effect of turbulence parameterization on...
TRANSCRIPT
Effect of Turbulence Parameterization on Assessment of
Cloud Organization
Luiz A. T. Machado1 and Jean-Pierre Chaboureau2
1- Instituto Nacional de Pesquisas Espaciais (INPE). Centro de Previsão de Tempo e
Estudos Climáticos (CPTEC), Cachoeira Paulista, Brazil
2- Laboratoire d'Aérologie, University of Toulouse/CNRS, Toulouse, France
Submitted to Mon. Wea. Rev.
July 2014
Corresponding author address:
LUIZ Augusto Toledo MACHADO Instituto Nacional de Pesquisas Espaciais (INPE)
Centro de Previsão de Tempo e Estudos Climáticos (CPTEC) Rodovia Pres. Dutra, km 40
Cachoeira Paulista/SP - 12630-000 Brazil
Abstract
This study evaluates the cloud and rain cell organization in space and time as forecasted by
a cloud-resolving model. The forecasted fields, mainly describing Mesoscale convective
complexes and cold fronts, were utilized to generate synthetic satellite and radar images
for comparison with Meteosat Second Generation and S-band radar observations. The
comparison was made using a tracking technique that computes the size and lifetime of
cloud and rain distributions and provides histograms of radiative quantities and cloud top
height. The tracking technique was innovatively applied to test the sensitivity of forecasts
to the turbulence parameterization. The simulations with 1D turbulence produced too
many small cloud systems and rain cells with a shorter lifetime than observed. The 3D
turbulence simulations yielded size and lifetime distributions those were very consistent
with the observations. Further tests were performed on the sensitivity to the cloud mixing
length. Cloud organization was very sensitive to cloud mixing length and the use of a very
small value increased the number of small cells, much more than the simulations with 1D
turbulence. With a larger mixing length, the total number of cells, mainly the small ones,
was strongly reduced. A small cloud mixing length led to more total column integrated
rain, ice and graupel and less cloud water than with a larger one. The vertical profiles of
turbulent kinetic energy for each type of turbulent parameterization show that the
scheme with 3D turbulence describes the cloud evolution very consistently with what was
observed by satellite.
1. Introduction 1
2
Meteorological models are increasingly being used at higher space-time resolution. 3
Today’s computer power makes it possible to run models at scales of a few kilometers in 4
operational mode and at scales of few hundreds of meters for research purposes. The 5
short range forecast of precipitation at high resolution has important applications in 6
several areas of benefit to society. The greater frequency of precipitation extremes with 7
climate change (Meehl et al, 2000) reinforces the need for such forecasts able to predict 8
high rates of precipitation and produced by models running at high resolution. However, 9
high-resolution models solving or nearly solving convective and turbulence processes 10
explicitly still need to improve and to be evaluated. 11
12
It is well known that numerical models have difficulty in reproducing precipitation features 13
precisely, mainly in tropical regions (Kidd et al. 2013). A comparison of precipitation 14
between model simulation and observation normally cannot be performed point by point 15
due to the limited predictability and the large variability of the rainfall field at small spatio-16
temporal scales. Another drawback is generally the lack of a raingauge network at high 17
resolution in the region of interest. The most convenient solution for comparing simulated 18
and observed cloud and rainfall fields is to use satellite and radar data, from which 19
essential information on cloud processes can be obtained. Machado et al. (2009) described 20
the use of the difference between infrared window and water vapor channels to estimate 21
penetrative cloud and cloud lightning activity. Rosenfeld et al. (2008) used a combination 22
of satellite channels to retrieve vertical profiles of cloud particle effective radius and 23
thermodynamic phase. Chong et al. (1987) retrieved the kinematic structure using dual 24
Doppler radar. 25
26
One of the alternatives to comparing model outputs with satellite or radar observations is 27
to use direct radiative transfer algorithms, applied to model output to reproduce synthetic 28
satellite or radar images. In this way, the skill of the model in reproducing the cloud 29
processes at high space time resolution can be estimated (see Ringer et al. (2003), 30
Chaboureau and Pinty (2006), Meirold-Mautner et al. (2007) among many others). The 31
evaluation of a specific satellite brightness temperature or a channel combination or the 32
three-dimensional field observed by radar are some of the variables that can be used in 33
the comparison between a synthetic field produced from high-resolution models and 34
observations at the scale of a few kilometers (satellite) to a few hundred meters (radar). 35
36
Precipitation field estimation by numerical weather forecast models needs to make 37
considerable progress before it can be used for streamflow forecasting (Shrestha et al. 38
2013). Recently, Varble et al. (2011) compared nine cloud-resolving model simulations with 39
scanning radar reflectivity and satellite infrared brightness temperature observations. 40
Although the study was applied for only one case, they concluded that models, in general, 41
overestimated convective area and rainfall and underestimated stratiform rainfall. Negri et 42
al. (2013) checked the principal components obtained from synthetic and observed 43
satellite images to evaluate their space and time variability. They showed that the cloud 44
resolving model captured the main modes of cloud cover variability well. However, the 45
model produced many more cloud systems than observed by Meteosat Second Generation 46
(MSG.). Recently, Caine et al. (2013) compared high-resolution simulations with radar data 47
and found that the model produced smaller, taller rainfall cells than were observed. 48
49
The reasons for these deficiencies are not well established. They include the poor 50
description of the three-dimensional initial field and some systematic model errors. In 51
particular, the high sensitivity of high–resolution simulations to the microphysical 52
parameterization is often pointed out as a model limitation (see Morrison et al., 2007 for 53
an evaluation of different microphysical models). Another very important process that 54
controls the turbulent mixing between the cloud and its environment is entrainment. As 55
such, it is one of the most sensitive and important unknown parameters in deep 56
convective schemes (Mapes and Neale 2011). Wang et al. (2007), Wu et al. (2009) and Lu 57
et al. (2013) are some of several studies addressing the entrainment effect on cloud 58
formation, strengthening and microphysical properties. At the kilometer scale, 59
entrainment is partly represented by the subgrid turbulence scheme as shown for the 60
boundary layer by Honnert et al. (2011). So a good representation of the turbulence is 61
essential in the description of cloud processes and cloud space-time organization. The 62
increase in Turbulent Kinetic Energy (TKE) facilitates convection by helping to raise parcels 63
up to their level of free convection. Conversely, high entrainment can act to inhibit the 64
convective processes, thus delaying the transition from shallow to deep convection. So far, 65
very few studies have focused on the effect of turbulence on cloud organization. 66
67
These recent studies on cloud-resolving models point out the model deficiencies in: 68
describing the space-time cloud organization, the high cloud top of small convective cells, 69
and the dependence of a specific model set up. Here we examine the representation of 70
turbulence in a cloud-resolving model by looking at its impact on the cloud organization in 71
comparison with satellite and radar observations. The dataset employed in this study was 72
collected during one of the seven field campaigns of the CHUVA project (Cloud processes 73
of tHe main precipitation systems in Brazil: A contribUtion to cloud resolVing modeling and 74
