Diptera: Tephritidae - PubAg - USDA

1 downloads 0 Views 378KB Size Report
Nov 1, 2009 - lengths of R. pomonella and R. zephyria in Washington state and to determine ...... Feder, J. L., T. A. Hunt, and G. L. Bush. 1993. The effect of.
SYSTEMATICS

Analysis of Body Measurements and Wing Shape to Discriminate Rhagoletis pomonella and Rhagoletis zephyria (Diptera: Tephritidae) in Washington State WEE L. YEE,1,2 PETER S. CHAPMAN,1 H. DAVID SHEETS,3

AND

THOMAS R. UNRUH1

Ann. Entomol. Soc. Am. 102(6): 1013Ð1028 (2009)

ABSTRACT Rhagoletis pomonella (Walsh) (Diptera: Tephritidae) is a quarantine pest of apple (Malus sp.) in Washington state that is almost identical morphologically to Rhagoletis zephyria Snow, a nonpest of apple. Historically, the longer ovipositor in R. pomonella has been used to separate it from R. zephyria, despite overlap in ovipositor lengths. Here, the objectives were to determine whether use of multiple body measurements and wing shape can improve discrimination of the species. Ovipositor lengths allowed 94.6% correct identiÞcation of female R. pomonella but only 7.0% correct identiÞcation of R. zephyria. We found that multiple body measurements and wing shape can better separate these species. Canonical variates analysis (CVA) of nine body measurements in female ßies largely separated the species. In contrast, CVA of nine body measurements poorly separated the males of these species. Discriminant analysis using nine body measurements classiÞed female R. pomonella and R. zephyria with 95.4 and 100% accuracy, respectively. Geometric morphometrics and CVA separated wing shapes between species in both sexes. Bookstein shape coordinates indicated that the wing of R. pomonella is more tapered at the tip than that of R. zephyria. Use of wing shape in an assignments test identiÞed female R. pomonella with 98.5% and female R. zephyria with 99.0% accuracy, and it correctly identiÞed 100% of ßies whose identities were questionable based on ovipositor lengths. Results indicate that use of multiple body measurements or wing shape is an improvement over the use of ovipositor length alone for identifying female R. pomonella and R. zephyria in Washington state. KEY WORDS apple maggot, ovipositor length, geometric morphometrics, canonical variates analysis

Apple maggot, Rhagoletis pomonella (Walsh), and Rhagoletis zephyria Snow (Diptera: Tephritidae) are sibling species within the R. pomonella species complex that are almost identical morphologically, although they have evolved to use completely different hosts (Bush 1966). Larvae of R. pomonella develop in fruit of plants in the Rosaceae (primarily apples, Malus sp., and hawthorns, Crataegus spp.), whereas those of R. zephyria develop in fruit of plants in the Caprifoliaceae (primarily snowberry, Symphoricarpos spp.) (Bush 1966, Yee 2008, Yee and Goughnour 2008). R. pomonella is not native to the PaciÞc Northwest of the United States (AliNiazee and Penrose 1981) and is a quarantine pest of apple in Washington state. There is zero tolerance for larval-infested fruit, affecting apple export to California as well as at least 13 of 32 markets abroad (Washington State University 2008). Apples are the major fruit crop in Washington, valued at $1.746 billion in 2007, with an estimated per hectare value of $27,279 over 63,990 ha harvested (USDA 1 USDAÐARS, Yakima Agricultural Research Laboratory, 5230 Konnowac Pass Rd., Wapato, WA 98951. 2 Corresponding author, e-mail: [email protected]. 3 Department of Physics, Canisius College, 2001 Main St., Buffalo, NY 14208.

2008). IdentiÞcations of trapped ßies support quarantine-related measures imposed by the Washington State Department of Agriculture (WSDA), including mass trapping near orchards to show the absence of the ßy. WSDA spends ⬇$300,000/yr for detection surveys for R. pomonella in Washington by using yellow sticky traps. The numbers of R. zephyria versus R. pomonella caught on traps can vary dramatically from site to site. Accurate identiÞcation of ßies in traps is needed in efforts to prevent the spread of R. pomonella into commercial orchards. Inaction because R. pomonella are misidentiÞed as R. zephyria could result in economic losses because it could increase the threat of R. pomonella moving into orchards and result in expanded quarantined areas. WSDA relies on the longer ovipositor length in female R. pomonella than R. zephyria and the different orientation of the surstyli in males to discriminate the two species. Westcott (1982) showed that in Oregon the ovipositors of R. pomonella are longer than those of R. zephyria, with little overlap, and that the surstyli of male R. pomonella are parallel, whereas those of R. zephyria diverge, providing nearly complete separation of males of the two species. However, there seems to be more variation in ovipositor length than previ-

1014

ANNALS OF THE ENTOMOLOGICAL SOCIETY OF AMERICA

ously realized, and some male R. zephyria may have parallel surstyli like R. pomonella (W.L.Y., unpublished data). In addition, surstylus shapes differ (Wasbauer 1963, Bush 1966), but it is unknown whether intermediate shapes exist. When use of one body measure or trait is questionable, the use of multiple body measurements in a morphometric analysis may be useful for discriminating species (e.g., Herna´ndez et al. 2007). In particular, it may reduce the problem of overlap for the ovipositor trait now in use. Newer statistical methods have not been investigated for discriminating R. pomonella and R. zephyria, notably geometric morphometrics, which has been useful in analyzing wing shape differences in other ßy species (Klingenberg et al. 1998, Santos et al. 2004, Gilchrist and Crisafulli 2006, Marsteller et al. 2009). The landmark-based geometric morphometric method captures information about the shape of structures independent of scale, in a way that allows visual depiction of the results from clustering or discriminant analyses, such as canonical variates analysis (Bookstein 1991, Rohlf and Marcus 1993, Zelditch et al. 2004). As early as 1932, Lathrop and Nickels (1932) provided an illustration that showed apparent differences in wing shapes of blueberry maggot, Rhagoletis mendax Curran, and R. pomonella (then called “forms” of R. pomonella), although they concluded “no structural characteristics whatever could be discerned by which the forms might be separated.” Westcott (1982) brießy footnoted that the wing of R. pomonella is “very slightly less broadly rounded apically” than that of R. zephyria but provided no further comment. More recently, separation of R. pomonella and R. zephyria was obtained using wing vein structure modeled via Bayesian and probability neural networks (Bi et al. 2007), but this method is rather abstract and seems difÞcult to apply. Here, the objectives were to 1) examine ovipositor lengths of R. pomonella and R. zephyria in Washington state and to determine whether 2) multiple body measurements and whether 3) wing shape analysis using geometric morphometrics can improve discrimination of the species. In addition, body size and wing shapes of R. pomonella from different hosts and areas in Washington were determined, as were wing shape differences between sexes. Possible relationships between body measurements or wing shape and ßy life histories and host plants are discussed, as is the practical use of results for ßy identiÞcation. Materials and Methods Ovipositor Lengths of Flies Caught on Traps. Because of the relative rarity of R. pomonella in central Washington (Yee 2008), most trapping occurred in western Washington, where populations are high (Tracewski et al. 1987, Yee and Goughnour 2008). In contrast, populations of R. zephyria are high on snowberry in both western and central Washington (Yee and Goughnour 2008, Yee 2008). Sticky yellow panels and red spheres baited with ammonium carbonate

Vol. 102, no. 6

(AC), which attracts both R. pomonella and R. zephyria, were placed in apple and SuksdorfÕs hawthorn trees [Crataegus suksdorfii (Sarg.) Kruscke] in Woodland and Vancouver in Clark County and ⬇18 km east of Washougal in Skamania County (45.88⬚ N, 122.75⬚ W and 45.38⬚ N, 122.40⬚ W, and 45.58⬚ N, 122.16⬚ W, respectively) in July to August 2001, 2003, 2005, and 2007. Snowberry plants were present near trapped trees at all sites. Flies were removed weekly from traps and stored in ethanol. Ovipositors of 2,070 females were measured (in 2001, 2003, 2005, and 2007: 704, 412, 447, and 507, respectively). Flies were also trapped in introduced hawthorn trees (unidentiÞed species of Crataegus) in Pendleton, OR (45.65⬚ N, 118.80⬚ W), in 2005 with sticky red spheres baited with AC and from black hawthorn trees (Crataegus douglasii Lindl.) in Wenas (46.50⬚ N, 120.57⬚ W) in central Washington in 2007 and 2008. Ovipositors of 143 ßies from Pendleton and 390 ßies from Wenas were measured. No snowberry plants were found at the Pendleton site, but many were found at the Wenas site. Hosts of captured ßies were unknown, so ovipositor lengths of all ßies at each site were pooled. Ovipositors and all other body parts (below) were measured using a stereomicroscope (Zeiss Stemi 2000; Carl Zeiss Light Microscopy, Gottingen, Germany) with a reticule with 100 divisions mounted in a 10⫻ objective (Carl Zeiss, W-P1, 10⫻/23, 45 50 43) (50⫻ maximal power) and a cold light ring mount. In accordance with Westcott (1982), the ovipositor was measured from its apex to the apex of the basal dorsal extension process. However, this process was not always present, in which case it was measured from its apex to the base of the ovipositor (Fig. 1A). Body Measurements of Flies Reared From Known Host Fruit. Flies were reared from fruit collected in three areas in Washington in 2006 and 2007: 1) southwestern Washington, consisting of sites around Vancouver and a site in Skamania County; 2) Puyallup (47.10⬚ N, 122.17⬚ W) north of area 1; and 3) central Washington, consisting of three sites: Nile Valley, Tampico (46.50⬚ N, 121.57⬚ W and 46.33⬚ N, 120.53⬚ W, respectively), and Wenas. Seven groups of ßies were measured: R. pomonella from 1) apple, southwestern Washington (SW Wash pom apple); 2) apple, Puyallup (W Wash pom apple); 3) ornamental hawthorn (Crataegus monogyna Jacq.), southwestern Washington (SW Wash pom orn haw); 4) ornamental hawthorn, Puyallup (W Wash pom orn haw); 5) SuksdorfÕs hawthorn, southwestern Washington (SW Wash pom Suks haw); R. zephyria from 6) snowberry, southwestern Washington (SW Wash zeph snw), and from 7) snowberry, central Washington (C Wash zeph snw). In addition, low numbers of an eighth group, R. pomonella reared from black hawthorn in Wenas in central Washington (C Wash pom black haw), were measured. Abbreviations for groups and collection sites are used henceforth in the methods and results sections for easier tracking. Two to 10 individual plants were sampled per site. Fruit were held for at least 30 d at 20 Ð27⬚C on screens placed on rubber tubs elevated above a layer of soil (or

November 2009

YEE ET AL.: DISCRIMINATION OF R. pomonella AND R. zephyria

1015

Fig. 1. (AÐE) Body measurements made of R. pomonella and R. zephyria for analyses. Parts shown are from a female R. pomonella (from SuksdorfÕs hawthorn); comparable in R. zephyria. (F) Right wing of female R. zephyria (from snowberry), showing 18 landmarks used for canonical variates analysis. Wings to same scale.

directly in tubs for apples). Larvae left fruit to pupariate. Collected puparia were kept at 3Ð 4⬚C for at least 6 mo and then transferred to 20 Ð27⬚C for adult emergence. Flies from apples or hawthorns and from snowberries were assumed to be R. pomonella and R.

zephyria, respectively, because host associations are a strong indicator of species (Bush 1966). There is no evidence that R. pomonella attacks snowberry nor that R. zephyria attacks apples or hawthorns in Washington (Yee 2008, Yee and Goughnour 2008).

1016

ANNALS OF THE ENTOMOLOGICAL SOCIETY OF AMERICA

Adults were allowed to fully sclerotize for 2 d after emergence and then frozen. From 40 to 131 females and 40 males from each group were measured. Lengths of 10 structures or distances between structures were measured (Fig. 1AÐE): 1) ovipositor length (females) or 2) genitalia length (males; apex of the surstylus to the posterior edge of the epandrium); 3) head width; 4) intereye distance; 5) wing length [from proximal thickening of the costa to the apex, as in Wasbauer (1963)]; 6) wing band two (medial crossband); 7) wing band three (subapical crossband); 8) right hind femur; 9) right hind tibia; and 10) right hind tarsus. Voucher specimens were deposited in the entomological collection at the USDAÐARS Yakima Agricultural Research Laboratory in Wapato, WA. Statistical Analyses of Body Measurements. Each of the nine body measurements within each sex was compared across populations using one-way analysis of variance (ANOVA), followed by Fisher least signiÞcance difference (LSD) test and a Bonferroni adjustment. Pearson correlation coefÞcients between ovipositor length and other body measurements were determined. Canonical variates analysis (CVA) (Zelditch et al. 2004) was conducted to describe differences between R. pomonella and R. zephyria within sexes by using the PROC CANDISC procedure in SAS (SAS Institute 1999). CVA is used to simplify descriptions of differences between groups, rather than among individuals. CVA Þnds the set of axes, or canonical variates (CVs), which are linear combinations of the original variables, that best discriminates between two or more groups by maximizing the between-group variance relative to the within-group variance (Zelditch et al. 2004). PROC VARCOMP in SAS was used to show how the proportion of total variance in the canonical variates created in the CVA was partitioned among species and populations or within populations (error). Both species and populations were treated as random effects for this analysis. PROC GLM in SAS was used to conduct a multivariate ANOVA (MANOVA) on the coefÞcient of variation that best discriminated groups, followed by the LSD test and Bonferroni adjustment. Discriminant analysis was conducted using the PROC DISCRIM procedure in SAS to assign each ßy a probability of it belonging to each of the seven ßy groups based on the distance of its discriminant function from that of each class mean. Specimens were assigned to the group that was least different (⫽highest P value) from them. The CROSSVALIDATION option in SAS was used to provide a rigorous, although conservative, classiÞcation accuracy and the CROSSLISTERR option was used to list each misclassiÞed observation. Sources of Flies for Wing Analyses. The same eight groups of female and male ßies used for body measurements were used for wing analyses, including the C Wash pom black haw group. Rearing methods were the same, but most ßies used for body measurements and for wing shape analysis were different. Digitization of Wings for Analyses. The right wing of each ßy was removed and mounted on a glass slide, with the upper side of the wing facing up (Figs. 1F and

Vol. 102, no. 6

Fig. 2. Right wings of male (A) R. pomonella (from ornamental hawthorn) and male (B) R. zephyria (from snowberry). Wings to same scale.

2). A 0.13Ð 0.17-mm-thick glass coverslip was placed over each wing and the edges of the coverslip were sealed to the slide with hot wax to ensure ßatness and stability of the wing. Wings were photographed at 20⫻ magniÞcation by using a Nikon Coolpix 4500 digital camera (Nikon Inc., Melville, NY) mounted on an MZ6 Stereozoom microscope (Leica, Wetzlar, Germany); light was projected from below the slide using a Þber optic light source. Images were saved in jpeg format at 1024 by 768 pixels resolution and then converted into tps format for digitization of landmarks using the program tpsUtil (Rohlf 2001) (http://life. bio.sunysb.edu/morph). In total, 18 landmark were digitized on each wing image (Fig. 1F) using the tps Dig2 program (Rohlf 2001). Tps Þles were then converted into the Integrated Morphometrics Package for Windows (IMP) format using the program CoordGen6f (http://www.canisius.edu/⬇sheets/ morphsoft.html). Statistical Analyses of Wing Shapes. The program CVAGen6o (⫽CVAGen hereafter) in IMP was used to conduct a CVA on the digitized data, after transformation of the raw data via Procrustes and thin-plate spline techniques described below. Specimens were superimposed on an iteratively estimated mean form using a Procrustes superimposition process to remove differences in the landmark conÞgurations caused by scale, rotation, and translation. The method is called a Generalized Least Squares Procrustes superimposition and is a standard approach in geometric morphometrics used to prepare data for subsequent analysis

YEE ET AL.: DISCRIMINATION OF R. pomonella AND R. zephyria

November 2009

(Rohlf and Marcus 1993, Zelditch et al. 2004, Claude 2008). CVAGen generated sets of linear axes (CVs) that were submitted to BartlettÕs test and used to determine the numbers of statistically distinct CVs that discriminate among groups (Zelditch et al. 2004). Thin-plate spline (tps) deformation grids were generated for each species or group using partial warps (Bookstein 1991) to visualize changes in landmark positions relative to the other landmarks, to provide a statistically convenient basis data set. Deformation grids illustrate the pattern of change in a structure (a deformation) relative to a mean (or reference) form. CoordGen6f generated Bookstein shape coordinates (Bookstein 1991) for graphical display to compare landmark positions for all specimens of the two species. CVAGen was also used to perform an assignment test using a CVA-distance based method, which determines the probability that the specimen is closer to the mean of the group to which it was assigned a priori than to the mean of another group (Zelditch et al. 2004, Nolte and Sheets 2005). For showing reliability in the assignments test, jackknife estimates were used (Efron 1982). The PROC CANDISC procedure in SAS was used to conduct a one-way MANOVA on CV1 scores generated from 32 partial warp scores in CVAGen, to test the hypothesis that differences in CV1 among groups were different. Pairwise comparisons were made using the LSD test followed by a Bonferroni adjustment. Fifty wings per group from Þve of the six groups of female R. pomonella were analyzed; only 18 females from the C Wash pom black haw group were analyzed as these ßies were relatively rare. Totals of 40 and 58 wings from female SW and C Wash zeph snw groups, respectively, were analyzed. For males, wings of 30 Ð 42 ßies were analyzed for each of seven groups, with six from the C Wash pom black haw group. For both sexes, analysis was Þrst conducted on all eight groups, followed an analysis done on two groups created by combining all ßies of each species. To test whether CVA can successfully classify unknown specimens, wings of 18 ßies reared from apple and hawthorns in SW Wash that had intermediate ovipositor lengths (0.88 Ð 0.98 mm) (Westcott 1982) were analyzed. Wing shape data were entered into a CVA Gen program option that compared known with unknown ßies by using a CVA-distance based method. Only one R. zephyria had an intermediate ovipositor length (0.91 mm long), but it was also entered as an unknown. Female Versus Male Wing Shapes. Differences between wing shapes in female and male ßies within species were determined. Differences would justify separating morphometric data by sex. Flies of the same species were combined across sites. Results Ovipositor Lengths of Flies Caught on Traps. The ovipositor overlap zone, or “problem area” mentioned by Westcott (1982) is 0.88 Ð 0.98 mm. Of the 2,070 ßies from southwestern Wash, 2.0% had ovipositors below,

1017

9.7% had ovipositors in, and 88.3% had ovipositors above this zone (Fig. 3A). Of the 143 ßies from Pendleton, 0, 22.4, and 77.6% had ovipositors below, in, and above the zone, respectively (Fig. 3A). These sites were dominated by R. pomonella and possibly no R. zephyria occurred at Pendleton. Of 390 ßies from central Washington, where R. zephyria was abundant, 75.6, 8.7, and 15.7% had ovipositors below, in, and above the overlap zone, respectively (Fig. 3A). Thus, depending on the site, 8.7Ð22.4% of ßies (either R. pomonella or R. zephyria) had ovipositors in the problem area. Body Measurements of Flies Reared From Known Host Fruit. Female R. pomonella reared from different hosts and from different areas differed in body measurements. Those from the SW Wash pom orn haw group were slightly smaller, but most measurements of all R. pomonella groups were larger than those of R. zephyria from SW and C Wash zeph snw groups (Table 1). However, the C Wash pom black haw ßies had the shortest ovipositors within R. pomonella, and showed much greater overlap with C Wash zeph snw than other groups (see below). SW Wash zeph snw ßies were smaller than C Wash zeph snw ßies based on Þve measurements (Table 1). Hind tibia length, intereye distance, head width, hind femur length, and wing length all were larger in female R. pomonella than female R. zephyria, by ⬇30% (Table 1). However, wing band two and three widths did not differ among all R. pomonella and R. zephyria. In R. pomonella, correlations between ovipositor length and hind tibia length (r ⫽ 0.404), intereye distance (r ⫽ 0.265), head width (r ⫽ 0.342), hind femur length (r ⫽ 0.416), and wing length (r ⫽ 0.408) were modestly positive and signiÞcant (P ⬍ 0.0001 in all cases). Similarly, in R. zephyria, there were modestly positive correlations (P ⬍ 0.0001 in all cases) between ovipositor length and hind tibia length (r ⫽ 0.368), intereye distance (r ⫽ 0.286), head width (r ⫽ 0.439), hind femur length (r ⫽ 0.413), and wing length (r ⫽ 0.385). Of the nine body measurements in females, ovipositor length differed most (40% greater in R. pomonella) between species and had the highest R2, suggesting that it is the most reliable single measure to separate the species, although head width also had a high R2 (Table 1). However, ovipositors of R. pomonella and R. zephyria (all groups) ranged from 0.74 to 1.32 mm and from 0.60 to 0.91 mm, respectively, and thus overlapped from 0.74 to 0.91 mm (Fig. 3B and C). Fully 94.6% of R. pomonella fell above this expanded overlap zone and would be identiÞed accurately, whereas only 7.0% of R. zephyria fell below the overlap zone and would be identiÞed accurately, with the remainder falling in the expanded overlap zone. Use of WestcottÕs overlap zone of 0.88 Ð 0.98 mm placed 84.1% of R. pomonella (excluding the C Wash pom black haw group) above the overlap and 97.7% R. zephyria (both groups) below the overlap. Of the 240 R. pomonella, 13.8% had ovipositors in the overlap region and 2.1% had ovipositors that were below it (0.74 Ð 0.87 mm) (Fig. 3B). Use of WestcottÕs zone placed only 28.6% of 14 C Wash pom black haw ßies above the zone. None

1018

ANNALS OF THE ENTOMOLOGICAL SOCIETY OF AMERICA 600

Vol. 102, no. 6

A. Trapped Flies SW Wash (2001, 2003, 2005, 2007, n = 2,070)

500 400

Pendleton, OR (2005, n = 143) C Wash (2007 and 2008, n = 390)

300 200 100 0 0.5

90

5-

0.6

0 0.6

6-

0.7

0 0.7

7-

0.8

2 0.8

8-

3 8 0.9 -0.9 4 0.9

1. 0

1. 0 5-

9 1. 1

1. 2 6-

0 1.2

7-

1.3

1 1.3

8-

1.4

2 1. 4

1. 5 9-

3

B. Reared Flies, SW and W Wash (2006 and 2007)

Numbers of Flies

80 From apple and haws (n = 240) From snowberry (n = 40)

70 60 50 40 30 20 10 0 0.5

5-

0.6

0 6 0.6

-0

.70

7 0.7

-0

.82

0.8

8-

3 98 0.9 4-0. 0.9

1. 0

1. 0 5-

9 1.1

6-

1. 2

0 1.2

7-

1.3

1 1. 3

1. 4 8-

2 1.4

9-

1.5

3

70

C. Reared Flies, C Wash (2006 and 2007) 60

From black haws (n = 14) From snowberry (n = 131)

50 40 30 20 10 0 0.5

5-

0. 6

0 0.6

6-

0. 7

0 0.7

7-

0.8

2 0. 8

3 98 0. 9 0 . 8- .940

1. 0

1. 0 5-

9 1. 1

6-

1.2

0 1.2

7-

1.3

1 1. 3

1. 4 8-

2 1. 4

9-

1.5

3

Ovipositor Length (mm) Fig. 3. Distributions of ovipositor lengths of R. pomonella and R. zephyria (A) caught on traps, (B) reared from Þeld-infested apple, hawthorn, and snowberry fruit from southwestern and western Washington, and from (C) central Washington.

of the 40 SW and W Wash R. zephyria had ovipositors in the overlap zone (Fig. 3B), whereas 3.1% of the 131 C Wash R. zephyria had ovipositors in this zone (Fig. 3C). Male R. pomonella were generally larger than male R. zephyria (Table 2). However, not all measurements, including genitalia length, differed between all groups

of R. pomonella and R. zephyria. Unlike females, males of the C and SW Wash zeph snw populations did not differ in size. Of the nine body measurements in males, hind tibia length differed most between species, but wing length, femur length, head width, and hind tarsus length also were larger in R. pomonella by ⬇20% (Table 2). However, this was less than the ⬇30% in fe-

226.43 ⬍0.0001 0.771 40.0

1.07AB (0.01) 1.07A (0.01) 1.02B (0.01) 1.05AB (0.01) 1.08A (0.01) 0.93 (0.02) 0.77C (0.01) 0.81C (0.00)

Ovipositor length

193.14 ⬍0.0001 0.742 34.1

1.70A (0.01) 1.73A (0.01) 1.52C (0.02) 1.70A (0.01) 1.60B (0.01) 1.47 (0.04) 1.29E (0.02) 1.36D (0.01)

Head width

108.54 ⬍0.0001 0.617 34.7

0.66A (0.01) 0.64A (0.01) 0.56C (0.01) 0.63AB (0.01) 0.61B (0.01) 0.56 (0.02) 0.49D (0.01) 0.51D (0.00)

Intereye distance

153.79 ⬍0.0001 0.696 29.7

4.38A (0.04) 4.50A (0.01) 3.93C (0.05) 4.43A (0.03) 4.16B (0.04) 4.06 (0.09) 3.47D (0.04) 3.60D (0.02)

Wing length

51.77 ⬍0.0001 0.435 30.3

0.82A (0.01) 0.84A (0.01) 0.70C (0.01) 0.86A (0.01) 0.74B (0.01) 0.73 (0.03) 0.66C (0.01) 0.73B (0.01)

Wing band 2

11.99 ⬍0.0001 0.151 14.3

0.40A (0.01) 0.40A (0.01) 0.35B (0.01) 0.40A (0.01) 0.36B (0.01) 0.35 (0.02) 0.35B (0.01) 0.36B (0.00)

Wing band 3

Host, area 1.47A (0.01) 1.51A (0.01) 1.39B (0.02) 1.39B (0.02) 1.40B (0.02) 1.41 (0.04) 1.27C (0.02) 1.25C (0.01) 31.31 ⬍0.0001 0.408 20.8

6.74 ⬍0.0001 0.129 11.1

Head width

0.47B (0.01) 0.50A (0.01) 0.47AB (0.005) 0.47B (0.01) 0.46B (0.004) 0.46 (0.00) 0.45B (0.01) 0.46B (0.005)

Genitalia length

12.93 ⬍0.0001 0.221 17.0

0.53AB (0.01) 0.55A (0.01) 0.51BC (0.01) 0.50BCD (0.01) 0.51BC (0.01) 0.51 (0.02) 0.48CD (0.01) 0.47D (0.01)

Intereye distance

26.57 ⬍0.0001 0.369 23.2

3.62AB (0.04) 3.83A (0.04) 3.45B (0.05) 3.52B (0.04) 3.50B (0.04) 3.40 (0.10) 3.17C (0.04) 3.11C (0.08)

Wing length

14.74 ⬍0.0001 0.245 17.5

0.70AB (0.01) 0.74A (0.01) 0.65CD (0.01) 0.68BC (0.01) 0.64CD (0.01) 0.69 (0.02) 0.63D (0.01) 0.66BCD (0.01)

Wing band 2

4.17 ⬍0.0001 0.084 6.2

0.32AB (0.01) 0.34A (0.01) 0.32B (0.004) 0.31B (0.01) 0.32B (0.005) 0.33 (0.01) 0.32B (0.004) 0.32B (0.005)

Wing band 3

Host fruit codes same as in Table 1. pom from C Wash were not included in statistical analyses because of low sample size. b df ⫽ 6, 273. Means within columns followed by same letters are not signiÞcantly different (P ⬎ 0.0024; LSD test, Bonferroni adjustment).

a

145.56 ⬍0.0001 0.684 31.8

1.45A (0.01) 1.45A (0.01) 1.26C (0.02) 1.42A (0.01) 1.34B (0.01) 1.28 (0.04) 1.10E (0.01) 1.17D (0.01)

Hind femur

122.61 ⬍0.0001 0.646 35.1

1.29A (0.01) 1.31A (0.01) 1.10C (0.02) 1.28A (0.01) 1.16B (0.01) 1.16 (0.04) 0.97E (0.01) 1.05D (0.01)

Hind tibia

92.36 ⬍0.0001 0.578 28.3

1.33A (0.02) 1.36A (0.01) 1.18B (0.02) 1.35A (0.01) 1.22B (0.01) 1.19 (0.03) 1.06D (0.01) 1.12C (0.01)

Hind tarsus

48.27 ⬍0.0001 0.515 23.1

1.28A (0.01) 1.33A (0.01) 1.19B (0.02) 1.19B (0.01) 1.22B (0.01) 1.20 (0.04) 1.09C (0.01) 1.08C (0.01)

Hind femur

35.62 ⬍0.0001 0.439 24.0

1.13B (0.01) 1.19A (0.01) 1.05C (0.02) 1.06C (0.01) 1.05C (0.01) 1.07 (0.04) 0.96D (0.01) 0.99D (0.01)

Hind tibia

32.35 ⬍0.0001 0.416 20.2

1.22A (0.01) 1.25A (0.01) 1.13B (0.02) 1.14B (0.01) 1.14B (0.01) 1.13 (0.03) 1.05C (0.01) 1.04C (0.01)

Hind tarsus

Mean (SE) measurements (millimeters) of male R. pomonella (pom) and R. zephyria (zeph) reared from host fruit collected in 2006 and 2007 from different areas in Washington state

pom (n ⫽ 40) apple, SW Wash pom (n ⫽ 40) apple, W Wash pom (n ⫽ 40) orn haw, SW Wash pom (n ⫽ 40) orn haw, W Wash pom (n ⫽ 40) Suks haw, SW Wash black haw, C Wash poma (n ⫽ 6) zeph (n ⫽ 40) snw, SW Wash zeph (n ⫽ 40) snw, C Wash One-way ANOVAb F P R2 Maximal % greater in R. pomonella than R. zephyria

Fly species

Table 2.

Apple, Malus domestica; orn haw, Crataegus monogyna; Suks haw, Crataegus suksdorfii; black haw, Crataegus douglasii; snowberry, Symphoricarpos albus. a pom from C Wash were not included in statistical analyses because of low sample size. b df ⫽ 6, 404. Means within columns followed by same letters are not signiÞcantly different (P ⬎ 0.0024; LSD test, Bonferroni adjustment).

F P R2 Maximal % greater in R. pomonella than R. zephyria

apple, SW Wash apple, W Wash orn haw, SW Wash orn haw, W Wash Suks haw, SW Wash black haw, C Wash snw, SW Wash snw, C Wash

Host, area

Mean (SE) measurements (millimeters) of female R. pomonella (pom) and R. zephyria (zeph) reared from host fruit collected in 2006 and 2007 from different areas in Washington state

pom (n ⫽ 40) pom (n ⫽ 60) pom (n ⫽ 50) pom (n ⫽ 40) AM (n ⫽ 50) poma (n ⫽ 5) zeph (n ⫽ 40) zeph (n ⫽ 131) One-way ANOVAb

Fly species

Table 1.

November 2009 YEE ET AL.: DISCRIMINATION OF R. pomonella AND R. zephyria 1019

1020

ANNALS OF THE ENTOMOLOGICAL SOCIETY OF AMERICA

Vol. 102, no. 6

Female Flies SW Wash pom apple W Wash pom apple SW Wash pom orn haw W Wash pom orn haw SW Wash pom Suks haw SW Wash zeph snw C Wash zeph snw

6

4

CV2

2

0

-2

-4

-6 -6

-4

-2

0

2

4

6

CV1 Fig. 4. Scatter plot from canonical variates analysis of body measurements of 411 female R. pomonella (pom) and R. zephyria (zeph) reared from Þeld-infested fruit collected from different areas in Washington in 2006 and 2007.

males, and R2 for all measures were lower than in females. CVA of Body Measurements. CVA divided female ßies into roughly two distinct groups, one for each species (Fig. 4). However, CV1 scores of SW Wash pom orn haw ßies overlapped those of R. zephyria, especially of the slightly larger C Wash zeph snw ßies. Mean CV1 scores between all female R. pomonella and R. zephyria differed, as did those of SW and C Wash zeph snw ßies (Table 3). CV1 discriminated groups better (F ⫽ 318.03; df ⫽ 6, 404; P ⬍ 0.0001; R2 ⫽ 0.825) than CV2 (F ⫽ 46.40; df ⫽ 6, 404; P ⬍ 0.0001; R2 ⫽ 0.408). The linear combination of centered variables CV1 (females) ⫽ 9.26⫻ ovipositor length ⫹ 5.13⫻head width ⫹ 1.44⫻ intereye distance ⫹ 0.76⫻ Table 3. Mean CV1 scores ⴞ SE based on nine body measurements of female and male R. pomonella (pom) and R. zephyria (zeph) from different hosts from western and central Washington state Group

Females

Males

SW Wash pom apple W Wash pom apple SW Wash pom orn haw W Wash pom orn haw SW Wash pom Suks haw SW Wash zeph snw C Wash zeph snw

2.146 ⫾ 0.158A 2.256 ⫾ 0.129A 0.762 ⫾ 0.141B 1.935 ⫾ 0.158A 1.802 ⫾ 0.141A ⫺2.977 ⫾ 0.158D ⫺2.349 ⫾ 0.087C

0.966 ⫾ 0.158AB 1.466 ⫾ 0.158A 0.104 ⫾ 0.158C 0.028 ⫾ 0.158C 0.460 ⫾ 0.158BC ⫺1.403 ⫾ 0.158D ⫺1.620 ⫾ 0.158D

Host fruit codes same as in Table 1. Means within columns followed by the same letter are not signiÞcantly different (P ⬎ 0.0024; LSD test, Bonferroni adjustment).

