This article describes the PATMOS-x cloud climate data record, focusing on the methods used to minimize inter-satellite artifacts.
The National Oceanic and Atmospheric Administration (NOAA) Advanced Very High Resolution Radiometer (AVHRR) Pathfinder Atmospheres–Extended (PATMOS-x) project provides a new satellite-based climate dataset that is now available to the public. As the name implies, the focus of PATMOS-x is on atmospheric applications including clouds and aerosols. PATMOS-x also includes the calibrated AVHRR observations and selected ancillary data, which allow other applications and climate records to be generated from the PATMOS-x data. The NOAA and European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT) data record from the AVHRR spans from 1978 to the present and should extend at least through 2020. The AVHRR record therefore constitutes the longest global record from a consistent set of satellite imaging sensors.
Since 2011, PATMOS-x data have been hosted at the NOAA National Climatic Data Center (NCDC) as part of the NOAA Climate Data Record (CDR) program. Since 2006, PATMOS-x has appeared as a cloud climate record in the Bulletin of the American Meteorological Society (BAMS) State of the Climate (SOC) annual report. Starting in October 2013, PATMOS-x data will be updated daily at NCDC, and monthly anomaly maps similar to those shown annually in the BAMS SOC report will be generated routinely.
One goal of this paper is to describe the characteristics of the AVHRR PATMOS-x dataset. PATMOS-x has participated in the recent Global Energy and Water Cycle Experiment (GEWEX) Cloud Assessment (CA), which (Stubenrauch et al. 2012, 2013) includes a rigorous intercomparison of PATMOS-x against other datasets. PATMOS-x has also participated in a number of initiatives designed to improve calibration techniques as well as to make the record suitable for climate applications. These include the European Space Agency (ESA) Cloud Climate Change Initiative (CCI) (Hollmann et al. 2013), where PATMOS-x participated in a round-robin data comparison; the EUMETSAT Cloud Retrieval Evaluation Workshops (CREW) (Roebeling et al. 2013); the World Meteorological Organization Sustained, Coordinated Processing of Environmental Satellite Data for Climate Monitoring (SCOPE-CM) pilot project; and the Global Space-Based Inter-Calibration System (GSICS) program. PATMOS-x is one of a handful of long-term satellite-based cloud records, of which the International Satellite Cloud Climatology Project (ISCCP) is likely the best recognized. ISCCP provides grid-averaged data with a spatial resolution of 2.5° and varying temporal resolutions. ISCCP also provides two-dimensional histograms of cloud optical depth and pressure, which are based on pixel-level results. These histograms have been used widely to study cloud-type occurrence and weather states (Rossow et al. 2005). Key differences between the PATMOS-x and ISCCP products include ISCCP being primarily derived from geostationary satellite measurements, while PATMOS-x is derived from polar orbiters; the ISCCP product algorithms are based on two channels (0.63 and 11 μm), while PATMOS-x also uses the 1.6-, 3.75-, and 12-μm channels. Finally, the PATMOS-x standard product is pixel level, meaning no averaging has been performed, though it has been fit to a standard global grid. Other AVHRR-based cloud records exist as well, such as that of the EUMETSAT Satellite Application Facility on Climate Monitoring (CM SAF) project (Karlsson et al. 2013).
Another goal of this paper is to demonstrate to what extent an AVHRR-derived cloud climate data record can be free of sensor artifacts. In particular, the AVHRR record extends over 17 sensors and three decades, so an important goal of any AVHRR-based climatology has to be the demonstration of intersensor stability.
Finally, we should note that there are several methods of accessing PATMOS-x data, though the primary method is through NCDC. Currently, NCDC hosts the global PATMOS-x calibrated radiances spanning 1979–2009 (at www.ncdc.noaa.gov/cdr/operationalcdrs.html). Beginning in October 2013 this record will include a suite of cloud products that will update continuously (with an approximate 1-week lag from real time). Each file in the record includes a global daily scene of a single satellite for an ascending or descending node. The entire record numbers 50,000 files and 10 TB in size. The file format is netCDF with internal compression applied. The data are pixel level subsampled to a global 0.1° × 0.1° grid, meaning it is not averaged in any way. The NCDC infrastructure supports the Thematic Realtime Environmental Distributed Data Services (THREDDS) and FTP access. A few years covering specific domains are available at the PATMOS-x website (http://cimss.ssec.wisc.edu/patmosx/data/). Those files are in Hierarchical Data Format, version 4 (HDF4), and may be accessed via anonymous FTP.
THE ADVANCED VERY HIGH RESOLUTION RADIOMETER.
The AVHRR has flown as the meteorological imaging sensor on the NOAA polar-orbiting satellites since 1978. The AVHRR has also flown on EUMETSAT Meteorological Operational (MetOp) satellites since 2006. Cracknell (1997) provides a thorough history of the AVHRR sensor and its applications. Only a brief review is given here. The AVHRR's spatial resolution is 1.1 km at nadir and its swath width is 2500 km. Only the MetOp satellites provide full-resolution AVHRR data (1.1 km) globally. From the NOAA satellites, only the global area coverage (GAC) data are available globally. GAC data are generated by averaging raw sensor counts over a 3 × 5 array spread over three scan lines. Only four (1.1 km) pixels on the central scan line are averaged to compute the GAC value. The approximate resolution of GAC data is therefore 1 km × 4 km. GAC data are used in the AVHRR PATMOS-x data. When processing MetOp, GAC data from National Environmental Satellite, Data, and Information Service (NESDIS) are used.
The AVHRR makes observations in six spectral bands (channel 1 = 0.63 μm, channel 2 = 0.86 μm, channel 3a = 1.6 μm, channel 3b = 3.75 μm, channel 4 = 11 μm, and channel 5 = 12 μm), all of which are used in PATMOS-x. The AVHRR has had three versions. AVHRR/1 provided channels 1, 2, 3b, and 4. AVHRR/1 sensors flew on Television and Infrared Observation Satellite-N (TIROS-N), NOAA-6, NOAA-8, and NOAA-10. AVHRR/2 saw the addition of channel 5 (Schwalb 1978). NOAA-7, NOAA-9, NOAA-11, NOAA-12, and NOAA-14 flew AVHRR/2 sensors (Schwalb 1982). AVHRR/3 added channel 3a, but as the name implies, it shared its space in the raw count dataset with channel 3b because the AVHRR/3 was still limited to recording only five channels. NOAA-15 through NOAA-19 and all MetOp satellites flew AVHRR/3 sensors. On MetOp-A and MetOp-B, channel 3a was on during daytime operation and channel 3b was on at night. The same procedure was followed with NOAA-17. On NOAA-15, NOAA-18, and NOAA-19 channel 3a was not used. On NOAA-16, the channel 3a/3b day/night switch was used from launch until May 2003. After May 2003, NOAA-16 did not use channel 3a. As described by Heidinger et al. (2001), channels 3a and 3b provide significantly different information, and these differences are often evident in the AVHRR PATMOS-x time series.
The calibration of the AVHRR sensor has evolved over time. The climate community is fortunate that the GAC data provide much of the information to diagnose the instrument's performance and to improve its radiometric calibration. In PATMOS-x, channels 1, 2, and 3a are calibrated using the method outlined in Heidinger et al. (2010, 2002). An examination of the stability of this calibration is provided by Zhao et al. (2011) and Cermak (2010). The thermal calibration in PATMOS-x is derived from the blackbody temperature measurements and space views using the pathfinder method (Rao et al. 1993). The specific methodology for applying the thermal calibration is taken from Goodrum et al. (2009). PATMOS-x calibration procedure makes no correction for the solar contribution to the 3.75-μm channel. It is important to realize that the AVHRR was not designed with quantitative climate research in mind. Research in more advanced approaches at AVHRR calibration (i.e., Mittaz et al. 2009) is ongoing and will certainly lead to improved climate records.
Another important aspect of the AVHRR is its observation times. NOAA has typically flown AVHRR in an afternoon or morning orbit. The afternoon orbits cross the equator on their ascending (northward) node at approximately 1330 local time (LT). The morning orbits cross the equator on their descending (southward) node at approximately 0730 LT. The AVHRRs in afternoon orbits include TIROS-N, NOAA-7, NOAA-9, NOAA-11, NOAA-14, NOAA-16, NOAA-18, and NOAA-19. The AVHRRs in morning orbits include NOAA-6, NOAA-8, NOAA-10, NOAA-12, and NOAA-15. Starting with NOAA-17 and all MetOp satellites, AVHRR data are available from a midmorning orbit where the AVHRR crosses the equator in its descending node at approximately 0930 LT. Complications arise from changes in the equator crossing times of individual AVHRR sensors due to satellite drift. Figure 1 shows the equator crossing times for various AVHRR instruments for both ascending and descending nodes (Ignatov et al. 2004). For the AVHRR/1 and AVHRR/2 satellites, the equator crossing time varied uniformly and drifted away from noon/midnight. The NOAA satellites after NOAA-15 were launched to drift toward noon/midnight for roughly 2 yr before drifting later in the day. The MetOp satellites fly in a controlled orbit. The initial postlaunch ascending equator-crossing times of the NOAA afternoon satellites varied from 1330 to 1500 LT. Given that diurnal cloudiness variation is associated with a well-defined solar forcing (Bergman and Salby 1996; Cairns 1995; Dai and Trenberth 2004; Gray and Jacobson 1977), accounting for the varying observation times is an important challenge for PATMOS-x and any other climate data record based on the AVHRR. PATMOS-x algorithms are designed to operate consistently at all times of day. In addition, Foster and Heidinger (2013) demonstrate techniques to account for changes in observation time for time series of cloud amount and other cloud properties.
Equatorial crossing time of the NOAA and MetOp polar-orbiting satellite series spanning 1979–2012.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
Equatorial crossing time of the NOAA and MetOp polar-orbiting satellite series spanning 1979–2012.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
Equatorial crossing time of the NOAA and MetOp polar-orbiting satellite series spanning 1979–2012.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
In addition to spectral coverage and diurnal sampling, another important and varying aspect of the AVHRR data record is its global coverage. Figure 2 shows the variation over time of the mean number of times AVHRR viewed the globe per day. The fraction of the globe viewed is taken from the global attributes in the PATMOS-x files and does not account for the multiple views per day at high latitudes nor does it weight the result by latitude. The gray line in Fig. 2 represents the maximum number of possible daily views given the satellites in orbit for a given month, while the black dots represent the actual value.
Variation of the monthly-mean number of global views per day offered by the AVHRR sensors (gray line) and included in PATMOS-x (black circles). Gray line is based on the dates of operation of the AVHRR sensors and does not include gaps in the AVHRR level-1b GAC archive. PATMOS-x numbers include gaps in the GAC archive and data that could not be successfully processed.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
Variation of the monthly-mean number of global views per day offered by the AVHRR sensors (gray line) and included in PATMOS-x (black circles). Gray line is based on the dates of operation of the AVHRR sensors and does not include gaps in the AVHRR level-1b GAC archive. PATMOS-x numbers include gaps in the GAC archive and data that could not be successfully processed.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
Variation of the monthly-mean number of global views per day offered by the AVHRR sensors (gray line) and included in PATMOS-x (black circles). Gray line is based on the dates of operation of the AVHRR sensors and does not include gaps in the AVHRR level-1b GAC archive. PATMOS-x numbers include gaps in the GAC archive and data that could not be successfully processed.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
The difference between the two curves is mainly caused by the lack of GAC data from the NOAA archive. This is especially true for TIROS-N and NOAA-6 and NOAA-8. Another cause of data voids are scan lines that are set to missing by the error handling within the PATMOS-x software. These events occur mainly during scan motor failures late in the life of most sensors.
For most of the record, NOAA flew an afternoon and a morning orbiting satellite and therefore provided four global views per day. With the advent of the midmorning orbit in 2002, that number became six global views per day. The continued operation of NOAA-14 from 2000 to 2002 explains the increase after the launch of NOAA-16 in 2000. The continued operation of NOAA-18 after the launch of NOAA-19 in 2009 accounts for the values of eight global views per day after 2009. The maximum data coverage occurred in 2009 when NOAA-15, NOAA-17, NOAA-18, and NOAA-19 and MetOp-A were functioning. NOAA-17 failed in late 2009, however. The largest data gaps occur for NOAA-6 and NOAA-8 before 1985. Evident in Fig. 2 is the lack of NOAA-10 data in 1991, loss of NOAA-11 in late 1994, and the loss of NOAA-15 for some months in 2000.
COMPARISON OF PATMOS-X TO PATMOS.
PATMOS-x was built on the legacy of the original PATMOS project (Jacobowitz et al. 2003; Stowe et al. 2002) that was part of the NOAA/NASA Pathfinder initiatives of the 1990s. PATMOS processed the AVHRR GAC data from the afternoon orbiting sensors from 1981 to 1999 (AVHRR/2). The only cloud product from PATMOS was total cloud amount generated using the Clouds from AVHRR, phase 1 (CLAVR-1) cloud detection scheme (Stowe et al. 1999). The selected monthly-mean PATMOS products with a resolution of 1.0° are available from the NOAA Comprehensive Large Array-Data Stewardship System (CLASS) system (www.class.noaa.gov). The PATMOS project funded the transfer of the entire AVHRR GAC tape archive from NCDC to the NESDIS Satellite Active Archive (SAA). The SAA was the forerunner of the NOAA CLASS system and CLASS continues to serve all AVHRR GAC data freely to the public.
PATMOS-x is an extension of PATMOS in several aspects. First, PATMOS-x has been modified to handle the NOAA-K, NOAA-L, and NOAA-M (NOAA KLM) and EUMESAT MetOp series of AVHRR sensors (AVHRR/3), which allows PATMOS-x to operate after 1999. Also, PATMOS-x incorporates the morning orbiting sensors including AVHRR/1 (NOAA-6, NOAA-8, and NOAA-10) and NOAA-12 (AVHRR/2) and NOAA-15 (AVHRR/3). PATMOS-x also includes the midmorning orbiting AVHRR/2 sensors (NOAA-17 and MetOp-A). The inclusion of the other orbits greatly improves the diurnal sampling of PATMOS-x over PATMOS. Last, PATMOS-x includes a full suite of quantitative cloud products that are similar in content to those provided by the National Aeronautic and Space Administration (NASA) Earth Observing System Moderate Resolution Imaging Spectroradiometer (MODIS) products. PATMOS included no quantitative cloud products except for the total cloud amount. While PATMOS generated level-3 data, PATMOS-x employs what we call the level-2b file format, which is a hybrid between the level-2 and level-3 processing levels. In way of explanation, the level-1b processing level is generally composed of raw counts with information necessary for calibration or counts converted to sensor units [e.g., brightness temperature (BT) and reflectance]; the GAC data described in “The Advanced Very High Resolution Radiometer” section are considered to be at the level-1b processing level. The level-2 processing level maintains the same measurement resolution and geolocation as that of level 1b, but sensor calibration has been applied and the suite of cloud products has been derived (cloud height, optical depth, particle size, etc.). The level-3 processing level involves a standardized gridding system and some sort of merging of the pixel-level measurements; for example, several adjacent swaths may be merged to form a larger spatial domain and grid-averaged values may be derived from multiple measurements. The level-2b format is a hybrid, as it is a pixel-level product that has been sampled and fit to a 0.1° global longitude–latitude grid. The benefits of this are that the dataset is reduced in size, while maintaining the ability to perform pixel-level analysis.
Three versions of PATMOS-x data have been released. Version 4 was hosted at the University of Wisconsin Cooperative Institute for Meteorological Satellite Studies (CIMSS) and generated for the initial GEWEX cloud climatology assessment in 2008. Version 5.2 data were generated in 2010 and these data reside at NCDC. Version 5.3 is in production now and will be hosted at NCDC by the end of 2014. The examples shown here come from version 6. With the release of the MODIS Collection 6 data, the PATMOS-x AVHRR solar reflectance calibration will be updated, and this will likely result in a new dataset delivery to NCDC in 2014.
AVHRR PATMOS-X PRODUCTS AND ALGORITHMS.
As the name implies, the PATMOS-x project is focused on the generation of atmospheric products including cloud and aerosol information. The algorithms are described here briefly, but further detail is available in the reference publications. Comparisons to other cloud datasets are included in these publications. Table 1 provides an overview of the current AVHRR PATMOS-x product suite for the version of the data delivered to NCDC. In Table 1, the products are divided into 10 areas. The first are the calibrated observations for all AVHRR channels. The solar reflectance channel data are provided in terms of the isotropic reflectances, and the thermal channel data are reported in terms of brightness temperatures. Attributes in the PATMOS-x level-2b files allow for the conversion of these observations into radiances. As described later in the next section, the PATMOS-x level-2b data are composed of sampled pixels. To include information on the original texture of the pixel-level data (level 2), the standard deviations of the 0.63- and 11-μm observations computed over a 3 × 3 pixel array centered on the level-2b pixel are included.
Types of products contained in PATMOS-x
The second type of product included in the AVHRR PATMOS-x files is ancillary data. As described in the algorithm references, PATMOS-x employs ancillary data from various sources, though the primary source is 6-hourly National Centers for Environmental Prediction (NCEP) Climate Forecast System Reanalysis (CFSR) (Saha et al. 2010). The fields included in the PATMOS-x files are intended to facilitate the interpretation of the PATMOS-x products and to allow for further stratification of climate analyses. The ancillary data include land/sea classif ication, land-cover type, a snow/ice mask, and a coast mask.
A large part of the PATMOS-x processing involves the estimation of clear-sky observations using CFSR along with other ancillary data to perform fast radiative transfer model (RTM) calculations. The fast infrared model employed in PATMOS-x is the Pressure-Layer Fast Algorithm for Atmospheric Transmittance (PFAAST) (Hannon and McMillan 1996). The performance of this process is used to ascertain PATMOS-x product quality. Clear-sky 11-μm brightness temperature and the assumed background surface temperature used in the RTM are included output. Inspection of these fields for artifacts allows users to screen potentially bad data from their analysis.
The cloud detection algorithm produces a cloud probability that spans from 0 to 1 (Heidinger et al. 2012). This is used to generate a four-level cloud mask (clear, probably clear, probably cloudy, and cloudy). In addition, a seven-level cloud type is included based on the cloud mask and other spectral tests (Pavolonis et al. 2005). The current cloud types are clear, probably clear, near surface, water phase, super-cooled water phase, opaque ice phase, cirrus, multilayer cloud, and deep convective cloud. After a cloud is classified as one of these types, the AVHRR version of the Algorithm Working Group (AWG) cloud height algorithm (ACHA) (Heidinger and Pavolonis 2009) is used to estimate cloud-top temperature, 11-μm emissivity, and an infrared microphysical index β. In addition to cloud-top temperature, cloud-top height and cloud-top pressure are also provided and deduced from the NWP profiles. During the day, the daytime cloud optical and microphysical properties (DCOMP) algorithm generates the cloud optical depth and cloud effective radius products (Walther and Heidinger 2012). Both DCOMP and ACHA provide error estimates and quality flags. The error estimates from DCOMP and ACHA are those generated automatically by the optimal estimation (Rodgers 1976) mathematics used in the algorithms. The cloud detection uncertainty is given by deviations in the cloud probability from 0 or 1. No error estimate is generated by the cloud typing approach.
PATMOS-x also provides some parameters not specific to cloud remote sensing, including surface temperature estimated from atmospherically corrected and emissivity-adjusted 11-μm observations. This parameter is intended for use in diagnosing errors in cloud detection. Inspection of the retrieved surface temperature to the background value should allow users to perform additional cloud screening. The aerosol products are generated using the lookup tables used in the official NESDIS real-time AVHRR aerosol processing (Ignatov and Stowe 2002) and include the optical depths computed separately for the 0.63-, 0.86-, and 1.6-μm channels. Aerosol products using a different methodology have also been generated from the PATMOS-x level-2b files (Zhao et al. 2008) that are also hosted by NOAA/NCDC. Last, information on Earth's radiation budget is provided by an estimate of the outgoing longwave radiation (OLR) and the cloud 0.63-μm albedo and transmission estimates from DCOMP.
