Issues in high resolution limited area data assimilation for quantitative precipitation forecasting

High resolution limited area data assimilation is a key aspect of improved quantitative precipitation forecasting. With advances in computer power it is now possible to envision operational models running at resolutions of O(1 km). To increase forecast skill, improvements in dynamics must be accompanied by improvements in initial conditions. Observations, such as radar winds and reflectivities, have the potential to provide enough high resolution information for good mesoscale analyses. A number of questions have to be addressed before this can become a reality. We review a number of issues in high resolution limited area data assimilation, including background and observation errors, balance, scale disparities, position errors and lateral boundary conditions. We go on to consider various methods for assimilation at high resolution and their relative advantages and disadvantages. In the operational context, the high computational cost of each method must be weighed against forecast improvements.