Document
P7.1 REAL-TIME CLOUD, RADIATION, AND AIRCRAFT ICING PARAMETERS FROM GOES OVER THE USA Patrick Minnis, Louis Nguyen, William Smith, Jr., David Young Atmospheric Sciences, NASA Langley Research Center, Hampton, VA 23681 Mandana Khaiyer, Rabindra Palikonda, Douglas Spangenberg, Dave Doelling, Dung Phan, Greg Nowicki, Kirk Ayers AS&M, Inc., Hampton, VA 23666 Patrick W. Heck CIMSS, University of Wisconsin-Madison, Madison, WI, USA 53706 Cory Wolff NCAR, Boulder, CO, USA 80301 1. INTRODUCTION 2. DATA & METHODOLOGY New sensors and retrieval algorithms have facilitated The USA domain covers 25°N - 50°N and 65°W - the development of more quantitative uses for geosta- 125°W. Datasets used here include half-hourly GOES- tionary (GEO) satellite data in weather and climate. For 10, and 12 4-km spectral radiances. The GOES-10 at weather applications such as data assimilation or air 135°W measures radiances at 0.65, 3.9, 10.8, and 12 safety, near real-time processing is needed to produce μm while the 12-μm channel on GOES-12 at 75°W was datasets in a timely fashion. It requires not only fast data replaced with a 13.3-μm channel. GOES-12 data are throughput, readily available on many computer analyzed over an area between 65°W and 105°W, while systems, but good calibration of the sensors to ensure the GOES-10 data cover 90°W to 125°W. The results accuracy of the products. Reliable calibration, formerly a are stitched together at 99°W. The Rapid Update Cycle chronic problem for some GEO imager channels, can be (RUC) analyses (Benjamin et al., 2004) provide hourly readily acquired using the more commonly operating, profiles of temperature and humidity at spatial well-calibrated research satellites (Minnis et al., 2002). resolutions of 40 and 20 km before and after April 2002, Satellite-derived cloud and radiation products such as respectively. The RUC data are used to convert the cloud phase or albedo have been available for a number retrieved cloud temperature Tc to cloud height z and c of years from satellites devoted to climate studies but correct radiances for atmospheric attenuation. Surface- have been generated only sporadically from operational type, clear-sky albedo, and surface emissivity maps are weather satellites. With improving forecast models and used to estimate the cloud-free radiances for a given the ever-present need for air safety information, such scene as described by Minnis et al. (2001, 2004).
products will be useful for model validation and The analysis procedure is outlined in Fig. 1, where assimilation and for diagnosing air-traffic hazards. the green boxes indicate relatively fixed input Minnis et al. (2001) adapted a set of algorithms used parameters and the light blue denotes the input varying on low-Earth orbit satellite data for Earth radiation at each time step. The data are processed as tiles (1° budget and cloud process studies to provide a variety of region). After estimating the clear-sky radiances for pixel-level products for the Atmospheric Radiation each tile (gray boxes in center), each pixel is classified Measurement (ARM) Program over the central United as clear or cloudy (tan box) based on a set of decision States of America (USA) in near real-time. Smith et al. trees using all four channels. If clear pixels are found in (2002, 2003) demonstrated that several of the derived the tile, they replace the original clear radiance field for products could be useful for diagnosing aircraft icing the tile and are used to derive the cloud properties for potential. Because icing can occur anywhere over the each cloudy pixel. For each clear pixel, the algorithm USA, the products should not be limited the central USA but should be available over the entire country. With the aid of the NASA Advanced Satellite Aviation-weather Products (ASAP) program, the ARM domain has recently been expanded to include the entire continental USA requiring the combination of 4-km data form both the East and West Geostationary Operational Environ- mental Satellites (GOES). This paper summarizes the current state of those products and their applicability to aircraft icing potential.
* Corresponding author address : Patrick Minnis, NASA Langley Research Center, MS 420, Hampton, VA 23681-2199. email: p.minnis@larc.nasa.gov.