to GPM (globAl precipitation measurement); see Machado et al. 2013 for a detailed 75
description). This study uses data collected in the CHUVA-SUL campaign, in Santa Maria 76
Rio Grande do Sul State, in the south of Brazil, from 15 November to 15 December 2012. 77
This is a region with a very high frequency of Mesoscale Convective Systems (Salio et al., 78
2007) and is close to the region where severe hailstorms occur most often (Cecil and 79
Blankenship, 2012). The present article describes a new methodology for comparing 80
simulations with satellite and radar observations using a tracking technique. This new 81
methodology allows us to evaluate cloud organization in space and time and to test the 82
effect of turbulence on the cloud organization. 83
84
Section 2 describes the simulations and the verification approach. Section 3 gives an 85
overview of the simulations in terms of brightness temperature distribution and cloud 86
space and time organization. Section 4 discusses the effect of different turbulence and 87
cloud mixing length parameterizations on the cloud organization and the turbulent kinetic 88
energy. Finally, section 5 summarizes our main findings. 89
90
2. Model and evaluation approach 91
92
a. Meso-NH simulations 93
94
During the CHUVA SUL campaign, the Meso-NH model (Lafore et al. 1998) version 4.9 was 95
run using the two-way interactive grid-nesting method (Stein et al. 2000) with two nested 96
grids, a horizontal grid mesh of 10 and 2 km and a vertical grid with 62 levels. The initial 97
and boundary conditions were provided by European Center for Medium-range Weather 98
Forecasts (ECMWF) analysis and forecasts issued at 1200 UTC each day, from which the 99
model was run for 36 h. The model includes parameterizations for turbulence (Cuxart et al. 100
2000), subgrid shallow convection (Pergaud et al. 2009), mixed-phase microphysics (Pinty 101
and Jabouille 1998), and subgrid cloud cover and condensate content (Chaboureau and 102
Bechtold 2005). The radiative scheme employed was the Rapid Radiative Transfer Model 103
(RRTM; Mlawer et al. 1997) for long-wave radiation and the two-stream formulation 104
originally employed by Fouquart and Bonnel (1986) for short-wave. The convection 105
scheme of Kain and Fritsch (1993), adapted to the Meso-NH model by Bechtold et al. 106
(2001), was activated for the 10-km grid, while convection was assumed to be explicitly 107
resolved for the 2-km grid (simulation using 500 by 500 grid points, see the domain in Fig. 108
1). From the model outputs, satellite brightness temperatures were computed using the 109
radiative transfer code RTTOV (Radiative Transfer for Tiros Operational Vertical Sounder) 110
version 8.7 (Saunders et al. 2005). Radar reflectivity was simulated using the methodology 111
described by Richard et al. (2003). 112
113
The turbulence scheme implemented in Meso-NH by Cuxart et al. (2000) is based on a 114
prognostic equation for the TKE together with conservative variables for moist non-115
precipitating processes. It can be used in both 3D and 1D mode, for large-eddy simulation 116
and mesoscale configurations, respectively. During the campaign, the model was run with 117
1D turbulence. In that case, it was assumed that the horizontal gradients and turbulent 118
fluxes were negligible compared to their vertical counterparts. The mixing length was 119
parameterized to represent the size of the largest eddies following Bougeault and 120
Lacarrère (1989). Sensitivity tests were run with the turbulence scheme in the 3D mode 121
using the Deardorff mixing length, which equals the grid size limited by the stability in the 122
entrainment zone. 123
124
b. Evaluation approach 125
126
The Meteosat Second Generation (MSG) images and the Constant Altitude Plan Position 127
Indicators (CAPPIs) of the S-band radar (10-cm wavelength) at Canguçu were used. The 128
MSG images employed in the present study were at the 10.8 m brightness temperature 129
(Tir), as these are the images mainly affected by the cloud top emission. The radar data 130
were from 2 to 15 km height, covering 250 km radius, with 1-km spatial resolution. The 131
radar was the most suitable instrument for comparisons with rainfall simulations because 132
of their very similar resolution and close relationship to the rainfall field. However, radar 133
data presents several non-precipitation echoes (clutters or others associated errors) that 134
are very difficult to filter out automatically. In order to avoid this echo-induced noise, we 135
used only rainfall cell pixels having reflectivity values larger than 20 dBZ. The region 136
covered by the radar was only part of the whole simulated region (around 25% of the area, 137
see Fig. 1) employed in the comparison with satellite images. 138
139
Six “golden” cases were chosen as the most important convective systems crossing the 140
field campaign experiment. All were associated with a large-scale cloud organization 141
forced by cold frontal systems penetrating into South America. The mesoscale systems 142
formed during these days were generally very well organized. The golden days studied 143
were Julian days 327, 333, 335, 338, 339 and 345. For each case, an hourly comparison 144
was performed with satellite data and, for some days, with radar (only a few days had a 145
complete data record). Each golden case consisted of a 36-h simulation from 1200 UTC. 146
When a case study is mentioned as a specific Julian day, it corresponds to 36 satellite or 147
radar images and model simulations, one each hour, projected over the 2-km resolution 148
grid model for satellite or 1-km grid resolution for radar comparisons. The comparison 149
using the radar horizontal field, CAPPI at 2 km height, was used to approximately represent 150
the precipitation field. For the evaluation using three-dimensional fields, observations 151
were computed using 1-km radar vertical resolution. 152
153
Simulation and observation were not compared case to case but from the statistical point 154
of view. Figure 1 shows the Tir from MSG and the Meso-NH simulation on 1 December 155
2012 (Julian day 335) at 2000 UTC (32-h lead time). For this particular case, the observed 156
and simulated Tir fields are very similar, showing a cold front covering Brazil, Argentina 157
and Paraguay, and moving northwards. In both images, cold cloud tops are embedded in 158
stratiform clouds surrounded by middle and low-level clouds. However, it can be seen that 159
the simulation shows more middle- and low-level clouds and a less marked cloud 160
organization than the observations. Due to the cloud organization into cold fronts or 161
mesoscale convective complexes forced by large-scale features, simulations generally 162
compared very well with observations, in spite of some lags in the system evolution. In a 163
visual inspection between model and observation, simulated Tir agreed with the 164
observation much better than the forecasted precipitation field did. 165
166
3. Results with standard turbulence scheme 167
168
a. Overall performance of the Meso-NH simulations 169
170
A first step in comparing satellite image and numerical model simulation consists in 171
utilizing the histogram technique, traditionally used in studies comparing model and 172
satellite data. For example, Chaboureau et al. (2008) employed this technique for the 173
evaluation of Meso-NH cloud fields and the retrieval of hydrometeor characteristics using 174
the model outputs. Figure 2 shows the Tir histogram for the MSG observations and the 175
Meso-NH simulations for the 36 images of each Julian day for all 6 golden days. The two 176
histograms compare well, with a peak for high Tir associated with clear sky and low-level 177
clouds, a nearly constant frequency between 260 K and 210 K and a fast decrease in the 178
Tir population after 201 K. The largest discrepancies in the simulations are the higher 179
number of pixels for the Tir range of clear sky and low-level clouds and the smaller 180
number of cold pixels between 200 and 240 K, in part as consequence of the relative larger 181
proportion of clear sky pixels. The population of very cold tops, lower than 200 K, is very 182
