Abstract

Chaff is a radar countermeasure typically used by military branches in training exercises around the United States. Chaff within view of the S-band WSR-88D beam can appear prominently on radar users’ displays. Knowledge of chaff characteristics is useful for radar users to discriminate between chaff and weather echoes and for automated algorithms to do the same. The WSR-88D network provides dual-polarimetric capabilities across the United States, leading to the collection of a large database of chaff cases. This database is analyzed to determine the characteristics of chaff in terms of the reflectivity factor and polarimetric variables on large scales. Particular focus is given to the dynamics of differential reflectivity ZDR in chaff and its dependence on height. In contrast to radar observations of chaff for a single event, this study is able to reveal a repeatable and new pattern of radar chaff observations. A discussion about the observed characteristics is presented, and hypotheses for the observed ZDR dynamics are put forth.

1. Introduction

Chaff is a military radar countermeasure consisting of small, metallic-coated silica fibers that are released in large quantities via aircraft or rockets to mask targets of interest to an enemy radar (Fig. 1; De Martino 2012). These fibers are cut to lengths that resonate at a desired frequency, meaning that chaff fibers are often cut on the order of centimeters (Hessemer 1961; Palermo and Bauer 1965). In cases observed by the Weather Surveillance Radar-1988 Doppler (WSR-88D) network, a range of lengths between 1 and 20 cm have been observed on the ground where they were deposited (as reported in various news stories about chaff observations on radar), although both longer and shorter lengths are possible depending on the mission type. Given these lengths of widely dispersed, metallic dipoles, chaff releases near centimeter-wave weather radars can cause considerable contamination of the radar volume (Hall et al. 1984; Zrnić and Ryzhkov 1999). Chaff releases are intended to disperse into “clouds” that can occupy large volumes and appear similar in nature to stratiform or convective precipitation (Harrison and Heinz 1963; Pinson 1975). This behavior is ideal for the military to use while attempting to mask aircraft, warships, and missiles from enemy detection (Pode 1960) but can cause problems for users of weather radar, who view the chaff as undesired clutter. An example of a chaff release on the Key West, Florida, (KBYX) WSR-88D is shown in Fig. 2; note that virtually all of the returns in the image are from chaff over areas of tens of thousands of square kilometers, essentially flooding the radar volume with clutter.

Fig. 1.

Individual strands of X-band chaff (cut to approximately 1.5 cm). Multiple clumps of chaff are also shown. The authors were only able to acquire X-band chaff for testing purposes, even though multiple chaff lengths are likely responsible for returns on the S-band WSR-88D network.

Fig. 1.

Individual strands of X-band chaff (cut to approximately 1.5 cm). Multiple clumps of chaff are also shown. The authors were only able to acquire X-band chaff for testing purposes, even though multiple chaff lengths are likely responsible for returns on the S-band WSR-88D network.

Fig. 2.

East–west-oriented chaff clouds as seen by the KBYX WSR-88D at 2001:04 UTC 12 Feb 2016 and 0.5° elevation angle. Variables shown are (a) reflectivity factor (dBZ), (b) differential reflectivity (dB), (c) copolar cross-correlation coefficient (unitless), and (d) differential phase (°).

Fig. 2.

East–west-oriented chaff clouds as seen by the KBYX WSR-88D at 2001:04 UTC 12 Feb 2016 and 0.5° elevation angle. Variables shown are (a) reflectivity factor (dBZ), (b) differential reflectivity (dB), (c) copolar cross-correlation coefficient (unitless), and (d) differential phase (°).

While chaff is primarily a military countermeasure, it is used for training exercises across the United States on a nearly daily basis. It is also still used for radar testing and calibration. Although the chaff in Fig. 2 is dispersed by aircraft in linear segments, chaff can often take the shape of small convective cells, stratiform regions, and everything between, including biological scatterers and combustion debris, because of its tendency to act as a tracer of the three-dimensional wind field (Schurr et al. 2003; Ryzhkov et al. 2005; Melnikov et al. 2008; Murphy et al. 2016; Kurdzo et al. 2017). In some cases, chaff can comingle with weather echoes and can even be ingested into ongoing weather, making it particularly difficult for radar users to distinguish between chaff and weather, especially with a static image.

Although chaff can resemble weather returns in the estimates of reflectivity factor Z, it can often be distinguished from weather using the polarimetric variables (Figs. 2b–d) because of its strongly aspherical shape. These variables include differential reflectivity ZDR, copolar cross-correlation coefficient ρHV, and differential phase ϕDP (Doviak and Zrnić 1993; Bringi and Chandrasekar 2001; Zhang 2016). The dual-polarimetric WSR-88D network allows for broad characterization of chaff properties across the United States. Trained forecasters usually can detect the difference between chaff and weather, especially through the use of the polarimetric variables, but the knowledge to determine what echoes are chaff starts with a determination of chaff characteristics. This characterization was first offered by Zrnić and Ryzhkov (2004, hereinafter ZR04), who used a polarimetric radar to determine the polarimetric properties of chaff releases in Oklahoma. Their study utilized one dataset at a single elevation and six subsequent times to create scatterplots of the polarimetric variables relative to Z. Since that time, no broader study of chaff characteristics has been conducted using S-band weather radars. With the proliferation of dual-polarimetric WSR-88D observations now at our disposal, this study aims to offer a comprehensive overview of the polarimetric properties and characteristics of chaff across the United States at different times, elevations, locations, and seasons.

Kim et al. (2013) utilized spatial and temporal clustering to detect and filter chaff using single-polarization radar data, and Yu et al. (2016) also used single-polarization data to create a tree-initialized genetic algorithm to train a fuzzy-logic scheme. Chaff has been studied within the meteorological community primarily as a wind tracer for storm and cloud entrainment (e.g., Jessup 1972; Moninger and Kropfli 1987; Reinking and Martner 1996; Jung and Albrecht 2014), clear-air tracing (Rowland 1976; Sauvageot et al. 1982), and cloud electrical studies (Kasemir et al. 1976; Rust and Krehbiel 1977; Helsdon 1980; Maddox et al. 1997). Some of the first observations of chaff on the pre-dual-polarimetric WSR-88D network were reported in Roeder (1995), in which the difficulty of determining chaff versus weather with single-polarimetric radar is acknowledged. Additional work about chaff with the WSR-88D was presented by Arnott et al. (2004), where distributions, fall speeds, and concentration of chaff particles were studied. Gillies and Nickling (2003) explored the entrainment of chaff by wind after fallout as a function of surface roughness, finding that chaff can occasionally be relofted after falling to the ground.

This study seeks to build upon the results from ZR04 to increase our knowledge of how chaff appears to the weather radar user. This paper takes the following form: an introduction to the data and methods used for generation of a chaff database is presented, along with the methods for processing and for human verification. The resulting properties of chaff are then presented, followed by a series of case studies that focus on the dynamics of ZDR in chaff. The observations of ZDR in chaff differ significantly in this study from those in ZR04, resulting in a focus on this area specifically. Discussion about these observations and characteristics follows with a presentation of three working hypotheses for negative ZDR in chaff (including both electrical and mechanical methods, as well as a combination of the two). Conclusions are then drawn about our results and how they can be used by the general WSR-88D user base.

2. Data and methods

a. Datasets and case studies

The collection of data for the chaff database spans the year of 2016 and includes 75 selected cases from 33 different WSR-88D units from all geographical regions of the continental United States. Each case consists of a single volume at the approximate midpoint of fallout (in time), with all volume coverage patterns (VCPs) represented. Approximately 20% of the cases occurred in clear-air VCPs (with 94% of these cases occurring in VCP 32), while the remaining 80% took place in precipitation VCP modes (with 64% of the cases occurring in VCP 12). In total, over 2.2 million individual resolution volumes are included [at legacy resolution (1° by 0.25 km); see section 2b]. The histograms and scatterplots in section 3a are derived from this dataset.

In addition to the single-volume cases that compose the master database, eight individual cases (from radars; see Table 1) were chosen to highlight ZDR dynamics and how they change with time and height. Because each case is different (e.g., varying release points and heights, thermodynamic and kinematic atmospheric conditions, and the presence of weather), a simple conglomeration of cases would not be appropriate to investigate trends with height. These cases span all elevations and times from the beginning of the chaff release to final fallout, and only cases that were serendipitously close to radar sites were chosen so as to sample at the lowest elevations possible. Since chaff clouds may be released at different times, elevations, and so on, only a single cloud for each case was selected for tracking. A total of 387 volumes were analyzed over these cases. The height analyses group data into 0.5-km bins regardless of point density. The time analyses are smoothed with a 5-by-5 moving-average filter that operates along the data and time dimensions.

