GOES ABI Detection of Thin Cirrus over Land

Theodore M. McHardy aAmerican Society for Engineering Education, Washington, D.C.
bNaval Research Laboratory, Monterey, California

Search for other papers by Theodore M. McHardy in
Current site
Google Scholar
PubMed
Close
https://orcid.org/0000-0002-5996-5320
,
James R. Campbell bNaval Research Laboratory, Monterey, California

Search for other papers by James R. Campbell in
Current site
Google Scholar
PubMed
Close
,
David A. Peterson bNaval Research Laboratory, Monterey, California

Search for other papers by David A. Peterson in
Current site
Google Scholar
PubMed
Close
,
Simone Lolli cCNR-IMAA, Tito Scalo, Italy

Search for other papers by Simone Lolli in
Current site
Google Scholar
PubMed
Close
,
Anne Garnier dScience Systems Applications Inc., Hampton, Virginia

Search for other papers by Anne Garnier in
Current site
Google Scholar
PubMed
Close
,
Arunas P. Kuciauskas bNaval Research Laboratory, Monterey, California

Search for other papers by Arunas P. Kuciauskas in
Current site
Google Scholar
PubMed
Close
,
Melinda L. Surratt bNaval Research Laboratory, Monterey, California

Search for other papers by Melinda L. Surratt in
Current site
Google Scholar
PubMed
Close
,
Jared W. Marquis eDepartment of Atmospheric Sciences, University of North Dakota, Grand Forks, North Dakota

Search for other papers by Jared W. Marquis in
Current site
Google Scholar
PubMed
Close
,
Steven D. Miller fCooperative Institute for Research in the Atmosphere, Colorado State University, Fort Collins, Colorado

Search for other papers by Steven D. Miller in
Current site
Google Scholar
PubMed
Close
,
Erica K. Dolinar bNaval Research Laboratory, Monterey, California

Search for other papers by Erica K. Dolinar in
Current site
Google Scholar
PubMed
Close
, and
Xiquan Dong gDepartment of Hydrology and Atmospheric Sciences, The University of Arizona, Tucson, Arizona

Search for other papers by Xiquan Dong in
Current site
Google Scholar
PubMed
Close
Free access

Abstract

This study develops a new thin cirrus detection algorithm applicable to overland scenes. The methodology builds from a previously developed overwater algorithm, which makes use of the Geostationary Operational Environmental Satellite 16 (GOES-16) Advanced Baseline Imager (ABI) channel 4 radiance (1.378-μm “cirrus” band). Calibration of this algorithm is based on coincident Cloud–Aerosol Lidar with Orthogonal Polarization (CALIOP) cloud profiles. Emphasis is placed on rejection of false detections that are more common in overland scenes. Clear-sky false alarm rates over land are examined as a function of precipitable water vapor (PWV), showing that nearly all pixels having a PWV of <0.4 cm produce false alarms. Enforcing an above-cloud PWV minimum threshold of ∼1 cm ensures that most low-/midlevel clouds are not misclassified as cirrus by the algorithm. Pixel-filtering based on the total column PWV and the PWV for a layer between the top of the atmosphere (TOA) and a predetermined altitude H removes significant land surface and low-/midlevel cloud false alarms from the overall sample while preserving over 80% of valid cirrus pixels. Additionally, the use of an aggressive PWV layer threshold preferentially removes noncirrus pixels such that the remaining sample is composed of nearly 70% cirrus pixels, at the cost of a much-reduced overall sample size. This study shows that lower-tropospheric clouds are a much more significant source of uncertainty in cirrus detection than the land surface.

© 2022 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Publisher’s Note: This article was revised on 3 March 2023 to correct the affiliation of coauthor Lolli, which was incorrect when originally published.

Corresponding author: Theodore M. McHardy, theodore.mchardy.ctr@nrlmry.navy.mil

Abstract

This study develops a new thin cirrus detection algorithm applicable to overland scenes. The methodology builds from a previously developed overwater algorithm, which makes use of the Geostationary Operational Environmental Satellite 16 (GOES-16) Advanced Baseline Imager (ABI) channel 4 radiance (1.378-μm “cirrus” band). Calibration of this algorithm is based on coincident Cloud–Aerosol Lidar with Orthogonal Polarization (CALIOP) cloud profiles. Emphasis is placed on rejection of false detections that are more common in overland scenes. Clear-sky false alarm rates over land are examined as a function of precipitable water vapor (PWV), showing that nearly all pixels having a PWV of <0.4 cm produce false alarms. Enforcing an above-cloud PWV minimum threshold of ∼1 cm ensures that most low-/midlevel clouds are not misclassified as cirrus by the algorithm. Pixel-filtering based on the total column PWV and the PWV for a layer between the top of the atmosphere (TOA) and a predetermined altitude H removes significant land surface and low-/midlevel cloud false alarms from the overall sample while preserving over 80% of valid cirrus pixels. Additionally, the use of an aggressive PWV layer threshold preferentially removes noncirrus pixels such that the remaining sample is composed of nearly 70% cirrus pixels, at the cost of a much-reduced overall sample size. This study shows that lower-tropospheric clouds are a much more significant source of uncertainty in cirrus detection than the land surface.

© 2022 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

Publisher’s Note: This article was revised on 3 March 2023 to correct the affiliation of coauthor Lolli, which was incorrect when originally published.

Corresponding author: Theodore M. McHardy, theodore.mchardy.ctr@nrlmry.navy.mil

1. Introduction

Historical estimates of global cirrus based on passive satellite radiometer were first estimated to be around 20% (Rossow and Schiffer 1999). However, with the launch of the Cloud–Aerosol Lidar with Orthogonal Polarization (CALIOP) aboard the Cloud–Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) satellite in 2006, cirrus occurrence could be more readily identified through active laser profiling. Based on these data, Mace and Zhang (2014) show that cirrus clouds in fact cover roughly 40% of Earth at any given time, while other studies have shown this value can approach 70% regionally (e.g., Lolli et al. 2017). This gross discrepancy between passive- and active-based cirrus fraction estimates arises from an inherent limitation of passive systems in distinguishing clouds with visible wavelength optical depths (COD) of less than ∼0.3 (Ackerman et al. 2008; Stubenrauch et al. 2013). Sassen and Cho (1992) refer to this subgenre of cirrus clouds as “optically thin.” Campbell et al. (2015) elaborate further by characterizing the cirrus cloud distribution as being weighted toward optically thin and even “subvisual” clouds (COD < 0.03; Sassen and Cho 1992). They further conclude, again based on CALIOP, that roughly 50% of all cirrus exhibit COD less than 0.3.

Because of the global prevalence of optically thin cirrus and the limited fidelity demonstrated by passive radiometric instruments in distinguishing them, contamination by cirrus is increasingly being recognized as a problem in level-2 meteorological, oceanographic, and cryospheric retrievals (i.e., lower-atmospheric or surface parameters residing below potential cirrus cloud layers when viewed from space). These issues can bias operational forms of predictive models through data assimilation applications. For example, sea surface temperature (Marquis et al. 2017) and aerosol optical depth (AOD; Chew et al. 2011) retrievals have both been shown to be susceptible to radiance contributions (contamination) from unscreened optically thin cirrus. Likewise, radiance contamination induced by cold and relatively large cirrus cloud ice crystals (10–100 μm on average) can introduce a host of biases in surface parameter retrievals (e.g., land and ocean surface temperature).

While spaceborne lidars have proven useful for compiling climatological information of cirrus (e.g., Sassen and Campbell 2001), their utility in distinguishing cloud presence relative to passive radiometers and mitigating their adverse effects is extremely restricted. This is because of their limited spatial and temporal coverage compared with relatively wide field-of-view imagers. A spaceborne lidar flying in a sun-synchronous orbit, such as CALIOP, has a small ground footprint (∼70 m) and a long revisit period (∼16 days; Winker et al. 2007); while a geostationary-based sensors like the Advanced Baseline Imager (ABI) provide 10-min updates of much of the western hemisphere. Similarly, ground-based lidars can be used to detect cirrus locally. However, their signals are frequently attenuated by optically thick lower cloud layers, limiting their ability to profile the upper troposphere and cirrus cloud heights. In addition, they are generally only deployed at land-based sites, which limits global coverage. The expense and logistics associated with deploying a network of instruments precludes their usage for widespread and robust global cirrus monitoring.

Given the proper spectral coverage and spatial resolutions, passive radiometric imagers can be effective in identifying more occurrences of thin cirrus clouds. This capability was first recognized in the early 1990s by Gao et al. (1993), who demonstrate that a channel near 1.38 μm can facilitate daytime cirrus cloud detection. A strong water vapor absorption feature around this wavelength attenuates signals from the lower troposphere, while leaving the solar radiation reflected by cirrus clouds relatively unaffected. Detection is possible because cirrus clouds are typically located above the majority of water vapor in the atmospheric column. This characteristic prompted the addition of the 1.375-μm band on the Moderate Resolution Imaging Spectroradiometer (MODIS) instruments (Gao and Kaufman 1995; Justice et al. 1998). Unfortunately, implementation of this channel exhibited multiple technical issues, including stray light contamination and detector cross talk (Dessler and Yang 2003). As a result of these issues, analyses shows that MODIS level-2 sensitivities to thin cirrus were similar to historical datasets, approaching the 0.30 COD threshold for optically thin clouds (Ackerman et al. 2008).

