- Email: [email protected]
Determination of cloud top height using meteorological satellite and radar data
Phys. Chem. Earth (B), Vol. 25, No. 10-12, pp. 1103-l 106.2000 Q 2000Elsevier Science Ltd All rights reserved 1464-1909/00/S - see front matter
Pergamon
PII: S1464-1909(00)00160-X
Determination Data
of Cloud Top Height Using Meteorological
R. Randriamampianina’,
J. Nagy2, T. Balogh’ and J. Ker6nyi3
‘Division for Numerical Weather Prediction, Hungarian Meteorological Service, Hungary. E-mail: [email protected] 2Division of Aerology and Radarmeteorology, Hungarian Meteorological Service, Hungary 3Satellite Research Laboratory, Hungarian Meteorological Service, H- 1675 Budapest Received
16 June 2000; accepted
H-1675
Budapest
P.O. Box 39,
H-1675
Budapest
P.O. Box 39,
P.O. Box 39, Hungary
30 June 2000
Abstract. At the Hungarian Meteorological Service we receive the digital Meteosat images every 30 minutes and operate a radar network, which consists of two MRL-5 and one DWSR-2500C Doppler weather radar. The MRL-5 radar measures at 3 and 10 cm wavelengths in different ranges (32, 64, 128 and 256 km), while the Doppler radar at 5 cm in 480 km range. In this study we derive the cloud top heights (CTH) from Meteosat images and compare them with echo top values calculated from Range Height Indicator (RHI) vertical radar pictures. A cloud detection method has been developed using a threshold technique based on infrared and visible Meteosat data. The cloud top height is calculated operationally by using the infrared brightness values and ECMWF predicted temperature profiles. We got good agreement between the calculated cloud top heights and radar measurements for one-layer clouds, while bigger differences were obtained for cases, when clouds at different levels presented. 0 2000 Elsevier Science Ltd. All rights reserved.
(Schmetz et al., 1993; Nieman et al., 1993; Nieman et al., 1997) and the infrared (IR) window method (Schmetz et al., 1993). We use the IR window method for the CTH assignment at this stage. This paper describes the cloud detection and CTH assignment schemes used, and discusses the results by comparing them with data derived from radar measurements.
2 Materials and methods 2.1 Database The Hungarian radar meteorological network consists of two MRL-5 radars, situated in Nyiregyhaza-Napkor and Szentgothard-Farkasfa, and a DWSR-2500C Doppler weather radar, situated in Budapest-LGrinc. The latter one is a new radar and is being calibrated at the time. Thus, for the Budapest station the measurements of the previous MRL-5 radar were used. In our computations we took into account only the data not further than 100 km from the MRL-5 but skipped the data from the nearest neighbourhood. For the CTH assignment using the IR window method ancillary data are required (Schmetz et al., 1993). We use the forecast of the European Centre for Medium-Range Weather Forecasts (ECMWF) for this purpose. Vertical radar pictures (measured at 3 and 10 cm wavelength in one direction) in different ranges (32, 64, 128 and 256 km) are measured hourly for their comparison with the detected CTH. The 12 UTC data of the period of January to April 1998 were used for this investigation to ensure data from all the three radar stations.
1 Introduction The nowcasting system recently worked out at the Hungarian Meteorological Service requires cloud top height (CTH) data as recent as possible. For this reason a reliable calculation scheme to determine the CTH is being worked out. Geostationary satellites (in our case Meteosat) provide consecutive images that allow us to follow the cloud motion. Many methods were developed to assign the CTH from the remotely sensed data, such as the CO* absorption method (Menzel et al., 1983), the histogram method (Derrien et al., 1988), the water vapour absorption method
Correspondence
Satellite and Radar
to: R. Randriamampianina 1103
R. Randriamampianina
1104
et al.: Determination
3 Results and discussion
2.2 Cloud detection The cloud detection by TLnczer (1997). method using the visible threshold (r)
of Cloud Top Height
technique is The method image from is computed
close to the one published is based on the threshold the visible channel. The as follows:
Comparing the CTHs derived from radar and satellite measurements we found comparable results when there was only one cloud type (Fig.2. and Fig.3.). 500045003500 4000
.-
n
Radar mes.: 1998 February 25 11 h 33mn UK
A
3000
-
8
2500 2000
.
