at I... x
.ll..-.!§.‘:.. .
a. i

 

 

 

 

 

 

 

 

 

 

 

 

 

   

£5 Eaéfiz V V , ,

.

y

 

 

THE‘JiS

QOCM

 

Michigan Sir-4.x“
University

‘wfl "

« LIBRARY 1
J

 

This is to certify that the

thesis entitled

EVALUATION AND FIELD APPLICATION OF
AN OPTICAL SENSOR THAT DETECTS
FLYING INSECTS

presented by

Meghan Suzanne French

has been accepted towards fulfillment
of the requirements for

M. S . Entomology

degree in

Gym/M

Major professor

 

 

Date 8’/7’ @/

0.7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

PLACE IN RETURN BOX to remove this checkout from your record.

TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJCIRC/Dateouepes-p. 15

 

EVALUATION AND FIELD APPLICATION OF AN OPTICAL SENSOR THAT
DETECTS FLYING INSECTS

By

Meghan Suzanne French

AN ABSTRACT OF A THESIS

Submitted to
Michigan State University
In partial fulfillment of the requirements
For the degree of
MASTER OF SCIENCE
Entomology
2001

Dr. James R. Miller

ABSTRACT

EVALUATION AND FIELD APPLICATION OF AN OPTICAL SENSOR THAT
DETECTS FLYING INSECTS

By

Meghan Suzanne French

An optical sensor that was designed to remotely detect flying insects (OFIDIS;
www.qubitsystems.com) was evaluated outdoors under full sunlight. Shadows cast by an
artificial insect (an electric toothbrush oscillating at 53 Hz over a 120° angle), were
detectable up to 35 m away from the sensor: they weakened with distance at a rate of
y=902°'°'1 1" (r2=0.9 at p < 0.001). Shadowed signals were undetectable when sunlight
intensity was reduced below 20,000 lux. Sunlight signals reflected from a mirror
attached to the oscillating toothbrush remained detectable and strong (y=-0.24x + 86; r2 =
0.33) beyond 25 m. The ability of this sensor to detect differences in abundance, as well
as density in flying insect populations, was evaluated in three replicated site types: old-
field, wetland, and parking lot. Insect signal abundance (based on differing fundamental
wing-beat frequencies of each signal) at both old-field and wetland locations were higher
than that of the parking lot and control (F=8.85; df=8; P>0.0001). Signal richness was
greatest in the old-field (R=259) and least rich in the parking lot (R=88). However,
diversity was greatest at the wetland (H' = 2.26; H'max = 2.7), showing a high richness that
was most evenly distributed from 100-1000 Hz. Combined with an appropriate data
logger, this year 2000 sensor is a promising development for remote sensing of insects

with wing-beat frequencies over 100 Hz, notably Diptera and Hymenoptera.

Dedicated with love to my mother and father whose encouragement and friendship
continuously guide me through life’s greatest challenges.

iii

ACKNOWLEDGEMENTS

Special thanks to my major professor, Dr. James R. Miller, whose creative and
insightful approach to science has both guided and inspired me. Also, a very special
thank you to my supportive committee members: Dr. Frederick Stehr of Entomology and
Dr. Kelly Millenbah of Fisheries and Wildlife.

Thank you to the members of the Miller lab: Piera Giroux, Eric Hoffman, Alicia
King and our faithful student helpers for helping me crunch numbers and listening to
practice talks. Dr. Larry Gut and Peter McGhee get big thanks for the use of their lab’s
Macintosh computer and assistance. Insect identification help from Gary Parsons is also
greatly appreciated.

My project would not have been possible without the support of the sensor’s
inventor, Dr. Aubrey Moore and the folks at Qubit Systems, Ontario, Canada, who loaned
the OFIDIS sensor and software. Thank you for the chance to evaluate your product as
applied to my deep interest in animal conservation.

Thanks to everyone in the Department of Entomology for sticking by me through
all the bumps in the road!

Support for this project was provided in part by a student grant given by the

Hutson Endowment within the Department of Entomology at Michigan State University.

iv

TABLE OF CONTENTS

LIST OF TABLES .................................................. viii
LIST OF FIGURES .................................................. ix
CHAPTER 1 Introduction to wing-beat frequencies, harmonics, and their
application to insect detection ............................... 1
The phenomenon to be explored ...................... 1
Mechanisms of insect wing beat ..................... 2
Ranges in insect wing-beat frequencies ................. 5
Factors influencing wing-beat frequency ............... 7

Wave properties if insect wing beats: Harmonic content . . . 12
Developments in remote sensing of insects ............. 17
Measuring wing-beat frequencies ..................... 18

The potential application of OFIDIS relative to other
insect monitoring techniques ....................... 22

Objectives of thesis research ........................ 25
CHAPTER 2 Evaluation of an optical sensor and waveform analysis software

system for detection and possible taxonomic characterization of
flying insects using their wing-beat frequency and harmonics:

Response to controlled, artificial stimulation .................. 26
Introduction ................................................. 26
Materials and Methods ......................................... 28

Recording techniques .................................... 28
Shadowed signal intensity over distance ...................... 28
Reflective signal intensity over distance ...................... 29
Effects of light intensity on signal strength .................... 30

Harmonic patterns as influenced by the variation of stimulus

approach to the sensor ................................... 30
Testing T Rex harmonic measurement capabilities using
standard tuning forks .................................... 30
Effects of signal duration on TRex detection .................. 31
Results and Discussion ......................................... 31
Measuring shadow signal intensity over distance ............... 31
Measuring reflective signal intensity over distance ............. 32
Effects of light intensity on signal strength ................... 33

Harmonic patterns as influenced by the variation of stimulus
approach to the sensor ................................... 35

Testing TRex harmonic measurement capabilities using

standard tuning forks .................................... 38
Effects of signal duration on TRex detection ................. 38
Relevance of these results for OFIDIS use in the field .......... 42
CHAPTER 3 Field application of the OFIDIS sensor: Measuring flying insect
signal abundance and diversity in three habitat types ........... 43
Introduction ................................................. 43
Materials and Methods ......................................... 45
Study design ........................................... 45
Research sites .......................................... 46
Equipment set-up ....................................... 47
Counts of signals from three habitat types .................... 47
Wing-beat analysis of signals from three habitat types .......... 48
Signal diversity analysis .................................. 48

vi

Insect wing-beat signal capture for 33 collected insect specimens . . 50

Results and Discussion ......................................... 50
Wing-beat signals from the field ........................... 50
Counts of signals from three habitat types .................... 62
Diversity in wing-beat signals from three habitats ............. 65
Variation of pattern of insect signals from the field ............ 73

CHAPTER 4 Summary of thesis research outcomes and suggestions for

building upon the ......................................... 75
LITERATURE CITED ............................................... 79
APPENDD( 1 Record of deposition of voucher specimens ................... 84
APPENDIX 2 Site maps .............................................. 90

vii

LIST OF TABLES

Table 1. Signal richness (R), diversity (H') and relative diversity (J ')
compared for all three sites used in signal diversity study
analysis .................................................... 66

Table 2. Time of day each 10 min recording was made at each site
on each day used in the signal diversity study ....................... 71

viii

LIST OF FIGURES

Figure 1. Direct ((a), (b)) and indirect ((c), (d)) flight mechanisms.
Dragonfly thorax during (a) upstroke and (b) downstroke
of the wings. House-fly thorax during (c) upstroke and (d)
downstroke of wings. Stippled muscles are those contracting
in each illustration. (Modified from Gullen and Cranston, 1994
with requested permission from Chapman and Hall Publishing ........ 3

Figure 2. Wing best frequency ranges for common insect orders.
Information from Brodsky 1995, Chapman 1998, and
Dudley 2000 was combined to make this figure. Orders
marked with an * are synchronous, while the others are
asynchronous .............................................. 6

Figure 3. Wing beat frequencies as they relate to respective log body
mass (g) for some insect orders and one hummingbird family
(Trochilidae). This was compiled by Dudley 2000 and modified
with permission from Princeton University Press .................. 8

Figure 4. Representation of the sum of the possible standing waves in a
stretched string ............................................. 14

Figure 5. A complex wave (c) results from the summation of the standing
waves (a) (Harmonic 1) and (b) (Harmonic 2) that comprise it ....... 16

Figure 6. The OFIDIS sensor is 3 cm in height, 6 cm in diam, and
weighs 126 g. It is waterproof and very durable; electronics are
contained within a protective case. Connector 1 delivers the
sensor’s output relative to resting voltage, while connector 2
delivers absolute voltage ..................................... 21

Figure 7. Shadowing toothbrush signal as a fiinction of distance. Panel A
shows an individual record with a typical signal-to-noise ratio.
Panel B shows mean signal intensity for 10 replications ............ 32

Figure 8. Reflective toothbrush signal as a function of distance. Panel A
shows an individual record with a typical signal-to-noise ratio.
Panel B shows mean signal intensity for 10 replications ............ 34

Figure 9. Shadowed signal intensity (A), mean light intensity at the sensor

face as a function of density of cloth screens (B), and signal
intensity relative to the maximum detected for each cloth treatment. . . 36

ix

Figure 10.

Figure 11.

Figure 12.

Figure 13.

Figure 14.

Figure 15.

Figure 16.

Figure 17.

Figure 18.

Figure 19.

Signal intensities shown as bandwidth (a) and harmonic spectra (b)
of three tuning forks at identical fundamental frequencies.

(+) = examples of when the arms of a fork were horizontal vs.

(-) = arms vertical with respect to the face of the sensor. The
intervals between maximum intensity bands in the harmonic spectra

are equivalent to the fundamental frequency of each tuning fork ..... 38
Harmonic patterns for 3 separate signals made by a 430 Hz tuning
fork as analyzed and displayed by TRex ........................ 39

Middle C (523.23 Hz) displayed over time (top panel of each pair)
and harmonic spectra measured by TRex . (A) and (B) = ca. 0.5 s
and (C) and (D) = 1.0 s, respectively ........................... 41

Cluster analysis of the TRex measured harmonics for Middle C
signal (n=33) of two durations. The arrow indicates the major
split between clusters ..................................... . 42

Types of data displayed by Canary 1.2.4 sound analysis software,

used to analyze insect signals in this research. An insect signal in

(A) can be expanded to show its entire waveform (B) and subjected

(B) to spectral analysis to make a spectrogram (C) of the

(C) harmonics. The intervals between each harmonic band are

(D) equal to the fundamental fiequency of the analyzed signal ....... 49

The individual waveforms for each of the 33 insects caught and
identified. Identification of each insect was taken to the most
specific taxonomic name using available keys .................... 51

Mean insect signal counts (d: SE) per 10 min sampling period across

all sampling dates. No signals were detected when the sensor was
shielded by the opaque cover (control). Treatments bearing a

common letter are not different at p = 0.05 ...................... 63

The number of insects heard per 10 min recording as a fimction of
habitat type and sampling date during summer 2000 .............. 64

A frequency distribution of all wing-beat frequencies measured on
3 dates at Dansville old-field ................................. 67

A frequency distribution of all wing-beat frequencies measured on
3 dates at Rose Lake flooding ................................ 68

1'{:Tnnfi

Figure 20. A frequency distribution of all wing-beat frequencies measured on
3 dates at the Kmart® parking lot .............................. 69

Figure 21. Insect signals and their harmonics of individuals between 225-45 Hz

(A) and 545-75 Hz (B). Two similar looking harmonic patterns have
(B) been denoted (*). They may be of the same taxon .............. 74

xi

Chapter 1
Introduction to insect wing-beat frequencies, harmonics, and their

application to insect detection

1. The Phenomenon to be explored

A female mosquito buzzes near your head on approach for a blood meal. The
regular, high frequency beating of her tiny wings generates the necessary lift and thrust
for this oriented flight, now near completion. However, these diminutive creatures are
not always sufficiently stealthy to secure an uncontested meal. A by-product of even tiny
beating wings is compression waves in air that, if generated in proximity to the human
car, are sufficiently energetic to set one’s eardrum vibrating synchronously with her wing
beats. The resultant hi gh-pitched ‘Vvhine-of-the-mosquito” is unmistakable to the
experienced human host, causing us to cringe and swat in an attempt to avoid the
imminent bite.

The core question to be explored in this thesis extends from the above commonly
experienced phenomenon, in which, unintended cues emitted by the actions of one
organism can sometimes be exploited as useful information by another. That question is:
Do flying insects produce sufficient physical cues during flight to enable notable
advances in rapid and remote detection, quantification, and possible identification by
humans extending their natural sensory capabilities via available tools and techniques of

the modern electronic age?

to... mml
‘ —

ll. Mechanisms of insect wing heat

A. flight powered by direct flight muscles. Insect wings are comprised of a short
basal and longer distal shaft separated by a pivot point (Figure l). “Primitive” insect
orders like Odonata and Blattodea, contract elevator muscles attached directly to the
basal shaft (Figure 1) to raise their wings. However, some authors (Brodsky 1995;
Chapman 1998) suggest that indirect muscles are involved in wing elevation in some
Odonata. There is full agreement that direct, depressor muscles, attached to the distal
shaft just beyond the pivots, power the down-stroke. A series of smaller muscles (not
shown in Figure 1) also attached directly to the wings govern lift, speed, and steering by
regulating wing rotation, angle of attack, and exact stroke plane (Gullen and Cranston
1994; Dudley 2000). The up and down movement of a wing within the stroke plane is
known as a wing beat.

Direct flight muscles are characterized as synchronous -- in the sense that delivery
of at least one action potential to the neuro-muscular junction is necessary for every
contraction (Dudley 2000). Since there is an upper limit to action potential frequencies
(rarely sustained at > 100/s) and that only a few neurons innervate single muscle cells
(Chapman 1998), wings powered entirely by direct muscles are restricted by this neural
frequency limit.

The flight system of insects powered by direct flight muscles is often
characterized as a primitive system (Brodsky 1995) associated with clumsy fliers like
stoneflies, lacewings and roaches. However, under certain selective forces like predation,
direct flight muscles can yield outstanding flight speeds and maneuverability, e.g.,

dragonflies.

