The sensitivity of the WRF-4DVar data assimilation system to the control variables: A study on heavy rainfall events over India
Abstract
The impact of different formulations of background error covariances (BECs) is examined for three heavy rainfall episodes over north India with a regional 4-dimensional variational (4DVar) data assimilation (DA) system. Three BEC formulations are analyzed, in which two of them employ stream function and velocity potential (ψ and χ) and the third one uses zonal and meridional velocity components (v and v) as momentum control variables. The uv -based formulation is completely univariate whereas, the correlations among the control variables are taken into account in the ψχ -based formulations through linear regression relations. Among the two ψχ -based BECs, one uses univariate relation and the other one uses multivariate relations for the moisture field. The multivariate relationship allows for impacting the moisture analysis through the assimilation of temperature or wind observations. Three experiments are carried out for each case with cyclic 4DVar assimilation. The conventional surface and upper-air observations are assimilated in combination with atmospheric motion vectors (AMVs) and ocean surface winds. Free forecast for 48 h is performed from respective final analysis fields for all the experiments. The results indicate that the uv -based analysis fields are closer to the observations. A comparative analysis of the 4DVar experiments with the 3DVar DA system provided a critical insight on the role of the 4DVar DA system on implicitly accounting for the multivariate correlations. The precipitation forecasts confirm the improved performance of the ψχ -based experiment, when multivariate nature of the humidity is incorporated. The time evolution of the intense rainfall episodes over the location of maximum rainfall are relatively well reproduced in the uv -based experiment. The results indicate that the inclusion of multivariate humidity variable in the BEC formulation does have a significant impact on suppressing the excessive overestimation in rainfall intensity.
- Publication:
-
Dynamics of Atmospheres and Oceans
- Pub Date:
- September 2022
- DOI:
- 10.1016/j.dynatmoce.2022.101304
- Bibcode:
- 2022DyAtO..9901304G
- Keywords:
-
- Data assimilation;
- 4DVar;
- WRF;
- Background error covariance matrix;
- Extreme rainfall events