similar in the observations and the simulations. These results indicate that the Meso-NH 183
simulations have more clear sky and low-level clouds, a smaller population of stratiform-184
convective clouds and a similar amount of deep convective cloud. This behavior is 185
reasonably similar from day to day. 186
187
b. Space-time organization of clouds 188
189
The novelty of the evaluation approach, made in the observation space, is to assess the 190
cloud organization, a property that cannot be evaluated with a Tir histogram. We checked 191
whether the simulated cloud and rain fields described the same organization as observed 192
by MSG and the S-band radar, respectively. Cloud organization was defined as clusters of 193
pixels with Tir lower than 235 K. Machado et al. (1998) discussed the use of this threshold 194
to represent the organization of convective clouds into clusters using satellite images. The 195
clusters of 235-K pixels were tracked using the “Forecast and Tracking the evolution of 196
Cloud Clusters” (Fortracc) technique, described by Vila et al. (2008). Fortracc is an 197
algorithm that tracks the MCS radiative and morphological properties, using infrared 198
satellite imagery or radar in a regular grid. The main components of this software are the 199
following: a cloud cluster detection method based on a size and temperature threshold; a 200
statistical module to determine morphological and radiative parameters of each 201
convective system; a tracking technique based on overlapping of convective system areas 202
in successive images; and a forecast module based on the evolution in previous time steps. 203
This last module was not employed in this study. The minimum size tracked using satellite 204
data was 60 pixels, which corresponds to an effective radius of around 9 km. The analysis 205
of radar rainfall cell size was applied using a 20-dBZ threshold following Machado et al. 206
(2002). Considering a Marshall Palmer distribution, a reflectivity value of 20 dBZ 207
corresponds to a rain rate of 1 mm/h. 208
209
Figure 3a shows the cloud size distributions observed by MSG and simulated by Meso-NH 210
for the 6 golden days. The Meso-NH cloud field has nearly twice as many small cloud 211
systems as are actually observed. This difference in cloud system numbers decreases as 212
the cloud system size increases. It results in a similar number of systems with effective 213
radius (an equivalent are circle: effective radius = (Area/)1/2) above 100 km. For the 214
largest convective systems, i.e. beyond an effective radius of 300 km, the model simulates 215
fewer systems than are observed. It should be noted that, even if the difference in the 216
number of the largest cloud systems is very small, their associated cloud cover is very 217
large. For instance, the cloud coverage of only two systems with 300 km effective radii 218
equals that of 1800 cloud systems with 10 km effective radii. Figure 3b shows the life cycle 219
duration of the convective systems. The simulations present many more short-lived 220
systems than are observed. As the life cycle duration increases, the difference falls. Beyond 221
8 hours, simulations and observations show nearly the same number of long-lived cloud 222
systems. Also, the large systems observed by satellite are larger than the simulated ones. 223
224
The difference in the space-time organization of clouds between simulation and 225
observation contrasts with the agreement obtained with the Tir histogram. To better 226
understand this behavior, Figure 4 shows the averaged Tir histograms for the small 227
(smaller than 50 km effective radius) and large (larger than 150 km effective radius) cloud 228
systems. As a cloud system is defined as a cluster of Tir lower than 235 K, its averaged Tir is 229
thus below 235 K. Nearly 75% of the observed small cloud systems have averaged Tir 230
higher than 230 K and very few have Tir lower than 220 K (Fig. 4a). This contrasts with the 231
larger number of small cloud systems with low Tir in the simulations. The comparison for 232
the large cloud systems presents a systematic, albeit less remarkable, discrepancy. Large 233
systems in the observations show higher averaged cloud top height (colder brightness 234
temperature) than in the simulations (Fig. 4b). 235
236
These results indicate that the model simulates a larger number of small systems with 237
deeper cloud top than is observed and a smaller number of large systems, less deep than 238
observed. One possible reason for this discrepancy is the too small entrainment rate for 239
small cloud systems, resulting in a large number of too-deep clouds. Such sensitivity of 240
cloud size to entrainment is well known; Simpson (1971) proposed a parameterization of 241
the entrainment rate with inverse dependence on the tower radius. An underestimation of 242
the mixing between convective cloud updrafts and their environment can lead to cloud 243
fields as seen in the simulations. Jensen and Del Genio (2006) examined the environmental 244
factors that could determine the depth of convective clouds and the environmental 245
parameters that could be related to the entrainment rate. They found that buoyancy close 246
to the surface, the major source of kinetic energy, was the main factor responsible for the 247
variability in entrainment rate. However, as their study was limited to the development of 248
cumulus congestus clouds, they could not classify the effect of this variability for a large 249
population of clouds that should experience different degrees of entrainment with size. 250
The present study does capture the effect of the turbulence parameterization in the cloud 251
size distribution. 252
253
Similarly, the smaller number of large systems can be explained by the entrainment being 254
too strong, which reduces the cloud size and top. However, the mechanisms leading to 255
large systems are much more complex than for the small systems because of the 256
mesoscale dynamics. The underestimated number of large systems could also be related to 257
some model limitations in simulating large convective systems with large stratiform cloud 258
decks. 259
260
4. Sensitivity of the cloud organization to the representation of turbulence 261
262
a. Effect of different turbulence parameterizations 263
264
Deardorff (1980) discussed the effect of the 3D turbulence on the cloud top entrainment 265
instability and proposed the basis for 3D turbulence simulation. As discussed in Honnert et 266
al. (2011), the 2-km grid spacing is at the beginning of the gray zone in which the 267
turbulence is partially solved and partially parameterized. In consequence, most of the 268
turbulence is parameterized and the turbulence scheme can have a strong impact on the 269
simulation of cloud-topped boundary layer via cloud top entrainment instability (Deardorff 270
1980). In the present case, prevailed by mesoscale convective system associated this deep 271
convection, the relationship between the updraft and entrainment follow a different 272
process and will be better addressed in the section describing the tracers simulations. 273
274
Up to this point, the results have been based on simulations using the turbulence scheme 275
in 1D mode. We now verify the impact of the entrainment with simulations using the 276
turbulence scheme in 3D mode run for Julian days 333 and 335. These days were chosen 277
because the 36-h sequence from the simulated day at 1200 UTC was fully covered by the 278
S-band radar at least once per hour. Even every hour is too long an interval to be used for 279
tracking rainfall cells (which have much shorter lifetimes than the cloud systems) but the 280
size description is independent of the time interval. Figure 5 shows the size distribution of 281
the cloud systems and the rainfall cells for observations and simulations using 1D and 3D 282
turbulence. It shows the considerable decrease in the number of small cloud systems and 283
rainfall cells using 3D turbulence. The distributions of cloud systems for observation and 284
simulations with 3D turbulence are very close, with slightly too many small cloud systems 285
