92nd American Meteorological Society Annual Meeting (January 22-26, 2012)

Wednesday, 25 January 2012
Validation of the Uniparthenope ARW-WRF Model for a Convective Case Study
Hall E (New Orleans Convention Center )
Angelo Zinzi, Università di Napoli , Naples, Italy; and R. Montella, G. Agrillo, A. Riccio, and G. Budillon

1.      Introduction

The ARW WRF model (Skamarok et al., 2008) is a widely used mesoscale model and has been chosen for the PROMETEO project at the University “Parthenope” of Naples (Italy) to provide detailed weather forecasting over the Campania region.

In this work 11th June 2011 has been identified as a case study, since a major convective event originated an extensive cloud coverage and large precipitations not forecasted by the operative version of the model used at that time.

2.      Model configurations

In particular, on the basis of works already published (e.g., Moscatello et al., 2008; Miglietta et al., 2010; Lara-Fanego et al., 2011), the configuration was the first described in Table 1.



Configuration 1

Configuration 2

Configuration 3


New Thomson



Longwave radiation


Shortwave radiation


Surface layer

MM5 similarity


5 layer





Table 1: The different parameterizations of the models here tested.

The model has been set-up with three two-way nested domains: the largest (named d01) comprising all the Europe region at a horizontal resolution of 27 km, the second (d02) centered on Italy with a resolution of 9 km and the last (d03), over Southern Italy, with a resolution of 3 km.

Has shown in Table 1, for this study only the microphysics scheme has been modified. Furthermore also the initialization date was changed, being set on 10th June Z00 and 11th June Z00. This resulted in six different model configurations, that have been tested by means of two different techniques.

The first one is the point-stat analysis provided by the MET (Model Evaluation Tool) with observation data selected by the CISL Research Data Archive ds337.0. In particular temperature at 2 meters and relative humidity at different atmospheric layers (i.e., 2 m, 850 and 500 hPa) have been used for comparison. These data have been compared with the simulations of the d02 domain at Z06, Z12 and Z18 of 11th June 2011.

The other analysis has been focused on the cloud coverage, using Eumetsat data. In particular, the cloud mask dataset has been compared with a similar variable computed from the model results in the domain d03.

3.      Data analysis

By means of the point-stat analysis no sensible difference has been found between the different model configurations, with a percentage of hits (i.e., points with values larger than a certain threshold value) always ranging from 50 to 70%. However these values can be indicative of a good forecasting accuracy for the selected fields.

On the contrary large differences arose from the analysis of the cloud coverage, whose average observed value from Z06 to Z18 is around 90%. In the first two configurations (whatever the initialization date) the simulations have average cloud coverage not larger than 25%, whereas the third microphysics allowed to reach a 70% cloud coverage mean value, much more similar to the observed one. In addition the percentage of hits is between 65 and 70% with this latter and lower than 35% in the other cases.

It is also worthy to note that the two model in Configuration 3 do not produce similar results. In particular they show different hourly trends, with that initialized on 10th June seeming to display a delay of 3 hours respect to the observations, even if with a cloud coverage value always 20% lower than the observed one. The same cannot be said for the other version of the model, that only displays a slight similarity if the delay reaches 6 hours.

By looking at Figure 1, it is clear that, taking into account the 3 hours delay described above, the model initialized on 10th June is able to reproduce with good accuracy the cloud coverage over a large fraction of the analyzed area, with major differences between model and observations only present over the sea.


Figure 1: Comparison between observations (left, Z12) and Model C initialized on 10th June 2011 (right, Z15), taking into account the 3 hours delay described in the text.

4.      Conclusions and future works

The work here proposed seems to demonstrate that, in the case of severe convective phenomena, the Millbrandt-Yau microphysics, initialized the day before the event, is strongly suggested for the ARW WRF model over Southern Italy.

In the future further improvements and tests of the model have to be performed. In particular a change of the horizontal resolution of the d03 domain that, with a 1 km resolution, should be able to run without cumulus parameterization.

References: Lara-Fanego et al., 2011, Evaluation of the WRF model solar irradiance forecasts in Andalusia (Southern Spain), Solar Energy, doi: 10.1016/j.solener.2011.02.014; Moscatello et al., 2008, Numerical Analysis of a Mediterranean “Hurricane” over Southeastern Italy, Monthly Weather Review, doi: 10.1175/2008MWR2512.1; Miglietta et al., 2010,             WRF model and ASAR-retrieved 10 m wind field comparison in a case study over Eastern Mediterranean Sea, Adv. Space Sci., doi: 10.5194/asr-4-83-2010; Skamarok et al., 2008, A description of the Advanced Research WRF Version 3, NCAR/TN-475+STR, Mesoscale and Microscale Meteorology Division, National Centre for Atmospheric Research, Boulder, USA.

Supplementary URL: