The International Satellite Cloud Climatology Project (ISCCP) B1 data, which were recently rescued at the National Oceanic and Atmospheric Administration’s National Climatic Data Center (NOAA/NCDC), are a resource for the study of the earth’s climate. The ISCCP B1 data represent geostationary satellite imagery for all channels, including the infrared (IR), visible, and IR water vapor sensors. These are global 3-hourly snapshots from satellites around the world, covering the time period from 1979 to present at approximately 10-km spatial resolution. ISCCP B1 data will be used in the reprocessing of the cloud products, resulting in a higher-resolution ISCCP cloud climatology, surface radiation budget (SRB), etc. To realize the promise of a higher-resolution cloud climatology from the B1 data, an independent assessment of the calibration of the visible band was performed. The present study aims to accomplish this by cross-calibrating with the intercalibrated Advanced Very High Resolution Radiometer (AVHRR) reflectance data from the AVHRR Pathfinder Atmospheres–Extended (PATMOS-x) dataset. Since the reflectance calibration approach followed in the PATMOS-x dataset is radiometrically tied to the absolute calibration of the National Aeronautics and Space Administration’s (NASA) Moderate Resolution Imaging Spectroradiometer (MODIS) imager instrument, the present intercalibration scheme yields calibration coefficients consistent with MODIS. Results from this study show that the two independent sets (this study and the ISCCP) of results agree to within their mutual uncertainties. An independent approach to calibration based on multiyear observations over spatially and temporally invariant desert sites has also been used for validation. Results reveal that for most of the geostationary satellites, the mean difference with ISCCP calibration is less than 3% with the random errors under 2%. Another result is that this extends the intercalibrated record to beyond what ISCCP provides (prior to 1983 and beyond 2009).
The International Satellite Cloud Climatology Project (ISCCP; for a list of acronyms, please refer to the appendix) was established in 1982 as part of the World Climate Research Programme (WCRP), with the objective of collecting and analyzing the satellite radiance values from both the Polar-orbiting Environmental Satellites (POES) and the geostationary Earth orbit (GEO) meteorological satellites to infer the global cloud climatology and its impact on the earth radiation budget. Data collection began in 1983 (Schiffer and Rossow 1983) by setting up of sector processing centers (SPCs) worldwide that received the full-resolution raw GEO data from Geostationary Operational Environmental Satellite (GOES), Geostationary Meteorological Satellite (GMS), and Meteosat (Geosynchronous Meteorological Satellite) as well as data from National Oceanic and Atmospheric Administration (NOAA) polar obiters. Because of the computing, storage, and processing limitations at the time, data were subsampled to 10 km and 3-hourly resolution (creating so-called B1 data; Knapp 2008b), and set aside for archiving. The ISCCP Global Processing Center (GPC) was instead provided with further subsampled reduced resolution (30 km and 3 hourly) data (B2), leaving the B1 data languishing in the archive until 2003. NOAA’s National Climatic Data Center (NCDC; Knapp 2008b) initiated the rescue efforts for the B1 data by developing documentation, read software, preliminary data quality testing, and calibration. Full details of scientific data stewardship operations, description, and evolution of the B1 data are provided by Knapp (2008a,b).
ISCCP B1 data will be used in the reprocessing of the cloud products, resulting in a higher-resolution ISCCP cloud climatology, improved cloud detection (Rossow and Schiffer 1991; Rossow and Garder 1993), cloud optical depth, precipitation, surface albedo (Govaerts et al. 2008), and surface radiation budget, etc. Among other applications, the B1 data have been successfully employed in hurricane research and precipitation monitoring in data-sparse regions (Knapp et al. 2011; Kossin et al. 2007; Knapp and Kossin 2007). The primary common channels among the earlier GEOs were the visible (0.67 μm) and the infrared (IR) window (11 μm). The IR water vapor channel at 6.7 μm became available on later satellites. The IR channels have already been calibrated by ISCCP (Desormeaux et al. 1993) and by Knapp (2008a), and the focus of this study is the recalibration of the visible channel.
