Convective Parameterization In a 2.5 km NWP Model: Improvements of Summer Precipitation Diurnal Cycle

Abstract:

A new high resolution Global Environmental Multiscale (GEM; Côté et al.) model version is currently developed at Environment Canada as part of the next operational High Resolution Deterministic Prediction System (HRDPS, Mailhot et al., 2010) with an horizontal grid-spacing of 2.5 km. At this resolution, the question of using or not a convective parameterization (CP) remains.

Continental precipitation diurnal cycle is a critical aspect of numerical weather prediction: when no CP is applied, delayed precipitation is often seen even at 2 km grid-spacing (Xu et al. 2002) whereas the use of CP often show too early precipitation (Rio et al. 2009). Moeng et al. (2010) have shown that models with horizontal grid-spacing of the order of km are missing important part of subgrid-scale moisture vertical fluxes throughout the deep convective layer. However, CP were generally developed for lower resolution models with hypotheses (i.e. convective quasi-equilibrium, local compensating subsidence) that are not always valid at the km scale. New approaches are proposed to relax those constraints (e.g. Arakawa et al., 2011, Kuell and Bott, 2008).

The Kain-Fritsch CP (Kain and Fritsch, 1990, 1993) is already used in two GEM versions with grid-spacing of 25 and 10 km (Charron et al. 2012). It has been adapted for the HRDPS by decreasing the convective updraft radius, minimum depth and adjustment time scale, while increasing the triggering threshold. It has been tested over a Pan-Canadian domain for 20 summer cases of 48h simulations.

It will be shown that, when using the adapetd Kain-Fritsch CP, a reduction in diurnal cycle biases is seen, particularly for large precipitation events, while it doesn't degrade other surface skill scores, except for trace precipitation bias. A squall line case (see figure) will be presented against observations, to illustrate how the precipitation timing, intensity and structure is modified with the use of Kain-Fritsch and how it is sensitive to the different Kain-Fritsch parameters.