Next Article in Journal
Investigation of Marine Wind Veer Characteristics Using Wind Lidar Measurements
Next Article in Special Issue
Concurrent Influence of Different Natural Sources on the Particulate Matter in the Central Mediterranean Region during a Wildfire Season
Previous Article in Journal
Trends and Variabilities of Thunderstorm Days over Bangladesh on the ENSO and IOD Timescales
Previous Article in Special Issue
Multi-Weather Evaluation of Nowcasting Methods Including a New Empirical Blending Scheme
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Atmosphere
Volume 11
Issue 11
10.3390/atmos11111177
atmosphere-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Evaluation of Two Cloud-Resolving Models Using Bin or Bulk Microphysics Representation for the HyMeX-IOP7a Heavy Precipitation Event

by
Diana Arteaga
1,
Céline Planche
1,*,
Christina Kagkara
1,
Wolfram Wobrock
1,
Sandra Banson
1,
Frédéric Tridon
2 and
Andrea Flossmann
1
1
Université Clermont Auvergne, Laboratoire de Météorologie Physique, LaMP/CNRS UMR 6016, F-63000 Clermont-Ferrand, France
2
Institute for Geophysics and Meteorology, University of Cologne, D-50923 Cologne, Germany
*
Author to whom correspondence should be addressed.
Atmosphere 2020, 11(11), 1177; https://doi.org/10.3390/atmos11111177
Submission received: 12 September 2020 / Revised: 26 October 2020 / Accepted: 27 October 2020 / Published: 31 October 2020
(This article belongs to the Special Issue Advances in Mesoscale Numerical Weather Prediction and its Applications)
Browse Figures
Versions Notes

Abstract

:
The Mediterranean region is frequently affected in autumn by heavy precipitation that causes flash-floods or landslides leading to important material damage and casualties. Within the framework of the international HyMeX program (HYdrological cycle in Mediterranean EXperiment), this study aims to evaluate the capabilities of two models, WRF (Weather Research and Forecasting) and DESCAM (DEtailed SCAvenging Model), which use two different representations of the microphysics to reproduce the observed atmospheric properties (thermodynamics, wind fields, radar reflectivities and precipitation features) of the HyMeX-IOP7a intense precipitating event (26 September 2012). The DESCAM model, which uses a bin resolved representation of the microphysics, shows results comparable to the observations for the precipitation field at the surface. On the contrary, the simulations made with the WRF model using a bulk representation of the microphysics (either the Thompson scheme or the Morrison scheme), commonly employed in NWP models, reproduce neither the intensity nor the distribution of the observed precipitation—the rain amount is overestimated and the most intense cell is shifted to the East. The different simulation results show that the divergence in the surface precipitation features seems to be due to different mechanisms involved in the onset of the precipitating system: the convective system is triggered by the topography of the Cévennes mountains (i.e., south-eastern part of the Massif Central) in DESCAM and by a low-level flux convergence in WRF. A sensitivity study indicates that the microphysics properties have impacted the thermodynamics and dynamics fields inducing the low-level wind convergence simulated with WRF for this HyMeX event.

1. Introduction

During autumn, the South of France is frequently affected by heavy precipitation that produce flash-floods and landslides [1,2] costing each year several billions euros of damages and fatalities. Mesoscale Convective Systems (MCSs) that stay over the same area during hours are the main reason for high rainfall totals [3,4]. These convective precipitating systems are associated with synoptic situations characterised by upper-level low pressure northwest of the precipitation area, generating low-level marine inflow charged with water vapour and directed towards the coastal mountainous regions [5,6]. Then, the onset of deep convection could be triggered, for example, by an orographic lifting of the moist air [7] among others or the presence of a low-level cold pool induced by the convective system itself [3].
To improve the forecast of these MCSs in Numerical Weather Prediction (NWP) models, the 10-year international HyMeX (Hydrological Cycle Experiment in the Mediterranean) program [8] organised a major field campaign over the north-western Mediterranean region from September to November 2012 [9]. During a special observation period (SOP1), a large set of instruments was deployed to characterize the MCSs. Several convective activities observed during SOP1 provide a good opportunity to evaluate the capability of mesoscale models to reproduce the observed MCSs. Indeed, using these data, an objective of HyMeX and the associated MUSIC project (MUltiscale process Studies of Intense Convection precipitation events in Mediterranean) was to compare model simulations with observations for the purpose of improving NWP model representation of convection and intense precipitation events.
The representation of the microphysics processes has been identified as one of the major components that strongly influence the forecasts of the MCSs at the kilometric scale [10]. Indeed, it is considered as a leading order contributor to NWP models precipitation forecast errors.
Heavy precipitation is typically connected with cloud fields that extent at least to the mid troposphere or higher altitudes. At these altitudes, the temperature ranges approximately from −5 to −35 °C and cloud and precipitation formation takes place in mixed phase conditions (i.e., presence of supercooled liquid water and ice). However, the microphysical properties of these clouds are still not completely understood. Indeed, in this temperature range, the formation of ice crystals occurs due to heterogeneous nucleation [11] and the parameterisations representing this microphysical process differ significantly from each other, leading to different results when applied in cloud models [12] among others. The properties of the ice hydrometeors depend on temperature and supersaturation. Their growth by vapour deposition as well as their collection efficiencies for riming and aggregation varies significantly with the crystal habit [13]. Thus, the ice microphysics and its representation in models have a significant impact on quantitative precipitation forecasts [14,15] and on cloud-dynamical interactions [16].
Microphysics representations are classified broadly into two types: (i) the bin schemes that predict the particle size distribution (PSD) by discretising it explicitly into multiple size (or mass) bins as in for example, References [17,18], and (ii) the bulk schemes that predict one or more bulk quantities assuming some prescribed form for the PSD [19,20,21] among others, often as a superposition of several three-parameter gamma distribution functions. The size distribution evolves continuously in microphysics schemes and the computational cost associated to bin schemes is high and thus currently not suitable for NWP models. The one-moment bulk approach, in which only one moment of the hydrometeor size distribution function is predicted, was first developed by Kessler for warm clouds [22]. A more detailed approach was subsequently developed predicting not only the mixing ratio (i.e., third moment of the hydrometeor size distribution function) but also the number concentration (i.e., zero-th moment of the size distribution function) of cloud droplets and rain [23,24]. Then, the bulk approach was extended to the ice phase using a similar separation between cloud ice and precipitating ice and adding different predefined categories of precipitation (e.g., snow, aggregates, graupel and hail) [19,20]. Recently, a novel approach has been introduced [25,26] replacing the different ice categories that are difficult to define with a representation of a gradual transition from small to large ice particles due to growth by water vapour deposition and aggregation, and from unrimed crystals to graupel due to riming. This latter representation of the microphysics is not yet used in the NWP models where one-moment or two-moment bulk schemes with predetermined ice categories are considered for example, References [27,28].
This study compared the performance of two regional models, WRF (Weather Research and Forecasting) and DESCAM (DEtailed SCAvenging Model), for one of the most intense convective event observed during the HyMeX SOP1 campaign, that is, the IOP7a convective system observed on the 26 September in 2012. The main objective of this study is to compare how WRF and DESCAM (where the microphysics is represented using, respectively, a bulk or a bin approach) reproduce this MCS over the South of France and highlight the mechanisms that are important in its formation and development. The paper is organised as follows—Section 2 provides a description of the case studied and its setup and configuration in the considered mesoscale models. Section 3, Section 4, Section 5 and Section 6 compare the observations and mesoscale simulation results obtained with DESCAM using the bin microphysics scheme see Reference [29] for additional results and with WRF using two different bulk microphysics schemes [19,20] focusing on precipitation features, thermodynamic profiles, microphysics and dynamics properties. Additional sensitivity studies performed to understand the differences in the convective system simulated with both models are presented in Section 7. Finally, the main findings are discussed in Section 8.

2. Methodology

2.1. The HyMeX-IOP7a (26 September 2012, Southern France)

Although several MCSs were observed during the HyMeX campaign, in this study we only focus on the IOP7a intense convective event developed over the Cévennes-Vivarais region (South of France) during the morning of the 26 September 2012. The synoptic conditions were favorable for the development of a convective event since an upper-level low pressure system was centered over the British Isles propagating eastwards to France. In association to a cold front located over Spain, a strong southwesterly flow passing over the Mediterranean was formed carrying moist, warm and unstable air masses towards the French mountainous coasts (i.e., the reliefs of the Cévennes-Vivarais) more details in Reference [30]. The convective system was observed over the mountain ridge during the morning with the most intense rainfall around 07:30–08:30 UTC. Figure 1 shows the composite 12-h cumulative rainfall retrieved from the S- and C-band radars of the French operational radar network ARAMIS (Application Radar à la Météorologie InfraSynoptique) [31]. The cumulative rainfall attains a maximum of approximately 100 mm.
During the HyMeX SOP1, the LaMP (Laboratoire de Météorologie Physique) X-band radar was deployed for a few months within the Cévennes-Vivarais Mountains (Figure 1b). It is designed to provide the precipitation field with high spatial (60 m) and temporal (30 s) resolution over a typical domain of a small catchment basin (about 36 km range) using a beam elevation of 1.5° [29,32]. Figure 2a,b show the radar reflectivity field observed by the X-band radar at different times that characterize the most intense period of the convective system over the radar observational area. The cloud system forms close to mountain ridge (dashed line on Figure 2a) and moves towards the north-north-east of the radar domain. The precipitating system is quite intense and the reflectivity field attains a maximum of 60 dBZ for many hours.