Table 1.

Individual cases chosen for analysis.

Individual cases chosen for analysis.
Individual cases chosen for analysis.

b. Processing methods and the ORPG simulator

The operational WSR-88D Open Radar Product Generator (ORPG) ingests data from the radar (level 2), processes the data, and produces the “official” Next Generation Weather Radar (NEXRAD) products (level 3) used by government agencies and the private sector. Each ORPG is assigned to operate with one radar in real time. An offline, research version of the ORPG, the Common Operations and Development Environment (CODE), does the same but instead ingests level-2 data from any live-feed NEXRAD data stream or playback from archived radar volumes. The products created with CODE are still “official” quality but are not the official products. ORPG and CODE include a specific processing path for handling dual-polarimetric data that includes a preprocessor and quality-control steps leading to the operational hydrometeor classification algorithm (HCA). CODE is very useful as the “sandbox” in which new algorithms and capabilities and resultant products are developed for eventual porting to the operational ORPG. CODE is not amenable to massive data processing, however, and therefore the ORPG simulator is needed.

To generate dual-polarimetric level-3 products at legacy resolution, an ORPG simulator was developed that ingests level-2 data and mimics the entire dual-polarimetric ORPG/CODE chain in a MATLAB (software package) environment for simple processing. In the spirit of CODE, a ϕDP level-3 product was also added since such is not currently included in operations but is needed for this work. In addition, texture fields of Z and ϕDP were included. The ORPG simulator allows rapid collection and archiving of chaff cases for possible implementation into other algorithms (e.g., an expanded HCA) in the future. Comparisons between our ORPG-simulated level-3 products and the level-3 products provided by the National Centers for Environmental Information show virtual agreement. The operational HCA (Park et al. 2009) in the WSR-88D ORPG requires that data be at legacy resolution (1° instead of 0.5° in azimuth). The dual-polarimetric preprocessor acts on the level-2 estimates, resulting in the dual-polarimetric level-3 estimates that have less noise and transformed resolution. Given this method and observations that the general trends shown throughout this paper are consistent with the level-2 data, only level-3 data are presented for the sake of brevity.

Because ZDR analysis is a critical aspect of this paper, it is prudent to discuss the ±7.9-dB limits to the ZDR field in the ORPG and how these limits may change our analysis in the future. When the ZDR product was originally developed, it was determined that the product would be limited to 1-byte encoding of ZDR per pulse resolution volume. A number of proposed binning schemes (scale and offset) were discussed by a product committee. The consensus was to have uniform decibel steps through the 256 bins of the product and that the maximum/minimum range of the current product was sufficient to account for the expected scatterers encountered. Since the implementation of dual-polarimetric capabilities on the WSR-88D network, a new consensus developed for an expanded range for ZDR. This likely will result in the product of the future being more than 1 byte for encoding or uneven bin (decibel step) distributions or both to cover an expanded range (WSR-88D Radar Operations Center 2016). This means that our analysis will likely need to be expanded to fully understand the range of ZDR in chaff in future builds of the ORPG.

c. Verification by humans

The core function of this study is to determine which echoes are chaff and to investigate their statistical properties. Therefore, a verification step ensures that only pulse resolution volumes that contain chaff are included in the analysis. To achieve this goal, a method using examination by human experts was developed. Using all of the available radar variables, as well as temporal loops and volumetric analysis, cells believed to be chaff were manually traced (using definitions described in the upcoming paragraphs). Using a graphical-user-interface tool, these traces were converted into masks that were applied to all available data to gather all instances of chaff into one large dataset, as well as the individualized case studies shown in section 3. For each resolution volume, the following information was saved: Z, ZDR, ρHV, ϕDP, elevation angle, VCP, and estimated altitude above radar level (ARL) on the basis of a 4/3 Earth-radius model (Doviak and Zrnić 1993). In addition, the two texture fields used in the operational HCA were calculated and saved: standard deviation of Z [SD(Z)] and standard deviation of ϕDP [SD(ϕDP)]. These standard deviations are calculated by comparing the output of a sliding smoothing window with the original data. For an explanation of the use of the term “texture” in this study, the reader is referred to Park et al. (2009).

Verification of chaff events has proven to be difficult since confirmation from the releasing agency or military branch is rarely provided. The original determination of chaff characteristics stemmed from literature-provided statistics (e.g., low ρHV, high ZDR, and a “falling” pattern; ZR04), discussions with forecasters and other radar users, and collection of social-media posts from National Weather Service accounts about chaff on their radar displays. Through an intense observation period of WSR-88Ds around the continental United States, a series of qualitative characteristics was determined to indicate the presence of chaff:

  1. relatively high (greater than 100° on average) ϕDP estimates with a high texture parameter,

  2. relatively low ρHV estimates (below 0.7, although higher values are interspersed),

  3. negative ZDR near the bottom of a chaff cloud and positive values near the top,

  4. exceptionally low estimates of spectrum width (below 3 m s−1), and

  5. a temporal dispersion pattern (from a release-time perspective) that acts as a wind tracer and progresses from an elevated “point” source to a lower-level “cloud” (in general, the chaff clouds disperse horizontally while slowly descending vertically, resulting in a broadening coverage in time and space).

Application of these rough guidelines led to the archiving of potential cases, and these cases were further vetted for adherence to the expected criteria and similarity from case to case.

3. Characteristics of collected data

a. Statistical properties of chaff

The 75 collected chaff cases were combined into one dataset for histogram analysis. Histograms of Z, ZDR, ρHV, and ϕDP are shown in Fig. 3. These histograms include data from all radar scan elevations. Reflectivity Z displays a range from −25 to 45 dBZ, with a plateau between 0 and 20 dBZ. A nearly linear increase in frequency of occurrence is evident between −25 and 0 dBZ, and a sharper decrease is seen after 20 dBZ. When compared with ZR04, a much wider range of Z values is evident, although note that outlier cases such as one event in Alabama (Murphy et al. 2016) can push this range even farther, with the potential for Z values over 60 dBZ.

Fig. 3.

Histograms of (a) reflectivity factor (dBZ), (b) differential reflectivity (dB), (c) copolar cross-correlation coefficient (unitless), and (d) differential phase (°) in chaff. Data were compiled using over 2.2 million resolution volumes of level-3 WSR-88D data from 75 cases in 2016.

Fig. 3.

Histograms of (a) reflectivity factor (dBZ), (b) differential reflectivity (dB), (c) copolar cross-correlation coefficient (unitless), and (d) differential phase (°) in chaff. Data were compiled using over 2.2 million resolution volumes of level-3 WSR-88D data from 75 cases in 2016.

Of all the polarimetric variables, ZDR displays perhaps the most striking difference from ZR04. Instead of predominantly positive values in ZR04, a range covering the entire spectrum of available ZDR values is seen. The concentration of positive ZDR values is higher than for their negative counterparts, but the negative component of ZDR values is conspicuous. A near-linear increase in probability of occurrence is evident from −7.9 to +1.5 dB in Fig. 3b, with a linear decrease from +1.5 to +7.9 dB. The characteristics of ZDR in chaff, specifically the differences between negative and positive values, are the subject of section 3b and will be given additional treatment in following sections.

Differences relative to ZR04 are also displayed in ρHV in that values close to 1 are somewhat common. Although the peak of ρHV values is in the 0.5–0.7 range, which agrees well with ZR04, values as low as 0.1 and as high as 1.0 are evident. This is counterintuitive because of the expected random “fluttering” nature of chaff, especially with the substantial range seen in other parameters (specifically ZDR and ϕDP). Numerous chaff clouds have displayed ρHV values as high as would be expected in raindrops (near 1.0), however, making characterization even more difficult. Of concern to the radar user is the fact that multiple parameters can vary significantly in time and space in a single chaff cloud; this fact is addressed in section 4.