(with radii less 100 km) and a slight lack of large cloud systems, with radii greater than 300 286
km (Fig. 5a). The favorable reduction in the number of small rainfall cells with 3D 287
turbulence appears to be a little overestimated (Fig. 5b). 288
289
The turbulence scheme in 3D mode considerably improved the representation of the cloud 290
and rainfall spatial organization. Figure 6 shows the histograms for the reflectivity at 2 km 291
and the 0 dBZ echo cloud top height. Note the large population of reflectivity simulated by 292
1D turbulence, higher than that observed by the radar (Fig. 6a). The use of 3D turbulence 293
has the effect of reducing this population although the number of very large reflectivity 294
values is greater for 3D turbulence than for 1D. The simulations with 3D turbulence reduce 295
the average but increase the statistical population of high reflectivity values, typically 296
associated with high intense convection. In general, the relative frequency of the cloud top 297
echo (0 dBZ) distribution is not very well represented by the model (Fig. 6b). It is 298
noteworthy that the cloud tops around 10-12 km are probably clouds associated with the 299
convective towers and stratiform cloud deck. The simulations with 3D turbulence present 300
a smaller frequency of clouds with cloud tops below 9 km, as expected because of the 301
larger entrainment. However, the higher population of the tallest clouds was not expected 302
and may be a consequence of non-linear interaction as discussed below. Also, the greater 303
number of middle-level clouds and the fewer deep-convective clouds produced with 1D 304
turbulence could be explained by the larger number of isolated, deep clouds and the fewer 305
deep convective clouds with large-scale organization. This is clearly shown in Fig. 6a with 306
fewer (more) low (high) reflectivity values in the simulations with 3D turbulence. 307
308
The turbulent scheme in 3D mode takes the horizontal turbulent fluxes into account, 309
which were neglected in the 1D mode. However, these two turbulent modes were used 310
with different cloud mixing length parameterizations, the 1D mode used the Bougeault 311
and Lacarrère (1989) mixing length while the 3D mode took the Deardorff mixing length. 312
The differences seen above are a combination of these two effects, i.e. the additional 313
horizontal turbulent fluxes and the change in the mixing length parameterization. To test 314
the impact of these two effects, further simulations were run using 1D and 3D turbulence 315
modes with the two clouds mixing length parameterizations (not shown). Whatever the 316
mixing length parameterization, the simulations with 1D turbulence did not show 317
significant differences in cloud organization, probably due to the slight vertical gradient of 318
the average potential temperature in the lower levels. In the simulations with 3D 319
turbulence, the cloud organization differed according to the cloud mixing length 320
parameterization. This shows that the larger horizontal potential temperature gradient at 321
the cloud boundaries had a significant impact in the simulation. Because of the importance 322
of the cloud mixing length in 3D turbulence this effect was investigated specifically. 323
324
b. Effect of different cloud mixing lengths 325
326
The turbulence scheme in 3D mode has considerable impact in the space-time of clouds 327
and rainfall organization. Now, the effect of the large-eddy length scale that acts directly in 328
the turbulence scheme needs to be evaluated, because it modulates the turbulence effect 329
by weighting the TKE. Emanuel (1994) shows that the interface between cloud and 330
environment undergoes small scale instabilities that enhance the mixing. As the effect of 331
turbulence undertaken a cloud-size dependence, as shown in the previous results, we 332
investigated the effect of the cloud mixing length multiplied by a constant coefficient (. 333
To conduct this analysis, we tested the effect of the mixing length scaled by a factor 5, i.e. 334
we multiplied the mixing length inside the clouds by 0.2 and by 5, forcing small and 335
large entrainment respectively. Figure 7 shows the cloud size distribution during the 36-h 336
simulation for Julian day 335. The effect of the scale factor is very significant. It leads to an 337
effect similar to the simulation with 1D turbulence when 0.2, increasing the number of 338
small systems, and it drastically reduces the number of cloud systems when 5. This 339
result shows the effect of the change in cloud mixing length for the simulations and its 340
impact with respect to the cloud size. Based on the previous results, we can suppose that 341
the modulation of this factor could depend on the size of the cloud system. For a small 342
cloud system, the factor applied to the mixing length should be very high. As the cloud size 343
increases, the factor would decrease, allowing the model to reproduce the large cloud 344
patterns observed by satellite. The very high values employed here were intended to 345
demonstrate the effect of this scale factor on the cloud mixing length. More adequate 346
values, certainly closer to a scale factor of 1, should be tested to obtain a more realistic 347
cloud organization. 348
349
The effect on microphysical parameterization of using the scale factors applied to the 350
mixing length was also evaluated. Figure 8 shows the histograms of the total column 351
integrated cloud water, rain, ice and graupel. The high scale factor (5.0) leads to a 352
smaller number of events of clouds containing rainfall, graupel and ice particles. However, 353
this simulation shows the strongest rainfall, graupel and ice values. Nevertheless the 354
increase in the entrainment decreases the number of rainfall and deep convection events. 355
When such an event occurs, it can be more intense than the cases with smaller 356
entrainment. The high scale factor corresponds to a high value of entrainment and 357
consequently clouds are inhibited by mixing with the environmental air. However, when a 358
deep convective event occurs, instability is higher, probably due to the smaller number of 359
clouds and neighboring deep convective cells stabilizing the atmosphere. The statistical 360
population of cloud water with a small amount of column integrated liquid water is 361
considerably smaller for the simulation using a large mixing length than with the small 362
one, as expected because of the greater entrainment of dry air. 363
364
To further assess the effect of the different parameters in the turbulence scheme, we used 365
the Equitable Threat Score (ETS) to quantify the ability of the model to forecast a cloud 366
event at the right place. The ETS measures the fraction of correct forecasts after 367
eliminating those that would occur simply due to chance. Values of ETS are, by definition, 368
less than 1, one being the perfect score and zero meaning that all successful forecasts can 369
be attributed to chance. Categorical scores such as ETS are widely used to compare models 370
or verify the impact of a change in model parameterization. We used the ETS to compare 371
the occurrence of high clouds using a brightness-temperature threshold of 260 K simulated 372
by Meso-NH and observed by MSG, following the methodology employed by Söhne et al. 373
(2008). Figure 9 shows the ETS for Julian day 335 for the 36-h Meso-NH simulations. The 374
simulation with 3D turbulence is by far the best and its score increases with time. At 36-h 375
lead time, its score is twice as high as the simulation with 1D turbulence. The score for the 376
simulation using 0.2 is very similar to the one with 1D turbulence. The simulation using 377
=5.0 sometimes shows the best ETS. This is probably a result of the small number of 378
clouds produced by the model using such a very large value of the cloud mixing length at a 379
time when there was no observed convection. 380
381
c. Turbulent kinetic energy 382
383
Within convective clouds, conversion from shear and buoyancy generation are the main 384
sources of TKE. Entrainment is the result of turbulence near the cloud edges where 385
