`to Create Digital Models of Hailstones
`
`IAN M. GIAMMANCO, BENJAMIN R. MAIDEN, HEATHER E. ESTES, TANYA M. BROWN-GIAMMANCO
`
`3D LASER SCANNING AND HAIL. Hail-
`storms account for more than $1 billion in annual
`insured property
`losses, and
`their
`increasing
`trend seen over the past two decades has outpaced
`advances in observation, forecasting, and mitigation
`of hail damage (Changnon et al. 2009; Roeder
`2012; Kunkel et al. 2013). In 2012, the Insurance
`Institute for Business and Home Safety (IBHS)
`began a comprehensive research program with the
`overarching goal to help mitigate property losses
`from severe hail. A component of this initiative
`included determining the properties of hailstones that
`must be accounted for in laboratory material impact
`tests such that the results of these standardized test
`methods would be reasonably predictive of real-world
`performance of building materials. Subsequently, this
`led to a field campaign to measure the physical and
`material properties of hail, and to explore emerging
`technologies to aid in this effort.
`It is well known that hailstones are found in a
`variety of nonhomogeneous shapes and can have
`large protuberances, which makes characterizing
`their true shape difficult using conventional means
`(i.e., caliper or ruler). Obtaining an accurate volume
`through physical measurements is also difficult, even
`when measuring multiple dimensions. In the past,
`record-breaking hailstones were kept in cold storage
`so a cast could be made of the hailstone. The impact
`craters of giant hailstones have also been examined
`and molds made of their shapes, as well (Knight
`
`AFFILIATIONS: GIAMMANCO, MAIDEN, ESTES, AND BROWN-
`GIAMMANCO—Insurance Institute for Business and Home Safety,
`Richburg, South Carolina
`CORRESPONDING AUTHOR: Ian M. Giammanco,
`igiammanco@ibhs.org
`
`DOI:10.1175/BAMS-D-15-00314.1
`©2017 American Meteorological Society
`
`and Knight 2001). While the process is effective in
`capturing the hailstone shape, it is cumbersome and
`time-consuming. A method was needed that provided
`accurate 3D measurement data without substantial
`contamination or melting of the hailstone prior to
`strength testing. The finescale, nonhomogeneous
`nature of hailstones provided the motivation to
`investigate how 3D laser scanners could be applied
`toward hail research.
`The emergence of 3D scanning technology has led
`to new research opportunities across a wide range of
`fields (e.g., medical, mechanical and civil engineering,
`archaeology, etc.) but with little application within
`physical meteorology. In the atmospheric sciences,
`measurement systems such as lidar, particle imagers,
`laser disdrometers, scintillometers, optical rain
`gauges, and visibility sensors come to mind when
`considering laser-based applications. These systems
`are focused on in situ measurement of atmospheric
`particles or rely on backscattered energy from these
`particles. For 3D laser scanners, most atmospheric
`particles are too small and their in situ collection is
`too difficult for a manually operated laser scanning
`system to be of use to map their shape. However,
`hailstones are large enough and their shape is
`complex enough for laser scans and the 3D models
`that are produced to be scientifically beneficial.
`Three-dimensional laser scanners are also efficient for
`collecting sizeable datasets to evaluate the complex
`shapes of hailstones. During field campaigns in 2015
`and 2016, a handheld 3D laser scanner was used
`successfully by IBHS to collect full digital 3D models
`of hailstones. It is believed this is the first time this
`technology has been used in this manner.
`
`EVOLUTION OF 3D SCANNER TECHNO-
`LOGY. The development of scanning technology
`to obtain accurate and precise measurements of
`objects began in the 1960s with advances in computer
`technology. Optical methods proved to be much
`
`AMERICAN METEOROLOGICAL SOCIETY
`
`JULY 2017
`
`| 1341
`
`Align EX1010
`Align v. 3Shape
`IPR2022-00145
`
`
`
`faster, did not require direct physical contact with
`specimens, and were well-suited for complex shapes.
`The foundational research that integrated both passive
`photogrammetric and active laser techniques was
`pioneered by the National Research Council of Canada
`(Mayer 1999). Modern systems apply an active laser
`and passive photogrammetric components to capture
`point-cloud data to produce the digitized 3D model.
`At each data point, the distance and angle from the
`object to the system is recorded in a scanner-relative
`coordinate system. For large objects, several footprints
`of data are needed to stitch together the full 3D shape.