   
  

., '53:? . Dorso-ventral indirect
(6) flight muscles

    

muscle . = pivot Stemite

point Longitudinal indirect
flight muscles

    

Direct depressor
muscle

(b, ‘ Id)

Figure 1. Direct ((a), (b)) and indirect ((c), (d)) flight mechanisms. Dragonfly
thorax during (a) upstroke and (b) downstroke of the wings. House-fly thorax
during (c) upstroke and (d) downstroke of wings. Stippled muscles are those
contracting in each illustration. (Modified from Gullen and Cranston, 1994
with permission from Kluwer Publishing)

 

 

B. Flight powered by indirect flight muscles. Most flying insects power the

strokes of their wings by massive indirect rather than direct muscles (Chapman 1998).
Indirect flight muscles, attached to the walls of the thorax, are comprised of opposing
muscle groups that distort the thorax in such a way as to raise and lower the wings of an
insect by means of a pivot (Figure 1). Contractions of a dorso-ventral set of indirect
flight muscles anchored to stemites and tergites of the thorax cause the thorax to
compress dorso-ventrally and the wing to rise. Contractions of a set of longitudinal
indirect flight muscles anchored to the anterior and posterior thorax causes the thorax to
shorten longitudinally and the wings to move downward. Rather complex sets of non-
massive direct flight muscles imbedded in the thorax and attaching directly near the wing
base control such features as wing rotation, angle of attack, and realized stroke-plane. In
essence, the indirect flight muscles generate power for flight, while the direct flight
muscles govern how that power will get translated into lift, speed, and steering. In
actuality, the wing-beat strokes of most insects are better represented as a ‘figure 8 ”
rather than a simple up-down stroke.

The opposing indirect muscle sets are activated by action potentials delivered by
motor neurons under control of a neural central pattern generator located in the thoracic
ganglia (Chapman 1998). The action potentials are delivered to the respective neuro-
muscular junctions in a coordinated pace that causes only alternating and never
simultaneous contractions of opposing muscle groups. In fact, the complementary nature
of these opposing muscle groups is so well refined that the stretch received upon
contraction of either one of the pair is oflen a sufficient stimulus for immediate counter-

contraction of the other. In this sense, this type of muscle system acts as a mutual

 

oscillator that, once activated, can sustain its impetus for a burst of cycles before needing
to be reinforced by another action potential.

Such flight muscles capable of multiple contractions per arriving action potential
are termed asynchronous (Dudley 2000). A distinct advantage of this arrangement is that
such insects can generate oscillation rates well in excess of the maximum possible rate
for sustained delivery of action potentials (Brodsky 1995). Of course, this condition of
having rates of muscle contraction exceed the limits for neural signal delivery can
develop only concurrently with extraordinary designs for energy-efficient muscle
contractions and oxygen-delivery systems. Both of these conditions are exemplified by
the flight muscles of the higher insect Orders (Coleoptera, Hymenoptera, and Diptera)
exhibiting highly ordered muscle fiber substructures (actin, myosin, and their
attachments), an extraordinary number of mitochondria (Dudley 2000), and dense and
regularized distribution of trachea throughout these muscles (Chapman 1998). While
both the direct, synchronous and indirect, asynchronous flight mechanisms are found
within fliers of extraordinary speeds and agility (e.g., dragonflies vs. bees and flies
respectively), the indirect design is judged superior energetically (Dudley 2000).

III. Ranges in insect wing-beat frequencies

Insect wing-beat frequencies, usually expressed as cycles per second (Hz), vary
depending on such factors as wing morphology, body size, body mass, and neuro-
muscular mechanisms of power generation. Wing-beat frequencies for most insect
Orders fall below 100 Hz (Figure 2). Non-Dipteran, aquatic insects have relatively
small ranges, all falling below 100 Hz. Some Lepidoptera have the lowest wing beat

frequencies ever recorded, with a total range typically between 4 Hz

 

nae—82oz»? 2a 8050 BE? fine—552:? Em _.. an 53, vac—SE 82:0 .8st $5 838 3 3:358 33 coon
>295 wen .mma 58%ch .32 EmvSm Eat core—585 .8020 83.5 .8888 no.“ 8922 352.0% “won 955 .N oSwE

31:833.“. Loom mE>>
83 com com oov com com 2: o
\\ _ n _ F _ _
LN 8290
memfiocoEeflI
998200

 

 

 

 

.Eofiocamfi

 

- $28081
999:5:
EofioEoI
.Emfiocoth

.Eofiofimos.

«998002

.9030502

.Smfloooi

yam—oceans.

.mouozflm
.mEEoEEocam

I @9855

\\ .Smcono

\\ _ _ A _ _ _
ooo F com com oov com CON oo _. o

 

 

 

to 80 Hz. Only the Hymenoptera and Diptera markedly exceed the wing-beat
frequencies of all other orders. No Hymenoptera beat their wings slower than ca. 70 Hz
and most exceed 100 Hz, while midges can have wing beats as high as 900 per second.
Diptera may be able to achieve such high wing-beat frequencies because they only need
to energize two wings, while stabilizing themselves with halteres.

Wing-beat frequency does not necessarily determine the speed of insect flight.
For instance, fast flying dragonflies are able to use an array of direct flight muscles to
rotate their wings in such a way as to highly vary their flight speed and direction while
defending their territory (Chapman 1998). But, they do this at low (well below 100 Hz)
wing-beat frequencies. Butterflies also have a low wing-beat frequency and move at
lower speeds than the dragonfly. However, butterflies are tremendous gliders, allowing
them to conserve energy while in flight.
IV. Factors influencing wing-beat frequency

A. Body mass. Dudley (2000) established that body mass within various insect
Orders is generally inversely correlated with wing-beat frequencies (Figure 3).
Specifically, this pattern was observed by Casey et a1. (1985) for contained, hovering
euglossine bees (Euglaema-Eufriesea spp.). They exhibited a significant correlation
(-0.347:I: 0.048x + 1.992; r2 = 0.902; p< 0.0001) with a body mass range of 0.1 g to 1.5 g
and a respective, correlated wing-beat frequency range of 90 Hz to 240 Hz. Female
mosquitoes whose mass increased after a blood meal were reported to decrease their
wing-beat frequencies from 437 Hz to 433 Hz, or by approximately 2.4% (Ogawa and

Kanda 1986).

Log Wing-Beat Frequency (Hz)

1 1 I 1 1 OOO

 

2.5 a

 

2.0-
“ 100

Blattaria
Coleoptera
Diptera
Hemiptera
Heterocera
Homoptera
Hymenoptera

1.5“

  

Mecoptera

DIOX9I>OO

1O

Neuroptera

O Odonata

0 Orthoptera

D Rhopalocera

V Trochilidae

1 l

-5 -4 -3

 

 

 

Log Body Mass (9)

Figure 3. Wing-beat frequencies as they relate to respective log body mass (g )
for some insect orders and one hummingbird family (Trochilidae). Data were
compiled and plotted by Dudley 2000. Permission to reproduce this figure
with some modification was granted by Princeton University Press.

(2H) Aouenbar :1 meg-Bum

B. Wing size. An inverse relationship with wing-beat frequency holds for insects
with large vs. those with smaller wings. Lepidoptera and Odonata have relatively large
wings and lower wing-beat frequencies compared to other Orders. Even within a
particular Genus, there can be considerable variation. J 005 et a1. ( 1991) studying free-
flying bumblebees showed mean wing-beat frequencies of ca. 155 Hz and increased by

.030; r2 = 0.88). Increased wing

up to 13 Hz per 2 mm change in wing length (y=166.4x
length was likewise correlated with slower wing beats in tethered, anopheline mosquitoes
(Ogawa and Kanda 1986). Specifically, their mean wing-beat frequency ranged from 250
Hz at 5 mm wing length to 550 Hz at 2.0 mm wing length.

C. Maturation. Wing-beat frequency of a given species of hemi-metabolous
insect is reported to change with maturity. For example, Australian locusts increased
their wing-beat frequencies two-fold during the first few days as adults (Altman 1975).

In addition, wing-beat frequencies of 2.day old male mosquitoes were observed to be up
to 300% faster than those immediately eclosed, an increase from 200 Hz to 600 Hz
(Ogawa and Kanda 1986). This trend was also observed by Moore et a1. (1986), who
showed that male Aedes aegypti (Say) increased their wing-beat frequencies by up to 150
Hz within their second day as adults.

D. Type of behayior. Insect behaviors affect wing-beat frequency. As insects take
off, land, switch directions, hover and glide they modify the angle of their wings and in
some cases, wing-beat frequencies. In experiments with tethered vs. free-flying
Australian locusts, the free-fliers beat their wings 13% (3 Hz) faster (Baker et al. 1981).

A study by Spangler (1993) compared four behavioral types of the honeybee (Apis

mellifora L.) to show that tasks can affect mean wing-beat frequencies. Specifically,

5

Arizona honeybee robbers had a mean wing-beat frequency of 235 Hz, while Arizona
foragers had a mean of 220 Hz (2% difference). In a study that compared freely flying
foraging and hovering bumblebees (J 005 et al.1991), mean wing-beat frequencies for
both groups differed by a mean of 4 Hz (3%).

These studies suggest that wing-beat frequency may vary more'between castes in
the social insects than between individuals behaving similarly. Dudley (2000) believes
that wing beat frequencies of insects within an “unchanging behavioral context” remain
consistent. He reasons that insects exhibiting a given behavior will settle on a common
wing-beat frequency that represents an energy optimization.

E. Thoracic temperature. Physiological and environmental factors can influence

 

insect wing-beat frequencies. Studies relating thoracic temperature to wing beat
frequency show mixed effects. For instance, Ortelli (1989) reported that during take off
flight, the beetle Dineutus americanus (Coleoptera: Gyrinidae) decreased its wing-beat
frequency as its thoracic temperature increased (y = -1 .81i1.76x + 136.35 $8.88). But
this was not the case in the other beetles he studied; either there were no effects of
temperature or slight increases in wing-beat frequency were observed as thoracic
temperatures increased. The greatest effects of temperature were observed in individuals
of four Coleopteran families Scarabaeidae, Elateridae, Cantharidae, and Chrysomelidae,
suggesting that variation is not taxon specific. Interestingly 3 of the 4 families had the
highest mean wing-beat frequencies of all of the beetles tested and also showed high
ambient temperature sensitivity in their wing beats. Japanese beetle (Popilliajaponica
Newman) was most affected by an increase in thoracic temperature; as temperature

increased from 33-43 °C wing beats increased from 113—140 Hz.

10

 

Studies with bumblebees (J 005 et a1. 1991) showed no correlation between wing-
beat frequency and thoracic temperatures, while studies with dragonflies (T etrgonneuria
cynosura) (May 1981) showed that even after they wanned-up for flight, body
temperatures continued to increase and so did their mean wing-beat frequencies by up to
20 Hz (50% increase). Even during free flight, dragonflies can have wing-beat frequency
changes up to 10 Hz. These studies seem to suggest that consistent wing-beat
frequencies are more readily found among insects, such as bees, that use asynchronous
rather than synchronous muscles.

F. Air temperature. Ambient air temperatures also affect wing-beat frequencies of
insects. They have had significant effects on the mean wing-beat frequencies of Centris
pallida (Hymenoptera: Anthrophoridae); a 40 Hz decrease occurred down to 210 Hz
between 25° and 35° (Roberts et a1. 1998). The authors suggested that this decrease
might have been due to water loss from the insect in response to the higher temperatures,
and /or the decrease in metabolic rate.

Foraging, stingless bees, (T rigona jaty F. Smith), increased their mean wing-beat
frequencies from 180-300 Hz between 22° and 32° C (Unwin and Corbet 1984).
However, only slight changes in mean wing-beat frequency occurred among three groups
of flies studied: Calliphora vicina (5.4d: 0.04 Hz), Drosophila melanogaster (2.91 0.04
Hz), and Musca domestica (2.93: 0.04 Hz) (Unwin and Corbet 1984). The results agreed
well with the original measurements made of these flies by Soltavalta (1947).

In other studies with honeybees, little or no effect of air temperature was seen
among individuals (Spangler 1992), but mean wing-beat frequencies of free flyers did

comprise a range of 152-278 Hz. Spangler believes that the differences in wing-beat

ll

frequencies of freely flying honeybees resulted fiom fluctuations in thoracic temperatures
occuring when insects are thenno-regulating before and during flight. This information
again suggests that constancy in wing-beat frequency can be influenced by behavioral
state, particularly warm-up.

While it is important to understand the factors that may be influencing the wing-
beat frequencies of insects, the context of the studies should be noted. Insects that are
confined (Moore 1996) or tethered (Baker et al. 1981) or hand-held (May 1981) may be
behaviorally constrained, unable to reach and maintain the maximum wing-beat
frequency they typically can in nature. These studies tell us something about fluctuations
in wing-beat frequencies, but they all have constrained the insects in some way, making
the possibility of fluctuations in wing beat greater.

The greatest constancy is likely to be found among insects with the indirect flight
muscles system equilibrated by sustained flight. However, species identifications purely
on the basis of fundamental wing-beat frequency are judged unlikely.

V. Wave properties of insect wing beats: Harmonic content

The flexible body parts of a flying insect represent a medium that can express
complex wave properties under the influence of the strongly oscillating flight muscles in
the thorax. In this context, membranous wings can be thought of as flexible wands
anchored at the pivot (Figure 1), and the muscles powering wing beat can be considered
on-going generators of highly regular and energetic waves imparted to the wing shaft, as
well as surrounding body parts. The physics of this system can be approximated by more
familiar models like a human hand regularly imparting waves into a stretched rope whose

other end is fixed, or to the plucked string of a familiar musical instrument. Let us now

12

 

 

consider some of the most important properties of such resonating systems in preparation
for more complex analysis of insect wing beat that goes beyond simple fundamental
wing-beat frequency.

Different musical instruments (e.g., guitar vs. banjo) sounding the identical note
(e.g., middle C) vary in tonal quality as influenced by each instrument’s size, shape,
mass, and material composition. Each emission begins with the pluck of a stretched
string (Figure 4 A). Initially, the guitar and banjo string can be expected to vibrate with a
common fundamental frequency assignable to the condition where the full string
alternates between being either bowed up or bowed down (Figure 4 B). However, this
condition quickly becomes more complicated (enriched) as the wave propagating along
the string reaches an end and is reflected back into the string repeatedly. Following well-
established physical laws governing all types of waves (Berg and Stork 1995), a series of
standing waves arises exhibiting nodes precisely at 1/2, 1/3, 1/4, etc. the total length of
the vibrating string (Figure 4 C - E). The loops of string between these respective nodes
vibrate at even integer multiples of the fundamental frequency as set by the total number
of loops of one given length. Thus, a length of string between nodes of short spacing
vibrates at a much higher fiequency than that between distantly spaced nodes. The term
harmonic and a number reflecting the fold increases over the fundamental are assigned to
these discrete ascending steps in vibration frequency (Figure 4).