To retrieve accurate ISCCP cloud properties and radiative budget parameters, it is essential that all meteorological geostationary satellites be calibrated to a consistent standard. The visible sensors of all meteorological satellites are calibrated at prelaunch in the laboratory, and they do not have onboard visible calibration, thus showing a decrease of sensitivity during the postlaunch period. The ISCCP performs calibration of the GEO visible sensors every month by normalizing to the concurrent Advanced Very High Resolution Radiometer (AVHRR) on the afternoon NOAA polar-orbiting weather satellite at the same viewing geometry (Desormeaux et al. 1993) and anchors on multiple ER-2 calibration underflights (Brest and Rossow 1992; Brest et al. 1997). The morning NOAA polar orbiters have not been used in the ISCCP calibration. Among the other vicarious calibration methods, several investigators have used varying techniques. These range from intersatellite matchups with the low-Earth-orbiting (LEO) sensors, such as the Visible and Infrared Scanner (VIRS) instrument on board the Tropical Rainfall Measuring Mission (TRMM) satellite (Minnis et al. 2002); AVHRR (Heidinger et al. 2010); Moderate Resolution Imaging Spectroradiometer (MODIS; Wu and Sun 2005); observation of stable targets, such as deserts (Uprety and Cao 2010), stars (Bremer et al. 1998; Chang et al. 2005) and the moon (Stone et al. 2013); and deep convective clouds (Ham and Sohn 2010). Some of the vicarious calibration methods employ radiative transfer model to estimate calibration coefficients from numerical counts over ocean (Knapp and Vonder Haar 2000) and ocean, desert, and clouds (Govaerts et al. 2001). Studies based on monitoring of targets (Grau et al. 2002) sometimes yield only the degradation rate of the sensor responsivity during their lifetime and not the absolute calibration. The techniques and studies cited here form only a handful of those selected from the extensive literature on calibration.
Since ISCCP began in 1983, there was no calibration prior to that. Among the LEO sensors, AVHRR spans the longest time series of measurements, coinciding with the GEO time period, and hence offers the best source of calibration. But the calibration of the LEO solar channels has been pursued by many researchers, including ISCCP (Brest and Rossow 1992; Brest et al. 1997). Recently, an improved and MODIS-compatible AVHRR visible-channel-calibrated climate data record (CDR) product (Heidinger et al. 2010) has become available at NCDC. This is one of the most intersensor-consistent AVHRR data records, similar in quality to the one used by ISCCP in its calibration. The present study reports results of cross calibration with the CDR product and comparison with the ISCCP calibration. In addition, we perform independent assessments of calibration through stable desert locations.
2. Matchup data
The ISCCP data to be calibrated are composed of the archived B1U (B1 Unified format) files (Knapp 2008b) processed from raw data files received from the different SPCs all over the world, which differ from the raw satellite data. For example, the number of bit counts used by the visible channel of different satellites is not uniform. It varied sometimes even for the same satellite, such as GOES-7 from 7- to 8-bit counts. All the visible channel bit counts have been converted to standard 10-bit counts through shifting the required number of bits to the left. For example, conversion of an 8-bit count to a 10-bit count is accomplished by shifting 2 bits to the left, which is equivalent to multiplying by 4.
The reference data used in this study are the solar reflectance calibration in the AVHRR Pathfinder Atmospheres–Extended (PATMOS-x) dataset (Heidinger et al. 2010). These represent a set of calibration coefficients for all of the AVHRR sensors (including those in the morning orbits) derived from a consistent approach traceable to the MODIS standard. Channel 1 (0.63 μm) reflectance values have been derived employing four different sources, including 1) simultaneous nadir overpass (SNO) with MODIS for the MODIS era (years 2000+), 2) an Antarctic target, 3) a Libyan Desert target, and 4) AVHRR to AVHRR SNOs. The Antarctic and desert targets have been assumed to be radiometrically stable over time and have been used as reference during the pre-MODIS era after they have been characterized by MODIS. Thus, the methodology employed in the calibration of the solar reflectance channel in the PATMOS-x data is radiometrically tied to the MODIS imager and guarantees continuity in satellite-to-satellite transitions.
a. Calibration terminology
The integrated radiance L (W m−2 sr−1) over the visible spectral channel is expressed in terms of the detector-measured raw counts X and instrumental spectral response function as
where is the wavelength (μm), is the intercept, and is the calibration slope. It is also conventionally expressed in terms of the counts to space as
with m being the calibration coefficient at prelaunch. However, for the pre–GOES variable (GVAR) format or the GOES series of satellites prior to GOES-8, with onboard Visible and Infrared Spin Scan Radiometer (VISSR) instrument, the conversion of telemetry count values to radiance employed the squared response,
In addition to the pre-GVAR (GOES-1 to GOES-7) data, the GMS satellite series (GMS-3 to GMS-5) operated by the Japanese Meteorological Agency (JMA) also followed the squared response.