Though the MODIS cirrus channels proved limited, scientific progress was still made in passive thin cirrus detection and cirrus property retrievals. Dessler and Yang (2003) use the MODIS Terra 1.375-μm reflectance to retrieve the COD of optically thin cirrus based on phase function calculations in pixels labeled as clear sky by the MODIS cloud mask. This research suggested that optically thin cirrus clouds may be much more prevalent than previously thought. Applying the methodology from Dessler and Yang (2003), Lee et al. (2009) compute COD and radiative forcing of cirrus clouds in the tropics during a 1-yr period and found that 40% of pixels classified as clear sky by the operational MODIS cloud mask were in fact thin cirrus pixels. Taking a slightly different approach, Meyer et al. (2004) retrieve cirrus COD via a bispectral method similar to the operational MODIS cloud microphysics algorithm. Their technique was limited by computational expense at the time. Meyer and Platnick (2010) later expanded on Meyer et al. (2004), where methods to account for surface reflectance and attenuation of the 1.375-μm signal by above-cloud precipitable water vapor are described, along with COD uncertainty estimates. More recently, Wang et al. (2014) use the 1.375-μm channel to derive the phase function of ice clouds. All of these studies focus on the retrieval of cloud properties, rather than on calibrated detection of cirrus clouds.

The launch of newest generation of Geostationary Operational Environmental Satellites (GOES) provided a new and exciting opportunity. The ABI aboard GOES-16 and GOES-17 contain a 1.378-μm band intended for daytime cirrus cloud applications. The ABI represents a significant improvement over MODIS for several reasons. First and foremost, the sensor meets its on-orbit performance requirements (Yu et al. 2017; Bartlett et al. 2018). As discussed, relative to MODIS, design specifications for ABI were critical for demonstrating sensitivities to the very low radiances attributable to scattering of sunlight by optically thin cirrus. In terms of band specifications, the ABI 1.378-μm channel is narrower than its counterpart on MODIS, with a range from 1.366 to 1.38 μm, which reduces noise. Also, while data from Aqua MODIS can be collocated with CALIOP, the sun-synchronous orbit has limitations. Variation in viewing and solar geometry is required to gain a full understanding of any cloud remote sensing application, and collocation between Aqua MODIS and CALIOP can only be done relatively easily at near-nadir viewing.

Geostationary platforms, on the other hand, can observe within their field-of-review the diurnal cycle of clouds, and CALIOP overpasses through this domain capture variations in viewing and solar geometry. Operationally, Aqua MODIS is hindered by its relatively long return period of 1 day (twice daily including Terra). These issues also inhibit the Visible Infrared Imaging Radiometer Suite (VIIRS) instruments which, while containing 1.38-μm cirrus bands of their own, are not able to be collocated with CALIOP for comparison as often as Aqua MODIS.

McHardy et al. (2021, hereafter MCH21) test the application of the new GOES-16 1.378-μm channel for detecting thin cirrus clouds. They developed and demonstrated a simple algorithm based on 1.378-μm radiance thresholds for discriminating between clear sky and cirrus clouds over ocean. This algorithm was calibrated using collocated CALIOP profiles to distinguish between clear and cloudy scenes, as well as optically thin and relatively opaque cirrus. By this algorithm, cirrus clouds exhibiting an optical depth less than 0.3 were detected at a rate of 84% and clear-sky pixels were correctly identified 96% of the time, despite the relatively straight-forward design of the algorithm. Further, they reported that column precipitable water vapor (PWV) is not a factor impacting cirrus cloud detection in over 95% of their case sample size over water.

The present study builds upon the results from MCH21 by applying the cirrus detection algorithm to scenes over land, which is significantly more complicated due to the inhomogeneous nature of PWV over land compared with ocean, as well as widely varying surface reflectance and terrain height. Underlying data sources and methodologies used here are similar to those in MCH21, including the collocation of ABI pixels with the CALIOP lidar to provide contextual column profile information. However, while MCH21 prioritizes discrimination between clear sky and thin cirrus clouds, this study focuses on identifying pixels where it is instead appropriate to reject them. That is, a full-factor recalibration of the radiance thresholds in MCH21 is redundant. Any relaxation of these values will come at the expense of the detection of thinner clouds, which drives against the point of the whole exercise. Rather, we seek to work only with ABI pixels where these thresholds can be applied confidently.

Over land, we are primarily concerned with pixels corresponding with the land surface and low-level liquid water clouds that, due to effective water vapor absorption through the column above the cloud, have an observed signal comparable to that of optically thin cirrus. In other words, targets with a higher reflectance at 1.378 μm than semitransparent cirrus, such as low-altitude liquid water clouds, could have their signal reduced by above-cloud water vapor such that their observed reflectance is of similar magnitude to that of the cirrus. Climatological cloud occurrence rates and profile heights are different over land than water, which impacts the retrievals in a manner not considered significant in the prior study. Overall, thin cirrus detection rates remain largely the same for those “quality controlled” pixels rendered by the new algorithm. We describe these new working constraints, however, and outline what we conclude to be the effective operating limits to where a 1.378-μm channel can be applied effectively over land.

2. Input data, calibration, and collocation

a. GOES-16 data

Following MCH21, one month of GOES-16 (East) data from August 2018 were selected for this study. Only one month was used here to limit computational expense—using the full disk images provides adequate variation in viewing/solar geometry and seasonality. Pixel data were limited to daytime by restricting the solar zenith angle (SZA) to <80°. To limit the effects of pixel expansion, the viewing zenith angle was confined to <80°. GOES-16 full disk images were available every 15 min during the study period used here. The GOES-16 ABI 1.378-μm channel radiance is the subject of this study and features a spatial resolution of 2 km at nadir, though pixels expand as the viewing zenith angle (VZA) increases (∼10 km2 for SZAs > 60°; Figs. 2 and 3 in MCH21). The specific range of the channel is from 1.366 to 1.38 μm (50% full-width at half maximum spectral response). This spectral width is narrower than the corresponding band on MODIS, which hones in on the strongest water vapor absorption, reducing the signal detected from the lower troposphere and the potential for background solar contamination. More information on GOES-16 and its 1.378-μm channel can be found in MCH21.

Downstream GOES-16 cloud-top height (CTH) and COD products are used as additional validation for case studies. Readers are directed to the algorithm theoretical basis documents for more information (Heidinger 2012; Walther et al. 2013), as the goal of this study is neither validation of nor comparison with these products. The GOES-16 ABI L2+ aerosol detection product (ADP; NOAA/NESDIS 2018), which is a binary aerosol mask (0 = no aerosol, 1 = aerosol), is used to screen pixels with significant aerosol loading present. The ADP uses the ABI cloud mask (ACM; Heidinger 2010) to ensure that aerosol detection is only performed for cloud-free pixels. The ACM does include a 1.378-μm reflectance threshold for cirrus detection. While this threshold value is currently specified too high to detect most thin cirrus (MCH21), care must still be taken to ensure that thin cirrus pixels are not removed when using the ADP. In MCH21, clear-sky pixels were screened for the presence of aerosol by requiring the ADP to have attempted a retrieval, but detected no aerosol (i.e., ADP = 0). This was done to ensure the best calibration of the cirrus detection algorithm. Here, pixels having an ADP value of 1 are removed from all subsamples (clear and cloudy). This means that pixels not having a valid ADP retrieval (potentially due to the occurrence of cloud) are not removed. This will be discussed further in section 3a.

b. CALIOP data

Data from the CALIOP v4.20 level-2 5-km cloud profile product (Vaughan et al. 2020; Winker et al. 2009) were used to classify clear-sky and cloudy pixels in this study. The Hybrid Extinction Retrieval Algorithm (HERA; Young and Vaughan 2009; Young et al. 2018) was used to derive 532-nm extinction coefficients. The profile products are reported at a uniform spatial resolution of 60 m vertically below altitudes of 20.2 km and 180 m vertically above altitudes of 20.2 km, and 5 km horizontally, over a nominal altitude range from 0.5 km below the surface up to 30 km in the atmosphere. The level-2, 5-km cloud profile product was chosen over higher resolution products due to the horizontal averaging required to resolve targets as optically thin as subvisual cirrus (Young and Vaughan 2009). In contrast to MCH21, COD and cloud classification values for individual layers were used in this study. Cloud classifications are contained in the feature classification portion of the Atmospheric Volume Descriptor (AVD). The cloud type categorization follows International Satellite Cloud Climatology Project definitions (ISCCP; Rossow and Schiffer 1999). Readers are directed to MCH21 and the v4.20 cloud profile product user’s guide for more information on the CALIOP data.