The collection of harmonics associated with a given fundamental vibration is also
referred to as its overtone series (Berg and Stork 1995). It turns out that the number and

relative intensity of harmonics varies for different instruments sounding the identical

13

A A - Stretched string
(resting position; 0 Hz)

 

 

. .t _, ,. 1loop (Ist harmonic;
8' C r > fundamental; X Hz)

 

 

, 2 loops (2nd harmonic;
C. . _. 0 2x Hz)

 

 

i .. 4‘ 3 loops (3rd harmonic;
D‘ i 0 \‘r 3x HZ)

 

4 loops (4th harmonic;
Em. 4x H2)

Figure 4. Representation of the lower harmonic series of possible standing waves
in a stretched string.

14

 

note. However, it is important to recognize that the fundamental vibration need not
always be the most intense, and that certain harmonics in an ascending series may appear
minor or absent. It is the relative patterns in harmonics that impart richness in tonal
quality to the human auditory system. Indeed, the overtone series for a guitar vs. banjo
sounding middle C are distinct. We listeners learn to associate unique patterns in
overtones to particular instruments or particular voices.

In actuality, the vibrations shown in Figure 4 B — E occur simultaneously. Thus,

the condition of the pure standing waves shown in Figure 4 would not be visible to the

 

careful observer of the plucked guitar or banjo string. The actual pattern in standing
waves along the length of string would be a summation even more complicated than the
simple summation shown in Figure 5 for a fundamental vibration (lSt harmonic) summed
with its 2nd harmonic. Fortunately, software for analyzing complex vibrations and
visualizing patterns in harmonics is now readily available (Canary® 1.2.4 at
http://birds.cornell.edu/BRP/SoundSoftware.html and TRex® 2.0 at
www.qubitsystems.com).

Harmonic theory can be applied to the waves generated by insect wings.
Knowing that insects differ in size, shape, mass and flight behaviors, we can expect
insects to differ somewhat in harmonic patterns of wing beat. The extent to which this is
true is partially revealed by the data of this thesis and of high importance to the future of

this type of research.

15

 

A A f\ f) A A fll
\/ \/ V \/ \/ \/

 

 

 

 

 

 

f\B/\ fl f 1 fl fl fl ll f l A
\f \f if \f \f l/ \f \f \/ \

 

 

 

 

 

 

 

C
i ure 5. A complex wave (C) results from the summation of the standing
onic l) and B) (Harrnoni rise it.

ves A) (Harm

VII. Developments in remote sensing of insects

Remote sensing may be defined as the collection of information about an object
without being in physical contact with the object (Sabins 1987). Sabins further stated
that the term remote sensing is restricted to methods that employ electromagnetic energy
as the means of detecting and measuring target characteristics. Based on this definition,
it seems plausible that anything from cameras to radar detectors used in the history of
entomology are indeed remote sensing instruments.

Entomologists have historically relied on aerial photographs and satellite imagery
to assess insect damage to crops and forests (Riley 1989). Distance photographs allow
researchers to observe changes over vast geographic areas that may not be as easily
summated by local measurement of damage. However, there may be disadvantages to
this more sophisticated methodology, including the financial and time costs of gathering
the data. Up-to-date satellite images of specific areas on earth are not always easily
available and usually come with a cost to the researcher.

Videography is commonly used to capture insect behavior as it applies to insect
movement. Aphid flight patterns have been monitored using camera techniques (Hardie
and Young 1997) and three-dimensional moth flight trajectories have been mapped for
fliers too low to the ground for accurate radar detection (Riley et a1. 1990). The response
of drones to queen bee pheromone has also been captured by videography (Loper et al.
1993).

Application of radar systems to the monitoring and detection of insect biota has
met with considerable success. Radar techniques have included Doppler, harmonic,

tracking, and millimetric-wavelength radar (Riley 1989). Radar systems have been used

17

 

at a macro-level to track the movement of migrating locusts (Riley 1980 and Schafer
1976 cited by Riley 1989) and to ascertain insect flight migration trajectories (Hobbs
1989; Rui-Lu et a1. 1989). Used in localized experiments, radar systems have detected
and documented moths following a pheromone plume (Riley et a1. 1998) when these
behaviors could not be captured on video because of the long flight distances involved.
Radar has also been used to remotely sense ground beetle movement (Mascanzoni and
Wallin 1986). Radar offers many benefits for monitoring insects, such as the ability to
collect and record real-time data. In addition, radar is thought not to harm the insect or
modify its behavior during a study. Unfortunately, the high cost of using radar reduces
the practicality of this tool for daily or seasonal insect monitoring.

VIII. Measuring wing-beat frequencies

Insect wing beats were first catalogued and published by Sotavalta (1947).

Amazingly, this researcher measured wing-beat frequencies over 100 insects by simply
listening to them and assigning frequencies based on his perfect pitch (Sotalvalta 1947).
This is an uncommon ability. More typically, quantifying wing-beat frequency has more
recently relied on microphones placed closely to the insects of interest (e.g., Ogawa and
Kanda 1986; Roberts et al. 1998). The recorded sounds are analyzed through an
oscilloscope or sound-analysis computer software. Microphones coupled to a system for
amplifying and recording signal are cheap and portable tools for researchers. However,
microphones can detect sounds from insect wings only when positioned within a short
distance (often less than 1 m). Intensity of the weak acoustical signal diminishes with the
square of distance. Therefore, microphone use has been limited to typically experiments

where the target insects are contained within small cages. Other interesting techniques

18

for recording insect wing beats included the use of high-speed film photography (Baker
1981) and electromyographic (EMG) recordings (Foster and Robertson 1992). Both of
which require researchers to be present during the study.

Detection of insect activity (including flight, but not wing-beat frequency) was
also made easier with the advent of electro-static actographs (Backlind and Ekeroot 1950
as cited by Berry 1972; Edwards 1960; Berry 1973) and later versions using infrared
(Eaton 1980) and radar (Buchan and Moreton 1981; Schaefer and Bent 1994). These
technologies are beneficial in that they can run in the researcher’s absence. More
esoterically, Macauley (1974) was able to use thermal gradients disturbed by flying
moths to detect insect flight. While these types of sensors were used only to measure
activity and not wing-beat frequencies, their automated data collection and circuitry
design became foundational in the development of newer, optical sensors.

Unwin and Ellington invented an optical tachometer in 1979. This device
detected and amplified minute changes in light intensity across the surface of a photo-
diode as modulated by the movement of the body parts of flying insects even decimeters
away from the sensor. From the optical tachometer signal, fundamental wing-beat
frequencies of insects could be extracted by oscilloscopic analysis.

More specifically, the device was comprised of a lens and a photo-diode
integrated into an amplified circuit board. The unit was battery powered and had a
simple electrical analogue output recorded over time on devices such as a tape recorder.
The system was used to detect wing-beat frequencies of bees and flies in Costa Rica
(Unwin and Corbet 1984) as well as various beetles during Oertli’s (1989)

thermoregulation work.

19

“fig

Dr. Aubrey Moore of Ottawa, Ontario, Canada has, over the last 15 years,
substantially improved optical tachometer technology. This research yielded the Optical
Flying Insect Detection and Identification System (OFIDIS) now manufactured and
distributed by Qubit Systems® (www.qubitsystems.com), of Kingston, Ontario, Canada
for use as a remote sensing, pest-monitoring tool for field use. The system has two main
components: 1) An optical sensor (Figure 6) with imbedded photocell inside, and 2) wave
analysis software called TRex 2.0. The sensor’s photocell is able to detect very rapid
changes in light intensity (beyond 1000 Hz) caused by insect wing beats in the presence
of sunlight or an artificial light source. Insects can either diminish light falling on the
sensor by shadowing it, or increase the light on the sensor’s surface by reflecting light
onto it. The millivolt electrical changes in the sensor circuitry can be amplified and
recorded as raw electrical signal using a digital recording device, such as a mini-disc
player. The individual signals can then be analyzed for their fundamental frequencies as
well as harmonic content. Dr. Moore postulates that harmonic patterns may be taxon-
specific, in which case a neural network database could be trained to recognize insects of
interest.

Moore’s initial studies using the precursor to the OFIDIS sensor were conducted
on mosquitoes (Moore et al. 1986) and aphids (Moore and Miller unpublished data). The
mosquito studies showed that the sensor and its analysis software were capable of
distinguishing between Aedes aegypti (L.) and A. triseriatus Say 84% of the time when
analysis was conducted only on the fundamental fi'equencies (also called primary
frequency or 1St harmonic). When analysis was done on the subsequent 4 harmonics of

each mosquito’s signal, 82% accuracy was achieved.

20

 

Figure 6. The OFIDIS sensor is 3 cm high, 6 cm in diam, and weighs
126 g. It is waterproof and very durable; electronics are contained
within a protective epoxy. Connector 1 delivers the sensor’s output
relative to a resting voltage, while connector 2 delivers absolute volt-
age.

21

During aphid studies in the laboratory, individual frequencies of 4 species
overlapped, but their means were significantly different (p = 0.05) from each other. This
outcome suggested that there is variation of wing beat frequencies within a species and
that using only wing-beat frequency to distinguish individual insects will not suffice for
identification. Moore suggested that harmonic content of the wing-beat frequency might
provide more individualized characterization for more accurate identification of that
insect. However, considerable work remained to be done on the OFIDIS system. For
example, the active space of the sensor was not yet characterized nor was the system
broadly tested in the field for recording a wide range of insects. Beginning in year 2000,
Dr. Moore and Qubit Systems sought collaborators to assist in this research.

IX. The potential application of OFIDIS relative to other insect monitoring
techniques

Insect detection and monitoring are fundamental to pest management and
conservation programs addressing habitat quality. Insect populations can rise and fall
dynamically within and between seasons due to many biotic (e. g., predator density) and
abiotic (e. g., degree day accumulation) factors (Southwood 1978; Samways 1994). This
reality can make insect detection and monitoring a challenge, especially since the patterns
associated with population fluctuations are not known for all species in need of control or
protection. An insect detection device that features real-time data collection, over a large
area, and that is portable and cheaper than radar or more detailed than aerial photographs
would benefit management programs.

A. Pest management. Flying insects can be detected using sticky traps with or

without lures, light traps, sweep netting, suction devices and through indirect means, such

22

 

as parasitoid emergence (Pedigo 1996). Unfortunately, initial detection and/or reaching
the action potential of a pest population usually warrant insecticide application and the
long-term monitoring necessary to delay spraying is halted (Dent 2000). This is often the
case even if natural control of the pest were possible. Stinner (1983) asserted that using
light, pheromone, and suction traps for early detection of pests is problematic because of
insufficient studies converting trap catch into accurate density estimates.

More recently, studies have been conducted to measure the reliability and
accuracy of trapping systems deployed in greenhouse and cropping systems. Heinz et al.
(1992) studied the time-efficiency use of sticky traps in greenhouses. They saw that a
major problem with trap deployment was not only the cost of the traps themselves, but
also the labor needed to sort and identify the arthropods caught. This unique study
established a specific number of traps, that when deployed, would give a reliable estimate
of pest density that could be used for pest management decisions.

Trap design must be conducive for collecting the highest number of insects of
interest within specific crops. Knodel and Agnello (1990) determined that non-sticky
pheromone traps were better than the sticky variety within New York orchards.
Leafhoppers, which are very quick, erratic flyers, are better collected through suction
trapping than through a beating tray (Henns et al. 1990). Yet for slower moving insects,
like lady beetles, visual inspection/ scouting was judged the most efficient sampling
technique so long as time was not a factor (Michels and Behle 1992). In addition, abiotic
factors such as temperature and light intensity will affect catch density. For instance, pea
aphids and their predators are most efficiently caught by sweep netting at late morning

hours (Schotzko and O’Keefe 1989).

23

I now join Dr. Moore in suggesting that several OFIDIS sensors placed within a
crop or orchard could serve a grower well by gathering real-time data about number and
possible type of flying insects. This optical sensor could be used in conjunction with
traps and scouting protocols to enhance monitoring activities. With an added light
source, like a laser beam, night monitoring could be achieved.

B. Insects as indicators of habitat gualig. Insects are regularly used as indicators
of habitat quality. Tiger and ground beetles, as well as butterflies, ants and flies play
important indicator roles in terrestrial habitat quality studies, while mayfly larvae may be
sampled and compared during water quality studies (New 1995). For the terrestrial
studies of insect indicators, or even for those conservationists trying to monitor for
protection, the OFIDIS sensor could prove useful. The OFIDIS sensor could be applied
to insect density and abundance studies within many land and aquatic environments.

Some conservationists have described and ranked the insects that would be good
indicators in certain regions of the world. Brown (1997) and Brazilian colleagues created
a list of valuable insect indicators of environmental changes in the Brazilian Atlantic
Forests. This list includes flying insect families and subfamilies from Hymenoptera
(Meliponinae, Euglossinae), Lepidoptera (Nymphalidae, Lycaenidae, Saturniidae,
Arctiidae), Diptera (Muscidae, T ephritidae, Phlebotominiae, Bibionidae), Coleoptera
(Elateridae, Cerambycidae, Cassidinae), Hemiptera (Pentatomidae), Homoptera
(Menbracidae), and many odonates. He further mentions that using an array of insect
indicators, like those mentioned above would better indicate the extent of disturbance.
Monitoring a collective group of insects within one region would avoid drawing

conclusions about habitat quality from monitoring one taxon. But, this can be viewed as

24

so time-consuming as to be judged impractical. The OF IDS sensor could contribute
valuable multi-species data when deployed in fields and waterways throughout an
ecological region being evaluated for animal diversity and habitat quality. Moreover, this
electromagnetically based system has considerable potential to operate remotely.

X. Objectives of thesis research. Study objectives were to: 1) Characterize the OFIDIS
sensor’s active space when operating under sunlight, and 2) Determine whether this
system capturing optical signals from insect wing beats could yield useful measures of
insect abundance and diversity in a few types of natural and disturbed areas in central

Michigan.

25

Chapter 2
Evaluation of an optical sensor and waveform analysis software system for detection
and taxonomic characterization of flying insects using wing-beat frequency and

harmonics: Response to controlled, artificial stimulation

Introduction

Insect detection and identification is fundamental to any conservation monitoring
program or pest management plan involving insects (Samways 1994; Pedigo 1996). For
example, the Optical Flying Insect Detection and Identification System (OFIDIS),
designed by A. Moore (Moore 1998) and marketed by Qubit Systems
(www.qubitsystems.com) of Kingston, Ontario, Canada, was created to detect and, in-so-
far-as possible, identify flying insects quickly and remotely by using their wing-beat
frequencies and wing-beat harmonic content as identification characters.

The OFIDIS sensor’s design builds on the optical tachometer invented by Unwin
and Ellington (1979), in which rapid, minute changes in light intensity across a photo-
diode increase or depress electrical signal output. This tachometer was used in studies
measuring how wing-beat frequency of bees and flies increased with ambient air
temperatures (Unwin and Corbet 1984) and how increased thoracic temperature in some
beetles caused an increase in wing-beat frequencies (Oertli 1989).

The OFIDIS sensor (Figure 6, chapter 1) includes a photocell that detects rapid
light changes across its surface caused by beating insect wings and moving body parts.
The photocell and additional circuitry are contained within in a 3 cm thick, 6 cm diam,

weatherproof case that collectively weighs 126 g. Electrical signals, created by an

26

insect’s shadow or reflection of light onto the photocell of the sensor, are detected as a
change in voltage. These electrical data can be stored on a recording device, such as a
tape recorder or mini-disc player. OF IDIS has accompanying waveform analysis
software (TRex 2.0) that displays waveforms and harmonic spectra from captured
signals. However, other waveform analysis programs, such as Canary 1.2.4, a well-
established program developed by the Bioacoustics Research Program at the Cornell
Laboratory of Ornithology, also display waveforms and harmonics of captured signals in
alternative forms.

Initial studies using an OFIDIS predecessor, designed by Moore, were conducted
on mosquitoes (Moore et al. 1986) and aphids (Moore, unpublished data). The sensor
and its analysis software were capable of distinguishing between laboratory-held Aedes
aegypti (L.) and A. triseriatus (Say) 82% of the time. Analysis was conducted on the
fundamental frequency (also called primary frequency or 1"t harmonic) and the 4
consecutive harmonics from each mosquito’s record. In studies of aphids in the
laboratory, individual frequencies of 5 species overlapped, but their means were different
from each other (P<0.05). These studies suggest that there may be sufficient variation of
wing-beat frequency and harmonics to distinguish among some taxa. Moore concluded
that using only the primary wing beat frequency (1St harmonic) of an insect is a less
discriminatory characteristic compared to using both the insect’s primary wing beat
frequency as well as additional harmonics.

Both the mosquitoes and the aphids used in the above studies were held within
flight arenas under artificial lighting. These conditions may have modified their flight

behavior by forcing more take-offs and landings and wing buzzing rather than

27

equilibrated free flight. Insects that are wanned-up and flying in their natural
environment might have less variation in wing-beat frequency and harmonics.

If particular insect taxa generate unique harmonic patterns with their wing beats,
it is possible that some could be identified and counted remotely, using the OFIDIS
system (Moore, per. comm). Recently, OFIDIS was offered to collaborators worldwide
for testing. Objectives of this research were to: 1) characterize active space for shadowed
and reflective stimuli detected by the OFIDIS sensor operating under sunlight, 2)
determine how light level affected signal intensity, and 3) evaluate the utility of TRex 2.0

and other wave analysis software in displaying the signals recorded from the sensor.

Material and Methods

Recording techniques. All data were recorded on a mini-disc player (Sony®
MZ-R3 7, Japan), using 74 min, Memorex® mini-discs. The sensor was connected to the
mini-disc player by an RCA cable (Radio Shack® 278-962). The mini-disc player’s
lineout was directly connected to the input of a Gateway Performance PC or Power Mac
with a 16 bits or higher soundcard. Light intensity readings were made with a Lutron®
Lux 100 photometer (Lutron Electronics, Taipei, Taiwan).

Various wave-analysis programs are available to researchers evaluating biological
sounds and electrical data. The OFIDIS wave-analysis program, TRex 2.0, has been
evolving concurrently with this research. Therefore, it was important to use additional
software to provide contrasting measurements of the data collected. Canary 1.2.4 was
chosen to accompany this research and provide a comparison to TRex.

Shadowed signal intensity over distance. A l x l m mirror was used to redirect

sunlight horizontally along a 50 m sidewalk before it struck the shaded wall of a building.

28

The OF IDIS sensor was placed in the sun-lit spot on the wall with its face perpendicular
to the redirected sunbeam. A helper maneuvered an oscillating (53 Hz at a 120° angle)
Braun® electric toothbrush (Type 4726, Kronberg, Germany) to which was affixed a
blackened 1 cm square mirror chip between the sensor and the mirror, breaking the beam
of light at l m intervals progressing either toward or away from the sensor. Mean
intensity for each signal (10 reps) at each distance was measured by Canary 1.2.4
software.

Reflective signal intensity over distance. In preliminary use of the OF IDIS
sensor in the field, I discovered that insect signals could still be detected even when the
sensor was facing away from the sun on a cloudless day. Although the inventor and
distributor have emphasized shadowing as the predominant mechanism of signal
generation, it became apparent that insects are capable of reflecting enough light onto the
sensor for their detection. This remained true even when the sensor was already fully
illuminated by sunlight.

All tests characterizing reflective signal were conducted after 2 pm when the sun
was not directly overhead. A 25 m rope was first stretched along the ground. The
OF IDIS sensor was placed at one end of this transect with its face perpendicular to the
ground and facing away from the sun. A hand-held, oscillating, electric toothbrush with
an attached 1 cm2 mirror was held 0.5 m above the ground and maneuvered to reflect
sunlight until it struck the sensor as sensed by the investigators via auditory output from a
small, amplified speaker (Radio Shack 277-1008C). This manipulation was done at
every 1 m interval while moving along the 25 m transect. Undoubtedly, insect body parts

are not perfect mirrors. However, measuring the response of the OFIDIS system under

29

this controlled and reproducible activation offered insights into how the system is likely
detecting insect signals when shadowing is impossible. Mean intensity of signal (10
reps) over distance was quantified using Canary 1.2.4 software. Data were expressed on
a relative scale (0-100%) relative to maximum signal.

Effects of light intensity on signal strength. The investigator stood on the
ground holding the OFIDIS sensor directly toward the sun in late moming on a cloudless
day. A helper stood on top of a ladder and placed embroidery hoops (25 cm in diameter)
holding one to six layers of polyester, shear cloth (20 intersections per cm) between the
sun and the oscillating, electric toothbrush held approximately 2 m from the sensor. Light
intensities adjacent to the sensor were recorded for each cloth density along with signal
intensity of the toothbrush operating only in the shadowed mode.

Harmonic patterns as influenced by the variation of stimulus approach to the
sensor. Tuning forks of 430, 440 and 523 Hz were struck and placed over the sensor in
fiill sunlight. Each fork was first displayed with both vibrating prongs shadowing the
sensor and then twisted so one prong was eclipsing the other over the sensor. Five
repetitions per position were recorded and each tuning fork signal’s harmonic spectrum
was analyzed and displayed using Canary 1.2.4.

Testing TRex harmonic measurement capabilities using standard tuning
forks. The sensor was placed face-up on the floor in a darkened room. An illuminated
laser pointer was securely placed 1 m above the face of the sensor. The laser pointer was
used because of lack of clear skies for several days in a row at the time of this test.
Tuning forks of frequencies 430, 440 and 523.25 Hz were individually struck and placed

in the laser beam. Five repetitions per tuning fork were recorded. TRex 2.0 analyzed

30

each recorded signal, displaying harmonics for each tuning-fork signal that were then
visually compared.

Effects of signal duration on TRex detection. The music synthesizer within
Creative Sound Studio® software was set to piano mode. Middle C (523 Hz) was played
at both 0.5 and 1 5 intervals and recorded as wave files on the computer using Creative®
recording options. This note was used because it matched the frequency of one of the
previously used tuning forks. Resultant wave files were played through TRex 2.0 to 3V
generate waveforms and harmonic spectra for each middle C piece of varying duration. i

The TRex waveforms were visually compared and their subsequent harmonic spectra

 

were compared through cluster analysis of Teach Me Data Analysis® (Lohniger 1999) to
determine if differences in harmonic patterns resulting from both the 0.5 and l s signals

would be separated into two distinct clusters.

Results and Discussion

Measuring shadow signal intensity over distance. Shadowed signals were
always most intense when the stimulus was immediately in front of the sensor. The
negative exponential decrease in signal intensity with distance from the sensor (Figure 7)
was to be expected for a small object casting a shadow on a target surface lit by sunlight.
The explanation is well known (Minnaert 1993) and involves the non-linear narrowing of
the umbra (darker part of the shadow) and the swelling of the penumbra, (lighter parts of
the shadow) as the eclipsing object is moved from the target toward the sun.

While shadowed signals from the toothbrush were sometimes detectable up to 45

m away both by the analysis software and the human ear upon acoustical playback, 30 m

31

 

 

Relative Signal Intensity
o

 

 

 

 

 

 

 

 

 

 

 

 

Mean Relative Signal Intensity (‘70) +/- SE

 

VIIIIIIIIIIIIIIUIIUIIIIIITIIIIIIIIII

0 5 1O 1 5 20 25 30 35 40
Distance (rn)

Figure 7. Shadowing toothbrush signal as a function of distance. Panel A
shows an individual record with a typical signal-to-noise ratio. Panel B shows
mean signal intensity for 10 replications.

32

} Noise

or greater signal was most often amidst equal or greater amplitude of noise caused by
atmospheric haze and wind. Without a filter option to eliminate this noise, shadowed
signals beyond 30 m cannot be reliably analyzed by TRex 2.0 or Canary 1.2.4.
Nevertheless, the OFIDIS sensor impressively has a much greater distance of detection
than microphones used in previous insect flight studies (Ogawa and Kanda 1986; Roberts
et a1. 1998).

Additionally, I noticed that weaker, distant signals produced fewer harmonics.
More testing of how strong a detected signal must be in order to allow a wave analysis
program to produce usable harmonic patterns is needed. As of this writing, Dr. Moore
has established a minimum threshold of signal intensity of 500 bits for TRex to analyze a
signal and display harmonics (Moore per. corn.)

Measuring reflective signal intensity over distance. Reflective signal from the
mirrored toothbrush diminished considerably less per unit distance than did shadowed
signal (Figures 7 and 8). A shallow and negatively linear slope was documented when
mean signal intensity was regessed on distance. The decrease over the first 2 m did
appear to be sharper than that at geater distances.

This modest decrease in signal intensity with distance is undoubtedly related to
modest decrease in light intensity reflected from a perfect mirror over distance. This
potential benefit in the insect sampling context needs to be contrasted against a lower
probability of hitting the sensor with light with increasing mirror distances. At distances
10 m and geater it often took the mirrored-tootlnbrush handler up to 30-60 s to hit the

sensor and sustain a signal, whereas it only took a few seconds when the handler was

33

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

A
Q‘
(I)
E
s
E O a 1 Al -IJL I
a W 7 n '
0
.2
5
0
a:
("ti 100
4 90- B
4.
gs 80-
.5 70-
§ 50 - y=-g.24x+86
a? 40-
§ 30-
g
a 20-
§ 10-
2 0 IIII I I [*Iil fiI I I I I I I I I I I ITfiI I I I I I I I I
0 5 10 15 20 25 30

Distance (m)

Figure 8. Reflective toothbrush signal as a function of distance. Panel A shows an
individual record with a typical signal-to-noise ratio. Panel B shows mean signal
intensity for 10 replications.

34

only 1 m away. These experiences suggested that although the signal strength of sunlight
may remain very strong over considerable distances, the probability of reflected light
striking the sensor would decrease appreciably with distance. Preliminary trials in which
the mirror was positioned off the transect line showed that the active space for reflective
signals is a broad cone rather than a cylinder. Although incomplete, this research has
established that combined active space for reflected plus shadowed signals is large
enough to make insect detection by the OFIDIS sensor a common rather than uncommon
occurrence.

Further research is needed to determine the exact distance limits for insect
detection by this sensor and the probability of detecting different-sized insects within this
limit. Such details will allow a user of the sensor to make more informed decisions about
which insects to measure (e. g., migatory aphids perhaps 100 m above the sensor or
pollinating bees moving between plants in a crop).

Effects of light intensity on signal strength. Shadowed signal intensity
decreased considerably once two cloth layers were placed between the sensor and the
toothbrush stimulus (Figure 9). As shown by information combined in panels (B) and
(C), a minimum sunlight intensity of 20,000 lux was necessary for the OFIDIS sensor to
detect a shadow-produced toothbrush signal from ca. 3 m. Cloud cover or other
conditions dimming light intensity will decidedly reduce the OFIDIS active space. Full
sunlight is a requirement for maximal detection of shadow-generated signals. Although
not formally tested here, decreasing light intensity will likely similarly diminish reflected

insect signals.

35

 

o
l

     

Relative Shadowed
Signal lntensnty

* lcloth * 2cloths " 3cloths "‘ 4cloths

 

 

 

 

 

 

 

 

 

 

 

 

60
gt” 50 - B
‘0 (I)
5 4
E i 40 n
E § 30
on- _
'3 O
o
c o 20 ‘
(U ‘— v: .......... E .......... § ....... T:
g 25, 10 -
0
~03 4 80 - C
a) +
> ’3‘ 60 -
z: o\
(u v
'3") 3., 40‘
a ‘Z’ 201
§ .9
C fl . f - . 4 . c : c
2. _ 0

Full 1cloth 2 cloth 3 cloth 4 cloth 5 cloth 6 cloth solid
Sun

Cloth Treatments

Figure 9. Shadowed signal intensity (A), mean light intensity at the sensor face as a
function of density of cloth screens (B), and signal intensity relative to the maximum
detected for each clotln treatments (C). Dashed line in (B) shows light intensity due
to diffuse sunlight indirectly striking the light meter from sources below the clotln.

36

Harmonic patterns as influenced by variation of stimulus approach to the sensor.
Each tuning fork signal became markedly less intense when the prongs were shifted from
horizontal to vertical with respect to the sensor face (Figure 10). The harmonics
produced by each tuning fork varied in their intensity and clarity in direct correspondence
with overall signal intensity (Figure 10). However, whenever there was sufficient signal
strength (e. g. 523 Hz) to produce harmonic spectra, the harmonic pattern was conserved,
i.e. in no cases were there reversals in intensity of harmonics. These results must be
taken with caution, but nevertheless, they support the idea that signals with identical
fundamental frequency can produce the same harmonic pattern.