The scaled radiance is , where is the annual mean incident solar flux at the top of the atmosphere weighted by the instrumental spectral response. Thus,
with being the extraterrestrial solar flux at wavelength . The spectral response functions used here are identical to the ones used by ISCCP. The value of the incident solar flux, F0, has been obtained from the Thuillier et al. (2003) spectrum (provided in Table 1) and is also nearly identical to the ones used by ISCCP. The instrument-independent bidirectional or isotropic reflectance R is derived from the scaled radiance by normalizing with the cosine of the incident solar zenith angle, , and correcting for the mean Earth–sun distance d as
or alternatively, the radiance can be expressed as
b. Regression analysis
Calibration can be performed by matching up concurrent (spatial and temporal) imagery from the GEO and AVHRR. While the GEO comprises the 10-km 3-hourly ISCCP B1U data, the AVHRR channel 1 reflectance is derived from the level 2b PATMOS-x, which is on a 0.1° latitude–longitude grid. The matching criteria used here are 1) observations that happen within 10 min (ensures similar solar zenith angles); 2) a solar zenith angle of less than 60°; 3) a difference in cosine of satellite view angles less than 0.05; and 4) for off-nadir view angles (cosine of view zenith angle less than 0.9), a difference in azimuth angles less than 15°. For nadir views (cosine of view zenith angle greater than 0.9) reflected radiation is assumed isotropic and so no azimuth restriction is imposed.
Spatial and temporal matching of data from the two sensors, which are at 10-km resolution each, on an exact pixel-to-pixel level can be challenging, especially when the images can be spaced as far apart as 10 min. To ensure better correspondence in the general features of the imagery, both GEO and PATMOS data have been aggregated to a 50-km domain. The domain averages of AVHRR channel 1 reflectance is matched with the domain average of 10-bit count value of the GEO visible sensor. The individual 50-km matched (GEO and AVHRR) reflectances R [Eq. (4)] are assumed to be identical. They are similar but can vary slightly due to GEO and AVHRR spectral band differences. So, the absolute radiance L (W m−2) for GEO visible sensor can be obtained from Eq. (4) by employing as the extraterrestrial incident solar flux integrated over the GEO visible spectrum. Values for F0 used in the present study for different GEO sensors are provided in Table 1. Figure 1 shows an example of the correlation between the GOES-observed raw counts and the radiance L derived from Eq. (4) for GOES-8 satellite for September 1995. A linear regression adequately describes the correlation. This regression analysis is repeated for each GEO, yielding a slope and intercept value for each month similar to the ISCCP calibration tables (http://isccp.giss.nasa.gov/docs/calib.html).
Figure 2 shows a comparison of the time series of radiance calculated from the ISCCP calibration coefficients and the present approach for a 10-bit count value of 500 for GOES-7, GOES-10, Meteosat-7, and GMS-3 satellites. For the present scheme, no calibration coefficients have been calculated for months where the regression correlation is less than 0.9, or the number of data points was less than 500, or too few matchups were available for solar zenith angles less than 60°. The sensor degradation for GOES-10 and Meteosat-7 (discussed in the following section) tends to flatten out with time. The radiance values for the GMS-3 satellite are systematically higher by about 4%–5% than ISCCP.
c. Annual degradation of sensor
The standard calibration equation at prelaunch is given by Eq. (2) in terms of a single calibration coefficient m. Although the rate of degradation of the sensor response increases linearly with time initially, there is evidence (e.g., GOES-10) to show that it decreases after sufficient time in orbit. Hence, the calibration coefficient can be considered to increase with time in a quadratic or exponential form,
where Y is the number of years since launch, and the expression in the second parentheses in Eq. (5) or the term in Eq. (6) is a measure of the annual degradation of the sensor response. Annual degradation is the standard metric used for comparing the relative degradation of different satellite sensors or different postlaunch calibration methods.