Pixels were considered “clear” for CALIOP COD equal to 0. Cloudy pixels were broken down into three categories based on layer COD and morphological cloud classification from CALIOP for analysis. For each CALIOP layer, a feature classification and a layer-based 532-nm extinction coefficient are provided. Therefore, the 532-nm COD attributable to each cloud type within the full column can be obtained from integrating column depth-normalized profile values. This information was used to collate the cloudy pixels into three categories based on the topmost cloud layer distinguished through feature classification: low-/midaltitude water clouds, high-altitude ice clouds (both transparent and opaque), and thin cirrus only. Pixels with additional cloud layers beneath cirrus clouds were not addressed individually because they are likely to be positively identified by the cirrus detection algorithm, and it would not be possible to distinguish between them and cirrus-only pixels operationally without the use of additional products. The CALIOP-derived CTH is considered the altitude of the highest layer that contains cloud within the column. Because CALIOP vertical profile information is provided in terms of altitude above mean sea level (MSL), it must be corrected to above ground level (AGL) when collocated with ABI. This is done by subtracting the surface elevation from the CALIOP layer heights obtained from the Global Multiresolution Terrain Elevation Data (GMTED2010; Danielson and Gesch 2011) product. This product also contains the land–sea mask used to distinguish between land and ocean pixels.

Column-integrated total PWV measurements are negligible on the calibration of the 1.378-μm channel radiance thresholds for stable cirrus cloud detection over ocean in MCH21. This study is predicated on the presumption that this will not be the case for cirrus detection algorithm over land due to continental air masses and elevated terrain. PWV is therefore computed from relative humidity profiles provided in the CALIOP cloud product, derived from the Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2; Gelaro et al. 2017). The layer-relative humidity is converted into mixing ratio, which is then integrated from the top of the atmosphere (TOA) down to either the surface, the CTH, or a specific altitude, all of which are considered in this study.

c. Colocation

Spatiotemporal collocation of CALIOP and GOES ABI data is complicated because the CALIOP L2 cloud profiles product footprint is essentially a one-dimensional curtain (∼70-m width; Winker et al. 2007) subsampled into 5-km “pixels” along track. In contrast, GOES-16 ABI is a geostationary-based imaging radiometer whose pixel sizes grow as VZA increases (MCH21). This means that as VZA increases, CALIOP-retrieved cloud properties may become less representative of the clouds within the corresponding GOES pixels. Pixel match-ups are achieved when a CALIOP orbit intersects a GOES pixel within 7.5 min of a GOES full disk scan. Results from subsamples with a collocation time (dt) of ±1 min are shown throughout this study as a means of reducing collocation error. Because CALIOP “pixels” are 5 km in length, it is possible that multiple GOES pixels pair with one CALIOP “pixel.” These pixels are considered independent data pairs. Examples of this collocation methodology under different viewing geometries are available in MCH21.

Parallax correction of GOES-observed cirrus data to the CALIOP nadir-view profile was vital in MCH21 because that study focused exclusively on differentiating between optically thin cirrus, which occur at high altitudes, and clear sky. Radiance detection thresholds were then derived from this overwater sample. Parallax correction is deemed unnecessary for over land application of the cirrus detection algorithm in this study, as the fundamental goal is different. This study focuses on characterizing the 1.378-μm channel’s response to the land surface and low-/midlevel clouds under varying moisture conditions. Error due to the parallax effect decreases as a target’s altitude decreases. Therefore, the horizontal displacement due to the parallax effect of pixels that are of interest in this study is relatively small. While it is possible under oblique viewing conditions for GOES to “see” a cloud from a neighboring pixel while CALIOP sees clear sky, this is impractical to correct for on a large-scale basis. The lack of parallax correction in this study will result in slightly decreased positive cirrus cloud detection rates compared with MCH21; however, this is not an actual decrease in detection rates and is due to the mismatching of the CALIOP and ABI footprints in this study. As false alarms and not positive detections are of interest here, the overall impact on this study by not correcting for parallax is considered minimal.

3. Detection algorithm for thin cirrus over land

The core of the cirrus detection algorithm used in this study is unchanged from MCH21. A flowchart is provided in Fig. 1, where the right side of the chart describes the process according to MCH21. First, the relationships between the clear-sky 1.378-μm radiance and viewing/solar geometry are characterized. The detection thresholds are computed using the 1.378-μm radiance as a function of airmass factor (AMF), which is defined as
AMF=[1cos(VZA)+1cos(SZA)].
This parameter replaces VZA and SZA to remove 1 degree of freedom. Regression lines are computed for one and two standard deviations above the mean radiance as a function of AMF. Any pixel with a 1.378-μm radiance greater than the regressed radiance value for its AMF value is deemed to be cloudy. The algorithm detects over 80% of optically thin cirrus (0.03 < COD < 0.3) and nearly 50% of subvisual cirrus (COD < 0.03). Multiple cirrus thresholds are presented in MCH21. The land-based application described herein uses
Threshold=0.266235+(0.022984×AMF),
where Threshold is the radiance above which a pixel is flagged as cloudy and AMF is the pixel-level AMF (MCH21). This equation was computed as two standard deviations above the mean 1.378-μm radiance as a function of AMF for collocated GOES CALIOP pixels within 1 min of each other.
Fig. 1.
Fig. 1.

An updated algorithm flowchart showing the methodologies for pixel rejection, based on Fig. 9. In MCH21.

Citation: Journal of Atmospheric and Oceanic Technology 39, 9; 10.1175/JTECH-D-21-0160.1

MCH21 also presents equations for estimating COD from the 1.378-μm radiance, which are intended for qualitative applications only. The following equation is used in that capacity here:
COD=10^{0.85082+[log10(R)×0.709307]},
where R is the 1.378-μm radiance. The estimated COD is only used in this study to determine if cirrus detection false alarms due to clear sky or low-/midlevel water clouds would appear thin or subvisual to the cirrus detection algorithm. This technique will be discussed further in section 3b.

Figure 2 provides example output of the previously developed cirrus detection algorithm (described in MCH21) applied over land and the corresponding visible (channel 2) imagery for the contiguous United Stated (CONUS). The cirrus detection algorithm produces a binary result (see Fig. 1). A value of 1 (hereafter, a “positive”) means that a pixel had an R value greater than its threshold R value. For clear-sky and low-/midlevel cloud pixels this is considered a “false alarm” detection. A value of 0 (hereafter, a “negative”) means the pixel had an R value below its threshold R value and is considered a “false negative” for cirrus cloud pixels. These definitions can be seen in Table 1.

Fig. 2.
Fig. 2.

(a) GOES ABI visible (channel 2) imagery for 2200:20 UTC 19 Jun 2021 for a sector centered over the CONUS and (b) the corresponding cirrus mask as described in MCH21.

Citation: Journal of Atmospheric and Oceanic Technology 39, 9; 10.1175/JTECH-D-21-0160.1

Table 1

Terminology for the binary result of the cirrus detection algorithm based on the case type.

Table 1

The results from applying the previously developed MCH21 thin cirrus detection algorithm (Fig. 1, right column) to pixels over land for the entire study period are shown in Table 2. Over-ocean results are shown as well for comparison. Samples (i.e., collocated ABI and CALIOP pixels) are divided into four subsamples which were further described in section 2. These include clear sky, low-/midlevel cloud, cirrus, and thin cirrus only. Values also shown are the total number of samples with the percentage of the total sample size that each subsample comprises in parentheses, the false alarm (for clear-sky and low-/midlevel cloud pixels) or the detection rate (for cirrus pixels), which is computed as the number of positive detections by the cirrus detection algorithm divided by the number of samples, for each subsample, and the thin rate, which is the number of false alarms with a COD estimated by using Eq. (3) of <0.3 divided by the number of false alarms, for each subsample. The impact of this will be discussed further in section 3b. Values are shown for both overland and overwater, and for collocation time limits of ±7.5 and ±1 min, for comparison. Note that due to lack of parallax correction and ADP filtering, these values will differ slightly from those in MCH21. It is immediately noticeable that the false alarm rate is higher for overland pixels than overocean pixels for both clear-sky and low-/midlevel cloud conditions. We will first address clear-sky contamination by examining how column PWV affects the cirrus detection algorithm.

Table 2

Results from applying the thin cirrus detection algorithm to overland pixels. Samples are divided into four subsamples—clear sky, low-/midlevel cloud, cirrus, and thin cirrus only, which are described in section 2.

Table 2

a. PWV thresholds to reduce clear-sky contamination

Thin cirrus detection using the 1.378-μm channel is possible because of water vapor absorption around 1.38 μm. Given sufficient column PWV, solar radiation reflected by the surface and low-/midlevel clouds is fully attenuated. Solar reflection by cirrus situated almost exclusively above the most significant depths of water vapor presence, in contrast, are imaged by the sensor over a dark, highly contrasting lower-level layer. To demonstrate how PWV impacts the transmission of radiation through the atmosphere, Gao and Kaufman (1995) computed two-way transmission values for five different atmospheric profiles, with PWV values ranging from over 4 cm to less than 0.5 cm, using a proposed band centered at 1.375 μm. VZAs of 0° and SZAs of 45° were used for these calculations.

From their exercise, it is clear that profiles with more water vapor (i.e., tropical and midlatitude summer) have smaller two-way transmissions than drier profiles (i.e., subarctic winter). Most important to note from these data is the propensity for lower PWV pixels to maintain two-way transmission values above zero near the surface, and substantial transmission values (∼20% or greater) at altitudes where liquid clouds are prevalent. This implies a nonnegligible potential for surface radiance contamination at 1.378 μm that must be reconciled, and possibly accounted for.

To investigate how PWV impacts cirrus detection for clear-sky pixels, and explicitly seek out the potential for surface contamination, the false alarm rate as a function of surface elevation and PWV is investigated. This is done, for collocation times of both ±1 and ±7.5 min (top and bottom of Fig. 3, respectively), by binning all clear-sky pixels by their PWV and elevation values. The sample densities (log scale, Figs. 3a,d), computed false alarm rates (Figs. 3b,e), and negative rates (Figs. 3c,f) are shown for each bin. Figure 3e reveals that nearly all clear-sky overland pixels corresponding with a total column PWV of less than approximately 0.5 cm are false alarms. We characterize this effect as being mostly due to surface contamination, suggesting that a PWV value of around 0.5 cm is enough to ensure atmospheric opacity with respect to the 1.378-μm channel for typical land surface reflectance values (∼0.2–0.4).

Fig. 3.
Fig. 3.

(a),(d) Sample density, (b),(e) false alarm rate, and (c),(f) negative rate for clear-sky overland samples binned by total-column PWV and surface elevation, with bin sizes of 0.1, respectively. Rows show collocation time limits of (top) ±1 and (bottom) ±7.5 min.

Citation: Journal of Atmospheric and Oceanic Technology 39, 9; 10.1175/JTECH-D-21-0160.1

To refine this threshold more definitively, we estimate a false alarm rate as a function of hypothetical low-end PWV threshold value [Fig. 4a (dt < 1) and Fig. 4b (dt < 7.5)], where all pixels exhibiting a PWV value less than the corresponding hypothetical threshold are removed from the sample. The sample size remaining after applying the thresholds is also shown in Fig. 4. By simultaneously minimizing the false alarm rate and maximizing the sample size, we can resolve the most appropriate PWV threshold value for application. The false alarm rate does not decrease significantly between PWV threshold values of 0.4 and 1 cm. Therefore, all pixels having a PWV value < 0.4 cm are recommended for removal from all samples. The results of applying this threshold are shown in Table 3. The clear-sky false alarm rate over land drops by over 2%, while cloudy samples are relatively unaffected. Ocean pixels are also not strongly impacted by a PWV threshold, as the PWV over the ocean is very rarely below 0.4 cm, as discussed in MCH21.

Fig. 4.
Fig. 4.

The false alarm rate (black lines, left y axes) and total number of samples (red lines, right y axes) remaining after removing pixels having a PWV less than the PWV threshold, which is varied (x axes) for clear-sky samples. Collocation time limits of (a) ±1 and (b) ±7.5 min are shown. The chosen PWV threshold value of 0.4 cm is indicated by the dashed lines.

Citation: Journal of Atmospheric and Oceanic Technology 39, 9; 10.1175/JTECH-D-21-0160.1

Table 3

As in Table 2, but after the removal of all pixels having a PWV value of <0.4 cm.

Table 3

Even after applying this initial PWV threshold, the clear-sky false alarm rate over land is still significantly higher than over ocean. This can, in part, be explained with the GOES aerosol and cloud mask products (ADP and ACM). As mentioned earlier, we elected not to require clear-sky pixels to correspond with an ADP value of 0, which means that pixels where the ADP did not make a valid retrieval are not removed. By inspecting the ADP values of removed clear-sky pixels over land, it was revealed that most did not have a valid ADP retrieval as opposed to having positively identified aerosol. Table 4 shows how using the ADP to filter clear-sky samples changes the overall statistics. When pixels without an ADP retrieval are removed, the clear-sky false alarm rate decreases to nearly 5%, which is similar to that in MCH21. For nearly all of these pixels, the ADP did not reflect a successful a retrieval because the ACM detected cloud.

Table 4

The number of cirrus detection algorithm negatives and positives for clear-sky pixels after applying various filters based on the GOES aerosol detection product (ADP). For a filter requiring that ADP ≠ 1, the number of pixels where the ADP did not make a retrieval are shown. For pixels where the ADP did not make a retrieval, the number of pixels where the nonretrieval was due to cloud detection by the GOES Advanced Baseline Imager (ABI) cloud mask are also shown.

Table 4

In addition to over 7000 positives being removed from the sample, almost 4000 negatives were also removed (pixels where cirrus algorithm returned 0). For these pixels, the CALIOP level-2 profiles and the cirrus detection algorithm both indicated that no cloud was present, while the GOES ACM did indicate cloud. This significant discrepancy between CALIOP and the ACM is ultimately due to how the CALIOP 5-km products are aggregated from finer-resolution observations. Cloud layers with cloud-top altitudes lower than 4 km that were detected by CALIOP at single-shot (i.e., 1/3-km) resolution are potentially cleared from the 5-km-layer product to improve the detection of aerosols at coarser spatial resolutions (Vaughan et al. 2005), and are not included in the 5-km cloud profile products. Thus, a column cloud optical depth equal to 0 in the 5-km product does not necessarily indicate that the column is cloud free. As a result, there are more false alarms due to low-level clouds and less due to the land surface under clear skies than indicated by Tables 2 and 3. This does not impact the major findings of this paper.

b. Challenges presented by variable cloud cover over land

Cloud-top heights for CALIOP-designated low-level liquid water cloudiness are fundamentally different over water and over land. Similar to clear sky, we seek applying a means for filtering these low-/midlevel clouds based on column water vapor. However, water vapor beneath clouds will not attenuate radiation reflected by the cloud-top to satellite, rendering the total column PWV inappropriate. This is especially true given the fact that in most pixels, most of the water vapor in a column is located near the surface. We therefore employ a modified above-cloud PWV parameter, which is the integrated column water vapor present between the clouds and the satellite.

Differences between land and water scenes are shown (in Figs. 6a and 6d) as function of sample densities of CTH versus above-cloud PWV. The relatively high fidelity in overwater algorithm performance demonstrated by MCH21 is a function of these two mechanisms at play. Another point of note is the difference in low-/midlevel false alarm rate between land and ocean (Table 3). The issue of false alarms due to lower-tropospheric water clouds was not discussed in MCH21, as the focus was on discriminating between clear sky and cirrus clouds while maximizing cirrus detection over water. Aside from surface reflectance considerations, however, the sensitivity of the zero-transmission heights across the simulations in Gao and Kaufman (1995) reflect the susceptibility of low-level atmospheric opacity to low-level liquid water presence. That is, low- to midlevel liquid water cloudiness is dominated by a singular mode confined mostly below 3 km (i.e., presumably marine stratocumulus; Fig. 5a). Also, above-cloud PWV values are relatively high, meaning that most low-level cloudiness is suppressed and the cirrus retrieval performs in a relatively stable manner.

Fig. 5.
Fig. 5.

(a),(d) Sample density, (b),(e) false alarm rate, and (c),(f) negative rate for low-/midlevel cloud samples binned by above-cloud PWV and CALIOP cloud-top height (CTH), with bin sizes of 0.1, for (top) ocean and (bottom) land.

Citation: Journal of Atmospheric and Oceanic Technology 39, 9; 10.1175/JTECH-D-21-0160.1

This single-mode cloud frequency profile distribution breaks down over land, however, presumably driven by convection and midlevel cloudiness associated with baroclinic midlatitude weather disturbances (Rossow and Schiffer 1999; Fig. 5d). Combined with a propensity for highly heterogeneous total-column PWV values, the potential for low-level liquid water cloud detection in the 1.378-μm channel is high. Importantly, we are not necessarily concerned with this issue as long as their measured radiance is of significant magnitude so as to not be mistaken for an optically thin cloud. It is entirely plausible that brightness temperature measurements taken from other channels could be integrated to overcome these low-cloud contaminants. But, for the purposes of 1.378-μm attribution alone, and thus in the absence of any other information, our goal is to suppress as much of this cloud that could be confused or misattributed as optically thin cirrus as possible.

c. Thresholds to reduce low-/midlevel cloud contamination

Similar to Fig. 3, Fig. 5 provides false alarm and negative rates, as well as the aforementioned sample density, for pixels containing low-/midlevel clouds over land and a comparison with ocean scenes. In this case, the height is the CTH (from CALIOP) and the PWV is for the above-cloud portion of the column only. The false alarm rate decreases as the above-cloud PWV increases, which contrasts with the clear-sky hard cutoff between false alarms and negatives at PWV values near 0.4 cm. This distinction is likely due to the differences in the variability of scattering properties (e.g., spectral reflectance) between clouds and the land surface. Comparing land with ocean, the false alarm rates due to low-/midlevel clouds over ocean are more consistent with respect to above-cloud PWV. In other words, the false alarm rates are less “noisy.” However, this is likely a second-order contributor to the difference in false alarm rates between land and ocean compared to the difference in cloud climatology between land and ocean.

Computing above-cloud PWV requires cloud-top height information from a source other than GOES data (in this case we used CALIOP). While level-2 products such as cloud-top height retrievals are highly relevant, they can significantly increase product latency. Computing above-cloud PWV in operational applications is therefore undesirable. Instead, the PWV integrated from the TOA down to a designated layer height H is used. While PWV calculated from CALIOP/MERRA-2 was used in section 3a, both PWV and water vapor mixing ratio profiles can be obtained from operational model analysis in near–real time. We consider employing these data as an acceptable amount of additional complication, as the cirrus detection algorithm’s latency will not be significantly increased using model data available in real time, which the Navy keeps via its Navy Global Environmental Model (NAVGEM; Hogan et al. 2014).

Selecting a PWV height threshold (THPWV,H) is achieved in a manner consistent with how the total-column PWV threshold was selected—by iterating through each potential THPWV,H value and simultaneously minimizing the contamination rate and maximizing the remaining sample size. However, the procedure is slightly more complicated here, as the optimal height H must also be determined. Also, instead of overall sample size, the focus is on maximizing the number of cirrus pixels remaining after applying THPWV,H, as preferentially removing low-/midlevel cloudy pixels will result in lower false alarm rates. To test each THPWV,H, the false alarm rates and the fraction of cirrus pixels remaining after applying various THPWV,H for selected THPWV,H values are shown in Fig. 6. PWV values range from 0.10 to 1.5 cm and H values range from 1 to 6 km. Here, the false alarm rate calculation (FAPWV,H) is modified as a function of THPWV,H. It is computed as the number of clear-sky false alarms after applying THPWV,H plus the number of low-/midlevel cloud false alarms after applying THPWV,H divided by the total number of samples after applying THPWV,H.

Fig. 6.
Fig. 6.

The fraction of cirrus pixels retained and the false alarm rate after applying THPWV,H. The PWV threshold value PWVH is indicated by the shape and the height threshold value H is indicated by the color.

Citation: Journal of Atmospheric and Oceanic Technology 39, 9; 10.1175/JTECH-D-21-0160.1

Figure 6 shows how PWV and H can be chosen to minimize FAPWV,H while preserving pixels containing cirrus clouds. Increasing either the PWV or H threshold values generally removes cirrus pixels and decreases the FAPWV,H. However, it is clear that changes in FAPWV,H and the number of cirrus pixels retained as functions of PWV and H are not linear or uniform. For example, for a PWV threshold of 0.75 cm (squares), the FAPWV,H and cirrus pixels retained decrease for H values ranging from 1 to 3 km. For H = 3.5 km, the cirrus pixels retained continues to decrease, while the FAPWV,H increases. This phenomenon becomes more pronounced as PWV increases. Another interesting finding is that widely different PWV and H combinations can produce similar results in terms of FAPWV,H and cirrus pixels retained. Although this suggests that the exact choice of THPWV,H may not be of utmost importance and is likely somewhat subjective, Fig. 6 does show that using a lower PWV and higher H produces slightly better results, i.e., a lower FAPWV,H and comparable cirrus pixels retained, than a higher PWV and a lower H.

Based on Fig. 6, a PWV value of 0.10 cm at an altitude of 6 km (red star in Fig. 6) is chosen as THPWV,H in order to provide further analysis using an example. While this choice of THPWV,H is somewhat arbitrary, this combination of PWV and H appears to be the best choice in terms of retaining as many cirrus pixels as possible. This threshold produces a FAPWV,H of 13.9% and retains 82.1% of the original cirrus pixels. The results after applying this first proposed threshold are shown in Table 5. After applying TH10cm,6km over land, the fraction of the sample that is made up of clear-sky and low-/midlevel cloud pixels decreases. The false alarm rates for these subsamples also decreases slightly. The same is true for over-ocean pixels. However, clear-sky pixels change more significantly over land, while low-/midlevel cloud pixels change more significantly over ocean. This is likely due to the differences in the cloud sample density between land and ocean (Fig. 5).

Table 5

As in Table 3, but after the removal of pixels having a PWV6km of <0.10 cm.

Table 5

For comparison, the results of applying a threshold using an H value of 6 km and a PWVH value of 0.40 cm are shown in Table 6. This second proposed threshold THPWV,H produces a FAPWV,H of 13% and removes approximately 80% of all cirrus pixels from the overall sample. This threshold does not improve false alarm rates for cloudy and clear-sky samples, but does significantly decrease the fraction of clear-sky pixels from the overall sample, such that the sample is composed of nearly 70% cirrus pixels. Remember that some low-/midlevel clouds may be classified as clear sky within the CALIOP data. Returning to Fig. 1, the left side of the flowchart describes how the new pixel-rejection methodologies would precede the primary cirrus detection algorithm.

Table 6

As in Table 3, but after the removal of pixels having a PWV6km < 0.40 cm.

Table 6

4. Example application

Figure 7 shows the ABI channel 4 radiance for the same scene as in Fig. 2. Two different scaling methods are used to highlight the sensitivity of the cirrus band and the general magnitude of values that semitransparent cirrus exhibit (e.g., in the Gulf of Mexico). The results of the cirrus detection algorithm after applying the total-column PWV threshold (PWV > 0.4 cm) and the conservative (PWV at H = 6 km > 0.10 cm) pixel rejection threshold for the scene are shown in Fig. 8a. The colored shading indicates whether detected clouds were estimated to be optically thin (COD < 0.3; green shading) or not (COD > 0.3, purple shading). Pixels not passing either PWV threshold test are colored gray. PWV and above-cloud PWV values were computed using NAVGEM analysis data from 0000:00 UTC 20 June 2021. Comparing the two images, it is apparent that the PWV threshold tests remove pixels with generally dry columns (e.g., Texas) and also pixels that are dry aloft but not at the surface (e.g., off the coast of California). This case is during Northern Hemisphere summer; therefore, the atmosphere over the Rocky Mountains is relatively moist and pixels are not screened out there. However, the Andes are almost entirely screened out (not shown), suggesting the PWV threshold tests will be effective during winter when high elevation surfaces are snow covered. Other regions that appear to be problematic are the tropical eastern Pacific, where there is a broken stratocumulus deck, and the Sierra Madres and southwestern United States, where there are high-altitude cumulus fields. In these situations, tests based on spatial correlation may be more effective in eliminating false alarms due to the noise-like appearance cumulus fields have in the cirrus detection mask.

Fig. 7.
Fig. 7.

GOES-16 1.38-μm radiance (W m−2 sr−1) at 2200:20 UTC 19 Jun 2021 for a sector centered over the CONUS using two different scaling methods to highlight the ability of this channel.

Citation: Journal of Atmospheric and Oceanic Technology 39, 9; 10.1175/JTECH-D-21-0160.1

Fig. 8.
Fig. 8.

Application of the cirrus detection algorithm described in this study at 2200:20 UTC 19 Jun 2021 for a sector centered over the CONUS applying the total-column PWV threshold and (a) the conservative layer–PWV threshold (PWV > 0.10 cm; H = 6 km) and (b) the aggressive layer–PWV threshold (PWV > 0.40 cm; H = 6 km). Positive detections estimated to have a COD < 0.3 are shown in green. Positive detections estimated to have a COD > 0.3 are shown in purple. Pixels removed by the PWV thresholds are indicated as gray.

Citation: Journal of Atmospheric and Oceanic Technology 39, 9; 10.1175/JTECH-D-21-0160.1

Figure 8b shows the results of the cirrus detection algorithm after applying the total-column PWV threshold and the aggressive (PWV > 0.75 cm, H = 5 km) layer–PWV threshold. Using this threshold significantly reduces the number of samples the cirrus detection is performed on. However, nearly all of the overland pixels remaining contain thick clouds based on Figs. 2 and 7. This figure is in agreement with the sample distribution by cloud type shown in Table 6. Figures 8a and 8b suggest that robust thin cirrus detection over land comes at the expense of much of sample, and that preserving clear-sky pixels over land without a false alarm rate that is higher than acceptable requires much more algorithm complexity.

5. Summary and conclusions

This study develops an over-land methodology for pixel rejection in applying the MCH21 thin cirrus detection algorithm designed for GOES-16 ABI channel 4 radiance (1.378 μm) over water. Consistent with MCH21, the new algorithm is calibrated based on collocated Cloud–Aerosol Lidar with Orthogonal Polarization (CALIOP) cloud profiles. MCH21 show that precipitable water vapor (PWV) did not significantly impact cirrus detection over ocean due to the dark surface, given a relatively homogenous distribution of channel 4 radiances over open ocean waters. However, over land, PWV becomes the primary consideration in tuning response due to brighter and elevated surface terrain and relatively drier columns.

Clear-sky false alarm rates (see Table 2) over land were examined as a function of total-column integrated PWV to investigate the impact of surface contamination of the 1.378-μm band. We demonstrate that nearly all pixels having a value less than 0.4 cm produced false alarms. This suggests that an initial over-land threshold of total-column integrated PWV of approximately 0.4 cm is sufficient to ensure atmospheric opacity with respect to the land surface over most surface altitudes and brightness conditions. Applying this threshold reduces the clear-sky false alarm rate and sample fraction of clear-sky pixels. However, we caution with regards to complications encountered with collocation between CALIOP and GOES and representativeness bias between the CALIOP 5-km profile products and clouds present within GOES pixels.

