You are on page 1of 9

ISSN 1024-8560, Atmospheric and Oceanic Optics, 2019, Vol. 32, No. 2, pp. 193–201. © Pleiades Publishing, Ltd.

, 2019.
Russian Text © G.A. Zarochentsev, K.G. Rubinstein, V.I. Bychkova, R.Yu. Ignatov, Yu.I. Yusupov, 2018, published in Optika Atmosfery i Okeana.

OPTICAL MODELS
AND DATABASES

Comparison of Several Numerical Methods for Fog Prediction


G. A. Zarochentseva, b, c, *, K. G. Rubinsteinb, c, V. I. Bychkovab, c, R. Yu. Ignatovb, c, and Yu. I. Yusupovb
aMoscow
State University, Moscow, 119991 Russia
b
Hydrometeorological Centre of Russia, Moscow, 123242 Russia
c
The Nuclear Safety Institute, Russian Academy of Sciences, Moscow, 115191 Russia
*e-mail: xztv@mail.ru
Received March 12, 2018; revised September 17, 2018; accepted September 25, 2018

Abstract—Several methods for visibility calculation for fog forecasting are discussed, including a method sug-
gested by the authors. We use the WRF-ARW model to obtain the necessary meteorological information. The
forecasts are estimated using data with a high spatial resolution from European Synoptic stations. The analysis
of the methods shows a generally satisfactory quality of forecasts for this phenomenon.

Keywords: fog, meteorological visibility, mesoscale modeling, transfer of moisture in the surface layer
DOI: 10.1134/S1024856019020180

In the general sense, fog is a dispersion medium Los Angeles Airport has been observed even at a rela-
formed by water droplets or small ice crystals in a cer- tive humidity of about 90% [7].
tain concentration, which reduces visibility to 1 km or It is important to keep in mind that characteristics of
less near the Earth’s surface [1]. Fog causes serious the wind regime in the surface air layer are also import-
problems at airports and on highways. A reliable fore- ant for fog formation. Under moderate or strong winds,
cast of this weather phenomenon can contribute to fog is formed only if the surface layer is cooled suffi-
public safety. ciently rapidly (temperature gradient exceeds 2°C/km),
In this work, the visibility range means the range at since the turbulent transfer of heat and moisture is
which a black object with angular dimensions of no less intensified and the fog dissipates at such wind speeds.
than 0.3° is seen against the sky in daylight hours [1]. Radiation fogs are most sensitive to the wind regime of
the surface layer; they are observable at characteristic
Fog is formed due to two processes [2]: an increase wind speeds of no more than 1 m/s, while advective
in the moisture content in air and a decrease in the fogs, according to measurements [8], have been
surface air temperature. Cooling fogs, in particular observed even under a surface wind speed of 6 m/s.
advective fogs, which are formed when warm air cools Today, synoptic forecasts of a favorable situation
during the movement along a cooler surface, are most for the development of radiation fogs are based on the
prevalent. Advective fogs have no characteristic spatial analysis of the presence and movement of anticyclone
or temporal localization and are observed mainly in centers and axes of the ridges, and advective fogs, on
warm sectors of cyclones [3]. the forecast of movement of warm atmospheric masses
(warm fronts). Soil moisture is an important parame-
Water vapor usually starts condensing when the air ter which affects fog formation. For example, due to
is saturated with moisture and the relative humidity is precipitation or floods, soil moisture may become an
close to 100%. However, as was ascertained in [4], in additional source of moisture for advection.
the absence of foreign particles in air (condensation
nuclei), complexes of molecules in the condensed There are also numerical methods for fog forecast-
phase can originate only as a result of random colli- ing. For radiation fogs, the forecasting methods by
sions of molecules in supersaturated air. The real A.S. Zverev [3] and N.V. Petrenko [9] are widely used
atmosphere always contains different impurities (par- at aeronautical weather services. The method of
ticles of dust, sand, salt). It is shown in [5, 6] that the Zverev is based on the forecast of the fog formation
condensation nuclei in the atmosphere are sufficiently temperature Tt, which can be represented as
active and numerous to induce moisture condensation Tt = Td − ΔTd − ΔTd', (1)
and the beginning of the process of fog formation
when the air has not attained saturation. Polluted air, where Td is the dew point; ΔTd is the possible decrease
for example, in urban conditions, is often more favor- in the dew point (additional cooling) for the forecast
able for the formation of fogs. For example, fog at the time; ΔTd' is the required change in the dew point for