2 .?P
l
&
JYE%iJ
0
+ A
where f, is the brightness value from the visible spectral channel at 12 UTC, already corrected for the solar elevation angle; fv,i, and f,,,, are the global minimum and maximum off, after filtration of the erroneous values, respectively. The cloud detection scheme is presented in Fig.1. This scheme is simplified according to the aim of the study - to derive the CTH. Thus, it should be more complicated for cloud classification and for cloud type specification cases.
. +
A
l
n
A
A
8
lSO0 .
I +
A
1000
t
:
500
A
A
l
:
.
,,
.
2
A
:
10
11
12
0 1
2
3 satdite
Fig. 2. Comparison satellite data
4
5
7
6
8
9
of derived heights using MRL-5 radar in Budapest and
6000 5800
.
Radar mes.: 1998 April 09 12h Olmn
5600
.
UTC
4,
l
. .
*
l l
5400
+
+
+
+
l
4800. sooo5200 . 2A 3
:
:
:
:
:
:
2
;
2
:
3
4
5
6
7
8
9
10
11
12
4600
.
4400
- d
13
pixels along the radar masureml?nt dimction
1
.it
4200 4000 1
Fig. 1. Cloud detection scheme. Where: q is the visible threshold; 6 = 0.3 from May to September and 6 = 0.5 from October to April; AT = T12 - Ts; Tlz and Tg - calibrated IR cloud brightness temperatures at 12 UTC and 9 UTC, respectively.
2.3 Cloud top height assignment The cloud top height assignment is based on the IR cloud brightness temperature. The ECMWF forecasts are used as supplementary data to derive the CTH. The level, where the brightness temperature fits the forecast temperature, represents the cloud top level (height). The calculation is accomplished for cloudy [cases [(c) and (d) in Fig.l.1 or probably cloudy [case (b) in Fig.1 .] pixels, The experiments showed that 6 has different values within the year. Thus, 6=0.3 for the period from May until September and 6=0.5 from October until April The assigned heights were compared with the maximum (Rad. Max) and the average (Rad. Avg) values of radar pixels in 3 km surrounding area. In general, the CTHs derived from satellite data are higher than the ones derived from radar data.
2
Fig. 3. Comparison satellite data
13
of derived heights using MRL-5 radar in Napkor and
In case of a multi-level cloud system big differences between the two derived CTHs (Figs.4 and 5a) were found due to the different measurement techniques of satellite and radar. Thus, satellite measures from above the top of the cloud while the radar detects from the ground. Moreover, the reflectivity of the clouds depends on the physical properties of the particles which they consist of (Mazin and Khrgian, 1989). The CTHs, detected by radar, depend on whether the cloud is composed of water droplets or ice. Hence, the radar, in general, is not able to detect the Cirrus clouds. On the other hand, transparent clouds - Cirrus - can not be clearly distinguished in the infrared pictures. Thus, for one-layer Cirrus clouds the CTHs, calculated from the satellite data are underestimated.
R. Randriamampianina
7000
“I
6000 5000
et al.: Determination
1105
of Cloud Top Height
f$adar mes. :1998 April 08 llh 51mn UTC
4000 3000 2mo
. Rad. Max
1000 0 1
2
3
5
4
satdlitepixels
Fig. 4. Comparison satellite data
6
7
8
9
10
12
11
13
along the I-&w mcassunlnel direclion
of derived heights using MRL-5 radar in Farkasfa and Fig. 5b. Visible digital Meteosat UTC
Regarding multi-level cloud systems, that include Cirrus clouds, the CTHs of the lower clouds are overestimated, because the Cirrus affects the detection of the cloud top temperature of the lower clouds. An example of such situation is given for the Farkasfa station in FigsSa-d. The CTHs, calculated from the satellite data (Fig.Sa) indicate the presence of clouds in pixels 16-2 1. 9000 ++*4*
8000 -. 7000 I 6000 0
*fit;... .
