High-Resolution Observations of Insects in the Atmospheric Boundary Layer

2176 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 25

ROBERT F. CONTRERAS AND STEPHEN J. FRASIER Microwave Remote Sensing Laboratory, University of Massachusetts—Amherst, Amherst, Massachusetts

(Manuscript received 5 September 2007, in final form 17 May 2008)

ABSTRACT

High spatial and temporal resolution S-band radar observations of insects in the atmospheric boundary layer (ABL) are described. The observations were acquired with a frequency-modulated continuous-wave

(FMCW) radar during the 2002 International H20 Project (IHOP_2002) held in Oklahoma in the months of May and June 2002. During the observational period the boundary layer was convective with a few periods of rain. Rayleigh scattering from particulate scatterers (i.e., insects) dominates the return; however, Bragg scattering from refractive index turbulence is also significant, especially at the top of the afternoon boundary layer. There is a strong diurnal signal in the insect backscatter: minima in the morning and at dusk and maxima at night and midafternoon. Insect number densities and radar cross sections (RCSs) are calculated. The RCS values range from less than 10Ϫ12 m2 to greater than 10Ϫ7 m2 and likewise have a strong diurnal signal. These are converted to equivalent reflectivity measurements that would be reported by typical meteorological radars. The majority of reflectivity measurements from particulate scatterers ranges from Ϫ30 to Ϫ5dBZ; however, intense point scatterers (Ͼ10 dBZ) are occasionally present. The results show that although insects provide useful targets for characterization of the clear-air ABL, the requirements for continuous monitoring of the boundary layer are specific to time of day and range from Ϫ20 dBZ in the morning to Ϫ10 to Ϫ5dBZ in the afternoon and nocturnal boundary layer (NBL).

1. Introduction sections in the range of 10Ϫ10–10Ϫ7 m2 at X band, whereas larger migratory insects exhibit cross sections Microwave radar observations of the clear-air (i.e., Ϫ greater than 10 5 m2. Observations by typical weather rain free) atmospheric boundary layer (ABL) are gen- radars, however, are unable to measure insect cross sec- erally accepted to exhibit two primary scattering tions, as single insects are rarely encountered within the sources. These are Bragg scattering from refractive in- large-resolution volumes of such systems. When insect dex turbulence and Rayleigh scattering from insects, population densities are sufficient, however, these birds, dust, or other airborne particles of sufficient di- weather radars do see insect echo as a volume scatter- ameter to be detected by radar. Bragg scattering coming target. Reported insect reflectivities vary widely, poses a substantial fraction of the backscatter for fre- with values often reported greater than 20 dBZ in con- quencies below 3 GHz, while Rayleigh scatter tends to vergence zones, convective rolls, and outflow bound- dominate for higher frequencies. Operational weather aries (Achtemeier 1991; Wilson et al. 1994; Russell and radars in the United States such as the Weather Sur- Wilson 1997; Lang et al. 2004). veillance Radar-1988 Doppler (WSR-88D) operate Over land, radar return from insects may be viewed near 3 GHz, where both scattering mechanisms are as either an undesired source of clutter, or as a desired, prevalent. albeit imperfect, tracer of the clear air. Achtemeier Radar cross sections of insects have been reviewed (1991) analyzed dual-polarized S-band observations to by Riley (1985) and Vaughn (1985). In general, small evaluate insects as a tracer of ABL motion. From the nonmigratory “weak flying” insects tend to have cross polarization ratio he inferred that insects were reori- enting themselves under certain conditions in response to air motion to avoid temperatures less than 10°–15°C. Corresponding author address: Robert Contreras, Microwave Remote Sensing Laboratory, University of Massachusetts— As a result they are not passive tracers during these Amherst, Amherst, MA 01003. periods, especially for the purpose of measuring verti- E-mail: [email protected] cal velocities. Vertical motion of insects has been ar-