Something not addressed, or ultimately considered noteworthy, in MCH21 was false alarms due to low-/midlevel cloud pixels, and is more thoroughly reconciled here. The number of these false alarms is much greater over land than over ocean due to inherent differences in cloud-top height occurrence frequency and total-column PWV. When looking at false alarms due to low-/midlevel cloud pixels as a function of above-cloud PWV, a similar relationship to clear-sky false alarms is seen. It appears that an above-cloud PWV of approximately 1 cm is sufficient to ensure that most low-/midlevel clouds are not detected by the 1.378-μm channel, which is significantly greater than what is necessary to filter clear-sky false alarms. Because the use of additional level-2 products, such as cloud-top height, is undesirable due to a significant increase in latency, a PWV threshold (THPWV,H) designed to remove false alarms due to low-/midlevel clouds was developed by computing the PWV for a layer between the top of the atmosphere (TOA) and a predetermined altitude H. After testing THPWV,H for all reasonable combinations of PWV and H values, it was shown that using a PWV of 0.10 cm integrated between the TOA and an altitude of 6 km was optimal for reducing the number of clear-sky and low-/midlevel cloud false alarms while also preserving the largest percentage of cirrus pixels.

Applying this threshold to the data, after also applying the total column PWV threshold of 0.4 cm, false alarm rates due to clear sky and low-/midlevel clouds were further reduced. However, the more significant change was the removal of clear-sky and low-/midlevel cloud pixels from the overall sample. An additional, more aggressive, threshold was also applied, which removed a large number of clear-sky and low-/midlevel cloud pixels and significantly reduced the overall false alarm rate; however, the reduction to the overall sample size was extreme. It is important to note that these PWV thresholds are dependent on the source from which they are obtained (e.g., NWP models, reanalysis). The specific values presented in this study should not be used as-is in an operational or scientific setting without calibration between PWV sources.

This study has shown that handling clear-sky contamination in the 1.378-μm channel is relatively simple, as the atmosphere becomes opaque to surface reflectance at relatively small PWV values. However, low- and midlevel clouds are a significant issue, especially over land. These may be handled with level-2 products such as operational cloud masks, though the methodology for collocating CALIOP and GOES made investigation of this difficult for this study. Now that the GOES ABI 1.378-μm channel has been calibrated for thin cirrus and clear-sky pixels over ocean and land, more complex methods for handling contamination due to aerosols and lower-tropospheric clouds can be explored. With enough data, the stricter thresholds presented here and in MCH21 can be used to detect cirrus clouds with high accuracy. Applications that make use of cirrus detection, such a large-scale cirrus microphysical property retrieval are currently being explored.

Acknowledgments.

This research was supported by the Office of Naval Research Grant N0001420WX00481. We graciously acknowledge the achievements of Naval Research Laboratory colleague Dr. Bo-Cai Gao in his groundbreaking work of demonstrating and advocating the potential for the 1.378-μm band as a means for identifying thin cirrus clouds.

Data availability statement.

GOES-16 data (radiances and ADP) used in this study are available at https://doi.org/10.7289/V5BV7DSR. CALIOP cloud profiles data are available at https://doi.org/10.5067/CALIOP/CALIPSO/LID_L2_05KMCPROSTANDARD-V4-20. Information on the GMTED data and access can be found at https://www.usgs.gov/land-resources/eros/coastal-changes-and-impacts/gmted2010?qt-science_support_ page_related_con50#qt-science_support_page_related_con.

REFERENCES

  • Ackerman, S. A., R. E. Holz, R. Frey, E. W. Eloranta, B. C. Maddux, and M. McGill, 2008: Cloud detection with MODIS. Part II: Validation. J. Atmos. Oceanic Technol., 25, 10731086, https://doi.org/10.1175/2007JTECHA1053.1.

    • Search Google Scholar
    • Export Citation
  • Bartlett, B., J. Casey, F. Padula, A. Pearlman, D. Pogorzala, and C. Cao, 2018: Independent validation of the Advanced Baseline Imager (ABI) on NOAA’s GOES-16: Post-launch ABI airborne science field campaign results. Proc. SPIE, 10764, 107640H, https://doi.org/10.1117/12.2323672.

  • Campbell, J. R., M. A. Vaughan, M. Oo, R. E. Holz, J. R. Lewis, and J. E. Welton, 2015: Distinguishing cirrus cloud presence in autonomous lidar measurements. Atmos. Meas. Tech., 8, 435449, https://doi.org/10.5194/amt-8-435-2015.

    • Search Google Scholar
    • Export Citation
  • Chew, B. N., J. R. Campbell, J. S. Reid, D. M. Giles, E. J. Welton, S. V. Salinas, and S. C. Liew, 2011: Tropical cirrus cloud contamination in sun photometer data. Atmos. Environ., 45, 67246731, https://doi.org/10.1016/j.atmosenv.2011.08.017.

    • Search Google Scholar
    • Export Citation
  • Danielson, J. J., and D. B. Gesch, 2011: Global multi-resolution terrain elevation data 2010 (GMTED2010). USGS Open-File Rep. 2011-1073, 26 pp., https://doi.org/10.3133/ofr20111073.

  • Dessler, A. E., and P. Yang, 2003: The distribution of tropical thin cirrus clouds inferred from Terra MODIS data. J. Climate, 16, 12411247, https://doi.org/10.1175/1520-0442(2003)16<1241:TDOTTC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Gao, B.-C., and Y. J. Kaufman, 1995: Selection of the 1.375-μm MODIS channel for remote sensing of cirrus clouds and stratospheric aerosols from space. J. Atmos. Sci., 52, 42314237, https://doi.org/10.1175/1520-0469(1995)052<4231:SOTMCF>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Gao, B.-C., A. F. H. Goetz, and W. J. Wiscombe, 1993: Cirrus cloud detection from Airborne Imaging Spectrometer data using the 1.38 μm water vapor band. Geophys. Res. Lett., 20, 301304, https://doi.org/10.1029/93GL00106.

    • Search Google Scholar
    • Export Citation
  • Gelaro, R., and Coauthors, 2017: The Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2). J. Climate, 30, 54195454, https://doi.org/10.1175/JCLI-D-16-0758.1.

    • Search Google Scholar
    • Export Citation
  • Heidinger, A. K., 2010: ABI cloud mask. NOAA/NESDIS/STAR Algorithm Theoretical Basis Doc., 93 pp.

  • Heidinger, A. K., 2012: ABI cloud height. NOAA/NESDIS/STAR Algorithm Theoretical Basis Doc., version 3.0, 79 pp., https://www.star.nesdis.noaa.gov/goesr/documents/ATBDs/Baseline/ATBD_GOES-R_Cloud%20Height_v3.0_July%202012.pdf.

  • Hogan, T. F., and Coauthors, 2014: The Navy Global Environmental Model. Oceanography, 27 (3), 116125, https://doi.org/10.5670/oceanog.2014.73.

    • Search Google Scholar
    • Export Citation
  • Justice, C. O., and Coauthors, 1998: The Moderate Resolution Imaging Spectroradiometer (MODIS): Land remote sensing for global change research. IEEE Trans. Geosci. Remote Sens., 36, 12281249, https://doi.org/10.1109/36.701075.

    • Search Google Scholar
    • Export Citation
  • Lee, J., P. Yang, A. E. Dessler, B. C. Gao, and S. Platnick, 2009: Distribution and radiative forcing of tropical thin cirrus clouds. J. Atmos. Sci., 66, 37213731, https://doi.org/10.1175/2009JAS3183.1.

    • Search Google Scholar
    • Export Citation
  • Lolli, S., and Coauthors, 2017: Daytime top-of-the-atmosphere cirrus cloud radiative forcing properties at Singapore. J. Appl. Meteor. Climatol., 56, 12491257, https://doi.org/10.1175/JAMC-D-16-0262.1.

    • Search Google Scholar
    • Export Citation
  • Mace, G. G., and Q. Zhang, 2014: The CloudSat radar‐lidar geometrical profile product (RL‐GeoProf): Updates, improvements, and selected results. J. Geophys. Res. Atmos., 119, 94419462, https://doi.org/10.1002/2013JD021374.

    • Search Google Scholar
    • Export Citation
  • Marquis, J. W., A. S. Bogdanoff, J. R. Campbell, J. A. Cummings, D. L. Westphal, N. J. Smith, and J. Zhang, 2017: Estimating infrared radiometric satellite sea surface temperature retrieval cold biases in the tropics due to unscreened optically thin cirrus clouds. J. Atmos. Oceanic Technol., 34, 355373, https://doi.org/10.1175/JTECH-D-15-0226.1.

    • Search Google Scholar
    • Export Citation
  • McHardy, T. M., and Coauthors, 2021: Advancing maritime transparent cirrus detection using the Advanced Baseline Imager “cirrus” band. J. Atmos. Oceanic Technol., 38, 10931110, https://doi.org/10.1175/JTECH-D-20-0130.1.

    • Search Google Scholar
    • Export Citation
  • Meyer, K., and S. Platnick, 2010: Utilizing the MODIS 1.38 μm channel for cirrus cloud optical thickness retrievals: Algorithm and retrieval uncertainties. J. Geophys. Res., 115, D24209, https://doi.org/10.1029/2010JD014872.