Radar mes.: 1998 April 14 Ilh 51mn UTC 3Ono-
g
2
2000 -. *% d! lOQO-.
1
2
3 4 5 6 satellite
Fig. Sa. Comparison and satellite data
7
6
9
: ::
; ; :::
in 1998 April 14, 12
4 Conclusions The CTH data derived from satellite and radar measurements are in reasonable agreement if a single cloud type is observed. In case of multi-level cloud system we get information about the cloud top height of the lowest and the highest clouds, which might be used in cloud classification studies. It is recommended to compare the CTH data derived from satellite measurements -besides the radar data- with other calculations.
n A
04 : : ; ; : ; : ::
picture observed
::
Acknowledgement. This study was supported Research Foundation (OTKA-TO3 1865).
by the Hungarian
National
10 11 12 13 14 15 16 17 18 19 20 :
pials alongthe radarmeasurementdir&ion
of derived heights using MRL-5 radar
References in Farkasfa
The visible picture (FigSb) proofs this assumption. The radar pictures, (FigsSc-d), however, show clear territories with separated clouds. We suppose that the differences between the CTHs for pixels 16-21, derived by different methods (Fig.Sa) were caused by a Cirrus that covered the whole territory. Concerning the IR window method, using predicted profiles instead of actual profiles in the CTH calculation scheme could cause systematic errors which might have increased the inaccuracy of the results. According to our investigation, it would be desirable to compare the CTH data derived from satellite and radar measurements with other calculations. Numerical weather prediction products or results of a physically based cloud model could be used for this purpose.
Derrien, M., Lavanant, L., and LeGleau, H.. Retrieval of the top temperature of semi-transparent clouds with AVHRR_ Proceedings of the IRS’88 Conference, 199-202, 1988. Mazin, I.P. and Khrgian, A.Kh. (editors), Handbook of clouds and cloudy atmosphere, 450-492, Gidrometeoizdat, 1989. (in Russian) Menzel, W.P., Smith, W.L., and Steward, T.R., Improved Cloud Motion Wind Vector and Altitude Assignment Using VAS, Bulletin of the American Meteorological Society, 22, 377-384. 1983. Nieman, S.J., Menzel, W.P., Hayden, C.M., Gray, D., Wanzong, ST., Velden, C.S., and Daniels, J., Fully Automated Cloud DriftWinds in NESDIS Operations, Bulletin of the American Meteorological Society, 78/6, 1121-I 133, 1997. Nieman, S.J., Schmetz, J., and Menzel, W.P., A Comparison of Several Techniques to Assign Heights to Cloud Tracers, Bulletin of the American Meteorological Society, 32, 15S9- 1568, 1993. Schmetz, J., Holmlund, K., Hoffman, J., Strauss, B., Mason, B., Gaertner, V., Koch, A., and van de Berget, L., Operational Cloud-Motion Winds from Meteosat Infrared Images. Journal of Applied Meteorology 32, 1206-1224, 1993. Tanczer, T., The analyses of the cloudiness based on satellite pictures, LPgkiir, 1997/f. 1 l-15, 1997 (in Hungarian).