DOI: 10.1175/2008JTECHA1059.1

JTECHA1059 DECEMBER 2008 CONTRERAS AND FRASIER 2177 gued by Angevine (1997) to contribute to observed vertical velocity biases in 915-MHz profilers. To evaluate the clear-air return of the WSR-88D radars, Wilson et al. (1994) analyzed observations from multiple radars located in different geographic regions and determined that insects are effective tracers of horizontal wind velocities during summer daylight hours. In addition, the authors showed that insects are useful in identifying microscale updrafts. Kusunoki (2002) observed insect echoes with C-band radar over the Kanto Plain in Ja- pan and found insects to be absent in air colder than 10°C. Zrnic´ and Ryzhkov (1998) and Lang et al. (2004) reported polarimetric signatures of insects with research S-band weather radars. Using an airborne verti- FIG. 1. The University of Massachusetts S-band FMCW radar cally pointing W-band radar, Geerts and Miao (2005b) deployed at the “Homestead” site during IHOP_2002. characterized insect scattering in the convective boundary layer (CBL) and argued that insects respond to (IHOP_2002) (Weckwerth et al. 2004). Similar to the sudden vertical motion and not to temperature. measurements of Richter et al. (1973) and others, the To investigate whether bird migrations contaminate radar measured clear-air backscatter from both re- clear-air wind estimates, Martin and Shapiro (2007) fractive index turbulence and from insects. The radar used S-, X-, and W-band radars and showed in two case often resolved individual insects that appeared as dis- studies that insects were the dominant source of return tinct “dot” echoes in displays of volume reflectivity. in the nocturnal boundary layer (NBL). Bachmann and Through fairly simple image processing techniques, it is Zrnic´ (2007) showed that even with contamination by possible to isolate such echoes from distributed Bragg migratory birds, useful wind and polarimetric informa- backscatter and consider their contributions to the tion from insects can be isolated using spectral analysis overall microwave echo. techniques. In section 2, the FMCW radar and its observations These studies have shown insects to be useful for are presented, as are the analysis techniques used to determining atmospheric motions in the absence of pre- isolate and characterize the insects. Once separated in- cipitation with the primary biases coming from migra- sect radar cross sections (RCSs) are calculated, the fre- tion and from insect response to vertical motion. The quency of occurrence of insects in the sampling volume range of equivalent radar reflectivity factor Ze cited in is determined, and assuming ergodicity, insect number the literature is large (Ϫ25 to 20 dBZ) and due to varia- densities are calculated. Following in section 3, the ver- tions in dominant insect size, type, and density. tical and diurnal distributions of insect RCS are quan- Recently, interest has grown in the development of tified, and the corresponding expected reflectivity fac- networks of short-wavelength (i.e., Յ3 cm) weather ra- tors (Ze) are calculated. We finish with a summary and dars to monitor conditions within the ABL where a discussion of radar requirements for continuous clear- longer-range radars lack sensitivity, owing to earth cur- air observations. vature, terrain blockage, and the large spacing between such radar systems (McLaughlin et al. 2007). The large 2. Methodology range of insect reflectivities, however, introduces un- certainty about the sensitivity requirements for these a. FMCW radar observations radars to monitor clear air. In this paper, we focus on Figure 1 shows the University of Massachusetts quantifying the clear-air echo of Rayleigh scattering in- FMCW radar originally described in I˙nce et al. (2003). sects in the CBL of the southern Great Plains. During 13 May–13 June 2002 the radar was located at To quantify scattering from insects, data are analyzed the “Homestead” site located approximately 12 miles from the University of Massachusetts S-band fre- east of the National Center for Atmospheric Research’s quency-modulated, continuous-wave (FMCW) radar. (NCAR) S-band dual-polarization Doppler radar (S- The radar was developed and is operated by the Uni- Pol) near Bryans Corner, Oklahoma. Several other pro- versity’s Microwave Remote Sensing Laboratory filing instruments were also located at this site including (MIRSL). It was deployed to the panhandle of Okla- a 915-MHz profiler, sodar, sounding systems, and sev- homa as part of the 2002 International H2O Project eral lidars.

