Browsing by Subject "Data assimilation"

Now showing 1 - 3 of 3

Convective-scale data assimilation of thermodynamic lidar data into the weather research and forecasting model
(2022) Thundathil, Rohith Muraleedharan; Wulfmeyer, Volker
This thesis studies the impact of assimilating temperature and humidity profiles from ground-based lidar systems and demonstrates its value for future short-range forecast. Thermodynamic profile obtained from the temperature Raman lidar and the water-vapour differential absorption lidar of the University of Hohenheim during the High Definition of Clouds and Precipitation for advancing Climate Prediction (HD(CP)2) project Observation Prototype Experiment (HOPE) are assimilated into the Weather Research and Forecasting model Data Assimilation (WRFDA) system by means of a new forward operator. The impact study assimilating the high-resolution thermodynamic lidar data was conducted using variational and ensemble-based data assimilation methods. The first part of the thesis describes the development of the thermodynamic lidar operator and its implementation through a deterministic DA impact study. The operator facilitates the direct assimilation of water vapour mixing ratio (WVMR), a prognostic variable in the WRF model, without conversion to relative humidity. Undesirable cross sensitivities to temperature are avoided here so that the complete information content of the observation with respect to the water vapour is provided. The assimilation experiments were performed with the three-dimensional variational (3DVAR) DA system with a rapid update cycle (RUC) with hourly frequency over ten hours. The DA experiments with the new operator outperformed the previously used relative humidity operator, and the overall humidity and temperature analyses improved. The simultaneous assimilation of temperature and WVMR resulted in a degradation of the temperature analysis compared to the improvement observed in the sole temperature assimilation experiment. The static background error covariance matrix (B) in the 3DVAR was identified as the reason behind this behaviour. The correlation between the temperature and WVMR variables in the background error covariance matrix of the 3DVAR, which is static and not flow-dependent, limited the improvement in temperature. The second part of the thesis provides a solution for overcoming the static B matrix issue. A hybrid, ensemble-based approach was applied using the Ensemble Transform Kalman Filter (ETKF) and the 3DVAR to add flow dependency to the B matrix. The hybrid experiment resulted in a 50% lower temperature and water vapour root mean square error (RMSE) than the 3DVAR experiment. Comparisons against independent radiosonde observations showed a reduction of RMSE by 26% for water vapour and 38% for temperature. The planetary boundary layer (PBL) height of the analyses also showed an improvement compared to the available ceilometer. The impact of assimilating a single lidar vertical profile spreads over a 100 km radius, which is promising for future assimilation of water vapour and temperature data from operational lidar networks for short-range weather forecasting. A forecast improvement was observed for 7 hours lead time compared with the ceilometer derived planetary boundary layer height observations and 4 hours with Global Navigation Satellite System (GNSS) derived integrated water vapour observations. With the help of sophisticated DA systems and a robust network of lidar systems, the thesis throws light on the future of short-range operational forecasting.
Microwave forward model for land surface remote sensing
(2015) Park, Chang-Hwan; Wulfmeyer, Volker
In order to improve hydro-meteorological model prediction using remote-sensing measurements the difference between the model world and the observed world should be identified. The forward model proposed in this study allows us to simulate the BT (brightness temperature) from the land surface model to compare with the observed microwave BT. The proposed dielectric mixing model is the key part of the forward model to properly link the model parameters and the BT observed by remote sensing. In this study, it was established that the physically valid computation of the effective dielectric constant should be based on the arithmetic average with consideration of the proposed universal damping factor. This physically based dielectric mixing model is superior to the refractive mixing model or semi-empirical/calibration model with RMSE values of 0.96 and 0.63 for the predicted real and imaginary parts, respectively, compared to the measured values. The RMSE obtained with the new model is smaller than those obtained by other researchers using refractive mixing models for operational microwave remote sensing. Once we determine the model uncertainty using this forward model, we can update the model state using the values obtained from the remote-sensing measurement. The challenging task in this process is to resolve the ill-posed inversion problem (estimation of multiple model parameters from a single BT measurement). This study proposes a simple partitioning factor based on model physics. Again, the forward model is crucial because these factors are required to be computed in BT space. In the case study involving the Schäfertal catchment area, the proposed forward model, including the new dielectric mixing model, and the proper partitioning factors computed from land surface model physics was able to successfully extract the refined soil texture information from the microwave BT measurements. The highly resolved soil moisture variability based on the refined soil texture will allow us to predict convective precipitation with higher spatial and temporal accuracy in the numerical weather forecasting model. Moreover, microwave remote sensing using the developed forward model, which provides the soil texture, soil moisture, and soil temperature with a fine scale resolution, is expected to open up new possibilities to examine the energy balance closure problem with unprecedented realism.
Model evaluation and data assimilation impact studies in the framework of COPS
(2012) Schwitalla, Thomas; Wulfmeyer, Volker
The goal of this thesis was the study of new approaches for improving and investigating quantitative precipitation forecasting (QPF), e.g., by optimizing model resolution, physics combination, and data assimilation. A forecasting system based on the Mesoscale Model 5 (MM5) was compared against other operational numerical weather prediction models from Meteo France, MeteoSwiss and the German Weather Service primarily with respect to daytime precipitation. First, a notable daytime dry bias was observed. It appears to be the result of a too small high-resolution domain and the switched-off convection parameterization from the second to the innermost domain. Even the application of a 4-dimensional variational data assimilation (4DVAR) with GPS slant total delays (STD) does not solve this problem due to inconsistent model physics between the 4DVAR and the forecasting model. Nevertheless, the MM5 is in good agreement with the shape of the observed diurnal cycle after the spin-up phase. As the development of the MM5 was suspended, a transition to the new Weather Research and Forecasting (WRF) model system was made after the D-PHASE period (end of 2007). This system features state-of-the-art physics packages and also a variational data assimilation system. As a new observing system, GPS Zenith Total Delay (ZTD) data from Central Europe were incorporated into the 3-dimensional variational data assimilation (3DVAR) system to further improve the initial water vapor field. A first study with this system revealed an improvement of the integrated water vapor RMSE of about 15% and a small but positive impact on the spatial and quantitative precipitation forecast. Additionally, the importance of assimilating upper air observations and the necessity to select a large, convection permitting model domain emerged. Finally a rapid update cycle (RUC) approach, comparable to operational forecast centers, has been developed for a convection-permitting configuration of the WRF model. The system is capable to assimilate radar observations from Germany and France, GPS-ZTD data and satellite radiances and can be applied even for near real-time applications. First experiments with this system show promising results in comparison to other operational models.