`Processing algorithms assimilate these footprints and
`remove duplicate data. Most current systems connect
`the point-cloud data by applying a nonuniform
`rational basis (NURB) spline fit. The result is faceted
`polygons (typically triangles), which produces the 3D
`surface. With advancements in reducing the size and
`cost of electronic components, small, single-operator,
`handheld units have become less cost-prohibitive
`for a wide range of research projects, including field
`studies and commercial applications.
`
`CAPABILITY SCOPING. The system selected
`to explore 3D scanning of natural hailstones was a
`handheld HandySCAN EXAscan system, manu-
`
`FIG. 1. Conceptual diagram of the laser, single-camera
`configuration, and triangulation coordinate system
`for a typical 3D laser scanning system, where d is the
`distance between the object and the scanner unit.
`Note that multiple cameras are used in hand-held
`systems, and the figure describes the configuration
`of one camera unit relative to the laser.
`
`factured by Creaform Inc. The system is a noncontact
`active scanner that employs a class II eye-safe laser
`to project a beam on a target. An array of cameras
`tracks the projected laser location, as shown in the
`conceptual diagram in Fig. 1. Its relatively small size,
`low weight (~1.5 kg), and simple operation by a single
`person made it ideal for use in a field vehicle, under
`
`FIG. 2. (a and b) Photographs of the scanner in operation. The positioning targets, hailstone mount, and
`turntable are annotated; (c) the 3D faceted surface created by processing the collected point-cloud data; (d)
`the full 3D model of the first hailstone captured with this system; (e) comparison with a sphere of the same
`maximum diameter; and (f) the 3D-printed cavity mold of this hailstone.
`
`1342
`
`|
`
`JULY 2017
`
`
`
`nonoptimal conditions (Fig. 2a,b). To operate the
`unit, information must be simultaneously collected on
`the unit’s position while it is scanning the specimen.
`Additionally, the unit must be calibrated prior to
`operation periods. The scanner is calibrated by using
`a plate with a grid of reflective targets, supplied by
`the manufacturer (reflective targets are identified in
`Fig. 2a,b). The precise dimensions and target locations
`of the plate are stored in the operating software, which
`is able to identify and adjust for any small bias errors.
`Small errors may result from temperature changes
`and the expansion and contraction of hardware
`components such that calibration is recommended
`prior to scanning sessions. The reflective positioning
`targets are also used to define the coordinate system
`with respect to the specimen being scanned. Targets
`are scanned separately (only one time) prior to
`data collection. The information is stored by the
`operating software and applied when scanning of
`the specimen is underway. The targets are adhered
`either to the specimen itself or to a mounting system
`such that the unit always has several positioning
`targets in its field of view (HandySCAN EXAscan
`requires at least three). If the minimum number of
`targets is not detected during data collection, the
`software will cease logging until they are identified to
`automatically avoid data gaps due to user error. The
`system has a trigger that toggles the laser projection,
`camera operation, and data collection.
`The unit has a maximum configurable resolution
`of 0.008 cm, an accuracy of ±0.004 cm, and a
`maximum sampling rate of 25 kHz. It is tethered to
`a laptop computer running Creaform’s VXelements
`software package to operate the scanner, view
`ongoing scans in real time, and store the data. The
`NURB spline-based polygon-mesh approach is used
`by VXelements to capture, process, and display the
`3D data. The processed dataset can then be quality
`controlled to synthetically fill in missing data, remove
`other objects that may have been in the field of view,
`and filter spurious returns. Once the data have been
`processed, additional analyses can be performed on
`the digital model to extract more information on
`the characteristics of the hailstone. Data can also be
`exported in a .STL file format for use in standard
`CAD packages or other computational analysis tools.
`
`LABORATORY ICE TESTING. The EXAscan
`system’s ability to detect and map ice surfaces was
`tested using ice spheres made with pure distilled
`water (very clear ice) and water with diffused carbon
`
`dioxide gas (bubble-filled, opaque ice). The ice
`spheres were then chipped or deformed to introduce
`small shape changes to evaluate the scanner’s
`ability to detect these deformations. It was quickly
`discovered during initial testing that ice surfaces
`are difficult mediums to effectively scan. Clear ice
`surfaces and ice surfaces coated with a large amount
`of liquid water scattered the projected laser such
`that it was not well defined on the object surface.