The distribution of ϕDP is of particular interest to radar users because it stands out from weather more than the other variables. While typical weather echoes in light rain may be well below 100° and increase in a linear fashion along the radial, ϕDP in chaff tends to be over 100° and can appear to be stochastic in nature, both in the temporal and spatial dimensions. These values are relative to the system phase, which on the WSR-88D is generally set to 60° to prevent negative phase values. This assumption is carried throughout the remainder of the paper. Note that when switching between VCPs slight changes in the reference phase can be made, but these changes have not been recognized as significant. The ϕDP values in chaff peak around 115° and stay relatively high in concentration past 200°—values that are rarely seen in rainfall. Similar to ZDR and ρHV, ϕDP in chaff covers the entire spectrum of available values, making it difficult to detect chaff from an algorithmic perspective.

Additional parameters calculated in the ORPG and readily accessible through the simulator can be useful in detecting chaff. Included in the HCA in particular are texture fields for Z [SD(Z)] and ϕDP [SD(ϕDP)]. These parameters represent running standard deviations of Z and ϕDP with different window sizes along a radial: five resolution volumes for Z and nine resolution volumes for ϕDP. They are ingested into the HCA to aid in discriminating between smooth and rough targets (in an image-processing sense), making the fuzzy-logic system more robust (Park et al. 2009). These parameters were calculated using the ORPG simulator and are shown in Fig. 4. SD(Z) shows relatively low values not dissimilar to light rain and snow/ice. This is due to the smooth nature of many chaff clouds in the Z field (i.e., a lack of strong Z gradients), likely a result of slow dispersion as the chaff traces the wind field. SD(ϕDP), on the other hand, displays very high values (approaching 100°), indicating a contextually “rough” surface with high variance between resolution volumes in the ϕDP field. Although not explicitly calculated by the ORPG, typical values for SD(ρHV) range from 0.05 to 0.3, also indicating a high value of texture and a wide range of ρHV values in close proximity.

Fig. 4.

Histograms of (a) standard deviation of reflectivity factor (dBZ), (b) “standard” standard deviation of differential phase (°), and (c) “full” standard deviation of differential phase (°) in chaff. The full standard deviation of differential phase removes the 100° limitation currently implemented in the ORPG. Data were compiled using over 2.2 million resolution volumes of level-3 WSR-88D data from 75 cases in 2016.

Fig. 4.

Histograms of (a) standard deviation of reflectivity factor (dBZ), (b) “standard” standard deviation of differential phase (°), and (c) “full” standard deviation of differential phase (°) in chaff. The full standard deviation of differential phase removes the 100° limitation currently implemented in the ORPG. Data were compiled using over 2.2 million resolution volumes of level-3 WSR-88D data from 75 cases in 2016.

Figure 4b shows the SD(ϕDP) as calculated in the current ORPG implementation [the so-called standard SD(ϕDP)]. This method limits the ϕDP differential threshold to 100° because of the lack of targets with high values of SD(ϕDP) and to avoid errors. When this restriction is removed, the true nature of SD(ϕDP) in chaff is exposed, as seen in Fig. 4c. This full SD(ϕDP) shows average values near 100°, with many values up to 200°. These observations offer a tremendous opportunity to discriminate between chaff and other targets, especially meteorological targets. Note that exceptionally low values of SD(ϕDP) (near 0) do still occur in chaff, and therefore the use of full SD(ϕDP) is not sufficient by itself to distinguish chaff from other targets. In addition, Kurdzo et al. (2017) found that exceptionally high values of SD(ϕDP) also occur in sea clutter, combustion debris, and during radio-frequency interference.

To offer direct comparisons with ZR04, scatterplots of Z versus ZDR and Z versus ρHV are presented in Figs. 5 and 6, respectively. Because of the substantial size of the dataset, a heat-map format was chosen to display trends instead of a traditional scatterplot. The “hotter” (yellower) colors indicate a stronger signal (more occurrences in this case), and the “cooler” (bluer) colors indicate fewer observations. Figure 5 shows Z versus ZDR, with a peak in the range of 10–20 dBZ for Z and in the range from 0 to +3 dB for ZDR. These values align well with the histograms shown in Fig. 3, although a slight trend is evident on a larger scale. In general, the trend seems to be from the top right of the plot toward the bottom left, indicating higher Z values when ZDR is higher and lower Z values when ZDR is lower. This result makes intuitive sense because Z in this case is estimated using the horizontal channel of the return signal. The trend does not seem to serve any real significance for chaff identification outside the data shown in Fig. 3.

Fig. 5.

Heat map of reflectivity factor (dBZ) vs differential reflectivity (dB) in all chaff events. Shading indicates number of occurrences per grid point. Data were compiled using over 2.2 million resolution volumes of level-3 WSR-88D data from 75 cases in 2016.

Fig. 5.

Heat map of reflectivity factor (dBZ) vs differential reflectivity (dB) in all chaff events. Shading indicates number of occurrences per grid point. Data were compiled using over 2.2 million resolution volumes of level-3 WSR-88D data from 75 cases in 2016.

Fig. 6.

As in Fig. 5, but for reflectivity factor (dBZ) vs copolar cross-correlation coefficient (unitless).

Fig. 6.

As in Fig. 5, but for reflectivity factor (dBZ) vs copolar cross-correlation coefficient (unitless).

The plot of Z versus ρHV shown in Fig. 6 displays a substantial focus in the 10–20-dBZ range for Z and in the 0.4–0.6 range for ρHV. This result deviates slightly from the ρHV plot in Fig. 3, which maintained exceptionally high levels almost entirely up to 1.0. Although this trend is evident in Fig. 6, a strong focus toward the center of the plot is clear as well. The sudden drop-off of values over 1.0 is due to the relative rarity of values over 1.0 in the WSR-88D ORPG (for any target), making the results appear somewhat discontinuous. There appears to be a potential trend related to lower values of Z indicating higher values of ρHV. This may be a useful clue in terms of why we see so many high ρHV values, but the data are not entirely clear. This trend is not evident in the ZR04 scatterplot.

Note that we maintain an underlying assumption of sufficiently stochastic fiber orientations. Although there may be areas of the chaff cloud with different tendencies than those found in other portions of the cloud (e.g., areas of predominantly negative or positive ZDR), the distribution of orientations is still assumed to be unpredictable and “random” about a mean. Variations from this randomness would certainly affect the data distributions, but one cannot control for these variations. For example, chaff-fiber orientations that vary between along radial and orthogonal could cause changes in distributions with time. In this case, a mean along-radial orientation would lead to lower Z values. Although we cannot quantify these effects, they should be considered when factoring in the contribution from all radar observables.

b. Time–height characteristics of ZDR in chaff

It was noticed early in the study that some cases contain almost entirely negative ZDR while others consist of almost entirely positive ZDR. When these individual chaff clouds were examined over their entire lifetimes, however, organized and repeatable transitions between negative and positive ZDR were observed. Different parts of the chaff cloud display consistently negative or positive ZDR, similar to Fig. 2b. Over time, it was discovered that there tends to be more negative ZDR values at lower elevations and more positive ZDR values at higher elevations. After additional examination, this observation was modified to indicate that the earlier stages of fallout (at the lowest levels) produce negative ZDR whereas the later stages (also at the lowest levels) produce positive ZDR (see Fig. 13, described in more detail below). These observations of time and height dependence led to the exploration of ZDR characteristics that is presented in this section.

It is not feasible to analyze the time and height characteristics in chaff for a conglomeration of cases, since each chaff drop occurs at a different elevation and under different atmospheric conditions, leading to different fallout characteristics and dispersion rates. Eight cases have been selected for further analysis, as listed in Table 1. Two different analyses are presented for each case; the first is an analysis of the ZDR distribution with height across the entire timespan of a single chaff cloud, and the second is an analysis of the ZDR distribution with time at the lowest available elevation angle throughout the lifespan of the same chaff cloud. The time intervals are not equal because of the use of differing VCPs. Single clouds that stayed separated from other clouds were selected so as to differentiate between release times and heights, and different stages of fallout were selected to accurately depict the ZDR trends in time and height.

The eight height analyses (cases A–H) are presented as violin plots in Figs. 7 and 8 and are labeled with a–h for identification purposes in the two figures. Each case displays different available ranges of elevations for analysis on the basis of the proximity to the radar and the VCP in use at the time. The most notable trend across the cases is the existence of higher (i.e., positive) ZDR aloft and lower (i.e., negative) ZDR toward the surface. This is prominently shown in all cases except C. Some cases (A, B, E, and F) display a smooth transition from strongly positive ZDR to strongly negative ZDR from top to bottom, and other cases (C, D, and G) maintain a significant proportion of positive ZDR in their distributions down to the lowest available elevation. Despite some cases lacking a distinctly clear low-level negative-ZDR signature, it is generally true throughout the cases that the higher elevations contain more positive ZDR values than do the lower elevations.