2.2. The DESCAM and WRF Models

The numerical simulations were performed with two different mesoscale models—the Clark model [33,34,35] coupled to the DESCAM bin microphysics scheme [18,32,36] and the WRF model (version 3.6.1) with two different bulk microphysics schemes [37].
The Clark model is a non-hydrostatic and anelastic cloud scale model. This model is an established tool for the simulation of airflow and the formation of clouds over complex terrain on small meteorological scales [38,39] and its potential for high horizontal and vertical resolution can provide important insights in the cloud evolution process [32,40]. The environmental variables are written in perturbation form, as detailed in Clark et al. [41].
The WRF v3.6.1 model is a non-hydrostatic atmospheric model. For this model, two dynamical solvers are available, but in this study, only the ARW solver is used (originally referred to as the Eulerian mass solver) [37]. The model serves a wide range of meteorological applications across scales ranging from meters to thousands of kilometers. Table 1 summarizes the pertinent physical characteristics of both models used in this study.
Moreover, in DESCAM, radiative cooling and heating rates in the clouds were not considered. Incoming radiative fluxes do only interact with the earth surface and thus determine the up-down welling fluxes of sensible and latent heat. The variability of the surface fluxes depends on the altitude, the direction of the terrain slope and the sun’s inclination angle see details in Reference [44]. In WRF, the longwave and shortwave radiations are, respectively, described by the Rapid Radiative Transfer Model (RRTM) [46], and the Dudhia schemes. Also, WRF considers the Yonsei University boundary layer scheme (YSU) [47]. As the horizontal resolution is greater than 3 km in the outermost domain, the Kain-Fritsch cumulus parameterization scheme [48] is applied in this domain in order to well represent the sub-grid-scale effects of convective and/or shallow clouds.
For this study, the used microphysics schemes in WRF are either the Morrison scheme [20] or the Thompson scheme [19] that are respectively a two-moment or a partially two-moment bulk microphysics scheme (more details about differences in the representation of the rain processes are available in References [49,50]). In contrast, the DESCAM microphysics scheme is a spectral (bin) scheme which simulates the number distribution functions for the aerosol particles and for drops. Both number distributions use a logarithmically equidistant spaced mass coordinate for wetted aerosol particles and drops, both with mass doubling for two subsequent bins. The model also considers the mass density distribution of aerosol material in cloud and rain drops. All functions are discretised in 39 bins covering the range from 1 nm to 7 μm for wet aerosol particles and 1 μm to 12 mm for drops. The ice phase is represented in a similar way with two additional distribution functions characterising the number of ice particles and the aerosol mass inside ice particles. As described in details in References [18,32], the DESCAM scheme considers the following processes from both liquid and ice cloud phases: the growth of drops by condensation and the ice growth by vapour deposition according to Pruppacher and Klett [13], the collision-coalescence and riming processes follow the scheme of Bott [51] considering the collection efficiencies of Hall [52] and the terminal velocities of Pruppacher and Klett [13], the ice melting, and the homogeneous and heterogeneous ice nucleation follow respectively the parameterisation of Koop et al. [53] and Meyers et al. [54]. The model also considers the aerosol particle activation and growth and the drop de-activation following the Köhler theory [13]. Note that this microphysics scheme allows the study of the impact of the aerosol particle loading on the intense rainfall of the HyMeX IOP7a event as in Reference [29].

2.3. Simulation Strategy

In order to evaluate the quantitative forecast of the precipitation for the IOP7a convective cloud system, the different high-resolution simulation results are compared to the available HyMeX observations. For all experiments, the configuration of the numerical domains is the same as in the work of Kagkara et al. [29]: it consists of three nested domains with an increasing horizontal resolution. The innermost domain is represented on Figure 1b. The different models were initialised with the IFS (Integrated Forecasting System) operational data (produced thanks to the Cy40r1 spectral general circulation model with parameterised processes) from the European Center for Medium-range Weather Forecasts (ECMWF) for the 26 September 2012 at 00:00 UTC and forced every 12 h. In DESCAM, the aerosol size distribution and chemical properties are initiated similarly than in the HymRef case of Reference [29] (properties derived from the French ATR-42 aircraft aerosols measurements). The vertical grid z is following terrain and non-equidistant with a different model top according to the nested domains in all experiments. However, the different model architectures do not allow us to use exactly the same gridding. Indeed, in WRF, the vertical spacing Δz next to the surface is about 60 m and increases to 305 m at 9 km for the innermost domain, whereas in DESCAM, Δz varies respectively from 40 m to 230 m. Table 2 details the numerical settings used for the different experiments.
Sensitivity studies are performed with WRF in order to better understand the differences between the observations and the simulation results obtained with DESCAM, WRF-THOM and WRF-MORR. All the physical characteristics (see Table 1) and the model domain configuration (see Table 2) are kept for these WRF tests except for the microphysics scheme and the initiation data. Indeed, in order to know if different initial WRF settings could provide simulation results more comparable to the observations, homologous experiments to WRF-THOM and WRF-MORR are performed initiating the model with the Era-Interim operational reanalyses (available every 6 h) from the ECMWF [55] instead of the IFS data (see Section 4). Moreover, to study the impact of the microphysics on the simulated convective system, the WRF model is used without any microphysics scheme (experiment called WRF-NOMIC in Table 2) (see Section 7).

3. Precipitation Features

Several radars deployed during the campaign permit to characterize the precipitation distribution of the IOP7a event down to sub-kilometric scales. Additional instruments, such as rain gauges, disdrometers, and micro rain radars provide observations locally and at raindrop scale. Hereafter, only bulk comparisons over the innermost domain are described but comparisons between DESCAM and observations at raindrop scale are available in Reference [29]. For these comparisons, the respective limitations and uncertainties of the observations have to be considered, for example, calibration and attenuation, perturbations of the signal, instrumental effects or locations of the stations.

3.1. Accumulated Rain

Figure 3 shows the 12-h cumulative rain retrieved from the ARAMIS radars network and simulated with DESCAM and WRF for the innermost domain. The rain rate R (in mm h−1) is derived from the observed radar reflectivity Z using a classical Marshall-Palmer Z − R relationship: Z[mm6m−3] = 200R1.6 [56]. The rain accumulation (in mm) is then computed from R and represented in Figure 3a. The observed accumulated rain attains a maximum of 105.3 mm (see Table 3) at 12:00 UTC and is distributed along an intense band that follows the relief of the Cévennes Mountains, that is, from the south-west to the north-east of the domain. In Reference [29], using the KED technique (geo-statistical technique merging radar observations and rain gauge measurements [57]), the distribution of the precipitation at the ground is similar to the ARAMIS retrieval but reaches a local maximum of 116 mm.
The accumulated rain amount simulated in DESCAM for a 12-h period is given in Figure 3b. We can see an intense band of precipitation located over the relief and oriented as in the ARAMIS observations. In addition to its location, the intensity of the maximum of rain is well reproduced in DESCAM since it attains a value of 119.5 mm. The area of intense precipitation at the surface observed to the east of Langogne is quite well reproduced by DESCAM. However, the area with a rain amount greater than 5 mm and lower than 30 mm is less spread in the simulations than in the ARAMIS observations. In both WRF experiments, the cumulative rain amount is also organised according to a band that is oriented in a direction similar to the observations (or DESCAM). However, this band of rain is shifted by approximately 20 km to the East. After 12-h of integration, the maximum of the accumulated rain amount for the main cell is largely overestimated in WRF-THOM and in WRF-MORR (Figure 3c,d).

3.2. Temporal Evolution of the Cumulative Rain

The temporal evolution of the cumulative rain (Figure 4) shows that the IOP7a convective system was particularly intense since the domain-averaged of rain increases by a factor of 10 from 06:00 to 08:00 UTC.
In DESCAM, the development of the system is delayed by approximately 1 hour (see also Reference [29]). So, in order to help the comparison of DESCAM results with the observations or the WRF experiments, the corresponding line is 1-h shifted in Figure 4. The intense period of rain is rather well reproduced in DESCAM but its duration was shorter than observed (only 1h30). In WRF-THOM and WRF-MORR, the temporal evolution of the mean cumulative rain amount is different. At 11:00 UTC the mean rain amount simulated in both experiments are close to the observations, but the model is not able to reproduce the flash rain period. Also, the precipitation starts earlier in the WRF simulations than in DESCAM or in the observations. After 12-h of integration with DESCAM, both the total and mean rain amount are underestimated whereas the rain area (considering points where the cumulative rain is ≥0.25 mm) is comparable to the observations (see Table 3; in Figure 3b lot of points correspond to a cumulative rain comprised between 0.25 mm and 5 mm see Reference [29]). Table 3 also confirms that the total rain is overestimated by approximately 40% in both WRF simulations. On the opposite, the mean rain amount is underestimated (by ≈22%) in WRF-THOM and in WRF-MORR because the irrigated area is much larger (+32%) than in the observations.

