2.1. Dynamics and Radar Detection of Meteorite Falls

Weather radars are not designed to detect meteorite falls but fortuitously serve quite well in this purpose. Meteorite falls occur when a meteoroid or asteroid impacts a planetary atmosphere in such a manner that material survives to reach the ground. Meteoroids are defined by the International Astronomical Union (IAU) as small objects between 30 μm and 1 m in diameter and asteroids are larger objects [15]. When one of these objects strikes a planetary atmosphere, a meteor, or short-lived trail of ionized and luminous gas, is created. Any meteor brighter than absolute visual magnitude −4 is called a fireball or bolide, and a meteor brighter than absolute visual magnitude −17 is termed a superbolide (ibid). The overwhelming majority of meteors do not produce a meteorite fall as the meteoroid is disaggregated by shock pressure and evaporated to produce dust and melt spherules. There is no reliable correlation between meteor brightness and whether a fall actually occurs, as this is dependent upon multiple variables to include infall velocity, infall angle, meteoroid mechanical toughness. In a typical meteorite fall, a bolide occurs that is sufficiently bright that a significant percentage of the local population witnesses it. The bolide itself typically lasts between ∼3 and 15 s depending on velocity and infall angle, during which time well in excess of 90% of the original mass of the meteoroid is reduced to dust, gas, and small melt spherules. A meteor first becomes luminous around 90 km altitude on Earth and, for an event of sufficient initial mass to generate a meteorite fall, remains luminous down to an altitude of approximately 20–30 km above the mean sea level (AMSL) This point is defined as the fireball terminus. At this point, the remaining mass decelerates from atmospheric drag to below 3 km/s and no longer transfers sufficient energy to surrounding gas to ionize it, marking an end to the optically luminous meteor. The portion of flight between the terminus and where they land on the ground is known as “dark flight.” The motion of any surviving meteorites is aerodynamically limited through dark flight and they decelerate quickly to their terminal velocity. Terminal velocity diminishes as the falling bodies experience higher atmospheric pressure at lower altitude, and so the falling meteorites decelerate all the way to the ground. Terminal velocity is also dependent upon the mass of the meteorites. Kilogram-mass meteorites with a density of 3.3 g/cm³ will reach the ground from a 25 km AMSL terminus and 45 degree infall angle in 2.25 min, whereas 1 g meteorites require 7.5 min to cover the same distance. Therefore, meteorites are aerodynamically size sorted on their way to the ground. Lateral forces due to wind may be significant as jet stream winds typically occur at altitudes encountered during dark flight. Larger meteorites in the kg mass range are relatively resistant to aerodynamic forces due to their short flight time and higher momentum, but meteorites in the 1–100 g mass range may move sufficiently that the distribution of meteorites on the ground (or “strewn field”) takes on an arcuate shape defined by dominant wind direction.

Weather radars are designed to detect weather, and so the emitting dish is usually pointed at an angle close to the ground. NEXRAD radars operate exclusively below 20° elevation and most dish rotations (or “sweeps”) occur below 10°, with most data collected below about 10 km altitude. The optically bright meteor occurs at higher altitudes ranging from about 90 km AMSL to a typical value of about 20–25 km AMSL. For this reason, weather radars rarely detect meteorites while they are optically bright, and weather radar detection is almost exclusively a dark flight detection method. For the same reason, weather radars also do not detect ionized air around the optically bright meteor; rather, it detects falling stones by direct reflection from the rocks themselves. Currently, weather radar is the only technique that can detect and analyze meteorites while in dark flight. Because small meteorites fall more slowly than large ones, smaller meteorites (1–100 g) are more likely to be detected by the relatively slow-turning radars than the fast-moving kg-mass meteorites. Therefore, weather radar preferentially detects meteorite falls larger than those composed of only a few stones and those composed of a large proportion of small meteorites.