Unauthenticated | Downloaded 09/27/21 04:03 AM UTC

Fig 1 live 4/C 2178 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 25

TABLE 1. FMCW system characteristics and sampling strategy. ary (clutter) echoes are subsequently removed in the signal processing. Frequency (f ) 2.94 GHz Range resolution (⌬R) 2.5 and 5.0 m Figure 2 shows the effective reflectivity factor Ze, which is defined as Range (Rmax) 2.5 and 5.0 km ␪ Beamwidth ( 1) 3.5° ␭4 Sampling frequency (fs)20Hz ϭ ␩ ͑ ͒ Ze 5 2 , 1 Averaging period (␶)5s ␲ |K| where ␭ is the radar wavelength and |K| 2 is a constant depending on the dielectric constant of the scattering Radar system characteristics for this deployment are medium (for water it is Ϸ0.9). The volume reflectivity ␩ summarized in Table 1. The radar operates at S band is defined as the radar cross section ␴ divided by the ␩ ϭ ␴ (10-cm wavelength), and during most of the experiment radar sampling volume V (i.e., /V). The factor Ze the radar was configured to operate with a vertical reso- is commonly used to characterize rain because it is pro- lution of 2.5 m and a frequency-sweep period of 50 ms. portional to the sixth power of the drop diameters. The

Individual vertical profiles of microwave echo were av- figure shows Ze for a 24-h period on 10 June 2002. The eraged over 100 sweeps to yield vertical profiles of re- abscissa is time [central daylight time (CDT ϩ 5 ϭ flectivity at 5-s intervals. The radar was periodically UTC)] and the ordinate is the altitude above ground operated with 5-m vertical resolution when the after- level (AGL). The figure shows the typical diurnal be- noon CBL grew particularly deep. The observational havior of clear-air return observed during the experi- period for the present analysis was from 13 May ment. During the morning the ABL is shallow (Ͻ500 through 13 June 2002. During this period there were m) with relatively low reflectivities. During the early two extended periods when the radar was not operated: afternoon, the depth of the ABL increases to 1.5–2.5 15–19 May and most of 30 May. Periods of rain were km, and the intensity of scattering increases. As can be excluded. seen by focusing on a 6-min period during the after- The radar is absolutely calibrated by means of an noon of 10 June (Fig. 3a), the increase in scattering is internal calibration loop in which a portion of the trans- due to both an increase in refractive index turbulence mitted signal is coupled into a surface acoustic wave and a greater quantity of insects in the convective (SAW) delay line and into the receiver prior to the boundary layer. low-noise amplifier. The resulting calibration signal ap- Our observations show that the number of insects in pears as a synthetic stationary target at the 1.5-km the afternoon boundary layer substantially increases range. The overall attenuation of this passive calibra- and is consistent with the observations of Geerts and tion loop is known, and so the signal can be equated to Miao (2005a) that show insect plumes to be collocated a reference reflectivity at that range. Drifts in transmit- with updrafts. At night the depth of the boundary layer ted power and/or receiver gain are reflected in varia- decreases and a secondary maximum in reflectivity oc- tions of this calibration signal. The synthetic target does curs due to nocturnal insects. Typically, there is strong not appear in the processed radar imagery, as station- return at the top of the CBL due to a combination of

FIG. 2. FMCW measured reflectivity for 10 Jun 2002. The ordinate is altitude AGL and the abscissa is CDT.

Unauthenticated | Downloaded 09/27/21 04:03 AM UTC

Fig 2 live 4/C DECEMBER 2008 CONTRERAS AND FRASIER 2179

FIG. 3. (a) Unfiltered volume reflectivity ␩ and (b) the pulse-to-pulse (or sweep to sweep) correlation coefficient. insects and refractive index turbulence, as well as The 1-dB threshold corresponds to two standard de- within insect layers in the NBL. viations above the average echo power for our averaged profiles of 100 samples. Pixels exceeding this local b. Signal processing threshold are retained, while all others are set to zero. To isolate the contribution of insects from the radar In regions of locally homogeneous Bragg backscatter, imagery, the following procedure is followed. First, a the probability of false alarm (identifying a pixel as 7 ϫ 7 median filter was applied over the two- containing an insect when it does not) is less than 2.5%. dimensional radar reflectivity image. This filter serves This assumes that the local averaged reflectivity esti- to remove isolated impulsive echoes that occupy fewer mate is described by a Gaussian random variable with a than half of the pixels within the filter’s interrogation normalized standard deviation, ␴/m ϭ 0.1, where m is window. The resulting median-filtered image is sub- the mean. tracted from the original to yield an image of positive Besides the morphological differences between in- (and negative) excursions about the median values. We sect dot echoes and distributed Bragg backscatter, the are interested only in the positive excursions that ex- two sources of scattering also exhibit differing Doppler ceed the median-filtered value by more than 1 dB. The characteristics. As described in I˙nce et al. (2003), fol- filtered and thresholded image is referred to as the in- lowing the FFT used for converting the detected fre- sect echo image. When this image is subtracted from quency domain signal to the time domain, the complex the original image, the result is the Bragg scatter image. (I and Q) samples corresponding to individual range bins The size of the median filter was chosen after some may be further processed for Doppler analysis in a man- experimentation on the reflectivity imagery. It was ner identical to that for pulsed radars. In this case the found that results were weakly dependent upon win- effective pulse repetition frequency is the sweep repeti- dow sizes between 5 ϫ 5 and 11 ϫ 11. The optimum tion frequency, which is 20 Hz for the data shown here. window size actually depends upon the density of insect This rather low pulse rate yields a narrow Nyquist velocity dot echoes. Ideally, the interrogation window always interval of only Ϯ0.5 m sϪ1. Measured vertical velocities includes at least one insect target. Small windows suffer are in many instances aliased; however, the spectrum from violating this condition frequently, while large width can still be interpreted. This is done via pulse- windows sacrifice spatial resolution. pair analysis as described in Doviak and Zrnic´ (1993).

Unauthenticated | Downloaded 09/27/21 04:03 AM UTC 2180 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 25

FIG. 4. (a) Unfiltered volume reflectivity ␩ due to both Rayleigh and Bragg scattering, (b) the separated Bragg component of the echo, and (c) the corresponding excursion image (i.e., the insect echo).

By analyzing the correlation coefficient between suc- of the interpretation of the scattering mechanism (i.e., cessive pulses (sweeps) averaged over the 5-s interval, Rayleigh or Bragg). The figure shows the filtering we find that the correlations for the isolated dot echoes method to be reasonably successful at separating the tend to be very nearly unity, while those for distributed scattering contributions. Shortcomings of this filter re- regions of backscatter tend to be significantly less than sult from the assumption that insect echoes occupy a unity (Fig. 3). The very high pulse-pair correlation for minority of pixels in the interrogation window. When the dot echoes is indicative of a narrow Doppler spec- insect densities become sufficiently large, this condition trum characteristic of a single, simple target, such as an may be violated, and insect echoes will appear as dis- individual insect. Distributed Bragg backscatter, which tributed scatter and will be removed. This occurs when is volume filling, produces a wider spectrum width be- the insect density is greater than about one insect in cause of multiple scattering centers within the pulse every two volume cells (or 1⁄2 V, where V is the radar volume. Very dense populations of insects will produce sampling volume). Additionally, when insect density is a similar signature (i.e., when multiple scatterers are greater than a few per resolution volume, the insects present). It is this evidence and the impulsive dot echo produce a signature that looks like Rayleigh fading. nature of the reflectivity field that supports the dot ech- Sharp boundaries of regions dominated by Bragg scat- oes as due to individual insect returns. tering may also be misidentified as due to insects when Figures 4a–cshow1hofunfilteredvolumereflectiv- the edge of a region occupies fewer than half of the ity ␩, the corresponding Bragg component of the echo, pixels within the interrogation window, although this and the corresponding insect component, respectively. occurs less frequently than the first effect. Volume reflectivity is used here because it is simply the To further illustrate the behavior of the filtering radar cross section per unit volume and is independent scheme, Fig. 5a shows an unfiltered volume reflectivity

Unauthenticated | Downloaded 09/27/21 04:03 AM UTC DECEMBER 2008 CONTRERAS AND FRASIER 2181

FIG. 5. (a) Unfiltered volume reflectivity ␩ image during the night of 10 Jun 2002. (b) Profile of mean ␩ over the same period. (c) Median-filtered insect ␩ image and the corresponding (d) mean profiles of unfiltered (black) and filtered (gray) ␩ over the period. image from 0314 to 0330 CDT 1 June 2002, and the ties. The latter description becomes meaningful only corresponding mean ␩ profile (Fig. 5b). During the when multiple scatterers are distributed throughout the time period, there was a particularly fine layer consist- radar’s resolution volume. When this is the case, the ing of numerous insects and also, possibly, a sharp re- volume reflectivity becomes independent of the radar’s fractive index gradient. Figure 5c shows the filtered im- resolution volume. In the present dataset, we isolate age for the same period and Fig. 5d compares the mean- individual insects or tightly clustered aggregates and as filtered (gray) and unfiltered (black) profiles. In this a result, “instantaneous” volume reflectivities are layer, the filtering procedure results in a decrease of the mostly a reflection of the small sampling volume. volume reflectivity by 5–6 dB. Outside the layer, the From the aforementioned insect echo images, we mean profile does not deviate much from the unfiltered first calculate the probability of encountering an insect mean profile. Thus, in cases of particularly dense insect within the radar scattering volume, P(Insect). This is densities, the filtering procedure tends to attenuate the simply the relative frequency of occurrence of an echo desired insect signal by a few decibels. within the time series for each height observation. Fig- It is important to reiterate that the motivation for this ure 6 shows the vertical profile of P(Insect) averaged research is to quantify the requisite radar sensitivity to over the entire observational period, regardless of hour. reliably measure the clear-air boundary layer. The pri- Overall, insects were encountered in roughly 10%–20% mary complication of our filter is that it underestimates of the volume cells with three distinct maxima evident echoes when insect densities are high and, therefore, in the profile: ϳ200, 1000, and 2100 m. the results presented here are conservative estimates. Ergodicity is necessarily assumed in this analysis such that the temporal average of detections within a small c. Analysis volume over 1 h is representative of the spatial average Since the identified insects are point targets, it is ap- in a substantially larger volume. The diurnal relative propriate to express their amplitudes in terms of radar frequency of encountering insects is shown in Fig. 7. cross section (RCS or ␴), rather than volume reflectivi- Here, P(Insect) has been composited by hour of the day

Unauthenticated | Downloaded 09/27/21 04:03 AM UTC 2182 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 25

over the observational period. The maximum relative frequency of about 0.25 occurs from the surface up to about 200 m AGL within the NBL and, as discussed, it may be an underestimate due to the appearance of dense nocturnal insect layers. However, the density of nocturnal insects is relatively high throughout the lower 1.5 km of the atmosphere; that is, they often extend well beyond the top of the NBL. The minimum probability of an insect occurs after sunrise until roughly 1000 CDT. A secondary maximum in the probability occurs in the afternoon and extends throughout the convective boundary layer. At about 1700 CDT, prior to the sun setting, convection subsides and the density of insects decreases. With nightfall the low-level nocturnal insect layer quickly develops. Given the known scattering volume with height, the insect number density, Nˆ ␷ , may then be estimated as

͑ ͒ P Insect Ni Nˆ ␷ ϭ ϭ , ͑2͒ V NV

where Ni is the number of insect detections and N is the total number of observations. The radar sampling vol- FIG. 6. Vertical profile of the probability of an insect in the ume, V, is given by radar sampling volume P(Insect) from 13 May to 13 Jun 2002 regardless of the hour of the day. 2␪ 2 R 1 V ϭ ⌬R , ͑3͒ 8 log2

FIG. 7. Diurnal composite of profiles of the probability of encountering an insect in FMCW sampling volume during IHOP_2002.

Unauthenticated | Downloaded 09/27/21 04:03 AM UTC DECEMBER 2008 CONTRERAS AND FRASIER 2183

FIG. 8. Diurnal composite of the number density of insects with S-band cross sections greater than 10Ϫ10 m2. The density is plotted as number per 104 m3.

where ⌬R is 2.5 or 5 m (depending on radar mode), R ␪ is range, and 1 is the half-power (one way) beamwidth of the radar, which is 3.5°. Figure 8 shows diurnal estimates of Nˆ ␷ for insects with an S-band RCS greater that 1 ϫ 10Ϫ10 m2 as the number of insects per 104 m3 (a cube 21.6 m per side) as in Riley (1992).

3. Results a. RCS: Gross and diurnal statistics For the pixels containing insect echoes, we calculate the distribution of observed insect radar cross sections. This is the conditional distribution of RCS values given that an insect is present, p(␴ԽInsect). The RCS value when no insect is present is, of course, zero. Figure 9a shows the distribution of insect cross sections as a function of height for all of the IHOP data. Based on the observed cross sections, the majority of insect backscatter is from “microinsects” with S-band RCS values from 1 ϫ 10Ϫ11 to 1 ϫ 10Ϫ9 m2. These cross-section measurements agree with those of smaller insects measured by Riley (1985) when one considers the RCS ratio of Rayleigh scattering insects at X and S bands,

FIG. 9. (a) PDF profiles of insect RCS measured during ␭ ϭ 3.2 Ϫ4 X Ϸ ͑ ͒ IHOP_2002, regardless of hour of the day. (b) Corresponding ͩ␭ ϭ ͪ 100. 4 S 10.2 cumulative distribution function profile (F).

Thus, X-band cross sections will exceed S-band cross sections by about 20 dB. Considering that the contours broader distributions of cross section are found below shown in Fig. 9a represent probability density functions 2000 m, and especially below 1200 m. These altitudes (PDFs) at each altitude with unit integrals over all RCS correspond to those in which there are substantial values, maxima above 2000 m indicate relatively nar- changes in insect densities over the typical day. Also row distributions of cross sections. It is notable that notable is the local maximum at about 200 m AGL,

Unauthenticated | Downloaded 09/27/21 04:03 AM UTC 2184 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 25

FIG. 10. (a)–(f) Profiles of cumulative distribution function F of RCS measured during the night [except for (f), which is morning] at the Homestead site in Oklahoma over the IHOP_2002 experiment: May and June 2002. which results from dense layers of nocturnal insects. Each panel in the figures corresponds to the indicated Below 200 m, the unexpected reduction in detected in- hour. At about 2300 CDT (Fig. 10b), a low-level maxi- sects is attributed to instrumental effects (clutter filter mum begins to develop at 200 m. As the night pro- cutoff and parallax of the antennas). gresses (Figs. 10c,d) the relative frequency of insects Figure 9b also shows profiles of F, the cumulative increases with roughly 10%–20% of RCS values greater distribution function of insect RCS. The figure shows than 1 ϫ 10Ϫ9 m2 at 0300 CDT. With sunrise, nocturnal that roughly 5%–10% of insects throughout the lower insects disappear and the frequency of occurrence de- 2.5 km of the atmosphere have S-band RCS values ex- creases by more than an order of magnitude. ceeding 1 ϫ 10Ϫ9 m2, which corresponds to 1 ϫ 10Ϫ7 m2 As shown in Figs. 11a,b, the number of insects below at X band, the largest of the “small” insects reported by 1500 m remains low in the morning. As the CBL grows, Riley (1985). starting late morning, insect densities increase (Figs. Figures 10 and 11 show RCS cumulative distribution 11e,d) and reach a maximum at about 1700 CDT (Fig. profiles for nighttime and daytime hours, respectively. 11e) between 500 and 1500 m. This maximum consists

Unauthenticated | Downloaded 09/27/21 04:03 AM UTC DECEMBER 2008 CONTRERAS AND FRASIER 2185

FIG. 11. (a)–(f) Profiles of cumulative distribution function F of RCS measured during daylight hours at the Homestead site in Oklahoma over the IHOP_2002 experiment: May and June 2002. (e) Densities of large insects reach a maximum at 1700 CDT from 500 to 1500 m AGL. of more than 10% of the measurements having cross which is the product of the number density and the Ϫ sections greater than 1 ϫ 10 9 m2, 5% greater than 1 ϫ average insect cross section. The expected reflectivity Ϫ8 2 ϫ Ϫ7 2 ͗ ͘ 10 m , and 1% greater than 1 10 m , which is in factor Ze calculated from Eq. (1) using the gross RCS the range of “large” insects reported by Riley (1985). distribution is shown in Fig. 12. The average profile of ͗ ͘ Ϫ Ze decreases from 7dBZ at 100 m AGL to about ͗ ͘ Ϫ ͗ ͘ b. Expected effective reflectivity factor Ze 22 dBZ at 2000 m. The Ze values should be the same for different wavelength radars as long as the scattering To convert the radar observations to volume reflec- is approximated by Rayleigh scattering (i.e., the wave- tivitys ␩, we calculate the average cross section (includ- length is much greater than the dimensions of the tar- ing all observations, many of which are zero) and divide get). by the radar resolution volume V. This is equivalent to Figure 13 shows the diurnal modulation of the average radar reflectivity factor. We find on average that ͗␩͘ ϭ ˆ ␴ ͑␴| ͒ ␴ ͑ ͒ N␷ ͵ p Insect d , 5 the maxima in reflectivity in the nocturnal boundary

Unauthenticated | Downloaded 09/27/21 04:03 AM UTC 2186 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 25

FIG. 13. Diurnal composite of expected effective reflectivity ͗ ͘ ͗ ͘ factor Ze during IHOP_2002; Ze is calculated using Eq. (1).

are notably smaller than those observed by Achtemeier (1991), Wilson et al. (1994), and others, who have fo- cused on episodic periods of intense clear-air features in the convective boundary layer where number densities may be especially high. The FMCW radar measurements during IHOP_2002 were made continuously from a fixed position and sampled “extreme” events FIG. 12. Vertical profile of the gross expected effective reflectivity only as they passed over the radar. The FMCW obser- ͗ ͘ ͗ ͘ factor Ze over IHOP_2002; Ze is calculated using Eq. (1). vations agree well with the coincident measurements of microinsects made by Geerts and Miao (2005b,a). Recent interest in the use of short-wavelength radar layer and in the afternoon convective boundary layer networks for monitoring of precipitation and winds in are 10–20 dBZ greater than the minima observed in the the ABL motivates the consideration of insects as trac- morning and at dusk. The maximum at night (the af- ers of the ABL winds. The analysis here suggests that ternoon) is confined to the lower 700 (lower 1100) m. the maximum sensitivity for continuous all-day monitoring of the ABL is approximately Ϫ25 to Ϫ20 dBZ, 4. Summary and discussion which is consistent with the sensitivity of most millimeter-wave cloud radars. However, targeted measure- The FMCW radar provides a unique view of the at- ments of the CBL in the afternoon (i.e., for monitoring mospheric boundary layer with very fine temporal and convective initiation) would require substantially less spatial resolutions. It is the fine space–time resolution sensitivity: Ϫ10 to 0 dBZ. Such sensitivity is not beyond that enables the segregation of point scatterers (in this the capabilities of short-wavelength meteorological ra- case, Rayleigh scattering insects) from distributed scat- dars. These results provide a framework for system de- terers and allows the study of insect number densities, sign and highlight the utility of making targeted obser- cross sections, vertical distributions, and average reflec- vations. tivity. Another important factor in determining the useful- The observed backscatter shows a strong diurnal sig- ness of clear-air X-band radar return is knowledge of nal with a low-level maximum in the NBL and another the presence, size, and spatial distribution of insects. maximum in the afternoon CBL between 500 and 1500 This is especially important because shorter-wave- m around 1700 CDT. RCS values at this maximum have length radars are insensitive to Bragg scattering turbu- values ranging from less than l ϫ l0Ϫ12 m2 to greater lence and, therefore, their utility depends upon there than l ϫ l0Ϫ7 m2 (or 0.4 mm2). These insects are “mi- being insects from which to scatter. As mentioned, in- croinsects” (Ͻ10 mm in diameter) and, when the wave- sects are absent over the ocean or when temperatures length of radiation is considered, are consistent with the are less that 10°C. Even if they are present, prior work cross-section measurements of Riley (1985). has shown a wide range of clear-air return (Ϫ25 to 20

However, the average reflectivity factor observations dBZ). As has been shown here, variations in Ze and

Unauthenticated | Downloaded 09/27/21 04:03 AM UTC DECEMBER 2008 CONTRERAS AND FRASIER 2187