3.3. X-Band Radar Reflectivity Fields

In order to evaluate the DESCAM performances for simulating rain at sub-kilometric scales, the computed radar reflectivities are compared to the high resolution X-band radar observations (see location in Figure 1 and Figure 3). Panels c and d of Figure 2 show the modelled radar reflectivity field obtained with DESCAM at different time steps corresponding to the same stages of development of the precipitating system shown in the observations (panels a and b), that is, considering the one hour delay. An additional qualitative comparison at 07:50 UTC is available in Reference [29]. To facilitate the comparison between model and radar observations, the simulation results are presented for the same Plan Position Indicator (PPI) as the X-band radar observations (i.e., using the same 1.5° beam elevation angle). Radar reflectivity is calculated from the simulation via the sixth moment of the hydrometeor size distribution and, herein, the normalised radar reflectivity Z d B Z (in dBZ) is compared to observations as in Reference [32]. As in the X-band radar observations, the precipitating cells are also organised along a band. The locations and intensities of the modelled precipitating cells are comparable to those observed (Figure 2). Furthermore, the modelled band of rain is slightly shifted to the east with a slight underestimation of its intensity at 07:20 UTC. Also, the different cells cover a larger horizontal extension than the observed one. This may be explained by the model resolution of 500 m which is coarse compared to the 60 m radial resolution of the X-band radar. Likewise, this can explain the weaker model maximum intensities with respect to those of the X-band radar.
A more quantitative comparative analysis is performed in Figure 5. The probability density functions (PDF) of the observed and simulated radar reflectivities for the entire lifetime of the cloud system over the X-band observational area from 06:40 to 10:40 UTC are compared. We chose a resolution of 1 dBZ for the PDF and restricted the study to the range 10 to 60 dBZ in order to exclude noisy data (below 10 dBZ) of the X-band radar.
Comparison of the PDFs confirm the reasonable agreement between the X-band radar measurements and the DESCAM model for the entire lifetime of the convective system. The differences in the 20–30 dBZ range may be due to an excessive correction of the radar attenuation. Indeed, attenuation is not negligible at X-band and it has been taken into account in the radar data by using the Hitschfeld and Bordan algorithm [58] which is well known for being unstable in case of large attenuation [59]. It results in extended areas of overestimated reflectivity as visible in Figure 2a,b (turquoise areas beyond most of the intense small cells).

4. Thermodynamics Properties

4.1. Temperature and Humidity Profiles

Figure 6 shows the profile of the thermodynamics conditions obtained by the radio-sounding launched at Nimes on the 26 September 2012 at 12:00 UTC (see location on Figure 1b), that is, at the end of the rain period. Note that the sounding at 06:00 UTC was not available. This sounding reveals that the conditions of the atmosphere, that is, unstable and slightly humid, are favourable for the formation of the convective cloud up to the temperature inversion at approximately 500 hPa. However, as the Nimes station is more to the south than the innermost domain area, and is at low altitude whereas the simulation domain covers high mountainous region, no direct comparison between the Nimes sounding and the domain-averages obtained for the different experiments is done. Unfortunately, no other observations from lidar or satellite are available for this specific location and period.
Figure 6 also shows the domain-averaged mean profiles of the temperature and the dew point temperature obtained with DESCAM and WRF using either the Thompson scheme (WRF-THOM) or the Morrison scheme (WRF-MORR) for the innermost domain between 07:00 and 12:00 UTC. The mean temperature profiles simulated with the three different experiments are close. DESCAM is slightly warmer (1–2 °C) than WRF between the 850–500 hPa levels. On the contrary, between 950 and 850 hPa, the temperature in DESCAM is slightly lower than in both WRF experiments. Above ≈800 hPa, the mean dew point temperature in DESCAM is higher than in both WRF experiments. Thus, DESCAM is wetter than WRF at upper-level. On the contrary, DESCAM shows a much drier low-level than WRF. The humidity and temperature differences in the profiles are determinant for convection evolution. Indeed, in the dryer environment of DESCAM, convection can be impeded by a more stable stratification and could contribute to the lower accumulated rain amount compared to WRF-MORR and WRF-THOM (Figure 3).

4.2. Impact of the Initial Thermodynamics Fields

In order to know whether different initial forcing data could provide a cumulative rain field with WRF more comparable to the observations, a test was performed initiating the model with the Era-Interim operational reanalyses (see details in Section 2) instead of the IFS data. This numerical configuration does not improve the results previously obtained for the precipitation production with IFS. The band of precipitation is again shifted 20 km to the East compared to the observations (not shown) and the overestimation of the rain amount is even amplified. Indeed, Table 3 shows that the mean rain amount is modified by +5% (or +14.7%) using the ERA-Interim forcing data in WRF-THOM (or in WRF-MORR) whereas the maximum of the accumulated rain is, respectively, decreased by −23.5% and −7%. Moreover, the total rain mass accumulated at the surface during the 12 h of integration is 242.1 Megatons (Mt) in WRF-THOM and 263.0 Mt in WRF-MORR, that is, an increase of +9.5% in WRF-THOM and +14.7% in WRF-MORR. The increase of the rain production in both WRF simulations forced with the ERA-Interim data is probably due to the fact that the humidity is higher in ERA-Interim, especially at lower levels of the atmosphere.

5. Microphysics Features

The mean vertical profiles of the mixing ratio of cloud water, rain and ice (i.e., suspended and precipitating species) are represented in the Figure 7. These profiles are obtained by averaging the corresponding fields over the innermost domain between 07:00 and 12:00 UTC in the different simulations. Regarding the rain species, the vertical distribution (Figure 7b) is quite similar for both WRF experiments and in DESCAM, except above 4 km where the mean mixing ratio is larger in DESCAM than in WRF but remains small. Moreover, the profiles of the cloud and the ice species show that the vertical distribution of the liquid- and the ice-water phases is different in DESCAM and in both WRF experiments. Indeed, in WRF, the cloud is mostly composed of ice, especially between 3 and 6 km where the ice mixing ratio is 4 to 6 times greater than the cloud mixing ratio. In contrast, in DESCAM, over the same range of altitudes, the ice mixing ratio is close to the cloud mixing ratio.
The large amount of ice in WRF-MORR (Figure 7c) could explain that the total rain amount simulated at the surface (see Table 3) is the highest in this experiment. As suggested, for example, by References [13,15,60], when the ice phase is involved the formation of the precipitation is more efficient within convective systems. We can also note that the convective cloud system is thicker at upper levels in DESCAM than in WRF (Figure 7a) (i.e., the cloud mixing ratio ( q c ) at high altitudes is continuously significantly larger in DESCAM than in WRF). This is probably due to the fact that the atmosphere is wetter in DESCAM than in WRF at levels above 500 hPa (see Figure 6). From these comparisons, it appears that the convective systems simulated in both WRF experiments and in DESCAM are quite different, from the cloud initiation to the evolution of the cloud and precipitation properties. However, for this case, we do not have observations of the amount and the distribution of the different water phases inside the cloud system.

6. Wind Fields

Figure 8 shows the horizontal and the vertical wind components simulated over the innermost domain in DESCAM, in WRF-THOM and in WRF-MORR at 0.5 km and at 2 km above sea level (asl), that is, respectively below the cloud system and where the cloud mixing ratio is ≥0.05 g kg−1 in all simulations (see Figure 7a). The two time steps represent the wind fields 1 hour after the beginning and at the end of the intense precipitation period. At 07:00 UTC, a south-easterly low-level wind reaches the eastern slopes of the Cévennes Mountains in DESCAM whereas in WRF-THOM and in WRF-MORR a wind convergence line (yellow solid lines in Figure 8) appears because of a weak south-westerly flow present on the lee side of the Mountains. At 09:00 UTC, the cloud system is well developed and Figure 8 shows the wind components at 2 km asl (i.e., close to the cloud base). A south-westerly wind flow is obtained in the different simulations with a quite homogeneous intensity over the entire domain. Nevertheless, in DESCAM, the wind comes more from the southern directions and intensifies over the relief of the Cévennes Mountains. Also, an organised band of updrafts is visible in all simulations but at a different location: close to the crest of the mountain in DESCAM and around 20 km to the east of the mountainous area in both WRF simulations. This band is located close to the maximum of the rain amount of the different simulations (compare Figure 3 with yellow dashed lines in Figure 8). This suggests that the initiation of the convection is induced by the low-level wind convergence in WRF-THOM and in WRF-MORR while it is induced by the relief itself in DESCAM.
Note that the horizontal wind simulated with WRF and DESCAM close to the surface and at 2 km height (Figure 8) are similar to the UHF radar observations (not shown) made from the Candillargues site (see position on Figure 1). However, this site is far from the area of the innermost domain and this comparison cannot be used to confirm neither the presence of the simulated convergence in WRF nor the impact of the relief found in DESCAM.
To better understand the build-up of the low-level wind convergence in WRF, Figure 9 represents the hourly evolution of the wind components in WRF-MORR at 0.5 km. Figure 9 shows that the south-easterly low-level wind reaching the east slopes of the Cévennes Mountains at 01:00 UTC (Figure 9a) gradually weakens and turns towards a south-westerly wind at 06:00 UTC (Figure 9f) over time. In addition, the two outermost domains (not shown) reveal that this south-westerly wind is confined to the foot of the Cévennes Mountains. Thus, the associated wind convergence obtained in WRF simulations seems to be induced by local processes. Note that similar results are obtained in WRF-THOM.
Section 5 and Section 6 demonstrate the influence of different vertical distributions of water species (cloud, rain, ice) for precipitation amount (Figure 3) and the shifting of the main convective band due to low-level convergence in the WRF simulations, compared to DESCAM. In Section 7, we will investigate the origin of the different low-level wind convergence (Figure 9) obtained in both WRF-MORR and WRF-THOM.

