Analysis of tropical cloud systems using a cloud-top height data by geostationary satellite split-window measurements trained with CloudSat data NISHI, Noriyuki (Kyoto University) HAMADA, Atsushi (University of Tokyo) HIROSE, Hitoshi (Kyoto Universtiy)
Abstract New tropical-subtropical cloud top dataset has been completed and opened in our web-site. – Almost real-time update – Archive is available since June 2005 – Area covered by MTSAT-1R and MTSAT2 You can get cloud top height with 1-hour and 0.04° resolution – Best precision is obtained for upper tropospheric cloud – cirriform clouds, nimbostratus, cumulonimbus…
Tb (11μm) Example image on the web (19Z 01Jul2010) Tb (11μm)-Tb(12μm)) Optical Thickness SD for CloudTop CloudTop SD for Optical Thickness
Outline of the data We have made lookup tables for estimating the cloud-top height and visible optical thickness of upper-tropospheric clouds by the infrared brightness temperature (T B ) at 10.8 μm (T11) and its difference from T B at 12 μm (ΔT) measured by a geostationary satellite Influence of satellite zenith angle (SZA) on the brightness temperature around the rim of satellite field-of-view is reduced by creating LUT separately for the regions with a width of 15˚ SZA
Features – Hourly estimates (day&night) within wide area can be obtained by using only geostationary satellite data – Estimates have (ideally) no bias, since lookup tables were trained with cloud radar measurements onboard CloudSat – Reliability of estimate at each point is offered at the same time (less than 1 km for the upper-tropospheric clouds)
Estimation of cloud parameter by split-window method Utilizing the diference of absorption coefficient between 11μm and 12μm Particularly for ice phase Applicable both daytime and night (with no use of visible channel) (After HITRAN2000)
Estimation of cloud parameter by split-window method By using the difference between two channel, we can distinguish the difference of cloud top height between two clouds with same T11 value but different cloud top and optical depth 255K 270K Tb 11μm Tb 12μm DENSE 255K 270K 265K DENSE THIN
Split-window T B -- cloud top height, optical thickness high thick T 11, T 12 : 11,12um Tb z T : cloud top height τ : visible optical thickness T := T 11 – T 12 Estimation by traditional method with only T 11 Observed value Model parameter
Problem in the method with using radiative transfer model Simplified too much Too sensitive to the variation of the model parameter We made lookup table by purely experically with cloud radar ・ We should prepare a different lookup table for each satellite.
Data 1: geostationary satellite MTSAT-1R, MTSAT-2 11μm, 12μm T B (brightness temperature) – T 11, T 12 TOA radiation affected by cloud and water vapor Resolution : 0.04 o lon/lat, every hour Area and period Jul 2006– 80 o E–160 o W , 45 o S–45 o N 10/ /06/07 学位申請論文公聴会
Data 2: CloudSat 94GHz ( wavelength ~ 3mm ) millimeter-wavelength radar Resolution: horizontal 1.4 x 2.5km (only just below) Sampling: Vertically 240m (resolution around 500m) Merit Cloud ice, cloud water can be observed, as well as light rain and snow Possible to get almost all the data in whole cloud layer Demerit Observation region is narrow (~2km , only just below) Only two local time 0130/1330LT Attenuation due to heavy rain 高度 (km) ~5km ~500km Kochi-Univ. 11/ /06/07 学位申請論文公聴会
Extraction of sample 1. Extracting cloud radar observations that pass a MTSAT grid within 60 seconds from MTSAT observation, and averaging all the radar observation in a MTSAT grid 2. Defining more than two vertically continuous cloud echo bins as cloud layer (using cloud mask data) 3. Calculating: MTSAT 4-point average: T 11, T CloudSat Cloud top height ( z top ) Optical thickness ( vis ) of the highest cloud layer → Regressing z top and vis with T 11 and T to obtain lookup table (LUT) ~ 5km MTSAT grid CloudSat swath 12/48
Lookup table 0˚ < Satellite Zenith Angle < 15˚ Dashed line: standard deviation (km) Cloud top estimate (km) CTOP revised version
Estimation error Dashed line: standard deviation (km) CTOP revised version
Example over the central Pacific Ocean BLACK: Our results (estimates, stdev) RED: Aqua/MODIS YDL2_06 (CO2-slicing) Model-based approaches tend to underestimate CTHs even for optically thick clouds such as cumulonimbus/nimbostratus
Veritical distribution of the cloud top Jul-Dec N-7.5N CTOP(revised version) Cloud top estimate (maskout: tau 288 and ΔT < 2.5) ) Large difference from CloudSat direct observation in the region of large zenith angle CloudSat: the highest of at least three continuous cloud mask CloudSat CTOP (revised version) Maximum around E and peak around km are well represented in CTOP
Upper tropospheric cloud top (11-17km) CloudSat CTOP (revised version) Jul-Dec N-7.5N Maximum over the maritime continent and presence of ITCZ/SPCZ are well represented in CTOP
CTOP statistics General distribution of the upper-tropospheric cloud top in CTOP is close to that of CloudSat CloudSat observation is sparse. We can use our CTOP dataset as a complementary one to study climatology and interannual variability of the upper tropospheric clouds. Some problems remain In large zenith angle region (edge of the MTSAT view), the CTOP distribution is somewhat different from that of CloudSat Due to the lack of the sample size in making LUT, the vercial distribution is still ‘noisy’
Revision to version 2 Improve the estimate for very thin clouds (High T 11 and large ΔT) – When making LUT, excluding the pixels that CloudSat indicates ‘no-cloud’ but T 11 is low Some technical revision – Matching the location of pixels of CloudSat and MTSAT, considering the zenith angle of MTSAT – Adjusting the parameters in local regression to extend the parameter range (T 11, ΔT) when making LUT – Avoiding the effect of small difference of viewing angle between T 11 and T 12 observations
Revision to version 2 Ver.1: when CloudSat shows no cloud, the cloud height is set at 0 km. Revised: do not use the pixel with CloudSat height less than 3 km. Ver.1 revised Regression curve around T11=270K Each point shows that each observation
Future plans Making lookup tables (LUTs) for other satellites – current geostationary satellites (MSG, GOES, etc.) – current orbital satellites (Aqua, NOAA, etc.) – past geostationary/orbital satellites for the study of climate before CloudSat launch – Extension to the mid-latitude with objective analyisis dataset Seeking the possibility to analyze clouds with lower cloud top (<11km) – Congestus clouds…
Conclusion New tropical-subtropical cloud top dataset has been completed and opened in our web-site. – Almost real-time update – Archive is available since June 2005 – Area covered by MTSAT-1R and MTSAT2 You can get cloud top height with 1-hour and 0.04°resolution – Estimation error is also utilized – Best precision is obtained for upper tropospheric cloud – cirriform clouds, nimbostratus, cumulonimbus…
Cloud top dataset (CTOP)
FAQ Why do you use CloudSat rather than CALIPSO – In many cases, cloud top observed by CALIPSO is subvisible cirrus near the cold point tropopause. IR method is not applicable to such an altitude where the temperature does not decrease with height. If we select only the cloud with large optical thickness, we can do the same analysis. However, the result is not so different with one with CloudSat data.
FAQ Is this dataset can be available for any clouds? – No. The main target is visible cirrus, nimbostratus, and cumulonimbus, which has their top is between km. – Subvisible cirrus around the cold point tropopause cannot be included, since the hypothesis that the temperature decreases with height cannot be correct there. So, IR method is not good for that. – Lower cloud (< 11km) is not good for our method. The estimation error is fairly large.