surrounding air mixes with cloud air parcels. TKE is associated with the cloud mass flux as 386
defined by Hohenegger and Bretherton (2011), the mass flux at cloud base is a function of 387
the ratio between convective inhibitions and mean planetary boundary layer TKE. The TKE, 388
and consequently the entrainment, are primarily controlled by the turbulence scheme and 389
the associated mixing length. Therefore evaluating the TKE allows the effect of the 390
different turbulence parameterizations and cloud mixing lengths on the entrainment and 391
cloud mass flux to be estimated. 392
393
We computed the profiles of TKE within clouds for all the simulations of Julian day 335. We 394
considered as cloud all grid points where the ice or liquid water mixing ratio was larger 395
than 10-6 kg kg-1. Figure 10 shows the 99th percentiles of the TKE profiles. These are the 396
99th percentiles inside the clouds and cannot be interpreted as the mean TKE, which is 397
biased by the greater or lesser convective activity adjusted by the rate of entrainment. 398
399
The simulations were ranked by their TKE profiles from low to high values in the following 400
sequence at nearly all levels: 0.2 scale factor, 1D turbulence, 3D turbulence and 5.0 scale 401
factor, as expected. The simulation with 1D turbulence and the scale factor of 0.2 were 402
very close, as was the simulation with 3D turbulence with the scale factor 5.0 in the sub-403
cloud layer. However, inside the cloud, the difference between individual simulations was 404
considerable. Very large (small) eddies at the boundary between the clouds and the 405
environment increased (reduced) the TKE inside clouds to unrealistically high (low) values. 406
Therefore, any adjustment in the scale factor should be a function of the pressure and 407
smaller than these boundary values. 408
409
To understand the effect of 1D and 3D turbulence in the cloud life cycle evolution, we 410
chose a medium sized convective system and followed its life cycle in the two simulations. 411
Of course, there were several nonlinear effects and the evaluation was not 412
straightforward. A convective system found in the simulation with 3D turbulence was not 413
exactly the same as in the simulation with 1D turbulence. We could compare the two 414
systems because they occurred at the same time and location. In the simulation with 3D 415
turbulence, the convective system was initiated at 1700 UTC on 30 November, reached the 416
maximum size of 80 km effective radius at 2300 UTC, and very rapidly dissipated at 0000 417
UTC on 1 December (Fig. 11c). In the simulation with 1D turbulence, the convective 418
system was initiated one hour later, at 1800 UTC, reached the maximum size of 60 km at 419
2000 UTC, and dissipated at 2300 UTC. The cloud systems were allowed to merge with 420
other systems during the life cycle and were considered dissipated if they merged with a 421
larger convective system. The different size or minimum brightness temperature and time 422
of initiation between the simulations cannot be directly associated with the different 423
turbulence parameterizations. However, as already discussed, the simulation with 1D 424
turbulence created too many small cloud systems in relation to the more organized 425
convection in the simulation with 3D turbulence. Both cloud organizations had nearly the 426
same size between 1800 and 2000 UTC. At 2100 UTC, the simulation with 3D turbulence 427
continued to increase the size of the convective system, whereas in the simulation with 1D 428
turbulence, the system was starting to decrease in size. Also, the cloud top was deeper 429
with 3D turbulence than with 1D. 430
431
Figures 11a and 11b present the evolution of the TKE profile during the life cycle. These 432
profiles are the maximum TKE value computed in an area, around the center of the 433
convective system, proportional to the effective radius of the system at the specific time. 434
In the simulation with 3D turbulence, the TKE profile shows the largest values around the 435
cloud base at the initiation followed by a large TKE increase in the middle troposphere in 436
agreement with a strong increase in the cloud size. Machado and Laurent (2004) described 437
a relationship between the cloud system area expansion, the upper level divergence and 438
the life cycle duration in the convective system initiation. In this example, it is clear that 439
the TKE increased with the system size in the first hours of the life cycle. This increase in 440
TKE was found up to 2000 UTC when the system was increasing in size and cloud top 441
height (cloud top temperature decreases, see the minimum brightness temperature). At 442
2100 UTC, the system reached the maximum size and the coldest cloud top and TKE was 443
drastically reduced. The system continued to increase in size but the cloud top decreased, 444
showing behavior similar to that observed statistically by Machado and Laurent (2004) 445
using satellite observations. This decrease in TKE up to the end of the life cycle associated 446
with the increase in effective radius shows that the simulation with 3D turbulence was 447
efficient to create a large stratiform and cirrus cloud deck after the system reached a 448
maximum of convective activity. The simulation with 1D turbulence showed a marked 449
increase in TKE at the initiation, associated with a large increase in the size of the 450
convective system. This is, however, different from the simulation with 3D turbulence, 451
probably due to smaller entrainment. The system reached the maximum cloud top height 452
one hour later and TKE was strongly reduced at this time. At 2000 UTC the system merged 453
with another system and TKE increased again. In the next few hours, TKE decreased as did 454
the size and cloud top. In the last hour the system merged with a larger system, increasing 455
cloud size and top. 456
457
The turbulence scheme in 3D mode proved to describe the cloud organization and the life 458
cycle of convective systems in cloud-resolving models very well. However, for lower 459
resolution models, such as regional models running at resolutions where setting the 460
turbulence scheme in 3D mode has no sense, one way to improve the representation of 461
cloud organization would be to develop a cloud mixing length parameterization, inside the 462
cloud, as a function of the cloud radius and height. If the cloud is small, a factor larger than 463
one should be applied near the surface, but should be gradually reduced with height. If the 464
cloud is large, the factor should be smaller than one and gradually increased with height. 465
466
5. Conclusion 467
468
This study has presented an innovative procedure for evaluating cloud organization in 469
space and time in cloud resolving models and assessing their representation of turbulence. 470
Comparisons were performed with satellite and radar data to study the morphology of the 471
cloud organization. The analysis employed used the data collected during the CHUVA-SUL 472
campaign, MSG images and Meso-NH simulations. Six main events of large-scale forcing 473
producing mesoscale convective cloud organization were analyzed. Overall, the total 474
histogram distribution of brightness temperature was quite similar between simulation 475
and observation for the six golden cases, although the model generally produced more 476
clear sky than was actually observed. Nevertheless, the cloud organization among the 477
different simulations was very different. 478
479
The simulations with 1D turbulence produced a larger number of small, tall convective 480
systems than observed. This was true for the cloud organization compared to satellite data 481
and for the rain cells compared against the radar data. The life cycle duration simulated 482
with 1D turbulence showed too large a population of short-lived cloud cells compared to 483
what was observed by MSG. The brightness temperature histogram, for small cloud 484
organization only, showed deeper simulated clouds than observed. Conversely, the 485
simulated large cloud clusters presented cloud tops that were statistically warmer than the 486
