On the Need for Perturbed LBCs in Limited-Area Ensemble Forecasts

Nutter (2003) showed that the use of coarsely resolved and/or temporally interpolated lateral boundary conditions (LBCs) is sufficient to cause underdispersive limited area model (LAM) ensemble forecasts. The effect is present in all typical modeling systems using “one-way” LBC forcing and suggests the need to apply statistically consistent, finescale LBC perturbations at every time step throughout the LAM simulations. In this paper, we present a method for introducing LBC perturbations that is designed specifically to help restore the loss of LAM ensemble dispersion using a simplified model system. The use of coarsely resolved and temporally interpolated LBCs restricts small scale error growth in LAMs because of the LBC “sweeping” effects that were documented by Nutter et al. (2004) using a carefully controlled approach. This limitation of LAMs has important consequences because it artificially inflates estimates of predictability limits which, in turn, decreases forecast uncertainty. The LBC sweeping mechanism associated with coarsely resolved LBCs has been considered previously by a number of authors including Errico and Baumhefner (1987), Vukicevic and Errico (1990), Paegle et al. (1997), Warner et al. (1997), Hou et al. (2001), and De Elı́a and Laprise (2002). Nutter et al. (2004) additionally considered error growth constraints introduced by temporal interpolation between relatively infrequent LBC updates and extended the analysis to quantify the impact on LAM ensemble dispersion. They concluded that it is difficult to construct an efficacious LAM ensemble system unless these LBC constraints on error growth are relaxed. To help improve the design of LAM ensemble systems, LBC perturbations are introduced in this paper to help counter LBC sweeping effects by creating a statistically consistent source of error growth along the lateral boundaries. The LBC perturbations are designed to amplify with time while coherently propagating into the domain. The perturbations are shown to help restore small-scale error growth that would otherwise be swept away through the downstream boundary.