Comparative Study on Cloud Parameter Estimation Among GOSAT / CAI , MODIS , CALIPSO / CALIOP and Landsat-8 / OLI with Laser Radar : Lidar as Truth Data

A comparative study on cloud parameter estimation among GOSAT/CAI, MODIS, CALIPSO/CALIOP and Landsat-8/OLI is carried out using Laser Radar: Lidar as a truth data. Optical depth, size distribution, as well as cirrus type of clouds are cloud parameters. In particular, cirrus cloud detection is tough issue. 1.38 μm channel is required for its detection. Although MODIS and Landsat-8/OLI have such channel, the other mission instruments, CAI and CALIPSO/CALIOP do not have such channel. As a truth data of cloud parameter, ground based Lidar is used in this comparative study. From the Lidar, backscattered echo signal and depolarization coefficient are obtained as a function of altitude. Therefore, cloud type, vertical profile can be derived from the Lidar data. CALIPSO/CALIOP is satellite based Lidar which allows observation of clouds from space. Although the directions of laser light emissions between CALIPSO/CALIOP and the ground based Lidar are different, their principles are same. Therefore, it is expected that CALIPSO/CALIOP data derived cloud parameters are similar to the ground based Lidar data derived cloud parameters. The experimental results show the aforementioned facts and are useful for improvement of cloud parameter estimation accuracy with several sensor data combinations. Keywords—Cirrus cloud; GOSAT/CAI; Landsat; LiDAR; Sky view camera; CALIPSO/CALIOP; topogramphic representation of 3D clouds


INTRODUCTION
Cloud detection is one of tough issues in satellite remote sensing in particular for cirrus cloud detections [1]- [16].In order to detect cirrus clouds, 1.38μｍ of wavelength channel is adopted for Moderate resolution of Imaging Spectrometer: MODIS1 and Landsat-8 Operational Land Imager: OLI2, etc. Green house gasses Observation Satellite / Cloud and Aerosol Imager: GOSAT 3 /CAI (Cloud and Aerosol Imager) 4 is dedicated sensor for cloud and aerosol retrievals.Because that GOSAT 5 /FTS (Fourier Transform Spectrometer 6 ) data is affected by clouds and aerosols, GOSAT/CAI is carried on the same platform of GOSAT satellite.
In order to check the capability of cirrus cloud detection, Light Detection and Ranging、Laser Imaging Detection and Ranging: Lidar data which allows measurement of back scattering ratio and depolarization ratio is used [17]- [29].Preliminary study results of cirrus cloud detection are reported by authors [30].The ground based Lidar is equipped at one of the GOSAT validation site which is situated at Saga University, Japan.Therefore, vertical profile of aerosol particles as well as cloud particles are detected which results in detection of aerosols and clouds including cirrus clouds.Vertical cloud structure can be retrieved with Cloud Aerosol Lidar and Infrared Pathfinder Satellite Observations: www.ijarai.thesai.orgCALIPSO/CALIOP 7 data.Therefore, detected cirrus clouds can be validated with CALIPSO/CALIOP data.In this paper, comparative study on cloud parameter estimation among GOSAT/CAI, MODIS, CALIPSO/CALIOP and Landsat-8/OLI is carried out using Lidar as a truth data.Optical depth, size distribution, as well as cirrus type of clouds are cloud parameters.Although MODIS and Landsat-8/OLI have such channel, the other mission instruments, CAI and CALIPSO/CALIOP do not have such channel.As a truth data of cloud parameter, the ground based Lidar is used in this comparative study.From the Lidar, backscattered echo signal and depolarization coefficient are obtained as a function of altitude.Therefore, cloud type, vertical profile can be derived from the Lidar data.CALIPSO/CALIOP is satellite based Lidar which allows observation of clouds from space.Although the directions of laser light emissions between CALIPSO/CALIOP and the ground based Lidar are different, their principles are same.Therefore, it is expected that CALIPSO/CALIOP data derived cloud parameters are similar to the ground based Lidar data derived cloud parameters.
In the next section, the proposed method for evaluation of cirrus detection accuracy of GOSAT/CAI, CALIPSO/CALIOP, MODIS and Landsat-8/OLI is described followed by experiments which includes method and procedure as well as the obtained experimental results.Finally, conclusion and some discussion are described.

II. RESEARCH BACKGROUND
Using spaceborne based FTS, column CO 2 and CH 4 in the atmosphere can be estimated.In the estimation of these atmospheric continuants, clouds and aerosols are influencing factors for estimation.Therefore, clouds and aerosol parameters, effective particle size distribution, optical thickness has to be clarified.In order to eliminate such FTS data which are suffered from clouds and aerosols have to be identified and cloud and aerosol masks have to be created.Using the mask pattern, non-confidential FTS data is removed from the estimation of CO 2 and CH 4 .Fig. 1

A. Lidar as Truth Data of Clouds
There are TCCON8 validation sites in the world.One of these is Saga University site in Japan.The location and outlook is shown in Fig. 3.

A. Landsat-8/OLI
There are 7 match-up data between the Lidar and Landsad-8/OLI images.Fig. 6 shows the match-up Landsat-8/OLI images.Color composite is assigned for red color to Band 5, for green color to Band 9, and for blue color to Band 6, respectively.Therefore, green colored pixels show existing of cirrus cloud because Band 9 (1.38 μm) is dedicated spectrum channel for cirrus cloud detection.

C. CALIPSO/CALIOP
The CALIPSO/CALIOP data of the orbit for the bottom right in the Fig. 5 are used for analysis.It is found the match up data of CALIPSO/CALIOP data (7 days out of 173 days).
One of examples of CLIPSO/CALIOP image data is shown in Fig. 9. Around the center of these images, the Saga University TCCON site is situated.It is clearly seen that there are cirrus clouds at around the altitude of 10-14 km above sea level.
On the other hand, Fig. 10 shows MODIS Band 31 imagery data which acquired almost same time on that day.Both CALIPSO/CALIOP and Lidar data (Fig. 4 (c)) are coincident.This is same thing for the other 6 days.Therefore, the match upped CALIPSO/CALIOP data show 100% perfect cirrus cloud detection accuracy.

D. GOSAT/CAI
The pixels of which the ratio of Band 4 (1.60um) and Band 3 (0.87um) is greater than 0.8 are identified as cirrus contaminated pixels.Band 4 is for detecting clouds while Band 3 is for removing surface reflectance influence.One of the examples of cirrus cloud detection is shown in Fig. 11.The match up ratio is around 26.6% (46 days out of 173 days) while cirrus detection accuracy is approximately 52% (24 days out of 46 days).
The results of the comparative study are summarized in the Table 2.  CALIPSO/CALIOP shows the best cirrus cloud detection accuracy (100%) followed by Aqua/MODIS, Landsat-8/OLI and GOSAT/CAI.CALIPSO/CALIOP is dedicated to cloud detection and has Lidar as mission instrument.Therefore, it is obvious that cirrus detection accuracy is excellent.Meanwhile, swath width of the CALIPSO/CALIOP is not so wide that entire globe coverage cannot be achieved within a day.On the other hand, swath width of Aqua/MODIS ensures entire globe coverage within a day for cirrus cloud detection accuracy is not so high though.Therefore, complementary use of Aqua/MODIS and CALIPSO/CALIOP would be recommendable combination for cirrus cloud distribution of entire globe within a day.

VI. CONCLUSION AND FUTURE WORKS
A comparative study on cloud parameter estimation among GOSAT/CAI, MODIS, CALIPSO/CALIOP and Landsat-8/OLI is carried out using Laser Radar: Lidar as a truth data.Optical depth, size distribution, as well as cirrus type of clouds are cloud parameters.In particular, cirrus cloud detection is tough issue.1.38μm channel is required for its detection.Although MODIS and Landsat-8/OLI have such channel, the other mission instruments, CAI and CALIPSO/CALIOP do not have such channel.As a truth data of cloud parameter, ground based Lidar is used in this comparative study.From the Lidar, backscattered echo signal and depolarization coefficient are obtained as a function of altitude.Therefore, cloud type, vertical profile can be derived from the Lidar data.CALIPSO/CALIOP is satellite based Lidar which allows observation of clouds from space.Although the directions of laser light emissions between CALIPSO/CALIOP and the ground based Lidar are different, their principles are same.Therefore, it is expected that CALIPSO/CALIOP data derived cloud parameters are similar to the ground based Lidar data derived cloud parameters.The experimental results show the aforementioned facts and are useful for improvement of cloud parameter estimation accuracy with several sensor data combinations.
Cirrus cloud detection accuracy of GOSAT/CAI and Landsat-8/OLI is evaluated with a ground based Laser Radar: Lidar data and sky view camera data.Also, the evaluation www.ijarai.thesai.orgresults are confirmed with CALIPSO/CALIOP data together with a topographic representation of vertical profile of cloud structure.Furthermore, origin of cirrus clouds is estimated with forward trajectory analysis.The results show that GOSAT/CAI derived cirrus clouds is not accurately enough due to missing of cirrus cloud detection spectral channel while Landsat-8 derived cirrus cloud.
In order to create cirrus cloud flag for GOSAT/FTS, it is recommendable to use GOSAT/FTS data derived spectrum.As shown in Fig. 1, the spectrum which is suffered from cirrus cloud has specific feature in the spectrum (Relatively large variance in the greater wave number region is observed for cirrus cloud contaminated FTS data).This cirrus cloud flag creation is one of future research works.Also, the complementary use of Aqua/MODIS and CALIPSO/CALIOP would be one of future research works for cirrus cloud distribution of entire globe within a day.
Further study is required for assess and evaluate the recommendable method for cirrus cloud detection as well as flagging to the acquired GOSAT/FTS data.

Fig. 3 .
Fig. 3. Saga University TCCON site for GOSAT validation B. Lidar Data Derived Backscatterin Ratio and Depolarization Ratio Examples of the Lidar data derived backscattering ratio and depolarization ratio are shown in Fig.4.Backscattering ratio shows the back scattering strength from aerosol and cloud particles while depolarization ratio shows degree of spherical of the particulates.

Fig. 4 ( 29 Fig. 4 .
Fig.4 (a) to (i) shows the lidar data derived backscattering ratio and depolarization ratio which are acquired on April 23, May 9, May 11, September 14, November 17, December 27 in 2014 and on January 20, March 25, June 29 in 2015, respectively.There are aerosol and cloud particles in the lower layer (boundary layer in troposphere) on April 23 2014 while it seems that there is cirrus clouds at around 12.5km, there are thin aerosol and cloud particles in the lower layer though.This is same thing for the lidar derived backscattering and depolarization ratios of May 11 2014.There seems to be cirrus cloud at around 12km, there are thin aerosol and cloud particles in the lower layer though.There are cirrus clouds at around 11.5km on September 14 2014.On the other hand, it seems that there are non-spherical aerosol and cloud particles at around 1.5km for the case of November 17 2014.Meanwhile, there are cirrus clouds at around 8km for the case of December 27 2014.There are low layer clouds at around 1km for the case of January 20 2015 while comparatively clear sky condition is confirmed on March 25 2015.On the other hand, there are cirrus clouds at around 7 km for the case of June 29 2015.

Fig. 5 .
Fig. 5. Examples of the CALIPSO/CALIOP satellite track Through a comparison between Fig.3 and Fig.5, it is truly rare to get match-up data between CALIPSO/CALIOP and the Lidar data (In these examples, only the satellite track which is shown in the bottom right is match-up to the ground based Lidar site).

Fig. 6 .
Fig. 6.Match-up data of Landsat-8/OLI imagery data with the Lidar data acquisition Although Landsat-8/OLI data show the existing cirrus clouds, the Lidar data show no existing of such cirrus cloud at all.Cirrus clouds are used to be existed at around 5 to 15 km depending on latitude and season.There is no such evidence at the altitude in the Lidar data.Clearly, there are some evidences of the existing of cirrus clouds in the Lidar data.Also, Landsat-8/OLI shows the same existing of cirrus clouds.Therefore, both Landsat-8/OLI and Lidar show good coincidence.

B
. Aqua/MODIS 96 days Aqua/MODIS data are match upped with GOSAT/FTS data.Cirrus clouds are detected with Aqua/MODIS data for 72 days out of 96 days.MODIS has 1.38 μm (cirrus detectable channel) and 660nm channel for removal of land surface influence in the cirrus clouds detection.Therefore, good cirrus cloud detection accuracy is expected.www.ijarai.thesai.orgMODIS cloud products (MYD03: Cirrus data and MYD06: Longitude / Latitude data) used for analysis.Fig.7 is an example of MYD03 of cirrus reflectance flag data.Black portions show cirrus clouds.On the other hands, Fig.7 shows Lidar data, Sky view camera imagery data, and MODIS imagery data of December 27 2014.Fig.8 (a) shows existence of the cirrus clouds and also MODIS imagery data of Fig.8 (b) shows existence of the cirrus clouds.

Fig. 9 .
Fig. 9.One of the examples of CALIPSO/CALIOP image data

Fig. 10 .
Fig. 10.MODIS Band 31 of brightness temperature at that time on that day (Lidar data of back scattering ratio and depolarization ratio at the time on May 11 2014 is shown in Fig.4 (c))

Table 1 shows
the summarized results of the experiments through comparison between Landsat-8/OLI and Lidar derived cirrus clouds.As a result through experiments, five coincidences are confirmed between Landsat-8/OLI and Lidar derived cirrus clouds out of seven cases.Therefore, approximately 71% of cirrus cloud detection accuracy is confirmed for Landsat-8/OLI data.