However, the estimation of the annual degradation based on the published prelaunch calibration coefficient is misleading for many reasons. One reason is that a satellite often does not become operational soon after launch. For instance, the GOES-10 satellite experienced a solar array anomaly after launch and was held in storage for nearly 2 years until needed as a replacement for GOES-8 or GOES-9. The degradation should only be computed during the satellite operational time period and not include the prelaunch calibration. The sensor will be subject to launch contamination and a postlaunch exposure to the harsh space environment, causing a change in the prelaunch calibration coefficient. Another reason concerns the value of space counts or offset. The published calibration coefficient is valid for a specified space count value . While NOAA and the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT) propose predetermined space count values, some other calibration studies (Moulin et al. 1996; Chun et al. 2012) use their own value. Further, in the processing of the ISCCP B1 data, raw data were handled by as many as 15 different SPCs all over the world. Each of them received, processed, and delivered the data in unique ways (Knapp 2008b) and formats. It is believed that some SPCs even assigned a zero value to space look, resulting in a relativized count value and a higher effective calibration coefficient than the one published. For example, there were as many as four different SPC providers for the GOES-8 East satellite, which operated between 1994 and 2003 (Fig. 3). Each SPC is calibrated independently in this study, and since each SPC may apply the relativization differently, it may change the overall space count between SPCs. One can notice the varying rates of degradation for both ISCCP and the calibration derived from the present approach. Calibration for the Meteorological Satellite Center (MSC) has a flatter response than others. The present results are seen to be in close agreement with those of ISCCP. There is a large gap during the Atmospheric Environment Service (AES) period in 1998 due to nonavailability of matchups at solar zenith angles below 60°.
We employ for the present study a modified form of Eq. (5) for the absolute radiance without the constraint of a predetermined space count (for reasons mentioned in the preceding paragraph) as
where the term represents the space clamp value or the space offset count . Representing the calibration coefficient as a quadratic function in the number of years lapsed since launch as C(Y), one may write Eq. (7) relating the sensor digital output with effective radiance as
and the prelaunch calibration coefficient becomes C(0) , and the time variation of the calibration coefficient may be expressed as
The expression in the second parentheses in Eq. (8) constitutes the gain. The time series of radiance data derived from the monthly sets of slope and intercept coefficients have been further processed into the quadratic formulation [Eq. (7)] in terms of the time lapsed since launch for the entire count range through employing the nonlinear Lavenberg–Marquart scheme. The coefficients (see Table 1) may then be used to derive the time-dependent variation of the calibration coefficient C(Y) [Eq. (8)] for both ISCCP and the present scheme. The statistically fitted time variation of the calibration coefficient since launch obtained from the present scheme is compared with that of ISCCP in Fig. 4 for a sampling of satellites.
Note, however, that the ISCCP processing allows, by design, differing time variations of the intercept and slope values, which are listed for each month and employed in the processing. The statistical fitting to the form represented by Eq. (8) is merely for overall comparison of the present scheme with that of ISCCP and does not capture the details of the shorter-term variations in the calibration.
The present results (Fig. 4) reveal close agreement with the ISCCP calibration. Among the four sensors depicted in Fig. 4, GOES-7 shows the largest annual degradation rate of about 7%. Table 1 lists the coefficients of Eq. (7) for all GEOs along with the accompanying error statistics provided in the last two columns. There was no ISCCP calibration prior to 1983 for GOES-1 to GOES-4. Also, the spectral response data were unavailable for these satellites and a value for the incident solar flux, F0 (250 W m−2), over the visible spectral domain was assumed to perform the calibration. Calibration of some of the earlier Meteosat visible sensors up to and including Meteosat-6 (Govaerts et al. 1999) also suffers from a lack of reliable sensor spectral response. Except for a few of the earlier Meteosat and GMS satellites, the bulk of the calibrations agree within 3% bias and with a low RMS error of less than 2%.
