NUMERICAL STUDY OF A NONNUMERICAL STUDY OF A NON--FORECASTED SEA BREEZE THUNDERSTORM IN THE EASTERN FORECASTED SEA BREEZE THUNDERSTORM IN THE EASTERN

IBERIAN PENINSULA. PART I. HIRLAM AND HARMONIE PRECIPITATION PERIBERIAN PENINSULA. PART I. HIRLAM AND HARMONIE PRECIPITATION PERFORMANCEFORMANCE

Acknowledgements: The authors would like to thank the AEMET, CEAM (data provided by David Corell), CHE, CHJ, IVIA, METEOCLIMATIC, SIAM, SIAR and SMC for raingauge data; the AEMET for the radar and lightning pictures; and José-Antonio García-Moya and Estrella Gutiérrez-Marco (AEMET) for providing the HIRLAM 6.1.2 outputs. The Instituto Universitario CEAM-UMH is partly supported by the Generalitat Valenciana (GV), Fundación Bancaja and the projects GRACCIE (Consolider-Ingenio 2010) and FEEDBACKS (Prometeo - GV). This research was undertaken in the frame of the BEST/2010/014 (GV)

*Corresponding author: Cesar Azorin-Molina, Pyrenean Institute of Ecology (Spanish Research Council), Department of Geoenvironmental Processes and Global Change,

Avda. Montañana 1005, 50059-Zaragoza, Spain. E-mail: cazorin@ipe.csic.es

This study is a numerical investigation of an isolated sea breeze thunderstorm that occurred over a preferential sea breeze convergence zone, the Iberian system mountains (Spain), on 7 August 2008. This sea breeze front with intense convective activity (heavy rainfall, hail and gusty winds) was not forecasted by the operational HIRLAM 6.1.2 (16.0-km and 5.0-km) numerical weather prediction. Two grid-spacing setups (5.0-km and 2.5-km) of the operational HIRLAM 7.2.2 and the non-hydrostatic spectral HARMONIE suite (2.5-km) are used to simulate isolated convection associated with sea breezes. The overall aim is to improve local forecasting accuracy for dangerous sea breeze convective phenomena, and particularly to estimate the ability of NWP to correctly simulate convective precipitation associated with sea breezes. The precipitation forecast performance of both HIRLAM and HARMONIE model setups is evaluated against high-density >500 raingauge measurements. The operational HIRLAM 6.1.2 version from the Spanish Meteorological Agency (AEMET) at 0.16 and 0.05 degrees missed the isolated sea breeze showers, whereas both the HIRLAM and the HARMONIE setups showed precipitation signals in the area where the most intense precipitation (Cati 45 mm) was observed. Advances in short-term forecasts and increased understanding of isolated convection associated with sea breezes could have practical applications for dangerous convective phenomena.S

umm

ary

Cesar Azorin-Molina (1)*, Sander Tijm (2), Elizabeth E. Ebert (3), Sergio-M. Vicente-Serrano (1), Maria-Jose Estrela-Navarro (4)

(1) Pyrenean Institute of Ecology, CSIC (Spanish Research Council), Department of Geoenvironmental Processes and Global Change, Zaragoza, Spain http://www.ipe.csic.es/(2) Royal Netherlands Meteorological Institute (KNMI), De Bilt, The Netherlands, http://www.knmi.nl/

(3) Centre for Australian Weather and Climate Research (CAWCR), Bureau of Meteorology, Melbourne, Australia, http://www.cawcr.gov.au/(4) Laboratory of Meteorology-Climatology, Mixed Unity CEAM-UVEG, Department of Physical Geography, University of Valencia, Valencia, Spain http://www.ceam.es/

Mod

else

t up

Sea

bre

eze

fron

t: 7

Aug

ust 2

008

HIR

LAM

& H

AR

MO

NIE

per

form

ance

Con

clus

.11th EMS /10th ECAM. Berlin, Germany, 12-16 September 2011

* A more detailed description of the HIRLAM and HAR MONIE model suites can be found at these URL, respe ctively:

http://hirlam.org/index.php?option=com_content&view =article&id=64:general-description-of-the-hirlam-mo del&catid=48:synoptic-scale-model-hirlam&Itemid=101

http://hirlam.org/index.php?option=com_content&view =article&id=65&Itemid=102

TABLE I. Model description, set up and initialization. The simulations were initialized at 1200 UTC 6 August 2008, and integrated for 48 h until 1200 UTC 8 August 2008, with model outputs at intervals of 1 hour. The dynamics time step is 120 s for the HIR-D5, and 60 s for the HIR-D2.5 and the HAR-D2.5 experiments.