observations. These results, for simulations with 1D turbulence, demonstrated that the 487
model produced a larger number of small, deep cells than observed and nearly the same 488
number of large cells, but shallower than in the observations. One possible explanation for 489
these discrepancies is the small entrainment produced by the simulations with 1D 490
turbulence, mainly for small cloud systems. This leads to too many small, deep cells. The 491
turbulence scheme in 1D mode seems to lack the description of the turbulent kinetic 492
energy and the cloud mixing process. 493
494
The turbulence scheme in 3D mode, using Deardorff mixing length parameterization, was 495
tested to check the effect on cloud organization. These simulations drastically reduced the 496
number of small cloud systems and rain cells, presenting a distribution very close to what 497
was observed. The comparison of the reflectivity histogram and cloud top distribution 498
between simulations with 1D and 3D turbulence modes showed simulations with 3D 499
turbulence producing fewer cases of cloud tops below 10 km height and reflectivity 500
smaller than 40 dBZ. For the deeper cells having very high reflectivity and tops higher than 501
12 km, the simulations with 3D turbulence show more relative frequency than the 502
simulations with 1D turbulence. Results comparing model and radar were similar to those 503
obtained by satellite. The simulations with 3D turbulence produced fewer small rain cells 504
and these cells were less deep than those produced by 1D turbulence. For the large rain 505
cells, associated with the high reflectivity and cloud top height, the simulations with 3D 506
turbulence produced rain cells deeper and with higher reflectivities than those produced 507
by 1D turbulence. 508
509
The simulations employing the 3D turbulence mode were highly sensitive to the cloud 510
mixing length parameterization. Therefore a sensitivity test was carried out utilizing a 511
cloud mixing length scaled by constant factors increasing and decreasing the cloud mixing 512
length. Two scale factors were considered, one five times smaller (scale factor =0.2) and 513
another five times larger (=5.0). The simulation using these adjustments had a strong 514
effect on the cloud organization. For the smaller scale factor, the number of small cells was 515
very high, higher than for the simulation with 1D turbulence. When the scale factor was 516
5.0, the total number of cells (mainly the small ones) was strongly reduced. As these 517
simulations using scale factors presented strong sensitivity, these simulations were used to 518
evaluate the effect on the cloud microphysics. Because the scale factor acts directly on the 519
entrainment (large scale factor permits large eddies that improve the mixing of 520
environmental air inside the clouds) an important marked change in the microphysical 521
variables was expected. Small cloud mixing length produced much more rain, ice and 522
graupel and a small amount of cloud water. As the entrainment was much larger for =5.0 523
these simulations produced more cloud water and less rain and ice species than the 524
simulations using smaller mixing length. However, it is interesting to see that, for the very 525
deep convective cases, i.e. the largest cloud water and rain and ice and graupel integrated 526
values, this trend was reversed for simulations using =5.0. With these simulations using a 527
large entrainment value, the convective events were much rarer but were more severe 528
than those with lower entrainment. This is probably a nonlinear effect which is difficult to 529
analyze. It appears that during the periods of strong inhibitions, only a small fraction of the 530
regions can have convective activity. This can favor the intensity of the deep convection 531
cell because that increases instability and cloud mass flux. 532
533
When we look at the general skill score of the simulations, the simulation with 3D 534
turbulence is much better than that with 1D turbulence and the lower scale factor is closer 535
to one dimension. Finally, the turbulent kinetic energy was examined to check the general 536
effect in each simulation. The 99% percentile profiles show a systematic increase of TKE 537
from the cloud base to the cloud top, from the lower scale factor, to 1D turbulence, 3D 538
turbulence and large scale factor. The value =5.0 shows a very strong TKE and =0.2 a 539
very small TKE. 540
541
The results obtained for a particular cloud system are very consistent with the statistics for 542
those observed by satellite in former studies. At initiation, the cloud system starts to 543
increase TKE close to the cloud base, increasing the TKE in the middle levels in the next 544
few hours as the system produces a strong area expansion. When the cloud system 545
reaches the minimum temperature, it continues to increase in size and the TKE is strongly 546
reduced until the cloud system dissipates. The simulation with 1D turbulence also shows 547
the same behavior, but resulting in a smaller system. Because of the large number of small 548
cells, the system merges with others, producing a less clear evolution. 549
550
The new methodology presented in this study allows us to evaluate the skill of the model 551
to describe the space-time organization of cloud and rainfall. Using this methodology, it 552
was possible to test the effect of different turbulent parameterizations and cloud mixing 553
lengths. The cloud mixing length affects the cloud organization so the adjustment of these 554
parameters can result in a much better description of the cloud and rain fields. Future 555
work will investigate this proposed parameterization in regional models. 556
557
Acknowledgement 558
559
This work was supported by FAPESP grant No. 2009/15235-8, CHUVA Project and CNPQ 560
No. 242659/2012-8 and 573797/2008 (Brazilian National Institute of Science and 561
Technology for Climate Change). This work was carried out in part at the Observatoire 562
Midi-Pyrénées – Laboratoire d´Aérologie. The authors also would like to thank all the 563
participants in CHUVA-SUL field campaign. 564
565
References
Bechtold, P., E. Bazile, F. Guichard, P. Mascart and E. Richard, 2001: A mass flux convection
scheme for regional and global models. Quart. J. Roy. Meteor. Soc., 127, 869-886.
Bougeault, P., and P. Lacarrère, 1989: Parameterization of orography-induced turbulence in
a meso-beta-scale model. Mon. Wea. Rev., 117, 1872-1890.
Caine, S., Todd P. Lane, P. T. May, C. Jakob, S. T. Siems, M. J. Manton, J. Pinto, 2013:
Statistical Assessment of Tropical Convection-Permitting Model Simulations Using a Cell-
Tracking Algorithm. Mon. Wea. Rev., 141, 557–581.
Cecil, D. J. and C. B. Blankenship, 2012: Toward a Global Climatology of Severe Hailstorms
as Estimated by Satellite Passive Microwave Imagers. J. Climate,25, 687–703.
Chaboureau J.-P., and P. Bechtold, 2005: Statistical representation of clouds in a regional
model and the impact on the diurnal cycle of convection during Tropical Convection, Cirrus
and Nitrogen Oxides (TROCCINOX). J. Geophys. Res., 110, D17103.
Chaboureau, J.-P. and J.-P. Pinty, 2006: Validation of a cirrus parameterization with
Meteosat Second Generation observations. Geophys. Res. Let., 33, L03815,
doi:10.1029/2005GL024725
Chaboureau J-P, Söhne N, Pinty JP, Meirold-Mautner I, Defer E, Prigent C, Pardo JR, Mech
M, Crewell S. 2008. A midlatitude cloud database validated with satellite observations. J.
Appl. Meteorol. Climatol. 47:1337-1353.
Chong, M., P. Amayenc, G. Scialom, J. Testud, 1987: A Tropical Squall Line Observed during
the COPT 81 Experiment in West Africa. Part 1: Kinematic Structure Inferred from Dual-
Doppler Radar Data. Mon. Wea. Rev.,115, 670–694.
Cuxart, J., Bougeault, P. and Redelsperger, J.-L., 2000: A turbulence scheme allowing for
mesoscale and large-eddy simulations. Q. J. R. Meteorol. Soc., 126, 1-30.
Emanuel K. A. 1994, Atmospheric convection, Chp8: Theory of mixing in cumulus cloud,
p215
Deardorff, J. W., 1980: Cloud Top Entrainment Instability. J. Atmos. Sci., 37, 131–147.
Fouquart Y and Bonnel B. 1986. Computations of solar heating of the Earth's atmosphere:
A new parametrization. Beitr. Phys. Atmos. 53: 35-62.