Fig. 7.

Violin plots of differential reflectivity (dB) distribution by height ARL in chaff cases A–D. Each case tracks a single chaff cloud over its entire lifetime at all available elevations. Vertical black and red lines indicate the mean and median values, respectively. Extensions beyond the ±7.9-dB limitation are due to the Gaussian kernel bandwidth and are not indicative of actual values beyond the limits. The highest elevations are not necessarily shown because of data sparseness.

Fig. 7.

Violin plots of differential reflectivity (dB) distribution by height ARL in chaff cases A–D. Each case tracks a single chaff cloud over its entire lifetime at all available elevations. Vertical black and red lines indicate the mean and median values, respectively. Extensions beyond the ±7.9-dB limitation are due to the Gaussian kernel bandwidth and are not indicative of actual values beyond the limits. The highest elevations are not necessarily shown because of data sparseness.

Fig. 8.

As in Fig. 7, but for cases E–H.

Fig. 8.

As in Fig. 7, but for cases E–H.

It is important to note that some of the elevations are much more densely populated than other elevations because of beam position differences and different fallout times and characteristics. In addition, this analysis does not take into account chaff density during fallout. This means that, if a chaff cloud disperses as it falls, the latter portions of the fallout will contribute an anomalously high percentage to the totals at lower elevations. Also, the extension in some cases beyond the ±7.9-dB limitation is due to a Gaussian kernel bandwidth inherent to violin plots and is not indicative of actual values beyond the limits.

To attempt to correct for the shortcomings in a purely height-based analysis, an analysis of the ZDR distribution with time at the lowest available elevation for each case is presented in Figs. 9 and 10. The same figure labels of a–h are used to allow comparison with the cases in Figs. 7 and 8. This analysis technique is meant to show the transition of the ZDR distribution from the beginning to the end of a single chaff cloud fallout at a single elevation. The added dimension of time helps to paint a clearer picture of the ZDR characteristics. In each case presented, the general trend is to observe negative ZDR values early in the fallout (bottom of each image) and positive ZDR values toward the tail end of the fallout (top of each image). This pattern indicates that the vertically oriented chaff fibers fall out more quickly than the horizontal chaff fibers do, which makes intuitive sense. It also leads to the question of why the fibers are becoming vertically oriented in the first place, since negative ZDR is indicative of vertical fallout; this question is addressed in section 4.

Fig. 9.

Differential reflectivity (dB) distribution by time in chaff cases A–D. Each case tracks a single chaff cloud over its entire lifetime at the 0.5° elevation angle. In each case, the early times display near-zero or negative ZDR and the later times display strong positive ZDR.

Fig. 9.

Differential reflectivity (dB) distribution by time in chaff cases A–D. Each case tracks a single chaff cloud over its entire lifetime at the 0.5° elevation angle. In each case, the early times display near-zero or negative ZDR and the later times display strong positive ZDR.

Fig. 10.

As in Fig. 9, but for cases E–H.

Fig. 10.

As in Fig. 9, but for cases E–H.

All of the cases display strongly positive ZDR values at the end of fallout. The beginning of the fallout varies, with some cases (B–D) displaying generally negative values at the first observation and the other cases starting close to zero (E and H) or just above zero (A) before becoming strongly negative. Most cases achieve a minimum high probability of ZDR in the range from −4 to −6 dB, but some cases (A, F, and H) extend all the way to −7.9 dB. All of the cases finish at nearly the maximum possible ZDR. Some cases (B, D–F, and H) display generally linear trends from negative to positive with time, and others (A, C, and G) exhibit more of a curvilinear shape that features a rapid transition from negative to positive ZDR. Regardless of the subtle differences, these observations were extremely common in this study. Although not fully ubiquitous, we estimate that the observation of negative ZDR occurs in over 80% of the chaff cases assembled in this study.

4. Discussion

a. Observations of reflectivity

Although some of the examined parameters are within the expected ranges for chaff, others seem somewhat counterintuitive. Reflectivity, for example, is lower than might be expected for highly reflective dipoles. As shown in Arnott et al. (2004), however, the cut length of chaff can vary returns significantly. If the cut length is not resonant at S band, the returns can be significantly lower than cuts made specifically for S-band radars. For this study, the cut lengths are not known. In cases such as that shown in Murphy et al. (2016), Z values over 60 dBZ were observed, however. Even considering the lack of substantial wind in that event, it is plausible to expect that different chaff lengths may be dropped in different scenarios. Although many of the chaff cases that we sampled were in the 5–25-dBZ range, some of the examples in late 2016 exhibited even lower values (below 0 dBZ); this situation is reflected in Fig. 3a and may be a result of different VCP strategies. In these cases, it is possible that yet another chaff length is being dropped. Without confirmation from those deploying the chaff, it is difficult to know the exact type and length of chaff being dropped. Understanding the range of possibilities is important for forecasters and other radar users.

Arnott et al. (2004) show strong dependence of Z on both chaff type and concentration. For one chaff tube per kilometer cubed (a common measure of release amount in the military; chaff is released in tubes that are detonated to invoke dispersion), for example, the difference between R129-type chaff (with approximately 3.6 × 106 fibers per tube), having up to 5.08-cm cut length, and R189-type chaff, with up to 1.88-cm cut length, is 35 dB (55 dBZ for R129 and 20 dBZ for R189). In addition, even the chaff types that are resonant at S band can decrease to 20 dBZ when the concentrations fall below 1 × 10−3 chaff tubes (or approximately 3.6 × 103 chaff fibers) per kilometer cubed. This can occur when less chaff is dropped, or if chaff is dropped over a longer area or over a longer time period. These variations lend themselves to explaining the wide variety of Z values sampled across the study cases.

b. Observations of correlation coefficient

Another counterintuitive observation is the pervasiveness of relatively high ρHV values that approach values one would see in stratiform rainfall (uniform small drops with ρHV values near 1.0). Although the sample size was small, ZR04 reported maximum ρHV values of roughly 0.6, leading to a large discrepancy with our dataset. The ρHV values are shown in Fig. 2c, and, similar to the ZDR values in Fig. 2b and across many other cases, the high ρHV values seem to cluster together in different areas of the cloud. In this particular case, the higher values are near the edges of the clouds, which could indicate a potential estimation artifact that is due to low signal-to-noise ratio (similar to areas of ρHV >1.0 in rain). This is not always the case, however. In a general sense, similar to the SD(ϕDP) field, the texture of ρHV can be exceptionally high in chaff, leading to values as low as 0.2 being within a few resolution volumes of values as high as 1.0. Cases of ρHV close to 1.0 in the middle of chaff clouds are common in our dataset. No clear relation or dependency between ρHV and any other variable was evident (plots not shown).

We postulate that the occasional values of high ρHV are due to low turbulence values, as evidenced by low spectrum widths in the data, allowing for occasionally near-uniform flutter angles as mentioned in ZR04. Arnott et al. (2004) also discuss lower Reynolds numbers leading to smaller fall speeds, possibly indicating more consistent flutter angles and orientations. Although we agree that the near-uniform-flutter-angle approximation is crude in most cases, certain areas of the cloud may exhibit sufficiently low turbulence to allow a closer approximation. This may result in anomalously high ρHV values in pockets of the cloud. It is also important to note that the observation of low spectrum width in most chaff cases may be due to the fact that chaff drops often take place in fair weather and without strong wind shear. In the few cases that were observed near thunderstorms, the values of spectrum width were still sufficiently small for our hypothesis to hold.

c. Observations of differential phase

The appearance of exceptionally noisy ϕDP estimates [SD(ϕDP) greater than 100°] in chaff is similar to that seen in clutter (Ryzhkov et al. 2005). The highly variable ϕDP estimates in clutter are leveraged in the existing HCA, specifically in the SD(ϕDP) function (Park et al. 2009). In a highly randomized distributed target, and especially in a target with significant changes in flutter angle acting as a dipole, we would expect the effect on ϕDP to be difficult to predict. For a given range bin, the chaff fibers that contribute most to the horizontal return likely differ from the fibers that contribute most to the vertical return. This is expected to result in random-phase sums of the horizontal and vertical echo components, making for a stochastic nature in ϕDP. In the case of chaff, assuming a lack of common alignment, the horizontal and vertical return components would be random in the sense that the return is a random-phase sum of the contributions from the randomly oriented fibers.

Through the modeling of thin wires/dipoles discussed in ZR04, we can assume that approximating chaff as a dipole (or thin cylindrical antenna) is sufficient when no clumping is involved. With this assumption, it is expected that the primary component of phase shift estimated at the radar is from propagation phase rather than from a combination of backscatter phase and propagation phase because the backscatter phase would be nearly equal in both planes for a thin target. Given the fact that the measured chaff samples (discussed later in this section) are not all fully conducting, there may be some backscatter differential phase contribution in the weaker return dimension from such fibers. This understanding does break down, however, when clumping is considered. Once the chaff can no longer be modeled as a dipole, backscatter phase may become a contributor. Because we suspect that clumping only occurs early in fallout and is not a significant contributor to the observed variables (discussed in the following sections), no attempt at modeling clumped chaff was made in this study.

d. Observations and dynamics of differential reflectivity

The most intriguing new finding in this study is the prevalence of negative ZDR in chaff toward lower altitudes. Previous studies have shown predominantly positive values of ZDR in chaff (ZR04; Schurr et al. 2003; Scharfenberg et al. 2005; Melnikov et al. 2008), and, in general, it would be expected that a dipole such as chaff would fall horizontally, assuming no manufacturer intentions to the contrary. This is due to the preferred aerodynamic torque generated by air resistance favoring a horizontal fall pattern, similar to the way a flat-plate ice crystal (or a feather) falls through the air (Weinheimer and Few 1987). We see a significant percentage of positive ZDR values (<60%), but a nontrivial number of negative ZDR values (>30%) have been observed in our cases. As shown in Figs. 710, these negative values seem to be most prevalent at lower elevations (below approximately 2 km) and early on (in the first half) during chaff fallout.

A vertically oriented dipole offers less aerodynamic resistance than a horizontally oriented dipole, for the same gravitational force. Accordingly, the former object will cut through the air at higher speeds than the latter. Given that we observe predominantly negative ZDR first at the lowest elevations (temporally), it makes sense that we are seeing the fastest-falling chaff particles first and as vertically oriented particles. By this logic, we would expect to see the fastest-falling (vertical) particles first, and the slowest-falling (horizontal) particles last; this agrees well with our observations. The relevant questions then become: What causes the chaff particles to become vertically oriented in the first place and at what altitude(s)? We have three working hypotheses with regard to vertically oriented chaff: influence by the fair-weather electric field, mechanical influences that are based on clumping and weight, and a combination of the two that involves distributions of chaff fibers with different lengths and electrical properties.

1) Hypothesis 1: The fair-weather electric field

Because most of the observations of vertically oriented chaff are in the lower levels of the atmosphere (Figs. 7 and 8), the fair-weather electric field (Chalmers 1967; Israël 1973; Ogawa 1985) is a logical first step in looking for vertical orientation. Although we do see evidence of negative ZDR at higher elevations, it is important to note that the concentration of negative ZDR is significantly skewed toward the lower elevations. Gish (1944) showed that the electric field approaches or exceeds −100 V m−1 close to the ground but drops off exponentially with altitude, with typical values of −25 V m−1 at 2 km AGL and −10 V m−1 at 5 km AGL (MacGorman and Rust 1998). This exponential decrease in the magnitude of the vertically oriented fair-weather electric field above the planetary boundary layer serves as our first hypothesis for orienting chaff vertically as it falls.

Estimates for the magnitude of the electric field needed to orient chaff fibers in the vertical direction are based on earlier theoretical calculations pertaining to ice crystals by Weinheimer and Few (1987). In this treatment, the target shapes are approximated by oblate and prolate spheroids. Key forcing agents are the aerodynamic and electrical torques acting on the falling radar targets. The key geometrical parameter is the axis ratio q. A chaff fiber is closely approximated by a prolate spheroidal conductor, with axis ratio q defined as the ratio of chaff fiber length to diameter. For chaff cut to be resonant at S-band wavelengths, the length L = λ/2 = 5 cm, and, given a diameter of 25 μm = 2.5 × 10−5 m, we have q = 5 × 10−2/2.5 × 10−5 = 2000. The aerodynamic torque acting on the chaff fiber in potential flow is [again following Weinheimer and Few (1987)]

 
formula

where ρa is the local air density, U is the fall speed of the chaff, d is the chaff length, and h′ is a nondimensional parameter that depends only on axis ratio q.

The corresponding electrical torque acting on the chaff fiber in a uniform electric field E is

 
formula

where ϵo is the permittivity of free space and f′ is another nondimensional factor that is dependent only on the axis ratio q. The matching of the aerodynamic and electrical torques is assumed to be the condition for electric field alignment of the chaff, with the long axis of the chaff parallel with the electric field. Equating Eqs. (1) and (2) and solving for the critical electric field (given the condition f′/h′ = f/h) gives

 
formula

The axis ratio for chaff (approximately 2000) is always substantially greater than values for ice crystals (approximately 2–20), and so the plot of f/h (q) in Weinheimer and Few (1987) needs to be extended to larger q, as shown in Fig. 11. This plot demonstrates why strongly elongated chaff (as compared with ice crystals) will be aligned in relatively weak electric fields and why fields of thunderstorm magnitude are not required. By inspection, for q = 2000, f/h is close to 2 × 105. Inserting appropriate values (ρa = 0.7 kg m−3, U = 0.2 m s−1, and ϵo = 8.85 × 10−12 F m−1) into Eq. (3) yields a critical value for E of 63 V m−1. This value is of the same order as the fair-weather electric field in the continental boundary layer and is about twice the fair-weather value in the near-surface air over oceans (Israël 1973).

Fig. 11.

Expanded prolate axis ratio, adapted from Weinheimer and Few (1987). While q increases, as with chaff rather than ice crystals, a weaker electric field is needed to orient the fibers vertically. The X, C, S, and L bands are marked with vertical lines.

Fig. 11.

Expanded prolate axis ratio, adapted from Weinheimer and Few (1987). While q increases, as with chaff rather than ice crystals, a weaker electric field is needed to orient the fibers vertically. The X, C, S, and L bands are marked with vertical lines.

Figure 11 also shows that, for large axis ratio q, the slope of the power-law plot for f/h is close to 2. Taken together with Eq. (3), this scaling implies that the critical electric field for chaff alignment will be inversely proportional to chaff length. In the radar context, this means that the critical field for chaff cut for resonance at C band will be approximately 2 times 63 V m−1 and for the longer chaff cut for L band will be approximately one-half of 63 V m−1. This finding may be key to interpreting the persistence of horizontal orientation of chaff in the late stages of the fallout, when shorter chaff fibers may be the primary radar target (described in the third hypothesis). Through experimentation in a vertical wind tunnel designed to levitate chaff fibers in the airflow, we have found that we can orient some X-band chaff fibers vertically in electric fields approaching 1 kV m−1. Given the square dependence of the electrical torque on aspect ratio, it can be asserted that longer-cut chaff would be more susceptible to vertical orientation on the basis of the fair-weather electric field. The results of Israël (1973) can be extrapolated to imply that these longer-cut chaff particles could be oriented vertically in electric fields approximately from −50 to −100 V m−1.

As further evidence that chaff fibers can be oriented vertically by electric fields, a chaff case on 2 August 2016 off the coast of Key West and near a thunderstorm was investigated in similar fashion to the cases shown in Figs. 7 and 8. In this case, the chaff release occurred roughly simultaneously in time and spatially close (within 10–20 km) to a convective thunderstorm. For the first half of fallout, the chaff maintained a presence adjacent to the convective cell; during the second half of fallout, the chaff became embedded within the convective updraft [in fact, the chaff was lofted nearly 15 kft (~4500 m) higher than the release height once embedded in the updraft]. Tracking of ZDR was performed in this cell, and ZDR was found to be primarily negative (strongly so at times), as seen in Fig. 12. Although the field strengths are likely orders of magnitude higher in the thunderstorm updraft than in the boundary layer in fair-weather conditions, the results in Fig. 12 suggest that virtually all of the chaff fiber lengths are oriented vertically rather than just the longer fiber lengths, even at exceptionally high altitudes. Note that only relatively high altitudes (>4 km) are shown for this case because of its range from the radar. We have made similar observations near and within convective storms numerous times during 2016. We were not able to find cases that were clearly located within clear-air thermals. Future work on this topic could search specifically for such cases and investigate whether the electric field is affected by clear-air updrafts and how lofting would differ from fallout.

Fig. 12.

Violin plot of differential reflectivity (dB) distribution by height ARL as seen by the KBYX WSR-88D on 2 Aug 2016. A single chaff cloud close to (and at times embedded within) a convective thunderstorm was tracked. The orientation of the chaff particles was predominantly negative, even at very high elevations, suggesting that a strong-enough electric field can orient chaff vertically. Vertical black and red lines indicate the mean and median values, respectively. Extensions beyond the ±7.9-dB limitation are due to the Gaussian kernel bandwidth and are not indicative of actual values beyond the limits.

Fig. 12.

Violin plot of differential reflectivity (dB) distribution by height ARL as seen by the KBYX WSR-88D on 2 Aug 2016. A single chaff cloud close to (and at times embedded within) a convective thunderstorm was tracked. The orientation of the chaff particles was predominantly negative, even at very high elevations, suggesting that a strong-enough electric field can orient chaff vertically. Vertical black and red lines indicate the mean and median values, respectively. Extensions beyond the ±7.9-dB limitation are due to the Gaussian kernel bandwidth and are not indicative of actual values beyond the limits.

These results are promising; not all of the fibers at the lowest levels of chaff clouds are vertically oriented, however. Recall that at the tail end of the chaff fallout the final particles are nearly ubiquitously horizontally oriented. On the basis of these observations, this hypothesis cannot fully explain the negative-ZDR phenomenon, especially if all of the fibers are assumed to be identical in nature.

2) Hypothesis 2: Mechanical effects

Our second hypothesis revolves around the mechanical aspects of chaff, including the tendency for chaff to “clump” as it is released from the tube (ZR04; Arnott et al. 2004; Murphy et al. 2016); Arnott et al. (2004) described this as a “bird nesting” effect. Chaff is generally coated with an anticlumping agent for this reason (Brandin et al. 1997). In our wind-tunnel tests, we found clumps of X-band chaff to fall at a rate that is similar to that of vertically oriented chaff, or about an order of magnitude higher than the horizontally oriented chaff. This scale of difference is similar to the fall speed differences we have observed in WSR-88D data between the bottom of the cloud and the top. In these cases, the clump was formed through a joining of multiple fibers near one end point, causing an effectively weighted end to orient the clump vertically. The clumping hypothesis could also aid in explaining why many of the cases examined start off with ZDR near zero before quickly becoming negative. The clumps may “average out” to a near-zero ZDR in a volume and may be followed through the height of the 0.5° radar plan position indicator by separate vertically oriented individual chaff particles immediately afterward.

The second piece to the mechanical theory is that chaff fibers may contain different thicknesses across the length of the fiber. This was proposed by Arnott et al. (2004), who saw this effect when making laser diffraction measurements of individual chaff fiber diameters. A “weighted” end of the chaff due to a different thickness would influence the fiber to fall in a vertical orientation. These differences in thickness and/or weighting may be intentional or may simply be a by-product of the manufacturing process caused by uncontrollable imperfections. If some percentage of the chaff was weighted in this fashion, it would quickly orient vertically and fall out faster, creating a “curtain” effect. This hypothesis would also explain why some of the chaff becomes vertically oriented at relatively high elevations, although it does not explain the eventual concentration of negative ZDR at lower elevations in Figs. 7 and 8.

3) Hypothesis 3: Size sorting and manufacturing differences

Our final hypothesis combines the first two hypotheses with the fact that we do not know what lengths of chaff are being dropped in each case. In addition, we do not know the conductivity distribution of these fibers. It is plausible to think that various lengths of chaff may be dropped at once to target multiple radar wavelengths. De Martino (2012, p. 290) claimed that a chaff payload is “usually a mix of different individual dipole lengths.” Imperfections may also exist in the chaff lengths cut for specific radar wavelengths, as well as their weighting and conductivities; the random samples of chaff in Fig. 1 suggest evidence for multiple fiber lengths. If this is the case, the fallout could come in three stages: first, the clumps would fall out, possibly more quickly than the vertically oriented individual particles, resulting in a ZDR near zero. Second, the longer (or more conductive) chaff particles would be more strongly influenced by the fair-weather electric field (given the quadratic dependence of electrical torque on chaff fiber length), and would therefore be more likely to be oriented vertically and fall out more rapidly than the remaining chaff.

Last, the remaining “short” chaff may not possess a sufficiently large aspect ratio to be oriented vertically by the electric field, even at low elevations, resulting in the slowest fallout and a horizontal orientation. In addition, any inconsistencies in weighting, whether by design or not, would add to this sorting effect. Jones (2000, p. 13-8) stated, “Generally, the longer cuts used for lower frequency radars fall faster than the shorter cuts used for higher frequency radars,” as well as “since a single cut length is restricted in effectiveness to a narrow range of frequencies, different lengths are normally packaged together to provide coverage over a wide range of frequencies” Jones (2000, p. 13-5), adding credence to this hypothesis. Researchers at the University of Alabama Huntsville have also confirmed that chaff drops at the ground often contain a variety of fiber lengths (R. Wade 2017, personal communication).

With respect to the electrical properties of chaff, our tests of the direct-current characteristics of chaff fibers showed significantly varying conductivities ranging from an open electrical circuit (infinite resistance) down to resistance on the order of 0.2 Ω, meaning there was certainly a wide distribution of electrical continuities. When electrical continuity along the strand is lacking, the effective fiber length is shortened, possibly adding to the hypothesized physical size-sorting effect. Note, however, that less-conductive chaff fibers would contribute less to the radar resolution volume returns because of their decreased resonance and reflective characteristics relative to the radar. This means that the radar will selectively “see” the more highly conductive dipoles.

A general schematic of our observed trends is provided in Fig. 13a. The initial release point diverges into negative ZDR toward the bottom (faster fallout speeds) and positive ZDR toward the top (slower fallout speeds). The negative ZDR is observed at the lowest levels and up through the midlevels of the cloud for the first half of the cloud fallout. The positive ZDR is observed at upper levels down through the midlevels. Toward the latter half of fallout, the positive ZDR makes it to the lowest levels and dominates the returns for the remainder of fallout. This scenario has repeated itself dozens of times in our cases across 2016. A conceptualization of the fallout and dispersion process that is based on consideration of the full space–time evolution of all chaff drops observed is shown in Fig. 13b. Three different chaff fiber orientations are depicted here: horizontally oriented fibers (in black), reorienting fibers (in red), and vertically oriented fibers (in green). This conceptual diagram illustrates the observed dispersion characteristics within the boundary layer, where more rapid dispersion and maximum horizontal extent are often observed during fallout. It is suspected that a marked increase in turbulence and eddy diffusivity within the boundary layer is responsible for expanded horizontal dispersion, independent of fiber orientation. As noted by Stull (1988, his Fig. 5.13), the daytime dissipation rates increase markedly into the boundary layer from the free troposphere above.

Fig. 13.

(a) Conceptual diagram of differential reflectivity trends in a single chaff cloud vs height and time. The near-zero ZDR (clumped; shown in blue) particles fall out first, followed by the negative-ZDR (vertically oriented) particles shown in green. The positive-ZDR (horizontally oriented) particles shown in gray fall out last because of their slower fall speeds. (b) Illustration of the distribution of orientations and their dispersion properties from the free troposphere through the boundary layer and to the ground. The increase in eddy diffusivity in the boundary layer is suspected as a cause for suddenly rapid dispersion at low elevations.

Fig. 13.

(a) Conceptual diagram of differential reflectivity trends in a single chaff cloud vs height and time. The near-zero ZDR (clumped; shown in blue) particles fall out first, followed by the negative-ZDR (vertically oriented) particles shown in green. The positive-ZDR (horizontally oriented) particles shown in gray fall out last because of their slower fall speeds. (b) Illustration of the distribution of orientations and their dispersion properties from the free troposphere through the boundary layer and to the ground. The increase in eddy diffusivity in the boundary layer is suspected as a cause for suddenly rapid dispersion at low elevations.

