CMC MJO Forecast Methodology


To get the combined EOF1 and EOF2 bases:

  1. The unfiltered daily NCEP data and the observed OLR from 1979 to 2001 are used.
  2. The seasonal cycle which is the first 3 harmonics of the daily climatology is removed.
  3. The 120-day mean of the last 120 days for each day is removed. By removing the previous 120-day average most of the interannual variability related to ENSO is removed. Even though we didn't remove the linear regression to ENSO using the SST1 as in WH2004, our resulting structures of EOF1 and EOF2 are almost identical to theirs: the temporal correlation between the daily PC1 with that of WH2004 over the 23 years is 0.97 , and 0.98 for PC2.

To get the observed RMM1 and RMM2:

  1. CMC analysis of U850, U200 and precipitation rate are used.
  2. Seasonal cycle and the last 120-day average are removed as in the combined EOF calculation.
  3. Project onto the combined EOF patterns (reverse sign for PR anomaly).

To get the forecast RMM1 and RMM2:

  1. Seasonal cycle is removed. The 120-day mean right before the forecast starts is subtracted from all forecast data.
  2. Ensemble means of U850, U200 and precipitation rate from the CMC ensemble forecast are used.
  3. Project onto the combined EOF patterns (reverse sign for PR anomaly).