Observation sensitivity for W4DPSAS and W4DVAR

[arango]

Two new weak constraint, 4DVar data assimilation algorithms have been developed to compute the sensitivity of the assimilation system (W4DPSAS and W4DVAR) to each observations. They measure the degree to which each observation contributes to the uncertainty of the estimate. This analysis can be used to determine the type of measurements that need to be made, where to observe, and when.

These are very complex algorithms. We will provide guidelines for their usage in the future. They are activated using either W4DPSAS_SENSITIVITY and W4DVAR_SENSITIVITY. These algorithms can be only used for a single outer loop. They include the regular data assimilation algorithm (W4DPSAS or W4DVAR) plus the adjoint of the system. More than one outer loop will require the tangent linear of the tangent linear model which we don't have.

Many thanks to Andy Moore for helping us code and test these algorithms.