In summary, PATMOS-x includes parameters not strictly classified as cloud or aerosol products with the intent to allow users to perform additional filtering and to generate additional products as desired. Figures 3 and 4 show selected products from the AVHRR PATMOS-x level-2b file for day 239 of year 2010 for the ascending node of NOAA-18 over the United States. Figure 3 shows selected radiometric products. Figure 3a shows the 0.63-mm reflectance, and Fig. 3b shows the 11-mm brightness temperature. PATMOS-x also contains some spatial statistics computed from the level-2 data in order to preserve information on the texture of the original data. Figure 3c shows the 11-mm standard deviation computed from a 3 × 3 array of level-2 pixels. Also included in PATMOS-x level-2b is the simulated clear-sky 11-μm brightness temperature. Figure 4 shows selected cloud products from PATMOS-x such as the cloud mask (Fig. 4a), the cloud type (Fig. 4b), cloud-top pressure (Fig. 4c), and the cloud optical depth (Fig. 4d). All of these data come from the standard 0.1° AVHRR PATMOS-x level-2b data format.
Example PATMOS-x 0.1° level-2b radiometric products viewed over the conterminous United States from NOAA-18 AVHRR from day 239 of year 2010. (a) The 0.63-μm reflectance, (b) the 11-mm BT, (c) the standard deviation of the 11-μm BT computed at the original resolution, and (d) a simulated estimate of the clear-sky 11-μm BT.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
Example PATMOS-x 0.1° level-2b radiometric products viewed over the conterminous United States from NOAA-18 AVHRR from day 239 of year 2010. (a) The 0.63-μm reflectance, (b) the 11-mm BT, (c) the standard deviation of the 11-μm BT computed at the original resolution, and (d) a simulated estimate of the clear-sky 11-μm BT.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
Example PATMOS-x 0.1° level-2b radiometric products viewed over the conterminous United States from NOAA-18 AVHRR from day 239 of year 2010. (a) The 0.63-μm reflectance, (b) the 11-mm BT, (c) the standard deviation of the 11-μm BT computed at the original resolution, and (d) a simulated estimate of the clear-sky 11-μm BT.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
Example PATMOS-x 0.1° level-2b cloud products viewed over the conterminous United States from NOAA-18 AVHRR from day 239 of year 2010. (a) The four-level cloud mask, (b) the cloud type, (c) the cloud-top pressure from ACHA, and (d) the optical depth from DCOMP.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
Example PATMOS-x 0.1° level-2b cloud products viewed over the conterminous United States from NOAA-18 AVHRR from day 239 of year 2010. (a) The four-level cloud mask, (b) the cloud type, (c) the cloud-top pressure from ACHA, and (d) the optical depth from DCOMP.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
Example PATMOS-x 0.1° level-2b cloud products viewed over the conterminous United States from NOAA-18 AVHRR from day 239 of year 2010. (a) The four-level cloud mask, (b) the cloud type, (c) the cloud-top pressure from ACHA, and (d) the optical depth from DCOMP.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
Known PATMOS-x limitations.
The PATMOS-x cloud detection, DCOMP, and ACHA algorithms all include uncertainty estimates that should be used in any analysis, and these are described in the references to these algorithms. However, it is important to summarize some of the known limitations at this time. The AVHRR lacks some spectral bands on MODIS that improve cloud optical depth estimation over snow, and therefore the AVHRR PATMOS-x DCOMP results over snow are much lower in quality than their MODIS equivalents and should be used with caution. In addition, cloud detection in polar regions is highly uncertain with the limiting spectral information from the AVHRR, and we advise caution in performing long-term trend analysis of the PATMOS-x AVHRR results in polar regions. The switch from channel 3a to 3b on NOAA-16 during 2003 has also introduced artifacts in time series using the DCOMP cloud effective radius. The AVHRR navigation accuracy has varied substantially and at times that is evident in the products especially near coastlines, and any climate analysis that involves small-scale regions with coastal boundaries should account for these added uncertainties.
PATMOS-X DATA PROCESSING.
One of the challenges facing any climate data record is reducing the data size to an acceptable value while maintaining the information content required for climate studies. This is particularly difficult for cloud climatologies since clouds vary at fine temporal and spatial scales, and have properties that vary in nonlinear ways. The technical aim in defining the PATMOS-x data format is to generate data of a manageable size that is flexible enough to serve a wide variety of applications.
Traditionally, cloud climate records are provided as spatially and perhaps temporally averaged fields referred to as level 3. In the GEWEX CA data (http://climserv.ipsl.polytechnique.fr/gewexca/), the level-3 spatial resolution is 1° × 1° and the temporal resolution is 1 month. As stated, averages of cloud properties over space and time can be misleading. For example, the vertical distribution of cloudiness is often bimodal, and direct averaging of cloud heights generates a mean value where very little cloud was ever observed. The negative impacts of averaging can be reduced by segregating the clouds by phase, height, and time of day. Following the GEWEX CA prescriptions for level-3, the PATMOS-x submission consisted of 92 cloud parameters.
In the PATMOS-x project, the size of the PATMOS-x level-2 files approached 80 TB, which far exceeded the storage requirements of the project at the time. However, PATMOS-x needed a flexible dataset to accomplish the goals of generating the GEWEX CA files that evolved over time. The PATMOS-x team therefore sought a solution that reduced the storage requirements while maintaining the flexibility to generate the GEWEX CA files in the short term and serve as a useful dataset to the community in the long term.
Level-2b description.
To resolve these issues, the PATMOS-x record uses the level-2b file format. PATMOS-x level-2b is generated by sampling the level-2 values on a fixed latitude and longitude grid with a nominal resolution of 0.1°. One level-2b file consists of either the ascending or descending orbits for a satellite in a single day and covers the entire globe. A global level-2b grid is composed of 6,485,401 points (3601 longitude × 1801 latitude). In contrast, a global view of AVHRR GAC data includes 40,082,000 pixels, while AVHRR FRAC data include 602,112,000 pixels, meaning the level-2b format reduces the number of GAC pixels by a factor of 6 and the number of full-resolution area coverage (FRAC) pixels by a factor of 92. However, because of the changing gridcell size on an equal angle spherical map projection and satellite swath geometry, this reduction is not equally distributed. Figure 5a shows the number of level-2 GAC pixels that falls in the range of a 0.1° grid point for the descending orbits of 1 day. Oversampling is up to a factor of 12 in the tropics at the highest sensor zenith. In polar regions the spatial extent of the 0.1° grid boxes decreases, meaning fewer measurements fall within each grid cell. However, multiple orbits over the polar region still lead to oversampling up to a factor of 16.
Pixel selection of level-2b data taken 1 Aug 2006. (a) Number of GAC level-2 observations for which nominal midpixel coordinates fall within a level-2b grid cell. (b) Observation time of descending NOAA-18 level-2b data. (c) Sensor zenith angle of descending NOAA-18 level-2b data.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
Pixel selection of level-2b data taken 1 Aug 2006. (a) Number of GAC level-2 observations for which nominal midpixel coordinates fall within a level-2b grid cell. (b) Observation time of descending NOAA-18 level-2b data. (c) Sensor zenith angle of descending NOAA-18 level-2b data.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
Pixel selection of level-2b data taken 1 Aug 2006. (a) Number of GAC level-2 observations for which nominal midpixel coordinates fall within a level-2b grid cell. (b) Observation time of descending NOAA-18 level-2b data. (c) Sensor zenith angle of descending NOAA-18 level-2b data.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
The level-2 pixels used for level-2b data are selected in a two-step process. The first step is the spatial selection of the pixels within a given orbit. The next step handles the selection of pixels in regions where multiple orbits overlap. For the first step the standard PATMOS-x product employs a “nearest neighbor” method that selects the pixel nearest to the level-2b position (within 0.1°). An alternative option randomly selects a pixel in the same circle and is referred to as the “random” option. The nearest neighbor method has the advantage of more homogeneous spacing between selected pixels. This allows a better use of textural or spatial analysis of clouds from level-2b data. However, random selection ensures that the level-2b pixels are representative of the entire region surrounding the level-2b point. This is not the case for the nearest neighbor method, which may preferentially underrepresent or overrepresent small-scale features such as coastlines or cities depending on their alignment with the level-2b grid.
The second level-2b sampling step deals with the selection of pixels in regions where multiple orbits overlap. In the AVHRR, the data from consecutive orbits touch at the equator. As the distance from the equator increases, the amount of overlap increases. Near the poles, all of the area is viewed multiple times by a single AVHRR sensor in 1 day. The standard PATMOS-x product employs a “nadir overlap” method that involves the preferential selection of pixels with a lower-viewing zenith angle (most nadir). The spatial cloud structures are not necessarily conserved with this method because the observations may come from several different hours of the day. To reduce this effect and avoid oscillating patterns in orbit selection, the earliest orbit of the day is initially set and only replaced if another orbit shows solar zenith angle more than 5° lower.
The standard PATMOS-x data hosted at NCDC use the nearest neighbor spatial sampling and the nadir overlap methods at a 0.1° resolution, and thus we will focus our analysis on the results of those methods and resolution. We should note, however, that the source code for the PATMOS-x processing system is publicly available and contains options for processing files using random pixel and orbit selection as well as variable spatial resolution. Figure 5 shows an example of this pixel selection for the descending orbit of NOAA-18 where local overpass time is approximately 0130 LT. Figure 5b illustrates the UTC for the given the pixel selection process. The spatial coherency and lack of oscillating patterns seen in overlapping orbit regions outside of the polar regions is a result of the 5° cushion used in conjunction with the “most nadir” orbit selection. Figure 5c shows the sensor zenith angle, which differs from nadir view (zenith equals 0) to up to 65° at the edges of a swath.
PATMOS-x retrieval algorithms were designed to be insensitive to viewing geometry, but the uncertainty in the ability to model atmospheric and cloud radiative transfer grows with increasing viewing zenith angle. The choice of taking the most nadir views in regions of orbital overlap makes the assumption that the selection of more accurate retrievals is more beneficial than the preservation of uniform viewing zenith angle sampling at all latitudes. This same decision was made in the PATMOS data. PATMOS-x users should be aware of the latitudinal variation of the viewing zenith angle distribution. These impacts can be reduced if users filter out data with viewing zenith angles larger than 30° when making climate analyses.
Impact of level-2b spatial sampling.
As stated above, 0.1° level-2b data represent a significant reduction in the number of values reported in the dataset. One obvious question is what impact this data culling has on the ability to preserve the distributions of the pixel values relative to those in the level-2 data. To answer this, a data granule from NOAA-18 was used to generate pixel distributions within each 1° × 1° box completely filled by the data. This granule was taken from an orbit from year 2006 that began at 0515 UTC and ended at 0703 UTC on day 218 and covered the region between 60°S–60°N and 60°–135°W and included 1.6 million level-2 pixels and 0.3 million level-2b pixels. The results of this analysis were repeated for several orbits and did not vary significantly. In each box, there were 121 level-2b values. The number of level-2 values ranged from 625 at nadir to 330 near the edge of scan for this example. The shift in level-2 pixel numbers is determined by the growth in pixel size with increasing sensor zenith angle. This analysis will be limited to the 0.63-μm reflectance and the 11-μm brightness temperature. It is assumed that these results apply to the PATMOS-x products as well. To compare the level-2 and level-2b distributions within each box, their first two moments were computed. In addition, the minimum and maximum values of the distributions were compared to judge the ability of level 2b to capture the full range of the level-2 values.
Figure 6 shows the results of comparing the first two moments and the dynamic range for the 0.63-μm reflectance values. The mean values are in very good agreement over the whole range. The variance results show a high correlation with little bias, though the scatter increases with increased variance. As expected for the 0.63-μm reflectance, the minimum values over the ocean should be clear sky and well captured by the level-2b sampling; Fig. 6 confirms this. The maximum reflectance would be provided by clouds, and because of their fine spatial structure, the potential for the level-2b sampling to miss the brightest cloud in the box is much more likely than missing the darkest value. For this wide swath of the Pacific, the mean reduction in the brightest 0.63-μm reflectance is roughly 6% with some values exceeding 20%. However, these pixels are rare enough to have little impact on the mean.
Scatterplots of metrics of the 0.63-mm reflectance distributions computed from 1° × 1° cells using level-2 pixels (x axis) and 0.1° level-2b pixels (y axis). Data taken from the ascending node of the NOAA-18 orbit named NSS.GHRR.NN.D06218.S0515.E0703.B0624344.WI. Analysis region is from 60°S to 60°N and from 60° to 150°W.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
Scatterplots of metrics of the 0.63-mm reflectance distributions computed from 1° × 1° cells using level-2 pixels (x axis) and 0.1° level-2b pixels (y axis). Data taken from the ascending node of the NOAA-18 orbit named NSS.GHRR.NN.D06218.S0515.E0703.B0624344.WI. Analysis region is from 60°S to 60°N and from 60° to 150°W.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
Scatterplots of metrics of the 0.63-mm reflectance distributions computed from 1° × 1° cells using level-2 pixels (x axis) and 0.1° level-2b pixels (y axis). Data taken from the ascending node of the NOAA-18 orbit named NSS.GHRR.NN.D06218.S0515.E0703.B0624344.WI. Analysis region is from 60°S to 60°N and from 60° to 150°W.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
Figure 7 shows the comparison of the moments and the ranges for the distributions of the 11-μm brightness temperatures within each 1° × 1° box. Whereas variation in the 0.63-μm reflectance drives variations in cloud optical depth, the variation in the 11-μm BT drives cloud temperature and emissivity. The distributions of 11-μm BT are expected to show less variation, and inspection of Fig. 7 confirms this. The variance metric is much better captured with 11-μm BT distributions than 0.63-μm reflectance. For 11-μm BT, maximum values will be associated with clear sky, and the level-2b maximum values compare well with the level-2 values. The minimum value of 11-μm BT is associated with clouds, and as was the case with the maximum 0.63-μm reflectance, the level-2b minimum 11-μm BT is more likely to deviate from the level-2 value. The minimum 11-μm BT is on average 1.3 K warmer than the level-2 minimum. In summary, Figs. 6 and 7 indicate that the level-2b sampling retains information necessary to recreate distribution of pixels at level-2 resolution. As the spatial scale of the box increases (i.e., 2.5° × 2.5°), the number of level-2b pixels increases, as does the ability to reproduce the level-2 distribution accurately.
As in Fig. 6, but for 11-μm BT.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
As in Fig. 6, but for 11-μm BT.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
As in Fig. 6, but for 11-μm BT.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
CONSISTENCY OF SELECTED AVHRR PATMOS-X TIME SERIES.
A cornerstone of environmental records suitable for climate applications is consistency. Changes in calibration, spatial, and temporal coverage, or in dependencies such as ancillary data and radiative transfer (RT) models, can have nontrivial effects on long-term trends. Consistency challenges faced by the AVHRR record include lack of onboard calibration and satellite drift (see Fig. 1). Significant work has been done to make the calibration consistent (Heidinger et al. 2010) and to correct for satellite drift (Foster and Heidinger 2013). The AVHRR record currently spans 17 separate sensors, so here we examine whether this consistency is maintained from satellite to satellite. To accomplish this we use transitional areas where concurrent satellite measurements are available. Figure 8 shows the AVHRR PATMOS-x mean monthly total cloudiness anomalies over the North Pacific domain (15°–35°N, 140°–120°W), and Fig. 9 shows 0.63-μm total-sky albedo, which is the cloud albedo weighted against surface albedo for clear-sky portions of a scene. The time series are normalized to account for phases of ENSO using a polynomial fit to the North Pacific sea level pressure (Trenberth and Hurrell 1994). Deseasonalized anomalies are calculated by removing monthly means. For much of the early record we only have one operational satellite, but by 1983 NOAA-7 and NOAA-8 are in orbit concurrently and we are able to begin evaluating overlapping satellite months. Over the course of the entire record there are over 450 overlapping satellite months with which to work. The difference in North Pacific cloudiness and albedo for each of these pairings was calculated as an absolute value, and the median and standard deviation of these differences, along with the correlation between satellite values, are found at the bottom of Figs. 8 and 9. For cloudiness the median absolute difference is 0.012, and the standard deviation is 0.013. The correlation between satellite pairings is 0.94 and was slightly higher before removing the ENSO signal. For albedo the median absolute difference and standard deviation is 0.018, while the correlation is 0.80. At least part of the decreased correlation in albedo is likely due to 3D effects associated with solar zenith angles. It should be noted that NOAA planned the local equatorial crossing time of its polar orbiters to maximize temporal coverage, meaning concurrent months are almost entirely composed of satellites in different orbits (generally 0130/1330 and 0730/1930 LT). The exceptions to this are NOAA-18 and NOAA-19, both flying in the 0130/1330 LT orbits, and NOAA-17 and MetOp-A, both in a 0930/2130 LT orbit. Synoptic-scale variability therefore is a factor when considering the difference between concurrent satellites. An estimate of natural variability is calculated by taking the standard deviation of the daily averages that go into each month. Shading in Figs. 8 and 9 represents this variability and shows that differences between concurrent satellites are small in comparison.
Time series of AVHRR PATMOS-x monthly cloudiness anomalies over the North Pacific. Shading represents uncertainty estimates, calculated using the standard deviation of the daily averages for each month. Time series has been normalized to the mean sea level pressure to account for ENSO effects, and the monthly averages have been removed to account for seasonal effects. The median, standard deviation, and correlation for all pairs of overlapping satellite months are located in the bottom-right corner. For months with more than two satellites available, all pairing combinations are included.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
Time series of AVHRR PATMOS-x monthly cloudiness anomalies over the North Pacific. Shading represents uncertainty estimates, calculated using the standard deviation of the daily averages for each month. Time series has been normalized to the mean sea level pressure to account for ENSO effects, and the monthly averages have been removed to account for seasonal effects. The median, standard deviation, and correlation for all pairs of overlapping satellite months are located in the bottom-right corner. For months with more than two satellites available, all pairing combinations are included.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
Time series of AVHRR PATMOS-x monthly cloudiness anomalies over the North Pacific. Shading represents uncertainty estimates, calculated using the standard deviation of the daily averages for each month. Time series has been normalized to the mean sea level pressure to account for ENSO effects, and the monthly averages have been removed to account for seasonal effects. The median, standard deviation, and correlation for all pairs of overlapping satellite months are located in the bottom-right corner. For months with more than two satellites available, all pairing combinations are included.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
As in Fig. 8, but for 0.63-μm total-sky albedo.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
As in Fig. 8, but for 0.63-μm total-sky albedo.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
As in Fig. 8, but for 0.63-μm total-sky albedo.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
Two additional points to consider when evaluating the consistency of the AVHRR PATMOS-x record: 1) are there biases in the record related to individual satellites, and 2) are differences in concurrent satellites, albeit small, occurring in a systematic way that would affect long-term trends? An example of the first sort might be the failure of specific channels on sensors, such as switching off the 3.75-μm channel in NOAA-16 for parts of 2002 and 2003. To test this we compare the North Pacific time series of cloud amount to an independent dataset; in this case we use the Interim European Cent re for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-Interim) (Dee et al. 2011), which does not use AVHRR measurements as part of its observation ingestion. The top panel of Fig. 10 shows the results of this comparison. The shape of the time series for AVHRR and ERA-Interim are in good agreement, but there is a positive offset in the AVHRR PATMOS-x record compared to that of the ERA-Interim of ~0.09. This is not unexpected, as what constitutes cloud in a record is more a function of the spectral sensitivity of a sensor than of common standards (though this is an interesting topic in and of itself that has many implications).
(top) Time series of monthly cloudiness for AVHRR PATMOS-x (black line) and the ERA-Interim (gray line) over the North Pacific. The slope of a linear fit to each time series is found at the bottom in the corresponding color. (bottom) Cloudiness difference for those months with overlapping satellite measurements. The difference is calculated by subtracting the satellite with the later launch date from the satellite with the earlier launch date.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
(top) Time series of monthly cloudiness for AVHRR PATMOS-x (black line) and the ERA-Interim (gray line) over the North Pacific. The slope of a linear fit to each time series is found at the bottom in the corresponding color. (bottom) Cloudiness difference for those months with overlapping satellite measurements. The difference is calculated by subtracting the satellite with the later launch date from the satellite with the earlier launch date.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
(top) Time series of monthly cloudiness for AVHRR PATMOS-x (black line) and the ERA-Interim (gray line) over the North Pacific. The slope of a linear fit to each time series is found at the bottom in the corresponding color. (bottom) Cloudiness difference for those months with overlapping satellite measurements. The difference is calculated by subtracting the satellite with the later launch date from the satellite with the earlier launch date.
Citation: Bulletin of the American Meteorological Society 95, 6; 10.1175/BAMS-D-12-00246.1
One point of interest in the top panel of Fig. 10 is that the AVHRR PATMOS-x record experiences a decrease in cloudiness of the magnitude −0.011 cloudiness per decade, while the ERA-Interim shows an increase of 0.013 cloudiness per decade. This relates to the second point of whether the small differences in concurrent satellites could occur in a systematic way so as to bias a long-term trend in the AVHRR PATMOS-x record. To test this we calculated the difference in measured total cloudiness between concurrent satellite pairs, but instead of looking at the absolute difference we subtracted the monthly-mean cloudiness of the satellite with the later launch date from that of the earlier launch date. The bottom panel of Fig. 10 shows the results of this analysis. We see that early in the record there are systematic differences between the satellites: NOAA-8 is consistently higher than NOAA-7, and NOAA-10 appears to be consistently higher than NOAA-9 and lower than NOAA-11. As the record progresses the differences tend to decrease and become more evenly distributed between positive and negative. In terms of long-term trend, a linear fit to this data shows a trend of −0.002 cloudiness per decade, which is close to a flat slope and much smaller than the −0.011 cloudiness per decade shown by the entire AVHRR PATMOS-x record. We conclude that differences in concurrent satellites are not driving the long-term trends, though further research as to the source of the difference in the early record satellites is suggested.
CONCLUSIONS.
PATMOS-x represents an evolution in the NESDIS effort to extract climate records from the AVHRR sensor record that began with PATMOS. PATMOS-x has migrated away from the traditional large-scale averages in level-3 data used in PATMOS-x to the mapped and sampled level-2b data format. The level-2b format offers the convenience of being mapped to a common grid, but has the drawbacks of duplication of data of one latitude column at the date line and oversampling in tropical and polar regions. The PATMOS-x data contents and format were designed to give climate researchers a flexible, powerful, and yet condensed set of climate data records. Being composed of data from 17 sensors, the demonstration of intersatellite consistency is critical. Analysis of the sensor-to-sensor biases and comparisons to time series from reanalysis data indicate that sensor-to-sensor differences are not driving factors in the PATMOS-x time series. AVHRR PATMOS-x data continue to be hosted by NCDC. Application of the PATMOS-x data concept to other sensors is being conducted with the hopes these other sensors can aid in the interpretation of the unique and multidecadal records from the AVHRR.
ACKNOWLEDGMENTS
We appreciate the support of the NOAA Climate Data Record (CDR) program. The views, opinions, and findings contained in this report are those of the author(s) and should not be construed as an official National Oceanic and Atmospheric Administration or U.S. government position, policy, or decision.
REFERENCES
Bergman, J. W., and M. L. Salby, 1996: Diurnal variations of cloud cover and their relationship to climatological conditions. J. Climate, 9, 2802–2820.
Cairns, B., 1995: Diurnal variations of cloud from ISCCP data. Atmos. Res., 37, 133–146.
Cermak, J., M. Wild, R. Knutti, M. I. Mishchenko, and A. K. Heidinger, 2010: Consistency of global satellite-derived aerosol and cloud data sets with recent brightening observations. Geophys. Res. Lett., 37, L21704, doi:10.1029/2010GL044632.
Cracknell, A. P., 1997: The Advanced Very High Resolution Radiometer (AVHRR). CRC Press, 968 pp.
Dai, A., and K. E. Trenberth, 2004: The diurnal cycle and its depiction in the Community Climate System Model. J. Climate, 17, 930–951.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, doi:10.1002/qj.828.
Foster, M. J., and A. K. Heidinger, 2013: PATMOS-x: Results from a diurnally corrected 30-yr satellite cloud climatology. J. Climate, 26, 414–425.
Goodrum, G., K. Kidwell, and W. Winston, cited 2009: NOAA KLM user's guide with NOAA-N, -N? supplement. NOAA. [Available online at http://www2.ncdc.noaa.gov/docs/klm/cover.htm.]
Gray, W. M., and R. W. Jacobson, 1977: Diurnal variation of deep cumulus convection. Mon. Wea. Rev., 105, 1171–1188.
Hannon, S. L. L. S., and W. W. McMillan, 1996: Atmospheric infrared fast transmittance models: A comparison of two approaches. Optical Spectroscopic Techniques and Instrumentation for Atmospheric and Space Research II, P. B. Hays and J. Wang, Eds., International Society for Optical Engineering (SPIE Proceedings, Vol. 2830), 94, doi:10.1117/12.256106.
Hansen, M., R. DeFries, J. R. G. Townshend, and R. Sohlberg, 2000: Global land cover classification at 1 km resolution using a decision tree classifier. Int. J. Remote Sens., 21, 1331–1365.
Heidinger, A. K., and M. J. Pavolonis, 2009: Gazing at cirrus clouds for 25 years through a split window. Part I: Methodology. J. Appl. Meteor. Climatol., 48, 1100–1116.
Heidinger, A. K., C. Cao, and J. T. Sullivan, 2002: Using Moderate Resolution Imaging Spectrometer (MODIS) to calibrate Advanced Very High Resolution Radiometer reflectance channels. J. Geophys. Res., 107, 4702, doi:10.1029/2001JD002035.
Heidinger, A. K., W. C. Straka, C. C. Molling, J. T. Sullivan, and X. Q. Wu, 2010: Deriving an inter-sensor consistent calibration for the AVHRR solar reflectance data record. Int. J. Remote Sens., 31, 6493–6517.
Heidinger, A. K., A. T. Evan, M. J. Foster, and A. Walther, 2012: A naive Bayesian cloud-detection scheme derived from CALIPSO and applied within PATMOS-x. J. Appl. Meteor. Climatol., 51, 1129–1144.
Heidinger, A. K., I. Laszlo, C. C. Molling, and D. Tarpley, 2013: Using SURFRAD to verify the NOAA single-channel land surface temperature algorithm. J. Atmos. Oceanic Technol., 30, 2868–2884.
Hollmann, R., and Coauthors, 2013: The ESA climate change initiative: Satellite data records for essential climate variables. Bull. Amer. Meteor. Soc., 94, 1541–1552.
Ignatov, A., and L. Stowe, 2002: Aerosol retrievals from individual AVHRR channels. Part I: Retrieval algorithm and transition from Dave to 6S radiative transfer model. J. Atmos. Sci., 59, 313–334.
Ignatov, A., I. Laszlo, E. Harrod, K. Kidwell, and G. Goodrum, 2004: Equator crossing times for NOAA, ERS and EOS sun-synchronous satellites. Int. J. Remote Sens., 25, 5255–5266.
Jacobowitz, H., L. L. Stowe, G. Ohring, A. Heidinger, K. Knapp, and N. R. Nalli, 2003: The Advanced Very High Resolution Radiometer Pathfinder Atmosphere (PATMOS) climate dataset: A resource for climate research. Bull. Amer. Meteor. Soc., 84, 785–793.
Karlsson, K.-G., and Coauthors, 2013: CLARA-A1: A cloud, albedo, and radiation dataset from 28 yr of global AVHRR data. Atmos. Chem. Phys., 13, 5351–5367, doi:10.5194/acp-13-5351-2013.
Lee, H.-T., A. Heidinger, A. Gruber and R. G. Ellingson, 2004: The HIRS Outgoing Longwave Radiation product from hybrid polar and geosynchronous satellite observations. Adv. Space Res., 33, 1120–1124.
Mittaz, J., A. Harris, and J. Sullivan, 2009: A physical method for the calibration of the AVHRR/3 thermal IR channels 1: The prelaunch calibration data. J. Atmos. Oceanic Technol., 26, 996–1019.
Moody, E. G., M. D. King, and S. Platnick, 2005: Spatially complete global spectral albedos: Value-added datasets derived from Terra MODIS land products. IEEE Trans. Geosci. Remote Sci., 43, 144–158.
Pavolonis, M. J., A. K. Heidinger, and T. Uttal, 2005: Daytime global cloud typing from AVHRR and VIIRS: Algorithm description, validation, and comparisons. J. Appl. Meteor., 44, 804–826.
Rao, C. R. N., J. T. Sullivan, C. C. Walton, J. W. Brown, and R. H. Evans, 1993. Nonlinearity corrections for the thermal infrared channels of the advanced very high resolution radiometer: Assessment and corrections. NOAA Tech. Rep. NESDIS 69, 38 pp.
Reynolds, R. W., N. A. Rayner, T. M. Smith, D. C. Stokes, and W. Wang, 2002: An improved in situ and satellite SST analysis for climate. J. Climate, 15, 1609–1625.
Rodgers, C. D., 1976: Retrieval of atmospheric temperature and composition from remote measurements of thermal radiation. Rev. Geophys. Space Phys., 14, 609–624.
Roebeling, R., B. Baum, R. Bennartz, U. Hamann, A. Heidinger, A. Thoss, and A. Walther, 2013: Evaluating and improving cloud parameter retrievals. Bull. Amer. Meteor. Soc., 94, ES41–ES44.
Rossow, W. B., G. Tselioudis, A. Polak, and C. Jakob, 2005: Tropical climate described as a distribution of weather states indicated by distinct mesoscale cloud property mixtures. Geophys. Res. Lett., 32, L21812, doi:10.1029/2005GL024584.
Saha, S., and Coauthors, 2010: The NCEP Climate Forecast System Reanalysis. Bull. Amer. Meteor. Soc., 91, 1015–1057.
Schwalb, A., 1978: The TIROS-N/NOAA A-G satellite series. NOAA Tech. Memo. NESS 95, 86 pp.
Schwalb, A., 1982: Modified version of the TIROS N/NOAA A-G satellite series (NOAA E-J)—Advanced TIROS N (ATN). NOAA Tech. Memo. NESS 116, 34 pp.
Stowe, L. L., P. A. Davis, and E. P. McClain, 1999: Scientific basis and initial evaluation of the CLAVR-1 global clear/cloud classification algorithm for the Advanced Very High Resolution Radiometer. J. Atmos. Oceanic Technol., 16, 656–681.
Stowe, L. L., H. Jacobowitz, G. Ohring, K. R. Knapp, and N. R. Nalli, 2002: The Advanced Very High Resolution Radiometer (AVHRR) Pathfinder Atmosphere (PATMOS) climate dataset: Initial analyses and evaluations. J. Climate, 15, 1243–1260.
Stubenrauch, C. J., W. Rossow, and S. Kinne, 2012: Assessment of global cloud datasets from satellites: A project of the World Climate Research Programme Global Energy and Water Cycle Experiment (GEWEX) radiation panel. WCRP Rep. 23/2012, 180 pp. [Available online at www.wcrp-climate.org/documents/GEWEX_Cloud_Assessment_2012.pdf.]
Stubenrauch, C. J., and Coauthors, 2013: Assessment of global cloud datasets from satellites: Project and database initiated by the GEWEX radiation panel. Bull. Amer. Meteor. Soc., 94, 1031–1049.
Trenberth, K. E., and J. W. Hurrell, 1994: Decadal atmosphere-ocean variations in the Pacific. Climate Dyn., 9, 303–319.
Walther, A., and A. K. Heidinger, 2012: Implementation of the daytime cloud optical and microphysical properties algorithm (DCOMP) in PATMOS-x. J. Appl. Meteor. Climatol., 51, 1371–1390.
Zhao, T. X.-P., I. Laszlo, W. Guo, A. K. Heidinger, C. Cao, A. Jelenak, D. Tarpley, and J. Sullivan, 2008: Study of long-term trend in aerosol optical thickness observed from operational AVHRR satellite instrument. J. Geophys. Res., 113, D07201, doi:10.1029/2007JD009061.
Zhao, T. X.-P., A. K. Heidinger, and K. R. Knapp, 2011: Long-term trends of zonally averaged aerosol optical thickness observed from operational satellite AVHRR instrument. Meteor. Appl., 18, 440–445.