Testing TRex harmonic measurement capabilities using standard tuning
forks. TRex displayed very similar harmonic patterns for each tuning fork that
produced strong enough signal (Figure 11). However, TRex declined analyzing up to
50% of the signals produced due to weak or clipped signal (those exceeding an amplitude
limit). At times, I could hear a clear signal upon acoustical playback, but TRex did not
detect it for unexplained reasons.

It appears that there is a trade-off inherent in TRex: fewer signals are worthy of
analysis, but those that are sufficiently intense, produce measurable harmonics. This
information is important for future training of a neural network capable of identifying a
particular insect signal within multiple signals (Moore 1998). Identifying specific insects
of interest will be the strength of OFIDIS, especially if the same consistencies of
harmonic patterns for a given taxon are observed and a reliable database is deveIOped in

which to compare field-acquired signals.

37

 

 

392 Hz

 

 

 

 

b 523 Hz + +

 

 

 

Time (s) after tuning fork was struck

Figure 10. Signal intensities shown as bandwidth (a) and harmonic spectra
(b) of three tuning forks vibrating at identical fundamental frequencies.
(+) = examples of when the arms of a fork were horizontal vs. (-) = arms
vertical with respect to the face of the sensor. The intervals between
maximum intensity of bands in the harmonic spectra are equivalent to the
fundamental frequency of each tuning fork.

38

Amplitude Relative to 1st Harmonic as 100%

100

100

100

lit 2r'nd 3r|d

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

l I j I I I 1

0 1000 2000 3000

Harmonic Spectrum (Hz)

Figure 11. Harmonic patterns for 3 separate signals made by a 430 Hz
tuning fork as analyzed and displayed by TRex software.

39

Effects of signal duration on TRex detection. Signal duration influenced the
harmonic spectra produced by TRex. When comparing harmonic spectra produced by
0.5 s signals versus 1.0 s signals (Figure 12), slight differences in the amplitude of each
harmonic distribution are apparent. These slight differences in harmonic amplitude
caused an almost perfect division of the short and long signals into two major
classification goups (Figure 13) when analyzed via cluster analysis using full frequency
distributions of harmonics (Figure 12). This outcome demonstrates that signals generated
by the same source can be over analyzed so as to detect differences due to subtleties in
behavior. Developers and users of OFIDIS need to guard against attributing all variations
in signal to authentic differences in signal generators.

Relevance of these results for OFIDIS use in the field. As reported herein, the
active space of the OFIDIS sensor is sufficiently large to detect many different flying
insects in varied environments under full sunlight conditions. Shadowing insects can be
detected up to 30 m away. However, the active space for capturing reflective insect
signals has not been officially delimited. As a safety margin against missing flying insect
signals, the OFIDIS sensor should be treated as having an unbounded active space for
reflective signals. For accurate comparative measures of insect abundance in differing

sites, the sensor should be pointed at open sky under full sunlight.

40

 

596858 a _ n R: as Q a no .8 u Ev Ea A3 .65 3 85808
88on 35:5:— 28 38 some he 6:3 not 0:5 .550 3.523% 22.me 98E GE 33 D 2sz .2 Raw:—

 

 

l‘

:4 4i .1 I J.— 1...:

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

41

 

 

 

 

 

 

 

 

 

eases , w

<

 

@8620 50.33 E% SE: 65 82865
Beta 2:. accuses—u 025 qummHE flea—ma U 2252 .5.“ 8808.3: 3568:. x33. 2.: .«o mam—ecu .5620 .2 oSwE

 

 

 

 

 

 

 

 

 

 

 

 

 

00m
A:

00m

 

 

42

Chapter 3
Field application of the OFIDIS sensor: Measuring flying insect signal abundance

and diversity in three field-site types

Introduction

Insect conservation is an important and on-going part of conservation biology and
endangered species protection. According to Pyle et al. (1981), the geatest cause of
insect extinction is the destruction of necessary habitat. The United States Endangered
Species Act (ESA) currently legally protects over twenty endangered and threatened
insect species (U SFWS 2000). Most of these insects are found in the Western USA states
of California and Oregon, which have undergone tremendous human population gowth
and land use changes since ESA inception in 1973. In addition, worldwide invertebrate
protection is broadly supported through the activities of the World Conservation Union
(IUCN), World Wildlife Fund, and various other conservation goups.

Critical to the process of protecting BSA-listed species and adding new species to
the threatened or endangered status list is acquisition of obtainable and reliable
population data. Without proof that a species is declining, the US. Department of the
Irnterior is unlikely to endorse protection (Houck 1993). Among the animal species
needing protection are some insects.

Insect populations fluctuate from year to year, ranging from heavy declines to
outbreaks (Samways 1994). For instance, the threatened California Bay checkerspot
butterfly (Euphydryas editha bayensis) has many isolated populations, all of which
fluctuate dramatically from year to year (Ehrlich and Murphy 1987). In these cases, it is

often unclear to researchers and wildlife managers as to why the insect population is

43

fluctuating and whether or not a decline is temporary or indicative of a potential loss of
the species. Unlike mammal studies, where data regarding gender ratios and dispersal
patterns are often well defined for managers devising an appropriate conservation plan,
insect studies often lack such resolution.

Insects serve as important environmental indicators of habitat quality. Steytler
and Samways (1995) examined the larval and adult Odonata colonization of a man-made,
urban lake in South Africa as a means of monitoring the lake’s water quality. After the
lake was constructed, Odonata species richness in the area doubled from 12 species to 26
species. These same authors (1996) used Odonata as indicators of habitat disturbance by
examining species richness along a river flowing through a forest, residential area, park
and city. Species richness was highest in the park and lowest in the forest, which had
been disturbed by commercial harvesting.

Collecting and identifying insect indicators is time consuming. Pearson and
Cassola’s (1992) use of tiger beetle species as indicators of regions high in biodiversity
relied heavily on accurate human observation. At their site in Peru, it took 50
observational hours to find 93% of the tiger beetles krnown to live there. In addition,
Integated Pest Management practices are often not implemented in place of traditional
pesticide control because of the pest scouting costs (both time and labor) to gowers
(Wearing 1988). Hence, there is a need for more rapid assessment of insect richness and
population. Oliver and Beattie (1993) and others are testing the methodology of rapid
assessment of biodiversity (estimates of species richness of insects using recognizable
taxonomic units (RTUs) or a morpho-species approach rather than actual identified insect

species) to help conservation agencies survey habitats faster. It is currently thought that

the inclusion of invertebrates in habitat assessment is excessively costly in botln time and
labor. New technologies capable of remote detection, identification, and counting of
insects might enhance such assessments.

In the current research, I deployed the OF IDIS sensor into three field-site types
(wetland, old-field, and parking lot) in an attempt to measure relative abundance and
diversity of flying insects. This research represents the first steps in determining: 1)
whether or not the OFIDIS sensor is a reliable tool for measuring flying insect activity
under a set of standard weather conditions, and 2) the extent to which captured insect
signals vary in fundamental frequency.

Materials and Methods

Study design. Research was conducted between July 7 and September 6 of
Summer 2000 at three field-site types: wetland, old-field, and parking lot. Each site type
was replicated three times for a total of 9 sites. Because insect activity was expected to
vary with temperature, humidity, and perhaps wind conditions, data collection and
analysis were blocked by day in the sense that each site was sampled on a given day and
replicates were accumulated over days. Order of visitations was randomized; but certain
sites were paired to consolidate driving time. Another restriction in this randomized
complete block experiment was that full sunlight was required for at least 9 of the 10
sampling min at each site. Past work on the OF IDIS sensor (Chapter 2) established that
its active space was proportional to light intensity. This limitation was costly, as
Michigan skies rarely proved cloud-free for the full 6 hours needed to complete one full
block of sampling. Nevertheless, seven days of insect signal data were collected and

analyzed.

45

Research sites. Locations were selected for their site type and were all in the
state of Michigan, within 25 mi of Michigan State University (See Appendix 2 for the
maps of the sites and their approximate areas).

The Rose Lake flooding site (42° 48.775 N, 84° 21.501 W; Shiawassee Co., MI)
is a wetland surrounded by forest. It is within the protected Rose Lake Wildlife Research
Area. Swimming, fishing, or boating is not commonly practiced in this body of water,
making it the most pristine of the wetland sites used in this study. Hewes Lake (42°
31.445 N, 84° 20.272 W; Ingham Co., MI) is located within the Dansville State Game
Area, which is surrounded by forest. Boating, fishing and swimming are permitted here.
Central Park pond (42° 43.795 N, 84° 25.310 W; Ingham Co., MI) is situated within a
park in a suburban setting and highly maintained by mowing regimes. People do fish
here, but the pond is too small for boating; swimming is not allowed.

The Rose Lake old-field site (42° 48.775 N, 84° 21.501 W; Clinton Co., MI) is
also within the Rose Lake Wildlife Research Area. This once-cultivated land is
undergoing secondary succession and is populated by a large variety of plants. It is the
largest old-field in this study. The Dansville old-field site (42° 48.770 N, 84° 23.450 W)
within the Dansville State Game Area is surrounded by forest. Legg Park old-field (42°
41.665 N, 84° 22.949 W; Ingham Co., MI) is surrounded by forest, railroads, and a gavel
road leading visitors to a gassy picnic area and forest trails.

The Kmart parking lot (42° 31.445 N, 84° 20.272 W; Ingham Co., MI) was the
largest of all the parking lots used in this study. Residential areas and businesses
surround it. Hannah Plaza parking lot (42° 43.165 N, 84° 27.646 W; Ingham Co., MI)

had the most car traffic activity of the lots. Manly Miles parking lot (42° 43.007 N, 84°

46

27.519 W; Ingham Co., MI) was the least busy of the lots, but did contain the most trees
and most abundant vegetative border.

Equipment set-up. The OFIFIS sensor (Figure 6) was fastened to the top of a l
m high geen garden stake that could be held vertical by insertion into the gound or a
supportive stand at parking lots. The sensor always faced directly upwards toward the
sunlit sky. Signals at each site were recorded on a mini-disc player (Sony® MZ-R3 7,
Japan) for 10 min. During recordings, I stood several meters away from the sensor so
that my possible attraction of deer flies and other insects did not affect the measurements.
Cloud cover for a total of more than 1 of the 10 min disqualified a sample. Wind speed
and local temperature at the initiation of each recording were measured with a combined
anemometer/temperature probe (Omega; Manchester, UK), while a Lutron® Lux 100
photometer (Lutron Electronics, Taipei, Taiwan) was used to measure light intensity.

Counts of signals from the three site types. I listened to each 10-min recording
played through speakers (Boston® Digital BA735) connected to a Gateway®
Performance PC and counted the number of audible insect signals. Count means :t SE
were calculated and gaphed in Excel; significance between means was tested by Proc
glm and means were separated by Least Means Square (SAS Institute, 2000). A Multiple
Linear Regession (R-Square Selection Method, SAS Institute 2000) was run to establish
the extent to which abiotic factors (date, time-of-day, wind, site type, location,
temperature, wind speed, and light intensity) affected count data.

Wing-beat analysis of signals from three site types. Signals collected from the
Rose Lake flooding, Dansville old-field, and the Kmart parking lot on the dates of July

20, August 16, and September 6, 2000 were used for this analysis. Dates were chosen to

47

achieve representative coverage of the 3-month field season. Each lO-min recording was
analyzed by Canary 1.2.4 sound analysis software (Cornell Bioacoustics Lab, Ithaca,
NY)) for fundamental frequency and harmonic content. The harmonic spectra for each
signal were printed and the fundamental frequencies of the wing-beat signals were
calculated by hand because no software was found that could perform these calculations
with desired precision (Figure 14). Calculations of fundamental frequencies consisted of
using a high-precision metric ruler to measure the mean interval between harmonics and
equating this distance to Hz in accordance with the y-axis scaling provided by the
software. In some records where the scatter plot of harmonics were too compressed or
smudged for measurements, the waveform (Figure 14B) was further expanded and the
fundamental frequency assigned by measurement scaled to the calibrated x-axis.
Precision of these fundamental wing-beat frequency measurements was quantified by
having the two researchers taking measurements each re-measure 20 signals randomly
selected within the range of 100-1000 Hz.

Signal diversity analysis. Frequency distributions of the wing-beat frequencies
from each site location were made so that signal richness, diversity (using the Shannon-
Weaver diversity index), and relative diversity could be compared between sites (Zar
1999)

Insect wing-beat signal capture for 33 collected insect specimens. Insects
from diverse flying taxa were collected from site locations and the Michigan State
University campus by sweep net. Individuals were placed in either a 1 x 1 m screened

cage with the sensor positioned pointing downward from the center of the top of or in a

48

Relative Intensity

Relative Intensity

Relative Intensity

 

 

 

 

 

 

*
- . - l A _ -‘l- -
o a- are i . —-— T-
B . Expanded individual
insect signal designated
by“

 

O
n
l .
|
l
l

 

 

 

 

 

 

 

i 0* |
c l ; . .
l . . :
0‘ g . . ‘ l I ' l
: ~ ., t ‘
'E l; J : § ' z ‘.
I :. : f . Ij. :
I a" i f ‘ 3" u ‘1‘ i
I
0 10 20
Time (mS)

Figure 14. Types of data displayed by Canary 1.2.4 sound analysis software, used
to analyze insect signals in this research. An insect signal in (A) can be expanded
to show its entire wave form (B) and subjected to spectral analysis to make a
spectrogam (C) of the harmonics. The intervals between each harmonic band are
equal to the fundamental frequency of the analyzed signal.

49

10 cm diam, clear container placed on top of the sensor. Insect signals were recorded in
full sunlight and the precise time they interacted with the sensor was noted for accurate
linking of signal to given specimen. Individual signals were processed by the Canary
1.2.4 sound analysis progarn to generate a printed waveform used to measure the
fundamental wing-beat frequencies, and, where possible, a spectrum of the harmonics.
The 33 insects were frozen and then preserved according to the Michigan State
University entomology museum procedures (Gary Parsons personal communication) and
submitted to the museum as voucher specimens.
Results and Discussion

Wing-beat signals from the field. A sample of the waveforms for each of the 33
collected insects is displayed in Figure 15, arranged in ascending wing-beat frequency.
Diptera and Hymenoptera were the most prevalent insect Orders collected, and their
wing-beat frequencies (Figure 15) fell well within the expected ranges for these Orders
(Figure 1). Harmonic spectra are not included here, as many of the signals did not yield
distinguishable harmonics under captive conditions. The reason(s) for this absence are
unkrnown and a bit disappointing, since I was able to see very rich harmonic spectra for
signals recorded during the abundance/diversity study and in the tuning fork studies
(Figure 10, Chapter 2). So far, insects with lower fundamental frequencies produce
harnnonics that can be undetectable. Therefore to obtain a fundamental wing-beat
frequency, calculations must be based on expanded oscilloscopic waveform.

