1. Radi AJJAJI : "Incrementality deficiency in ARPEGE 4d-var assimilation system"

This contribution is very long, since presenting a synthesis of the work achieved along the 3 last years.

The three following papers (with a fourth one in the next Newsletter) try to understand erroneous humidity analyses noticed in punctual cases over some mountainous areas and in the desert. Two selected cases are thoroughly examined, one situation over Sahara region (June 26th, 2000), where 4d-var assimilation introduced a great fictitious moistening, and one over northern Spain, where the 4d-var analysis was far drier than the reality (April 24th, 2001).

The first paper tries to find an explanation in the incremental approach (more precisely the multiple-truncation incremental scheme used operationally in France). A series of experiments consisting in running several 15-days assimilation suites with changes in the different ingredients involved in 4d-var (with the French ARPEGE setups : incremental, multi-incremental, non incremental, simplified physics in the last inner loop, simplified physics in all loops, adiabatic runs, etc...). The results suggest that the problem of wrong humidity analysis is somewhere else. A number of hypotheses are expressed, which require more experimental studies.

The second paper tries to formulate the dynamics of the Eulerian tangent-linear model used in the minimisation. This theoretic formulation shows the potential terms likely to be responsible, in the adjoint model, for the large humidity increments. These terms are then carefully examined in terms of order of magnitude. They allow to explain partly the problem by indicating strange magnitudes for some fields, such as pressure gradients over mountainous areas and vertical velocities over southern Sahara.

The third paper tries to find an explanation in the B covariance matrix used in 4d-var, by examining the order of magnitude of the implied background-error standard deviations, in terms of the different control variables, focusing on specific humidity. This study shows reasonable values at the beginning of the assimilation window.