A convective-Scale Ensemble for Meiyu Season Precipitation over the Yangtze-Huaihe Basin of China Based on the WRF

With the increase in computing power, and wide spread development and applications of ensemble prediction systems (EPSs) at global and regional scale, increasingly more emphases have been placed on convective-scale ensemble prediction in the recent years. Convective-scale EPSs use convection-resolving or convection-allowing grid spacing, typically no more than 4 km. It's well recognized that convective-scale forecasting is multi-scale one, as it contains the convective processes and their larger-scale environment. So in order to gain more skillful convective-scale ensemble forecasts, it is necessary to sample the uncertainty of convective-scale as well as the larger-scale.

In the Yangtze-Huaihe Basin of China, precipitation from the Meiyu season that typically spans part of June and July makes up a significant portion of the annual rainfall in the region and is often the cause of major flooding. Meiyu precipitation systems involve multiple scales, including the synoptic scale, the mesoscale and convective scale. Perturbations representing the initial condition uncertainties should include uncertainties at all of these scales. To improve quantitative precipitation forecasting, and to provide probabilistic information associated with such forecasts, an experimental EPS is set up for the region with a 3 km grid spacing based on the WRF model and tested for precipitation episodes during the Meiyu seasons of 2013 and 2014. Multiple combinations of physical parameterization schemes, including those of microphysics and land surface parameterizations, are used in ensemble members to account for model uncertainties. As the first step of our study, two convective-scale ensemble prediction systems with large-scale initial and lateral boundary perturbations have been designed. One uses the forecasts from the NCEP global ensemble forecast system (GEFS) directly as the initial conditions (ICs) and lateral boundary conditions (LBCs) to drive the convective-scale ensemble forecasts, which we call it WRF_GEFS, while the other uses the global forecast system (GFS) analysis and forecasts as control ICs and LBCs, respectively, and constructs the perturbed ICs and LBCs by adding the GEFS perturbations, which is named WRF_LPFS. Preliminary results are concluded as follows: 1. Convective-scale EPS is more skillful in forecasting Meiyu precipitation than the global EPS, when verification against rain gauge observations; 2. Although with the same perturbations, the WRF_LPFS based on the finer GFS forecasts is more skillful than the WRF_GEFS; 3. While sampling large scale uncertainties is a necessary first step towards achieving skillful convective-scale forecasts, the forecast spread resulting from GEFS initial and boundary perturbations is insufficient for the convective-scale EPS because of insufficient sampling of smaller scale uncertainties.

Efforts to design multiple perturbations that better represent uncertainties with precipitation systems during Mei-yu season are under way.

Keywords: convective-scale; Ensemble prediction system (EPS); multi-scale; Meiyu