`Subsequently, the photogrammetric camera tracking
`functionality could not resolve the true location of
`the projected laser. This resulted in large gaps in the
`digitized model. Performance was improved when
`opaque, bubble-filled ice was tested, but this required
`long scanning durations and revisiting scanned
`areas to capture a complete model. To reduce the
`amount of scatter, a light dusting of a fine powder
`(i.e., athlete’s foot spray) was used, enabling the
`system to adequately track the projected beam and
`map the ice surfaces. At times, compressed air was
`also used to help remove any liquid water on the
`surface of the hailstone. Although this introduces
`a foreign substance onto the hailstone similar to
`an immersion test, compressive strength testing
`yielded no detectable influence between coated and
`uncoated laboratory ice spheres. The method is still
`more practical than immersion testing in a field
`setting, especially when considering substances used
`in past research (i.e., liquid mercury). During this
`initial testing, it was also determined that full scans
`can be completed in less than 1 min at low sampling
`resolutions, while higher resolution scans can take
`2–3 min to complete. The length of time needed for
`a complete scan was determined to be suitable for a
`pilot field application to help mitigate the melting of
`stones while they were being scanned.
`
`SCANNING HAILSTONES IN THE FIELD.
`The scanner system was pilot tested in the field for the
`first time in 2015 to determine if it would be effective
`for use during the 2016 field measurement program.
`Calibration was performed after the target storm
`was selected but prior to data collection. This helped
`mitigate any measurement errors from temperature
`changes and possible expansion and contraction of
`hardware components during transit. Hailstones were
`collected from a target thunderstorm following its
`passage across an identified roadway. Liquid water
`present on the surface of the hailstone was quickly
`wiped clean or blown off using compressed air prior
`to the powder application.
`
`AMERICAN METEOROLOGICAL SOCIETY
`
`JULY 2017
`
`| 1343
`
`
`
`To allow the operator to quickly scan the full
`volume, a custom mount was designed and 3D-printed
`to support the stone. The acrylonitrile butadiene
`styrene (ABS) plastic material helped reduce melting
`resulting from the direct contact between the
`hailstone and the supports. The mount used three
`points of contact to support the stone with as little
`interference as possible. Reflective positioning targets
`were permanently fixed to the mount to calibrate
`the scanner position relative to the mount and
`allow for the mount to be placed on a turntable. The
`reflective targets allow the unit to “know” its relative
`position in three-dimensional space. The turntable
`allowed the hailstone to be rotated so that the sides
`of the stones could be scanned without the operator
`needing to move frequently within the vehicle. The
`mount also allowed enough space between supports
`so the bottom of a hailstone could be captured by the
`operator simply turning the unit to allow the laser to
`pass across the underside of the hailstone. The support
`mount is detected during scanning, but is removed
`in data processing, leaving just the 3D model of the
`hailstone. An example of a hailstone being scanned
`in the field and the resulting 3D model can be seen at
`https://vimeo.com/167924554. Before a hailstone was
`scanned, specimens were photographed, measured
`with a caliper, and weighed.
`
`2015 PILOT FIELD TESTING. The system was
`first deployed during a period of active severe weather
`in the Central Plains on 15–18 September 2015. The
`field team intercepted a supercell thunderstorm near
`Atchison, Kansas, on 17 September 2015, which
`produced a relatively high bulk concentration of
`small hail (<2 cm). Attempts to collect a full scan
`of small hailstones (<1 cm) were unsuccessful
`due to the original design of the prototype mount
`(corrected in a later version). The hailstones were
`too small to effectively support as they began to melt.
`Fortunately, a larger hailstone (2.5 cm in diameter)
`was gathered, and a successful scan was made. The
`data were processed to remove scanner interference,
`synthetically fill any small data gaps, and produce
`the full 3D model (Fig. 2c,d). It is believed that this
`was the first successful 3D laser scan of a hailstone.
`The scan of this particular hailstone was completed
`in approximately 3 min at a resolution of 0.008 cm
`and used a maximum sampling rate of 25 kHz. The
`fully scanned hailstone had a mass of 2.50 g, and a
`maximum diameter of 2.504 cm. The diameter was
`defined as the longest straight line between two points,
`
`which passed through the center of the hailstone
`model. The volume, determined from the model, was
`3.654 cm3, which was 54% less than a sphere of the
`same diameter (Fig. 2e). The volume coupled with the
`measured mass yielded a bulk density of 0.68 g cm−3.
`The digitized 3D model was used to 3D-print a cavity
`mold based upon the highly detailed hailstone shape
`(Fig. 2f), and demonstrated the linkage between 3D
`scanning and printing technology. The success of
`integrating the digital hail model into a CAD design
`application and 3D-printing a model highlighted
`the ability to duplicate natural hailstone shapes and
`their intricate details in a laboratory setting. This,
`coupled with exploration of diffused gas ice mixtures,
`could lead to the re-creation of laboratory hailstones
`that match the physical and material properties of
`hailstones observed in the field.
`
`2016 FIELD MEASUREMENT PROGRAM
`AND ANALYSIS. The 2016 field measurement
`program focused on obtaining 3D models of hailstones
`and performing corresponding compressive strength
`tests. The efforts produced 42 digital hailstone models
`collected primarily from supercell thunderstorms
`in the Southern Great Plains of the United States in
`May and June of 2016. A subset of scanned hailstones,
`showing the variety of shapes that were captured, is
`shown in Fig. 3. The high-resolution models allowed
`for an accurate volume estimate to be obtained for
`each hailstone. It is acknowledged that some melting
`may have occurred prior to collection, and/or liquid
`water contained within small cavities in the hailstone
`may have drained, resulting in a small bias. It is also
`possible that protuberances may have been rounded
`off because of melting or impact with the ground.
`When compared with hailstone densities estimated
`using physical measurements and shape assumptions,
`the errors are expected to be reduced.
`Throughout historical literature, summarized by
`Knight and Knight (2001), hailstones are commonly
`referred to as “hard” or “soft” with no quantification of
`their strength. It is frequently assumed that hailstone
`strength and their damage potential scales with bulk
`density (Knight et al. 2008). The true relationship
`between density and strength is unknown at this time.
`The use of 3D-scanned hailstones combined with
`recent advances in the ability to test hailstones for their
`compressive strength can help clarify the relationship
`and determine if laboratory impact tests must replicate
`it in order to accurately produce a true correlation with
`real-world performance of building materials.
`
`1344
`
`|
`
`JULY 2017
`
`
`
`observations also showed that hailstone densities
`trend closer to pure ice (0.9 g cm−3) as they get larger.
`Three hailstones exhibited a density greater than
`0.9 g cm−3 and were characterized by nearly all clear
`ice with no visible layering structure. The hailstones
`also had notable protuberances. The high density of
`these stones raises the question of whether “super”
`density ice occurs in hailstones or if this was the result
`of a measurement error. It is possible that some mass
`loss between measurement and scanning occurred
`such that the density estimate contained an error;
`however, the maximum diameter measured using
`a caliper was within 0.04 cm for all three hailstones
`when compared to the scanner-based diameter. The
`scale used has a precision of 0.01 g, but any shaking or
`movement of the scale could have introduced some
`source of measurement error. The use of this system
`in the field will help improve the understanding of
`hailstone bulk density distributions and determine if
`high-density and/or low-density hailstones are more
`prevalent than historical literature would suggest.
`It was clear that the measured hailstones departed
`from spherical shapes with increasing diameter,
`which is in agreement with recent field observations
`(Heymsfield et al. 2014) (Fig. 4c).
`low-density
`Throughout historical
`literature,
`hailstones were often associated with being soft and of
`low strength. There has been little quantitative analysis
`to substantiate this expectation or to investigate a
`potentially different relationship. The datasets collected
`through 3D scanning and compressive strength testing
`allowed for a preliminary examination of how the two
`variables may be related. The relationship between the
`measured peak forces showed a general linear trend,
`with a larger force required for higher densities (not
`shown). However, the peak force must be scaled by
`the area of the plane in which the force was applied to
`produce an appropriate measure of strength. As shown
`in Fig. 4d, the slight linear trend was toward weaker
`hailstones with higher bulk densities. It is noted that
`the sample size shown here is only 42 hailstones, and
`larger datasets are needed. The ability to evaluate
`these properties is a notable advance that will foster
`new
`research
`toward understanding hailstone
`characteristics and determining their properties that
`affect damage potential.
`
`RESEARCH APPLICATIONS OF 3D HAIL-
`STONE MODELS. The first effort to 3D-scan
`hailstones was successful in proving the system
`could be operated efficiently in the field, collect a
`
`JULY 2017
`
`| 1345
`
`FIG. 3. Collection of several hailstone models showing
`the variety of shapes captured during the 2016 field
`program. The date and general location are provided
`for each hailstone.
`
`The 3D-scanned hailstones were subjected to
`compressive strength testing, which applies an
`increasing compressive force with a strain rate on the
`order of 10−1 s−1 to the hailstone until it fractures. The
`peak force at the time of fracture is captured and then
`scaled by the cross-sectional area (i.e., plane) in which
`the force was exerted to produce an estimate of uniaxial
`compressive stress. The compressive stress was used
`as a proxy to represent the hardness property of the
`hailstone (Giammanco et al. 2015). These stones were
`also examined with respect to the diameter-to-mass
`relationship, bulk density, their volume normalized
`by that of a sphere with the same maximum
`diameter, and their compressive strength (Fig. 4). The
`
`AMERICAN METEOROLOGICAL SOCIETY
`
`
`
`FI G. 4. Observations of
`3D -scanned hailstones
`derived from the digital
`models and physical mea-
`surements for (a) mass as
`a function of maximum
`diameter and the curve
`for an ice sphere with a
`density of 0.9 g cm−3 (solid
`orange); (b) bulk density
`as a function of maximum
`diameter; (c) ratio of the
`hailstone volume to the
`volume of a sphere of the
`same ma ximum diam -
`eter; and (d) compressive
`stress (used as a proxy for
`strength/hardness) as a
`function of bulk density.
`
`quality number of 3D models, and allow for further
`strength testing. The ability to collect these digital
`representations of natural hailstones will open the
`door to new investigations of their shapes and material
`properties. By eliminating the need for contact
`or immersion methods of determining hailstone
`densities, quantifying the relationship between bulk
`density and hailstone strength was possible. This
`capability will help improve ice-based laboratory
`impact test methods to ensure they are representing
`the necessary properties of natural hail. Cavity molds
`can also be used to determine if hailstone shape
`influences the type of damage for different materials.
`The aerodynamics of hailstones is an area that could
`also benefit from 3D hail model datasets. Digitized
`hail shapes could also be leveraged to explore the
`aerodynamic drag characteristics of hail through
`experimental and computational methods, which
`are vital to ensuring that proper kinetic energies are
`used in material impact tests. Current test standards
`use impact kinetic energies determined through
`assumptions that drag coefficients for spherical
`shapes can be used for natural hailstones (Heymsfield
`et al. 2014). Assumptions regarding hailstone drag,
`terminal velocities, and kinetic energies are also
`made within hydrometeor parameterization schemes
`for numerical weather prediction models (Morrison
`et al. 2015). The aerodynamic applications described
`here could be leveraged to improve the hail-related
`portions of these schemes. Experiments could also
`
`shed light on the tumbling of hailstones, which can
`complicate radar detection, especially for dual-
`polarimetric radars (Straka et al. 2000). An improved
`understanding of this effect may provide the ability
`to extract more detailed hydrometeor information
`(i.e., mean shape, concentration) from the dual-
`polarimetric moments.
`The use of 3D laser scanning systems continues
`to grow rapidly across a wide range of fields. Until
`now, their use in the atmospheric sciences has been
`limited. The pilot investigation presented here has
`shown how the technology can be used effectively to
`understand the characteristics of hail beyond what is
`considered in historical studies. These data will foster
`new research into the aerodynamics of hailstone
`shapes, the relationship between strength and density,
`radar hail detection, and hail damage severity. Each of
`these applications rely on the accurate representation
`of hail, and can be used to improve material impact-
`testing practices,
`improve hailstorm postevent
`characterizations, and develop new risk assessment
`methods through numerical modeling efforts. Each
`will ultimately aid in mitigating the large amount of
`property loss that occurs each year from severe hail.
`
`ACKNOWLEDGMENTS. Funding for the field pro-
`gram and analysis efforts are provided by the Insurance
`Institute for Business and Home Safety through the annual
`operating budget. We would like to acknowledge Kevin
`
`1346
`
`|
`
`JULY 2017
`
`
`
`Outz from Matrix CAD Design, Inc. for his invaluable
`assistance in proof-of-concept testing of 3D scanners for
`this application.