193
194 ZAROCHENTSEV et al.

the formation of fog with visibility range of less than hensive and quite universal method which allows
1 km (is 0–1°C in the positive and >2–3°C in the neg- numerical fog forecasting with good prognostic char-
ative temperature range). acteristics.
The method of Petrenko [9] makes it possible to Below, if visibility less than 1 km is forecasted, we
specify the temperature of fog formation considering consider this as the fog forecast. Four methods for vis-
the vertical gradient of the dew point in the 0–300 m ibility range forecasting are compared, as well as the
layer, but the specification does not go beyond 1°C in method we suggest.
most cases. Petrenko also developed a method for Method I. The most physically substantiated
forecasting advective fogs depending on the horizontal method. It has been developed by researchers of the
temperature gradient along a nine-hour trajectory of National Center for Atmospheric Research (NCAR),
air particles, the dew point deficit at the beginning of United States, [16] and is a research component of the
this trajectory, and the expected wind speed at the postprocessing of global weather forecasts. It is based
forecast point (at the weather vane level) for on the fact that the luminance of an object, which
Sheremetyevo Airport. L.A. Klyuchnikova presents reaches the observer position at an altitude of 2 m, is
the method for forecasting advective fogs in the coastal the sum of the object luminance weakened by the
zone 25–100 km wide depending on the water–land Bouguer–Lambert–Beer law during propagation
temperature difference [10]. There are also other through an absorbing medium, hydrometeors (water
semiempirical calculation methods for fog forecasting droplets, snowflakes, and ice crystals) in our case, and
developed for specific geographic regions accounting the brightness of the medium layer between the object
for the topography and meteorological features, for and the observation point. For a horizontally homoge-
example [3, 11]. neous medium,
Many semiempirical formulas allow calculating the
visibility range in fog from its water content [3, 12, 13]. I (L) = I 0 exp(−β L) + I b,∞[1 − exp(−β L)], (3)
Zverev [3] suggested the definition of the horizontal I0 is the luminance of the object, and I b,∞ is the back-
visibility range L (m) in terms of the condensed water
and the size of suspended particles based on measure- ground luminance I b as L → ∞.
ment data: The ratio K (L) = [I b − I (L)] I b is called the illu-
minance contrast; the visibility range is the distance
L = 2.3 × 104 r (2) L = Lth at which the illuminance contrast attains the
δm
threshold value K (Lth ) = ε th (ε th = 0.05 at present).
(r is the droplet radius, cm; δm is the condensed water, The sum of power functions of mass concentrations of
g/m3). hydrometeors (water droplets Сcw, rain droplets Сrw,
However, measurement of the radius of droplets ice crystals Сci, and snow crystals Сsn, g/m3) is used as
and the condensed water is a complex problem. At the extinction coefficient β. The weight factors have
meteorological stations, these parameters are not been selected according to [16–20]:
measured; therefore, calculations by this method are
difficult. According to Mason [14], there is no satis- − ln ε p
factory method for measuring water content which L= ,
meets the following requirements: the capability of β (4)
measuring both in the supercooled and non-super- β = 144.7Ccw
0.88
+ 1.1Crw + 163.9Cci + 10.4Csn .
0.75 0.78