R. Randriamampianina
1106
et al.: Determination
of Cloud Top Height
pw : B.BBI _.__-.__.____
FARKASFll PPI:1.5O Chl:Scm
IpIE:
1998/84/14 12:m:.29 Ran:128
:
nn,,,
Fig. 5c MRL,-5 Weather radar picture, 1998 April 14 12h 0 1nmUTC in Farkasfa
FfiRKASFfi RHI:111.0* Chl:3cm
Fig. 5d Vertical cross-section
1998,84,14 11:51:42 Ran : 64
(RHI), 1998 April 14 11h 5 lmn UTC in Farkasfa
I.
dBZ
dBz -1e * mmfh
re 28 38 48 58 68
Pergamon
PII: S1464-1909(00)00160-X
Determination Data
of Cloud Top Height Using Meteorological
R. Randriamampianina’,
J. Nagy2, T. Balogh’ and J. Ker6nyi3
‘Division for Numerical Weather Prediction, Hungarian Meteorological Service, Hungary. E-mail: [email protected] 2Division of Aerology and Radarmeteorology, Hungarian Meteorological Service, Hungary 3Satellite Research Laboratory, Hungarian Meteorological Service, H- 1675 Budapest Received
16 June 2000; accepted
H-1675
Budapest
P.O. Box 39,
H-1675
Budapest
P.O. Box 39,
P.O. Box 39, Hungary
30 June 2000
Abstract. At the Hungarian Meteorological Service we receive the digital Meteosat images every 30 minutes and operate a radar network, which consists of two MRL-5 and one DWSR-2500C Doppler weather radar. The MRL-5 radar measures at 3 and 10 cm wavelengths in different ranges (32, 64, 128 and 256 km), while the Doppler radar at 5 cm in 480 km range. In this study we derive the cloud top heights (CTH) from Meteosat images and compare them with echo top values calculated from Range Height Indicator (RHI) vertical radar pictures. A cloud detection method has been developed using a threshold technique based on infrared and visible Meteosat data. The cloud top height is calculated operationally by using the infrared brightness values and ECMWF predicted temperature profiles. We got good agreement between the calculated cloud top heights and radar measurements for one-layer clouds, while bigger differences were obtained for cases, when clouds at different levels presented. 0 2000 Elsevier Science Ltd. All rights reserved.
(Schmetz et al., 1993; Nieman et al., 1993; Nieman et al., 1997) and the infrared (IR) window method (Schmetz et al., 1993). We use the IR window method for the CTH assignment at this stage. This paper describes the cloud detection and CTH assignment schemes used, and discusses the results by comparing them with data derived from radar measurements.
2 Materials and methods 2.1 Database The Hungarian radar meteorological network consists of two MRL-5 radars, situated in Nyiregyhaza-Napkor and Szentgothard-Farkasfa, and a DWSR-2500C Doppler weather radar, situated in Budapest-LGrinc. The latter one is a new radar and is being calibrated at the time. Thus, for the Budapest station the measurements of the previous MRL-5 radar were used. In our computations we took into account only the data not further than 100 km from the MRL-5 but skipped the data from the nearest neighbourhood. For the CTH assignment using the IR window method ancillary data are required (Schmetz et al., 1993). We use the forecast of the European Centre for Medium-Range Weather Forecasts (ECMWF) for this purpose. Vertical radar pictures (measured at 3 and 10 cm wavelength in one direction) in different ranges (32, 64, 128 and 256 km) are measured hourly for their comparison with the detected CTH. The 12 UTC data of the period of January to April 1998 were used for this investigation to ensure data from all the three radar stations.
1 Introduction The nowcasting system recently worked out at the Hungarian Meteorological Service requires cloud top height (CTH) data as recent as possible. For this reason a reliable calculation scheme to determine the CTH is being worked out. Geostationary satellites (in our case Meteosat) provide consecutive images that allow us to follow the cloud motion. Many methods were developed to assign the CTH from the remotely sensed data, such as the CO* absorption method (Menzel et al., 1983), the histogram method (Derrien et al., 1988), the water vapour absorption method
Correspondence
Satellite and Radar
to: R. Randriamampianina 1103
R. Randriamampianina
1104
et al.: Determination
3 Results and discussion
2.2 Cloud detection The cloud detection by TLnczer (1997). method using the visible threshold (r)
of Cloud Top Height
technique is The method image from is computed
close to the one published is based on the threshold the visible channel. The as follows:
Comparing the CTHs derived from radar and satellite measurements we found comparable results when there was only one cloud type (Fig.2. and Fig.3.). 500045003500 4000
.-
n
Radar mes.: 1998 February 25 11 h 33mn UK
A
3000
-
8
2500 2000
.
