Friday, March 15, 2013


DModel/DT is calculus-based slang that meteorologists use to describe a trend in model forecasts.  For example, the GFS forecast from 0600 UTC 12 March showed a less than favorable forecast for snow in the Wasatch late Saturday night and Sunday morning, with a short-weve trough over the Pacific Northwest and Utah under the influence of a short-wave ridge.  Subsequent forecasts, however, have pushed the short-wave trough further east and sagged the storm-track southward, resulting in snow over the Wasatch late Saturday night and Sunday morning.  This model trend is illustrated in the DModel/DT loop below, which is based on 0600 UTC initialized GFS forecasts from the past four days all verifying at 1200 UTC 17 March (Sunday).   Note how each successive forecast produces more precipitation over northern Utah.

You have to be careful when using DModel/DT because the existence of a trend doesn't necessarily mean that the trend will continue in future forecasts.  In longer range forecasts, it's not uncommon to see the forecasts flop back and forth from run to run.  Nevertheless, DModel/DT can be a useful forecast tool and, in this case, might explain why the chances for precipitation on Sunday have been on the increase the past few days.


  1. The ECMWF has been showing precipitation for several days now and has actually been trending drier.

  2. Jim, have you seen this paper that shows dProg/dt to have no skill?

    1. Yes. I think a key point of that paper was that *extrapolation* of forecast trends was shown to have little value, which I agree with.