Downstream influence of mesoscale convective systems. Part 1: influence on forecast evolution

Abstract

Mesoscale convective systems (MCSs) are difficult to forecast due to their inherent‐60 unpredictability and development from scales that are subgrid in typical global models. Here the impacts of model representation of convection on MCS structure and downstream forecast evolution are examined using two configurations of the Met Office Unified Model: the convection‐permitting (4.4‐km grid spacing) limited‐area Euro4 and convection‐parametrizing (25‐km grid spacing) Global configurations. MCSs are associated with a characteristic potential vorticity (PV) structure: a positive PV anomaly in the mid‐troposphere and negative PV anomalies above and to the side of it. Convection‐permitting models produce larger‐amplitude MCS PV anomalies than convection‐parametrizing models. These differences are shown to persist after coarse graining the output from a Euro4 simulation to the 25 km grid spacing of the Global configuration for a case study from July 2012, and are largest in magnitude and extent in the upper troposphere. The effect of the poor representation of this PV structure by convection‐ parametrizing models on forecasts is investigated by adding “MCS perturbations”, calculated as differences between the coarse‐grained Euro4 and the Global outputs, to five‐day Global configuration forecasts. Upper‐level MCS perturbations lead to greater forecast differences than those at middle levels, though using perturbations at all levels yields the greatest impact. For the first 36 hr, differences grow on the convective scale related to the MCS and its influence on a developing UK cyclone, despite perturbation amplitudes initially reducing. Subsequently, differences grow rapidly onto the synoptic scale and by five days impact the entire Northern Hemisphere. MCS perturbations slow the eastward movement of Rossby waves due to ridge amplification. Thus, perturbing convection‐parametrizing models to include PV anomalies associated with MCSs produces synoptic‐scale forecast differences implying that the misrepresentation of the PV structures associated with MCSs are a potential source of forecast errors.

A simple ensemble approach for more robust process‐based sensitivity analysis of case studies in convection‐permitting models

“…In particular, such model configuration changes can modify the timing and location of the weather event of interest (or even prevent it occurring). For example, Done et al (2006) demonstrated that a mesoscale convective system simulated with explicit convection can have a very different structure (and associated precipitation rates) to that simulated using parametrized convection, even at the same resolution, and Clarke et al (2019) demonstrate that misrepresentation of mesoscale convective system structure by convective parametrization schemes can impact downstream forecasts. Given the chaotic and indeterminate nature of the atmosphere, the possibility for numerical artefacts (e.g.…”

Downstream influence of mesoscale convective systems. Part 2: Influence on ensemble forecast skill and spread

“…We also found in Clarke et al . () that increasing the amplitude of the perturbations had no effect on the spatial extent of the impact downstream, but did cause the differences to be larger in magnitude (similarly, also found in Zhang et al . (); see their figure 5).…”

“…In a companion article (Clarke et al . ) that used the same case study, the PV anomalies associated with two MCSs, termed MCS‐A and ‐B, that formed over France were shown to be more intense, even after coarse‐graining to the same resolution, in the output from the Euro4 configuration simulation than in that from the Global configuration simulation. This result suggests that the poor representation of the MCSs by convection‐parametrizing models can impact synoptic‐scale medium‐range flow evolution.…”

“…We suggest that the scv scheme could have been more effective by using winds and temperatures associated with negative PV anomalies at upper levels instead of zero‐PV anomalies, since we found in Clarke et al . () that the negative PV anomalies at upper levels have the largest impact on the downstream forecast. It is apparent that taking account of the unpredictability of MCSs (or other deep convection) and their misrepresentation by convection‐parametrizing models, particularly at upper levels, where the largest errors may develop due to divergent outflow from convection, is required to produce an ensemble with a large enough spread.…”

“…A case study from July 2012, chosen as a typical example of European MCSs and including a MCS that tracked over the UK, is used. In the companion study by Clarke et al (2019), hereafter "Part 1", it has been shown that the PV structure associated with the UK-tracking MCS in the convection-permitting simulation output is more intense, even after coarse-graining to the resolution of a convection-parametrizing simulation, than the equivalent PV structure in the output of that convection-parametrizing simulation. Here, the derived MCS perturbations are added to an ensemble of convection-parametrizing model simulations to determine the impact on ensemble forecast skill and spread.…”