7. Diabatic Effects

The forcing data (IFS analyses) used in order to study the IOP7a intense convective event are the same for both the DESCAM and the WRF simulations. Thus, the low-level wind convergence obtained in WRF experiments cannot be due to the initial synoptic conditions, or the relief representation since the resolutions are identical in the different experiments. Moreover, the wind convergence is also produced (not shown) when the WRF model is initiated with the ERA-Interim reanalyses (Section 4.2), implying that the origin of these dynamical features is inherent to the WRF model.
To understand the origin of this low-level wind convergence in WRF, an experiment called WRF-NOMIC (details in Table 1 and Table 2), where none of the cloud and precipitation microphysics processes occur, is performed to untangle the dynamics and the microphysics effects. The comparison between the wind properties obtained in WRF-NOMIC at 0.5 km asl and 07:00 UTC (Figure 10a) with those obtained in DESCAM, WRF-THOM and WRF-MORR at the same altitude and time (Figure 8a,c,e) reveals that a south-easterly low-level wind reaches the east slopes of the Cévennes Mountains in WRF-NOMIC, as in DESCAM: until this time the wind convergence does not appear in WRF-NOMIC. This sensitivity study indicates that the microphysical processes impact the dynamics properties locally.
Moreover, the thermodynamics properties simulated in WRF-NOMIC are compared to DESCAM, WRF-THOM and WRF-MORR experiments within a model volume (located near the wind convergence area) with a base of 12 × 10 km2 and a vertical depth of 6 km (illustrated by the rectangle in Figure 10b). In this column, the calculated mean profile of temperature (Figure 11) is quite similar in WRF-MORR and WRF-THOM. On the contrary, in WRF-NOMIC, the temperature is lower than in the other WRF experiments all along the profile. As cloud processes are switched off in WRF-NOMIC then the latent heat release is reduced since no phase changes occur (i.e., condensation or freezing).
A further study of the the mean temperature profile simulated in WRF-NOMIC and the three other experiments (i.e., DESCAM, WRF-MORR, and WRF-THOM) indicates that the most important discrepancies appear at around 5 km and within the two first km of the atmosphere (Figure 11). At 5 km, the local warming is more important in the WRF experiments than in DESCAM because the condensed water content is larger. Indeed, the ice mixing ratio is two times larger in WRF-MORR than in DESCAM (Figure 7c). In the lower atmosphere, the presence (or the absence) of precipitation (see Figure 3) in the studied volume induces a less (or more) intense local warming in WRF (or in DESCAM). As the atmosphere is subsaturated in this layer (Figure 6), the precipitation can evaporate [49,50]. Thus, the latent heat consumed during the evaporation process induces colder environment temperatures in WRF-MORR and WRF-THOM than in DESCAM.
Figure 10b–d show the atmospheric temperature difference ( Δ T ) between WRF-MORR and WRF-NOMIC at 07:00 UTC (same time than in Figure 11) and at three specific altitudes: at 0.5 km, 2 km and 5 km which correspond to, respectively, the altitude of the wind convergence, and the altitudes where the amount of the liquid and the ice phases are the highest in WRF-MORR (see Figure 7). As expected, especially at 2 km and 5 km, the latent heat released associated to the cloud formation processes (condensation and freezing) generally induces warmer environment temperatures in WRF-MORR (extended red areas) and only small portions show an atmospheric cooling in WRF-MORR (blue color). For example, Figure 10c shows that the temperature difference reaches ≈4.5 °C over the relief at 2 km height.
Figure 12 represents the hourly evolution of the temperature difference ( Δ T ) between WRF-MORR and WRF-NOMIC at 0.5 km. In combination with Figure 10b, this shows the temporal evolution of the Δ T field between 01:00 and 07:00 UTC. Figure 12 also represents the temporal evolution of the cloudy areas at 0.5 km and the surface accumulated rain in WRF-MORR. When clouds form at 02:00 UTC, the temperature locally (i.e., close to the slopes of the Cévennes mountains) becomes warmer in WRF-MORR (Figure 12b). At the same time, precipitation is already visible over the reliefs. Then, with time, the cloud field becomes more extended over the valley and the latent heat released during the cloud condensation induces a warmer air (Figure 12c,d). These thermodynamic conditions impact the horizontal wind direction building-up an up-valley wind along the windward side of Cévennes. This creates a barrier flow (such as in Reference [61]) leading to a low-level convergence which initiates convection afar from the mountains (see Figure 9f).
At 05:00 UTC, the cloud base is at higher altitude (i.e., altitude where q c > 0.01 g kg−1, not shown) and the precipitation starts in the valley close to the convergence line. Later, when the precipitation intensifies (Figure 12f), it locally impacts the wind speed (Figure 9e,f) and the temperature of the atmosphere close to the surface. Indeed, as shown in Reference [49], below the cloud base where the atmosphere is sub-saturated, the precipitation can evaporate. Thereby the latent heat consumed during this process induces a decrease of the air temperature (see blue patches in Figure 12e,f).
The low-level wind convergence follows the modification of the temperature field induced by the cloud formation at the beginning of the simulation. The microphysics processes (especially, those involving a phase change) in WRF substantially impact the dynamics of this case study and does not allow to reproduce the observed accumulated precipitation field at the surface.

8. Conclusions

In this study a highly resolved 3D cloud model with detailed (bin) microphysics called DESCAM and the WRF model with different bulk microphysics schemes are used to simulate an intense convective cloud system. This cloud system was observed during the HyMeX campaign that took place in the South of France during the autumn 2012. The convective cells were initiated over the Cévennes Mountains (France) in the early morning of the 26 September and persisted there for about 5 hours with a very intense period of rain from 6:00 to 9:00 UTC.
The radar reflectivity fields and cumulative rain amount simulated with DESCAM show a reasonable agreement with high resolution X-band radar observations and with the rain amount retrieved from the ARAMIS French radars network, even if for the lowest rain amounts, the irrigated area is smaller in DESCAM than in the observations (for additional details see Kagkara et al. [29]). On the contrary, in the WRF simulations, the rain amount is largely overestimated and the main irrigated region is shifted 20 km to the East compared to the observations.
In order to identify the origin of the differences between model results, we investigate the thermodynamics, the microphysics and the dynamics features. The mean thermodynamics profiles in DESCAM and WRF show large moisture differences, especially at low-level. The much drier low-level in DESCAM probably affects the instability of the atmosphere and the intensity of convection explaining the lower precipitation amount in DESCAM than in WRF. Moreover, regarding the microphysics, the mean vertical profiles for cloud water, rain and ice contents simulated in the two models are different, especially for the cloud water and the ice contents. Indeed, between 3 and 6 km height, the cloud water content is approximately 3 times higher in DESCAM than in WRF. On the contrary, the ice content is up to 2 times higher in WRF than in DESCAM over the same vertical layer. These differences in the vertical distribution of the cloud ice- and the liquid-water phases impact the formation of the precipitation. Consequently, the larger amount of ice present in WRF simulations induces larger rain amounts than in DESCAM simulations; this is consistent with more-efficient formation of precipitation when ice is involved [13] among others.
Regarding the dynamics features, the triggering of the convective system is due to the complex relief of the Cévennes Mountains in DESCAM and by a low-level flux convergence located 10–15 km upstream of the mountainous area in WRF. An additional sensitivity study shows that if the microphysics is switched off in WRF, the low-level flux convergence disappears. A further analysis of the temperature fields obtained in the WRF simulations with or without a microphysics scheme reveals that the latent heat released during the cloud formation at the beginning of the simulations locally impacts the atmospheric properties. Indeed, the local temperature differences help to build-up an up-valley wind along the windward side of the Cévennes Mountains leading to a low-level convergence which induces the convection initiation afar the mountains slopes (i.e., in the Rhone Valley). Thus, the low-level flux convergence present in WRF simulations is induced by diabatic effects of the microphysical processes.
As shown by Reference [50], inaccuracies in the DSD (Drop Size Distribution) representation induce underestimation of the evaporation rate within the stratiform region of an intense squall line case in WRF, impacting then not only the rain rate but also the strength of the cold pool associated to this system. In the intense convective system studied in this paper, the condensation process strongly impacts the dynamics fields inducing a precipitation field different from the one observed. So, in order to better estimate this intense precipitation event, it is important to advance our understanding of the involved microphysics processes by studying other convective systems. Moreover, a statistical approach, using various model settings such as in References [62,63], could further confirm the robustness of our simulation results. Furthermore, in future works, to make progress on the estimate of the impact of the diabatic effects of the microphysics processes on the dynamics of convective systems, it would be helpful to perform sensitivity studies using a detailed (bin) microphysics scheme coupled to the WRF model. Accordingly, the implementation of the DESCAM microphysics module into the WRF model is in progress.

Author Contributions

C.P. and W.W. performed the bulk and bin simulations; S.B. installed the WRF model on the different supercomputers; D.A., C.K., W.W. and C.P. analysed the observations and model data; F.T. assisted in the analysis and interpretation of the radar data; A.F. is the LaMP PI of the MUSIC project and contributed to the paper; C.P. wrote the paper with the help of all the other authors. All authors have read and agreed to the published version of the manuscript.

