Optimal configuration and resolution for the first convection‐permitting ensemble of climate projections over the United Kingdom

Abstract

Convection‐permitting models (CPMs) provide a better representation of sub‐daily precipitation statistics and convective processes on both climate and weather forecasting timescales. This study aims to establish what horizontal resolution and physical process settings are suitable for the first ensemble of convection‐permitting climate projections for the United Kingdom. For this purpose, 12‐year‐long simulations were run using three different convection‐permitting resolutions (grid spacing of 4 km, 2.2 km and 1.5 km) each with several configurations. The focus is on the ability to represent sub‐daily precipitation due to its potential for high impacts on society through flooding. The analysis shows that the use of a finer grid spacing within the convection‐permitting regime improves the representation of sub‐daily precipitation but the added value decreases as the grid becomes increasingly finer, while the computational costs substantially increase. Changes in the representation of physical processes can have as much impact as the grid spacing and can lead to similar differences in the diurnal cycle as between convection‐parametrized models and CPMs. Although the 4km model shows some key deficiencies that make it unreliable for climate projection, the 2.2km model performs as well as the 1.5km model for most of the metrics examined with considerable benefits in terms of computer cost. The optimal configuration and resolution identified here is being used for an ensemble of simulations run within the UK Climate Projections 2018 (UKCP18) project, which allows, for the first time, an estimate of uncertainty in future changes at convection‐permitting scale and providing more reliable climate change projections at local and hourly scales.

Changing Spatial Structure of Summer Heavy Rainfall, Using Convection‐Permitting Ensemble

“…The CPM simulation for historical Lewis et al, 2019). The hourly rainfall characteristics closely matches observations in terms of the diurnal cycle, heavy rainfall amount, event duration, hourly extremes (Kendon et al, 2019;Fosser et al, 2019) and its spatial structure Kendon et al, 2012) (also in Text S4).…”

Europe-wide precipitation projections at convection permitting scale with the Unified Model

“…We note that the use of an intermediate nest simulation improves the spin-up of small-scale atmospheric features, and alleviates the precipitation biases for smaller domain simulations (Fosser et al. 2019 ). Beyond Europe, large-domain CPM simulation has also been conducted for North America (Prein et al.…”

“…No intermediate nest is used; the 25 km "N512" HadGEM3 and 79 km ERA-Interim data directly drives the 2.2 km CPM model with downscaling ratios of ≈ 1 ∶ 11 and 1:36, respectively. • Recent UKMO CPM work suggests that nested simulation configurations with reduced downscaling ratios improve the ERA-Interim hindcast simulation by improving the spin up of small scale features (Fosser et al 2019). That recommendation comes in hindsight, and is untested for different resolution jumps and for a continental-scale CPM simulation where the buffer between the lateral boundaries and area of interest (i.e.…”

“…Operational numerical weather prediction (NWP) has shown that small domain size simulations often lead to inferior simulations due to a large spin-up region with negative precipitation biases extending considerably into the simulation domain (Lock et al, personal communication). We note that the use of an intermediate nest simulation improves the spinup of small-scale atmospheric features, and alleviates the precipitation biases for smaller domain simulations (Fosser et al 2019). Beyond Europe, large-domain CPM simulation has also been conducted for North America (Prein et al 2017b) although this is a pseudo-global warming simulation (Schär et al 1996); it is driven by reanalysis with a climate change signal superposed.…”

Convection‐Permitting Models Offer Promise of More Certain Extreme Rainfall Projections

“…Each RCM member is in turn driven by a member of the Hadley Centre global coupled (HadGEM3-GC3.05) perturbed parameter ensemble at 60-km resolution. Based on previous literature (e.g., Déqué et al, 2012), an ensemble size of 12 members was considered large enough for a first evaluation of the uncertainties in future changes and this was all that could be afforded with the available computational resource, for the selected CPM resolution and simulation length (Fosser et al, 2020).…”