Session NWP1 – EMS2011-397

HIRLAM (HIR-D5) HIRLAM (HIR-D2.5) HARMONIE (HAR-D2.5)

MODEL VERSION HIRLAM 7.2.2 HIRLAM 7.2.2 HARMONIE 36h1.2

MODEL TYPE Three-dimensional hydrostatic grid-point The same as the HIR-D5 Non-hydrostatic spectral model

DYNAMICAL CORE Semi-implicit semi-Lagrangian discretization of the multi-level primitive equations . Hybrid coordinate in the vertical

The same as the HIR-D5Two-time level semi-implicit semi-Lagrangian discretizationof the fully elastic equations . Hybrid coordinate in the vertical

HORIZONTAL GRID-SPACING 0.05º , i.e., 5-km 0.025º , i.e., 2.5-km 2.5-km

INTEGRATION AREA Unique domain with 300x306 horizontal grid points (latitudeby longitude) and the size of 1495 km x 1525 km

Unique domain with 400x406 horizontal grid points (latitudeby longitude) and the size of 997.5 km x 1012.5 km

Unique domain with 400x400 horizontal grid points (latitude by longitude) and the size of 1000 km x 1000 km

VERTICAL DOMAIN60 hybrid levels with enhanced resolution (27 levels in thelowest 3 km) in the planetary boundary layer. The model topis at 10 hPa and the lowest model level is around 32 m.

The same as the HIR-D560 hybrid levels with enhanced resolution (27 levels in thelowest 3 km) in the PBL. The model top is at 1 hPa and thelowest model level is around 10 m.

INITIAL&BOUNDARY CONDIT. ECMWF operational boundaries archive data (0.5º x 0.5º) The same as the HIR-D5 The same as the HIR-D5

FIG. 1. (a) HIRLAM model configuration showing a largest (HIR-D.5; blue solid line) and smallest (HIR-D2.5; green dashed line) domain with horizontal grid lengths of 5.0 km and 2.5 km, respectively, and also HARMONIE model configuration with the smallest domain (HAR-D2.5; red solid line) and 2.5 km grid length. The map shows location (magenta dots) of the high-density rain-gauge network for verification model output. (b) Terrain map of the Iberian Peninsula (IP) showing the complex topography of the study area (magenta rectangle) and the steep Iberian System mountains (ISM). The black solid line corresponds to the vertical cross sections over 40.8ºN presented in the poster nº EMS2011-400.

FIG. 2. (a) Sea-level pressure (in hPa), and geopotential (in m) and air temperature (in C) at (b) 500 and (c) 850 hPa on 7 August 2008 at 1200 UTC from the NCEP/NCAR reanalysis data. The mean sea-level pressure (the contour interval is 2 hPa), 850 and 500-hPa geopotential heights (the contour interval is 40 hPa) are shown with solid lines. The air temperature (the contour interval is 2 C) is shown with dashed lines.

FIG. 3. Thermal infrared (IR 12.0, spectrum 10 at 11.0-13.0 µm) Meteosat Second Generation (MSG) satellite image over the Iberian Peninsula between 1300 and 1830 UTC on 7 August 2008. Note that the MSG satellite picture at 1530 UTC was not available and replaced by the previous MSG satellite data at 1515 UTC. The color scale is shown in the lower right corner of the composite. Source: EUMETSAT

FIG. 4. Doppler radar images acquired from the radar located at Cullera (Valencia, Spain) between 1300 and 1830 UTC on 7 August 2008. Dark circle indicates the observation range of the Doppler radar.

FIG. 5. Lightning strikes detected between 1200 and 2400 UTC on 7 August 2008.

* The episode chosen for this numerical study occurred over the Iberian system mountains on 7 August 2008 , in the eastern part of the Iberian Peninsula.

* The storm corresponded to a typical sea breeze front that developed explosively underweakly defined synoptic-mesoscale forcing (see poster nº EMS2011-400)

* The inflow of moisture driven by well-developed thermally local circulations at the lowerlevels mainly triggered this sea breeze storm.

* The storm was isolated & stationary as shown in the satellite, radar, and lightning imagery.

* Produced severe weather with heavy showers, hail and straight-line winds.

Sea breeze thunderstorm characteristics

(a) Weak surface pressure gradient in thewestern Mediterranean basin that allowed

thermally driven flows to develop.

(b) Prevailing moderate to strong westerlyand southwesterly flows at 500 hPa , with a

short wave trough of maritime polar airmass (around -12C)

(c) Strong warm ridge of tropical air mass at 850 hPa (at about 20C) over the western

Mediterranean basin

As a consequence of the strong upward air motion at the SBF ( see poster nº EMS2011-400), solid precipitation in the form of large hail of around 3.5 cm caused significant agriculture damage, particularly in some fruit groves and irrigated areas for vegetables, and also damaged vehicles.

* The long-lived sea breeze front lasted approximately 5 hours and consisted of groups of Cb clouds and precipitating cells .

* This Cb storm showed a minimum cloud top temperature of 212 K and strong Doppler radar echoes greater than 65 dBZ between 1530 UTC and 1630 UTC, delineating heavy precipitating cell and severe turbulence along the sea breeze front.

