Title: | ICTS Data Assimilation Program: Numerical Weather Prediction and Data Assimilation (Lecture 1) |
Speaker: | Dr. Martin Weissmann, Hans-Ertel-Centre for Weather Research, Meteorologisches Institut, LMU, Germany |
Time/Place: | 11:00 - 12:00 FSC1217, Fong Shu Chuen Library, HSH Campus, Hong Kong Baptist University |
Abstract: | Improvements in data assimilation, both concerning data assimilation methodology and the use of additional satellite observations, have been a major contributor to advances in numerical weather prediction (NWP). Modern global NWP models assimilate several million satellite observations per day from a variety of different instruments. Traditionally, two different approaches are used for data assimilation in meteorology, variational methods and ensemble Kalman filters. Both exhibit certain advantages and disadvantages and currently all major weather centers either deploy or work towards hybrid methods for their global NWP models that combine the advantages of both. In addition to global NWP models, many weather centers deploy regional NWP models. These are usually run for a limited domain size, but using a higher horizontal resolution of a few km to explicitly represent atmospheric convection that is a primary forecast event. This poses significant additional challenges for data assimilation as the short life time and stochastic nature of convection require to assimilate very frequent and dense remote-sensing observations as well as to account for the inherent forecast uncertainty. The presentation will review basic concepts of data assimilation and the current status of global and regional data assimilation systems. Additionally, it will provide an overview of the global observing network and special regional observations as well as of ongoing research efforts in meteorological data assimilation. |
Title: | ICTS Data Assimilation Program: The Assimilation of Novel Remote-Sensing Instruments in Km-Scale Weather Prediction Models (Lecture 2) |
Speaker: | Dr. Martin Weissmann, Hans-Ertel-Centre for Weather Research, Meteorologisches Institut, LMU, Germany |
Time/Place: | 14:00 - 15:00 FSC1217, Fong Shu Chuen Library, HSH Campus, Hong Kong Baptist University |
Abstract: | Appropriate initial conditions for convective-scale (km-resolution) NWP models are a crucial requirement for accurate weather forecasts. Due to highly nonlinear dynamics even small initial condition errors can quickly grow to significant forecast errors. Data assimilation for such models is a particularly challenging task given stochastic nature of convection and the subsequent need to incorporate temporally and spatially dense observations. Ensemble systems that conduct multiple parallel forecasts to estimate their uncertainty are a promising approach to address the limited predictability of convection. The German Weather Service therefore currently develops the Km-scale ENsemble Data Assimilation system KENDA. The HErZ research group at LMU works on the refinement of this experimental system, specifically concerning the inclusion of cloud-related satellite observations from geostationary satellites. Clouds are the first signal of atmospheric convection that can be observed area-wide, but their assimilation and foresting is challenging due to unresolved scales, double-penalty errors and strong nonlinearity. Including frequent and dense observations from a variety of observing systems in data assimilation also crucially requires efficient tools to monitor the contribution of different observations to forecast quality that is usually referred to as observation impact. While observation impact can in principle be evaluated by conducting parallel numerical forecast experiments, this becomes infeasible for a large number of different observations. Our research group implemented, refined and evaluated a method to estimate observation impact efficiently based on information from the data assimilation system and correlations of ensemble forecasts to avoid the need for parallel experiments. For the first time, we demonstrated that the method provides reasonable estimates of observation impact in a convective-scale modelling system. |
Title: | Estimating A Large System of Seemingly Unrelated Regressions Using Penalized Maximum Likelihood Estimation |
Speaker: | Dr. FAN Qingliang, Wang Yanan Institute of Studies in Economics, Xiamen University, China |
Time/Place: | 11:00 - 12:00 FSC1217, Fong Shu Chuen Library, HSH Campus, Hong Kong Baptist University |
Abstract: | In this paper, we propose using a shrinkage estimator, named penalized quasi-maximum likelihood estimator, to estimate a large system of seemingly unrelated regressions (SUR) where the number of equations is relatively big compared to sample size. We derive the asymptotic properties of the penalized quasi-maximum likelihood estimators for both error covariance matrix and model coefficients. In particular, we derive the convergence rate of the estimated covariance matrix in terms of Frobenius norm and established sparsistency of the covariance matrix estimator. Asymptotic distributions of the coefficient estimators are also derived. Our simulation results show that when the number of equations is relatively big compared to sample size and the error covariance matrix has sparsity, the penalized likelihood estimator performs much better than traditional estimators, such as OLS, FGLS, MD, and using sample covariance estimator and MLE. We also apply our estimation approach to the study of state level public capital returns in the United States. The method developed in this paper is widely applicable in high dimensional SUR models such as asset pricing models with firm level characteristics. (Joint work with ”Bibo Jiang and Guangming Pan”.) |
Title: | Image Restoration: A Wavelet Frame Based Model for Smooth Functions and Beyond |
Speaker: | Dr. CAI Jianfeng, Department of Mathematics, University of Iowa, USA |
Time/Place: | 17:00 - 18:00 FSC1217, Fong Shu Chuen Library, HSH Campus, Hong Kong Baptist University |
Abstract: | In this talk, we present a new wavelet frame based image restoration model that explicitly treats images as piecewise smooth functions. It estimates both the image to be restored and its singularity set. It can well protect singularities, which are important image features, and provide enough regularization in smooth regions at the same time. This model penalizes the ℓ2- norm of the wavelet frame coefficients away from the singularity set, while penalizes the ℓ1-norm of the coefficients on the singularity set. To further understand the piecewise smooth nature of the obtained solutions, we connect it to a variational model on the space of piecewise smooth functions and prove rigorously that the discrete model converges to the variational model as image resolution goes to infinity. Our numerical studies show that the proposed model is the right one for image restorations, when the underlying solutions are piecewise smooth. |
Title: | Diffusions and Heat Kernels |
Speaker: | Prof. Ovidiu Calin, Department of Mathematics, Eastern Michigan University, USA |
Time/Place: | 11:00 - 12:00 FSC1217, Fong Shu Chuen Library, HSH Campus, Hong Kong Baptist University |
Abstract: | One way of finding heat kernels for second partial differential operators is to compute the transition density of the associated diffusions. This talk will apply this method for a few elliptic and subelliptic operators, such as Laplacians, Kolmogorov, Heisenberg, Grushin, sublaplacian on S^3 and H^3, etc. |
We organize conferences and workshops every year. Hope we can see you in future.
Learn MoreProf. M. Cheng, Dr. Y. S. Hon, Dr. K. F. Lam, Prof. L. Ling, Dr. T. Tong and Prof. L. Zhu have been awarded research grants by Hong Kong Research Grant Council (RGC) — congratulations!
Learn MoreFollow HKBU Math