Nevertheless, data begin to establish a database for wing-beat waveform and

50

 

 

 

 

 

 

 

 

 

 

 

 

52‘
g Coleoptera: Cerambycidae Typocerus velutinus (Olivier) 12 Hz
5
2 ° "
.9
In
0
>
23
E 1
O
a: 0 75 150
Time (m8)
9?
g Lepidoptera: Lymantriidae Lymanm'a dispar (Linnaeus) 24 Hz
5
i o -
C ”WV“ .
.9 f U u l M I
(I)
Q .
>
33
fl
3 I
n: 0 600 1200
Time (m8)
.é‘
0
g Neuroptera: Chrysopidae 28 Hz
.5
2 ° ‘
.9
a).
0
>
3
a .
m 0 125 250
Time (m8)

Figure 15. The individual waveforms for each of the 33 insects caught
and identified. Identification of each insect was taken to the most spe-
cific taxonomic name with the possible keys available.

51

 

Odonata: coenagrionidae 37 Hz

0 -W-IWMMMW

 

 

 

Relative Signal Intensity

I
0 100 200

 

 

 

 

 

 

 

 

Time (m8)

5‘

g Lepidoptera: Pvralidae Paraponvx sp. 41 Hz

g twin/WW. AM

3 o - r

5., fl W Ulfvl

to

e

>

a

3 I

[‘3’ o .5 1
Time (m8)

5."

g Diptera: Tabanidae Chrysops sp. 42 Hz

5

i o -

.s‘.» . "n

"’ l

o

>

'5

.L' .

é’ 0 so 160
Time (m8)

Figure 15. Cont’d.

52

 

Lepidoptera: Pterophoridae Platyptilia sp. 46 Hz

WMWMWW-W\W

 

 

 

o 350 700
Time (mS)

Relative Signal Intensity

 

Odonata: Coenagrionidae 58 Hz

O'Wx

 

 

 

0 so 120
Time (mS)

Relative Signal Intensity

 

° inwwm‘wwmww

0 80 160
Time (mS)

 

 

 

Relative Signal Intensity

Figure 15. Cont’d.

53

Relative Signal Intensity

Relative Signal Intensity

Relative Signal Intensity

 

Hemiptera: Miridae Lygus sp. 101 Hz

W4 nWNNN-WW

l

 

 

 

I
0 125 250

 

Time (mS)
Diptera: Sciomyzidae Tetanocera sp. 133 Hz

*WWMMJWWW»

 

 

 

I

0 50 100
Time (mS)

 

Coleptera:Curculionidae Apion centrale Fall 137 Hz

 

 

 

I

0 50 100
Time (mS)
Figure 15. Cont’d.

54

Relative Signal Intensity

Relative Signal Intensity

Relative Signal Intensity

-M1VWWWWWWWW

 

Hymenoptera: Apidae Bombus sp. 137 Hz

 

 

 

I
0 125 250

Time (mS)

 

Hemiptera: Miridae Lygus sp. 149 Hz

 

 

 

I
0 125 250

 

 

 

 

Time (m8)
”IIMMAMIWWIMMWMW
o Titian: (m8) 120

Figure 15. Cont’d.

55

Relative Signal Intensity

Relative Signal Intensity

Relative Signal Intensity

 

Diptera: Svrphidae

 

Platvchierus sp. 175 Hz

 

 

 

 

o MWWWWWWW

I
225 450
Time (mS)
Diptera: Syrphidae Toxomerus marginatus (Say) 175 Hz

 

 

I
150 300

 

OHWWWMMWW

 

Time (mS)
Hymenoptera: Halictidae Halictus sp. 175 Hz

 

 

Figure 15. Cont’d.

1

15 30

Time (mS)

56

Relative Signal Intensity

Relative Signal Intensity

Relative Signal Intensity

 

Hymenoptera: Sphecidae Ectemm‘us maculosus (Gmelin) 182 Hz

AMMIMMNHJIJ‘IAWAAVWM

 

 

 

35 70

Time (mS)

. l 03“-. ‘l

 

 

Diptera: Syrphidae Toxomerus marginatus (Say ) 182 Hz

fi

 

 

 

1

0 60 120
Time (mS)

 

Hemiptera: Lygaeidae Zeridoneus sp. 185 Hz

WNMWWWWWH

1

0 75 150
Time (m8)

1

 

 

 

Figure 15. Cont’d.

57

 

 

Diptera: Calliphoridae 189 Hz

0 'MWW‘WWWWWW‘

 

 

 

0 60 120
Time (mS)

Relative Signal Intensity

 

Hymenoptera: Anthrophoridae 204 Hz

0 ‘WWVWW

 

 

 

Relative Signal Intensity

 

 

 

 

0 30 50

Time (m8)

.3!

E Hymenoptera: Apidae Apis mellfera Linnaeus 217 Hz

.5

.9

(I) .

0

>

35

3 .

é’ o 115 230
Time (mS)

Figure 15. Cont’d.

58

Relative Signal Intensity Relative Signal Intensity

Relative Signal Intensity

 

Hymenoptera: Halictidae

 

 

Augochlorella sp.

217Hz

flfllhwwwnnww

 

 

j

 

 

0 "MW

50 100
Time (mS)
Hymenoptera: Apidae Bombus sp. 217 Hz

 

 

150
Time (m8)

320

 

Homoptera: Cicadellidae

0 J -— M‘MMMANWfiu‘W

 

227Hz

 

 

Figure 15. Cont’d.

1130
Time (mS)

59

200

Relative Signal Intensity

Relative Signal Intensity

Relative Signal Intensity

 

Diptera: Dolichopodidae Condylosrylus sp. 23 8H2

sA-NAWWWWAMWMW

 

 

 

T

 

 

 

 

40 80
Time (m8)
7 Hymenoptera: Sphecidae Tachytes sp. 250Hz
l 1
«WM 11W 111 ill. M 1m
0 55 110
Time (mS)

 

Hymenoptera: Halictidae

'WWWMWWWWWWNWWWM

 

 

 

o 50 100
Time (mS)

Figure 15. Cont’d.

60

Relative Signal Intensity

Relative Signal Intensity

Relative Signal Intensity

 

Diptera; Dolichopodidae

 

Dolichopus sp.

° 111111 111111111th

294Hz

 

 

 

 

 

 

 

 

0 50 100
Time (mS)
Diptera: Dolichopodidae Conaylostylus sp. 500Hz
o .-
I
0 75 150
Time (mS)
Diptera: Chironomidae 909Hz

oawwwwmwmu

 

 

 

Figure 15. Cont’d.

I

140

Time (mS)

61

280

frequencies as sorted by insect taxon (Figure 15). It is reassuring that all wing-beat
frequencies recorded here for a given taxa agee with the values compiled in Figure 2
(Chapter 1) for insects generally.

Counts of signals from three field sites. The count data were normal as
determined by the Shapiro-Wilk (P=0.0836, W= 0.9557) test and blocking was effective
(F = 6.00, P = 0.0001 when H>1) (SAS Institute, 2000). Remarkably high numbers of
insect signals per 10 min sample were detected for all site types (Figure 16). This was
true even for parking lots, originally included as a type of negative control. Moreover,
there was a difference among the sites (F = 8.85, df = 8, p < 0.0001). The equivalent
field and wetland means proved different fiom the parking lots and the control (Figure
16). It is encouraging to see that a hard number can be obtained to verify the extent to
which large areas of pavement and buildings lower the numbers of flying insects.
Notably, these results establish for the first time that optical tachometer devices such as
the OF IDIS sensor are capable of collecting data useful in making statistical comparisons
between habitats.

Interesting count fluctuations occurred at some of the study sites across the season
(Figure 17). Among the old-field sites (Figure 178), Dansville regularly had the highest
count of the three fields. One 10-min recording contained an amazing 1,215 insect
signals (Figure 17A). Perhaps this occurred because at Dansville old-field, the sensor
was close to forest edge. Also, the Dansville site had dense populations of flowering
plants, which may have lured many visitors to the site that subsequently crossed the

sensor.

62

600

 

 

 

 

 

a a
1.1.1
U)
4" 500 1
an
IE
1:
h
o
3 400 -
n:
3
3
IE
2' 300 -
o
F
L
o
a b
E 200 -
3
O
U
u
8
In 100 -
E
:
a
g c
0 ‘ I 1
Control Parking Lot Old-Field Wetland

Location Type

Figure 16. Mean insect signal counts (t SE) per 10 min sampling period across
all sampling dates. No signals were detected when the sensor was shielded by
and opaque cover (control). Treatrnernts bearing a common letter are not different
at p = 0.05.

63

Insect Counts

1200
1000 -

1200 -
1000 a

1200 _
1000 ~

1A

 

800 -
600 ~
400 -
200 -

 

Wetlands

 

 

 

800 .1
600 ~
400 1
200 -

 

 

 

 

 

800 7
600 -
400 -
200 -

o

 

Parking Lots

 

 

199 202 206 229 233 237 254

Julian Days

0 Rose Lake
Flooding

+ Hewes
Lake

+ Central
Park Pond

o Legg Park
Field

--I— Dansville
Field

+ Rose Lake
Field

0 Kmart
Parking Lot

-I— Manley
Miles
Parking Lot

—t-— Hannah
Plaza
Parking Lot

Figure 17. The number of insects detected per 10 min recording as a function of
habitat type and sampling date during summer 2000.

64

In the multiple linear regression analysis, location (irrespective of site type)
explained the most variation (12:0.36) compared to the other variables: time-of day
(r2=0.24), site types (r2=0.22), temperature (12:0.22), date (r2: 0.14), light intensity (r2:
0.03), and wind speed (12 <0.001).

Some interactions are evident between variables of the MLR model. For
example, wind and location together (r2=0.3 8) account for more variation than either
factor did alone. These data show, not surprisingly, that certain location types are prone
to more wind, (e. g., the Kmart parking lot). However, wind alone explained very little
variation (r2=0.07) in the entire model, suggesting that this optical sensor was equally
effective at detecting insects under windy and calm conditions.

Diversity in wing-beat signals from three site types. The standard deviation
for repeated by-hand measures of fundamental frequency was 2.5 Hz. This precision was
considerably better than measurements made by Canary 1.2.4, which showed 10-20 Hz
differences within one signal’s estimation. Figures 18-20 present (at a 2 Hz resolution)
frequency distributions of all fundamental wing-beat frequencies detected at each field
location for 3 dates. It is important to recognize that with a measurement SD of 2.5 Hz
adjacent bars do not represent statistically different measures of wing-beat frequency, but
that those separated by 5 Hz would do so.

If there were no differences in signal richness or diversity between site types, one
would expect similar frequency distributions across sites (Figures 18-20). However,
signal richness was the greatest at the Dansville old—field site (Table 1). The Rose Lake
flooding site was 83% as rich as the Dansville old-field site, while the Kmart parking lot

was only 41% as rich as Rose Lake flooding and 34% as rich as Dansville old-field.

65

Table 1. Signal richness (R), diversity (H') and relative diversity (J ')
compared for all three sites used in diversity study analysis

 

 

Site R H' J '
Dansville Old- Field 259 2.15 .80
Rose Lake Flooding 215 2.26 .84
Kmart Parking Lot 88 1.87 .69

 

H'max = 2.7 for all sites

H0: Equal abundance among k (500) categories for all sites when the
maximum diversity is 2.7

66

.22.?20 2:355 “a 836 m no 35308 385329 Eon—ms? =e Lo :25“:me 55:35 .3 oSwE

oooF

. _
, _

 

com

b _ b

__dfi=

£8 22. I
52 7593 a
5m .38.

31:05:09”. mem-m:_>>

com oou com com oov

3.33:1: _ mama :5,

 

2
.1
._

a

OOON U_m_n_-U_O m=_>wcmo

com

com

 

 

o0?

ro

lo;
-me
..ow
,.mN
..om
..mm

..ov

 

 

mv

slenpgAipui 40 JeunnN

67

.3600: 8.3 08% 3 8:6 n so @2388 aflocoscoa Beams; :5 Lo :ocsflhmmv 5532,.” .3 Saw;

31:05:09". “momma;

ooo_. com com com com 00

— b _

£8 22. I
52 use},
so .38

 

cow

b

o

 

m oov com CON
_ _ r __ __ 1:2: tifigfizjjf

88 9.68: 9.3 80m

to_.

[or

low

lmN

[om

{mm

lov

 

 

mv

SanplAlpU| ,io JeqwnN

68

.8— wfifiwa @525. 05 3 8me m co c.2338 momouoscoa Eon m5? =m mo cozsfibflc 5:26on .om 83w:—

31:05:02... ammm-m:_>>
003 com com com com com oov com com 8.. o

 

. . . P p b i _ _»__ I ; __:_._: .fi _T_ij:_u._: . . O
, - m

- S N

n

. B w

B

I om m:

- mm m

M.

£8 22.- - om nnv

£26.63 ole

£38m; 1 mm S
88 6.. eczema. :95. 8

 

 

 

ms

69

Signal richness alone might indicate that Dansville was the most diverse of the three
sites. However, a better measurement of diversity is an index of the distribution of data
compared to the total number of categories. To do so, I applied the Shannon-Weaver
index to the frequency distribution data. A higher H' value (Table 1) indicates a more
diverse data set based on the number of categories filled and the evenness of the data.
Rose Lake flooding had the highest H' value of 2.26 (H'max = 2.27), ranking it the most
diverse site (Table 1). Values of H' are much easier to compare when they are converted
to a proportion of maximum diversity of the entire data set (J '). The Rose Lake flooding
site was 84% as diverse as the entire data set (Table 1), while Dansville old-field was
80% and Kmart parking lot 69% as diverse as H', respectively.

The composite distribution of individual signals (n=1,088) and their frequencies
at the Dansville old-field site (Figure 18) showed that insects with wing-beat frequencies
between 100 Hz and 300 Hz (peak = 216 Hz) were most prevalent. Based on (Figure 2,
Chapter 2), I suspect that these came from Diptera and Hymenoptera along with perhaps
some Coleoptera and Homoptera. These signal frequencies were also common to the old-
fields (Figure 19) and the parking lots (Figure 20), where there was something even more
peculiar- no insects were detected having a wing-beat frequency above 600 Hz. To be
sure that this was not an affect of the time-of -day at which parking lots were sampled
(Table 2), I analyzed a separate Kmart lO-min recording made at 4:35 pm on July 17th
2000. Only one insect signal exceeded 600 Hz (663Hz) within the entire track.