4. Independent validations
Apart from the cross calibration with other satellites, there are several other vicarious or indirect calibration methods (Wu and Sun 2005). These involve use of empirical distribution functions (Crosby et al. 2005) and viewing of sources external to the satellite sometimes combined with radiative transfer modeling, etc. The external sources comprise stellar bodies like the moon or stars (Chang et al. 2005; Stone et al. 2013), or views of stable Earth targets, like deserts or deep convective clouds, etc. In the present study, use of the moon (Stone et al. 2013) and stable desert targets has been employed to monitor the response of the GEO visible channel.
a. Lunar calibration
The surface reflectance of the moon is exceptionally stable, considered invariant to under 1 part in 108 per year (Kieffer and Stone 2005) and thus well suited as a calibration target for the visible wavelengths. The moon regularly appears in the field of view of GEO imagers, and lunar images are captured periodically in the space viewing window. A methodology on the use of the moon as a stable radiometric source in the visible wavelengths has been developed at the U.S. Geological Survey (USGS) in Flagstaff, Arizona (Stone et al. 2013). A further independent assessment of ISCCP radiometric calibration for geostationary imagers by comparison of observations against the lunar reference and subsequent development of the imagery into long-term CDRs is under way by the USGS group. As of this writing, the coefficients for the time dependence of the calibration coefficient, similar to Eq. (8), have been provided for GOES-8 to GOES-13 visible imagers. Figure 5 shows a comparison of the time series of the calibration coefficient for GOES-8 to GOES-12 satellites with results from the present scheme, ISCCP, and lunar calibration overlaid. The discontinuous jumps observed in the curves for GOES-8 and GOES-12 are due to data received from different SPCs as discussed with reference to Fig. 3 earlier. However, for ISCCP the fit was done for the entire lifetime of the sensor and it is possible that these jumps were smoothed out. Nevertheless, the agreement is seen to be very good between all of them and even better between results from the present scheme and the lunar calibration.
b. Desert targets
Pseudoinvariant desert sites (i.e., those with relatively constant surface reflectance properties) have often been used to monitor and derive the calibration of in-orbit visible and infrared sensors (Rao and Chen 1995; Heidinger et al. 2003). The basic assumption underlying this method is that the reflectance of stable desert targets is spatially and temporally uniform and remains constant over time. Any trend detected by satellites over the sites can be attributed to the degradation of the instrument sensor with time.
In this study, we have chosen three desert sites (the Sonoran, Libyan, and the Simpson Deserts over the North American, African, and Australian continents, respectively) covering the orbits of GOES, Meteosat, and GMS series of satellites. Angal et al. (2010) have shown that deserts with partial vegetation exhibit seasonal changes in reflectance.
1) Site descriptions
The Sonoran Desert (32.25°N, 114.65°W) is a large flat land with partial vegetation located near the Arizona–Mexico border. Its altitude varies from sea level to about 1.2 km MSL and has been recommended as one of the best pseudoinvariant calibration sites over North America (Rao et al. 2003; Uprety and Cao 2010). It receives about 50–125 mm of precipitation from early summer to fall and during the late winter period every year. The Libyan Desert (22°N, 28.5°E) is a high-reflectance site located in Africa at an elevation of about 110 m MSL. It is made up of sand dunes with no vegetation and very little aerosol loading, if any. The site experiences minimal cloud cover with excellent spatial, spectral, and temporal uniformity. It has extensive spatial coverage and is widely recommended in calibration monitoring (Heidinger et al. 2003). Selection of a site over the Asia–Pacific region with convenient solar and view geometries for the GMS series of satellites proved to be quite challenging. After examining several desert sites over East Asia, the Simpson Desert (26.05°S, 137.15°E) in Australia was determined to be the best choice for the GMS and the Multifunctional Transport Satellite (MTSAT or MTS-1) series, which fulfilled the criteria of good spatial and temporal stability, brightness, and a high frequency of cloud-free observations with low solar and view zenith angles (Chun et al. 2012).
2) Characterization of desert sites
The entire time span (1979–2009) of the AVHRR PATMOS (Heidinger et al. 2010) data record available at the NOAA/NCDC archive, comprising 21 years of solar channel reflectance time series, has been employed to examine the spatial and temporal stability of the chosen deserts. The desert surface reflectance may exhibit bidirectional reflectance distribution function (BRDF)-related uncertainties at high solar zenith and view angles. Therefore, only data with solar zenith and satellite view angles below 50° were used. Although data with cloud fractions above 5% have not been used in this analysis, only those reflectance values with a standard deviation of less than 5% for the considered domain have been employed to reduce the effect of cloud contamination. Reflectance data from channel 1 for all years satisfying the above-mentioned criteria have been plotted for each desert site in Figs. 6–8. The top panel of each figure shows a regression line through the scatterplot representing the mean annual reflectance, while the bottom panels depict the histogram distribution. The Libyan and Sonoran Deserts do not exhibit a seasonal signal and the reflectance values remain stable without any discernible trend. However, the Simpson Desert (Fig. 8) displays a larger scatter than the other two, with a mean reflectance of about 29.4%. Also, the time series shows very little data between days 100 and 200 of each year, and a seasonal dip in the mean reflectance between about days 120 and 220. It was not immediately apparent what caused this dip in reflectance. The uncertainties relating to the characterization of the Simpson Desert are larger than those of the Sonoran and the Libyan Desert sites.
3) Calibration from desert monitoring
The reflectance characteristics for the three desert sites extracted from the AVHRR measurements (Figs. 6–8) may be employed to derive an independent calibration for the GEO satellites. We choose GOES-10, Meteosat-7, and GMS-5 satellites and examine the time series of observed raw counts over the Sonoran, Libyan, and Simpson Deserts, respectively. Cloudy scenes have been eliminated by discarding count values with a standard deviation of greater than 5%. Again, only solar zenith angles less than 50° have been used and any BRDF characteristics of the surface have been ignored in this instance. Equation (4) can be employed to determine the distribution of radiance over the desert region. Application of the nonlinear curve-fitting [Eqs. (7) and (8)] technique as before yields the time-dependent variation of the calibration coefficient and these are shown in Fig. 9 for all three satellites. The error statistics (mean bias and RMS) shown for the desert case are with respect to the present scheme and show favorable values of less than 3% and 2%, respectively.
A caveat of the desert-based calibration approach outlined here is its validity over a narrow count range. An interesting case in point here is the comparison of our calibration for the MTSAT-1R satellite visible sensor with that of an independent study (Chun et al. 2012), which also employed observations over the Simpson Desert in Australia. The latter study obtained the surface spectral reflectance of the Simpson Desert from the MODIS BRDF distribution functions after tuning against the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) spectral library and then input into a radiative transfer model. The simulated top-of-the-atmosphere radiance values were then applied to the calibration of the MTS-1 visible sensor for the period November 2007–June 2008. The published (Chun et al. 2012) monthly sets of slope and intercept values have been used to derive a time-dependent variation of the calibration coefficient as described before [Eq. (8)], and these are compared with results from the present scheme in Fig. 10. A space offset count of zero has been used for the MTS-1 satellite (D. R. Doelling 2013, personal communication). Although the results for the time variation of the calibration coefficient shown in Fig. 10 (solid line) here considered the full range of count values, it needs to be borne in mind that the slope and intercept values provided for individual months in the Chun et al. study considered only a partial range of count values (0–500) corresponding to the albedo value over the desert surface. The 17% underestimation of the calibration coefficient with reference to those from the present calibration study, which is applicable over the entire range of count values (0–1023), is consistent with the reported bias of up to 20% by Chun et al. (2012). In other words, results from the Chun et al. study seem to be valid only for the lower count range typical of deserts. Repeating the above-mentioned analysis after restricting the count range to be under 500 results in a dramatic reduction of the mean bias (Fig. 10, magenta-colored curves), consistent with the cited study. In fact, the bias swings from +17% to −3%, implying that the desert-target-based approach to calibration by itself will not be adequate and would result in an incorrect offset count with larger errors at the higher end of the count range. It is also relevant to note here that this situation seems unique to the MTS-1 satellite (Doelling et al. 2015). The MTS-1 visible image is known to be blurred by small radiance contributions from the surrounding 500 km for a given pixel, and it is necessary to mitigate this blurring effect to realize a linear sensor response (Khlopenkov et al. 2015; Doelling et al. 2015).