Fig. 1. Schematic real-time processing flow.
th 13 AMS Conf. on Satellite Oceanography and Meteorology, 20-24 September 2004, Norfolk, VA estimates surface skin temperature, the outgoing for re = 5 μm, and longwave radiation (OLR), and the clear-sky VIS and shortwave (ASW) albedos (not indicated in Fig. 1). IP = 0.138 ln( LWP ) – 0.024, (2) When the solar zenith angle SZA is less than 82° (daytime), cloudy pixels are analyzed with the visible for re = 16 μm. Linear interpolation between the results infrared solar-infrared split-window technique (VISST; of (1) and (2) are used for pixels with re between 5 and Minnis et al., 1995), which matches the observed values 16 μm. Pixels with larger or smaller values of re are with theoretical models of cloud reflectance and assigned the appropriate extreme value. The intensity of emittance (Minnis et al. 1998). At night, the cloud icing is classified as light or moderate-severe if LWP is -2 properties are determined using the solar-infrared less or greater than 440 gm , respectively. This infrared split-window technique (SIST), an improved approach to estimating the probability for icing is version of the 3-channel nighttime method of Minnis et considered as a preliminary technique because it is al. (1995). It uses 3.9, 10.8, and 12.0-μm data for based on only 11 days of PIREPS and GOES-12 data GOES-10 and only the first two channels for GOES-12. taken during February 2004. It will be evaluated using The methods estimate effective cloud temperature Tc , an additional 18 and 11 days of GOES-12 and 10 data, cloud height z and thickness h , phase, optical depth OD , respectively.
effective droplet radius r e or effective ice crystal diameter De , and LWP or ice water path IWP . Pixels 3. RESULTS with Tc < 273 K and identified as liquid water are Figure 2 shows an example of the USA results for designated as supercooled liquid (SLW) clouds. Other GOES-10/12 imagery taken at 1845 UTC, 15 March properties related to icing include h , O D , and LWP .
2004. The pseudocolor RGB image (Fig. 2a) reveals the Cloud thickness, used to find cloud base height, is various cloud types in a single image. In this type of estimated using the empirical parameterizations based image, red is assigned to the visible reflectance, the on ARM cloud radar and satellite data ( Chakrapani et temperature difference between the 3.9 and 11-μm al., 2001). The top-of-atmosphere VIS albedo, ASW, channels determines the green intensity, and the 11-μm and OLR are also computed for each cloudy pixel. The temperature T provides the blue intensity on an inverse results are output for each tile (red box in Fig. 1).
scale. Snow-free clear areas like much of the western These results are then used to estimate the aircraft USA and northern Mexico are green, blue, or tan while icing probability for each pixel in a tile. The potential for clear snow-covered areas such as Colorado, Wyoming, aircraft icing depends on many factors related to the the Sierra Nevada Mountains, western Iowa, northern particular aircraft and the weather conditions. Some Minnesota, and Quebec are typically bright or dark pink.
aircraft will accumulate ice in certain conditions while High clouds are generally white (e.g, Montana), grey other planes will remain ice-free in the same cloud.
(e.g., Sea of Cortez), or some shade of magenta (e.g., These aircraft-related factors are not considered here. A North Carolina to Michigan), while low or midlevel necessary condition for icing is the presence of clouds are often white (e.g., Oregon) or a shade of supercooled liquid water, relatively large droplets, peach or orange (e.g., Texas and Arkansas). The and/or large concentrations of droplets or high liquid retrieved cloud phase image (Fig. 2b) shows clear areas water content (LWC). SLW can be discriminated from in green, warm liquid water clouds in dark blue, SLW warm clouds using Tc while the concentration of large clouds in light blue, and ice clouds in red. Some of the droplets should be related to re . LWC can be estimated scattered clouds over the western US are actually as the ratio of LWP/ h . However, since both LWP and h misclassified clear areas because the clear-sky albedo depend on OD , only LWP is used here as a proxy for map has not yet been optimized for the GOES VIS LWC. Smith et al. (2002, 2003) found some weak spectral band or they occur near the edge of snow positive dependencies of icing intensity on LWP and re , fields. SLW clouds cover nearly all of the northeastern and a weak negative dependency on Tc using matched USA. The derived values of re (Fig. 2c) are typically VISST and in situ aircraft data.
between 7 and 11 μm but greater values occur off the Minnis et al. (2004b) developed a probability based Washington coast and over southeastern Texas. In method to classify the icing potential for each pixel some overlapped conditions along the edges of ice based on pilot reports (PIREPS). No pixels classified as -2 clouds (e.g., Pennsylvania, Wisconsin) or for relatively cloudy with Tc > 273 K or LWP < 40 gm , as clear, or as thin clouds over snow (northern Nevada), the VISST an ice cloud with O D < 5 are considered as icing often retrieves a large value of re because the 3.9-μm potential candidates. Ice clouds with O D > 5 are radiance is diminished due to absorption by ice in the considered as indeterminate pixels because the nature form of large cirrus crystals or snow grains. The cloud of any clouds below the ice cloud cannot be discerned.
