Reducing the risk associated with natural disasters is very important for the protection of people's lives and welfare.In this paper,some problems in whole links of monitoring,prediction and disaster management of meteorological disasters are identified and potential improvements are discussed.In the monitoring aspect,issues discussed include:needs in improving the accuracy and vertical resolution of meteorological elements using existing polar-orbiting and geostationary operational meteorological satellites;setting up small satellite networks to enhance monitoring in the tropics and lower-latitudinal zones,to provide necessary data for improved numerical weather prediction and for early detection of hurricanes,typhoons and other disastrous weather systems;improving Doppler radar to provide more cloud and rain structure information and micro-physical parameters;and using the so-called targeted observations in urgent situations for obtaining dense and detailed data using movable observation equipment.In the prediction aspect,there are problems with using bidirectional nesting of multiple models with different resolutions;parameterizing smaller scale processes and some diagnostic variables (output from the model predictions,such as precipitation);using the coupled atmosphericoceanic model and its complex coupled boundary layer;and methods for assimilating 4D data.There are also new and unsolved problems concerning predictions of typhoons,heavy rains,tornados and wind gusts.In the disastermanagement aspect,requirements include:development of numerical models for predicting disaster situations and the related risks of losses and the mathematical formulations and numerical models for optimal management decisions and actions for risk reduction of disasters,as well as real-time regulation.For efficient management of meteorological disasters,supercomputing plays a key role,and close cooperation with other branches of science and technology are very important.