RCS throughout the day can span many orders of mag- I˙nce, T., S. J. Frasier, A. Muschinski, and A. L. Pazmany, 2003: nitude, a range that calls for broader observations. An s-band frequency-modulated continuous-wave boundary Broader observations of insect backscatter would not layer profiler: Description and initial results. Radio Sci., 38, 1072, doi:10.1029/2002RS002753. only be of value to entomologists but would provide Kusunoki, K., 2002: A preliminary study of clear-air echo appear- radar scientists and engineers with guidance in devel- ances over the Kanto Plain in Japan from July to December oping operational radars capable of monitoring the 1997. J. Atmos. Oceanic Technol., 19, 1063–1072. clear-air boundary layer using insects. Lang, T., S. Rutledge, and J. Stith, 2004: Observations of quasi- symmetric echo patterns in clear air with the CSU–CHILL polarimetric radar. J. Atmos. Oceanic Technol., 21, 1182– Acknowledgments. The authors gratefully acknowl- 1189. edge the helpful comments of the anonymous review- Martin, W. J., and A. Shapiro, 2007: Discrimination of bird and ers. This work was supported primarily by the Engi- insect radar echoes in clear air using high-resolution radars. J. neering Research Centers Program of the National Sci- Atmos. Oceanic Technol., 24, 1215–1230. ence Foundation under Award 0313747 to the Center McLaughlin, D. J., E. A. Knapp, Y. Wang, and V. Chandrasekar, for Collaborative Adaptive Sensing of the Atmosphere 2007: Distributed weather radar using x-band active arrays. Proc. Radar Conf., Boston, MA, IEEE, 23–27, doi:10.1109/ (CASA). Any opinions, findings, and conclusions or RADAR.2007.374185. recommendations expressed in this material are those Richter, J. H., D. R. Jensen, V. R. Noonkester, J. B. Kreasky, of the authors and do not necessarily reflect those of M. W. Stimmann, and W. W. Wolf, 1973: Remote radar sens- the National Science Foundation. ing: Atmospheric structure and insects. Science, 180, 1176– 1178. Riley, J. R., 1985: Radar cross section of insects. Proc. IEEE, 73, REFERENCES 228–232. Achtemeier, G. L., 1991: The use of insects as tracers for “clear- ——, 1992: A millimetric radar to study the flight of small insects. air” boundary layer studies by Doppler radar. J. Atmos. Oce- Electron. Commun. Eng. J., 4, 43–48. anic Technol., 8, 746–765. Russell, R. W., and J. W. Wilson, 1997: Radar-observed “fine Angevine, W. M., 1997: Errors in mean vertical velocities mea- lines” in the optically clear boundary layer: Reflectivity con- sured by boundary layer wind profilers. J. Atmos. Oceanic tributions from aerial plankton and its predators. Bound.- Technol., 14, 565–569. Layer Meteor., 82, 235–262. Bachmann, S., and D. Zrnic´, 2007: Spectral density of polarimetric Vaughn, C. R., 1985: Birds and insects as radar targets: A review. variables separating biological scatterers in the VAD display. Proc. IEEE, 83, 205–227. J. Atmos. Oceanic Technol., 24, 1186–1198. Weckwerth, T. M., and Coauthors, 2004: An overview of the in- Doviak, R. J., and D. S. Zrnic´, 1993: Doppler Radar and Weather ternational H2O project (IHOP_2002) and some preliminary Observations. 2nd ed. Academic Press, 458 pp. highlights. Bull. Amer. Meteor. Soc., 85, 253–277. Geerts, B., and Q. Miao, 2005a: Airborne radar observations of Wilson, J. W., T. M. Weckwerth, J. Vivenkanandan, R. M. the flight behavior of small insects in the atmospheric con- Wakimoto, and R. W. Russell, 1994: Boundary layer clear-air vective boundary layer. Environ. Entomol., 34, 361–377. radar echoes: Origin of echoes and accuracy of derived winds. ——, and ——, 2005b: The use of millimeter Doppler radar ech- J. Atmos. Oceanic Technol., 11, 1184–1206. oes to estimate vertical air velocities in the fair-weather con- Zrnic´, D., and A. Ryzhkov, 1998: Observation of insects and birds vective boundary layer. J. Atmos. Oceanic Technol., 22, 225– with a polarimetric radar. IEEE Trans. Geosci. Remote Sens., 246. 36, 661–668.

Unauthenticated | Downloaded 09/27/21 04:03 AM UTC