This high-frequency gap was analyzed by a X2 test to assign a probability of
finding no high-frequency (over 600 Hz) signals in the Kmart parking lot if its signal

distribution were in fact identical to the Dansville old-field or the Rose Lake Flooding

70

Table 2. Time of day each 10 min recording was made at each site on
each date in year 2000

 

 

 

Site Date Time

Dansville Field July 20th 12:59 pm
August 16th 12:43 pm

Sept. 6th 3:35 pm

Rose Lake Flooding July 20th 2:08 pm
August 16th 4:30 pm

Sept. 6th 11:10 am

Kmart Parking Lot July 20th 11:36 am
August 16th 11:35 am

Sept. 6th 1:36 pm

 

71

sites. With the old-field serving as the expected values, the probability of finding no
signals above 600 Hz in the Kmart parking lot proved < 0.001 (X2 = 11.39, df = 1). I thus
conclude that parking lots are devoid of high-frequency fliers. When Rose Lake flooding
served as the expected values for the Kmart comparison, there should have been a 99.9%
chance of finding at least one signal > 600 Hz (X2 = 20.0, df = 1), if the these
distributions were not different.

At the old-field sites and all wetland sites, I visually observed some Odonata and
Lepidoptera flying over the sensor throughout the season. Their signals, known to fall
below 100 Hz, were mostly undetectable by Canary 1.2.4. , that could display harmonics
for signals only as low in frequency as 78 Hz. Unfortunately, this suggests that the
OF IDIS sensor is not well-suited for detecting insects with wing-beat fi'equencies below
80 Hz. This includes most aquatic insects (except for Diptera), Lepidoptera, some
Coleoptera and Odonata.

When the distribution data in (Figures 18-20) are compared to known wing-beat
frequencies for all insects (Figure 2, Chapter 1), the OFIDIS sensor appears to be
primarily a flying Hymenoptera and Diptera detector. Nevertheless, it is still a highly
useful instrument that may allow scientists to gain some perspective on the vast
complexity of pollination systems, natural enemy densities, and insect migration.

In The Forgotten Pollinators, (Buchmann and Nabhan 1996), the complex and
mutualistic relationships between plants and their pollinators are well described. The
bear-claw poppy, which grows only in gypsum soils in Southern Utah, is one clear
example of how unique and fragile these relationships can be with the endangered.

Entomologists, such as these authors and Dr. Tepidino from Utah State University

72

(mentioned by the authors) have attempted to determine the pollinators of these rare
flowering plants. Only recently did a student of Tepidino collect the solitary bee (Perdita
mercom's Griswold) that apparently is this plant’s pollinator. Further observations of the
bear claw poppy indicated that this bee is truly loyal to this plant, despite all of the other
flowers surrounding it.

So how does the OFIDIS sensor fit into this scenario? The wing-beat and
harmonic signature of this bee could be determined, and entered into a neural network so
that long hours of recorded insect data could be played and the solitary bee counted.
Such data could reveal the visitation rates to the flowers and seasonality of pollination.
Such information could aid efforts to restore and protect the bear-claw poppy.

Variation of harmonic patterns of insect signals from the field. Harmonic
patterns for the insects used in the abundance and diversity studies were both visible and
clearly separated, so that it was possible to measure primary frequencies directly from
harmonic spectra. Many signals shared a common fundamental wing-beat fi'equency.
For example, in the Dansville old-field site 40 signals fell within the 216- 217 Hz range
for the dates included in the diversity test. Signals with similar frequency could then be
examined (Figure 21). Indeed variation was detected within these selected signals. Some
harmonic patterns revealed their darkest band (most intense harmonic) at the first
harmonic, while other had their darkest band as the second or third harmonic. It is not
known at this time whether these differences represent signal variation generated by a
common type of insect or signals from different insect types. Although I was able to

match certain signals to each other by simply listening to them and could show that

73

matching signals showed similar harmonic patterns, I would like to see this matching

process become an automated software function.

74

Chapter 4

Summary of thesis research outcomes and suggestions for building upon them

I. Research summary

The research presented in this thesis has significantly advanced understanding of
the performance and utility of the Model 2000 Optical Flying Insect Detection and
Identification System (OFIDIS) under field conditions using sunlight. The sensor was
maximally sensitive when operating under full sunlight. Under a cloudless sky, the
active space for shadowed signals proved to be a cylinder extending from the sensor face
some 25 meters toward the face of the sun. That for reflected signals was characterized
as a cone extending widely from the face of the sensor and likely beyond 25 meters.
Flying insect populations were successfully sampled during 10 min intervals at multiple
times during Year 2000 under a protocol where this sensor was pointed directly upward
toward the unobstructed and cloudless sky at replicated wetland, old-field, and parking
lot sites in Central Michigan. Data were analyzed using commercially available wave-
analysis software packages (TRex 2.0 and Canary 1.2.4) to yield insect counts and
fundamental wing-beat frequencies accurate to i 2.5 (SD) Hz. Plots of fundamental
frequency of insect signals against abundance revealed statistically discemable
differences in insect composition as well as abundance between some sites. As expected,
insect abundance and diversity were lowest at large parking lots, selected to serve as
negative controls. However, all locations achieved a diversity index of 0.69 or greater,
suggesting even over parking lots, many different flying insects are present. On average,
insect abundance and diversity proved equivalent for the wetland and old-field sites. In

addition to fundamental fiequencies, patterns of harmonics were obtained for many of the

75

insect signals above but not below 100 Hz. However, because the identities of the insects
generating the field signals were known in only a few cases, information from these
harmonic records could not be assessed for correlations to particular taxa.

Collectively, this research validates, for the first time, the concept that OFIDIS or
similar optical tachometer systems can be used to quantify insect abundance and diversity
under field conditions. Moreover, this equipment is judged to have high potential for
operating remotely. This study yielded an extensive database of insect signals rich in
harmonics. However, until further work is done relating these signals to identified
insects, little can be concluded here about the utility of OF IDIS as an identification
system for insects generally. When multiple records were captured from defined
oscillators (tooth brush, tuning fork, or defined digital electronic signals), sufficient
variation in harmonic patterns due to signal duration and intensity was found to raise
caution about over-interpreting subtle differences in harmonics between individual
records. According to its designer, the main intent of OFIDIS was detection and
identification of particular insects of interest (e.g., key insect pests) for which extensive
background on specific flight signals (including their variations) could be generated to
serve as a reliable reference for identifications. Nevertheless, the current research shows
that the OFIDIS system can also capture quantitative data of broader ecological interest.
As demonstrated here, this system is appropriate for capturing data of the type useful to
conservation biology and management of habitat quality.

11. Suggestions for improving OFIDIS
The following suggestions are offered for possible improvements in or expansion

of OFIDIS utility for studies such as this one: 1) Because the fundamental wing-beat

76

frequencies of many insect Orders fall below 100 Hz (see Figure 1, p. 5), the sensitivity
of the sensor and/or the responsiveness of analysis software to low frequencies might be
improved. The OF IDIS equipment used here appeared to be better suited to detecting
higher- rather than lower-frequency signals. Perhaps filter systems could be implemented
to better separated low-frequency signals from noise, which seemed to be more prevalent
at lower than higher frequencies. 2) If it were difficult to design one sensor optimally
sensitive to all insect signal frequencies, it might be desirable to employ one low- and one
high-frequency sensor simultaneously to cover the full range of insect wing-beat
frequencies with higher confidence. 3) Design sensors that are range-specific to reduce
the amount of extraneous data collected. This would save time during signal analysis,
especially if signal recording were triggered only by certain frequencies. 4) Same signals
vary in their intensity, which showed evidence of producing variation in their harmonic
spectra. Finding a way to standardize all captured signals prior to harmonic analysis
might reduce this variation. 5) A powerfirl and long-range artificial light source should
be developed as a substitute for sunlight, which in most locations varies widely overtime.
For example, an ordinary laser pointer was found promising in preliminary trials; its
beam projected over many meters to generate a large active space for the sensor both
during daytime as well as night. Clearly, it would be advantageous for OF IDIS users to
be able to sample flying insects consistently over 24 h, under varying weather conditions.
A laser beam appropriately configured with the OFIDIS sensor could extend wing-beat
sampling to shade-seeking (e.g., mosquitoes) as well as nocturnal insects (e. g., moths), in

addition to those flying in full sunlight as sampled here.

77

III. The core question revisited

The OFIDIS system is a tremendous advancement in insect detection. As a device
that can successfully converts subtle insect flight cues into electrical information that is
rich in specific information about each insect, it has proven to be a useful tool of the
modern electric age. I believe that this non-invasive method of insect detection will
prove valuable to conservation biology researchers aiming to monitor specific insects and
those who use insects to measure environmental quality. It will also be useful to the pest-

management community protect both crops and human health.

78

LITERATURE CITED

Altman, J .S. 1975. Changes in the flight motor pattern during the development of the
Australian plague locust, Chortoicetes termimfera. Journal of Comparative Physiology
97:127-142.

Baker, P.S., M. Gewecke, and RJ. Cooter. 1981. The natural flight of the migratory
locust, Locusta migratoria L. Journal of Comparative Physiology 141 :233-237.

Berg, R. E. and D. Stork. 1994. The Physics of Sound. Prentice Hall, New Jersey.

Berry, 1. L. 1973. Improved system for measuring flying activity of insects by detecting
static charges. Journal of Economic Entomology 66(3): 820-821.

Brodsky, A. 1994. The Evolution of Insect Flight. Oxford University Press, Oxford.

Brown, K. S. 1997. Diversity, disturbance, and sustainable use of Neotropical forests:
insects as indicators for conservation monitoring. Journal of Insect Conservation 1:25-
42.

Buchan, P. B., and R. B. Moreton. 1981. Flying and walking of small insects (Musca
domestica) recorded differentially with a standing-wave radar actograph. Physiological
Entomology 6:149-155.

Buchmann, S. and G. Nabhan. 1996. The Forgotten Pollinators. Island Press,
Washington, DC.

Casey, T. M. and M. May 1985. Flight energetics of euglossine bees in relation to
morphology and wing stroke frequency. Journal of Experimental Biology 116:271-289.

Chapman, R. F. 1998. The Insects. Cambridge University Press, Cambridge, UK.

Dudley, R. 2000. The Biomechanics of Insect Flight. Princeton University Press,
Princeton, New Jersey.

Dent, D. 2000. Insect Pest Management. CABI Publishing, Wallingford, England.

Eaton, J. L. 1980. An infi'ared LED-based electronic actograph for monitoring insect
flight activity. Annals of the Entomological Society of America 73:744-746.

Edwards, D. K. 1960. A method for continuous determination if displacement activity in
a group of flying insects. Canadian Journal of Zoology 38:1021-5

79

Ehrlich, P. R. and R. Murphy. 1987. The structure and dynamics of butterfly
populations. Symposia of the Royal Entomological Society of London. Academic Press,
London.

Foster, J. A., and R. M. Robertson. 1992. Temperature dependency of wing-beat
frequency in intact and deafferented locusts. Journal of Experimental Biology 165:295-
3 12.

Gullen, P.J. and RS. Cranston. 1994. The Insects. Chapman and Hall Press, London.

Hardie, J. and S. Young. 1997. Aphid flight-track analysis in three dimensions using
video techniques. Physiological Entomology 22:116-122.

Heinz, K., M. Parrella, and J. Newman. 1992. Time-efficient use of yellow sticky traps
in monitoring insect populations. Journal of Economic Entomology 85:2263-9.

Herrns, D., D. Nielsen, and T. D. Syndor. 1990. Comparison of two methods for
sampling ar boreal insect populations. Journal of Economic Entomology 83: 869-74.

Hobbs, S. E. 1989. An airborne radar technique for studying insect migration. Bulletin
of Entomological Research 79:693-704.

Houck, O. 1993. The Endangered Species Act and the implementation by the US.
Departments of Interior and Commerce. Colorado Law Review 64:277-370.

Joos, B., P.A Young, and TM Casey. 1991. Wingstroke frequency of foraging and
hovering bumblebees in relation to morphology and temperature. Physiological
Entomology 16:191-200.

Knoldel, J. J. and A. Agnello. 1990. Filed comparisons of nonsticky and sticky
pheromone traps for monitoring fruit pests in Western New York. Journal of Economic
Entomology. 83(1): 197-204.

Lohniger H. 1999. Teach Me Data Analysis. Springer-Verlag Press, New York.

Loper, G. M., W.W. Wolf, and OR. Taylor. 1993. Radar detection of drones responding
to honeybee queen pheromone. Journal of Chemical Ecology. 19(9):]929-1938.

Macaulay, E. D. M. 1974. A simple insect flight recorder. Entomology Experimental
Applications 15(2):252-254.

Mascanzoni, D., and H. Wallin. 1986. The harmonic radar: a new method of tracing
insects in the field. Ecological Entomology. 11:387-390.

May, M. L. 1981. Wingstroke fi'equency of dragonflies (Odonata: Anisoptera) in
relation of temperature and body size. Journal Comparative Physiology 144:229-240.

80

Michels, G]. and R.W. Behle. 1992. Evaluation of sampling methods for lady beetles
(ColeopterazCoccinellidae) in grain sorghum. Journal of Economic Entomology.
85(6)2251-7.

Minnaert, M. G. J .1993. Light and Color in the Outdoors. Springer-Verlag Press, New
York.

Moore, A., J. R Miller, B. Tabashnik, and SH. Gage 1986. Automated identification of
flying insects by analysis of wingbeat frequencies. Journal of Economic Entomology
79:1703-1706.

Moore, A. 1991. Artificial neural network trained to identify mosquitoes in flight.
Journal of Insect Behavior 4(3):391-396.

New T. R. 1995. An Introduction to Invertebrate Conservation Biology. Oxford
University Press, Oxford, England.

Oertli, J. J. 1989. Relationship of wing beat frequency and temperature during take-off
flight in temperature-zone beetles. Journal of Experimental Biology 145:321-338.

Ogawa, K., and T. Kanda. 1986. Wingbeat frequencies of some anopheline mosquitoes
of East Asia (Deptera: Culicidae). Applied Entomological Zoology 21(3):430-435.

Oliver, 1. and A. Beattie. 1993. A possible method for the rapid assessment of
biodiversity. Conservation Biology 7:562-8.

Pedigo, LP. 1996. Entomology and pest management. Prentice Hall, New Jersey.

Pearson, D. and F. Cassola. 1992. World-wide species richness patterns of tiger beetles
(Coleoptera: Cicindelidae): Indicator taxon for biodiversity and conservation studies.
Conservation Biology. 62376-91.

Pyle, R. 1981. Insect conservation. Annual Review of Entomology 26:233-58.

Riley, J. R. 1989. Remote sensing in entomology. Annual Review of Entomology
34:247-271.

Riley, J. R., A.D. Smith and B.W. Bettany. 1990. The use of video equipment to record
in three dimensions the flight trajectories of Heliothis armigera and other moths at night.
Physiological Entomology 15:73-80.

Riley, J. R., P. Valeur, A.D. Smith, D.R. Reynolds, G. M. Poppy, and C. Lofstedt. 1998.

Harmonic radar as a means of tracking the pheromone-following flight of male moths.
Journal of Insect Behavior 11(2):287-295.

81

Roberts, S. P., J .F. Harrison, and NP. Hadley. 1998. Mechanisms of thermal balance in
flying Centris pallida (Hymenoptera: Anthophoridae). The Journal of Experimental
Biology 201:2321-2331.

Rui-Lu C., X.Z. Bao, V.A. Drake, R.A. Farrow, S.Y. Wang, Y. J. Sun, B. P. Zhai. 1989.
Radar observations of the spring migration into northeastern China of the oriental
arrnywonn moth, Mythimna separata, and other insects. Ecological entomology 14:149-
162. '

Sabins, F. F. 1987. Remote Sensing: Principles and Interpretation. W. H. Freeman and
Co., New York.

Samways, M. 1984. Insect Conservation Biology. Chapman and Hall Press, London.

Samways, M. and S. Steytler. 1996. Dragonfly (Odonata) distribution patterns in urban
and forest landscapes, and recommendations for riparian management. Biological
Conservation 79:279-88.

Schaefer, G. W., and GA. Bent. 1984. An infra-red remote sensing system for the active
detection and automatic determination of insect flight trajectories (IRADIT). Bulletin of
Entomology Research 74:261-278.

Schotzko, D. J. and L. E. O’Keefe. 1989. Comparison of sweep net, D-vac, and absolute
sampling, and die] variation of sweep net sampling estimates in lentils for pea aphid
(Homotera: Aphidaie), nabids (Hemitera: Nabidae), lady beetles (Coleoptera:
Coccinellidae), and lacewings (Neuroptera: Chrysopidae). Journal of Economic
Entomology 82(2): 491-506.

Sotavalta, O. 1947. The flight-tone (wing-stroke frequency) of insects. Acta. Entomol.
Penn. 4: 1-1 14.

Southwood, T.R.E. 1978. Ecological Methods. Chapman and Hall Press, London.
Spangler, H. G. 1992. The influence of temperature on the wingbeat fi'equencies of free-
flying honey bees, Apis mellifera L. (Hymenoptera: Apidae). Bee Science 2(4): 1 81-

1 86.

Spangler, H. G. 1993. Are the wingbeat frequencies of honey bees an indicator of
populations or behavior? American Bee Journal. 134:53-5.

Steytler, NS. and M. Samways 1994. Biotope selection by adult male dragonflies
(Odonata) at an artificial lake created for insect conservation in South Afiica. Biological
Conservation 72:381-6.

Stinner, RE. 1983. Dispersal and movement of insect pests. Annual Review of
Entomology 28:319-35.

82

Unwin, D. M., and S. A. Corbet. 1984. Wingbeat frequency, temperature and body size
in bees and flies. Physiological Entomology 92115-121.

Unwin, D. M., and GP. Ellington. 1979. An optical tachometer for measurement of the
wing-beat frequency of free-flying insects. Journal of Experimental Biology 82:377-
378.

Wearing, C.H. 1988. Evaluating the [PM implementation process. Annual Review of
Entomology 33:17-38.

Zar, J. 1999. Biostatistical Analysis. Prentice Hall, New Jersey.

83

Appendix 1

Record of Deposition of Voucher Specimens‘
The specimens listed on the following sheet(s) have been deposited in the named museum(s) as
samples of those species or other taxa, which were used in this research. Voucher recognition
labels bearing the Voucher No. have been attached or included in fluid-preserved specimens.
Voucher No.: 2001-01
Title of thesis or dissertation (or other research projects):

Evaluation and field application of an optical sensor that detects flying insects

Museum(s) where deposited and abbreviations for table on following sheets:

Entomology Museum, Michigan State University (MSU)

Other Museums:

Investigator’s Name(s) (typed)
Mgghan French

 

 

 

Date August 3. 2001

*Reference: Yoshimoto, C. M. 1978. Voucher Specimens for Entomology in North America.
Bull. Entomol. Soc. Amer. 24: 141-42.

Deposit as follows:
Original: Include as Appendix 1 in ribbon copy of thesis or dissertation.

Copies: Include as Appendix 1 in copies of thesis or dissertation.
Museum(s) files.
Research project files.

This form is available from and the Voucher No. is assigned by the Curator, Michigan State
University Entomology Museum.

84

Appendix 1.1

Voucher Specimen Data

5 Pages

of

1

 

Page

28 base

|

 

5.934. F memo

 

\OON..N3.\N\ 1.1! K . 30%
. 5:62 Em

 

3.29.5 29w c8222 9.. c_ .588

cocoa“. Panama

 

5. 956.8% 3%.. gone 0.... 823.3.
Foéoow .02 5:025

 

About. @952 $569.33.
3.338: L. «$05 .2368 can.

 

oo-__....r Fm
_. moi 228.com. 3.6:; 9.3 omom

.. oo-_:_.- 3
no? 5533. 8:22. 9.3 meow.

_. oo-c:.a-ow
teen. Ems.
oo-.:7m~
v we? 228.com 952.3 9.3 $01
oo.._2rmm
F no? 523.com. 02.2.? 9.3 omom
00-5..- 5

P

we? scheme”. o~__u._>> 9.3 33.

do 86995
$2538 ”9990

do 6888mm
32.9%. ”92323..
325.5280 ”5280
$2335 ”998502

338:5... Looms 5.6.5.4
omega—Lose... fibfioufio.

.525. 3559. 9:80qu
ongoEEmLoo ”928200

 

 

 

 

 

 

 

 

 

 

 

 

 

m m s
u e .m r m h. e ho. e 8.83% can now:
m are m. M m w m m. m w .0 8.02.8 mcoEBoam .0. Son .33 :96. 55° 8 3.025
M .m m m M A P N m E
no .3832

85

Appendix 1.1

Voucher Specimen Data

5 Pages

Page 2 of

 

 

 

 

 

 

 

00.00- 5

00..< 20.0003. 0...0..3 9.0.. 000m.

00.< 500000”. 02.0.. 3 9.0.. 000m.

0.00 8.050
flit 0.8
$0003.). 30.05.25
3.29.5 0.0.0 $95.: 05 c. .088 6:9“. egg
.0. 0:08.000... 00.0.. 0>0n0 0:. 002000”.
3-38 .02 .0623 .80... 3.0502 €209.82.
5000000: .. 0.00:0 .0:0...000 00:.
8-5..-00 .% 25.80
. 0:00 30.). 0029.. 0.0.0058»:
00.2.- .0 =0”. 0.9.000 00.3.

000.02.00.30 ”00.08.00

.00 800000.02
002360.00 ”0.0.0.0

 

 

 

 

 

 

 

 

 

 

 

 

 

8-50-3 .00 0:93
P 00..< 500000.. 02.0.. 3 9.0.. 000.... 00025. “0.0.0.60...
8-50-8
F 0:00 30.2 002530.000 0.0:000
8-30-00.
F 0:00 30.2 002530.000 ”0.00000
8.578 .00 0.00000
F 00? 50.00001 02.0.. 3 9.0.. 000m. 0000020905 ”0.0.3200.
m .W s
u e m r m e h e 09.00000 0:0 000:
m a m. m w m W W m w .0 00.02.00 20.20000 .0. 0.00 .003 :96. .050 .0 00.00%
M m w 0 A A P N m E
”.0 000052

86

0.00 5.050

#50022 >00.0Eo.:w
0.0.9.5 0.0.0 80.5.2 9.. c. 28000
5. 058.0000 00.0.. 0500 0:. 00>.0001

 

roan/$2 .. 0.00

 

 

 

Po- woom .02 .0:0:o>

:0:0.n. :0:m0.2

.0003. 3.0502 05.000005.
3.000000: 2 0.00:0 .0:0...000 00D.

 

 

Appendix 1.1
Page 3 of 5 Pages

Voucher Specimen Data

F

Rig-mu
00.4. :050001 02.0..3 9.0.. 0001

00-:00-ON

0:00 30.2
20-000-NN

0:00 30.2
00407....”

002 :050001 02.0.. 3 9.0.. 0001

09.2.- PM
002 :0000001 02.2.3 9.0.. 0001
8-504..”
002 :050001 02.0..3 9.0-. 0001

.2050. 0.005005 00.:E0.0m
0020050 ”0:050:03:

.00 00.0.00...
002.020... ”0:050:03:

$00. 00.05902. 0:00:55...
000E030 ”0.0.0.0

.00 030303.05.
002:0...0 ”0.0.0.0

.00 0205.?
002555.. ”0:050:03...

0022.... ”0.0.0.50...

 

 

 

 

 

 

 

 

deposited
Larvae
Eggs

Museum
Nymphs

where
Adults
Adults
Pupae

Other

00200000 0:0 000:
5 00.00.50 0:08.800 5. 0.00 .003

 

:98. 5:5 5 00.0000

 

 

O

 

 

87

Appendix 1.1

Voucher Specimen Data

0.00

5.050

 

5-9.4.-.. F 0.00

 

0.00022 30505.5

 

0.202.... 0.0.0 0005.2 o... c. .0800
.0. 0000..0000 00.0.. 0>000 0... 0020001

0000.... 00.505.

 

 

3. room .02 .000:0>

.000... @0502 0.209.002.
3.0000000 .. 0.0000 .000...000 00:.

 

 

 

Page 4 of 5 Pages

00-50-00.
00.< 00.00001 0.._0_.>> 00.0.. 0001

ooéErow 0000 2w.)—
oo-0.....-o~ 0000 am.)—
oo-0:0-oN 0000 05.2
ooéErow 0000 Ems.
RYE—(mm

00.< 00.00001 0.._2.>> 00.0.. 0001

8-0:..-mN
00.< 00.00001 02.2.3 00.0... 0001

002.0255 ”0.0.0005:
.00 0000.8

000.0< ”05.000005...
.00 0..05...000..<

002.0..0... ”05.000005.

000000... 0.0....00. 0.0...
000.0< ”0.0.0000E>I
002500955. ”05.000005...
000..0..0.._0o ”0.0.0.0

.00 000000.00N
000.000.... ”05.0.0.0...

 

 

 

 

 

 

 

 

 

00-3.30... 0.00. 00.00.90... 0.00055...
. 002 5.0800 000...... 9.0.. 080 000090 0.0.0.0
m m e m e 00.00000 000 000..
e s r .
m .r. W e m m m. M m w .0 00.00.50 000....0000 5.0.00 .000.. 00x0. .050 .0 00.0000
M m w W A A P N m E

 

 

O

H.

50:52

 

 

88

Appendix 1.1

Voucher Specimen Data

0.00 5.0.00

 

3.9.3.0 w 0.00

 

05000.2 30505.00

 

00.02.... 0.0.0 000.5... 0... c. .0800

0000.”. 0000M...

 

5. 00085000 00.0.. 0500 0.... 0020001
.0. .000 .02 .00000>

 

.000... @0502 0.209.002.
2.0000000 .. 0.0000 502.500 003.

 

 

Page 5 of 5 Pages

00.0..—(Nu
F 0000 305.
8-0:..-NN
F 0000 205.
F 09.0.3.0

00.< 00.00001 0...2.>> 00.0.. 0001

002050905 0.0.05
.00 005000000
000089050 0.0.00

.00 000000.50
00200000250 ”0.0.0.0

 

 

 

 

 

 

 

 

 

8-57.0
. 00.< 00.00001 0.._2.>> 00.0.. 0001 0025.5... ”05.000000...
8...... _.m .00 00.0.0005
P 00.< 00.00001 0...0..>> 00.0.. 0001 00200000 05.000005...
.00 005000.50
. $02.00 200 00... 002808.00 0.0.0.0
m m m e 00.00000 000 0000
m .fl. r. 06 K. e D. .
m e W m w Nu. m m. m w 5 00.00.50 000....0000 5.0.00 .000. 00x0. .005 .0 00.0000
M m m 0 A A P N m E

 

 

 

”.0 50.002

 

89

Appendix 2

90

00:5. —_.

    

0.3.3.5. .8838: 8
238.. .0820... E... .5083. 6.5080 I
.020.— 00050000 .00
3.0
§

300.5500. 0:- 600.200 00.05.0080
.5828“. 0180.00.83". 03508.0

 

 
    

 

a” v“ ‘.0““ .
g ‘ eta
< 0‘"
EC :68!

0 ‘ 50000

v.0 ._0_0_._.0_0 0.01

91

00:! to N... a N6

 

 

 

 

 

 

 

 

 

 

 

 

 

.3803 0.2280. .020". 02.8.0020 g

       

‘¢(u

000.0002 000.000 . 001 0.00 _0..000

  

92

. ...mo=s_ cc . 5.2. .. o .. m...

 

 

 

 

258.5. .5...»
.8523.
.25.: .3835... 38813: x/,

233... «cl—IEO; v:- .:o§¢ «Ear—20 I
gm see—COO .
.32.an 23.2.8. .83“. Bio—32¢ ” .

59:35. .m=_>m:mn “Euwfia 2.59:5 can 3.3 330...

 

93

 

 

 

.5528: U H.
ggvgwcoz //‘ w

({(G
1.

17”.. All. L ‘>_.

:mmEu

_.s_ 63:.

 

‘‘‘‘‘

¢.. .

. . I. ¢..,

‘ “AN. 5 .. ..

¢ inc $ .
\ N

<

20 U:

 

mm:

5

  

oo.

      

   

25.95. 28888: ‘ <
2:5... 282...; as. ecu-8: €512“. I
38252. a... 5828 .3250 .
.3823. 2.886. .85”. Essa-2n U H u .

 

   

 

 

 

 

   

 

(<(. ......

   

(<<<....... ..

  

<<¢¢....... ..

 

 

  

 

.3... 9:28 5.5.

95

 

59:25..

 

G

mez

 

[NH
Md G M6

<<<<<<

97