    • Search Google Scholar
    • Export Citation
  • Meyer, K., P. Yang, and B. C. Gao, 2004: Optical thickness of tropical cirrus clouds derived from the MODIS 0.66 and 1.375-μm channels. IEEE Trans. Geosci. Remote Sens., 42, 833841, https://doi.org/10.1109/TGRS.2003.818939.

    • Search Google Scholar
    • Export Citation
  • NOAA/NESDIS, 2018: ABI aerosol detection product. NOAA/NESDIS NOAA/NESDIS/STAR Algorithm Theoretical Basis Doc., version 3.0, 84 pp., https://www.star.nesdis.noaa.gov/goesr/documents/ATBDs/Baseline/ATBD_GOES-R_Aerosol_Detection_v3.0_Jan2019.pdf.

  • Rossow, W. B., and R. A. Schiffer, 1999: Advances in understanding clouds from ISCCP. Bull. Amer. Meteor. Soc., 80, 22612288, https://doi.org/10.1175/1520-0477(1999)080<2261:AIUCFI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Sassen, K., and B. S. Cho, 1992: Subvisual–thin cirrus lidar dataset for satellite verification and climatological research. J. Appl. Meteor., 31, 12751285, https://doi.org/10.1175/1520-0450(1992)031<1275:STCLDF>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Sassen, K., and J. R. Campbell, 2001: A midlatitude cirrus cloud climatology from the Facility for Atmospheric Remote Sensing. Part I: Macrophysical and synoptic properties. J. Atmos. Sci., 58, 481496, https://doi.org/10.1175/1520-0469(2001)058<0481:AMCCCF>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • 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, 10311049, https://doi.org/10.1175/BAMS-D-12-00117.1.

    • Search Google Scholar
    • Export Citation
  • Vaughan, M. A., D. M. Winker, and K. A. Powell, 2005: CALIOP Algorithm Theoretical Basis Document—Part 2: Feature detection and layer properties algorithms. NASA Rep. PC-SCI-202, 87 pp., https://www-calipso.larc.nasa.gov/resources/pdfs/PC-SCI-202_Part2_rev1x01.pdf.

  • Vaughan, M. A., and Coauthors, 2020: Cloud–Aerosol Lidar Infrared Pathfinder satellite observations data management system data products catalog. NASA Doc. PC-SCI-503, 273 pp., https://ntrs.nasa.gov/api/citations/20220003549/downloads/CALIPSO_DPC_Rev4x94.pdf.

  • Walther, A., W. Straka, and A. K. Heidinger, 2013: ABI Algorithm Theoretical Basis Document for Daytime Cloud Optical and Microphysical Properties (DCOMP). NOAA/NESDIS/STAR Algorithm Theoretical Basis Doc., version 3.0, 66 pp., https://www.star.nesdis.noaa.gov/goesr/documents/ATBDs/Baseline/ATBD_GOES-R_Cloud_DCOMP_v3.0_Jun2013.pdf.

  • Wang, C., P. Yang, A. Dessler, B. A. Baum, and Y. Hu, 2014: Estimation of the cirrus cloud scattering phase function from satellite observations. J. Quant. Spectrosc. Radiat. Transfer, 138, 3649, https://doi.org/10.1016/j.jqsrt.2014.02.001.

    • Search Google Scholar
    • Export Citation
  • Winker, D. M., W. H. Hunt, and M. J. McGill, 2007: Initial performance assessment of CALIOP. Geophys. Res. Lett., 34, L19803, https://doi.org/10.1029/2007GL030135.

    • Search Google Scholar
    • Export Citation
  • Winker, D. M., and Coauthors, 2009: Overview of the CALIPSO mission and CALIOP data processing algorithms. J. Atmos. Oceanic Technol., 26, 23102323, https://doi.org/10.1175/2009JTECHA1281.1.

    • Search Google Scholar
    • Export Citation
  • Young, S. A., and M. A. Vaughan, 2009: The retrieval of profiles of particulate extinction from Cloud-Aerosol Lidar Infrared Pathfinder Satellite Observations (CALIPSO) data: Algorithm description. J. Atmos. Oceanic Technol., 26, 11051119, https://doi.org/10.1175/2008JTECHA1221.1.

    • Search Google Scholar
    • Export Citation
  • Young, S. A., M. A. Vaughan, A. Garnier, J. L. Tackett, J. D. Lambeth, and K. A. Powell, 2018: Extinction and optical depth retrievals for CALIPSO’s version 4 data release. Atmos. Meas. Tech., 11, 57015727, https://doi.org/10.5194/amt-11-5701-2018.

    • Search Google Scholar
    • Export Citation
  • Yu, F., X. Wu, X. Shao, B. Efremova, H. Yoo, H. Qian, and B. Iacovazzi, 2017: Early radiometric calibration performances of GOES-16 Advanced Baseline Imager. Proc. SPIE, 10402, 104020S, https://doi.org/10.1117/12.2275195.

    • Search Google Scholar
    • Export Citation
Save
  • Ackerman, S. A., R. E. Holz, R. Frey, E. W. Eloranta, B. C. Maddux, and M. McGill, 2008: Cloud detection with MODIS. Part II: Validation. J. Atmos. Oceanic Technol., 25, 10731086, https://doi.org/10.1175/2007JTECHA1053.1.

    • Search Google Scholar
    • Export Citation
  • Bartlett, B., J. Casey, F. Padula, A. Pearlman, D. Pogorzala, and C. Cao, 2018: Independent validation of the Advanced Baseline Imager (ABI) on NOAA’s GOES-16: Post-launch ABI airborne science field campaign results. Proc. SPIE, 10764, 107640H, https://doi.org/10.1117/12.2323672.

  • Campbell, J. R., M. A. Vaughan, M. Oo, R. E. Holz, J. R. Lewis, and J. E. Welton, 2015: Distinguishing cirrus cloud presence in autonomous lidar measurements. Atmos. Meas. Tech., 8, 435449, https://doi.org/10.5194/amt-8-435-2015.

    • Search Google Scholar
    • Export Citation
  • Chew, B. N., J. R. Campbell, J. S. Reid, D. M. Giles, E. J. Welton, S. V. Salinas, and S. C. Liew, 2011: Tropical cirrus cloud contamination in sun photometer data. Atmos. Environ., 45, 67246731, https://doi.org/10.1016/j.atmosenv.2011.08.017.

    • Search Google Scholar
    • Export Citation
  • Danielson, J. J., and D. B. Gesch, 2011: Global multi-resolution terrain elevation data 2010 (GMTED2010). USGS Open-File Rep. 2011-1073, 26 pp., https://doi.org/10.3133/ofr20111073.

  • Dessler, A. E., and P. Yang, 2003: The distribution of tropical thin cirrus clouds inferred from Terra MODIS data. J. Climate, 16, 12411247, https://doi.org/10.1175/1520-0442(2003)16<1241:TDOTTC>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Gao, B.-C., and Y. J. Kaufman, 1995: Selection of the 1.375-μm MODIS channel for remote sensing of cirrus clouds and stratospheric aerosols from space. J. Atmos. Sci., 52, 42314237, https://doi.org/10.1175/1520-0469(1995)052<4231:SOTMCF>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Gao, B.-C., A. F. H. Goetz, and W. J. Wiscombe, 1993: Cirrus cloud detection from Airborne Imaging Spectrometer data using the 1.38 μm water vapor band. Geophys. Res. Lett., 20, 301304, https://doi.org/10.1029/93GL00106.

    • Search Google Scholar
    • Export Citation
  • Gelaro, R., and Coauthors, 2017: The Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2). J. Climate, 30, 54195454, https://doi.org/10.1175/JCLI-D-16-0758.1.

    • Search Google Scholar
    • Export Citation
  • Heidinger, A. K., 2010: ABI cloud mask. NOAA/NESDIS/STAR Algorithm Theoretical Basis Doc., 93 pp.

  • Heidinger, A. K., 2012: ABI cloud height. NOAA/NESDIS/STAR Algorithm Theoretical Basis Doc., version 3.0, 79 pp., https://www.star.nesdis.noaa.gov/goesr/documents/ATBDs/Baseline/ATBD_GOES-R_Cloud%20Height_v3.0_July%202012.pdf.

  • Hogan, T. F., and Coauthors, 2014: The Navy Global Environmental Model. Oceanography, 27 (3), 116125, https://doi.org/10.5670/oceanog.2014.73.

    • Search Google Scholar
    • Export Citation
  • Justice, C. O., and Coauthors, 1998: The Moderate Resolution Imaging Spectroradiometer (MODIS): Land remote sensing for global change research. IEEE Trans. Geosci. Remote Sens., 36, 12281249, https://doi.org/10.1109/36.701075.

    • Search Google Scholar
    • Export Citation
  • Lee, J., P. Yang, A. E. Dessler, B. C. Gao, and S. Platnick, 2009: Distribution and radiative forcing of tropical thin cirrus clouds. J. Atmos. Sci., 66, 37213731, https://doi.org/10.1175/2009JAS3183.1.

