A comparison of two off‐line soil analysis schemes for assimilation of screen level observations

Abstract

[1] Two analysis schemes are developed within an off-line version of the land surface scheme ISBA for the initialization of soil water content and temperature in numerical weather prediction models. The first soil analysis is based on optimal interpolation that is currently operational in a number of weather centers. The second soil analysis is an extended Kalman filter (EKF) which will allow the assimilation of satellite observations. First, it is shown, by comparing the Kalman gain of both analysis schemes, that it is possible to assimilate screen level temperature and relative humidity in an off-line system. This is of great interest for future combined assimilations of conventional and satellite data. The reduced computing time in running the land surface scheme outside the atmospheric model makes Kalman filter approaches compatible with operational requirements. The methodology for coupling the land surface data assimilation with the atmospheric analysis system is explained in order to highlight the existing feedbacks between the two systems (in comparison to fully decoupled land data assimilation systems). The linearity of the observation operator Jacobians estimated by finite differences and the relevance of the soil prognostic variables to be initialized are assessed. Finally, the two systems are compared over western Europe for the month of July 2006 by assimilating screen level temperature and relative humidity every 6 h. The EKF has been simplified by keeping the covariance matrix of background errors constant. The two soil analysis schemes behave similarly in response to screen level atmospheric errors. The EKF is superior in identifying situations where the near-surface atmosphere is sensitive to soil perturbations, which leads to better use of observations. Over France, the capability of both systems to moisten the soil when rain events are absent from the forcing is demonstrated.

Study of the Jacobian of an extended Kalman filter for soil analysis in SURFEXv5

“…When the local soil moistureatmospheric boundary layer feedback is weak, for example in situations of weak radiative forcing or strong advection, the screen-level observations do not provide any information about errors in the soil. Therefore, it would be useful to also include other soil observation types in the soil analysis, for example remotely sensed soil moisture (Draper et al, 2009(Draper et al, , 2011. OI uses analytically derived coefficients, making it difficult to include new observation types in this technique.…”

“…Results indicate that OI and the EKF have similar gain coefficients and increments. The EKF has been extended to include other observation types, like AMSR-E soil moisture retrievals (Draper et al, 2009), radar precipitation information (Mahfouf and Bliznak, 2011), and ASCAT surface soil moisture (Mahfouf, 2010;de Rosnay et al, 2012).…”

“…An EKF has been developed for SURFEX by Mahfouf et al (2009), assimilating screen-level temperature and relative humidity to correct soil moisture and soil temperature. Results indicate that OI and the EKF have similar gain coefficients and increments.…”

“…The Jacobian describes the sensitivity of the screen-level observations to changes in the soil prognostic variables. Mahfouf et al (2009) suggest calculating the Jacobian with a finite differences approach, using a reference run and one perturbed run for each of the soil prognostic variables (i.e. a run with an initial surface where one of the prognostic variables has been perturbed).…”

Sensitivity of Soil Moisture Analyses to Contrasting Background and Observation Error Scenarios

“…For example, Muñoz-Sabater et al [12] used an ensemble of model integrations to produce an estimate of the background error. A more physically-based relationship between relevant land parameters was tried by Mahfouf et al [13] to obtain representative errors of soil moisture. They specified the diagonal terms of the soil moisture background error covariance matrix as a function of the water holding capacity (field capacity minus wilting point), thus related to soil texture.…”