2 .?P
l
&
JYE%iJ
0
+ A
where f, is the brightness value from the visible spectral channel at 12 UTC, already corrected for the solar elevation angle; fv,i, and f,,,, are the global minimum and maximum off, after filtration of the erroneous values, respectively. The cloud detection scheme is presented in Fig.1. This scheme is simplified according to the aim of the study - to derive the CTH. Thus, it should be more complicated for cloud classification and for cloud type specification cases.
. +
A
l
n
A
A
8
lSO0 .
I +
A
1000
t
:
500
A
A
l
:
.
,,
.
2
A
:
10
11
12
0 1
2
3 satdite
Fig. 2. Comparison satellite data
4
5
7
6
8
9
of derived heights using MRL-5 radar in Budapest and
6000 5800
.
Radar mes.: 1998 April 09 12h Olmn
5600
.
UTC
4,
l
. .
*
l l
5400
+
+
+
+
l
4800. sooo5200 . 2A 3
:
:
:
:
:
:
2
;
2
:
3
4
5
6
7
8
9
10
11
12
4600
.
4400
- d
13
pixels along the radar masureml?nt dimction
1
.it
4200 4000 1
Fig. 1. Cloud detection scheme. Where: q is the visible threshold; 6 = 0.3 from May to September and 6 = 0.5 from October to April; AT = T12 - Ts; Tlz and Tg - calibrated IR cloud brightness temperatures at 12 UTC and 9 UTC, respectively.
2.3 Cloud top height assignment The cloud top height assignment is based on the IR cloud brightness temperature. The ECMWF forecasts are used as supplementary data to derive the CTH. The level, where the brightness temperature fits the forecast temperature, represents the cloud top level (height). The calculation is accomplished for cloudy [cases [(c) and (d) in Fig.l.1 or probably cloudy [case (b) in Fig.1 .] pixels, The experiments showed that 6 has different values within the year. Thus, 6=0.3 for the period from May until September and 6=0.5 from October until April The assigned heights were compared with the maximum (Rad. Max) and the average (Rad. Avg) values of radar pixels in 3 km surrounding area. In general, the CTHs derived from satellite data are higher than the ones derived from radar data.
2
Fig. 3. Comparison satellite data
13
of derived heights using MRL-5 radar in Napkor and
In case of a multi-level cloud system big differences between the two derived CTHs (Figs.4 and 5a) were found due to the different measurement techniques of satellite and radar. Thus, satellite measures from above the top of the cloud while the radar detects from the ground. Moreover, the reflectivity of the clouds depends on the physical properties of the particles which they consist of (Mazin and Khrgian, 1989). The CTHs, detected by radar, depend on whether the cloud is composed of water droplets or ice. Hence, the radar, in general, is not able to detect the Cirrus clouds. On the other hand, transparent clouds - Cirrus - can not be clearly distinguished in the infrared pictures. Thus, for one-layer Cirrus clouds the CTHs, calculated from the satellite data are underestimated.
R. Randriamampianina
7000
“I
6000 5000
et al.: Determination
1105
of Cloud Top Height
f$adar mes. :1998 April 08 llh 51mn UTC
4000 3000 2mo
. Rad. Max
1000 0 1
2
3
5
4
satdlitepixels
Fig. 4. Comparison satellite data
6
7
8
9
10
12
11
13
along the I-&w mcassunlnel direclion
of derived heights using MRL-5 radar in Farkasfa and Fig. 5b. Visible digital Meteosat UTC
Regarding multi-level cloud systems, that include Cirrus clouds, the CTHs of the lower clouds are overestimated, because the Cirrus affects the detection of the cloud top temperature of the lower clouds. An example of such situation is given for the Farkasfa station in FigsSa-d. The CTHs, calculated from the satellite data (Fig.Sa) indicate the presence of clouds in pixels 16-2 1. 9000 ++*4*
8000 -. 7000 I 6000 0
*fit;... .