Hohenegger, C. and Bretherton, C.S, 2011: Simulating deep convection with a shallow
convection scheme, Atmos. Chem. Phys., 11, 10389-10406.
Honnert R., V. Masson, and F. Couvreux, 2011: A diagnostic for Evaluating the
Representation of Turbulence in Atmospheric Models at the Kilometric Scale. J. Atmos.
Sci.,68(12), 3112-3131
Jensen, Michael P. and A. D. Del Genio, 2006: Factors Limiting Convective Cloud-Top
Height at the ARM Nauru Island Climate Research Facility. J. Climate,19, 2105–2117.
Kain JS, Fritsch JM. 1993. Convective parameterization for mesoscale models: the Kain-
Fritsch Scheme. Meteorol. Monogr. 46: 165-170.
Kidd, C., E. Dawkins, and G. Huffman, 2013: Comparison of precipitation derived from the
ECMWF operational forecast model and satellite precipitation data sets. J. Hydrometeor.
doi:10.1175/JHM-D-12-0182.1, in press.
Lafore, J. P., J. Stein, N. Asencio, P. Bougeault, V. Ducrocq, J. Duron, . Fischer, P. Hereil, P.
Mascart, J. P. Pinty, J. L. Redelsperger, E. Richard, and J. Vila-Guerau de Arellano, 1998: The
Meso-NH Atmospheric Simulation System. Part I: Adiabatic formulation and control
Simulations. Annales Geophysicae, 16, 90-109
Lu, C., Y. Liu, S. Niu, S. Krueger, and T. Wagner, 2013. Exploring parameterization for
turbulent entrainment-mixing processes in clouds, J. Geophys. Res. Atmos., 118, 185–194,
Machado L. A. T., W. B. Rossow, R. L. Guedes, and A. W. Walker, 1998: Life cycle variations
of mesoscale convective systems over the Americas. Mon. Wea. Rev., 126, 1630–1654.
Machado, L. A. T., H. Laurent, and A. A. Lima, 2002. Diurnal march of the convection
observed during TRMM-WETAMC/LBA, J. Geophys. Res., 107(D20), 8064.
Machado, L. A. T., Laurent, H, 2004. The convective system area expansion over Amazonia
and its relationships with convective system life duration and high-level wind divergence.
Monthly Weather Review, v.132, p.714 - 725.
Machado, L. A. T.; Lima, W. F. S.; Pinto Jr, O.; Morales, C. B. 2009. Relationship between
cloud-ground discharge and penetrative clouds: a multi-channel satellite application.
Atmos. Res. v.93, p. 304-309.
Machado, L. A. T. et al, 2013. The CHUVA Project - How different can convection be across
the tropics. Bulletin of the American Meteorological Society – paper proposal accepted, in
submission process.
Mapes, B. E. and R. B. Neale, 2011: Parameterizing convective organization. J. Adv. Model.
Earth Syst.,3, M06004, 20 pp.
Meehl, Gerald A., Francis Zwiers, Jenni Evans, Thomas Knutson, Linda Mearns, Peter
Whetton, 2000: Trends in Extreme Weather and Climate Events: Issues Related to
Modeling Extremes in Projections of Future Climate Change. Bull. Amer. Meteor. Soc.,81,
427–436
Meirold-Mautner, I., C. Prigent, E. Defer, J.-R. Pardo, J.-P. Chaboureau, J.-P. Pinty, M. Mech,
and S. Crewell, 2007. Radiative transfer simulations using mesoscale cloud model outputs:
comparisons with passive microwave and infrared satellite observations for mid-latitudes,
64, 1550-1568, J. Atmos. Sc.
Mlawer EJ, Taubman SJ, Brown PD, Iacono MJ, Clough SA. 1997. Radiative transfer for
inhomogeneous atmospheres: RRTM, a validated corrolated-k model for the longwave. J.
Geophys. Res. 102: D16663-16682.
Morrison, H. and W. W. Grabowski, 2007: Comparison of Bulk and Bin Warm-Rain
Microphysics Models Using a Kinematic Framework. J. Atmos. Sci.,64, 2839–2861.
Negri, R. G., L. A. T. Machado, S. English and M. Forsythe, 2013: Combining cloud resolving
model with satellite for cloud process model simulation validation. J. Appl. Meteor. Clim.,
in revision stage.
Pergaud, J., V. Masson, S. Malardel, and F. Couvreux, 2009: A parameterization of dry
thermals and shallow cumuli for mesoscale numerical weather prediction, Bound.-Layer.
Meteor., 132, 83-106
Pinty, J.-P. and P. Jabouille, 1998: A mixed-phase cloud parameterization for use in
mesoscale non-hydrostatic model: simulations of a squall line and of orographic
precipitations. Proc. Conf. of Cloud Physics, Everett, WA, USA, Amer. Meteor. soc., Aug.
1999, 217 – 220.
Richard E, Cosma S, Tabary P, Pinty JP, Hagen M. 2003. High-resolution numerical
simulations of the convective system observed in the Lago Maggiore area on 17
September 1999 (MAP IOP 2a). Q. J. R. Meteorol. Soc.129: 543-563.
Ringer,M. A., Edwards,J. M., and Slingo,A., 2003: Simulation of satellite channel 18
radiances in the Met Office Unified model. Q. J. R. Meteorol. Soc., 129, 1169-1190.19
Rosenfeld, D., W. L. Woodley, A. Lerner, G. Kelman, and D. T. Lindsey 2008, Satellite
detection of severe convective storms by their retrieved vertical profiles of cloud particle
effective radius and thermodynamic phase, J. Geophys. Res., 113, D04208,
Salio, P., M. Nicolini, E. J. Zipser, 2007: Mesoscale Convective Systems over Southeastern
South America and Their Relationship with the South American Low-Level Jet.Mon. Wea.
Rev.,135, 1290–1309.
Saunders, R., Matricardi, M., Brunel, P., English, S., Bauer, P., O’Keeffe, U., Francis, P., and
Rayer, P.: RTTOV-8 Science and validation report, Tech. rep., NWP SAF Rep., 41 pp., 2005
Shrestha,D. L., Robertson,D.E., Wang,Q.J., Pagano,T.C., and Hapuarachchi,H.A.P.
2013:Evaluation of numerical weather prediction model precipitation forecasts for short-
term streamflow forecasting purpose, Hydrol. Earth Syst. Sci.
Simpson, J., 1971: On Cumulus Entrainment and One-Dimensional Models. J. Atmos.
Sci.,28, 449–455.
Söhne, N., J.-P. Chaboureau, and F. Guichard, 2008, Verification of cloud cover forecast with
satellite observation over West Africa, Mon.Weather Rev., 136, 4421-4434
Stein J., E. Richard, J.P. Lafore, J.P. Pinty, N. Asencio and S. Cosma, 2000: High -resolution
non-hydrostatic simulations of flash-flood episodes with grid-nesting and ice-phase
parametrization. Meteorol. Atmos. Phys., 72, 101-110.
Varble, A., A. M. Fridlind, E. J. Zipser, A. S. Ackerman, J.-P. Chaboureau,J. Fan, A. Hill, S. A.
