6A.1 Near-Surface Wind Data Assimilation using a Geo-Statistical Observation Operator – Results from Observing System Experiments

Tuesday, 30 June 2015: 10:30 AM
Salon A-2 (Hilton Chicago)
Joël Bédard, Université du Québec à Montréal (UQAM), Montréal, QC, Canada; and S. Laroche and P. Gauthier

Handout (2.9 MB)

Although many observations describing the wind field in the lower troposphere are available, very few are assimilated over land mainly due to sub-grid scale topographic interactions with the flow. This study aims at improving short-term tropospheric analyses and forecasts by assimilating near-surface wind observations over land in the ensemble-variational data assimilation system of Environment Canada. A novel geo-statistical observation operator (GMOS) has been developed. It takes advantage of the correlation between resolved scales and unresolved scales to correct the stationary and isotropic components of the systematic and representativeness error associated with local geographical characteristics (e.g. surface roughness or coastal effects). The GMOS operator has been tested and compared with the results obtained from a conventional bilinear interpolation scheme (Bilin). By attributing higher weights to the most representative grid-points, GMOS better represent the meteorological phenomena onsite. As background states and observations are generally more consistent, GMOS produces relatively smaller innovations and analysis increments than Bilin. Thus, it is less prone to generate strong perturbations in the resulting analysis. Different observing system experiments (OSE) have been performed over the month of February 2011. The resulting analyses and subsequent 48h forecasts have been verified against near-surface wind observations, independent radiosonde profiles and analyses.

Results from non-cycling OSE (assimilating only near-surface wind observations) show that GMOS eliminates biases and significantly reduces representativeness errors as well as observation error correlations, mainly over complex terrain. Due to the background-error covariances, near-surface wind observations impact the lower part of the atmosphere. Results also show that flow-dependent background error covariances from ensembles provide better vertical information propagation than static error statistics. Overall, the analysis fit to non-assimilated collocated radiosonde observations is improved when assimilating wind observations from surface stations, which indicates that they have some ability to reconstruct the vertical structure of the atmosphere. The evaluation of forecasts against radiosonde observations show that very short-term wind predictions are significantly (slightly) improved when using GMOS (Bilin). However, the local impact decreases over time and is only significant for 6h lead time or shorter. A detailed analysis indicates that, when using dynamic (static) error statistics, the impact on the forecasts persists (decays). Flow dependent background error statistics modify the analysis increments of both wind and mass fields in a coherent way through multivariate covariances. Pressure gradient forces are generated and counterbalance the vertical diffusion and orographic blocking schemes. On the other hand, static error covariances have no significant impact on the mass fields and the boundary layer parameterization schemes diffuse the increments locally. Results show that Bilin degrades wind, temperature and geopotential height fields, while GMOS improves the wind speed field up to a 48h lead time.

Results from fully cycling OSEs in a near-operational context (near-surface wind observations are assimilated along with the operational assimilation dataset) indicate that GMOS (Bilin) leads to small (large) perturbations in the analysis. Bilin does not provide a proper model state comparison with the observation and most of its increments are diffused by the ABL parameterization. On the other hand, results show that the GMOS increments are in agreement with the model state and the information persists in the system. The GMOS (Bilin) forecasts and analyses are thus more (less) coherent than those from the control experiment. A detailed analysis of a particular case indicates that, three low pressure systems propagating from the North Atlantic towards Eastern Europe are well (badly) forecasted when assimilating near-surface wind observations with the GMOS (Bilin) operator. In agreement with the verification against observations, the evaluation against own analyses shows that the Bilin experiment significantly degrades forecasts of all fields whereas GMOS improves them up to a lead time of 48h.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner