1
00:00:06,076 --> 00:00:08,130
All right, so today
I'm going to be talking

2
00:00:08,130 --> 00:00:12,030
about how effective dogs versus humans are

3
00:00:12,030 --> 00:00:15,480
at being able to detect
spotted lanternfly egg masses

4
00:00:15,480 --> 00:00:18,430
not the adults but the actual egg masses.

5
00:00:18,430 --> 00:00:21,111
And if I can advance this, there we go.

6
00:00:21,111 --> 00:00:23,490
So there's a huge team
that's been involved

7
00:00:23,490 --> 00:00:26,670
in this work and I'd like to
first acknowledge the co-PIs

8
00:00:26,670 --> 00:00:30,490
on the project, Carrie Brown-
Lima, and Ann Hajek, as well

9
00:00:31,806 --> 00:00:34,117
as everyone else on this slide
that's contributed either

10
00:00:34,117 --> 00:00:37,980
with working with us as
collaborators or in the field.

11
00:00:37,980 --> 00:00:40,710
So, many thanks to the
large team and funding

12
00:00:40,710 --> 00:00:43,210
by the Cornell Atkinson
Center for Sustainability.

13
00:00:44,610 --> 00:00:47,940
For folks who don't know much
about spotted lanternfly,

14
00:00:47,940 --> 00:00:50,310
I'll give you a very brief
background about them.

15
00:00:50,310 --> 00:00:53,940
They're thought to have been introduced

16
00:00:53,940 --> 00:00:55,890
from Asia on these stone slabs.

17
00:00:55,890 --> 00:00:58,380
They lay their egg masses
on smooth surfaces.

18
00:00:58,380 --> 00:00:59,790
So you can see the picture

19
00:00:59,790 --> 00:01:02,370
on the upper right are
the actual egg masses.

20
00:01:02,370 --> 00:01:06,150
These are the egg masses you can see

21
00:01:06,150 --> 00:01:08,250
on that stone slab.

22
00:01:08,250 --> 00:01:09,360
In 2014,

23
00:01:09,360 --> 00:01:11,910
and they arrived in Pennsylvania.

24
00:01:11,910 --> 00:01:13,350
And then despite a lot

25
00:01:13,350 --> 00:01:15,480
of efforts that have
been ongoing to try to

26
00:01:15,480 --> 00:01:17,850
prevent the spread,
they've rapidly spread.

27
00:01:17,850 --> 00:01:21,090
I'll show a map in the
next slide of that spread.

28
00:01:21,090 --> 00:01:22,530
So they're plant hoppers

29
00:01:22,530 --> 00:01:25,080
and they have these
piercing, sucking mouth parts

30
00:01:25,080 --> 00:01:28,860
and they feed on and damage
agricultural species,

31
00:01:28,860 --> 00:01:32,970
primarily grapes, apples, hops, peaches.

32
00:01:32,970 --> 00:01:34,904
And one of the characteristics

33
00:01:34,904 --> 00:01:37,830
of them is they excrete
something called honey dew.

34
00:01:37,830 --> 00:01:40,500
And once they excrete that
honey dew, for example,

35
00:01:40,500 --> 00:01:42,870
on the leaves, then there's
a sooty mold that grows

36
00:01:42,870 --> 00:01:45,960
on top of them, on top of the honey dew.

37
00:01:45,960 --> 00:01:49,620
And that further damages
both the crop itself

38
00:01:49,620 --> 00:01:50,643
and the leaves.

39
00:01:51,510 --> 00:01:53,243
In addition to the damage of the crops,

40
00:01:53,243 --> 00:01:56,310
they also swarm in very large numbers

41
00:01:56,310 --> 00:01:58,444
and can be quite the annoyance

42
00:01:58,444 --> 00:02:00,513
if you're trying to do anything outside.

43
00:02:01,740 --> 00:02:03,240
This is the distribution room.

44
00:02:03,240 --> 00:02:06,540
This was actually October 26th

45
00:02:06,540 --> 00:02:08,820
and they have spreads since October 26th.

46
00:02:08,820 --> 00:02:10,920
I just checked the new map yesterday

47
00:02:10,920 --> 00:02:12,360
which I didn't upload here

48
00:02:12,360 --> 00:02:14,196
but there are actually new infestations,

49
00:02:14,196 --> 00:02:17,640
especially this new one
popped up right here.

50
00:02:17,640 --> 00:02:19,920
So that's now considered
an infestation rather

51
00:02:19,920 --> 00:02:21,420
than just an occurrence.

52
00:02:21,420 --> 00:02:26,420
So, the initial spotted
lanternfly introduction

53
00:02:26,490 --> 00:02:29,843
was right here in Berks
County, Pennsylvania in 2014.

54
00:02:29,843 --> 00:02:31,740
Now you can see they're

55
00:02:31,740 --> 00:02:35,587
in York, Massachusetts, Michigan,
Indiana, North Carolina.

56
00:02:35,587 --> 00:02:39,810
So, rapidly spreading in Vermont.

57
00:02:39,810 --> 00:02:42,330
You can see in the
[Indistinct] area right there,

58
00:02:42,330 --> 00:02:46,653
an individual detection.

59
00:02:48,718 --> 00:02:50,283
No invasion as of this point.

60
00:02:52,050 --> 00:02:54,750
And if we're thinking about
managing invasive species,

61
00:02:54,750 --> 00:02:57,690
particularly if we're thinking
about just single individuals

62
00:02:57,690 --> 00:03:00,420
or very small populations,
we want to try to be

63
00:03:00,420 --> 00:03:03,900
in this realm of eradication
or maybe containing

64
00:03:03,900 --> 00:03:05,760
the population.

65
00:03:05,760 --> 00:03:07,620
In Pennsylvania and New Jersey now,

66
00:03:07,620 --> 00:03:08,790
they're beyond the point

67
00:03:08,790 --> 00:03:11,130
of trying to contain or eradicate.

68
00:03:11,130 --> 00:03:13,800
This is where there is
like asset protection mode.

69
00:03:13,800 --> 00:03:16,770
But when we started this
project in New York,

70
00:03:16,770 --> 00:03:19,860
we're at this stage of
not having infestations.

71
00:03:19,860 --> 00:03:23,520
And so we were thinking
about how do we prevent them

72
00:03:23,520 --> 00:03:25,830
from coming into the
state from Pennsylvania.

73
00:03:25,830 --> 00:03:27,150
They're like knocking on our door

74
00:03:27,150 --> 00:03:28,680
and we don't want them to come in.

75
00:03:28,680 --> 00:03:32,250
That's what Vermont's
situation is right now, right?

76
00:03:32,250 --> 00:03:34,260
And so, we started to think about

77
00:03:34,260 --> 00:03:37,140
these early detection,
rapid response approaches

78
00:03:37,140 --> 00:03:39,250
that might be suitable for use in New York

79
00:03:40,110 --> 00:03:42,300
but it's challenging

80
00:03:42,300 --> 00:03:43,740
because

81
00:03:43,740 --> 00:03:47,070
the egg masses are very camouflaged

82
00:03:47,070 --> 00:03:48,600
and they're very difficult to find.

83
00:03:48,600 --> 00:03:51,720
So this tree right here
has an egg mass on it.

84
00:03:51,720 --> 00:03:52,653
Can anyone see it?

85
00:03:54,300 --> 00:03:55,490
No, probably not.

86
00:03:55,490 --> 00:03:57,150
I'm gonna zoom in,

87
00:03:57,150 --> 00:03:57,983
now,

88
00:03:57,983 --> 00:03:59,160
Can you see it there?

89
00:03:59,160 --> 00:04:01,443
Still probably not is my guess.

90
00:04:02,610 --> 00:04:03,660
There it is right there.

91
00:04:03,660 --> 00:04:05,850
So they're extremely,
extremely camouflaged,

92
00:04:05,850 --> 00:04:07,080
very hard to find.

93
00:04:07,080 --> 00:04:09,270
So even humans where you know

94
00:04:09,270 --> 00:04:10,718
it is on a tree,

95
00:04:10,718 --> 00:04:13,410
it's very easy to overlook them.

96
00:04:13,410 --> 00:04:15,990
And so given the complexity
with trying to find them,

97
00:04:15,990 --> 00:04:17,047
what we did is we said,

98
00:04:17,047 --> 00:04:20,790
"Well, can we use dogs to
try to find the egg masses?"

99
00:04:20,790 --> 00:04:22,590
Something that's very camouflage.

100
00:04:22,590 --> 00:04:24,540
Dogs can use their olfactory senses,

101
00:04:24,540 --> 00:04:27,720
we can train the dogs
to smell the egg masses.

102
00:04:27,720 --> 00:04:28,553
And especially

103
00:04:28,553 --> 00:04:29,700
that honey dew

104
00:04:29,700 --> 00:04:32,836
is something that they smell as well.

105
00:04:32,836 --> 00:04:34,805
Can we train these dogs to be able to

106
00:04:34,805 --> 00:04:36,510
to find the egg masses?

107
00:04:36,510 --> 00:04:38,700
So our detection dog team here,

108
00:04:38,700 --> 00:04:40,770
we used Dia and Fagan

109
00:04:40,770 --> 00:04:44,220
and they were our trustee field workers

110
00:04:44,220 --> 00:04:46,473
in addition to their handlers.

111
00:04:47,430 --> 00:04:50,970
We surveyed 20 vineyards in New Jersey

112
00:04:50,970 --> 00:04:53,490
and Pennsylvania and we
surveyed them during the winter.

113
00:04:53,490 --> 00:04:55,980
So that's the only time of the year when

114
00:04:55,980 --> 00:04:58,848
the individuals, the
species isn't moving, right?

115
00:04:58,848 --> 00:05:02,400
So there are no adults in
the environment at that time.

116
00:05:02,400 --> 00:05:04,380
It's just these egg masses.

117
00:05:04,380 --> 00:05:06,733
So we could think about the
population being closed at

118
00:05:06,733 --> 00:05:08,400
that period of time.

119
00:05:08,400 --> 00:05:12,930
So, this was the original core
in Harrisburg, Pennsylvania.

120
00:05:12,930 --> 00:05:15,990
We're just east of that area.

121
00:05:15,990 --> 00:05:18,270
And where we were in New York

122
00:05:18,270 --> 00:05:19,770
and here's Cornell right up here.

123
00:05:19,770 --> 00:05:22,450
And this is the the border where

124
00:05:23,640 --> 00:05:24,873
the area of concern is.

125
00:05:26,550 --> 00:05:29,190
For the survey methods,
we surveyed 20 vineyards

126
00:05:29,190 --> 00:05:32,040
and then the surrounding natural
areas and by natural areas

127
00:05:32,040 --> 00:05:33,840
usually this was like the forested area,

128
00:05:33,840 --> 00:05:36,990
kind of the upland area that
was adjacent to the vineyards.

129
00:05:36,990 --> 00:05:39,060
We had 12 different transects

130
00:05:39,060 --> 00:05:40,920
so that we're within the vineyard area.

131
00:05:40,920 --> 00:05:43,500
And the average length of
those was about 20 meters.

132
00:05:43,500 --> 00:05:45,200
And then we had 20 transects that were

133
00:05:45,200 --> 00:05:47,520
in these adjacent forests.

134
00:05:47,520 --> 00:05:50,880
Those were about 26 meters in length.

135
00:05:50,880 --> 00:05:53,020
We had subunits that we surveyed

136
00:05:54,159 --> 00:05:56,940
and the subunits were kind of
nested within these transects.

137
00:05:56,940 --> 00:05:58,980
And so, here is for example, a transect

138
00:05:58,980 --> 00:06:01,360
in the vineyard and you can see

139
00:06:03,497 --> 00:06:04,860
the vines right here.

140
00:06:04,860 --> 00:06:06,060
And this is a metal pole.

141
00:06:06,060 --> 00:06:08,580
I don't know if it's a little bit dark

142
00:06:08,580 --> 00:06:10,830
but there's metal poles
that are usually between

143
00:06:10,830 --> 00:06:13,200
these vines kind of holding them up.

144
00:06:13,200 --> 00:06:14,820
Approximately every meter is where

145
00:06:14,820 --> 00:06:15,990
these vines or poles are.

146
00:06:15,990 --> 00:06:18,030
So we considered those as sub-units.

147
00:06:18,030 --> 00:06:20,817
And then also in the forest
transect, we had these transects

148
00:06:20,817 --> 00:06:22,080
and we were able to divide 'em

149
00:06:22,080 --> 00:06:23,763
up into these one meter segments.

150
00:06:24,750 --> 00:06:27,150
We repeated the survey
two times with humans

151
00:06:27,150 --> 00:06:29,490
and we repeated the survey
two times with dogs.

152
00:06:29,490 --> 00:06:31,350
So every transect we measured,

153
00:06:31,350 --> 00:06:33,030
the same transect we measured twice

154
00:06:33,030 --> 00:06:34,590
with both humans and dogs.

155
00:06:34,590 --> 00:06:36,210
And then we gave the humans

156
00:06:36,210 --> 00:06:38,940
and the dogs an unlimited search time.

157
00:06:38,940 --> 00:06:41,310
And something important is that all

158
00:06:41,310 --> 00:06:42,480
of the sites that we went

159
00:06:42,480 --> 00:06:45,300
to had a known visible infestation.

160
00:06:45,300 --> 00:06:46,500
We knew that there were going

161
00:06:46,500 --> 00:06:48,068
to be egg masses there.

162
00:06:48,068 --> 00:06:50,970
We didn't tell the dogs or
the handlers where they are.

163
00:06:50,970 --> 00:06:52,560
We didn't tell the humans where the are,

164
00:06:52,560 --> 00:06:54,360
we just knew that that vineyard

165
00:06:54,360 --> 00:06:56,983
had some kind of egg masses on it.

166
00:06:56,983 --> 00:06:58,740
And the reason that we did that is

167
00:06:58,740 --> 00:07:00,201
because we're trying to
compare the difference

168
00:07:00,201 --> 00:07:01,650
between humans and dogs.

169
00:07:01,650 --> 00:07:02,820
So we needed to go to sites

170
00:07:02,820 --> 00:07:05,580
that had something to be able to detect.

171
00:07:05,580 --> 00:07:09,750
So, this isn't a very
early detection scenario

172
00:07:09,750 --> 00:07:11,250
per se

173
00:07:11,250 --> 00:07:13,923
because we knew it had
some kind of infestation.

174
00:07:15,480 --> 00:07:18,810
For those of you who don't
know about occupancy modeling,

175
00:07:18,810 --> 00:07:20,430
I'll give you a little overview of that

176
00:07:20,430 --> 00:07:24,330
because that's the method that
we use to analyze these data.

177
00:07:24,330 --> 00:07:26,460
The occupancy modeling accounts

178
00:07:26,460 --> 00:07:29,970
for imperfect detection
of organisms and surveys

179
00:07:29,970 --> 00:07:32,700
and the picture or the
figure at the bottom,

180
00:07:32,700 --> 00:07:34,140
I'm going to go through

181
00:07:34,140 --> 00:07:35,400
what I mean by that.

182
00:07:35,400 --> 00:07:37,200
Basically, it uses presence data.

183
00:07:37,200 --> 00:07:38,700
We'll denote that as a one

184
00:07:38,700 --> 00:07:42,123
and non detection data and
we'll denote that as a zero.

185
00:07:43,753 --> 00:07:45,390
And we go out and we survey an area,

186
00:07:45,390 --> 00:07:46,440
we're gonna go out and we're going

187
00:07:46,440 --> 00:07:48,190
to survey for spotted lanternflies.

188
00:07:49,230 --> 00:07:52,020
We go out and either one
of two things happens.

189
00:07:52,020 --> 00:07:53,730
We either see the species,

190
00:07:53,730 --> 00:07:56,280
we see the spotted lanternfly or we don't.

191
00:07:56,280 --> 00:07:58,803
So these are our field
observations right here.

192
00:08:00,250 --> 00:08:01,710
If we go out and we see
the spotted lanternfly,

193
00:08:01,710 --> 00:08:05,310
there's only one kind
of biological reason why

194
00:08:05,310 --> 00:08:07,200
we could have seen that spotted lanternfly

195
00:08:07,200 --> 00:08:09,630
because it was actually there, right?

196
00:08:09,630 --> 00:08:10,710
But if we go out

197
00:08:10,710 --> 00:08:13,290
and we survey the area and
we don't see the species,

198
00:08:13,290 --> 00:08:15,480
one of two things might be going on.

199
00:08:15,480 --> 00:08:17,084
The biological reality might be

200
00:08:17,084 --> 00:08:19,080
well, it's actually not there.

201
00:08:19,080 --> 00:08:20,392
But a second possibility is

202
00:08:20,392 --> 00:08:22,560
the spotted lanternfly

203
00:08:22,560 --> 00:08:25,200
is actually there, but
we just didn't see it.

204
00:08:25,200 --> 00:08:29,790
So, this issue right here
is where the problem comes

205
00:08:29,790 --> 00:08:32,760
in because this is our
imperfect detection.

206
00:08:32,760 --> 00:08:35,040
So as humans or as dogs,

207
00:08:35,040 --> 00:08:38,250
we don't always see
the biological reality.

208
00:08:38,250 --> 00:08:40,770
So that is what we'll call
field observation error

209
00:08:40,770 --> 00:08:42,720
or this imperfect detection.

210
00:08:42,720 --> 00:08:46,080
And these occupancy
models allow us to account

211
00:08:46,080 --> 00:08:48,376
for this imperfect detection

212
00:08:48,376 --> 00:08:51,240
and we can estimate two parameters.

213
00:08:51,240 --> 00:08:53,217
We can estimate the
probability of occupancy,

214
00:08:53,217 --> 00:08:55,977
the probability that a
site has that species

215
00:08:55,977 --> 00:08:57,330
and then the probability

216
00:08:57,330 --> 00:08:59,760
of detection given that species is there,

217
00:08:59,760 --> 00:09:02,100
what's our probability of
being able to detect it?

218
00:09:02,100 --> 00:09:03,270
And that's what we did

219
00:09:03,270 --> 00:09:05,793
in the comparison of the
humans versus the dogs.

220
00:09:06,900 --> 00:09:09,780
So we used this
multi-scale occupancy model

221
00:09:09,780 --> 00:09:11,850
and the two scales are the transect level

222
00:09:11,850 --> 00:09:14,280
and then the subunit level
within that transect.

223
00:09:14,280 --> 00:09:15,674
The transects,

224
00:09:15,674 --> 00:09:18,420
we can think about those as being related

225
00:09:18,420 --> 00:09:21,390
to the probability of
invasion from a source.

226
00:09:21,390 --> 00:09:22,873
So, if you're in the vineyard,

227
00:09:22,873 --> 00:09:25,500
it's most likely to be invaded

228
00:09:25,500 --> 00:09:26,950
from the forest site because

229
00:09:26,950 --> 00:09:29,770
the spotted lanternfly is associated

230
00:09:29,770 --> 00:09:32,900
with true species like tree of
heaven that are likely coming

231
00:09:32,900 --> 00:09:35,790
from the surrounding natural areas.

232
00:09:35,790 --> 00:09:38,400
And then the subunits
within that transect,

233
00:09:38,400 --> 00:09:41,460
each individual vine or
pole or one meter segment

234
00:09:41,460 --> 00:09:44,070
within the forest can
be thought to be related

235
00:09:44,070 --> 00:09:46,410
to the intensity of infestation.

236
00:09:46,410 --> 00:09:49,800
So how many of those one
meter segments within

237
00:09:49,800 --> 00:09:51,449
that transect

238
00:09:51,449 --> 00:09:54,690
have the spotted lanternfly egg masses?

239
00:09:54,690 --> 00:09:56,850
And then we could estimate
the detection probability

240
00:09:56,850 --> 00:10:00,390
of the subunits and
compare humans versus dogs.

241
00:10:00,390 --> 00:10:02,580
So here are the occupancy results.

242
00:10:02,580 --> 00:10:04,380
First off, at the transect level,

243
00:10:04,380 --> 00:10:07,140
so this is an individual
transect in the vineyard

244
00:10:07,140 --> 00:10:10,380
or an individual transect
within the forest area.

245
00:10:10,380 --> 00:10:15,030
The occupancy of egg masses
was very high in both of them.

246
00:10:15,030 --> 00:10:20,030
Within the vineyards, 0.94 and
then within the forest, 0.85.

247
00:10:20,520 --> 00:10:23,220
So most of the transects
that we were surveying

248
00:10:23,220 --> 00:10:25,050
had egg masses on them.

249
00:10:25,050 --> 00:10:27,923
However, if we go down
to the subunit level,

250
00:10:27,923 --> 00:10:30,540
this is within those one meter segments,

251
00:10:30,540 --> 00:10:32,610
the vineyard had an occupancy

252
00:10:32,610 --> 00:10:37,203
at the subunit level of
0.47 and the forest at 0.13.

253
00:10:38,280 --> 00:10:41,220
And so despite very high occupancy at

254
00:10:41,220 --> 00:10:44,913
both the vineyards and the
forest at the transect level,

255
00:10:46,181 --> 00:10:47,070
the forest had much lower,

256
00:10:47,070 --> 00:10:50,580
we could think of it as infestation level

257
00:10:50,580 --> 00:10:54,085
within the individual one meter transects.

258
00:10:54,085 --> 00:10:56,400
And this is showing the subunit occupancy

259
00:10:56,400 --> 00:10:58,800
on the right, we have
the occupancy probability

260
00:10:58,800 --> 00:11:01,914
on the Y axis and then
we're comparing vines,

261
00:11:01,914 --> 00:11:04,290
the metal poles within the vineyard.

262
00:11:04,290 --> 00:11:06,000
So those are kind of combined

263
00:11:06,000 --> 00:11:07,890
into what we're calling the vineyard

264
00:11:07,890 --> 00:11:09,360
and then the forested sites.

265
00:11:09,360 --> 00:11:11,490
So two things that you'll see, one is

266
00:11:11,490 --> 00:11:15,750
that the higher infestation
sites had a higher probability

267
00:11:15,750 --> 00:11:17,190
of occupancy.

268
00:11:17,190 --> 00:11:18,360
And then the second thing

269
00:11:18,360 --> 00:11:21,780
is that the metal poles
had a higher probability

270
00:11:21,780 --> 00:11:24,600
of occupancy than did the vines.

271
00:11:24,600 --> 00:11:27,810
And so I mentioned that the
spotted lanternfly likes to

272
00:11:27,810 --> 00:11:30,510
lay their egg masses on smooth surfaces

273
00:11:30,510 --> 00:11:35,510
and the metal poles are smooth
relative to the rough vines.

274
00:11:35,790 --> 00:11:36,688
So they,

275
00:11:36,688 --> 00:11:40,110
kind of have a say a "preference"

276
00:11:40,110 --> 00:11:43,776
for these vines or for
the poles over the vines.

277
00:11:43,776 --> 00:11:48,231
We can also look at the distance

278
00:11:48,231 --> 00:11:49,590
from forest.

279
00:11:49,590 --> 00:11:53,010
So, we have the vineyard
transects that were laid out.

280
00:11:53,010 --> 00:11:54,480
We surveyed these transects

281
00:11:54,480 --> 00:11:56,310
but we could measure the distance

282
00:11:56,310 --> 00:11:58,710
from that transect to that upland forest

283
00:11:58,710 --> 00:12:00,990
and look at the probability of occupancy

284
00:12:00,990 --> 00:12:04,440
as a function of how far
away it is from the forest.

285
00:12:04,440 --> 00:12:06,044
And our prediction would be

286
00:12:06,044 --> 00:12:09,860
that the occupancy should
be higher the closer you are

287
00:12:09,860 --> 00:12:11,520
to the forests because that's

288
00:12:11,520 --> 00:12:13,320
where the adults would be coming

289
00:12:13,320 --> 00:12:15,180
from out of the forest because

290
00:12:15,180 --> 00:12:18,540
of that association with with
tree of heaven for example.

291
00:12:18,540 --> 00:12:20,250
And that's exactly what we saw.

292
00:12:20,250 --> 00:12:22,550
This is the occupancy probability

293
00:12:22,550 --> 00:12:25,440
and distance to forest on the x axis.

294
00:12:25,440 --> 00:12:27,870
And the closer you are to the forest,

295
00:12:27,870 --> 00:12:30,183
the higher the probability of occupancy.

296
00:12:31,080 --> 00:12:32,850
As you get away from the forest,

297
00:12:32,850 --> 00:12:35,370
within those transects,
there was a lower probability

298
00:12:35,370 --> 00:12:37,833
that there was going
to be egg masses there.

299
00:12:37,833 --> 00:12:39,180
So this has implications
for where you might think

300
00:12:39,180 --> 00:12:42,960
about searching for egg
masses within vineyards.

301
00:12:42,960 --> 00:12:47,040
If you look at this,
somewhere around 75 meters

302
00:12:47,040 --> 00:12:49,632
or closer, would be really
good places to search

303
00:12:49,632 --> 00:12:53,553
since those have the highest
occupancy probabilities.

304
00:12:55,350 --> 00:12:56,940
This is consistent with a study

305
00:12:56,940 --> 00:13:00,390
in Pennsylvania where they
were looking at egg masses

306
00:13:00,390 --> 00:13:03,240
and they found that 44% of them were

307
00:13:03,240 --> 00:13:06,780
within 15 meters of the
edge of the vineyard.

308
00:13:06,780 --> 00:13:09,813
So again, there's strong
association with the forest.

309
00:13:11,160 --> 00:13:13,260
Now I'm switching to
the detection results.

310
00:13:13,260 --> 00:13:16,560
This is the difference between
the humans and the dogs.

311
00:13:16,560 --> 00:13:18,570
And if you look

312
00:13:18,570 --> 00:13:21,690
at the figure on the right,

313
00:13:21,690 --> 00:13:24,160
this is detection
probability on the Y axis.

314
00:13:24,160 --> 00:13:26,490
We have the humans searching the vineyards

315
00:13:26,490 --> 00:13:28,290
and the dogs serving the vineyards,

316
00:13:28,290 --> 00:13:29,580
humans searching the forest

317
00:13:29,580 --> 00:13:31,500
and then dogs surveying the forest.

318
00:13:31,500 --> 00:13:34,530
And what you can see here is
that within the vineyards,

319
00:13:34,530 --> 00:13:37,017
humans had a higher detection probability

320
00:13:37,017 --> 00:13:38,580
than did the dogs.

321
00:13:38,580 --> 00:13:42,810
They had 1.8% higher
detection probabilities

322
00:13:42,810 --> 00:13:44,220
than did the dogs.

323
00:13:44,220 --> 00:13:47,340
Now humans use visual search.

324
00:13:47,340 --> 00:13:50,430
The picture that I showed you
on the tree was hard to see.

325
00:13:50,430 --> 00:13:52,717
But when you're in a vineyard
and you're up close to

326
00:13:52,717 --> 00:13:57,540
a vine, a grapevine
that is only this high,

327
00:13:57,540 --> 00:14:00,480
it's really easy to see
all of the egg masses

328
00:14:00,480 --> 00:14:01,650
that are there.

329
00:14:01,650 --> 00:14:04,470
However, when you go out into the forest,

330
00:14:04,470 --> 00:14:06,690
we can see that the dogs are much better.

331
00:14:06,690 --> 00:14:09,240
The dogs had a much higher
detection probability

332
00:14:09,240 --> 00:14:12,270
and did 3.4 times better
than did the humans.

333
00:14:12,270 --> 00:14:13,140
And the reason why is

334
00:14:13,140 --> 00:14:15,960
because dogs can use
their olfactory sensors

335
00:14:15,960 --> 00:14:19,080
and we as humans are pretty
useless in that context.

336
00:14:19,080 --> 00:14:20,958
It's really hard brushy vegetation,

337
00:14:20,958 --> 00:14:24,180
we don't know where to
search, we can't see anything.

338
00:14:24,180 --> 00:14:25,890
Yeah, it's just like stems

339
00:14:25,890 --> 00:14:28,593
and leaves and it's
like kind of impossible.

340
00:14:30,240 --> 00:14:32,250
We can also think about
the search efficiency.

341
00:14:32,250 --> 00:14:33,950
So we had detections,

342
00:14:33,950 --> 00:14:37,350
how well we're doing at detecting them

343
00:14:37,350 --> 00:14:39,480
but if you think about
how much time it took,

344
00:14:39,480 --> 00:14:41,913
this turns into an issue of costs.

345
00:14:42,864 --> 00:14:45,360
So, the dogs actually ended up searching

346
00:14:45,360 --> 00:14:47,970
for a longer period of time than humans.

347
00:14:47,970 --> 00:14:51,419
And in the vineyards, the
humans had more detections

348
00:14:51,419 --> 00:14:56,419
and the time that it took
didn't overcome the dogs.

349
00:14:57,447 --> 00:14:59,790
And so the human, the number of detections

350
00:14:59,790 --> 00:15:02,630
per hour was 31 detections
per hour for humans,

351
00:15:02,630 --> 00:15:05,190
24 detections per hour for dogs.

352
00:15:05,190 --> 00:15:08,340
So humans were better in
terms of their efficiency.

353
00:15:08,340 --> 00:15:10,830
In the forest, dogs were
still slightly better

354
00:15:10,830 --> 00:15:12,870
despite the longer time that it took them

355
00:15:12,870 --> 00:15:13,980
to search these areas.

356
00:15:13,980 --> 00:15:16,590
So it took dogs 7.6 detections

357
00:15:16,590 --> 00:15:18,990
or they did 7.6 detections

358
00:15:18,990 --> 00:15:23,010
per hour and humans had
6.7 detections per hour.

359
00:15:23,010 --> 00:15:24,930
However, I'll note that

360
00:15:24,930 --> 00:15:27,360
I think that the number of detections

361
00:15:27,360 --> 00:15:30,060
and not the efficiency is
probably more important when

362
00:15:30,060 --> 00:15:32,520
we're thinking about early
detection situations where

363
00:15:32,520 --> 00:15:35,275
it's really important to
be able to detect those

364
00:15:35,275 --> 00:15:36,948
individuals.

365
00:15:36,948 --> 00:15:40,233
So probably the cost of the search time

366
00:15:40,233 --> 00:15:44,223
is worth it relative to the the gain.

367
00:15:45,180 --> 00:15:48,750
And then infestation level,
this is really not surprising.

368
00:15:48,750 --> 00:15:51,150
The higher infestation
sites had higher detection

369
00:15:51,150 --> 00:15:53,670
probability, so there's just
more egg masses that are

370
00:15:53,670 --> 00:15:55,619
available to be detected.

371
00:15:55,619 --> 00:15:58,410
Finally, what does this mean in practice,

372
00:15:58,410 --> 00:16:01,170
if you're going out there
and doing some searches.

373
00:16:01,170 --> 00:16:04,590
One, I would recommend
that you search vineyards

374
00:16:04,590 --> 00:16:05,970
that are closer to the forest,

375
00:16:05,970 --> 00:16:07,890
within about 75 meters.

376
00:16:07,890 --> 00:16:10,830
Second one is searching
larger trees near the edges,

377
00:16:10,830 --> 00:16:14,610
especially at the higher
elevations because the adults

378
00:16:14,610 --> 00:16:18,720
will be coming down from those
higher elevation tree sites,

379
00:16:18,720 --> 00:16:21,750
searching the metal poles and
the vineyards is a good idea.

380
00:16:21,750 --> 00:16:23,990
And then using dogs to
search the vineyards

381
00:16:23,990 --> 00:16:26,547
in these early detections
situations is still

382
00:16:26,547 --> 00:16:27,600
a good approach.

383
00:16:27,600 --> 00:16:30,623
We were, again, we were
surveying in what would be

384
00:16:30,623 --> 00:16:33,450
considered a moderate to
high level infestation sites.

385
00:16:33,450 --> 00:16:35,226
And then the dogs were particularly good

386
00:16:35,226 --> 00:16:38,280
in the forested sites where humans kind

387
00:16:38,280 --> 00:16:39,113
of fell apart

388
00:16:39,113 --> 00:16:43,410
because our visual search is
impeded in those forest sites.

389
00:16:43,410 --> 00:16:47,820
So thank dogs have a great
utility in these forested sites.

390
00:16:47,820 --> 00:16:49,590
Thanks to all the vineyard owners

391
00:16:49,590 --> 00:16:51,510
for access to their land.

392
00:16:51,510 --> 00:16:54,821
And then finally, I
will take any questions.

393
00:16:54,821 --> 00:16:57,488
(students clap)

394
00:17:01,470 --> 00:17:03,480
[Moderator] Somebody in the
chat does have a question.

395
00:17:03,480 --> 00:17:05,943
She asked, did the handlers ever report,

396
00:17:08,397 --> 00:17:10,800
Did the handlers ever report

397
00:17:10,800 --> 00:17:15,387
that the dogs alert for
lanternfly when not working?

398
00:17:15,387 --> 00:17:19,184
And if it gets widespread
and established with the dogs

399
00:17:19,184 --> 00:17:22,170
then would the dogs always be working?

400
00:17:22,170 --> 00:17:25,545
Yeah. So this issue of the dogs,

401
00:17:25,545 --> 00:17:27,420
so I can probably for folks

402
00:17:27,420 --> 00:17:30,237
that don't know how the
dogs work, I go over them

403
00:17:30,237 --> 00:17:34,200
but basically you're training
the dogs to detect something

404
00:17:34,200 --> 00:17:37,800
and the dogs are motivated
by like any normal dog,

405
00:17:37,800 --> 00:17:39,300
they're motivated by the attention,

406
00:17:39,300 --> 00:17:41,610
they're motivated by some kind of reward.

407
00:17:41,610 --> 00:17:44,970
So every time that they find an egg mass,

408
00:17:44,970 --> 00:17:45,837
you have to reward them

409
00:17:45,837 --> 00:17:47,811
and you throw them a
ball and you know, like a

410
00:17:47,811 --> 00:17:49,850
think about a yellow lab or something,

411
00:17:49,850 --> 00:17:51,300
they're just like, "Oh yeah."

412
00:17:51,300 --> 00:17:52,920
and then they go get the ball.

413
00:17:52,920 --> 00:17:55,320
They wanna keep working because you know

414
00:17:55,320 --> 00:17:57,453
their handler just rewarded them.

415
00:17:58,440 --> 00:18:00,810
So, this issue about the number

416
00:18:00,810 --> 00:18:03,270
of detections is particularly important

417
00:18:03,270 --> 00:18:06,960
because if they are everywhere,

418
00:18:06,960 --> 00:18:08,640
you're gonna have to keep rewarding them

419
00:18:08,640 --> 00:18:09,990
and rewarding them and rewarding them.

420
00:18:09,990 --> 00:18:11,730
So we did have to reward them.

421
00:18:11,730 --> 00:18:14,970
The dogs are basically
pulled when the search

422
00:18:14,970 --> 00:18:16,118
is done.

423
00:18:16,118 --> 00:18:17,033
So, in terms of,

424
00:18:17,033 --> 00:18:20,553
are they still finding them
when they're not working?

425
00:18:21,649 --> 00:18:23,640
There's a signal basically

426
00:18:23,640 --> 00:18:26,580
that the handler handler
will use to tell the dog,

427
00:18:26,580 --> 00:18:28,551
like, "Okay, now it's work time."

428
00:18:28,551 --> 00:18:30,933
So there isn't necessarily that issue.

429
00:18:31,909 --> 00:18:33,287
[Moderator] You have one more minute

430
00:18:33,287 --> 00:18:36,939
if you want to take this last question.

431
00:18:36,939 --> 00:18:39,480
[Student] Will you be publishing

432
00:18:39,480 --> 00:18:41,670
this somewhere accessible?

433
00:18:41,670 --> 00:18:43,440
Yep, we're writing it up right now.

434
00:18:43,440 --> 00:18:46,829
And so I'd say we'll send
it out to a journal probably

435
00:18:46,829 --> 00:18:49,383
in the next two months,
something like that.

436
00:18:51,497 --> 00:18:52,900
[Moderator] Thank you.

437
00:18:52,900 --> 00:18:55,567
(students clap)