Funding

This work is a contribution to the HyMeX program and MUSIC project, supported by Grants ANR-14CE1-0014 and MISTRALS/Hymex. C.K is funded by the MUSIC project. D.A. is funded by the Challenge 4 of the Clermont I-Site CAP20-25 project. The model calculations have been done on French computer facilities of the Institut du Développement des ressources en Informatique Scientifique (IDRIS) CNRS at Orsay, the Centre Informatique National de l’Enseignement Supérieur (CINES) at Montpellier under the project 940180 and the Centre Régional de Ressources Informatiques (CRRI) at Clermont-Ferrand.

Acknowledgments

The authors acknowledge Meteo-France for supplying the data and the HyMeX database teams (ESPRI/IPSL and SEDOO/OMP). The authors especially acknowledge Doerenbecher Alexis for the High-resolution operational radiosoundings, Nimes product and Labatut Laurent for the French Radar composite 5 mm cumulative rainfall in mm product. For the X-band radar data, the authors thank Joel Van Baelen, Sandra Banson and Yves Pointin for their help in assessing the data. X-band data were obtained from the HyMeX program, sponsored by Grants MISTRALS/HyMeX and ANR-2011-BS56-027 FLOODSCALE project. Data of the UHF radar situated at Candillargues are available via the database. We acknowledge the CNRM/GAME, UMR3589, CNRS/Meteo-France for measurements providing and the Laboratoire d’Aérologie, Université de Toulouse, UMR CNRS 5560 for data processing.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Ricard, D.; Ducrocq, V.; Auger, L. A climatology of the mesoscale environment associated with heavily precipitating events over a northwestern Mediterranean area. J. Appl. Meteor. Climatol. 2012, 51, 468–488. [Google Scholar] [CrossRef]
  2. Llasat, M.C.; Llasat-Botija, M.; Petrucci, O.; Pasqua, A.A.; Rosselló, J.; Vinet, F.; Boissier, L. Towards a database on societal impact of Mediterranean floods within the framework of the HYMEX project. Nat. Hazard. Earth Syst. 2013, 13, 1337–1350. [Google Scholar] [CrossRef] [Green Version]
  3. Ducrocq, V.; Nuissier, O.; Ricard, D. A numerical study of three catastrophic precipitating events over Southern France. Part II: Mesoscale trigerring and stationarity factors. Quart. J. R. Meteor. Soc. 2008, 134, 131–145. [Google Scholar] [CrossRef]
  4. Buzzi, A.; Davolio, S.; Malguzzi, P.; Drofa, O.; Mastrangelo, D. Heavy rainfall episodes over Liguria of autumn 2011: Numerical forecasting experiments. Nat. Hazard. Earth Syst. 2014, 14, 1325–1340. [Google Scholar] [CrossRef] [Green Version]
  5. Nuissier, O.; Joly, B.; Joly, A.; Ducrocq, V.; Arbogast, P. A statistical downscaling to identify the large-scale circulation patterns associated with heavy precipitation events over southern France. Quart. J. R. Meteor. Soc. 2011, 137, 1812–1827. [Google Scholar] [CrossRef]
  6. Duffourg, F.; Ducrocq, V. Origin of the moisture feeding the Heavy Precipitating Systems over Southeastern France. Nat. Hazard. Earth Syst. 2011, 11, 1163–1178. [Google Scholar] [CrossRef] [Green Version]
  7. Bousquet, O.; Smull, B. Observations and impacts of upstream blocking during a widespread orographic precipitating event. Quart. J. Roy. Meteor. Soc. 2003, 129, 391–409. [Google Scholar] [CrossRef]
  8. Drobinski, P.; Ducrocq, V.; Alpert, P.; Anagnostou, E.; Béranger, K.; Borga, M.; Braud, I.; Chanzy, A.; Davolio, S.; Delrieu, G.; et al. HyMEX: A 10-year multidisciplinary program on the Mediterranean water cycle. Bull. Am. Meteor. Soc. 2014, 95, 1063–1082. [Google Scholar] [CrossRef]
  9. Ducrocq, V.; Braud, I.; Davolio, S.; Ferretti, R.; Flamant, C.; Jansa, A.; Kalthoff, N.; Richard, E.; Taupier-Letage, I.; Ayral, P.A.; et al. HyMeX-SOP1: The field campaign dedicated to heavy precipitation and flash flooding in the Northwestern Mediterranean. Bull. Am. Meteor. Soc. 2014, 95, 1083–1100. [Google Scholar] [CrossRef]
  10. Clark, A.J.; Weiss, S.J.; Kain, J.S.; Jirak, I.L.; Coniglio, M.; Melick, C.J.; Siewert, C.; Sobash, R.A.; Marsh, P.T.; Dean, A.R.; et al. An Overview of the 2010 Hazardous Weather Testbed Experimental Forecast Program Spring Experiment. Bull. Am. Meteor. Soc. 2012, 93, 55–74. [Google Scholar] [CrossRef]
  11. Kanji, Z.A.; Ladino, L.A.; Wex, H.; Boose, Y.; Burkert-Kohn, M.; Cziczo, D.J.; Krämer, M. Overview of ice nucleating particles. Meteor. Monogr. 2017, 58, 1. [Google Scholar] [CrossRef] [Green Version]
  12. Hiron, T.; Flossmann, A.I. A Study of the Role of the Parameterization of Heterogeneous Ice Nucleation for the Modeling of Microphysics and Precipitation of a Convective Cloud. J. Atmos. Sci. 2015, 72, 3322–3339. [Google Scholar] [CrossRef]
  13. Pruppacher, H.R.; Klett, J.D. Microphysics of Clouds and Precipitation; Second Revised and Expanded Edition with an Introduction to Cloud Chemistry and Cloud Electricity; Springer: Dordrecht, The Netherlands; Heidelberg/Berlin, Germany; London, UK; New York, NY, USA, 2010; p. 954. [Google Scholar]
  14. Thompson, G.; Rasmussen, R.M.; Manning, K. Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. Part I: Description and sensitivity analysis. Mon. Weather Rev. 2004, 132, 519–542. [Google Scholar] [CrossRef] [Green Version]
  15. Planche, C.; Marsham, J.H.; Field, P.R.; Carslaw, K.S.; Hill, A.A.; Mann, G.W.; Shipway, B.J. Precipitation sensitivity to autoconversion rate in a numerical weather-prediction model. Quart. J. R. Meteor. Soc. 2015, 141, 2032–2044. [Google Scholar] [CrossRef] [Green Version]
  16. McFarquhar, G.M.; Zhang, H.; Heymsfield, G.; Halverson, J.B.; Hood, R.; Dudhia, J.; Marks, F. Factors Affecting the Evolution of Hurricane Erin (2001) and the Distributions of Hydrometeors: Role of Microphysical Processes. J. Atmos. Sci. 2006, 63, 127–150. [Google Scholar] [CrossRef] [Green Version]
  17. Khain, A.P.; Rosenfeld, D.; Pokrovsky, A. Aerosol impact on the dynamics and microphysics of deep convective clouds. Quart. J. R. Meteor. Soc. 2005, 131, 2639–2663. [Google Scholar] [CrossRef] [Green Version]
  18. Flossmann, A.I.; Wobrock, W. A review of our understanding of the aerosol—Cloud interaction from the perspective of a bin resolved cloud scale modelling. Atmos. Res. 2010, 97, 478–497. [Google Scholar] [CrossRef]
  19. Thompson, G.; Field, P.R.; Rasmussen, R.M.; Hall, W.D. Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. Part II: Implementation of a new snow parameterization. Mon. Weather Rev. 2008, 136, 5095–5115. [Google Scholar] [CrossRef]
  20. Morrison, H.; Thompson, G.; Tatarskii, V. Impact of cloud microphysics on the development of trailing stratiform precipitation in a simulated squall line: Comparison of one- and two-moment schemes. Mon. Weather Rev. 2009, 137, 991–1007. [Google Scholar] [CrossRef] [Green Version]
  21. Vié, B.; Pinty, J.P.; Berthet, S.; Leriche, M. LIMA (v1.0): A quasi two-moment microphysical scheme driven by a multimodal population of cloud condensation and ice freezing nuclei. Geosci. Model Dev. 2016, 9, 567–586. [Google Scholar] [CrossRef] [Green Version]
  22. Kessler, E. On the distribution and continuity of water substance in atmospheric situations. Meteorol. Monogr. 1969, 32, 1–84. [Google Scholar]
  23. Cohard, J.M.; Pinty, J.P. A comprehensive two-moment warm microphysical bulk scheme. I: Description and tests. Quart. J. R. Meteor. Soc. 2000, 126, 1815–1842. [Google Scholar] [CrossRef]
  24. Khairoutdinov, M.; Kogan, Y. A New Cloud Physics Parameterization in a Large-Eddy Simulation Model of Marine Stratocumulus. Mon. Weather Rev. 2000, 128, 229–243. [Google Scholar] [CrossRef]
  25. Morrison, H.; Milbrandt, J.A. Parameterization of ice microphysics based on the prediction of bulk particle properties. Part I: Scheme description and idealized tests. J. Atmos. Sci. 2015, 72, 287–311. [Google Scholar] [CrossRef]
  26. Milbrandt, J.A.; Morrison, H. Parameterization of ice microphysics based on the prediction of bulk particle properties. Part III: Introduction of Multiple Free Categories. J. Atmos. Sci. 2016, 73, 975–995. [Google Scholar] [CrossRef]
  27. Pinty, J.P.; Jabouille, P. A mixed-phase cloud parameterization for use in a mesoscale non hydrostatic model: Simulations of a squall line and of orographic precipitation. In Proceedings of the Conference on Cloud Physics, Everett, DC, USA, 17–21 August 1998; pp. 217–220. [Google Scholar]
  28. Wilson, D.R.; Ballard, S.P. A microphysically based precipitation scheme for the UK Meteorological Office Unified Model. Quart. J. R. Meteor. Soc. 1999, 125, 1607–1633. [Google Scholar] [CrossRef]
