Cloud-dependent piecewise assimilation based on a hydrometeor-included background error covariance and its impact on regional Numerical Weather Prediction

Abstract

The background error covariance (B) behaves differently and needs to be carefully defined in cloudy areas due to larger uncertainties caused by models’ inability to correctly represent complex physical processes. This study proposes a new cloud-dependent B strategy by adaptively adjusting the hydrometeor-included B in the cloudy areas according to the cloud index (CI) derived from the satellite-based cloud products. The adjustment coefficient is determined by comparing the error statistics of B for the clear and cloudy areas based on the two-dimensional geographical masks. The comparison highlights the larger forecast errors and manifests the necessity of using appropriate B in cloudy areas. The cloud-dependent B is then evaluated by a series of single observation tests and three-week cycling assimilation and forecasting experiments. The single observation experiments confirm that the cloud-dependent B allows cloud dependency for the multivariate analysis increments and alleviates the discontinuities at the cloud mask borders by treating the CI as an exponent. The impact study on regional numerical weather prediction (NWP) demonstrates that the application of the cloud-dependent B reduces analyses and forecasts bias and increases precipitation forecast skills. Diagnostics of a heavy rainfall case indicate that the application of the cloud-dependent B enhances the moisture, wind, and hydrometeors analyses and forecasts, resulting in more accurate forecasts of accumulated precipitation. The cloud-dependent piecewise analysis scheme proposed herein is extensible, and a more precise definition of CI can improve the analysis, which deserves future investigation.

The Potential Value of Combining Assimilation of Water Vapor Radiances and Cloud Liquid/Ice Water Path From ABI on Analyses and Forecasts for Hurricane Irma (2017)

“…However, the ABIAS of MSLP and MWS are more complex for both experiments, and a smaller intensity bias of the EXP_ALL experiment can be seen after 36 hr, which the mean ABIAS of intensity also confirms. It should be noted that despite the additional assimilation of polar-orbiting satellite clear-sky radiance and ABI clear-sky radiance to the EXP_CLR experiment in this study, there seems to be a minor improvement in the forecast results compared to the control experiment in Meng et al (2021aMeng et al ( , 2021b, which only assimilates conventional observations. This could be mainly attributed to the difference in the choice of the cumulus parameterization scheme.…”

“…The released regional GSI v3.7 does not have the hydrometeor control variables, which makes it difficult to assimilate CWP. Therefore, the specific humidity is chosen to transform into the hydrometeor proxies by a series of built‐in formulas assuming that the ratios of water species (including the mixing ratios of water vapor, cloud water, ice, rain, snow, and graupel) remain unchanged during the assimilation iterations (Meng et al., 2021a, 2021b). The assumption can be described by $$$\left\{\begin{array}{c}{Q}_{liquid}^{k}={Q}_{v}^{k}\cdot {R}_{L}^{k}\\ {Q}_{ice}^{k}={Q}_{v}^{k}\cdot {R}_{I}^{k}\end{array}\right.$$$ where $${Q}_{liquid}^{k}$$ (unit: kg/kg) is the sum of the liquid‐phase water cloud, $${Q}_{ice}^{k}$$ (unit: kg/kg) denotes the ice‐phase water cloud, $${Q}_{v}^{k}$$ is specific humidity, which is one of the control variables in the GSI, $${R}_{L}^{k}$$ and $${R}_{I}^{k}$$ are the ratios between liquid/ice cloud and $${Q}_{v}^{k}$$, that is, $$\frac{{Q}_{L}^{k}}{{Q}_{v}^{k}}$$ and $$\frac{{Q}_{I}^{k}}{{Q}_{v}^{k}}$$, respectively, and k indicates the k th model level.…”

“…Hence, it is crucial to accurately characterize the initial moisture conditions by the observations in the hurricane's inner core to produce intensity forecasts of TCs skillfully. The CWP from ABI could be a needed source of observation information for the hurricane inner-core areas and have been successfully assimilated into GSI with the newly developed observation operators built on the water vapor control variables (Meng et al, 2021a(Meng et al, , 2021b. Nevertheless, the mechanistic understanding regarding the CWP assimilation, especially combined with ABI clear radiances, on hurricane track and intensity forecasts has not been thoroughly investigated.…”

“…In addition, Tropical Rainfall Measuring Mission (TRMM)/GPM hydrometeor retrievals have been assimilated into the Gridpoint Statistical Interpolation (GSI) and have shown considerable advantages in hurricane forecasting (Wu et al, 2016;Wu & Zupanski, 2017). Recently, the Cloud Water Path (CWP), including the LWP and IWP, derived from the Advanced Baseline Imager (ABI) (Schmit et al, 2005) onboard the new generation geostationary environmental satellites (GOES-R series) has been assimilated with the new observation operators developed by Meng et al (2021aMeng et al ( , 2021b, improving the hurricane track forecasts. However, their study focuses on validating the LWP & IWP observation operators, not on utilizing all-sky measurements of ABI.…”

Satellite All-sky Infrared Radiance Assimilation: Recent Progress and Future Perspectives

“…Based on the observation error estimation, regularization could be used to better balance the contributions from observations and background (Li and Huang, 1999) to the increment of the atmospheric state, especially under cloudy skies. In addition, appropriately estimating the background error covariance matrix under cloudy skies is very important for successful assimilation (Meng et al, 2021b). In ensemble data assimilation, this necessitates use of adaptive covariance inflation methods.…”