Defining the Influence of Horizontal Grid Spacing on Ensemble Uncertainty within a Regional Modeling Framework

Abstract

Numerical weather prediction (NWP) models are limited with respect to initial and boundary condition data and possess an incomplete description of underlying physical processes. To account for this, modelers have adopted the method of ensemble prediction to quantify the uncertainty within a model framework; however, the generation of ensemble members requires considerably more computational time and/or resources than a single deterministic simulation, especially at convection-allowing horizontal grid spacings. One approach to solving this issue is the development of both a large and small horizontal grid spacing model framework over the same domain for ensemble and deterministic simulations, respectively. This approach assumes that model grid spacing has no influence on model uncertainty; therefore, the objective of this paper is to quantify the influence of horizontal grid spacing on the statistical spread of NWP model ensembles over a regional domain. A series of 24-h simulations using the Weather Research and Forecast (WRF) Model are generated over a static domain with horizontal grid spacings of 35, 25, 15, and 9 km, using both a stochastic kinetic energy backscatter scheme and a multiphysics ensemble approach. Results indicate that horizontal grid spacing does influence the magnitude of uncertainty within an ensemble, although the exact magnitude and type of statistical relationship (direct versus inverse) varies by case. As such, at shorter lead times (<12 h) the dominant atmospheric process associated with each event and the type of ensemble being used outweigh the individual impacts of horizontal grid spacing on ensemble spread.

A Hydraulic MultiModel Ensemble Framework for Visualizing Flood Inundation Uncertainty

“…), perturb initial conditions of a single model (Toth and Kalnay ; Dyer et al. ), or vary the physics of the single model (Stensrud et al. ; Du et al.…”

“…; Dyer et al. ). A challenge with multi‐physics ensembles is that the underlying distribution cannot be attributed to a common probability density function (PDF) or climatology.…”

“…The greater range of outcomes, or lower agreement among members, coincides with higher uncertainty in the model prediction. Three common approaches to generating an ensemble are to use statistical models to generate a time-lagged ensemble (Regonda et al 2013), perturb initial conditions of a single model (Toth and Kalnay 1997;Dyer et al 2016), or vary the physics of the single model (Stensrud et al 2000;Du et al 2004;Dyer et al 2016). A challenge with multi-physics ensembles is that the underlying distribution cannot be attributed to a common probability density function (PDF) or climatology.…”

An Examination of the Impact of Grid Spacing on WRF Simulations of Wintertime Precipitation in the Mid-Atlantic United States

“…Since computing resources are limited, it is important to know how much additional quality (e.g., Tracton 2008) would be gained by reducing grid spacing from convection-parameterizing to convection-allowing and/or cloud-allowing. For instance, it is recognized that model forecast errors grow more quickly as grid spacing decreases (Weygandt and Seaman 1994;Roebber et al 2004;Dyer and Zarzar 2016), leading more quickly to forecast displacement errors.…”