wing length ⫺ 0.79⫻ wing band 2 ⫺ 1.72⫻ wing band 3 ⫹ 0.85⫻ femur ⫺ 0.77⫻ tibia ⫺ 1.85⫻ tarsus separated female R. pomonella groups and R. zephyria most effectively, based on raw canonical coefÞcients (SAS Institute 1999). Even greater overlap between species is seen if the Þve C Wash pom black haw ßies are included (data not shown). CVA did not clearly divide male ßies into species because there was much overlap in observations (Fig. 5). However, mean CV1 scores of all male R. pomonella differed from those of R. zephyria (Table 3). There was no difference between R. zephyria populations. CV1 discriminated groups better (F ⫽ 52.65; df ⫽ 6, 273; P ⬍ 0.0001; R2 ⫽ 0.536) than CV2 (F ⫽ 18.17; df ⫽ 6, 273; P ⬍ 0.0001; R2 ⫽ 0.285). The combination of centered variables CV1 (males) ⫽ 3.82⫻ genitalia length ⫹ 2.62⫻ head width ⫺ 6.22⫻ intereye distance ⫹ 0.59⫻ wing length ⫺3.61⫻ wing band 2 ⫺ 2.59⫻ wing band 3 ⫹ 12.28⫻ femur ⫺ 2.57⫻ tibia ⫹ 1.15⫻ tarsus separated male R. pomonella groups and R. zephyria most effectively. In comparisons of females (Fig. 4) with males (Fig. 5), it was evident that body measurements were better at discriminating females than males of R. pomonella and R. zephyria. Discriminant Analyses (DAs) Using Body Measurements. Use of discriminant analysis (DA) (Table 4) resulted in 95.4% of 240 total female R. pomonella across groups (excluding Þve C Wash pom black haw ßies) being correctly classiÞed, with 11 ßies misclassiÞed, whereas only 84.1% were identiÞed correctly based on ovipositor length alone using the Westcott

YEE ET AL.: DISCRIMINATION OF R. pomonella AND R. zephyria

November 2009

1021

Male Flies SW Wash pom apple W Wash pom apple SW Wash pom orn haw W Wash pom orn haw SW Wash pom Suks haw SW Wash zeph snw C Wash zeph snw

6

4

CV2

2

0

-2

-4

-6 -6

-4

-2

0

2

4

6

CV1 Fig. 5. Scatter plot from canonical variates analysis of body measurements of 280 male R. pomonella (pom) and R. zephyria (zeph) reared from Þeld-infested fruit collected from different areas in Washington in 2006 and 2007.

overlap zone (⬎0.98 mm in R. pomonella). In addition, 71.7% of the 38 R. pomonella with ovipositors in the Westcott zone were correctly identiÞed using DA. Eight of the 11 misclassiÞed R. pomonella were from the SW Wash pom orn haw group (pom3, Table 4) and had ovipositor lengths of 0.74 Ð 0.93 mm. DA also resulted in 100% of 171 total female R. zephyria being correctly classiÞed (Table 4), again better that the 97.7% so classiÞed based on ovipositor length alone

(⬍0.88 mm in R. zephyria). MisclassiÞcations of females using DA arose from small R. pomonella grouped as R. zephyria and not from large R. zephyria grouped as R. pomonella. MisclassiÞcation of males of R. pomonella and R. zephyria occurred at higher rates than for females. DA of male ßies (Table 4) resulted in 89.5% of R. pomonella being correctly classiÞed, but only 75.0% of R. zephyria being correctly classiÞed. Although male

Table 4. Discriminant analysis classification of 411 female and 280 male R. pomonella (pom) and R. zephyria (zeph) based on nine body measurements; flies misclassified to species in bold A priori assignment Female ßies pom1 pom2 pom3 pom4 pom5 zeph1 zeph2 Male ßies pom1 pom2 pom3 pom4 pom5 zeph1 zeph2

A posteriori assignments 关numbers (%)兴 pom1

pom2

pom3

pom4

pom5

zeph1

zeph2

Totals

11 (27.50) 17 (28.33) 3 (6.00) 1 (2.50) 7 (14.00) 0 (0) 0 (0)

9 (22.50) 14 (23.33) 0 (0.00) 20 (50.00) 2 (4.00) 0 (0) 0 (0)

2 (5.00) 1 (1.67) 22 (44.00) 1 (2.50) 12 (24.00) 0 (0) 0 (0)

7 (17.50) 24 (40.00) 2 (4.00) 16 (40.00) 5 (10.00) 0 (0) 0 (0)

10 (25.00) 4 (6.67) 15 (30.00) 1 (2.50) 23 (46.00) 0 (0) 0 (0)

1 (2.50) 0 (0) 6 (12.00) 1 (2.50) 1 (2.00) 15 (37.50) 100 (76.34)

0 (0) 0 (0) 2 (4.00) 0 (0) 0 (0) 25 (62.50) 31 (23.66)

40 60 50 40 50 40 131

5 (12.50) 7 (17.50) 6 (15.00) 7 (17.50) 9 (22.50) 1 (2.50) 0 (0)

11 (27.50) 28 (70.00) 6 (15.00) 9 (22.50) 0 (0) 0 (0) 0 (0)

5 (12.50) 1 (2.50) 9 (22.50) 6 (15.00) 10 (25.00) 3 (7.50) 1 (2.50)

8 (20.00) 1 (2.50) 4 (10.00) 6 (15.00) 4 (10.00) 6 (15.00) 6 (15.00)

9 (22.50) 1 (2.50) 7 (17.50) 6 (15.00) 14 (35.00) 2 (5.00) 1 (2.50)

2 (5.00) 1 (2.50) 5 (12.50) 3 (7.50) 3 (7.50) 16 (40.00) 11 (27.50)

0 (0) 1 (2.50) 3 (7.50) 3 (7.50) 0 (0) 12 (30.00) 21 (52.50)

40 40 40 40 40 40 40

pom1, SW Wash apple; pom2, W Wash apple; pom3, SW Wash orn haw; pom4, W Wash orn haw; pom5, SW Wash Suks haw; zeph1, SW Wash snw; zeph2, C Wash snw. Host fruit codes same as in Table 1.

1022

ANNALS OF THE ENTOMOLOGICAL SOCIETY OF AMERICA

Vol. 102, no. 6

SW Wash pom apple W Wash pom apple SW Wash pom orn haw W Wash pom orn haw SW Wash pom Suks haw C Wash pom black haw SW Wash zeph snw C Wash zeph snw 6

Females 4

CV2

2

0

-2

-4 -6

-4

-2

0

2

4

6

8

CV1

Fig. 6. Scatter plot from canonical variates analysis of wing shapes of six groups of female R. pomonella (pom) and two groups of R. zephyria (zeph) reared from Þeld-infested fruit from different areas in Washington in 2006 and 2007.

P ⫽ 0.0021 for axes 1Ð5). Bookstein coordinates of the two species (data not shown) and thin plate spline deformation grids (data not shown) demonstrate that the wing of female R. pomonella is more tapered at the tip than that of female R. zephyria, giving it an overall narrower appearance than the wing of R. zephyria, which is broader at the terminus (Fig. 1E and F). Jackknifed groupings from the CVA-distance based method for all eight female ßy groups (Table 5), by using all distinct CVA axes, showed that more individuals were usually assigned to their a priori groups than to other groups, but that there were many misclassiÞcations among conspeciÞc populations of R. pomonella, and some incorrect conspeciÞc assignments between the two R. zephyria populations. Based on the CVA plot (Fig. 6) and Table 5, wings of 11% (two of 18) of C Wash pom black haw ßies resembled

R. pomonella were generally larger than male R. zephyria (Table 2), this analysis indicated that there were relatively fewer small male R. pomonella and relatively larger male R. zephyria producing the overlap, the opposite of the pattern seen in females. Species classiÞcation of males by using DA was much weaker than of females. Wing Shapes of Female Flies. CVA separated the wing shapes of six populations of female R. pomonella and the two populations of R. zephyria along axis 1 (CV1), although there was overlap between the two species (Fig. 6); the overlap was due to a few C Wash pom black haw ßies falling between the R. pomonella and R. zephyria clusters (Fig. 6). CVAGen using BartlettÕs test indicated there were actually Þve statistically distinct axes or CVs that contributed to discrimination among the eight groups (P ⬍ 2.2204e⫺016 to Table 5.

Jackknifed groupings of female R. pomonella (pom) and R. zephyria (zeph) wing shapes from CVA-distance based method

A priori assign. pom1 pom2 pom3 pom4 pom5 pom6 zeph1 zeph2

A posteriori assignments pom1

pom2

pom3

pom4

pom5

pom6

zeph1

zeph2

Total

16 6 10 8 5 1 0 0

2 24 1 16 4 2 0 0

9 4 19 5 12 0 0 0

10 11 4 15 2 2 0 0

10 1 13 1 25 0 0 0

3 4 2 5 2 10 0 1

0 0 0 0 0 2 29 13

0 0 2 0 0 0 11 44

50 50 50 50 50 18 40 58

pom1, SW Wash apple; pom2, W Wash apple; pom3, SW Wash orn haw; pom4, W Wash orn haw; pom5, SW Wash Suks haw; pom6, C Wash black haw; zeph1, SW Wash snw; zeph2, C Wash snw. Host fruit codes same as in Table 1. Flies misclassiÞed to species in bold.

YEE ET AL.: DISCRIMINATION OF R. pomonella AND R. zephyria

November 2009

1023

Table 6. MANOVA results testing wing shape variability among female R. pomonella (pom) and R. zephyria (zeph) groups in CV1 axis scores from CVA Hypothesis of no species or group effects Test statistica

Value

F

Num df

Den df

P

WilksÕ lambda Source Model Error Corrected total

0.074197 df 7 358 365

136.23 Sum of squares 2272.98 358.00 2630.98

14 Mean square 324.72 1.0

714 F 324.71

⬍0.0001 P ⬍0.0001

Group SW Wash pom apple W Wash pom apple SW Wash pom orn haw W Wash pom orn haw SW Wash pom Suks haw C Wash pom black haw SW Wash zeph snw C Wash zeph snw

Mean CV1⫾ SE ⫺1.7044 ⫾ 0.1414CD ⫺1.9441 ⫾ 0.1414D ⫺1.6617 ⫾ 0.1414CD ⫺1.5104 ⫾ 0.1414CD ⫺1.2392 ⫾ 0.1414C 0.4045 ⫾ 0.2357B 4.4039 ⫾ 0.1581A 3.7854 ⫾ 0.1313A

Host fruit codes same as in Table 1. a Results of PillaiÕs Trace, Hotelling-Lawley Trace, and RoyÕs Greatest Root were also signiÞcant (P ⬍ 0.0001). Mean CV1 values followed by the same letter are not signiÞcantly different (P ⬎ 0.0018; LSD test, Bonferroni adjustment).

and were classiÞed as R. zephyria. However, overall, 98.5% (n ⫽ 268) of female R. pomonella and 99.0% (n ⫽ 98) of R. zephyria were correctly identiÞed to species (Table 5). MANOVA of CV1 (Table 6) indicated differences among groups, with R2 ⫽ 0.864 and with wing shapes of all R. pomonella being different from those of R. zephyria. Wing shapes of ßies from the W Wash pom apple and SW Wash pom Suks haw groups differed, and wing shape of the C Wash pom black haw group was different from wing shapes of all other R. pomonella (Table 6). A low P value from least squares means (P ⫽ 0.0028 before Bonferroni adjustment) suggested subtle differences between the two R. zephyria groups as well. CV2 also discriminated groups, but not as well as CV1 (MANOVA of CV2: F ⫽ 42.65; df ⫽ 7, 358; P ⬍ 0.0001; R2 ⫽ 0.455). (CV2s are not reported in the remainder of the article.) When the six populations of female R. pomonella were combined (including C Wash pom black haw) as one group (n ⫽ 268) and the two female R. zephyria populations were combined as another group (n ⫽ 98), 96.6% of R. pomonella and 100% of R. zephyria were correctly identiÞed to species. MANOVA (CV1) indicated a difference between species (F ⫽ 1874.19; df ⫽ 1, 364; P ⬍ 0.0001; R2 ⫽ 0.837). The 18 “unknown” female R. pomonella that fell in the overlap zone were all correctly identiÞed as R. pomonella using the eight groups (assignments table not shown). They were assigned to Þve groups of R. pomonella, but not to the C Wash pom black haw group. The one unknown female R. zephyria was also correctly identiÞed (table not shown). Wing Shapes of Male Flies. As with females, CVA separated the wing shapes of the six populations of male R. pomonella and the two populations of male R. zephyria, and a few ßies from the C Wash pom black haw group again fell between the two major species clusters (Fig. 7). BartlettÕs test indicated there were four statistically distinct axes or CVs that were effective group discriminators (P ⬍ 2.2204e⫺016 to P ⫽

0.0004 for axes 1Ð 4). Bookstein coordinates and thin plate spline deformation grids for the two species (data not shown) indicate that the wing of male R. pomonella is more tapered near the tip of the wing and is narrower than that of male R. zephyria (Fig. 2). Jackknifed groupings from the CVA-distance based method using all eight male ßy groups (Table 7) showed there were misclassiÞcations within species, but that 98.8% (n ⫽ 160) of R. pomonella and 96.4% (n ⫽ 83) of R. zephyria were correctly identiÞed to species. MANOVA (CV1) showed that there were signiÞcant differences among groups (Table 8), with R2 ⫽ 0.847. R. pomonella groups differed from R. zephyria groups based on mean testing and the C Wash pom black haw group differed from W Wash pom orn haw and SW Wash pom apple groups (Table 8). When CVA of wing shape was performed on six male R. pomonella groups combined as one group (n ⫽ 160) and on the two R. zephyria groups combined as the other group (n ⫽ 83), 98.8% of R. pomonella and 96.4% R. zephyria were correctly identiÞed to species. MANOVA (CV1) indicated a difference between species (F ⫽ 1230.55; df ⫽ 1, 241; P ⬍ 0.0001; R2 ⫽ 0.836). Female Versus Male Wing Shapes. CVA of wings of female and male R. pomonella (Fig. 8A) showed that wing shapes of the sexes differed, which was conÞrmed using MANOVA (CV1) (F ⫽ 790.12; df ⫽ 1, 426; P ⬍ 0.0001; R2 ⫽ 0.650). This was also true of wings of female and male R. zephyria (Fig. 8B) (F ⫽ 537.41; df ⫽ 1, 179; P ⬍ 0.0001; R2 ⫽ 0.750). Examination of Figs. 1E and 2A and of 1F and 2B and of Bookstein shape coordinates (data not shown) indicated that differences in female and male wings within both species were very subtle, with differences in positions of landmarks four and Þve relative to 3. The apportionment of variation in canonical variates developed in the CVA from both body lengths and wing measurement were compared using a variance component analysis (Table 9). Males and females differ in the pattern of variation for body length measures with a low of 14% of the variation due to species

1024

ANNALS OF THE ENTOMOLOGICAL SOCIETY OF AMERICA

Vol. 102, no. 6

SW Wash pom apple W Wash pom apple SW Wash pom orn haw W Wash pom orn haw SW Wash pom Suks haw C Wash pom black haw SW Wash zeph snw C Wash zeph snw 6

Males 4

CV2

2

0

-2

-4 -6

-4

-2

0

2

4

6

8

CV1

Fig. 7. Scatter plot from canonical variates analysis of wing shapes of six groups of male R. pomonella (pom) and two groups of R. zephyria (zeph) reared from Þeld-infested fruit from different areas in Washington in 2006 and 2007.

ovipositor length to conÞrm the species of ßies. Based on the measurements of ßies reared from host fruit (see below), most of the trapped ßies in the overlap zone were probably R. pomonella, but this will depend on the mix of species at a given location. Of nine body measurements in females, ovipositor length contributed the most to discriminating R. pomonella and R. zephyria reared from fruit, supporting previous work (Bush 1966, Westcott 1982). However, there is more interspecies overlap in ovipositor lengths than reported previously, consistent with data from trapped ßies whose hosts were unknown in this study (Fig. 3A). SpeciÞcally, the upper limit of the observed overlap zone was 0.91 mm, the longest ovipositor of R. zephyria seen in our study. The shortest ovipositor we observed for R. pomonella reared from black hawthorn was 0.74 mm, representing a new

differences in females. In contrast, for male body lengths and male and female wing shapes, roughly two thirds of the variation was associated with species differences. When the C Wash pom black haw population was added to the wing shape analyses the variance components did not change appreciably for either sex (not shown). Wing shape data were more consistent across the sexes and showed a higher proportion of the variance due to species differences than among populations. Discussion More than 20% of Þeld-trapped R. pomonella or R. zephyria in Washington state had ovipositors in the Westcott (1982) overlap zone of 0.88 Ð 0.98 mm. This indicates there is need to use measures other than Table 7.

Jackknifed groupings of male R. pomonella (pom) and R. zephyria (zeph) wing shapes from CVA-distance based method

A priori assign. pom1 pom2 pom3 pom4 pom5 pom6 zeph1 zeph2

A posteriori assignments pom1

pom2

pom3

pom4

pom5

pom6

zeph1

zeph2

Total

7 4 4 6 6 0 0 0

3 9 3 12 5 1 1 0

7 1 10 2 12 0 2 0

7 12 2 8 4 0 0 0

5 3 9 2 3 1 0 0

0 3 3 0 1 3 0 0

1 0 0 0 0 1 27 13

0 0 0 0 0 0 11 29

30 32 31 30 31 6 41 42

pom1, SW Wash apple; pom2, W Wash apple; pom3, SW Wash orn haw; pom4, W Wash orn haw; pom5, SW Wash Suks haw; pom6, C Wash black haw; zeph1, SW Wash snw; zeph2, C Wash snw. Host fruit codes same as in Table 1. Flies misclassiÞed to species in bold.

YEE ET AL.: DISCRIMINATION OF R. pomonella AND R. zephyria

November 2009

1025

Table 8. MANOVA results testing wing shape variability among male R. pomonella (pom) and R. zephyria (zeph) groups in CV1 axis scores from CVA Hypothesis of no species or group effects Test statistica

Value

F

Num df

Den df

P

WilksÕ lambda Source Model Error Corrected total

0.094090 df 7 235 242

75.55 Sum of squares 1303.62 235.00 1538.62

14 Mean square 186.23 1.00

468 F 186.23

⬍0.0001 P ⬍0.0001

Group SW Wash pom apple W Wash pom apple SW Wash pom orn haw W Wash pom orn haw SW Wash pom Suks haw C Wash pom black haw SW Wash zeph snw C Wash zeph snw

Mean CV1 ⫾ SE ⫺2.1213 ⫾ 0.1826C ⫺1.5265 ⫾ 0.1768BC ⫺1.6316 ⫾ 0.1796BC ⫺1.8113 ⫾ 0.1826C ⫺1.4557 ⫾ 0.1796BC ⫺0.3266 ⫾ 0.4082B 2.9653 ⫾ 0.1562A 3.4026 ⫾ 0.1543A

Host fruit codes same as in Table 1. Results of PillaiÕs Trace, HotellingÐLawley Trace, and RoyÕs Greatest Root were also signiÞcant (P ⬍ 0.0001). Mean CV1 values followed by the same letter are not signiÞcantly different (P ⬎ 0.0018; LSD test, Bonferroni adjustment). a

lower limit for for the overlap zone and a signiÞcant problem for diagnosis because such ßies would be identiÞed as R. zephyria, demonstrating the need for improved discrimination procedures based on multiple body measurements. Thus, this new overlap zone is shifted to smaller sizes from the Westcott zone. The possibility that rearing method affects ovipositor length or general body size needs to be addressed. Bush (1966) and Westcott (1982) both cite Benjamin (1934), who suggested that deterioration of plums may reduce body size in R. pomonella. In the current study, mean ovipositor lengths of Þve R. pomonella groups (excluding central WA ßies) were 5Ð10% shorter than those of Oregon ßies caught in traps, which had a mean length of 1.14 mm (range, 0.92Ð1.38 mm), suggesting some rearing effect, but R. zephyria in the current study had similar ovipositor lengths as Oregon R. zephyria, which had a mean length of 0.81 mm (range, 0.72Ð 0.88 mm) (Westcott 1982). To determine whether there are effects of rearing from excised fruit on body size, future studies should allow ßies to develop in known host fruit in the Þeld, possibly by collecting puparia underneath isolated host plants or by bagging infested fruit on the tree just before larvae exit them, allow pupariation in the bags. Hind tibia length, intereye distance, head width, hind femur length, and wing length seemed to distinguish most female R. pomonella from R. zephyria, but none was better than ovipositor length. Ovipositor lengths were only weakly correlated with these measures, suggesting these measures can be used independently to help support identiÞcations of problematic ßies. There seemed to be no signiÞcant correlation between ovipositor length and wing length in Oregon R. pomonella (Westcott 1982). Body measurements of male R. pomonella and male R. zephyria differ, with tibia, wing, and femur length seeming to be the most reliable for separating males of the species. However, the differences in all body measurements between male R. pomonella and R. zephyria are smaller than those between females of the two species.

Discriminant analysis using multiple body measurements showed that small female R. pomonella are more likely to be misidentiÞed as R. zephyria than large R. zephyria are as R. pomonella. The observed greater size range in R. pomonella may be correlated with the high adaptability of this species to different hosts, although this variability was less in males than in females for unknown reasons. R. zephyria from southwestern Washington were slightly smaller than those from central Washington. Whether this is environmentally or genetically induced needs further study. We show for the Þrst time that wing shape is a good character to discriminate between R. pomonella and R. zephyria in Washington state, with ⱖ98.5% accuracy, even when the problematic C Wash pom black haw ßies are included. Wing shape should also discriminate R. pomonella and R. zephyria from elsewhere in North America, based on published photographs of R. pomonella wings (Bush 1966) and on neural network analysis in these species (Bi et al. 2007). Wing shape analysis also may be useful in discriminating other ßies in the pomonella species group of uncertain status, such as the “ßowering dogwood ßy,” “sparkleberry ßy,” “plum ßy,” and “mayhaw ßy” (Berlocher 2000). Use of wing shape to discriminate R. pomonella and R. zephyria provides higher successful classiÞcation rates than use of ovipositor length alone and is similar to or slightly better than use of multiple body measurements. Wing shape may have an advantage in that it may not be affected by rearing conditions as size might be. That ⬎70% of the variation observed in the Þrst canonical variate axis among ßies was due to differences between species and virtually none was due to population within species variation strongly supports the idea that wing shape is less sensitive to the size variation demonstrated among populations of both species. It would be valuable to compare results of geometric morphometrics with other types of wing analyses using vein measures or structure (Adsavakulchai et al. 1999, Bi et al. 2007) for the discrimination of tephritid fruit ßies, as similar re-

1026

ANNALS OF THE ENTOMOLOGICAL SOCIETY OF AMERICA

Vol. 102, no. 6

6

A. Rhagoletis pomonella

Females Males

4

CV2

2

0

-2

-4 -6

-4

-2

0

2

4

6

CV1 6

B. Rhagoletis zephyria

Females Males

4

CV2

2

0

-2

-4 -6

-4

-2

0

2

4

6

CV1

Fig. 8. Scatter plots from canonical variates analysis of wing shapes of (A) female and male R. pomonella and (B) female and male R. zephyria.

sults from multiple methods would substantiate conclusions about wing shape differences among sibling species. The subtle but signiÞcant differences in wing shapes of female central Washington R. pomonella from other R. pomonella and between ßies from SuksdorfÕs hawthorn in southwestern Washington and from apple in Puyallup suggest that host plant or area affected wing Table 9. Variance components for variates for the first canonical axes of the CVA for both length measures and geometric analyses of wing shapes for each sex of flies Variance component Popa Species Error a

Male length

Female length

Male wing shape

Female wing shape

Value

%

Value

%

Value

%

Value

%

0.122 2.047 0.912

4 66 30

0.833 0.186 0.351

61 14 26

0.031 2.847 0.978

1 74 25

0.011 2.626 0.992

0 72 27

Excludes population C Wash pom black haw.

shape. It is possible that there are host races of ßies in the two regions, given that ßies even in sympatry have evolved into host races in the eastern United States (Feder et al. 1993). Within Drosophila subobscura Collin, wing shape parameters can evolve swiftly independent of wing size (Santos et al. 2004), raising the possibility that R. pomonella wing shapes also could have evolved over relatively short periods. The mean wing shapes of female and male ßies differed, which justiÞes having separate analyses for males and females. Analysis using geometric morphometrics is highly sensitive, as we detected wing shape differences between sexes of R. pomonella, whereas use of wing length/width ratios as a measure of shape apparently could not (in Illinois ßies; mean ratios ⫾ SE: males, 8.34 ⫾ 0.10; females, 8.31 ⫾ 0.04) (Sivinski and Dodson 1992). It is unknown whether there is functional signiÞcance to the wing shape difference between sexes. Male R. pomonella and R. zephyria show minimal wing movements, if any, related to mat-

November 2009

YEE ET AL.: DISCRIMINATION OF R. pomonella AND R. zephyria

ing (Prokopy and Bush 1973, Tracewski and Brunner 1987), unlike in some other tephritid ßies. Our Þndings are important for the Washington apple industry because they indicate that use of multiple body measurements or wing shape provides a significant improvement over the use of ovipositor length alone for identifying female R. pomonella and R. zephyria, thus potentially reducing misidentiÞcations. If ovipositor lengths and surstlyi orientation are ambiguous, wing shape analysis probably should be used, followed by body measurements if ambiguity remains. In practice, a wing shape data base from known ßies can be maintained by a regulatory agency such as WSDA and used to classify problematic ßies (our data can be provided upon request). Morphometric programs for analysis and classiÞcation of unknowns used in this study are available online (Rohlf 2001, http:// www.canisius.edu/⬇sheets/morphsoft.html). With training, it takes ⬍30 min to photograph a wing of a ßy, digitize it, subject it to analysis, and classify the ßy. One caveat for using the method is that R. pomonella in central WA, at least in our sample sites in wild areas from black hawthorn, may not be identiÞed as accurately as ßies from western WA using our current wing data. This in part caused the 1.5% probability of misdiagnosis of R. pomonella. The possibility of using shapes of a combination of structures such as wings, ovipositors, and surstyli to discriminate 100% of ßies is currently being investigated. Acknowledgments We thank Janine Jewett, Robert Goughnour, Tom Coate, and Meralee Nash for invaluable Þeld and laboratory assistance; Ed Lisowski (Washington State Department of Agriculture) for assistance in the initial phases of the project; Ed Lisowski and David Horton (USDAÐARS) and anonymous reviewers for helpful comments on the manuscript; and the Washington Tree Fruit Research Commission for partial funding of the project.

References Cited Adsavakulchai, A., V. Baimai, W. Prachyabrued, P. J. Grote, and S. Lertlum. 1999. Morphometric study for identiÞcation of the Bactrocera dorsalis complex (Diptera: Tephritidae) using wing image analysis. Biotropica 13: 37Ð48. AliNiazee, M. T., and R. L. Penrose. 1981. Apple maggot in Oregon: a possible new threat to the Northwest apple industry. Bull. Entomol. Soc. Am. 27: 245Ð246. Benjamin, F. H. 1934. Descriptions of some native trypetid ßies with notes on their habits. U.S. Dep. Agr. Tech. Bull. 401. U.S. Dep. Agr., Washington, DC. Berlocher, S. H. 2000. Radiation and divergence in the Rhagoletis pomonella species group: inferences from allozymes. Evolution 54: 543Ð557. Bi, C., M. C. Saunders, and B. A. McPheron. 2007. Wing pattern-based classiÞcation of the Rhagoletis pomonella species complex using genetic neural networks. Int. J. Comput. Sci. Appl. 4: 1Ð14. Bookstein, F. L. 1991. Morphometric tools for landmark data: geometry and biology. Cambridge University Press, New York.

1027

Bush, G. L. 1966. The taxonomy, cytology, and evolution of the genus Rhagoletis in North America (Diptera, Tephritidae). Bull. Mus. Comp. Zool. 134: 431Ð562. Claude, J. 2008. Morphometrics with R. Springer, New York. Efron, B. 1982. The jackknife, the bootstrap and other resampling plans. Society for Industrial and Applied Mathematics, Philadelphia, PA. Feder, J. L., T. A. Hunt, and G. L. Bush. 1993. The effect of climate, host plant phenology, and host Þdelity on the genetics of apple and hawthorn-infesting races of Rhagoletis pomonella. Entomol. Exp. Appl. 69: 117Ð135. Gilchrist, A. S., and D. C. Crisafulli. 2006. Using variation in wing shape to distinguish between wild and mass-reared individuals of Queensland fruit ßy, Bactrocera tryoni. Entomol. Exp. Appl. 119: 175Ð178. Herna´ ndez, E. J., D. W. Roubik, and G. Nates-Parra. 2007. Morphometric analysis of bees in the Trigona fulviventris group (Hymenoptera: Apidae). J. Kans. Entomol. Soc. 80: 205Ð212. Klingenberg, C. P., G. S. McIntyre, and S. D. Zaklan. 1998. Left-right asymmetry of ßy wings and the evolution of body axes. Proc. R. Soc. Lond. 265: 1255Ð1259. Lathrop, F. H., and C. B. Nickels. 1932. The biology and control of the blueberry maggot in Washington County, ME. Technical Bulletin 275. 1Ð76. U.S. Dep. Agric., Washington, DC. Marsteller, S., D. C. Adams, M. L. Collyer, and M. Condon. 2009. Six cryptic species on a single species of host plant: morphometric evidence for possible reproductive character displacement. Ecol. Entomol. 34: 66 Ð73. Nolte, A. W., and H. David Sheets. 2005. Shape based assignment tests suggest transgressive phenotypes in natural sculpin hybrids (Teleostei, Scorpaeniformes, Cottidae). Front. Zool. doi:10.1186/1742-9994-2-11. (http:// www.frontiersinzoology.com/content/2/l/11). Prokopy, R. J., and G. L. Bush. 1973. Mating behavior in Rhagoletis pomonella (Diptera: Tephritidae). IV. Courtship. Can. Entomol. 105: 873Ð 891. Rohlf, F. J. 2001. tpsDig (version 1.3 1). (http://life.bio. sunysb.edu/morph). Rohlf, F. J., and F. L. Marcus. 1993. A revolution in morphometrics. Trends Ecol. Evol. 8: 129 Ð132. Santos, M., P. F. Iriarte, W. Ce´spedes, J. Balanya` , A. Fontdevila, and L. Serra. 2004. Swift laboratory thermal evolution of wing shape (but not size) in Drosophila subobscura and its relationship with chromosomal inversion polymorphism. J. Evol. Biol. 17: 841Ð 855. SAS Institute. 1999. SAS/STAT䉸 userÕs guide, version 8. SAS Institute, Cary, NC. Sivinski, J. M., and G. Dodson. 1992. Sexual dimorphism in Anastrepha suspensa (Loew) and other tephritid fruit ßies (Diptera: Tephritidae): possible roles of developmental rate, fecundity, and dispersal. J. Insect Behav. 5: 491Ð506. Tracewski, K. T., and J. F. Brunner. 1987. Seasonal and diurnal activity of Rhagoletis zephyria Snow. Melanderia 45: 27Ð32. Tracewski, K. T., J. F. Brunner, S. C. Hoyt, and S. R. Dewey. 1987. Occurrence of Rhagoletis pomonella (Walsh) in hawthorns, Crataegus, of the PaciÞc Northwest. Melanderia 45: 19 Ð25. [USDA] U.S. Department of Agriculture. 2008. National agricultural statistics service. (http://www.nass.usda. gov/Statistics_by_State/Washington/Historic_Data/fruit/ apples.pdf).

1028

ANNALS OF THE ENTOMOLOGICAL SOCIETY OF AMERICA

Wasbauer, M. S. 1963. Taxonomic discrimination of Rhagoletis pomonella from the eastern United States and Rhagoletis zephyria from California. Calif. Dep. Agric. T-9: 1Ð 6. Washington State University. 2008. Quarantine pests & phytosanitary certiÞcation. Tree Fruit Research & Extension Center, Wenatchee, WA. (http://entomology. tfrec.wsu.edu/Cullage_Site/Pages/QpestInfo.html). Westcott, R. L. 1982. Differentiating adults of apple maggot, Rhagoletis pomonella (Walsh) from snowberry maggot, R. zephyria Snow (Diptera: Tephritidae) in Oregon. PanPac. Entomol. 58: 25Ð30. Yee, W. L. 2008. Host plant use by apple maggot, western cherry fruit ßy, and other Rhagoletis species (Diptera:

Vol. 102, no. 6

Tephritidae) in central Washington state. Pan-Pac. Entomol. 84: 163Ð178. Yee, W. L., and R. B. Goughnour. 2008. Host plant use by and new host records of apple maggot, western cherry fruit ßy, and other Rhagoletis species (Diptera: Tephritidae) in western Washington state. Pan-Pac. Entomol. 84: 179 Ð193. Zelditch, M. L., D. L. Swiderski, H. D. Sheets, and W. L. Fink. 2004. Geometric morphometrics for biologists: a primer. Elsevier Academic, San Diego, CA. Received 8 May 2009; accepted 4 August 2009.