  29. Kagkara, C.; Wobrock, W.; Planche, C.; Flossmann, A. The sensitivity of intense rainfall to aerosol particle loading—A comparison of bin-resolved microphysics modeling with observations of heavy precipitation from HyMeX IOP7a. Nat. Hazards Earth Syst. Sci. 2020, 20, 1469–1483. [Google Scholar] [CrossRef]
  30. Hally, A.; Richard, E.; Ducrocq, V. An ensemble study of HyMeX IOP6 and IOP7a: Sensitivity to physical and initial and boundary condition uncertainties. Nat. Hazards Earth Syst. Sci. 2014, 14, 1071–1084. [Google Scholar] [CrossRef] [Green Version]
  31. Tabary, P. The new French operational radar rainfall product. Part I: Methodology. Weather Forecast. 2007, 22, 393–408. [Google Scholar] [CrossRef]
  32. Planche, C.; Wobrock, W.; Flossmann, A.I.; Tridon, F.; VanBaelen, J.; Pointin, Y.; Hagen, M. The influence of aerosol particle number and hygroscopicity on the evolution of convective cloud systems and their precipitation: A numerical study based on the COPS observations on 12 August 2007. Atmos. Res. 2010, 98, 40–56. [Google Scholar] [CrossRef] [Green Version]
  33. Clark, T.L.; Hall, W.D.; Coen, J.L. Source Code Documentation for the Clark-Hall Cloud-Scale Model Code Version G3CH01; NCAR Tech. Note NCAR/TN-426+STR; NCAR: Boulder, CO, USA, 1996; p. 182. [Google Scholar]
  34. Clark, T.L. Block-iterative method of solving the non-hydrostatic pressure in terrain-following coordinates: Two-level pressure and truncation error analysis. J. Appl. Meteor. 2003, 42, 970–983. [Google Scholar] [CrossRef]
  35. Coen, J.L. Modeling Wildland Fires: A Description of the Coupled Atmosphere-Wildland Fire environment model (CAWFE); NCAR Tech. Note NCAR/TN-500+STR; NCAR: Boulder, CO, USA, 2013; p. 38. [Google Scholar]
  36. Planche, C.; Wobrock, W.; Flossmann, A.I. The continuous melting process in a cloud-scale model using a bin microphysics scheme. Quart. J. R. Meteor. Soc. 2014, 140, 1986–1996. [Google Scholar] [CrossRef]
  37. Skamarock, W.C.; Klemp, J.B.; Dudhia, J.; Gill, D.O.; Barker, D.M.; Duda, M.G.; Huang, X.Y.; Wang, W.; Powers, J.G. A Description of the Advanced Research WRF Version 3; NCAR Tech. Note NCAR/TN-475+STR; NCAR: Boulder, CO, USA, 2008; p. 125. [Google Scholar]
  38. Clark, T.L.; Hall, W.D.; Kerre, R.M.; Middleton, D.; Radke, L.; Ralph, F.; Neiman, P. On the origins of aircraft damaging clear-air turbulence during the 9 December 1992 Colorado downslope windstorm: Numerical simulations and comparison with observations. J. Atmos. Sci. 2000, 57, 1105–1131. [Google Scholar] [CrossRef]
  39. Wobrock, W.; Flossmann, A.I.; Farley, R.D. Comparison of observed and modeled hailstone spectra during a severe storm over the Northern Pyrenean foothills. Mon. Weather Rev. 2003, 67-68, 685–703. [Google Scholar]
  40. Leroy, D.; Wobrock, W.; Flossmann, A.I. The role of boundary layer aerosol particles for the development of deep convective clouds: A high-resolution 3D model with detailed (bin) microphysics applied to CRYSTAL-FACE. Atmos. Res. 2009, 91, 62–78. [Google Scholar] [CrossRef]
  41. Clark, T.L. Numerical simulations with a three dimensional cloud model: Lateral boundary condition experiments and multi-cellular severe storm simulations. J. Atmos. Sci. 1979, 36, 2191–2215. [Google Scholar] [CrossRef] [Green Version]
  42. Danko, D.M. The digital chart of the world. Geoinfo Syst. 1992, 2, 29–36. [Google Scholar]
  43. Danielson, J.J.; Gesch, D.B. Global Multi-Resolution Terrain Elevation Data 2010 (GMTED2010); U.S. Geo-logical Survey Open-File Report 2011-1073; Earth Resources Observation and Science (EROS) Center: Sioux Falls, SD, USA, 2011; p. 26. [Google Scholar]
  44. Planche, C.; Wobrock, W.; Flossmann, A.I.; Tridon, F.; Labbouz, L.; VanBaelen, J. Small scale topography influence on the formation of three convective systems observed during COPS over the Vosges Mountains. Meteorol. Z. 2013, 22, 395–411. [Google Scholar] [CrossRef] [Green Version]
  45. Tewari, M.; Chen, F.; Wang, W.; Dudhia, J.; LeMone, M.A.; Mitchell, K.; Ek, M.; Gayno, G. Implementation and verification of the unified NOAH land surface model in the WRF model. In Proceedings of the 20th Conference on Weather Analysis and Forecasting/16th Conference on Numerical Weather Prediction, Seattle, DC, USA, 12–16 January 2004; pp. 11–15. [Google Scholar]
  46. Mlawer, E.J.; Taubman, S.J.; Brown, P.D.; Iacono, M.J.; Clough, S.A. Radiative transfer for inhomogeneous atmospheres: RRTM, a validated k-correlated model for the longwave. J. Geophys. Res. 1997, 102, 663–682. [Google Scholar] [CrossRef] [Green Version]
  47. Hong, S.Y.; Noh, Y.; Dudhia, J. A New Vertical Diffusion Package with an Explicit Treatment of Entrainment Processes. Mon. Weather Rev. 2006, 134, 2318–2341. [Google Scholar] [CrossRef] [Green Version]
  48. Kain, J.S. The Kain–Fritsch Convective Parameterization: An Update. J. Appl. Meteor. 2004, 43, 170–181. [Google Scholar] [CrossRef] [Green Version]
  49. Tridon, F.; Planche, C.; Mroz, K.; Banson, S.; Battaglia, A.; Baelen, J.V.; Wobrock, W. On the realism of the rain microphysics representation of a squall line in the WRF model. Part I: Evaluation with multifrequency radar Doppler spectra observations. Mon. Weather Rev. 2019, 147, 2787–2810. [Google Scholar] [CrossRef] [Green Version]
  50. Planche, C.; Tridon, F.; Banson, S.; Thompson, G.; Monier, M.; Battaglia, A.; Wobrock, W. On the realism of the rain microphysics representation of a squall line in the WRF model. Part II: Sensitivity studies on the rain Drop Size Distributions. Mon. Weather Rev. 2019, 147, 2811–2825. [Google Scholar] [CrossRef] [Green Version]
  51. Bott, A. A flux method for the numerical solution of the stochastic collection equation. J. Atmos. Sci. 1998, 55, 2284–2293. [Google Scholar] [CrossRef]
  52. Hall, W.D. A detailed microphysical model within a two-dimensional dynamic framework: Model description and preliminary results. J. Atmos. Sci. 1980, 37, 2486–2506. [Google Scholar] [CrossRef]
  53. Koop, T.N.H.P.; Luo, B.; Tsias, A.; Peter, T. Water activity as the determinant for homogeneous ice nucleation in aqueous solution. Nature 2000, 406, 611–615. [Google Scholar] [CrossRef]
  54. Meyers, M.P.; Demott, P.J.; Cotton, W.R. New primary ice nucleation parameterizations in an explicit cloud model. J. Appl. Meteor. 1992, 31, 708–721. [Google Scholar] [CrossRef] [Green Version]
  55. Dee, D.P.; Uppala, S.M.; Simmons, A.J.; Berrisford, P.; Poli, P.; Kobayashi, S.; Andrae, U.; Balmaseda, M.A.; Balsamo, G.; Bauer, P.; et al. The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. R. Meteor. Soc. 2011, 137, 553–597. [Google Scholar] [CrossRef]
  56. Tabary, P.; Augros, C.; Champeaux, J.L.; Chèze, J.L.; Faure, D.; Idziorek, D.; Lorandel, R.; Urban, B.; Vogt, V. Le réseau et les produits radars de Météo-France. La Météorologie 2013, 83, 15–27. [Google Scholar] [CrossRef] [Green Version]
  57. Boudevillain, B.; Delrieu, G.; Galabertier, B.; Bonnifait, L.; Bouilloud, L.; Kirstetter, P.; Mosini, M. The Cevennes-Vivarais Mediterranean Hydrometeorological Observatory database. Water Resour. Res. 2011, 47. [Google Scholar] [CrossRef]
  58. Marzoug, M.; Amayenc, P. Improved range-profiling algorithm of rainfall rate from a spaceborne radar with path-integrated attenuation constraint. IEEE Trans. Geosc. Remote Sens. 1991, 29, 584–592. [Google Scholar] [CrossRef]
  59. Delrieu, G.; Hucke, L.; Creutin, J.D. Attenuation in Rain for X- and C-Band Weather Radar Systems: Sensitivity with respect to the Drop Size Distribution. J. Appl. Meteorol. 1999, 38, 57–68. [Google Scholar] [CrossRef]
  60. Stephens, G.L.; Vane, D.G.; Boain, R.J.; Mace, G.G.; Sassen, K.; Wang, Z.; The CloudSat Science Team. The CloudSat mission and the A-Train: A new dimension of space-based observations of clouds and precipitation. Bull. Am. Meteor. Soc. 2002, 83, 1771–1790. [Google Scholar] [CrossRef] [Green Version]
  61. Davolio, S.; Volonté, A.; Manzato, A.; Pucillo, A.; Cicogna, A.; Ferrario, M.E. Mechanisms producing different precipitation patterns over north-eastern Italy: Insights from HyMeX-SOP1 and previous events. Q. J. R. Meteorol. Soc. 2016, 142, 188–205. [Google Scholar] [CrossRef] [Green Version]
