Pan-European climate at convection-permitting scale: a model intercomparison study

Abstract

We investigate the effect of using convection-permitting models (CPMs) spanning a pan-European domain on the representation of precipitation distribution at a climatic scale. In particular we compare two 2.2 km models with two 12 km models run by ETH Zürich (ETH-12 km and ETH-2.2 km) and the Met-Office (UKMO-12 km and UKMO-2.2 km). The two CPMs yield qualitatively similar differences to the precipitation climatology compared to the 12 km models, despite using different dynamical cores and different parameterization packages. A quantitative analysis confirms that the CPMs give the largest differences compared to 12 km models in the hourly precipitation distribution in regions and seasons where convection is a key process: in summer across the whole of Europe and in autumn over the Mediterranean Sea and coasts. Mean precipitation is increased over high orography, with an increased amplitude of the diurnal cycle. We highlight that both CPMs show an increased number of moderate to intense short-lasting events and a decreased number of longer-lasting low-intensity events everywhere, correcting (and often over-correcting) biases in the 12 km models. The overall hourly distribution and the intensity of the most intense events is improved in Switzerland and to a lesser extent in the UK but deteriorates in Germany. The timing of the peak in the diurnal cycle of precipitation is improved. At the daily time-scale, differences in the precipitation distribution are less clear but the greater Alpine region stands out with the largest differences. Also, Mediterranean autumnal intense events are better represented at the daily time-scale in both 2.2 km models, due to improved representation of mesoscale processes.

Appendix

1.1 Daily datasets

FRANCE: SAFRAN (8 km)  Systeme d’Analyse Fournissant des Renseignements Atmospheriques á la Neige (SAFRAN) is a precipitation analysis for continental France that uses an optimal interpolation method. One of the main features of SAFRAN is that the analyses are performed over climatically homogeneous zones, which are areas of irregular shape covering a surface usually smaller than 1000 km and where the horizontal climatic gradients (especially for precipitation) are weak. SAFRAN estimates one value of each parameter for each zone at several altitude levels. Within the zone, analyzed parameters depend only on elevation and aspect. First, SAFRAN performs a quality control of the observations. This is an iterative procedure based on the comparison between observed and analyzed quantities at the observation location. There were 3675 measurement stations for 2004/2005. The precipitation analysis is performed daily at 0600 UTC, to include in the analysis the numerous rain-gauges that measure precipitation on a daily basis (in particular in the climatological and snow networks). The first guess is a very simple and constant field. An hourly separation is then performed, but in this study we use the daily precipitation amount. Further description can be found in Quintana-Segui et al. (2008).

ALPS: APGD_EURO4M (5 km)  The Alpine rain-gauge dataset typically comprises 5500 observations on any day of the period 1971–2008. The analysis is based on a first guess for a day that is the long-term mean precipitation (period 1971–1990) of the relevant calendar month. The precipitation-elevation relationship is calculated locally and taken into account in this first guess. Then an anomaly is computed for every grid point using the stations located within a radius that depends on the station density. It can be up to 60 km from the grid point. The dataset has a 5 km resolution, but its effective resolution is closer to 10–15 km. The dataset is provided by the Federal Office of Meteorology and Climatology MeteoSwiss. The dataset incorporates local precipitation topography relationships at the climatological time-scale, which aims at reducing the risk of systematic underestimates at high elevations but does not correct for any gauge undercatch, which is comparatively larger during episodes with strong wind and during weather with low rainfall intensity or with snowfall. Sevruk and Zahlavova (1994) and Richter (1995) estimated measurement errors ranging from 7% (5%) over the flatland regions in winter (summer) to 30% (10%) above 1500 m in winter (summer). Further description can be found in Isotta et al. (2014).

SPAIN: Spain02 (0.11$^\circ$)  Daily precipitation gridded dataset developed for peninsular Spain and the Balearic Islands using 2756 quality-controlled stations over the time period from 1971–2010 (Herrera et al. 2012). The grid was produced by applying the kriging method in a two-step process. First, the occurrence was interpolated using a binary kriging and, in a second step, the amounts were interpolated by applying ordinary kriging to the occurrence outcomes. The elevation is not explicitly included in the development of the dataset because the available dense gauge network represents the orography corresponding to the 0.11$^\circ$ grid appropriately. Explicit comparison of Spain02 with the E-OBS dataset shows better performance of Spain02 to represent extreme events of daily precipitation in the region of Valencia regarding the amount and spatial distribution of precipitation (Herrera et al. 2012).

UK: UKCPOBS (5 km)  The National Climate Information Centre daily UK gridded precipitation dataset (Perry et al. 2009) spans the period 1958–present day, and from 1990 uses approximately 2500–3500 surface gauge observations. Quality control is performed through computerized and manual comparisons of individual daily station values against the daily all-station average and daily values from nearby stations. Any stations that have failed quality control are excluded from the computation of the gridded values. The gridding of the gauge data to a 5 km$\times$5 km grid uses a cubic inverse-distance weighting interpolation using stations within 50 km radius of the grid box.

CMORPH 1.0 (0.25$^\circ$) The CMORPH (NOAA Climate Prediction Center morphing method, Joyce et al. 2004) algorithm uses the relatively high-resolution IR information to infer the hydrometeorological position between two consecutive PMW estimates. IR maps are used to derive cloud system advection vectors (CSAVs) to propagate PMW rainfall estimates. Such propagation is performed forward and backward for each time step using information provided by the CSAVs. Final values are achieved by averaging forward and backward rainfall analyses proportionally to step distance.

1.2 Hourly datasets

Nimrod (UK)  Gridded hourly radar data for the UK at 5 km resolution are available from the Nimrod database (Golding 1998) for the period 2003-present-day. There are many issues with radar (clutter, anaprop, bright band, beam attenuation), and in particular radar data are known to systematically underestimate heavy rainfall amounts. The Met Office calibrates radar against rain gauges and employs algorithms to take account of known issues but some problems cannot be fully rectified. One of these is that the hourly gauges used in the calibration are relatively sparse, and thus are not able to fully correct for locally-varying effects such as attenuation.

Germany  The hourly precipitation data set assembled by Paulat et al. (2008) is used. It features a horizontal grid spacing of 7 km and an effective horizontal resolution of 14–28 km. The time period of the dataset is 2001–2008 (8 years). To assemble this dataset, measurements from rain gauges have been gridded as daily sums, following the procedure by Frei and Schär (1998). Afterwards, the daily sums were disaggregated into hourly values using rain rate retrievals from radar (Wuest et al. 2010). Beyond uncertainties arising from rain-gauge undercatch, gridding procedures (Frei et al. 2003), and weather radar measurements (Wüest et al. 2010), possible inconsistencies between gauge observations and radar restricts the data set to 92% of the possible days, at the respective grid points (Paulat et al. 2008).

Switzerland  RdisaggH is an experimental precipitation data set for Switzerland which provides gridded, radar-disaggregated rain-gauge observations (Wüest et al. 2010). In order to obtain hourly data, a gridded daily product was disaggregated into hourly sums, using information from weather radar fields. The resulting dataset has a grid-spacing of 0.01$^\circ \times$ 0.01$^\circ$ covers Switzerland and is available for the time period May 2003–2010.