A total of 256 strikes(16 positives; 240 negatives) were

detected between1200 and 2400 UTC

on 7 August 2008

FIG. 6. Large hail up to 3.5 cm in diameter with concentric rings that fell in Catí (Castellón, Spain) at 1726 UTC on 7 August 2008. Pictures taken by: Alex Carbo-Meseguer.

Wea

ther

fore

cast

THE OPERATIONAL HIRLAM 6.1.2

VERSION FAILED IN PREDICTING THIS

SEA BREEZE THUNDERSTORM.

THE OFFICIAL WEATHER

FORECAST ISSUED FOR THE VALENCIA

REGION WAS:“ Mostly sunny with few

Cumulus clouds over the high Iberian mountains.

Light westerly winds during the morning, with afternoon

sea breezes. Maximum temperatures weakly

increasing on the coast”.

SYNOPTIC STATION 24-h PREC. (in mm)

Cati- Meteoclimatic 45.0

Vallivana-CEAM 27.8

Morella-CHE 15.8

Cati-CHJ 11.2

Vallibona-CHJ 11.2

El Boixar 10.0

Xert-CHJ 8.2

Michavilla-CHE 8.2

Pobla Benifasá 6.8

Morella-Herbés-CEAM 6.0

Sant Rafael del Riu 6.0

AR

C_p

rec

_20

080

807

FIG. 7. 24-h accumulated rainfall between 06 UTC 7 August 2008 till 06 UTC 8 August 2008 simulated from (a) the HIR-D16 (0.16º , i.e., 16-km length) and (b) the HIR-D5 (0.05º , i.e., 5-km length) operational HIRLAM 6.1.2 version from the AEMET.

(a) (b)

FIG. 8. Accumulated rainfall recorded between 08 UTC 7 August 2008 till 08 UTC 8 August 2008. The number of rain-gauges used is shown in the lower left corner of the image (n= sample size). The map was produced using a Kriging interpolation.

n=533

TABLE II. Total precipitation measured between 08 UTC 7 August 2008 till 08 UTC 8 August 2008 for those stations with >5mm.

FIG. 9. Accumulated rainfall between 06 UTC 7 August 2008 till 00 UTC 8 August 2008 simulated from (a) the HIR-D5, (b) the HIR-D2.5, (c) the HAR-D2.5, and (d) the HAR-D2.5 with an increase in fall speed of precipitation by a factor of 5. Forecasted maps are plotted using the same boundaries and colors for comparison against the observed rainfall map shown in Figure 8 .

(a) (b) (c) (d)

73.4 mm

FIG. 10. Time series of observed and modelled precipitation in Vallivana-CEAM, Morella-CHE and Michavilla-CHE from 0700 till 2300 UTC on 7 August 2008. PRECNNB corresponds to precipitation at grid in the model closest to raingauge, and MPREC5 to the maximum modelled precipitation in area with radius of 5-km around raingauge station. Note that modelled precipitation for HIR-D2.5 exceeds the maximum y-axis for the MPREC5 – Michavilla-CHE graph.

PRECNNB PRECNNB PRECNNBMPREC5 MPREC5 MPREC5

(a)

1300 UTC 1330 UTC 1400 UTC

1430 UTC 1500 UTC 1530 UTC

1600 UTC 1630 UTC 1700 UTC

1730 UTC 1800 UTC 1830 UTC

(1) The operational HIRLAM 6.1.2 version (STRACO convection/condensation scheme) performed worse for this isolated sea breeze thunderstorm than more recent versions with updated STRACO or the Kain Fritsch/Rasch-Kristjansson (KF) convection/condensation scheme. It is widelyrecognized that inadequate representation of convective cloud processes is the greatest source of uncertainty in the NWP.

(2) Both HIRLAM and HARMONIE simulations succeded capturing the occurrence of convective showers over the Iberian System mountains. However, the HAR-D2.5 run forecasted the precipitation at the right place with a reasonable amount, the HIR-D5 run predicted a weak precipitationsignal (underestimation) and displaced too far the northwest, and the HIR-D2.5 run simulated greater amount of rainfall (overestimation) over a larger area placed far to the north. A large part of the overestimation of HIR-D2.5 in comparison to HAR-D2.5 is associated with the fact that the fallspeed is infinity in the HIRLAM model suite, i.e., precipitation is put on the ground directly, which is quite normal in hydrostatic models. The impact of this is mimicked through increasing the fall speed of precipitation in the HAR-D2.5 run (Fig 9d), showing a strong increase in precipitation.

(3) Times series revealed that HAR-D2.5 performed better the time/amount occurrence of this isolated sea breeze storm than HIR-D5 and HIR-D2.5, particularly taking the maximum modelled precipitation in a radius of 5-km around the raingauge station.

(4) The role of strong low-level sea breeze convergence in initiation of this non-forecasted sea breeze thunderstorm under weak synoptic-mesoscale forcing in the eastern region of the Iberian Peninsula is investigated in the poster nº EMS2011-400.

(b)