  62. Keil, C.; Baur, F.; Bachmann, K.; Rasp, S.; Schneider, L.; Barthlott, C. Relative contribution of soil moisture, boundary-layer and microphysical perturbations on convective predictability in different weather regimes. Q. J. R. Meteorol. Soc. 2019, 145, 3102–3115. [Google Scholar] [CrossRef]
  63. Flack, D.L.A.; Gray, S.L.; Plant, R.S. A simple ensemble approach for more robust process-based sensitivity analysis of case studies in convection-permitting models. Q. J. R. Meteorol. Soc. 2019, 145, 3089–3101. [Google Scholar] [CrossRef]
Atmosphere 11 01177 g001 550
Figure 1. 12-h cumulative rainfall retrieved from the French radars network ARAMIS: (a) over France and (b) over the Cévennes-Vivarais region. The rectangles present the area of the Cévennes-Vivarais region (solid), which corresponds to region presented in the panel (b), and the domains d01, d02 and d03 (dashed) used in the different simulations. Note that the innermost domain (d03) is presented in panel (b). The S- and C-band operational radars (Bollène, Nimes and Sembadel) covering the innermost domain (d03) of the simulations are represented by the grey small filled squares (in panels (a) and (b) and by letters B, N and S (in panel (b). The red and blue crosses show respectively the position of the X-band radar of the LaMP (Laboratoire de Météorologie Physique) and the UHF radar of the CNRM (Centre National de Recherches Météorologiques - Météo-France) located at Candillargues. The sounding used in this study was launched from the Nimes station represented by the letter N in panel (b). The topography is represented in the panel (b) from sea level to 2000 m height every 400 m.
Figure 1. 12-h cumulative rainfall retrieved from the French radars network ARAMIS: (a) over France and (b) over the Cévennes-Vivarais region. The rectangles present the area of the Cévennes-Vivarais region (solid), which corresponds to region presented in the panel (b), and the domains d01, d02 and d03 (dashed) used in the different simulations. Note that the innermost domain (d03) is presented in panel (b). The S- and C-band operational radars (Bollène, Nimes and Sembadel) covering the innermost domain (d03) of the simulations are represented by the grey small filled squares (in panels (a) and (b) and by letters B, N and S (in panel (b). The red and blue crosses show respectively the position of the X-band radar of the LaMP (Laboratoire de Météorologie Physique) and the UHF radar of the CNRM (Centre National de Recherches Météorologiques - Météo-France) located at Candillargues. The sounding used in this study was launched from the Nimes station represented by the letter N in panel (b). The topography is represented in the panel (b) from sea level to 2000 m height every 400 m.
Atmosphere 11 01177 g001
Atmosphere 11 01177 g002 550
Figure 2. Radar reflectivity observed with the LaMP X-band radar (a,b) and simulated with DESCAM (c,d) at different time steps of the mature stage of the cloud system (i.e., at 07:20 UTC (a,c) and at 08:00 UTC (b,d)), considering the 1 h delay (see details in Section 3.2). The representation of these figures is the Plan Position Indicator (PPI) Radar Image. The model results outside the observational area are not represented. The dashed line indicates the mountain ridge.
Figure 2. Radar reflectivity observed with the LaMP X-band radar (a,b) and simulated with DESCAM (c,d) at different time steps of the mature stage of the cloud system (i.e., at 07:20 UTC (a,c) and at 08:00 UTC (b,d)), considering the 1 h delay (see details in Section 3.2). The representation of these figures is the Plan Position Indicator (PPI) Radar Image. The model results outside the observational area are not represented. The dashed line indicates the mountain ridge.
Atmosphere 11 01177 g002
Atmosphere 11 01177 g003 550
Figure 3. 12-h cumulative rainfall retrieved from the ARAMIS radar network (a) and simulated with DESCAM (b), WRF-THOM (c) and WRF-MORR (d) for the innermost domain. The red cross and the black circle show, respectively, the location of the X-band radar and its observed area. The topography is represented from sea level to 2400 m height every 200 m.
Figure 3. 12-h cumulative rainfall retrieved from the ARAMIS radar network (a) and simulated with DESCAM (b), WRF-THOM (c) and WRF-MORR (d) for the innermost domain. The red cross and the black circle show, respectively, the location of the X-band radar and its observed area. The topography is represented from sea level to 2400 m height every 200 m.
Atmosphere 11 01177 g003
Atmosphere 11 01177 g004 550
Figure 4. Temporal evolution of the mean accumulated rain amounts in mm in the innermost domain from simulations and retrieved from ARAMIS observations. The averages are obtained considering the grid points where the cumulative rain is ≥0.25 mm. The line representing the DESCAM results is shifted to consider the 1 hour delay in the rain development.
Figure 4. Temporal evolution of the mean accumulated rain amounts in mm in the innermost domain from simulations and retrieved from ARAMIS observations. The averages are obtained considering the grid points where the cumulative rain is ≥0.25 mm. The line representing the DESCAM results is shifted to consider the 1 hour delay in the rain development.
Atmosphere 11 01177 g004
Atmosphere 11 01177 g005 550
Figure 5. Probability density function of the X-band radar observations from 06:40 to 10:40 UTC and modelled with DESCAM over the same period every 20 min.
Figure 5. Probability density function of the X-band radar observations from 06:40 to 10:40 UTC and modelled with DESCAM over the same period every 20 min.
Atmosphere 11 01177 g005
Atmosphere 11 01177 g006 550
Figure 6. Mean vertical profiles of the temperature (solid lines) and the dew point temperature (dashed lines) between 07:00 and 12:00 UTC for the innermost domain with DESCAM, WRF-THOM and WRF-MORR simulations and obtained by the radio-sounding launched at Nimes on the 26 September 2012 at 12:00 UTC.
Figure 6. Mean vertical profiles of the temperature (solid lines) and the dew point temperature (dashed lines) between 07:00 and 12:00 UTC for the innermost domain with DESCAM, WRF-THOM and WRF-MORR simulations and obtained by the radio-sounding launched at Nimes on the 26 September 2012 at 12:00 UTC.
Atmosphere 11 01177 g006
Atmosphere 11 01177 g007 550
Figure 7. Mean vertical profiles of the mixing ratio of cloud (a), rain (b) and ice (c) obtained over the innermost domain between 07:00 and 12:00 UTC with DESCAM, WRF-THOM and WRF-MORR simulations. The mean profiles are given with ±1 standard deviation.
Figure 7. Mean vertical profiles of the mixing ratio of cloud (a), rain (b) and ice (c) obtained over the innermost domain between 07:00 and 12:00 UTC with DESCAM, WRF-THOM and WRF-MORR simulations. The mean profiles are given with ±1 standard deviation.
Atmosphere 11 01177 g007
Atmosphere 11 01177 g008 550
Figure 8. Horizontal and vertical wind components simulated over the innermost domain in DESCAM (a,b), WRF-THOM (c,d) and WRF-MORR (e,f) below the cloud system (at 0.5 km height) (a,c,e) and at the cloud base (at 2 km height) (b,d,f). The colour scale represents the vertical wind for which the updrafts and the downdrafts are given with, respectively, positive (red) and negative (blue) values. The horizontal wind is indicated by the arrows. The two time steps represent the wind field 1 h after the onset (i.e., 7:00 UTC) and at the end (i.e., 9:00 UTC) of the intense precipitation period. The yellow solid and dashed lines indicate respectively the position of the wind convergence in the WRF simulations (solid) and the position of the precipitation band (dashed) in the different simulations (see Figure 3).
Figure 8. Horizontal and vertical wind components simulated over the innermost domain in DESCAM (a,b), WRF-THOM (c,d) and WRF-MORR (e,f) below the cloud system (at 0.5 km height) (a,c,e) and at the cloud base (at 2 km height) (b,d,f). The colour scale represents the vertical wind for which the updrafts and the downdrafts are given with, respectively, positive (red) and negative (blue) values. The horizontal wind is indicated by the arrows. The two time steps represent the wind field 1 h after the onset (i.e., 7:00 UTC) and at the end (i.e., 9:00 UTC) of the intense precipitation period. The yellow solid and dashed lines indicate respectively the position of the wind convergence in the WRF simulations (solid) and the position of the precipitation band (dashed) in the different simulations (see Figure 3).
