models struggled in simulating this cut-off point by comparing their forecast amounts (below) to the map of actual rainfall totals above.
Interestingly, the model which did the worst job simulating the cut-off was the European, which significantly overforecast rain in the D.C. area and much of Virginia. The NAM model did the best job, though it still overdid the rain in the D.C. area somewhat.
While the European model didn’t handle the dividing line well, it probably had the best simulation of the heavier rain to the north into Pennsylvania and New York whereas the NAM underestimated amounts. The GFS, generally, overestimated amounts across the board but did a nice job simulating the western cut-off to the rain from central Ohio to western New York state.
Summer is tricky time for models to predict precipitation amounts because they do not capture highly variable and random small-scale features like thunderstorms very well. In this case, the models were aided by the presence of the front, since it helped focus the rain/storms. But, because the front was diffuse, getting the edges right proved too difficult to overcome.
As a forecaster, I probably should’ve better recognized that we were close to the murky edge and played up the uncertainty in the rain chances a bit more.