Verification statistics for yesterday's MESO MM5 runs are shown below. The statistics shown are Mean Absolute Error (MAE) and bias for surface temperature, dewpoint, mixing ratio, surface pressure, wind speed, wind direction, and wind vector difference.

The model variables are interpolated to the surface station locations for the comparisons. Corrections are made to the model temperature and pressure to try to account for differences between model elevation and station elevation. Similarity theory is used to calculate the MM5 surface temperature and wind speeds. The statistics are computed for every available observation within the domain, excluding a boundary region of 2 dx. An observation is thrown out only if it flunks a very gross limits check (235C < T < 340C; wind speed > 50 m/s; mixing ratio > 25 g/kg; 200C < TD < 340C; 1050 mb < P < 500 mb). MAE is defined as sum(abs(ob-model))/nobs. Bias is defined as sum(model-ob)/nobs. A positive bias thus means the model is too warm/high/moist/fast.

ETA CRNT is the 0000 UTC ETA run - same initialization time as the MM5 runs. ETA PREV is the previous 1200 UTC ETA run, verified at its 12-48 hr forecast times. The 1200 UTC ETA run is that used as the first-guess for the MM5 initial analysis and for the lateral boundary conditions.

The verification software is new and evolving - there may still be errors in it! I would like to add upper air verification against raobs and precipitation verification at some point.

T, TD, MR, P MAEs for yesterday's MESO MM5 forecasts




Dir, Sp, Vec MAEs for yesterday's MESO MM5 forecasts




T, TD, MR, P Biases for yesterday's MESO MM5 forecasts




Dir, Sp, Vec Biases for yesterday's MESO MM5 forecasts




Back to MESO MM5 main menu