`
`FOR FURTHER READING
`
`American Society for Testing of Materials, 2010: Stan-
`dard test method for hail impact resistance of aero-
`space transparent enclosures. ASTM, 8 pp. [Available
`at www.astm.org/Standards/F320.htm.]
`Beraldin, J.-A., F. Blais, L. Cournoyer, M. Rioux, S. H.
`El-Hakim, R. Rodella, F. Bernier, and N. Harrison,
`1999: 3D Digital Imaging and Modeling, Proc. Second
`International Conf. on 3D Digital Imaging and Model-
`ling, Ottawa, Canada, 34–43.
`Besl, P. J., 1988: Range Imaging Sensors. Mach. Vis. Appl.,
`1, 127–152, doi:10.1007/BF01212277.
`Changnon, S. A., D. Changnon, and S. D. Hilberg, 2009:
`Hailstorms across the Nation: An Atlas about Hail
`and Its Damages. Illinois State Water Survey, 95 pp.
`FM Approvals, 2005: Specification test standard for im-
`pact resistance testing of rigid roofing materials by im-
`pacting with freezer ice balls (FM 4473). FM Approv-
`als, 8 pp. [Available online at www.fmapprovals.com
`/approval-standards?filter=0.]
`Giammanco, I. M., T. M. Brown, R. G. Grant, D. L.
`Dewey, J. D. Hodel, and R. A. Stumpf, 2015: Evalu-
`ating the hardness characteristics of hail through
`compressive strength measurements. J. Atmos. Oce-
`anic Technol., 32, 2100–2112, doi:10.1175/JTECH
`-D-15-0081.1.
`Heymsfield, A. J., I. M. Giammanco, and R. L. Wright,
`2014: Terminal velocities and kinetic energies of
`natural hailstones. Geophys. Res. Lett., 41, 8666–8672,
`doi:10.1002/2014GL062324.
`
`Kim, H., and J. N. Keune, 2007: Compressive strength of
`ice at impact strain rates. J. Mater. Sci., 42, 2802–2806,
`doi:10.1007/s10853-006-1376-x.
`Knight, C. A., and N. C. Knight, 2001: Hailstorms.
`Severe Convective Storms. C. A. Doswell III, Ed.,
`AMS, 223–254.
`—, P. T. Schlatter, and T. W. Schlatter, 2008: An un-
`usual hailstorm on 24 June 2006 in Boulder, Colo-
`rado. Part II: Low density growth of hail. Mon. Wea.
`Rev., 136, 2833–2848, doi:10.1175/2008MWR2338.1.
`Knight, N. C., 1986: Hailstone shape factor and its
`relationship to radar interpretation of hail. J. Cli-
`mate Appl. Meteor., 25, 1956–1958, doi:10.1175/1520
`-0450(1986)025<1956:HSFAIR>2.0.CO;2.
`Kunkel, K., and Coauthors, 2013: Monitoring and under-
`standing trends in extreme storms. Bull. Amer. Meteor.
`Soc., 94, 499–514, doi:10.1175/BAMS-D-11-00262.1.
`Laurie, J. A. P., 1960: Hail and its effects on buildings.
`Council for Scientific and Industrial Research Rep.
`176, 12 pp.
`Mayer, R., 1999: Scientific Canadian: Invention and In-
`novation From Canada’s National Research Council.
`Raincoast Books, 192 pp.
`Morrison, H., A. Morales, and C. Villanueva-Birriel,
`2015: Concurrent sensitivities of an idealized deep
`convective storm to parameterization of microphys-
`ics, horizontal grid resolution, and environmental
`static stability. Mon. Wea. Rev., 143, 2082–2104,
`doi:10.1175/MWR-D-14-00271.1.
`Roeder, P., Ed., 2012: Severe weather in North America:
`Perils, risks, and insurance. Knowledge Series: Natu-
`ral Hazards, Munich RE Rep., 274 pp.
`Straka, J. M., D. S. Zrnic, and A. V. Ryzhkov, 2000: Bulk
`hydrometeor classification and quantification us-
`ing polarimetric radar data: Synthesis of relations.
`J. Appl. Meteor., 39, 1341–1372, doi:10.1175/1520
`-0450(2000)039<1341:BHCAQU>2.0.CO;2.
`
`AMERICAN METEOROLOGICAL SOCIETY
`
`JULY 2017
`
`| 1347
`
`
`
`ORDER ONLINE: bookstore.ametsoc.org or see the order form at the back of this issue
`
`