Historically, two major factors influence whether meteorites can be recovered from a meteorite fall, namely, identifying whether a meteor actually produced a meteorite fall and whether the fall site can be accurately determined. Weather radar directly resolves both of these issues. The radars only detect falling meteorites, thereby indicating when a fall actually occurred and not just when a bright fireball did. Weather radars feature spatial resolution on a scale of tens to hundreds of meters depending on distance from the radar itself. This is a small value compared with the areal extent of a typical meteorite fall, meaning that radars not only detect meteorite falls but can also image them sufficiently to present a complete picture of their location and extent. An additional step is needed to calculate the actual landing site of meteorites from their detected location at a given altitude, and this is done with dedicated “dark flight” models such as Jörmungandr [16] used by the author.

2.2. Direct Detection Techniques

The primary discriminating factor in identifying a meteorite fall in weather radar data is the observation of signatures at the time and place identified by other assets. These include eyewitness accounts from sources such as the American Meteor Society [17], U.S. Department of Defense sensors for treaty monitoring [18], or the NOAA Geostationary Lightning Mapper instrument [19, 20] on the latest family of environmental monitoring satellites in geostationary orbit. Weather radars detect meteorite falls through a combination of factors based mostly on the unique dynamical properties of meteorite falls versus the other items detectable by radar. NEXRAD radars generate six different data types, two of which are direct measurements (reflectivity and Doppler velocity) and four are derived datasets calculated from the radar returns. At this time, no unique signatures of meteorite falls have been reliably identified in the four calculated product sets. Meteorites are directly detectable in reflectivity measurements (Figure 1). Reflectivity is effectively a measurement of total reflecting surface area (radar cross section or RCS) and meteorite falls do not generate a signature that is fundamentally unique from other reflectors. Reflectivity values of falling meteorites do not tend to be strong and can be masked by concurrent precipitation returns or lost in ground clutter at low elevation angles. The various NEXRAD scan strategies routinely collect data at higher elevations where meteorite falls can more reliably be detected and classified. This enables direct identification of meteorite signatures in radar imagery over weather, birds, aircraft, or other signatures. Typical weather radar detections of meteorites often exceed 10 km AMSL, with rare detections exceeding 20 km AMSL. The latter can occur when a radar happens to be at an optimal distance and operating in a VCP that features high-elevation sweeps. Radar signatures at 20 km AMSL or greater include a wide range of unsorted meteorite masses. Therefore, estimating mass from these signatures should be treated with caution as the altitude is high enough that the meteorites may not have size segregated yet.

In Doppler velocity imagery, meteorite falls can produce a unique signature of vigorous turbulence over the small area of the fall as larger (100 s of grams and greater) meteorites pass through the radar’s interaction volume at supersonic speeds (Figure 2). Reflectivity signatures can be detected over a time period of approximately 90s after the bolide to 10 minutes afterward. The velocity signature is only prominent for 2-3 min after the bolide. This is a diagnostic feature of meteorite falls that can be used to identify them even in “noisy” radar images.

2.3. Strewn Field Modeling

When meteorite falls are identified in radar data, author MF distributes information to the general public as a NASA public service. This includes building a strewn field that describes an area where meteorites landed on the ground. This can be done to show an estimate of meteorite masses or to show the expected concentration of meteorites on the ground regardless of size. The authors use the Jörmungandr dark flight model [16] which generates a mass-estimated strewn field. Jörmungandr takes advantage of NOAA radiosonde (“weather balloon”) data that is reported as wind direction and speed with respect to altitude. The user locates the radiosonde data from the site nearest in both time and location to the bolide, retrieves it from one of several public sources, and enters it into Jörmungandr. These balloons are released from sites around the world synchronized at 0000 UTC and 1200 UTC. The model performs a triplicate extrapolation of the data to produce a higher resolution dataset. Jörmungandr then calculates the force vectors acting upon the falling meteorite for each radiosonde altitude and calculates the meteorite’s new location at the next-lowest altitude value. This is repeated for a given range of altitudes. The calculation starts by examining flight from the bolide terminus down to the radar signature to generate an estimate of meteorite mass and the meteorite velocity at the height of the radar signature. It also back calculates an estimated site where it first entered “dark flight” or nonluminous flight, by finding the location and altitude where the meteorite’s velocity last equaled 3 km/s. This velocity value is used as the standard value indicating where the moving meteorite is no longer fast enough to be optically luminous. A second step then examines flight from the radar signature to the ground to produce a landing site in Universal Transverse Mercator projection using the MGS84 ellipsoid. This is repeated for all meteorite signatures, producing a range of detected masses and geolocated landing sites. The next step in producing a strewn field uses Google Earth, in large part because the NOAA Weather and Climate Toolkit data used to explore radar data includes the ability to directly export radar images to Google Earth. Once the landing sites for all observed meteorites are computed, an image is created of all the radar signatures at the ground level in locations defined by their calculated landing sites. As a final step, a single point is identified as a common location where all meteorites entered dark flight, and this site is used to calculate the simplified overall strewn field. This is done by finding the centroid of all back-calculated luminous flight terminus sites for all the observed meteorites and treating it as the site where all the falling meteorites began their flight. From this site, decades of masses are plugged into the model (i.e., 1 g, 10 g, 100, 1 kg, and 10 kg) and their landing sites calculated. The simplified color-scale strewn field is built by drawing roughly equal-sized polygons around each mass’ landing site. The purpose of this strewn field is to provide a quick turnaround, simplified indication of where meteorites of a range of masses can be located and is intentionally simplified. In reality, fragmentation usually occurs at multiple points along a bolide’s path each producing meteorites of different masses with overlapping landing ellipses. On the ground, meteorite masses do not follow a perfect gradient but show some mixing although there is a general trend from a “small end” where low-mass meteorites landed to a “big end” where a smaller number of larger meteorites can be found. In practice, the simplified strewn field works well because it can be produced in a matter of hours, is easily understood by the layperson, and has proven to accurately depict the area on the ground where meteorites can be found. Recently, improvements have been made on meteorite recovery through software refinement [23] and by demonstrating the use of drones and machine learning for rapidly identifying meteorites on the ground [24–27]. This allows rapid reconnaissance of large strewn fields, lessening the need for large search parties and allowing faster recovery before rain or other weather can affect the waiting meteorites.

2.4. Mass Calculation and Mass Frequency Distribution

Weather radars are sensitive to the cross section of objects but have diminished capacity to distinguish between multiple small meteorites and a single large rock. Therefore, it is not possible to detect the mass of falling meteorites directly. Future research into the data from NEXRAD following the upgrade to dual polarization beam structure may provide an initial estimate of size or distribution of objects within the sample volume. In addition, further study of the dual polarization datasets may improve upon this distinction with a more complicated integration of the calculated radar products which are designed to be sensitive to object shape. For now, however, the mass can be estimated by calculating the size of a meteorite that would travel between two known points in the meteorite’s flight path. If the altitude and seconds-accurate time of the bolide terminus is known or can be estimated, then that serves as the first point. The second point can be obtained from radar data which reports a millisecond-accurate time and meters-accurate altitude. This is possible because falling meteorites are aerodynamically size sorted in descent, minimizing the range of masses present in a given radar signature. Mathematical details on how Jörmungandr operates have been described elsewhere [16], but in essence, it breaks down the meteorite motion and effective wind into vector components for each altitude measurement in the radiosonde data and calculates forces upon the meteorite and then new vectors which are applied to the next altitude measurement. This continues to the bottom of the altitude range defined by the user. Mass calculation is performed by iterating the mass until the elapsed time of transit between the two points matches the measured value. For many bolides, the exact terminus altitude is not known and a default value of 25 km is assigned. Errors in this value have a direct consequence on calculated meteorite masses, so accurate terminus values are preferred. Terminus altitude values are most commonly provided via triangulation from multiple camera views of the bolide but may also be available from Department of Defense sensors (e.g., [24]) but with uncertain resolution. Terminus altitude is especially difficult to obtain for daytime bolides since fewer cameras are available or are of low resolution due to diminished definition of the bolide against a bright sky. Many all-sky meteor cameras also do not operate during the day. There is a need for a rapid, perhaps automated, system for determination of terminus altitude for bolides in both day and night.

The range of masses detectable with NEXRAD radar is useful information, both for appreciating the range of meteorites detected in any given fall and for comparing S-band performance to other radar types used worldwide. A compilation of mass measurements seen over fifteen meteorite falls reveals the typical meteorite detection mass range for the NEXRAD system (Figure 3). The histogram of detected masses shows that meteorite detection is optimized in the 1–100 g for the S-band NEXRAD radars. This arises from several factors. The size range of meteorites is determined by fragmentation and strongly favors production of small masses. Also, large masses (≥ kg mass) fall rapidly through the radars’ interaction volume, diminishing the likelihood of detection. On the small mass end of the histogram, sensitivity of the radars drops off linearly with decreasing size, which partially explains the decrease in detection frequency. It is probably more significant that very small particles (≤ 0.1 g) fall very slowly and are readily dispersed by both winds and convection from the hot meteor plasma. The combination of these factors produces the distribution in Figure 3 and different radar wavelengths will affect it primarily in differences in reflectivity versus the meteorite size.

Once mass estimates are obtained for signatures in all the various radar sweeps for a given fall, they can be compiled into a particle size distribution (PSD). PSDs are very useful for exploring the physical properties of their progenitor material, and in this case, the parent meteoroid for these meteorite falls. Since we have mass values (m) in hand, the second set of values needed to build a PSD is the number (n) of each meteorite mass present. This can be obtained in principle from reflectivity data. Reflectivity of a single meteorite of known mass is estimated and then the total observed reflectivity is divided by the reflectivity of a single rock to obtain the number of meteorites present. Making this calculation for all the m values builds an m vs. n curve, defining a PSD. The total mass of the fall can be obtained by integrating under this curve. In principle, this is straightforward, but there are currently unsolved difficulties. In weather radar terms, these calculations are performed for rain, hail, and other hydrometeors to obtain the rain rate and total rain/snow/hail fall amounts. Attempting this with meteorite signatures produces unreasonably large values, probably because the reflectivity of meteorites is currently unknown and a value for terrestrial basalt is substituted. In order to produce quantitative values for the number of meteorites present, direct measurements are needed on reflection properties of fusion-crusted large meteorites under 10 cm wavelength (S band) irradiation. While the refractive index could be estimated using terrestrial rocks, there are two complications that prevent that relatively easy solution. For one, meteorites contain Fe–Ni metal and sulfides that are not found in terrestrial analogs and which should have a significant effect on radar reflectance properties. More importantly, meteorites are coated with a thin melted layer called fusion crust which should dominate the reflection and the reflectance properties of fusion crust. Until such a time that these measurements are complete, we are restricted to producing relative PSDs where n has been calculated uniformly across all radar measurements (Figure 4) that are usable with the caveat that they can be improved on in the future. Figure 4 shows PSDs for fifteen meteorite falls presented in log–log format with linear curve fits. The higher up a curve sits on this graph, the higher the total fall mass. The slope of the curve is important in that it describes the proportion of large meteorites present versus small ones (or vice versa as per preference) and serves as a proxy for whole-body meteoroid mechanical strength. Figure 5 shows the curve fit slope values in Figure 4 in the histogram form. The upper histogram is for all fifteen meteorite falls and shows a multimodal distribution where the main body of falls have a slope of around −1.5. There is a subset with slopes ranging down to −2.3, indicating significantly more small meteorites present than larger ones. These bodies can be interpreted as “crumbly” relative to most falls, composed of material that fragmented extensively such that survival of small meteorites in the bolide is preferred. There is more than one way this can happen, such as a heavily fractured parent meteoroid due to impact shock history. The meteoroid may also come from impact-gardened surface regolith, which was present on the near surface of an asteroid in the past and is composed mostly of smaller material. When we look at only those falls for which we know the meteorite type (Figure 5(b)), there is no correlation between the meteorite type and “crumbly” behavior. This indicates that “crumbly” bodies are not due to the properties of a particular type of meteorite but are more likely due to impact processing or origin as regolith. Again, until the fusion crust reflectivity issue is resolved, these values are qualitative only but are useful for comparative purposes.

2.5. Spatial Distribution of Falls

Meteorite falls within the NEXRAD coverage area (Figure 6) reveal some interesting details. Perhaps, most striking is the lack of detected falls in the Pacific Northwest and upper Midwest and another smaller cluster of states in Appalachia (Tennessee, Kentucky, North Carolina, and West Virginia). For the northwestern states in particular, this may be a function of population density. While the radars themselves should perform equally as well in these areas as in the rest of the NEXRAD coverage area, the number of bolide eyewitnesses will be fewer in these regions. The possibility exists that meteorite falls await discovery in the NEXRAD archives for these regions. A systematic investigation of eyewitness reports with a low threshold for the number of reports may reveal new falls, as may a closer look at GLM bolide reports in these regions. Future data mining techniques such as artificial intelligence (AI)-driven event recognition may be fruitful in the sparse recovery regions in particular. Aside from discussion of these regions themselves, the fact that areas of the map appear to be “missing” meteorite falls indicates that the true rate of falls is somewhat higher than the current dataset indicates.

2.6. Weather Radars Around the World

The World Meteorological Organization, operated by the United Nations, states that more than 1100 weather radars are in service with national weather bureaus in more than 90 countries [29]. Additional radars are in service with military agencies, TV stations, and other entities but access to their data is difficult for the public to obtain. If we assume the areal coverage is similar, the world’s weather radars with public-accessible data amounts to roughly six times as many meteorites observed in radar data than NEXRAD alone (∼1100 minus NEXRAD’s 160 radars versus 160 NEXRAD radars). Since the first meteorite fall detected by NEXRAD in 1997 (the Worden, MI fall), NEXRAD has detected 32 recovered meteorite falls and 20 more probable falls where no meteorites were recovered. This amounts to an annual recovery rate of 1.2 falls/year for recovered falls and 1.9 falls/year including unrecovered probable falls. If all the world’s radars could be utilized for meteorite recovery, we could expect seven recovered meteorite falls per year observed in weather radar data and eleven observed falls per year if unrecovered falls are included. This simple calculation uses current recovery rates with the NEXRAD network although the previous chapter illustrates that there are likely additional falls to account for by optimizing the weather radar recovery protocol.

Significant challenges to implement global weather radars exist and must be contended with. The radars are built by at least 32 different manufacturers, each with its own software and many with proprietary data formats (WMO Database). The radars operate over a mixture of X (2.75–3.5 cm), C (3.75–7.5 cm wavelength), and S (7.5–15 cm) bands, and the sensitivity of these radars to meteorites in general, or meteorites of various sizes is currently unexplored. The radar data are the property of national weather bureaus of the various host nations, each with their own access rules and restrictions. In addition, each bureau retains its own policies pertaining to data retention and archiving. Some weather bureaus present the data as calculated data products such as the rainfall rate which are useless for identifying meteorite falls, although they may retain the unprocessed radar reflectivity data in archive. Very few nations make their radar data available in a near-real-time available manner similar to NEXRAD, and most require filing a request by email to the appropriate agency with a wait period involved. However, if these obstacles can be overcome, there should be an appreciable improvement in meteorite recovery time and the number of meteorites recovered.