4DVAR posterior error covariance matrix

[arango]

The weak constraint algorithms W4DPSAS and W4DVAR were updated to include the posterior (analysis) error covariance matrix, Ea, for the estimated initial conditions. The posterior error covariance matrix is very large, expensive to compute and store. However, the dominant EOFs of this matrix are easily computed by using an iterative Lanczos algorithm. These EOFs contain information about the patterns of largest uncertainty in the 4DVAR estimate.

The analysis error can be activated with the POSTERIOR_EOFS option. Many thanks to Andy Moore for his help in coding and testing this option.