Forecast model bias correction in ocean data assimilation G. Chepurin, Jim Carton, and D. Dee* Univ. MD and *GSFC Bias in ocean data assimilation Two-stage bias correction algorithm –Bias model –Results from a series of 30-yr assimilation experiments Manuscript available: {
Bias is the difference between the state forecast and the true state
Time-mean bias along equator 20C “Cold tongue is too cold, while the thermocline in the central basin is too diffuse”
Annual cycle of mixed layer bias in subtropics (10N-30N) Dec June “Too hot in summer, too cold in winter”
Time-evolution of forecast error along equator “Forecast error is episodic, linked to ENSO” 100m Time Mixed layer
Two stage algorithm to correct systematic aspects of forecast error Stage I Stage II
Three-term bias forecast model Time-mean bias Annual cycle bias ENSO-linked bias
Correcting time- mean bias along Pacific Eq This is business as usual This is what results when time-mean bias is modeled 20C
Correcting time-mean bias Corr time-mean bias
Correcting annual cycle bias Business as usualAnnual cycle bias correction Dec June
Annual cycle of forecast error before correction
Annual cycle of forecast error after correction Before After
Correcting ENSO bias before after Cor EOF1,SOI = 0.7
Summary of the impact of bias correction time mean +annual cycle +ENSO variability RMS (fcst-obs) ML temp Thermocline depth
RMS (fcst – obs)
Conclusions Half of the {forecast – observation} differences in high variability regions are due to bias. The largest contribution is time-mean followed by annual cycle and interannual variability. Two-stage correction works well in addressing these. Manuscript available: {
The End