L W P (Fig. 2e) reaches extremely high values, The approach estimates the icing probability for the -2 exceeding 400 gm , around Lake Michigan and the remaining pixels is estimated as Pacific Northwest. More commonly, LWP is less than -2 200 gm . The cloud top heights (Fig. 1d) range between IP = 0.147 ln( LWP ) – 0.084, (1) 8 and 10 km over northern Florida and in the Northwest.
Most of the cloud tops are below 3 km. The bases (Fig.
th 13 AMS Conf. on Satellite Oceanography and Meteorology, 20-24 September 2004, Norfolk, VA Fig. 2. Selected cloud and aircraft icing parameters from GOES10 and 12, 1845 UTC, 18 March 2004. Gray indicates ice clouds in (a) and (e) and indeterminate icing in (g) and (h). Color bar ranges: (b) dark blue, warm clouds; light blue, SLW; red or pink, -2 ice clouds (c) 5 (blue) to 21 μ m (red), > 21 μ m, dark red; (d & f) 0 – 10 km (purple –red), 10 –16 km; (e) 0-300 gm , purple to -2 -2 light green; 300 – 400 gm , yellow; 400 – 1000 gm , orange to dark red; (g) low – blue, medium – yellow, high – red; (h) low – blue, moderate/severe - red.
2f) for most of the low clouds are estimated at 1 km or to-high values off the Baja coast are most likely an less while the high cloud bases are between 3 and 7 artifact of thin cirrus over warm low-level clouds (Fig.
km. Most of the SLW clouds have icing probabilities 2b) that are sometimes misinterpreted as SLW clouds.
between low (0-33%) and medium (34-67%) with high (> The potential icing intensity is generally light except in 67%) probabilities around Lake Michigan. The medium- the Northwest and around Lake Michigan (Fig. 2h).
th 13 AMS Conf. on Satellite Oceanography and Meteorology, 20-24 September 2004, Norfolk, VA Southeast are a few kilometers higher than the ASOS values. These are the same areas classified as indeterminate by the icing algorithm because it is not possible to determine if a low cloud is below the high cloud observed form the satellite. The ASOS results indicate that a low cloud deck is present under the ice clouds and based on the properties of low clouds in the surrounding areas, significant icing potential is likely in those regions where the cloud base is too high relative to the ASOS. Thus, it appears that merging of the ceilometer data with the VISST results could be valuable for estimating the icing potential for indeterminate cases.
In situ data from a variety of field measurements as well as PIREPS have been used to validate the satellite icing measurements (Smith et al., 2002, 2003; Minnis et Fig. 3. Cloud base heights from ASOS ceilometers, 1900 al., 2004b). For example, Smith et al. (2000) showed UTC, 18 March 2004.
that when overlying cirrus clouds were absent, the VISST retrieved SLW in 98% of the PIREPS reports of The cloud property and icing methodologies are positive icing indicating that the satellite can provide the currently being applied in near-real time to GOES-10 first condition necessary to identify icing conditions.
and 12 data every half hour during the daytime and Determining the other conditions becomes more difficult.
hourly at night. The results from each satellite are Thus, validation and improvement of the prototype available separately and stitched together at the URL, algorithm has begun by comparing the satellite http://www-angler.larc.nasa.gov/satimage/products .html, retrievals with in situ data from field programs and in image or digital formats. The vertical extent of the additional PIREPS.
potential icing clouds can be estimated from the cloud- All PIREPS taken within a half hour of 1545 and top and base altitudes of the clouds.
2145 UTC over the GOES-12 domain during February 2004 and over the GOES-10 domain from 1 –11 4. DISCUSSION February 2004 were compared with the VISST results.
The average cloud properties and icing probabilities The example results presented above serve as were computed for 3 x 3 pixel arrays centered on the samples of the products currently being generated, but position given in each PIREP. The results, based on the algorithms used to derive each product are 4665 matches, are summarized in Table 1 for the cloud continually being updated as validation studies provide properties. Mean cloud O D , L W P , and thickness all new information or advanced methods become increase with increasing icing intensity and the average available. Validation efforts have demonstrated that cloud temperature is lower for the cases with positive properties like cloud optical depth, particle size, and ice PIREPS icing. The mean effective cloud droplet sizes, and liquid water path (e.g., Young et al., 1998; Dong et however, are not significantly different for icing and no- al., 2002; Min et al., 2004) are reasonably well icing cases. Although the standard deviations are large, correlated with and similar in magnitude to in situ and especially for LWP, the mean results confirm the LWP active remote sensing retrievals. The validations are basis of the icing probability algorithm. The dependence continuing.
on re appears to be weaker than seen in earlier Figure 3 shows the cloud bases measured with the comparisons, but the temperature dependence seems Automated Surface Observing Systems (ASOS) to be stronger and the cloud thickness means are well ceilometers at 1845 UTC, 18 March 2004. Many of the separated for icing and no icing cases.
VISST cloud base heights in Fig. 2f are in relatively good agreement ( + 0.5 km) with the ASOS data over Table 1. Mean VISST cloud properties corresponding to many areas such as Texas, Wisconsin, New York, PIREPS icing reports during February 2004.
Washington, and Pennsylvania. In other areas where Icing Intensity None Light Mod/Sev clouds are high or icing is indeterminate, the agreement is mixed. The bases are in good agreement over OD 18.1 34.0 39.1 Florida, Idaho, and North Dakota, but not over Quebec, North Carolina, and Georgia. Over Quebec, the re (μm) 12.2 11.8 12.4 retrieved cloud tops are lower than the ASOS bases, -2 indicating that the VISST is overestimating the cloud OD LWP (gm ) 269 382 460 resulting in too little correction of T to determine Tc . This h (km) 1.41 2.1 2.4 error is likely a result of the snow cover over Quebec, which causes a large uncertainty in the clear-sky Tc (K) 269 263 263 albedo. On the other hand, the retrieved bases over the th 13 AMS Conf. on Satellite Oceanography and Meteorology, 20-24 September 2004, Norfolk, VA The PIREPS reported icing in 54.5% of the cases. change that would require adjustment of the cloud-top When the PIREPS reported icing, the satellite algorithm altitude also.
produced 27% icing, 22% indeterminate, and 5.5% no Only daytime data have been considered so far.
icing relative to the total number of PIREPS. When the While the SIST has demonstrated some skill in PIREPS indicated no icing, the satellite algorithm discriminating between optically thin and thick clouds at produced 22.6% no icing, 8.5% indeterminate, and night, the utility of the resulting products for icing 14.6% positive icing probabilities. Thus, in 5.5% of the classification has not yet been examined. Most of the cases, the GOES missed the icing (false negatives) indeterminate cases were found to be a combination of while producing 14.6% false positives. Many of the false a high ice clouds over an icing cloud. Better detection of negatives result from relatively small LWP values and multilayered clouds using multispectral IR data or may be affected by the time differences between the matched microwave and VIS data over water surfaces satellite image and the PIREP report location. The false (Huang et al., 2004) would help minimize the number of positives are the result of using probability estimates indeterminate cases. Other ways to eliminate the and may coincide with some of the aircraft-related indeterminate cases include using the ceilometer data factors noted earlier. along with RUC soundings to locate a lower-level cloud Because the indeterminate cases account for 30% deck or, in a more universal approach, using empirical of the observations, it is desirable to account for the relationships between clouds and the RUC profiles (Yi potential of clouds underneath the ice clouds. The Cloud et al. 2004) to diagnose clouds underneath the satellite- Icing Potential (CIP) product (Bernstein et al. (2004) observed cirrus clouds. Similar methods are already already integrates RUC, PIREPS, and ceilometer data being used to develop the current CIP product and to estimate icing probability. Blending the satellite could be adapted to work in a conditional probability results with the CIP would be an attractive means for scenario with the satellite retrievals. False returns taking the indeterminate data into account. For caused by thin cirrus clouds over warm, low cloud decks assimilating the products into forecast models, it would can also be minimized by using the multispectral IR be optimal to produce a 3-D dataset by accounting for methods to detect thin cirrus clouds. Such techniques multilayered clouds. Such an approach would fill out the typically rely on the 12-μm data, which are currently not satellite data in the vertical using multilayered detection available GOES-12. Hopefully, future GOES imagers methods (e.g., Kawamoto et al., 2002), ceilometer cloud will return the 13.3-μm channel on GOES-12 to the base estimates, and RUC data linked to cloud radar original 12-μm channel.