Atmosphere 11 01177 g008
Atmosphere 11 01177 g009 550
Figure 9. Hourly evolution of the horizontal and vertical wind components simulated from 01:00 to 06:00 UTC (a–f) over the innermost domain in WRF-MORR at 0.5 km height. The colour scale represents the vertical wind for which the updrafts and the downdrafts are given with, respectively, positive (red) and negative (blue) values. The horizontal wind is indicated by the arrows. The yellow line indicates the position of the horizontal wind convergence simulated at 07:00 UTC in the WRF simulation (see Figure 8).
Figure 9. Hourly evolution of the horizontal and vertical wind components simulated from 01:00 to 06:00 UTC (a–f) over the innermost domain in WRF-MORR at 0.5 km height. The colour scale represents the vertical wind for which the updrafts and the downdrafts are given with, respectively, positive (red) and negative (blue) values. The horizontal wind is indicated by the arrows. The yellow line indicates the position of the horizontal wind convergence simulated at 07:00 UTC in the WRF simulation (see Figure 8).
Atmosphere 11 01177 g009
Atmosphere 11 01177 g010 550
Figure 10. Horizontal and vertical wind components simulated over the innermost domain in WRF-NOMIC below the cloud system (at 0.5 km height) and at 7:00 UTC (a) and temperature difference ( Δ T ) between WRF-MORR and WRF-NOMIC at 0.5 km (b) and at two other altitudes: 2 km and 5 km (c,d), that is, the altitudes where the amount of the condensed water phases are the highest (see Figure 7). The positive values indicate that the temperature is warmer in WRF-MORR than in WRF-NOMIC. The yellow segment indicates the position of the low-level wind convergence simulated in WRF-MORR at 07:00 UTC.
Figure 10. Horizontal and vertical wind components simulated over the innermost domain in WRF-NOMIC below the cloud system (at 0.5 km height) and at 7:00 UTC (a) and temperature difference ( Δ T ) between WRF-MORR and WRF-NOMIC at 0.5 km (b) and at two other altitudes: 2 km and 5 km (c,d), that is, the altitudes where the amount of the condensed water phases are the highest (see Figure 7). The positive values indicate that the temperature is warmer in WRF-MORR than in WRF-NOMIC. The yellow segment indicates the position of the low-level wind convergence simulated in WRF-MORR at 07:00 UTC.
Atmosphere 11 01177 g010
Atmosphere 11 01177 g011 550
Figure 11. Mean vertical profile of the temperature obtained in the column represented in the Figure 10b for DESCAM, WRF-MORR, WRF-THOM and WRF-NOMIC simulations at 07:00 UTC.
Figure 11. Mean vertical profile of the temperature obtained in the column represented in the Figure 10b for DESCAM, WRF-MORR, WRF-THOM and WRF-NOMIC simulations at 07:00 UTC.
Atmosphere 11 01177 g011
Atmosphere 11 01177 g012 550
Figure 12. Hourly evolution of the temperature difference ( Δ T ) between WRF-MORR and WRF-NOMIC at 0.5 km height from 01:00 to 06:00 UTC (a–f). The positive values indicates that the temperature is warmer in WRF-MORR than in WRF-NOMIC. The gray areas represent the zone where the cloud mixing ratio is greater than 0.001 g kg−1. The rain accumulation at the surface (in WRF-MORR) is represented from 10 mm every 10 mm by the solid blue lines. The yellow segment indicates the position of the wind convergence found at 07:00 UTC in the WRF simulations (see Figure 8).
Figure 12. Hourly evolution of the temperature difference ( Δ T ) between WRF-MORR and WRF-NOMIC at 0.5 km height from 01:00 to 06:00 UTC (a–f). The positive values indicates that the temperature is warmer in WRF-MORR than in WRF-NOMIC. The gray areas represent the zone where the cloud mixing ratio is greater than 0.001 g kg−1. The rain accumulation at the surface (in WRF-MORR) is represented from 10 mm every 10 mm by the solid blue lines. The yellow segment indicates the position of the wind convergence found at 07:00 UTC in the WRF simulations (see Figure 8).
Atmosphere 11 01177 g012
Table
Table 1. Description of the most pertinent DESCAM and Weather Research and Forecasting (WRF) physical characteristics for this study (assumptions for microphysics, aerosol and convection representation; databases and resolution used for the topography; surface properties).
Table 1. Description of the most pertinent DESCAM and Weather Research and Forecasting (WRF) physical characteristics for this study (assumptions for microphysics, aerosol and convection representation; databases and resolution used for the topography; surface properties).
DESCAMWRF
Microphysicsspectral (bin) representation [18,36]bulk representation [19,20]
Aerosolspectral (bin) representationnot considered
Convectionresolvedresolved
Turbulence1-order turbulence closure1.5-order turbulence closure
TopographyGTOPO30 data base (resolution: 30’) [42]GMTED2010 data base (resolution: 30’) [43]
Surfaceconstant heat & moisture fluxes [44]Unified Noah Land Surface Model [45]
Table
Table 2. Description of the model domain configuration as well as initiation and forcing conditions for DESCAM, WRF-THOM, WRF-MORR and WRF-NOMIC experiments.
Table 2. Description of the model domain configuration as well as initiation and forcing conditions for DESCAM, WRF-THOM, WRF-MORR and WRF-NOMIC experiments.
DESCAMWRF-THOMWRF-MORRWRF-NOMIC
Dynamics coreClark modelWRF-ARW
Microphysics schemeDESCAMThompson schemeMorrison schemeNo scheme
Initiation & ForcingIFS operational data & every 12 h
Number of domains3 nested domains
Horizontal resolutions8 km, 2 km and 0.5 km
Vertical resolution40–230 m60–305 m
Number of levels8072
Table
Table 3. Cumulative rain properties observed at 12:00 UTC for the ARAMIS radar network and obtained after 12 h of simulation using either the DESCAM or the WRF model. The WRF results are given for the different microphysics schemes (Thompson and Morrison) and initial data (IFS and ERA-Interim) used in this study. The maximum (in mm/12 h) and the total amount (in Megatons) of rain are obtained considering all the grid points of the innermost domain whereas the mean (in mm/12 h) and the area (in km2) of rain are obtained considering only the grid points where the cumulative rain is ≥0.25 mm.
Table 3. Cumulative rain properties observed at 12:00 UTC for the ARAMIS radar network and obtained after 12 h of simulation using either the DESCAM or the WRF model. The WRF results are given for the different microphysics schemes (Thompson and Morrison) and initial data (IFS and ERA-Interim) used in this study. The maximum (in mm/12 h) and the total amount (in Megatons) of rain are obtained considering all the grid points of the innermost domain whereas the mean (in mm/12 h) and the area (in km2) of rain are obtained considering only the grid points where the cumulative rain is ≥0.25 mm.
ARAMIS
OBS		DESCAM
IFS		WRF-THOM
IFS		WRF-MORR
IFS		WRF-THOM
ERA-Interim		WRF-MORR
ERA-Interim
Rain max. (mm/12 h)105.3119.5227.7181.7174.1169.2
Mean rain (mm/12 h)17.5510.0913.6113.9914.8516.05
Total rain mass (Mt/12 h)158.6105.6221.5229.3242.1263.0
Rain area (×103 km2)9.0410.7811.9311.9812.8211.54
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Arteaga, D.; Planche, C.; Kagkara, C.; Wobrock, W.; Banson, S.; Tridon, F.; Flossmann, A. Evaluation of Two Cloud-Resolving Models Using Bin or Bulk Microphysics Representation for the HyMeX-IOP7a Heavy Precipitation Event. Atmosphere 2020, 11, 1177. https://doi.org/10.3390/atmos11111177

AMA Style

Arteaga D, Planche C, Kagkara C, Wobrock W, Banson S, Tridon F, Flossmann A. Evaluation of Two Cloud-Resolving Models Using Bin or Bulk Microphysics Representation for the HyMeX-IOP7a Heavy Precipitation Event. Atmosphere. 2020; 11(11):1177. https://doi.org/10.3390/atmos11111177

Chicago/Turabian Style

Arteaga, Diana, Céline Planche, Christina Kagkara, Wolfram Wobrock, Sandra Banson, Frédéric Tridon, and Andrea Flossmann. 2020. "Evaluation of Two Cloud-Resolving Models Using Bin or Bulk Microphysics Representation for the HyMeX-IOP7a Heavy Precipitation Event" Atmosphere 11, no. 11: 1177. https://doi.org/10.3390/atmos11111177

APA Style

Arteaga, D., Planche, C., Kagkara, C., Wobrock, W., Banson, S., Tridon, F., & Flossmann, A. (2020). Evaluation of Two Cloud-Resolving Models Using Bin or Bulk Microphysics Representation for the HyMeX-IOP7a Heavy Precipitation Event. Atmosphere, 11(11), 1177. https://doi.org/10.3390/atmos11111177

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop