SEARCH

SEARCH BY CITATION

Keywords:

  • clear sky;
  • CALIOP/CALIPSO;
  • MODIS/Aqua

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Analysis
  5. 3. Results
  6. 4. Summary and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

[1] We derived the global distribution of the clear-sky occurrence rate during the daytime using Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) and Moderate-resolution Imaging Spectrometer (MODIS) data. Our goal was to estimate the frequency of likely successful CO2 and CH4 column concentration retrievals by the Greenhouse gases Observing Satellite (GOSAT). Clear-sky regions over land were frequently found around desert regions throughout the year, and at northern mid-latitudes in the winter hemisphere. Overall, clear-sky regions covered approximately 11% of the globe annually, on average. From a comparison of the CALIOP and MODIS cloud data for June 2007, we found that MODIS overestimated the clear-sky ratio by approximately 5%, except for in the tropics.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Analysis
  5. 3. Results
  6. 4. Summary and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

[2] The Greenhouse gases Observing Satellite (GOSAT) will be launched during Japan's fiscal year 2008 (i.e., before 31 March 2009). Orbiting at an altitude of 666 km with an orbit repetition frequency of 3 days, GOSAT will measure concentrations of green house gases, such as CO2 and CH4, as well as cloud and aerosol concentrations on a global scale [Yokota et al., 2004]. GOSAT will carry two instruments: the Thermal And Near-infrared Sensor for carbon Observation Fourier Transform Spectrometer (TANSO FTS) and the TANSO Cloud and Aerosol Imager (CAI). The shortwave infrared (SWIR) detectors of TANSO FTS will measure the visible and near-infrared solar spectra (Band 1: 0.76 μm; Band 2: 1.6 μm; Band 3: 2.0 μm) scattered by the ground, clouds, and aerosol. One of the main aims of the GOSAT project is to obtain data of sufficient quality to retrieve three month averages and subcontinental-scale column-density concentrations of CO2 and CH4 with random errors of less than 1 and 2%, respectively, for clear-sky conditions.

[3] The retrieval of CO2 and CH4 column concentrations from the TANSO FTS SWIR spectra faces several sources of error, such as those caused by cirrus, aerosol, and ground-surface albedo. However, because of the difficulties associated with dealing explicitly with such sources of error during the retrieval process of CO2 and CH4, the effects of clouds on observational data obtained from remote-sensing campaigns are often ignored; thus uncertainties can be significant. Therefore, our primary aim is to derive the CO2 and CH4 column concentrations for columns with a suitable albedo (0.1 < α < 0.5) within the instantaneous field of view (IFOV) of TANSO FTS (diameter ∼ 10 km) during the daytime, under clear-sky conditions (the clear-sky conditions used in the present study are defined below).

[4] TANSO CAI will provide cloud-coverage information within the IFOV of the FTS. If this cloud information indicates the presence of relatively thick clouds (optical thickness τ > 1.0), the spectra from TANSO FTS will not be processed any further. If τ < 1.0, the CO2 and CH4 columns will be retrieved under the assumption of thin cirrus. In the retrieval process, the TANSO FTS 2.0 μm band can be used to obtain the parameters of cirrus clouds (i.e., their heights and optical thicknesses), for a minimum τ of 0.01 (Y. Yoshida, GOSAT, personal communication, 2008).

[5] Although numerous groups have investigated cloud parameters [e.g., Hong et al., 2007], few studies have focused on the clear-sky distribution on a global scale. Miller et al. [2007] used MODIS cloud-mask data for clear-sky analysis. By comparing MODIS and in-situ measurements, Ackerman et al. [2008] concluded that MODIS is relatively insensitive to clouds with τ < 0.4. Recently, it was proven that satellite lidar data, particularly the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP), can detect relatively thin clouds and aerosol with an optical thickness as low as 0.01 [e.g., McGill et al., 2007] on a global scale, enabling more accurate determinations of the sky conditions (i.e., cloudy or clear). By analyzing the clear-sky distribution derived from the CALIOP data, we can predict which regions will exhibit small errors in the retrieved CO2 and CH4 columns; these data can also help with the selection of optimal GOSAT validation sites.

[6] The aim of the present study was to investigate the seasonal distribution of clear skies using the CALIOP cloud-product data. We also used MODIS cloud data (mainly cloud-mask data) for comparison with the clear-sky distribution derived from the CALIOP measurements and to obtain more wide-ranging information about clear skies which cannot be detected by the small along-track footprint (∼100 m; see below) of CALIOP.

2. Data Analysis

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Analysis
  5. 3. Results
  6. 4. Summary and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

[7] To derive accurate clear-sky information on a global scale, we mainly used the cloud-product data from CALIOP onboard the Cloud Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) satellite [Winker et al., 2007]. CALIPSO was launched on 28 April 2006. It orbits the Earth about 15 times a day at an altitude of 705 km and with an inclination of 98.2 degrees. The CALIOP laser foot print is ∼70 meters in diameter; the laser pulse repetition frequency is 20.16 Hz [Winker et al., 2007]; and the vertical resolutions in the upper and lower troposphere are ∼60 and 30 meters, respectively [Anselmo et al., 2007]. Three receiver channels measure a 1064 nm backscatter signal and two orthogonally polarized components of a 532 nm backscatter signal to obtain the heights and optical parameters of the clouds and aerosols. The present study used the top and bottom heights of the cloud layer and the number of cloud layers based on measurements with 1 and 5 km along-track resolutions obtained during the daytime from 1 January to 31 December 2007. Thermal effects affecting the CALIOP daytime signal have been removed from the current version of the cloud-layer product (Version 201) [Powell et al., 2008].

[8] We also used the daytime clear-sky data with a horizontal resolution of 1 km derived from the cloud-mask data of MODIS, onboard the Aqua satellite. We used the MAC06S0 products available from the Goddard Earth Sciences (GES) Data and Information Services Center (DISC) (http://gcmd.gsfc.nasa.gov/records/GCMD_GES_DISC_MAC06S0_v2.html). The MODIS cloud data were used to obtain cloud information of a greater horizontal extent than that supplied by CALIOP because the MODIS swath is a few hundred kilometers wide, while the CALIOP laser footprint is only ∼90 meters across. The orbit of CALIPSO is only a few hundred kilometers to the east of Aqua's path at the time of their respective Equator crossings on the dayside of the orbit, and it follows Aqua by approximately two minutes. The MODIS cloud mask data are provided in four categories: ‘Confident Cloudy’, ‘Probably Cloudy’, ‘Probably Clear’, and ‘Confident Clear’. We used Confident Clear (Cloudy) as an indicator of clear (cloudy) skies. For the clear-sky analysis, we removed the data for which the cirrus-cloud flag, obtained from the 1.38 μm band, indicated the presence of cirrus clouds and contaminating shadows. Most (∼90–95%) cirrus clouds were found in the data flagged Confident or Probably Cloudy, but a small percentage of cirrus clouds contaminated the data marked Confident or Probably Clear.

3. Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Analysis
  5. 3. Results
  6. 4. Summary and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

[9] This study defines the fraction of the number of clear arrays (i.e., no cloud signal) to the total number of arrays (based on the cloud-product data) as the clear-sky frequency measured by CALIOP. Thin cirrus with a τ of smaller than 0.01 might be included in this frequency. This condition for the cirrus detected by CALIOP meets that of the cirrus clouds retrieved from the band-2 spectrum of the TANSO FTS SWIR. The clear-sky frequency was investigated during the daytime because the TANSO FTS SWIR will measure sunlight reflected by the ground surface, clouds, and aerosols. To calculate the clear-sky frequency from the CALIOP data, we only used those data identified as good by the Feature-finder Quality Check. Using this definition, we found that approximately 99% of all CALIOP data were of good quality. We also used the fraction of ‘Confident Clear’ data to the total data volume as the clear-sky frequency measured by MODIS. For our investigation of the clear-sky frequency, we first aggregated CALIOP or MODIS pixels into larger arrays of different sizes and shapes, as described in this section, and labeled each array as ‘clear’ if all of the aggregated pixels in the array were flagged as cloud-free.

3.1. Annual and Seasonal Variations of the Clear-Sky Frequency Derived From CALIOP Measurements

[10] Figure 1 shows the annual mean distribution of the clear-sky frequency derived from CALIOP measurements with a resolution of 5 km. In preparing Figure 1, we aggregated two 5 km-resolution CALIOP data points taken along the satellite's track (i.e., spanning 10 km along the track) to determine the clear-sky frequency because the diameter of the TANSO FTS IFOV is approximately 10 km. The clear-sky frequency was high in desert and high-mountain regions throughout the year (Figure 1). Generally, the clear-sky frequency over sea, particularly over convective regions, was less than 10%. In the northern summer (not shown), the Asian monsoon and north Pacific regions were almost always cloudy. On the other hand, in the northern winter (not shown), the clear-sky frequency over land at northern middle latitudes was high, and the equivalent frequency in South America was low.

image

Figure 1. Annual mean distribution of the daytime clear-sky frequency [%] derived from the CALIOP/CALIPSO at 10-km along track with 5 km resolution data (Version 201). Data were averaged from runs from January to December 2007.

Download figure to PowerPoint

[11] Table 1 shows the seasonal variations of the clear-sky frequency at 10 km along-track resolution in each latitudinal band using the 5 km-resolution CALIOP data. On average, the clear-sky frequencies in the tropics and southern middle latitudes were lower than those in other latitudinal bands. The amplitudes of the seasonal variations of the clear-sky frequencies in subtropical regions were also determined to be higher by about 8% than those at other latitudes, possibly because the north-south movement of convective activity in the tropics affects the seasonal variations of clouds and clear skies in subtropical regions. The annual global average of the clear-sky frequency at 10 km along-track resolution derived from the CALIOP measurements was approximately 11.2%.

Table 1. Summary of the Clear-Sky Frequency From CALIOP Observationsa
 DJFMAMJJASONAnnual
  • a

    Observations are given in % and are at 10 km along-track resolution using daytime data from 2007, with 5 km resolution (Version 201). NH (SH), Northern (Southern) Hemisphere; DJF, December to February; and MAM, March to May; JJA, June to August; SON, September to November. Annual indicates January to December 2007.

Tropics (15°S–15°N)9.176.749.479.208.65
Subtropics in NH (15°N–30°N)17.8415.029.4912.8713.81
Subtropics in SH (15°S–30°S)8.139.1314.2212.3510.96
Middle Latitude in NH (30°N–60°N)16.7813.1713.0615.7314.69
Middle Latitude in SH (30°S–60°S)8.577.397.598.578.03
Globe (86°S–86°N)11.8610.7010.7311.3211.15

3.2. Spatial Variation of the Clear-Sky Frequency

[12] To test the spatial representativeness of the clear-sky frequency obtained from the CALIOP data, we calculated the clear-sky frequency for varying along-track aggregation lengths using the 1 km-resolution CALIOP data from June 2007 (Figure 2a). Just as for the CALIOP data, the clear-sky frequencies for varying aggregation lengths were derived from the contiguous aggregated MODIS data obtained with a 1 km × 1 km grid resolution during the same period. Figure 2a shows that the clear-sky frequencies derived from CALIOP (solid line) in each latitudinal band decreased exponentially with increasing aggregation length. Particularly in the tropics (star), the clear-sky frequency decreased from 33% for an aggregation length of 1 km to 10% for 20 km. In addition, the global value of the clear-sky frequency (open circle) for an aggregation length of 10 km was approximately 18% based on the CALIOP data, while it was approximately 25% based on the MODIS observations (dashed line in Figure 2a). The clear sky-frequencies derived from the MODIS data were approximately 5% higher than those based on the CALIOP measurements in all latitudinal bands, except in the tropics for shorter aggregation lengths.

image

Figure 2. (a) Distribution of the clear-sky frequency for aggregation lengths from 1 to 20 km in each latitudinal band derived from CALIOP data of 1 km resolution during daytime in June 2007 (Version 201), as shown by the solid lines. The dashed lines indicate the daytime clear-sky frequency derived from the MODIS cloud-mask data. (b) Distribution of the clear-sky frequency as a function of pixel area, from 1 km2 to 121 km2, derived from MODIS data with a 1 km × 1 km grid resolution during daytime in June 2007. The star, open square, asterisk, triangle, closed circle, and open circle represent the tropics, northern and southern subtropics, northern and southern middle latitude, and global average, respectively.

Download figure to PowerPoint

[13] In addition to the clear-sky frequency obtained by aggregating the along-track pixels, we also investigated the clear-sky frequency obtained by aggregating the pixels into squares of different sizes, using the 1 km × 1 km-resolution data from MODIS (Figure 2b). The clear-sky frequency decreased exponentially with increasing aggregated pixel size. For a pixel area of 78 km2, which is almost the same as the TANSO FTS IFOV, the global value of the clear-sky frequency (open circle) was approximately 20%. However, note that MODIS frequently overestimates the clear-sky frequency, as shown in Figure 2a.

3.3. Comparison of MODIS and CALIOP Data

[14] Figure 3 shows the latitudinal distributions of the frequencies of clear and cloudy skies in June 2007 derived from the CALIOP and MODIS data. The clear- and cloudy-sky distributions derived from MODIS (for the combined ‘Confident’ and ‘Probably’ categories) exhibit almost the same patterns as the equivalent distributions from CALIOP. In the tropics, the values of the cloudy and clear frequencies from CALIOP and MODIS were similar. In contrast, there were systematic biases between these distributions at higher latitudes; for example, the clear-sky frequency from MODIS was always higher than that from CALIOP, with the difference reaching a maximum of 10% in the southern subtropics. The latitudinal distribution of the frequency of ‘Probably Clear’ skies from MODIS was almost constant at approximately 5%, which implies a bias in the MODIS measurements. On the other hand, the frequency of ‘Probably Cloudy’ skies from MODIS was high over the cloudy regions between 15°S and 30°N, implying a high uncertainty associated with the MODIS data over cloudy regions.

image

Figure 3. Latitudinal distributions of the clear- and cloudy-sky frequencies [%] derived from the CALIOP and MODIS data during daytime from 1 to 30 June 2007. The thick solid (dotted) line shows the CALIOP clear-sky frequency (i.e., the combined frequency including both the ‘Confident’ and ‘Probably’ clear-sky frequency from MODIS). The thin solid (dotted) line represents the CALIOP cloudy-sky frequency (the ‘Confident’ and ‘Probably’ cloudy-sky frequency from MODIS). The long-dashed line represents the MODIS ‘Probably Cloudy’ category, the dot-dashed line indicates the MODIS ‘Probably Clear’ category.

Download figure to PowerPoint

[15] Horizontal maps showing of difference between the monthly mean clear-sky frequencies from MODIS and CALIOP in June 2007 (not shown) reveal: (i) the MODIS clear-sky frequency in the convective region over the ocean was lower than the CALIOP clear-sky frequency; and (ii) while MODIS predicted a higher clear-sky frequency than did CALIOP for land with a high clear-sky frequency, such as desert regions.

4. Summary and Discussion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Analysis
  5. 3. Results
  6. 4. Summary and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

[16] We investigated the distribution of the daytime clear-sky frequency using the CALIOP cloud-layer product data, with the aim of identifying optimal regions to retrieve CO2 and CH4 column concentrations from future GOSAT measurements. We could obtain accurate clear-sky information from CALIOP, a satellite-borne lidar that can detect thin clouds (0.01 < τ < 1.0). Also, we compared the clear-sky frequency from CALIOP with that derived from the MODIS cloud-mask data to gain information (particularly on the horizontal distribution) that could not be derived from CALIOP due to its small footprint and sparse sampling.

[17] The analysis of the CALIOP data shows that clear skies occur frequently over desert regions throughout the year and over land at northern mid-latitudes during the winter season. Over the ocean, the southern part of the tropical eastern Pacific has a relatively high clear-sky frequency throughout the year. The global clear-sky frequency derived from the CALIOP data was approximately 11%. Comparing the MODIS cloud-mask data to the cloud-layer data from CALIOP revealed that outside the Tropics the clear-sky (cloudy-sky) frequencies were overestimated (underestimated) by approximately 5% by MODIS data. We believe that this overestimation is caused by the insensitivity of MODIS to thinner clouds.

[18] Currently, in the nominal planned observation mode, GOSAT will obtain approximately 30,000 observational points in three days (data in the SWIR regime will only be taken during the daytime). In other words, if the clear-sky ratio is constant at about 11%, about 3300 data points will be available every three days for the retrieval of CO2 and CH4 column concentrations from the SWIR spectra if other conditions are met (e.g., the signal-to-noise ratio must be greater than 300, and the surface albedo must be between 0.1 and 0.5). That is, at maximum, GOSAT will be able to obtain SWIR spectra for CO2 and CH4 retrievals from approximately 10 observation points in each 7.5° × 7.5° (horizontal) grid box in three days over areas with high clear-sky frequencies, such as desert regions. On the other hand, over regions with low clear-sky ratios (less than 10%), such as over ocean, only two data points will be available in a grid box of the same size in three days. Because the uncertainty in the measured CO2 concentration becomes smaller in regions where many observations are available, the clear-sky distribution causes variations in the qualities of the retrieved CO2 and CH4 data.

[19] At present, several sites are being considered for validation of the forthcoming GOSAT measurements. For reliable validation, it is necessary to choose a site where we can obtain as many reliable data points as possible simultaneously from GOSAT and in situ measurements, for comparison. For example, we are currently working to establish candidate GOSAT validation sites at Tsukuba, Japan 140.1°E, 36.1°N), and Darwin, Australia 130.9°E, 12.4°S) (I. Morino and O. Uchino, GOSAT, personal communication, 2008). From the present study, we derived annual mean clear-sky frequencies of approximately 8% over Tsukuba and 15% over Darwin. However, note that other conditions, such as the roughness of the topography and the albedo, must also be considered when choosing a validation site, in addition to the clear-sky frequencies. We will investigate such issues in future research.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Analysis
  5. 3. Results
  6. 4. Summary and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

[20] We would like to thank Mark Vaughan at Science Systems and Applications Inc. (SSAI) for his critical comments and valuable suggestions. We would also like to thank two anonymous reviewers for their critical comments and valuable suggestions. This research is conducted for the GOSAT project of NIES, Japan.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Analysis
  5. 3. Results
  6. 4. Summary and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Analysis
  5. 3. Results
  6. 4. Summary and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information
FilenameFormatSizeDescription
grl25274-sup-0001-t01.txtplain text document1KTab-delimited Table 1.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.