Finally, Fig. 11 shows a plot of over 25 years of reflectance over the Sonoran Desert (top panel of Fig. 11) and Libyan Desert (bottom panel of Fig. 11) derived using the monthly intercept and slope values from both ISCCP and present study. The mean and one standard deviation spread over the life of each sensor are shown. The agreement is observed to be excellent for the GOES series, while there are differences for the Meteosat-3 and Meteosat-5 where the ISCCP-derived reflectance values are larger than those from the present method. This is consistent with the differences shown in Table 1. Yet, the agreement is still within the estimated mutual uncertainties for the most part.
5. Summary and concluding remarks
The recent rescue of the ISCCP-B1 data (Knapp 2008b; Knapp et al. 2011) has provided climate researchers with a dataset rich in climate information for the period 1978 through the present. This dataset will be the basis of a future reprocessing of the ISCCP cloud climatology, resulting in a higher spatial resolution of the cloud properties and surface radiation budget. The infrared channels have been already calibrated, and the present study takes advantage of a recent CDR with well-calibrated AVHRR visible channel to verify and fill in gaps in the existing GEO visible calibration. Calibration of some of the pre-1983 GOES visible channel has been hampered due to nonavailability of the spectral response functions. Other independent modes of calibration, such as lunar views and pseudoinvariant target sites, have been used for validation. The time-dependent variation of the calibration coefficients determined for the ISCCP calibration agrees within 3% bias for most of the satellites studied here.
This work was supported by NOAA’s Climate Data Record Program through the Cooperative Institute for Climate and Satellites–North Carolina under Cooperative Agreement NA09NES4400006. The merged sets of ISCCP calibration coefficients (monthly intercept and slope values) were obtained from Joseph Ferrier at the Goddard Institute for Space Studies (NASA), New York, New York. The lunar calibration coefficients were provided by Thomas Stone of the U.S. Geological Survey (USGS) office, Flagstaff, Arizona. The authors are deeply indebted to Bill Rossow and Joe Ferrier for providing many valuable insights during the analysis of the results and to the anonymous reviewers, who helped significantly in improving the quality of the paper.
Glossary of Acronyms
Many of the acronyms used in the text were defined; additional expansions of institutions and model names are available online at http://www.ametsoc.org/PubsAcronymList.
AES Atmospheric Environment Service, Downsview, Ontario, Canada
AVHRR Advanced Very High Resolution Radiometer flown on NOAA polar orbiters
BRDF Bidirectional Reflectance Distribution Function
CDR Climate Data Record
CSU Colorado State University, Ft. Collins, Colorado, USA
ESA European Space Agency EUM EUMETSAT
EUMETSAT European Organization for the Exploitation of Meteorological Satellites
GEO Geostationary Earth Orbit Satellites (GOES, Meteosat, GMS)
GMS Geostationary Meteorological Satellite operated by JMA
GOES Geostationary Operational. Environmental satellite operated by NOAA GOESame as GOES
GVAR GOES Variable format
IR Infrared (11 µm) channel
ISCCP International Satellite Cloud Climatology Project
JMA Japan Meteorological Agency, Tokyo, Japan
LEO Low-Earth Orbit Satellite (TRMM)
MET Geosynchronous Meteorological Satellite operated by EUMETSAT
MSC Meteorological Satellite Center
MODIS Moderate Resolution Imaging Spectroradiometer imager
MTSAT (or MTS-1) Multifunction Transport Satellite
NASA National Aeronautics and Space Administration
NOA NOAA SPC
NOAA National Oceanic and Atmospheric Administration
POES Polar-orbiting Environmental Satellite (NOAA AVHRR, METOPS, etc)
PATMOS-x AVHRR Pathfinder Atmospheres Extended (PATMOS-x) dataset
SNO Simultaneous Nadir Observation
SPC Satellite Processing Center of ISCCP
TRMM Tropical Rainfall Measurement Mission
VIRS Visible Infrared Sensor on board the TRMM satellite
VISSR Visible Infrared Spin-Scan Radiometer flown on GMS and GOES