data (Minnis et al., 2004c). This technique for The satellite icing algorithms are just one part of a developing a 3-D dataset needs further exploration. comprehensive aircraft icing program being developed It is critical to properly place the clouds in the right by NASA, NOAA, and the FAA. Ultimately, the results vertical location. For the 855 times when VISST and will be combined with PIREPS, model forecasts, and PIREPS reported icing, the aircraft mean altitude was other data within the CIP to provide a near-real time within the vertical boundaries from VISST in 70% of the optimized characterization of icing conditions for pilots cases. The VISST cloud-top height was too high in 26% and flight controllers.
of the cases and too low 4% of the time.
ACKNOWLEDGEMENTS 4. CONCLUDING REMARKS This research was supported by the NASA Aviation Safety Program through the NASA Advanced Satellite A preliminary new, physically based method for real- Aviation-weather Products Initiative. Additional support time estimation of the probability of icing conditions has was provided by the Environmental Sciences Division of been demonstrated using merged GOES-10 and 12 U.S. Department of Energy Interagency Agreement DE- data over the continental United States and southern AI02-97ER62341 through the ARM Program.
Canada. The algorithm produces pixel-level cloud and radiation properties as well as an estimate of icing REFERENCES probability with an associated intensity rating Because icing depends on so many different variables, such as Benjamin, S. G., D. Dévényi, S. S. Weygandt, K. J.
aircraft size or air speed, it is not possible to achieve Brundage, J. M. Brown, G. A. Grell, D. Kim, B. E.
100% success with this or any other type of approach.
Schwartz, T. G. Smirnova, and T. L. Smith, and G. S.
This initial algorithm, however, shows great promise for Manikin, 2004: An hourly assimilation-forecast cycle - diagnosing aircraft icing and putting it at the correct the RUC. Mon. Wea. Rev. , 132 , 495-518.
altitude within 0.5 km most of the time. Much additional Bernstein, B.C., F. McDonough, M.K. Politovich and research must be completed before it can serve as a B.G. Brown, 2004: CIP: a physically-based, integrated reliable input for the operational CIP. The delineation of approach to the diagnosis of in-flight aircraft icing, the icing layer vertical boundaries will need to be Part I: Algorithm description. Submitted, Weather and improved using either the RUC or balloon soundings or Forecasting .
ceilometer data to adjust the cloud base height, a Chakrapani, V., D. R. Doelling, A. D. Rapp, and P.
Minnis, 2002: Cloud thickness estimation from GOES- th 13 AMS Conf. on Satellite Oceanography and Meteorology, 20-24 September 2004, Norfolk, VA 8 satellite data over the ARM SGP site. Proc. 12th Minnis, P., W. L. Smith, Jr., L. Nguyen, M. M. Khaiyer, ARM Science Team Mtg , April 8-12, St. Petersburg, D. A. Spangenberg, P. W. Heck, R. Palikonda, B. C.
FL, 14 pp. Available at http://www.arm.gov/docs/ Bernstein, and F. McDonough, 2004b: A real-time documents/technical/conf_0204/chakrapani-v.pdf. satellite-based icing detection system. Proc. 14th Intl.
Dong, X., P. Minnis, G. G. Mace, W. L. Smith, Jr., M. Conf. Clouds and Precipitation , Bologna, Italy, 18-23 Poellot, R. T. Marchand, and A. D. Rapp, 2002: July.
