Final Report Summary - JOINT-ASSIMILATION (Joint assimilation of satellite aerosol, cloud, and precipitation observations in numerical models to support climate and hydrologic applications)
Project objectives
The objective of this project is to use joint aerosol cloud and precipitation assimilation to understand the impact of aerosols on clouds and precipitation, and to develop better methodologies for short precipitation forecasting. To achieve these objectives, a Kalman filter assimilation framework has been developed. The cloud-resolving model used within the framework is the Weather Research and Forecasting Model (WRF). A satellite radiance simulator, which is able to produce radiometer satellite observations in the microwave and visible/infrared domains, has been developed and coupled to the WRF. The satellite radiance simulator allows for the direct assimilation of passive microwave and visible/infrared observations into the WRF.
Work performed
The WRF-based assimilation framework has been investigated through its application to the reforecast of several flash-flood-producing storms that occurred recently in South-eastern Europe (e.g. Romania, Italy and Crete). To correct for biases in the dynamics and microphysics of the ice phase, an expectation maximisation procedure has been developed; i.e. height-dependent hydrometeror bias correction factors have been introduced to account for the systematic differences between simulated and observed brightness temperatures. While cloud microphysical schemes have continuously improved over the years, they still tend to produce relationships between precipitation fluxes above and below the freezing level that are systematically different from those derived from radar observations. These deficiencies in cloud microphysical schemes can result in systematic differences between simulated and observed brightness, which, unless corrected, produce biases in the assimilated surface precipitation. Thus, an expectation maximisation procedure was developed in the project to make use of independent information about the vertical distribution of hydrometeors provided by independent space and ground radars. More specifically, bias correction factors defined as the ratio of model simulated hydrometeor water contents to the expected water contents were introduced into the forward brightness temperature operators. These factors were assumed to be known in the assimilation step. In a subsequent step, they were updated to make the vertical covariances of simulated reflectivity consistent with those derived from independent radar observations. This procedure was found to properly mitigate microphysical biases.
To investigate the impact of aerosol information assimilation on precipitation forecasting, a WRF cloud microphysics scheme has been modified to account for variability in the initial distribution of cloud condensation nuclei. In its original formulation, the WRF double-moment microphysics scheme assumes a single (hard-coded) value for the cloud condensation nuclei concentration. The scheme has been modified to be initialised by 3D cloud condensation nuclei concentrations based on Moderate Resolution Imaging Spectroradiometer (MODIS) retrievals. Although the application of this procedure is deemed to improve the analysis of observed precipitating systems, its impact on the forecast could not be reliably quantified due to other types of uncertainties. Nevertheless, the ability to initialise the WRF double-moment scheme based on actual observations is an important component of this project.
Main results
The application of the assimilation procedure revealed that for systems characterised by weak large-scale forcing, the incorporation of microwave brightness temperatures improves the very short-term precipitation forecast (up to two hours), but it has negligible impact on long-term precipitation forecast. However, for systems characterised by strong large-scale forcing and exhibiting spatial organisation assimilation of microwave brightness temperature, it has a longer lasting impact (6 to 12 hours).
WRF microphysical schemes improve with each release and this project revealed that microphysics has significant impact on the dynamics. Flash-flood-producing storms are usually characterised by very little propagation, their stationarity being not only the result of large-scale dynamics but also of the interaction between incoming low-level flow and the storm's cold pool. Too much (or too little) evaporation, which depends on the initial distributions of cloud condensation nuclei, can significantly affect the strength of the cold pool and its interaction with the incoming flow. When the assimilated observations do not include ground observations (which was the case in this project), it is difficult to separate the effect of microphysics from that of large-scale dynamics, but the inclusion of additional data to better characterise the large-scale conditions is expected to reveal the importance of assimilating cloud condensation nuclei information.
The objective of this project is to use joint aerosol cloud and precipitation assimilation to understand the impact of aerosols on clouds and precipitation, and to develop better methodologies for short precipitation forecasting. To achieve these objectives, a Kalman filter assimilation framework has been developed. The cloud-resolving model used within the framework is the Weather Research and Forecasting Model (WRF). A satellite radiance simulator, which is able to produce radiometer satellite observations in the microwave and visible/infrared domains, has been developed and coupled to the WRF. The satellite radiance simulator allows for the direct assimilation of passive microwave and visible/infrared observations into the WRF.
Work performed
The WRF-based assimilation framework has been investigated through its application to the reforecast of several flash-flood-producing storms that occurred recently in South-eastern Europe (e.g. Romania, Italy and Crete). To correct for biases in the dynamics and microphysics of the ice phase, an expectation maximisation procedure has been developed; i.e. height-dependent hydrometeror bias correction factors have been introduced to account for the systematic differences between simulated and observed brightness temperatures. While cloud microphysical schemes have continuously improved over the years, they still tend to produce relationships between precipitation fluxes above and below the freezing level that are systematically different from those derived from radar observations. These deficiencies in cloud microphysical schemes can result in systematic differences between simulated and observed brightness, which, unless corrected, produce biases in the assimilated surface precipitation. Thus, an expectation maximisation procedure was developed in the project to make use of independent information about the vertical distribution of hydrometeors provided by independent space and ground radars. More specifically, bias correction factors defined as the ratio of model simulated hydrometeor water contents to the expected water contents were introduced into the forward brightness temperature operators. These factors were assumed to be known in the assimilation step. In a subsequent step, they were updated to make the vertical covariances of simulated reflectivity consistent with those derived from independent radar observations. This procedure was found to properly mitigate microphysical biases.
To investigate the impact of aerosol information assimilation on precipitation forecasting, a WRF cloud microphysics scheme has been modified to account for variability in the initial distribution of cloud condensation nuclei. In its original formulation, the WRF double-moment microphysics scheme assumes a single (hard-coded) value for the cloud condensation nuclei concentration. The scheme has been modified to be initialised by 3D cloud condensation nuclei concentrations based on Moderate Resolution Imaging Spectroradiometer (MODIS) retrievals. Although the application of this procedure is deemed to improve the analysis of observed precipitating systems, its impact on the forecast could not be reliably quantified due to other types of uncertainties. Nevertheless, the ability to initialise the WRF double-moment scheme based on actual observations is an important component of this project.
Main results
The application of the assimilation procedure revealed that for systems characterised by weak large-scale forcing, the incorporation of microwave brightness temperatures improves the very short-term precipitation forecast (up to two hours), but it has negligible impact on long-term precipitation forecast. However, for systems characterised by strong large-scale forcing and exhibiting spatial organisation assimilation of microwave brightness temperature, it has a longer lasting impact (6 to 12 hours).
WRF microphysical schemes improve with each release and this project revealed that microphysics has significant impact on the dynamics. Flash-flood-producing storms are usually characterised by very little propagation, their stationarity being not only the result of large-scale dynamics but also of the interaction between incoming low-level flow and the storm's cold pool. Too much (or too little) evaporation, which depends on the initial distributions of cloud condensation nuclei, can significantly affect the strength of the cold pool and its interaction with the incoming flow. When the assimilated observations do not include ground observations (which was the case in this project), it is difficult to separate the effect of microphysics from that of large-scale dynamics, but the inclusion of additional data to better characterise the large-scale conditions is expected to reveal the importance of assimilating cloud condensation nuclei information.