Feasibility of short-range numerical weather prediction using observations from a network of profilers

A series of observing system simulation experiments was conducted to investigate the feasibility of short-range numerical weather prediction using a network of profilers. A mesoscale model was used to generate data sets that mimic observations from a network of profilers and from an array of rawinsondes. The sensitivity of the model forecast to the characteristic measurement errors of a number of hypothetical profiler networks was tested. The results demonstrate that profiler wind observations would have a positive impact on short-range numerical weather prediction with a simple static initialization. It was also found that forecasts based on retrieved temperatures (calculated from profiler wind data) are significantly better than those based on direct radiometric temperature measurements (using climatology as the first guess for radiometric retrieval). However, the temperature fields from either radiometric measurements or from thermodynamic retrieval need further improvement before they can be as accurate as the radiosonde temperature observations for model initialization. Various hypothetical networks, each having a regular array of stations at a separation of 360 km, provided the initial conditions for short-range numerical forecasts. These predictions can be ranked by performance in the following order: 1) profiler wind with radiosonde temperature and moisture; 2) mixed profiler and rawinsonde wind with rawinsonde temperature and moisture; 3) rawinsonde wind, temperature, and moisture; 4) profiler wind and moisture with retrieved temperature; and 5) profiler wind, temperature and moisture. It was found that, with a domain of 4320 x 2880 km centered at 40 degrees N and a grid spacing of 40 km, accuracy in both the wind field and the temperature field is needed to define the initial state of the model properly. Even within the mesoscale range, the wind field and the temperature field adjust to each other during the course of the model integration. This is because temperature and wind errors associated with observing systems are often projected onto several different vertical modes at a wide range of horizontal scales, both larger and smaller than the Rossby radius of deformation, thus forcing the mutual adjustment of wind and mass fields. These conclusions are considered tentative because only one synoptic situation was tested with a simple static initialization procedure. Further modelling studies should utilize a four-dimensional data assimilation tec