cooled state, low inertia, and representativeness. Method II is a semiempirical method developed by
Attempts to measure the condensed water were made researchers of the National Center for Environmental
by Houghton and Radford [13]; they obtained values Prediction (NCEP), United States, and used as a
from 0.1 to 0.22 g/m3. component of visibility forecast in the RUC (Rapid
Thus, most existing methods for fog forecasting Update Cycle) model. The method uses the value of
either have a precise gridding or reduce the fog fore- relative humidity at an altitude of 2 m (RH, %) as a
cast to the equally difficult problem of forecasting the predictor. A decrease in the visibility with an increase
absolute content of condensed water in the surface air in the surface humidity in accordance with [21] is
layer. In [15], fog forecast methods are compared only taken into account via the equation

( )
locally at several points; the comparison results make
it difficult to judge the degree of universality of these L = 60 exp −2.5 min 0.8; RH − 0.15  . (5)
methods.  100 
In this work, we compare the forecasts of the hori- Method III is a semiempirical method developed
zontal visibility range as a function of the concentra- by researchers of the Hungarian Meteorological Ser-
tion of hydrometeors, surface humidity, wind speed, vice (OMSZ) [22] and used to forecast fogs at Buda-
and temperature. The main tasks are to estimate the pest airport. The vertical temperature gradient in the
potential of the methods, to analyze their advantages boundary layer (i.e., a measure of atmospheric stabil-
and disadvantages, and, finally, to suggest a compre- ity), the dew point deficit near the surface, and the

ATMOSPHERIC AND OCEANIC OPTICS Vol. 32 No. 2 2019


COMPARISON OF SEVERAL NUMERICAL METHODS FOR FOG PREDICTION 195

wind speed at the top of the boundary layer are used in Eqs. (5) and (6) so that the visibility function is less
the method as predictors: than 1 km for characteristic values of the air humidity,
dew-point deficit, temperature gradient, and wind
L = −1.33 + 0.9 ( Tsf − T850 + (Tsf − Tdsf ) + W850 ) , (6) speed recorded and forecasted at points where fog was
where Tsf and T850 are the temperatures at an altitude observed:

( )
of 2 m and on the 850-hPa isobaric surface, respec-
tively, K; Tdsf is the dew point at an altitude of 2 m, K; L = 60 exp −5.31 RH − 0.15  , (9)
W850 is the wind speed on the 850-hPa isobaric sur-  100 
face, m/s. L = −0.1596 + 0.108
(10)
Меthod IV is a semiempirical method developed by × [ Tsf − T850 + (Tsf − Tdsf ) + W850 ] .
a research team of the US National Oceanic and
Atmospheric Administration (NOAA) and used to Calculations with Eqs. (4) and (7)–(10) require
forecast visibility in aircraft flight zones [23]. As pre- knowledge of such quantities as the relative humidity,
dictors, the relative humidity at an altitude of 2 m and wind speed, dew point, and moisture content in the
the difference between the temperature and dew point lower air layer. To forecast the values of these quanti-
at this level are used: ties, the hydrodynamic mesoscale WRF-ARW model
[24] was used; the computational domain included
Tsf − Tdsf
L = 9656 1.75
. (7) Europe and European Russia; the spatial step of the
RH model was 18 km. The forecasted data were recorded
Меthod V is a semiempirical method we developed every three hours, which corresponded to the time
based on the analysis of data on fogs from 12 meteoro- resolution of observations. To describe the parame-
logical stations of the Moscow Region network for ters of underlying surface of the territory under study,
2000–2013. The number of fog events studied was the MODIS database with a resolution of 0.5° was
4847. The analysis made it possible to identify two pre- used. The fields of the global GFS model with a spa-
dictors of fog formation: high relative humidity of the tial resolution of 0.25° were used as input data for the
surface air at an altitude of 2 m and a low wind speed WRF-ARW model.
at an altitude of 10 m. The method for calculating the Previous work [25, 26] on optimizing the WRF-ARW
visibility range is based on the use of the discriminant model for the most successful forecasting of meteoro-
function of two arguments, relative humidity and wind logical processes in the surface air layer (turbulent
speed. The function tends to zero at a humidity of conditions, heat and moisture exchange of the lower
more than 97.5% and wind speed from 0 to 3.5 m/s: air layer with the surface, and formation and fallout of

( )
precipitation) allowed us to determine the following
arctan(RH − 99.5)
L = 5.5 1 − configuration of the model. To describe the processes
2 π (8) associated with convection, the Grella scheme [27] is
 1 arctan(W10 − 3.5)  used, which is an ensemble of parametrizations, where
× + ,
2 π  different external parameters are used to calculate
convection. Microphysical processes in clouds are cal-
where RH is the relative humidity at 2 m, %; W10 is the
culated according to [28], the processes occurring in
wind speed at the altitude of the weather vane, m/s.
the planetary boundary layer, according to [29]. The
To analyze the quality of the fog forecast, the hori- analysis of the quality of calculation of turbulent heat
zontal visibility range was calculated using the five and moisture fluxes [25] showed that the method for
abovementioned methods. Data from synoptic sta- calculating processes in the boundary layer [29]
tions of all European countries were used as standards, ensures a high quality forecast of turbulence parame-
since the European synoptic network has a high spatial ters. To calculate the shortwave and longwave radia-
resolution, as well as good quality of archive data. The tion, the scheme described in [30] is used. The
following periods were chosen for estimation: NOAH-LSM model of the land surface [31] is used for
9:00 August 20–0:00 August 22, 2015; parameterization of soil processes. It allows account-
9:00 October 24–0:00 October 26, 2015; ing for a mosaic snow cover and snow aging (the snow
9:00 December 19–0:00 December 21, 2015; density is represented as a function of time). The
mosaic structure of snow is taken into account in the
9:00 February 5–0:00 February 7, 2017. NOAH parameterization during calculation of the
During those periods, 6836 fog events and albedo and thermal conductivity; the corresponding
141776 episodes without fog were observed on the ter- values are taken with the weight of the share of snow-
ritory under study. covered surface and the share of green vegetation.
The preliminary analysis of forecasts by Eqs. (4)–(8) A possible numerical description of vegetation makes
has shown that the visibility function in fog exceeds a it possible to characterize the moisture regime of the
threshold of 1 km for methods II and III. Therefore, to surface layer with higher accuracy. The heat exchange
universalize these methods without changing their between the snow cover and the atmosphere and sub-
physical basis, we have changed the coefficients in limation and melting at the upper and lower snow

ATMOSPHERIC AND OCEANIC OPTICS Vol. 32 No. 2 2019


196 ZAROCHENTSEV et al.

boundaries are calculated based on the surface balance quality of the forecast of the relative humidity in the
equation. A qualitative description of the processes of surface air layer has been assessed.
heat and moisture exchange with the underlying sur- Figure 2 shows the difference in the relative humid-
face is extremely important in fog forecasts, since the ity at an altitude of 2 m according to the WRF-ARW
surface is a source of additional moisture. model data (RHf) and according to observation data at
Four forecasts were calculated using the WRF- meteorological stations (RHo). It is seen in Fig. 2a that
ARW model for the above time intervals for 48 hours, the difference RHf – RHo does not exceed 10% on
the first nine of which were not taken into account to average over the entire region. In Fig. 2b, the fore-
data adjustment (spin up—the time required for stabi- casted humidity deficit attained 40% in the Balkans
lization of hydrodynamic mesoscale WRF-ARW and northern Italy. The moisture error is systematic on
model results). the second day of the forecast, and the absolute value
During the preliminary analysis of observation of RHf – RHo is 18% on average. This large values
data, it has been ascertained that most meteorological might well be due to incorrect initial values in the
stations do not transmit the horizontal visibility range, moisture fields. Such large errors can strongly affect
unlike the weather code at the observation time, which the result of visibility calculations by all the methods.
is sent by most meteorological stations of the Euro- Hence, we can conclude that the forecast of the visi-
pean Synoptic stations. Therefore, to assess the fore- bility range on the second day by all the methods is
cast quality, we used weather data at the observation incorrect, as is seen in region F1 in Fig. 1b, due to the
time: if fog was observed at a meteorological station error in the humidity forecast (Fig. 2b).
and the visibility range was forecasted to be less than For both forecasts, the total success rate, the success
1 km (calculated at the station point using linear inter- rates of the forecast of presence and absence of fogs, the
polation), the forecast was considered successful, oth- warning and the percentage of false alarms were calcu-
erwise unsuccessful. lated (Table 1). Despite different approaches to selec-
Figure 1 shows the visibility range calculated by all tion of predictors of fog formation, the analysis of the
five fog forecasting methods and fog observation data forecast quality shows close estimates of the forecast
at meteorological stations. It is seen that the European success rate for all the methods under study for a period
Synoptic stations are quite dense on the territory under of 48 hours. A high percentage of false alarms points out
study, which, on the one hand, improves its data cov- that a model with this spatial resolution can produce
erage, and, on the other hand, can contribute to false only an averaged value if the observation points are
alarms when forecasting fogs. within the same calculation cell. However, all the meth-
Calculations by all the methods provided the fore- ods show high success rates of the forecast of fog
cast of higher quality for February 6, 2017, than for absence, since fog is a rare phenomenon and is
October 25, 2015. Methods I–IV (Fig. 1a) forecasted observed in only 4.8% of cases for the time intervals
false alarms for the Scandinavian Peninsula and selected.
Transcaucasia; method V (Fig. 1a) gave the smallest Table 1 shows the estimates of the fog forecast
number of false alarms. quality at the time of observation at the stations; if the
Several characteristic regions have been distin- fog was forecasted with a delay or in advance, as com-
guished for both dates in Fig. 1: region F, with preva- pared to observations, this was considered a false
lence of fog, according to data from meteorological sta- alarm. Therefore, Fig. 3 shows another assessment of
tions, and region N, where one can state the absence of the fog forecast quality, which takes into account its
fog with reasonable certainty. For February 6, 2017, possible phase shift relative to the observation data.
methods II–IV forecasted fogs not only in region F, but The relative duration of fogs is the ratio of the fore-
also quite far from it, while forecasts by methods I and casted duration of the phenomenon to its duration
V were closer to observations. The smallest number according to observations. Coincidence of forecasted
of false alarms for region N was also produced by duration of the phenomenon and its duration accord-
methods I and V. ing to observations at meteorological stations is shown
by the green color. If the weather code is not indicated
Regions F and N are divided into two parts in Fig. 1b. in the data from a weather station at the time of obser-
For region F2, fog is the most successfully forecasted by vation, the absence of fog is accepted for this the sta-
methods II, IV, and V; their forecasts quite accurately tion. If fog was not forecasted for the period under
coincide with observations. For region F1, method V did study and was not observed at a meteorological sta-
not allow forecasting fog at all. For region N1, the small- tion, we believe that a method correctly forecasts the
est number of false alarms was produced by methods III duration of the phenomenon, or, more precise, its
and V, and for region N2, by method V (see Fig. 1b). absence. For methods I–IV (Fig. 3a), the largest error
Therefore, it is difficult to say which method forecasts in fog duration corresponds to the Scandinavian Pen-
the most successfully in this case. insula and the Transcaucasia (see Fig. 1). Note that
To determine the cause of the different success of the error of method V (Fig. 3a) is also the smallest in
the methods for the above two cases (see Fig. 1), the this case. It follows from the estimate of the relative

ATMOSPHERIC AND OCEANIC OPTICS Vol. 32 No. 2 2019


COMPARISON OF SEVERAL NUMERICAL METHODS FOR FOG PREDICTION 197

(a) Method I (b)


63 N
N F2
57 N1
N2
51
F
45
F1
39

Method II
63
N
F2
57 N1
N2
51
F
45
F1
39

Method III
63 N
F2
57 N1
N2
51
F
45
F1
39

Method IV
63 N
F2
57 N1
N2
51
F
45
F1
39

Method V
63 N
F2
57 N1
N2
51
F
45
F1
39

0 5 10 15 20 25 30 35 40 45 50 55 60 65 0 5 10 15 20 25 30 35 40 45 50 55 60 65 E
Fig. 1. Maps of the visibility range calculated by five methods (dashed areas) for fog forecast and fog observation data at meteo-
rological stations (red dots correspond to fog events, black dots, to absence of fog) for (a) 06:00 on February 6, 2017, and (b) 06:00
on October 25, 2015.

ATMOSPHERIC AND OCEANIC OPTICS Vol. 32 No. 2 2019


198 ZAROCHENTSEV et al.

(a) (b)
RHf – RHo RHf – RHo
N N
60° 60°

50° 50°

40° 40°

0° 15° 30° 45° 60° E 0° 15° 30° 45° 60° E

–40 –20 0 20 40 % –40 –20 0 20 40 %


Fig. 2. Maps of the difference between the values of relative humidity at an altitude of 2 m calculated from the WRF-ARW model
data (RHf) and observations at weather stations (RHo) at (a) 06:00 on February 6, 2017, and (b) 06:00 on October 25, 2015.

Table 1. Assessment of the fog forecast success


Method Total success rate, % Fog presence success rate, % Warning, % Fog absence success rate, % False alarm, %
Region F (February 5–7, 2017)
I 34 17 76 86 83
II 59 22 67 90 78
III 58 17 43 84 83
IV 51 21 75 90 79
V 72 17 17 84 83
Region N (February 5–7, 2017)
I 62 13 69 99 87
II 89 9 72 99 91
III 89 6 47 99 93
IV 86 7 73 99 93
V 94 8 33 92 91
Region F1 (October 24–26, 2015)
I 64 26 32 78 74
II 72 30 10 76 70
III 65 16 18 76 84
IV 70 33 15 76 67
V 62 24 23 75 76
Region F2 (October 24–26, 2015)
I 38 8 47 87 92
II 57 12 49 93 88
III 71 15 47 94 85
IV 49 12 65 95 88
V 56 13 63 93 87
Region N1 (October 24–26, 2015)
I 53 2 32 96 97
II 79 4 34 97 95
III 85 6 28 97 94
IV 72 5 44 97 95
V 88 11 30 97 89
Region N2 (October 24–26, 2015)
I 49 3 68 98 97
II 57 4 55 98 96
III 64 3 40 98 96
IV 49 4 64 98 96
V 82 6 30 98 94

ATMOSPHERIC AND OCEANIC OPTICS Vol. 32 No. 2 2019


COMPARISON OF SEVERAL NUMERICAL METHODS FOR FOG PREDICTION 199

(a) Method I (b)


N N
60 60

50 50

40 40

Method II

60 60

50 50

40 40

Method III

60 60

50 50

40 40

Method IV

60 60

50 50

40 40

Method V

60 60

50 50

40 40

0 15 30 45 60 E 0 15 30 45 60 E

0 0.4 0.8 2.0 6.0 10.0 0 0.4 0.8 2.0 6.0 10.0

Fig. 3. Maps of the ratio of fog duration by all the five methods to the fog duration according to weather station observations on
(a) February 5–7, 2017, and (b) October 24–26, 2015.

ATMOSPHERIC AND OCEANIC OPTICS Vol. 32 No. 2 2019


200 ZAROCHENTSEV et al.

duration of fogs in Fig. 3b that methods III and V 9. N. V. Petrenko, “Improvement of the technique for
showed the smallest errors, and methods I and IV, the forecasting advective fog and visibility in this fog,” Tr.
largest. It can also be seen that the regions with the Gidromet. SSSR, Is. 162, 34–45 (1975).
maximal errors in fog duration coincide with regions F 10. L. A. Klyuchnikova, “About the advective fog forma-
and N shown in Fig. 1. tion,” Tr. GGO, No. 60, 122 (1956).
In general, the analysis of estimates of the fog fore- 11. Z. E. Babenko, Avtoref. Candidate’s Dissertation in
cast quality shows the greatest fog forecast success by Geography (Middle Asian Regional Research Institute
methods I, III, IV, and V. named after V.A. Bugaev, Tashkent, 1961).
12. H. Koschmieder, “Measurements of visibility at dan-
Let us sum up. The analysis of five methods for zig,” Mon. Weather. Rev. 58 (11), 439–444 (1930).
calculation of the horizontal visibility range for fog
forecasting showed a generally satisfactory quality of 13. H. G. Houghton and W. H. Radford, On the Measure-
ment of Drop Size and Liquid Water Content in Fogs and
forecasts with the use of the WRF-ARW model, pri-
Clouds (Massachusetts Institute of Technology and
marily in the lowland regions. The best results were Woods Hole Oceanographic Institution, Massachu-
shown by methods I and V. The fog duration was best setts, 1938).
forecasted by methods III and V. The total success
rate is the best for methods III and V, and warning, 14. D. B. Meison, G. T. Nikanorova, and V. S. Protopop-
ova, Physics of Clouds (Gidrometeoizdat, Leningrad,
for methods I and IV. 1961) [in Russian].
The forecast by method I, based on the concentra- 15. C. H. Bang, J. W. Lee, and S. Y. Hong, “Predictability
tion of hydrometeors in the surface air layer, and by experiments of fog and visibility in local airports over
method V, which takes into account not only the air Korea using the WRF model,” J. KOSAE 24 (E2), 92–
humidity, but also the wind regime in the atmospheric 101 (2008).
boundary layer, can be considered the most success- 16. M. T. Stoelinga and T. T. Warner, “Nonhydrostatic,
ful. In the future, we intend to supplement methods I mesobeta-scale model simulations of cloud ceiling and
and V with antipredictors to prevent false alarms and visibility for an East Coast winter precipitation event,”
to improve the accuracy of the fog forecast. J. Appl. Meteorol. 38 (4), 385–404 (1999).
17. B. A. Kunkel, “Parameterization of droplet terminal
velocity and extinction coefficient in fog models,”
ACKNOWLEDGMENTS J. Clim. Appl. Meteorol. 23 (1), 34–41 (1984).
The work was partly supported by the Russian 18. S. A. Rutledge and P. Hobbs, “The mesoscale and
Foundation for Basic Research (grant nos. 16-05- microscale structure and organization of clouds and
00704, 18-35-00044 mol_a, and 16-05-00822 a). precipitation in midlatitude cyclones. VIII: A model for
the "seeder-feeder” process in warm-frontal rain-
bands," J. Atmos. Sci. 40 (5), 1185–1206 (1983).
REFERENCES 19. J. R. Stallabrass, “Snow property measurement work-
1. S. P. Khromov and L. I. Mamontova, Meteorological shop,” in Proc. National Research Council Associate
Dictionary (Gidrometeoizdat, Leningrad, 1974) [in Committee on Geotechnical Research. Tech. memoran-
Russian]. dum (Canada, 1985), no. 140, p. 389–410
20. J. S. Marshall and W. M. Palmer, “The distribution of
2. L. T. Matveev, Atmospheric Physics (Gidrometeoizdat,
raindrops with size,” J. Meteorol. 5, 165–166 (1948).
Leningrad, 1965) [in Russian].
21. P. E. Bieringer, M. Donovan, F. Robasky, D. A. Clark,
3. A. S. Zverev, Synoptical Meteorology (Gidrometeoizdat, and J. Hurst, “A characterization of NWP ceiling and
Leningrad, 1977) [in Russian]. visibility forecasts for the terminal airspace,” in
4. C. T. R. Wilson and J. J. Thomson, “Condensation of 12th Conf. Aviation, Range, and Aerospace Meteorol-
water vapour in the presence of dust-free air and other ogy, Atlanta, GA, 2006.
gases,” Proc. Roy. Soc., London 61 (369–377), 240– 22. F. Wantuch, “Visibility and fog forecasting based on
242 (1897). decision tree method,” Idojárás 105, 29–38 (2001).
5. S. Petersen, The Weather Analysis and Forecasting 23. J. A. Doran, P. J. Roohr, D. J. Beberwyk, G. R. Brooks,
(Gidrometeoizdat, Leningrad, 1961) [in Russian]. G. A. Gayno, R. T. Williams, J. M. Lewis, and
6. M. A. Kohler and M. M. Richards, “Multicapacity R. J. Lefevre, “The MM5 at the Air Force Weather
basin accounting for predicting runoff from storm pre- Agency—New products to support military opera-
cipitation,” J. Geophys. Res. 67 (13), 5187–5197 tions,” in 8th Conf. Aviation, Range, and Aerospace
(1962). Meteorology, Dallas, Texas, 1999.
24. W. C. Skamarock, J. B. Klemp, J. Dudhia, D. O. Gill,
7. M. Neiburger and M. G. Wurtele, “On the nature and D. M. Barker, W. Wang, and J. G. Powers, A Description
size of particles in haze, fog, and stratus of the Los of the Advanced Research WRF Version 2. (Mesoscale
Angeles region,” Chem. Rev. 44 (2), 321–335 (1949). and Microscale Meteorology Division, National Cen-
8. J. J. George, Weather Forecasting for Aeronautics (Aca- ter for Atmospheric Research, Boulder, Colorado,
demic Press, London, 1960). USA, 2005). doi 10.5065/D68S4MVH