Comparison of stratus cloud properties deduced from Minnis, P., Y. Yi, J. Huang, and J. K. Ayers, 2004: surface, GOES, and aircraft data during the March Relationships between meteorological conditions and 2000 ARM Cloud IOP. J. Atmos. Sci. , 59 , 3256-3284. cloud properties determined from ARM data.
Huang, J., P. Minnis, B. Lin, Y. Yi, M. M. Khaiyer, R. F. Submitted J. Geophys. Res.
Arduini, and G. G. Mace, 2004: Advanced retrievals of Minnis, P., D. F. Young, S. Sun-Mack, Q. Z. Trepte, R.
multilayered cloud properties using multi-sensor and R. Brown, S. Gibson, and P. Heck, 2004a: Diurnal, multi-spectral measurements. Accepted, J. Geophys. seasonal, and interannual variations of cloud Res. , 10.1029/2004JD005101. properties derived for CERES from imager data. Proc.
th Kawamoto, K., P. Minnis, W. L. Smith, Jr., and A. D. 13 AMS Conf. Satellite Oceanogr. and Meteorol ., Rapp, 2002: Detecting multilayer clouds using Norfolk, VA, Sept. 20-24, CD-ROM, P6.10.
th satellite solar and IR channels. Proc. 11 AMS Conf. Smith, W. L., Jr., P. Minnis, B. C. Bernstein, A. D. Rapp, Cloud Physics. , Ogden, UT, June 3-7, CD-ROM, and P. W. Heck, 2002: Supercooled liquid water cloud JP1.18. properties derived from GOES: Comparisons with in tth Min, Q, P. Minnis, and M. M. Khaiyer, 2004: situ aircraft measurements. 10 AMS Conf. Aviation, Comparison of cirrus optical depths from GOES-8 and Range, and Aerospace Meteorol. , Portland, OR, May surface measurements. J. Geophys. Res. , 109 , D15, 13-16, 89-92.
D15207 10.1029/2003JD004390. Smith, W. L., Jr., P. Minnis, B. C. Bernstein, F.
Minnis, P., D. P. Garber, D. F. Young, R. F. Arduini, and McDonough, and M. M. Khaiyer, 2003: Comparison of Y. Takano, 1998: Parameterization of reflectance and supercooled liquid water cloud properties derived effective emittance for satellite remote sensing of from satellite and aircraft measurements. Proc. FAA cloud properties. J. Atmos. Sci. , 55 , 3313-3339. In-Flight Icing/De-icing International Conference , Minnis, P., et al., 1995: Cloud Optical Property Retrieval Chicago, IL, June 16-20, CD_ROM, 2003-01-2156.
(Subsystem 4.3). "Clouds and the Earth's Radiant Smith, W. L., Jr., P. Minnis, and D. F. Young, An icing Energy System (CERES) Algorithm Theoretical Basis product derived from operational satellite data. Proc.
Document, Volume III: Cloud Analyses and Radiance AMS 9th Conf. Aviation, Range, and Aerospace Inversions (Subsystem 4)", NASA RP 1376 Vol. 3 , pp. Meteorol. , Orlando, FL, 11-15 Sept., 256-259, 2000.
135-176. Yi, Y., P. Minnis, J. Huang, and J. K. Ayers, 2004: Cloud Minnis, P., L. Nguyen, D. R. Doelling, D. F. Young, W. detection using measured and modeled statge th F. Miller, and D. P. Kratz, 2002: Rapid calibration of parameters. Proc. 13 AMS Conf. Satellite Oceanogr.
operational and research meteorological satellite and Meteorol ., Norfolk, VA, Sept. 20-24, CD-ROM, imagers, Part I: Evaluation of research satellite visible P6.11.
channels as references. J. Atmos. Oceanic Technol. , Young, D. F., P. Minnis, D. Baumgardner, and H.
19 , 1233-1249. Gerber, 1998: Comparison of in situ and satellite- Minnis, P., et al., 2001: A near-real time method for derived cloud properties during SUCCESS. Geophys.
deriving cloud and radiation properties from satellites Res. Lett. , 25 , 1125-1128.
for weather and climate studies. Proc. AMS 11th Conf. Satellite Meteorology and Oceanography , Madison, WI, Oct. 15-18, 477-480.
th 13 AMS Conf. on Satellite Oceanography and Meteorology, 20-24 September 2004, Norfolk, VA