An example of average ZDR values for each elevation scan from the 5 August 2016 San Diego, California (the KNKX radar), chaff event (case C from Figs. 7 and 9) is shown in Fig. 14. This case matches the conceptual figure well, with a transition in time and height between negative and positive ZDR. Figure 14 also aids in explaining why some of the violin plots in Figs. 7 and 8 (cases C, D, and G) do not contain solely negative values of ZDR at the lowest elevations. Because positive ZDR occurs in the fallout at the latter stages, and these times generally consist of wider dispersion (i.e., more data points), it can be seen that the violin plots in Figs. 7 and 8 do not tell the entire story; the time plots in Figs. 9 and 10 help to fill this gap.

Fig. 14.

Average ZDR (dB) for each elevation scan during the 5 Aug 2016 KNKX chaff event (case C in Figs. 7 and 9). The ZDR trends match closely with the conceptual schematic shown in Fig. 13, with negative ZDR falling out first, followed by positive ZDR toward the tail end. The sloping effect is due to the changing beam height as the chaff moves toward (and then slightly away from) the radar. Note that two radar scan volumes are missing just after 2100 UTC.

Fig. 14.

Average ZDR (dB) for each elevation scan during the 5 Aug 2016 KNKX chaff event (case C in Figs. 7 and 9). The ZDR trends match closely with the conceptual schematic shown in Fig. 13, with negative ZDR falling out first, followed by positive ZDR toward the tail end. The sloping effect is due to the changing beam height as the chaff moves toward (and then slightly away from) the radar. Note that two radar scan volumes are missing just after 2100 UTC.

5. Conclusions

Chaff is a common source of clutter for weather radar users and is visible on radar displays around the United States during military training exercises on a nearly daily basis. Knowledge of chaff characteristics is integral to the user’s ability to decipher meteorological versus nonmeteorological targets, especially in cases of mixed chaff and weather. A large database of radar estimates in suspected chaff clouds has been developed, encompassing 75 cases across much of 2016 from 33 different WSR-88D units. These estimates were used to generate histograms and heat maps of the polarimetric variables in chaff as well as the texture fields used in the current HCA implementation.

The primary new finding of this study is the existence of an abundant amount of negative ZDR in chaff. Time and height trends of the ZDR distribution in eight cases were presented, along with a discussion of the tendencies across the other cases in the database. Three hypotheses were put forth about the existence of negative ZDR in chaff, including the effects of the fair-weather electric field, mechanical clumping and weighting, and size sorting via the electric field and a distribution of chaff sizes and conductivities. These updated observations of chaff and its characteristics will be helpful to forecasters and radar users to determine which echoes are likely to be chaff using the polarimetric variables on the WSR-88D. In addition, these findings can help in the eventual design and implementation of a chaff class in the existing WSR-88D HCA or a chaff-detection algorithm for separating chaff from echoes of greater meteorological interest.

Future work on the distribution of ZDR in chaff will need to take into account any inherent ZDR bias across the WSR-88D network. Although the ZDR values in chaff often reach the ends of the currently reported spectrum (±7.9 dB), ZDR biases in the individual radars could alter the distribution across the entire dataset and could certainly alter the ZDR trends of individual cases. Since polarimetric calibration of the WSR-88Ds is still a challenge, this type of analysis was not possible for this study, but it may be possible in the future. In addition, there are planned changes to the ZDR product in the ORPG/CODE to allow for ranges beyond ±7.9 dB. This analysis will need to be updated for these new possibilities once this is put into place. An expansion of ZR04’s modeling of thin wires might be attempted to model the contribution of clumped chaff to the radar signature. Because a dipole antenna cannot be used to model clumped chaff, such an effort would allow for differentiation between areas of clumping and nonclumping, resulting in potential refinement of the conceptual diagram that is shown in Fig. 13.

Acknowledgments

This document is approved for public release: distribution is unlimited. This material is based upon work supported by the Federal Aviation Administration under Air Force Contract FA8702-15-D-0001. Any opinions, findings, conclusions, or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the Federal Aviation Administration. The authors thank the three anonymous reviewers who aided significantly in the improvement of the paper as well as John Cho for reviewing an early version of the paper and providing very useful insight. The authors also thank Dusan Zrnić for providing the chaff used in the wind-tunnel tests and offering discussions on backscatter differential phase estimates. Discussions with John Cho, Valery Melnikov, Alexander Ryhzkov, Chris Schultz, Ryan Wade, Andy Weinheimer, and Dusan Zrnić led to a better understanding of the negative-ZDR issue and its possible causes. Yen-Jung Wu provided the plot of the expanded prolate axis ratio. The authors also acknowledge assistance from Patrick Arnott, Dave Ebel, Gabe Elkin, and Jim Flavin in our quest to learn more about chaff.

REFERENCES

REFERENCES
Arnott
,
W. P.
,
A.
Huggins
,
J.
Gilles
,
D.
Kingsmill
, and
J.
Walker
,
2004
: Determination of radar chaff diameter distribution function, fall speed, and concentration in the atmosphere by use of the NEXRAD radar. Desert Research Institute Tech. Rep., 30 pp.
Brandin
,
R.
, and Coauthors
,
1997
: Environmental effects of self-protection chaff and flares. U.S. Air Force Tech. Rep., 260 pp, https://ntrl.ntis.gov/NTRL/dashboard/searchResults/titleDetail/PB98110620.xhtml.
Bringi
,
V. N.
, and
V.
Chandrasekar
,
2001
: Polarimetric Doppler Weather Radar: Principles and Applications. Cambridge University Press, 664 pp.
Chalmers
,
J. A.
,
1967
: Atmospheric Electricity. Vol. 11, International Series of Monographs in Natural Philosophy, Pergamon Press, 515 pp.
De Martino
,
A.
,
2012
: Introduction to Modern EW Systems. Artech House, 417 pp.
Doviak
,
R. J.
, and
D. S.
Zrnić
,
1993
: Doppler Radar and Weather Observations. Dover Publications, 562 pp.
Gillies
,
J. A.
, and
W. G.
Nickling
,
2003
:
Entrainment of radio frequency chaff by wind as a function of surface aerodynamic roughness
.
J. Air Waste Manage. Assoc.
,
53
,
206
216
, https://doi.org/10.1080/10473289.2003.10466138.
Gish
,
O. H.
,
1944
:
Evaluation and interpretation of the columnar resistance of the atmosphere
.
Terr. Magn. Atmos. Electr.
,
49
,
159
168
, https://doi.org/10.1029/TE049i003p00159.
Hall
,
M. P. M.
,
J. W. F.
Goddard
, and
S. M.
Cherry
,
1984
:
Identification of hydrometeors and other targets by dual-polarization radar
.
Radio Sci.
,
19
,
132
140
, https://doi.org/10.1029/RS019i001p00132.
Harrison
,
C.
, and
R.
Heinz
,
1963
:
On the radar cross section of rods, tubes, and strips of finite conductivity
.
IEEE Trans. Antennas Propag.
,
11
,
459
468
, https://doi.org/10.1109/TAP.1963.1138074.
Helsdon
,
J. H.
,
1980
:
Chaff seeding effects in a dynamical-electrical cloud model
.
J. Appl. Meteor.
,
19
,
1101
1125
, https://doi.org/10.1175/1520-0450(1980)019<1101:CSEIAD>2.0.CO;2.
Hessemer
,
R.
,
1961
:
Scatter communications with radar chaff
.
IRE Trans. Antennas Propag.
,
9
,
211
217
, https://doi.org/10.1109/TAP.1961.1144981.
Israël
,
H.
,
1973
: Fields, Charges, Currents. Vol. 2, Atmospheric Electricity, Israel Program for Scientific Translations, 796 pp.
Jessup
,
E. A.
,
1972
:
Interpretations of chaff trajectories near a severe thunderstorm
.
Mon. Wea. Rev.
,
100
,
653
661
, https://doi.org/10.1175/1520-0493(1972)100<0653:IOCTNA>2.3.CO;2.
Jones
,
J.
,
2000
: Electronic warfare fundamentals. U.S. Department of Defense Tech. Rep., 351 pp., https://ia800206.us.archive.org/7/items/Electronic-Warfare-Fundamentals/Electronic-Warfare-Fundamentals.pdf.
Jung
,
E.
, and
B.
Albrecht
,
2014
:
Use of radar chaff for studying circulations in and around shallow cumulus clouds
.
J. Appl. Meteor. Climatol.
,
53
,
2058
2071
, https://doi.org/10.1175/JAMC-D-13-0255.1.
Kasemir
,
H.
,
F.
Holitza
,
W.
Cobb
, and
W.
Rust
,
1976
:
Lightning suppression by chaff seeding at the base of thunderstorms
.
J. Geophys. Res.
,
81
,
1965
1970
, https://doi.org/10.1029/JC081i012p01965.
Kim
,
Y. H.
,
S.
Kim
,
H.-Y.
Han
,
B.-H.
Heo
, and
C.-H.
You
,
2013
:
Real-time detection and filtering of chaff clutter from single-polarization Doppler radar data
.
J. Atmos. Oceanic Technol.
,
30
,
873
895
, https://doi.org/10.1175/JTECH-D-12-00158.1.
Kurdzo
,
J. M.
,
B. J.
Bennett
,
M. S.
Veillette
,
D. J.
Smalley
,
E. R.
Williams
, and
M. F.
Donovan
,
2017
: WSR-88D chaff detection and characterization using an optimized hydrometeor classification algorithm. 18th Conf. on Aviation, Range, and Aerospace Meteorology, Seattle, WA., Amer. Meteor. Soc., J2.5, https://ams.confex.com/ams/97Annual/webprogram/Manuscript/Paper301430/ARAM%202017.pdf.
MacGorman
,
D. R.
, and
W. D.
Rust
,
1998
: The Electrical Nature of Storms. Oxford University Press, 422 pp.
Maddox
,
R. A.
,
K. W.
Howard
, and
C. L.
Dempsey
,
1997
:
Intense convective storms with little or no lightning over central Arizona: A case of inadvertent weather modification?
J. Appl. Meteor.
,
36
,
302
314
, https://doi.org/10.1175/1520-0450(1997)036<0302:ICSWLO>2.0.CO;2.
Melnikov
,
V. M.
,
D. S.
Zrnić
,
R. M.
Rabin
, and
P.
Zhang
,
2008
:
Radar polarimetric signatures of fire plumes in Oklahoma
.
Geophys. Res. Lett.
,
35
,
L14815
, https://doi.org/10.1029/2008GL034311.
Moninger
,
W. R.
, and
R. A.
Kropfli
,
1987
:
A technique to measure entrainment in cloud by dual-polarization radar and chaff
.
J. Atmos. Oceanic Technol.
,
4
,
75
83
, https://doi.org/10.1175/1520-0426(1987)004<0075:ATTMEI>2.0.CO;2.
Murphy
,
T. A.
,
R. A.
Wade
, and
B. C.
Carcione
,
2016
:
Observations and operational considerations of the 4 June 2013 chaff event in northern Alabama
.
J. Oper. Meteor.
,
4
,
34
45
, http://dx.doi.org/10.15191/nwajom.2016.0403.
Ogawa
,
T.
,
1985
:
Fair-weather electricity
.
J. Geophys. Res.
,
90
,
5951
5960
, https://doi.org/10.1029/JD090iD04p05951.
Palermo
,
C. J.
, and
L. H.
Bauer
,
1965
:
Bistatic scattering cross section of chaff dipoles with application to communications
.
Proc. IEEE
,
53
,
1119
1121
, https://doi.org/10.1109/PROC.1965.4113.
Park
,
H. S.
,
A. V.
Ryzhkov
,
D. S.
Zrnić
, and
K.-E.
Kim
,
2009
:
The hydrometeor classification algorithm for the polarimetric WSR-88D: Description and application to an MCS
.
Wea. Forecasting
,
24
,
730
748
, https://doi.org/10.1175/2008WAF2222205.1.
Pinson
,
C. C.
,
1975
: Chaff cloud signature. Proc. Int. Symp. on Electromagnetic Compatibility, San Antonio, TX, IEEE, https://doi.org/10.1109/ISEMC.1975.7567810.
Pode
,
L.
,
1960
:
An approach to the study of automatic target tracking
.
Trans. Amer. Inst. Electr. Eng. Part 1
,
78
,
806
811
, https://doi.org/10.1109/TCE.1960.6368472.
Reinking
,
R. F.
, and
B. E.
Martner
,
1996
:
Feeder-cell ingestion of seeding aerosol from cloud base determined by tracking radar chaff
.
J. Appl. Meteor.
,
35
,
1402
1415
, https://doi.org/10.1175/1520-0450(1996)035<1402:FCIOSA>2.0.CO;2.
Roeder
,
W. P.
,
1995
: Operational impacts and identification of chaff on weather radar. Preprints, 27th Conf. on Radar Meteorology, Vail, CO, Amer. Meteor. Soc., 373–375.
Rowland
,
J. R.
,
1976
:
Clear air convective behavior revealed by radar chaff
.
J. Appl. Meteor.
,
15
,
521
526
, https://doi.org/10.1175/1520-0450(1976)015<0521:CACBRB>2.0.CO;2.
Rust
,
W. D.
, and
P. R.
Krehbiel
,
1977
:
Microwave radiometric detection of corona from chaff within thunderstorms
.
J. Geophys. Res.
,
82
,
3945
3950
, https://doi.org/10.1029/JC082i027p03945.
Ryzhkov
,
A. V.
,
T. J.
Schuur
,
D. W.
Burgess
,
P. L.
Heinselman
,
S. E.
Giangrande
, and
D. S.
Zrnić
,
2005
:
The Joint Polarization Experiment: Polarimetric rainfall measurements and hydrometeor classification
.
Bull. Amer. Meteor. Soc.
,
86
,
809
824
, https://doi.org/10.1175/BAMS-86-6-809.
Sauvageot
,
H.
,
G.
Lafon
, and
M.
Oruba
,
1982
:
Air motion measurements in and around thermal plumes using radar and chaff
.
J. Appl. Meteor.
,
21
,
656
665
, https://doi.org/10.1175/1520-0450(1982)021<0656:AMMIAA>2.0.CO;2.
Scharfenberg
,
K. A.
, and Coauthors
,
2005
:
The Joint Polarization Experiment: Polarimetric radar in forecasting and warning decision making
.
Wea. Forecasting
,
20
,
775
788
, https://doi.org/10.1175/WAF881.1.
Schurr
,
T.
,
A.
Ryzhkov
, and
P.
Heinselman
,
2003
: Observations and classification of echoes with the polarimetric WSR-88D radar. NOAA National Severe Storms Laboratory Tech. Rep., 45 pp, https://www.nssl.noaa.gov/publications/wsr88d_reports/JPOLE_Obs_and_Classification_of_Echoes_Report.pdf.
Stull
,
R. B.
,
1988
: An Introduction to Boundary Layer Meteorology. Kluwer Academic, 666 pp.
Weinheimer
,
A. J.
, and
A. A.
Few
,
1987
:
The electric field alignment of ice particles in thunderstorms
.
J. Geophys. Res.
,
92
,
14 833
14 844
, https://doi.org/10.1029/JD092iD12p14833.
WSR-88D Radar Operations Center, 2016: Interface control document for the product specification. Tech. Rep. 2620003W, 198 pp
, https://www.roc.noaa.gov/WSR88D/PublicDocs/NewTechnology/B17_2620003W_draft.pdf.
Yu
,
J.
,
H.
Lee
,
Y.
Jeong
, and
S.
Kim
,
2016
: Identifying chaff echoes in weather radar data using tree-initialized fuzzy rule-based classifier. Int. Conference on Fuzzy Systems, Vancouver, BC, Canada, IEEE, 2317–2324, https://doi.org/10.1109/FUZZ-IEEE.2016.7737982.
Zhang
,
G.
,
2016
: Weather Radar Polarimetry. CRC Press, 304 pp.
Zrnić
,
D. S.
, and
A. V.
Ryzhkov
,
1999
:
Polarimetry for weather surveillance radars
.
Bull. Amer. Meteor. Soc.
,
80
,
389
406
, https://doi.org/10.1175/1520-0477(1999)080<0389:PFWSR>2.0.CO;2.
Zrnić
,
D. S.
, and
A. V.
Ryzhkov
,
2004
:
Polarimetric properties of chaff
.
J. Atmos. Oceanic Technol.
,
21
,
1017
1024
, https://doi.org/10.1175/1520-0426(2004)021<1017:PPOC>2.0.CO;2.

Footnotes

© 2018 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).