Systematic Model Error

Diagnostic

Model bias for U/V/T/Q/Ps computed from gen_be with the NMC method (24h minus 12h forecasts) for AMPS1 domain over the period from 10/01/2006 to 10/29/2006.
Temperature bias near top of model over Antarctica region.

• Indicator = NMC method (24h - 12h forecast)
• Location = at top model level and (to a smaller extent) at the 4 levels below.
• Shape = Vortex type with opposite signs b/w Antarctica plateau and southern oceans
• Variance = large, due to rotation of vortex. Contaminates gen_be -> leading eigenvectors for T_u are located near model top.
• Investigate 36h -24h and 12h - 00h vs. 24h - 12h forecasts to check if we are facing a constant bias, growing error or diurnal effect.

Possible explanations

• Radiation
  Investigate different radiation schemes. Goddard SW scheme uses ozone (in stratosphere) (option 2). LW default scheme = option 1.
• Ozone
  Investigate correlation b/w bias and ozone field.
• Dynamics
  Investigate correlation b/w bias and height field.