Impact Studies of Introducing a Large-Scale Constraint into the Kilometer-Scale Regional Variational Data Assimilation

Framework design and observation selection are mainly for meso- and small-scale analyses; hence, kilometer-scale data assimilation (DA) systems often suffer from insufficient large-scale analysis capabilities. This work adds an extra large-scale constraint to the cost-function of the GRAPES (Global/Regional Assimilation and Prediction System) regional 3-km variational DA framework to study the impacts of introducing large-scale information of the global system on the kilometer-scale DA and forecast. Results of numerical experiments in one month show that the introduction of a large-scale constraint can greatly improve the analysis and forecast capabilities of the synoptic situation field, increase the precipitation forecast scores, and reduce the analysis and forecast error of the 2 m temperature and 10 m wind. Furthermore, results of the quantitative precipitation sensitivity tests show that the large-scale constraint of the temperature and humidity field is a crucial factor in improving the precipitation scores. Results also indicate that the humidity field constraint is important for reducing the precipitation false alarm and improving the TS (Threat Score) scores for short-term precipitation forecast, while the temperature field constraint is important for improving the TS scores for longer forecast ranges. In addition, under the condition of introducing the large-scale constraint, the analysis and prediction results of the experiment with the full cycling scheme (no cold start during one month of cycling) are equivalent to that of the experiment with a partial cycle (daily cold start). This laid a good foundation for the GRAPES kilometer-scale system to adopt the full cycling scheme to further simplify the cycle process and reduce the calculation consumption.