McFarlane, J.-P. Pinty, and B. Shipway 2011, Evaluation of cloud-resolving model
intercomparison simulations using TWP-ICE observations: Precipitation and cloud
structure, J. Geophys. Res., 116, D12206.
Vila, D. A., L. A. T. Machado, H. Laurent and I. Velasco, 2008: Forecast and Tracking the
Evolution of Cloud Clusters (ForTraCC) Using Satellite Infrared Imagery: Methodology and
Validation. Wea. Forecasting,23, 233–245.
Wang, Y., L. Zhou, and K. P. Hamilton, 2007: Effect of convective entrainment/detrainment
on simulation of tropical precipitation diurnal cycle. Mon. Wea. Rev., 135, 567–585.
Wu, J., A. D. Del Genio, M.-S. Yao, and A. B. Wolf, 2009. WRF and GISS SCM simulations of
convective updraft properties during TWP-ICE, J. Geophys. Res., 114, D04206.
FIGURE CAPTIONS
FIG. 1. MSG brightness temperature at 10.8 m on 1 December 2012 at 2000 UTC for (a)
observation and (b) simulation at forecast range of 32 h. Figure b) also shows the radar
coverage (circle line).
FIG. 2. Tir histogram of the 36 images of the 6 golden days obtained from the observations
and the simulations using the 1D turbulence scheme.
FIG. 3. Organization of clouds (Tir < 235 K) observed by MSG and simulated by Meso-NH for
the 6 golden days (during the 36-h period from 1200 UTC of the specific day). (a) Size
distribution and (b) life cycle duration.
FIG. 4. Tir histogram of cloud cells (Tir < 235 K) with effective radius (a) smaller than 50 km
and (b) larger than 150 km observed by MSG and simulated by Meso-NH for the 6 golden
days.
FIG. 5. Cloud organization from observation and Meso-NH simulations with 1D and 3D
turbulence for Julian days 333 and 335 (during the 36-h period from 1200 UTC of the
specific day). (a) Size distribution of cloud (Tir < 235 K) observed by MSG and simulated by
Meso-NH. (b) Size distribution of rain cells (reflectivity larger than 20 dBZ) observed by the
S-band radar and simulated by Meso-NH.
FIG. 6. Reflectivity from S-band radar and Meso-NH simulations with 1D and 3D turbulence
for Julian days 333 and 335 (during the 36-h period from 1200 UTC of the specific day).
Histograms of (a) reflectivity at 2-km altitude and (b) echo cloud top height using 0 dBZ
threshold.
FIG. 7. Organization of clouds (Tir < 235K) observed by MSG and simulated by Meso-NH
with 1D and 3D turbulence and for mixing length multiplied by 0.2 and 5.0, for Julian day
335 (for the 36-h period from 1200 UTC of the specific day).
FIG. 8. Microphysical variables simulated by Meso-NH for simulation with 3D turbulence for
mixing length multiplied by 0.2 and 5.0, during the 36-h period from 1200 UTC of Julian
day 335. Column integrated (a) cloud water and ice, (b) rain water and graupel.
FIG. 9. Equitable Threat Score (ETS), for the occurrence of high clouds using a brightness
temperature threshold of 260 K, simulated by Meso-NH for with 1D and 3D turbulence and
for mixing length multiplied by 0.2 and 5.0, during the 36-h period from 1200 UTC of Julian
day 335.
FIG. 10. 99th percentile profiles of turbulent kinetic energy, for cloud cover events,
simulated by Meso-NH with 1D and 3D turbulence and for mixing length multiplied by 0.2
and 5.0, during the 36-h period from 1200 UTC of Julian day 335.
FIG. 11. Profiles of turbulent kinetic energy simulated with (a) 1D and (b) 3D turbulence
every hour in the life cycle of a convective system and (c) time evolution of its effective
radius and Tir minimum.
566 567
FIG. 1. MSG brightness temperature at 10.8 m on 1 December 2012 at 2000 UTC for 568
(a) observation and (b) simulation at forecast range of 32 h. Figure b) also shows the 569
radar coverage (circle line). 570
571
572
573
574
575 FIG. 2. Tir histogram of the 36 images of the 6 golden days obtained from the 576
observations and the simulations using the 1D turbulence scheme. 577
578
579
580
581
582
583
584 585
FIG. 3. Organization of clouds (Tir < 235 K) observed by MSG and simulated by Meso-NH 586
for the 6 golden days (during the 36-h period from 1200 UTC of the specific day). (a) 587
Size distribution and (b) life cycle duration. 588
589
590
591
592
593 594
595
596
FIG. 4. Tir histogram of cloud cells (Tir < 235 K) with effective radius (a) smaller than 50 597
km and (b) larger than 150 km observed by MSG and simulated by Meso-NH for the 6 598
golden days. 599
600
601
602
603
604 605
606
FIG. 5. Cloud organization from observation and Meso-NH simulations with 1D and 3D 607
turbulence for Julian days 333 and 335 (during the 36-h period from 1200 UTC of the 608
specific day). (a) Size distribution of cloud (Tir < 235 K) observed by MSG and 609
simulated by Meso-NH. (b) Size distribution of rain cells (reflectivity larger than 20 610
dBZ) observed by the S-band radar and simulated by Meso-NH. 611
612
613
614
615
616 617
618
FIG. 6. Reflectivity from S-band radar and Meso-NH simulations with 1D and 3D 619
turbulence for Julian days 333 and 335 (during the 36-h period from 1200 UTC of the 620
specific day). Histograms of (a) reflectivity at 2-km altitude and (b) echo cloud top 621
height using 0 dBZ threshold. 622
623
624
625
626 627
628
FIG. 7. Organization of clouds (Tir < 235K) observed by MSG and simulated by Meso-NH 629
with 1D and 3D turbulence and for mixing length multiplied by 0.2 and 5.0, for Julian 630
day 335 (for 36-h period from 1200 UTC of the specific day). 631
632
633
634
635 636
637
638
FIG. 8. Microphysical variables simulated by Meso-NH for simulation with 3D turbulence 639
for mixing length multiplied by 0.2 and 5.0, during the 36-h period from 1200 UTC of 640
Julian day 335. Column integrated (a) cloud water and ice, (b) rain water and graupel. 641
642
643
644
645 646
647
FIG. 9. Equitable Threat Score (ETS) simulated by Meso-NH for with 1D and 3D 648
turbulence and for mixing length multiplied by 0.2 and 5.0, during the 36-h period from 649
1200 UTC of Julian day 335. 650
651
652
653
654
655 656
657
FIG. 10. 99th percentile profiles of turbulent kinetic energy, for cloud cover events, 658
simulated by Meso-NH with 1D and 3D turbulence and for mixing length multiplied by 659
0.2 and 5.0, for the 36-h period from 1200 UTC of Julian day 335. 660
661
662
663 664
665
FIG. 11. Profiles of turbulent kinetic energy simulated with (a) 1D and (b) 3D turbulence 666
every hour in the life cycle of a convective system and (c) time evolution of its effective 667
radius and Tir minimum. 668
669