    • Search Google Scholar
    • Export Citation
  • Lolli, S., and Coauthors, 2017: Daytime top-of-the-atmosphere cirrus cloud radiative forcing properties at Singapore. J. Appl. Meteor. Climatol., 56, 12491257, https://doi.org/10.1175/JAMC-D-16-0262.1.

    • Search Google Scholar
    • Export Citation
  • Mace, G. G., and Q. Zhang, 2014: The CloudSat radar‐lidar geometrical profile product (RL‐GeoProf): Updates, improvements, and selected results. J. Geophys. Res. Atmos., 119, 94419462, https://doi.org/10.1002/2013JD021374.

    • Search Google Scholar
    • Export Citation
  • Marquis, J. W., A. S. Bogdanoff, J. R. Campbell, J. A. Cummings, D. L. Westphal, N. J. Smith, and J. Zhang, 2017: Estimating infrared radiometric satellite sea surface temperature retrieval cold biases in the tropics due to unscreened optically thin cirrus clouds. J. Atmos. Oceanic Technol., 34, 355373, https://doi.org/10.1175/JTECH-D-15-0226.1.

    • Search Google Scholar
    • Export Citation
  • McHardy, T. M., and Coauthors, 2021: Advancing maritime transparent cirrus detection using the Advanced Baseline Imager “cirrus” band. J. Atmos. Oceanic Technol., 38, 10931110, https://doi.org/10.1175/JTECH-D-20-0130.1.

    • Search Google Scholar
    • Export Citation
  • Meyer, K., and S. Platnick, 2010: Utilizing the MODIS 1.38 μm channel for cirrus cloud optical thickness retrievals: Algorithm and retrieval uncertainties. J. Geophys. Res., 115, D24209, https://doi.org/10.1029/2010JD014872.

    • Search Google Scholar
    • Export Citation
  • Meyer, K., P. Yang, and B. C. Gao, 2004: Optical thickness of tropical cirrus clouds derived from the MODIS 0.66 and 1.375-μm channels. IEEE Trans. Geosci. Remote Sens., 42, 833841, https://doi.org/10.1109/TGRS.2003.818939.

    • Search Google Scholar
    • Export Citation
  • NOAA/NESDIS, 2018: ABI aerosol detection product. NOAA/NESDIS NOAA/NESDIS/STAR Algorithm Theoretical Basis Doc., version 3.0, 84 pp., https://www.star.nesdis.noaa.gov/goesr/documents/ATBDs/Baseline/ATBD_GOES-R_Aerosol_Detection_v3.0_Jan2019.pdf.

  • Rossow, W. B., and R. A. Schiffer, 1999: Advances in understanding clouds from ISCCP. Bull. Amer. Meteor. Soc., 80, 22612288, https://doi.org/10.1175/1520-0477(1999)080<2261:AIUCFI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Sassen, K., and B. S. Cho, 1992: Subvisual–thin cirrus lidar dataset for satellite verification and climatological research. J. Appl. Meteor., 31, 12751285, https://doi.org/10.1175/1520-0450(1992)031<1275:STCLDF>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Sassen, K., and J. R. Campbell, 2001: A midlatitude cirrus cloud climatology from the Facility for Atmospheric Remote Sensing. Part I: Macrophysical and synoptic properties. J. Atmos. Sci., 58, 481496, https://doi.org/10.1175/1520-0469(2001)058<0481:AMCCCF>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • 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, 10311049, https://doi.org/10.1175/BAMS-D-12-00117.1.

    • Search Google Scholar
    • Export Citation
  • Vaughan, M. A., D. M. Winker, and K. A. Powell, 2005: CALIOP Algorithm Theoretical Basis Document—Part 2: Feature detection and layer properties algorithms. NASA Rep. PC-SCI-202, 87 pp., https://www-calipso.larc.nasa.gov/resources/pdfs/PC-SCI-202_Part2_rev1x01.pdf.

  • Vaughan, M. A., and Coauthors, 2020: Cloud–Aerosol Lidar Infrared Pathfinder satellite observations data management system data products catalog. NASA Doc. PC-SCI-503, 273 pp., https://ntrs.nasa.gov/api/citations/20220003549/downloads/CALIPSO_DPC_Rev4x94.pdf.

  • Walther, A., W. Straka, and A. K. Heidinger, 2013: ABI Algorithm Theoretical Basis Document for Daytime Cloud Optical and Microphysical Properties (DCOMP). NOAA/NESDIS/STAR Algorithm Theoretical Basis Doc., version 3.0, 66 pp., https://www.star.nesdis.noaa.gov/goesr/documents/ATBDs/Baseline/ATBD_GOES-R_Cloud_DCOMP_v3.0_Jun2013.pdf.

  • Wang, C., P. Yang, A. Dessler, B. A. Baum, and Y. Hu, 2014: Estimation of the cirrus cloud scattering phase function from satellite observations. J. Quant. Spectrosc. Radiat. Transfer, 138, 3649, https://doi.org/10.1016/j.jqsrt.2014.02.001.

    • Search Google Scholar
    • Export Citation
  • Winker, D. M., W. H. Hunt, and M. J. McGill, 2007: Initial performance assessment of CALIOP. Geophys. Res. Lett., 34, L19803, https://doi.org/10.1029/2007GL030135.

    • Search Google Scholar
    • Export Citation
  • Winker, D. M., and Coauthors, 2009: Overview of the CALIPSO mission and CALIOP data processing algorithms. J. Atmos. Oceanic Technol., 26, 23102323, https://doi.org/10.1175/2009JTECHA1281.1.

    • Search Google Scholar
    • Export Citation
  • Young, S. A., and M. A. Vaughan, 2009: The retrieval of profiles of particulate extinction from Cloud-Aerosol Lidar Infrared Pathfinder Satellite Observations (CALIPSO) data: Algorithm description. J. Atmos. Oceanic Technol., 26, 11051119, https://doi.org/10.1175/2008JTECHA1221.1.

    • Search Google Scholar
    • Export Citation
  • Young, S. A., M. A. Vaughan, A. Garnier, J. L. Tackett, J. D. Lambeth, and K. A. Powell, 2018: Extinction and optical depth retrievals for CALIPSO’s version 4 data release. Atmos. Meas. Tech., 11, 57015727, https://doi.org/10.5194/amt-11-5701-2018.

    • Search Google Scholar
    • Export Citation
  • Yu, F., X. Wu, X. Shao, B. Efremova, H. Yoo, H. Qian, and B. Iacovazzi, 2017: Early radiometric calibration performances of GOES-16 Advanced Baseline Imager. Proc. SPIE, 10402, 104020S, https://doi.org/10.1117/12.2275195.

    • Search Google Scholar
    • Export Citation
  • Fig. 1.

    An updated algorithm flowchart showing the methodologies for pixel rejection, based on Fig. 9. In MCH21.

  • Fig. 2.

    (a) GOES ABI visible (channel 2) imagery for 2200:20 UTC 19 Jun 2021 for a sector centered over the CONUS and (b) the corresponding cirrus mask as described in MCH21.

  • Fig. 3.

    (a),(d) Sample density, (b),(e) false alarm rate, and (c),(f) negative rate for clear-sky overland samples binned by total-column PWV and surface elevation, with bin sizes of 0.1, respectively. Rows show collocation time limits of (top) ±1 and (bottom) ±7.5 min.

  • Fig. 4.

    The false alarm rate (black lines, left y axes) and total number of samples (red lines, right y axes) remaining after removing pixels having a PWV less than the PWV threshold, which is varied (x axes) for clear-sky samples. Collocation time limits of (a) ±1 and (b) ±7.5 min are shown. The chosen PWV threshold value of 0.4 cm is indicated by the dashed lines.

  • Fig. 5.

    (a),(d) Sample density, (b),(e) false alarm rate, and (c),(f) negative rate for low-/midlevel cloud samples binned by above-cloud PWV and CALIOP cloud-top height (CTH), with bin sizes of 0.1, for (top) ocean and (bottom) land.

  • Fig. 6.

    The fraction of cirrus pixels retained and the false alarm rate after applying THPWV,H. The PWV threshold value PWVH is indicated by the shape and the height threshold value H is indicated by the color.

  • Fig. 7.

    GOES-16 1.38-μm radiance (W m−2 sr−1) at 2200:20 UTC 19 Jun 2021 for a sector centered over the CONUS using two different scaling methods to highlight the ability of this channel.

  • Fig. 8.

    Application of the cirrus detection algorithm described in this study at 2200:20 UTC 19 Jun 2021 for a sector centered over the CONUS applying the total-column PWV threshold and (a) the conservative layer–PWV threshold (PWV > 0.10 cm; H = 6 km) and (b) the aggressive layer–PWV threshold (PWV > 0.40 cm; H = 6 km). Positive detections estimated to have a COD < 0.3 are shown in green. Positive detections estimated to have a COD > 0.3 are shown in purple. Pixels removed by the PWV thresholds are indicated as gray.

All Time Past Year Past 30 Days
Abstract Views 351 60 0
Full Text Views 241 133 9
PDF Downloads 204 88 13