ATMOSPHERIC AND OCEANIC OPTICS Vol. 32 No. 2 2019


COMPARISON OF SEVERAL NUMERICAL METHODS FOR FOG PREDICTION 201

25. V. I. Bychkova, R. Yu. Ignatov, and K. G. Rubinshtein, 29. P. Bougeault and P. Lacarrere, “Parameterization of
“The analysis of thermal fluxes on surface from WRF- orography-induced turbulence in a mesobeta-scale
ARW model calculations in polar regions,” Uch. Zap. model,” Mon. Weather. Rev. 117 (8), 1872–1890
Ros. Gos. Gidromet. Univ., No. 20, 42–55 (2011). (1989).
26. M. M. Smirnova, Avtoref. Candidate’s Dissertation in 30. R. West, D. Crisp, and L. Chen, “Mapping transforma-
Mathematics and Physics (Moscow State University, tions for broadband atmospheric radiation calcula-
Moscow, 2014). tions,” J. Quant. Spectrosc. Radiat. Transfer 43 (3),
191–199 (1990).
27. G. A. Grell, Y. H. Kuo, and R. J. Pasch, “Semiprog- 31. M. B. Ek, K. E. Mitchell, Y. Lin, E. Rogers, P. Grun-
nostic tests of cumulus parameterization schemes in the mann, V. Koren, Gayno, and J. D. Tarpley, “Imple-
middle latitudes,” Mon. Weather. Rev. 119 (1), 5–31 mentation of Noah land surface model advances in the
(1991). National Centers for Environmental Prediction oper-
28. J. A. Milbrandt and M. K. Yau, “A multimoment bulk ational mesoscale Eta model,” J. Geophys. Res.:
microphysics parameterization Part I. A proposed Atmos. 108 (D22), 16 (2003).
three-moment closure and scheme description,”
J. Atmos. Sci. 62 (9), 3065–3081 (2005). Translated by O. Ponomareva

ATMOSPHERIC AND OCEANIC OPTICS Vol. 32 No. 2 2019

You might also like