The Met Office convective‐scale ensemble, MOGREPS‐UK

Case studies remain an important method for meteorological parameter sensitivity process studies. However, these types of study often use just a few case studies (typically up to three) and are not tested for statistical significance. This approach can be problematic at the convective scales, since uncertainty in the representation of an event increases, and the variability in the atmosphere arising from convective‐scale noise is not routinely taken into account. Here we propose a simple ensemble method for performing more robust sensitivity analysis without the need for an operational‐style ensemble prediction system and demonstrate it using a case study from the 2005 Convective Storm Initiation Project. Boundary‐layer stochastic potential temperature perturbations with Gaussian spatial structure are used to create small ensembles to examine the impact of increasing cloud droplet number concentration (CDNC) on precipitation. Whilst there is a systematic difference between the experiments, such that increasing the CDNC reduces the precipitation, there is also an overlap between the different ensembles implying that convective‐scale variability should be taken into account in case study process‐based sensitivity studies.

Section: Introductionmentioning

confidence: 97%

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

Flack

Gray

Plant

2019

“…, ), Deutscher Wetterdienst (DWD) and the UK Met Office (Hagelin et al. , ), where member selection is obviated by setting up the global ensembles with the same number of members as the respective limited‐area ensembles. Other centres, such as the Spanish Meteorological Service (García‐Moya et al.…”

Section: Summary and Discussionmentioning

confidence: 99%

“…() and Hagelin et al. () at the UK Met Office, Gebhardt et al. () at Deutscher Wetterdienst, and Bouttier et al.…”

Section: Introductionmentioning

confidence: 99%

Clustering and selection of boundary conditions for limited‐area ensemble prediction

Bouttier

Raynaud

2018

Limited‐area ensemble predictions can be sensitive to the specification of lateral boundary conditions, which are often built by subsampling larger ensembles. Using the operational PEARP and AROME‐EPS ensembles, we compare several subsampling methods, including random selection, representative members, and a new selection method. The tests show that the algorithms used for the clustering and the member selection have a significant impact on the resulting ensembles. Clustering‐based methods are shown to outperform random subsampling, mostly (but not only) because they change the ensemble spread. Cluster sizes can be highly variable, which can complicate ensemble interpretation. We present a subsampling algorithm that has little impact on performance scores, but better preserves ensemble spread and produces nearly equally likely members by limiting cluster size variability.

“…Hagelin et al . () found that ensemble size matters mostly for the spatial details in the precipitation areas, so perhaps influencing the smallest spatial scales. To check if the impact on the spread of the precipitation is dependent upon the size of the ensemble, we have recalculated dFSSmean for three sets of ensemble members from EC‐SINGV (Figure ), approximately equal in size to UM‐SINGV.…”

Section: Spatial Ensemble Spreadmentioning

confidence: 97%

“…Significant, albeit small, differences can still hold even for an ensemble size of 30 members (again in agreement with studies over midlatitudes as in Raynaud and Bouttier () and Hagelin et al . ()). The growth in errors happens during the daytime and night‐time convection and the impact of ensemble size is largest at these times.…”

Section: Monthly Analysis For October 2017mentioning

confidence: 99%

Extreme rainfall sensitivity in convective‐scale ensemble modelling over Singapore

Porson

Hagelin

Boyd

et al. 2019

Self Cite

A convective‐scale ensemble system was developed to predict the occurrence of heavy convective rainfall around Singapore with a focus on the prediction of high‐impact events. The new ensemble SINGV‐EPS has been nested within two global ensembles, MOGREPS‐G (UK Met Office) and EC‐ENS (ECMWF). Predicting the occurrence of convective rainfall in an area such as Singapore is challenging and this article discusses the use of the convection‐permitting ensemble to characterize the uncertainties in the prediction of such localized heavy rainfall. First, verification of wind, temperature, and precipitation is performed for a month‐long period to assess the relative performance of each ensemble. This reveals differences, but no robust signal to say one is better than the other. The results are not statistically significant and not all variables are consistently better with one ensemble or the other. Secondly, the precipitation characteristics of SINGV‐EPS are analysed from probabilities of precipitation and variability among the ensemble members. SINGV‐EPS is sensitive to the choice of the global ensemble providing the initial conditions and boundaries. The results suggest there is benefit, in some cases, from combining the two ensembles. Thirdly, the spread of the ensemble precipitation is analysed using the dispersion Fractions Skill Score (dFSS). We compare the impact of the initial perturbations and the perturbations in lateral boundary conditions in both nesting options. The initial perturbations dominate in the beginning of the forecasts, with influence up to T+24 h, and are associated with an upscale growth of the uncertainties. The impact of the parent ensemble and lateral boundary conditions dominate at the end of the forecast and tend to influence larger scales more.