Sequential Assimilation of Soil Moisture from Atmospheric Low-Level Parameters. Part II: Implementation in a Mesoscale Model

Abstract

A sequential assimilation technique based upon optimum interpolation is developed to initialize soil moisture in atmospheric models. Soil moisture increments are linearly related to forecast errors of near-surface atmospheric temperature and relative humidity. Part I has shown that soil moisture can be estimated from surface characteristics (vegetation coverage, soil texture). In this part, the behavior of the method is examined within a three-dimensional mesoscale model. The model includes a realistic land surface parameterization that relates soil moisture to atmospheric variables. Results reveal that after 48-h assimilations soil moisture has converged near reference values by blending atmospheric quantities in the algorithm. The convergence rate is almost independent of the first guess. Sensitivity studies show that the observational errors modulate the efficiency of the process and that results with an analytic formulation of the optimum coefficients are close to those obtained with a Monte Carlo method. These conclusions are of practical interest for an implementation in operational models.