Radar mes.: 1998 April 14 Ilh 51mn UTC 3Ono-
g
2
2000 -. *% d! lOQO-.
1
2
3 4 5 6 satellite
Fig. Sa. Comparison and satellite data
7
6
9
: ::
; ; :::
in 1998 April 14, 12
4 Conclusions The CTH data derived from satellite and radar measurements are in reasonable agreement if a single cloud type is observed. In case of multi-level cloud system we get information about the cloud top height of the lowest and the highest clouds, which might be used in cloud classification studies. It is recommended to compare the CTH data derived from satellite measurements -besides the radar data- with other calculations.
n A
04 : : ; ; : ; : ::
picture observed
::
Acknowledgement. This study was supported Research Foundation (OTKA-TO3 1865).
by the Hungarian
National
10 11 12 13 14 15 16 17 18 19 20 :
pials alongthe radarmeasurementdir&ion
of derived heights using MRL-5 radar
References in Farkasfa
The visible picture (FigSb) proofs this assumption. The radar pictures, (FigsSc-d), however, show clear territories with separated clouds. We suppose that the differences between the CTHs for pixels 16-21, derived by different methods (Fig.Sa) were caused by a Cirrus that covered the whole territory. Concerning the IR window method, using predicted profiles instead of actual profiles in the CTH calculation scheme could cause systematic errors which might have increased the inaccuracy of the results. According to our investigation, it would be desirable to compare the CTH data derived from satellite and radar measurements with other calculations. Numerical weather prediction products or results of a physically based cloud model could be used for this purpose.
Derrien, M., Lavanant, L., and LeGleau, H.. Retrieval of the top temperature of semi-transparent clouds with AVHRR_ Proceedings of the IRS’88 Conference, 199-202, 1988. Mazin, I.P. and Khrgian, A.Kh. (editors), Handbook of clouds and cloudy atmosphere, 450-492, Gidrometeoizdat, 1989. (in Russian) Menzel, W.P., Smith, W.L., and Steward, T.R., Improved Cloud Motion Wind Vector and Altitude Assignment Using VAS, Bulletin of the American Meteorological Society, 22, 377-384. 1983. Nieman, S.J., Menzel, W.P., Hayden, C.M., Gray, D., Wanzong, ST., Velden, C.S., and Daniels, J., Fully Automated Cloud DriftWinds in NESDIS Operations, Bulletin of the American Meteorological Society, 78/6, 1121-I 133, 1997. Nieman, S.J., Schmetz, J., and Menzel, W.P., A Comparison of Several Techniques to Assign Heights to Cloud Tracers, Bulletin of the American Meteorological Society, 32, 15S9- 1568, 1993. Schmetz, J., Holmlund, K., Hoffman, J., Strauss, B., Mason, B., Gaertner, V., Koch, A., and van de Berget, L., Operational Cloud-Motion Winds from Meteosat Infrared Images. Journal of Applied Meteorology 32, 1206-1224, 1993. Tanczer, T., The analyses of the cloudiness based on satellite pictures, LPgkiir, 1997/f. 1 l-15, 1997 (in Hungarian).
R. Randriamampianina
1106
et al.: Determination
of Cloud Top Height
pw : B.BBI _.__-.__.____
FARKASFll PPI:1.5O Chl:Scm
IpIE:
1998/84/14 12:m:.29 Ran:128
:
nn,,,
Fig. 5c MRL,-5 Weather radar picture, 1998 April 14 12h 0 1nmUTC in Farkasfa
FfiRKASFfi RHI:111.0* Chl:3cm
Fig. 5d Vertical cross-section
1998,84,14 11:51:42 Ran : 64
(RHI), 1998 April 14 11h 5 lmn UTC in Farkasfa
I.
dBZ
dBz -1e * mmfh
re 28 38 48 58 68