SPINNING-WING DECOYS IN MINNESOTA 993

Effects of spinning-wing decoys onflock behavior and hunting

vulnerability of mallards in Minnesota

Michael L. Szymanski and Alan D. Afton

Abstract Waterfowl managers in Minnesota and other states are concerned that increased kill ratesassociated with the use of spinning-wing decoys (SWDs) may negatively affect local breed-ing populations of mallards (Anas platyrhynchos). Accordingly, we conducted 219 experi-mental hunts to evaluate hunting vulnerability of mallards to SWDs during the 2002 duckseason in Minnesota. During each hunt, we tested 2 SWD treatments: 1) SWDs turned OFF(control), and 2) SWDs turned ON (experimental) during alternate 15-minute sampling peri-ods that were separated by 5-minute buffer periods. We found that mallard flocks (>1 duck)were 2.91 times more likely to respond (i.e., approach within 40 m of hunters), and sizes ofresponding mallard flocks were 1.25 times larger, on average, when SWDs were turned ONthan OFF. Mallards killed/hour/hunter/hunt averaged 4.71 times higher (P<0.001) whenSWDs were turned ON than OFF. More hatch-year (HY) and after-hatch-year (AHY) mal-lards were killed when SWDs were turned ON than OFF; however, AHYs were relatively lesslikely than were HYs to be killed with SWDs turned ON. We found no evidence that SWDsreduced crippling or allowed hunters to harvest relatively more drakes than hens. Using aworst-case scenario model, we predicted that if 47% and 79% of Minnesota hunters hadused SWDs in 2000 and 2002, respectively, Minnesota mallard harvests would haveincreased by a factor of 2. However, increasing use of SWDs by northern hunters may resultin a partial redistribution of annual mallard harvests if naive ducks are harvested upon initialexposures to SWDs, and those ducks that survive become habituated to SWDs, as suggest-ed by our results. Our study was confined to a single hunting season in Minnesota and thusdid not assess whether vulnerability of mallards to hunters using SWDs varied among yearsor geographically. A multi-year, flyway-wide study is needed to make stronger and more rig-orous inferences regarding potential changes in harvest distribution and annual harvest ratesof mallards due to increasing use of SWDs by hunters in North America.

Key words Anas platyrhynchos, crippling, duck hunting, flock behavior, hunting vulnerability, mal-lard, Minnesota, spinning-wing decoys

Waterfowl managers in Minnesota and other Successful nesting females and many hatch-yearstates are concerned that local mallard (Anas (HY) mallards are present on or near brood marsh-platyrhynchos) breeding populations may be more es at the beginning of the hunting season invulnerable to hunters using spinning-wing decoys Minnesota and may be especially vulnerable to(hereafter SWDs) than are migrant ducks and thus hunters (Gilmer et al. 1977, Kirby et al. 1989).local breeding populations may be negatively However, many hunters believe that SWDs are ben-affected by the use of SWDs (Szymanski 2004). eflcial because they increase hunter effectiveness

Address for Michael L. Szymanski: School of Renewable Natural Resources, Louisiana State University, Baton Rouge, LA 70803,USA; present address: North Dakota Game and Fish Department, 100 N. Bismarck Expressway, Bismarck, ND 58501, USA.Address for Alan D. Afton: United States Geological Survey, Louisiana Cooperative Fish and Wildlife Research Unit, Louisiana StateUniversity, Baton Rouge, LA 70803, USA; e-mail: AAfton@lsu.edu.

Wildlife Society Bullplin 2005. 'MM i:'l'j i • 1 Uil I I'ocr rc-ii-n-i-rl

994 VnMlth s.ic/c-M liiilk'ttn 2

by reducing crippling and enabling hunters to bet-ter select drakes over hens (Szymanski 2004).

Field studies in California (Eadie et al. 2001),Missouri (Humburg et al. 2002), and Manitoba(Caswell and Caswell 2004) indicated that SWDsincreased vulnerability of mallards to hunters.Eadie et al. (2001) reported that effectiveness ofSWDs declined throughout the 1999-2000 huntingseason in California because naive or HY duckswere harvested early in the season and becauseSWD effectiveness was diluted later in the seasonwhen a larger proportion of hunters used them.Accordingly, SWDs subsequently were prohibiteduntil 30 November in California (California Fish andGame Commission 2001). Similarly, use of SWDs onall public waters was restricted until the Saturdaynearest 8 October in Minnesota beginning in 2002(Minnesota Statutes 2002). Knowledge of the vul-nerability of mallards to hunters using SWDs isneeded to determine whether and when SWDsshould be further restricted in Minnesota.Accordingly, our general objectives were to quanti-fy effects of SWDs on 1) flock behavior of mallards,2) hunter success and effectiveness, 3) harvest com-position (by age and sex), and 4) determinewhether any of these effects differed between thefirst and second halves of the hunting season.

More specifically, we predicted that if mallardswere more vulnerable to hunters using SWDs, thenflock responses, sizes of responding flocks, and killrates of mallards would be higher when SWDs wereturned "ON" than "OFF" (c/. Olsen and Afton 2000,Caswell et al. 2003, Caswell and Caswell 2004).Similarly, we predicted that if ducks approachedcloser to hunters when SWDs were turned ON,then crippling rates and crippling proportions ofmallards would be lower when SWDs were turnedON than OFF. Given that HY ducks generally aremore vulnerable to hunters than are after-hatch-year (AHY) ducks (Anderson 1975, Cox et al. 1998,Pace and Afton 1999), we predicted that if mallardswere more vulnerable to hunters using SWDs, thenproportionally more AHY mallards would be killedwith SWDs turned ON than OFF. We also predictedthat if SWDs enabled hunters to better selectdrakes over hens, then proportionally more drakeswould be killed when SWDs were turned ON thanOFF based on hunter preference (Metz and Ankney1991). Finally, given apparent conflicting resultsregarding temporal variation in vulnerability ofducks to SWDs (Eadie et al. 2001, Caswell andCaswell 2004) and restrictions of SWDs early in the

hunting seasons in California and Minnesota, weexamined whether vulnerability of mallards toSWDs was relatively greater during the first half ofthe season in Minnesota.

Study areaWe conducted experimental hunts in 17

Minnesota counties, from 28 September-26November 2002 (Figure 1). We selected counties toconduct experimental hunts that ranked highestfor mallard and total duck harvests during years1995-2000 (United States Fish and Wildlife Service,unpublished harvest data; Szymanski 2004); we con-ducted a few hunts in other counties where stateand federal wildlife managers observed large con-centrations of mallards during the 2002 huntingseason. Specific counties (number of hunts) inwhich we conducted experimental hunts wereBecker (n = 20), Big Stone (n = 22), Clay (n = 1),Douglas (M = 4), Grant (ra = 4), Houston (n=15), LacQui Park (n = 4), Marshall (n=3), OtterTail (w = 51),Pope (n = 29), Stearns (n = 3), Stevens (n = 2), Swift(n = 4),Todd (n = 2), Traverse (n = 8), Wabasha (n =11), and Winona (n- 36).

MethodsTechnician training

We trained 3 research technicians and familiar-ized them with experimental hunt protocols ineastern North Dakota 1 week prior to the 2002Minnesota duck season. We used ducks harvestedin North Dakota to train technicians in aging andsexing techniques (Hochbaum 1942, Carney 1992)and terminated training when qualitative daily com-parisons of flock observations and ages and sexesrecorded were accurate and similar among allobservers (Szymanski 2004). Accordingly, weassumed that observer bias did not influence ourresults.

Hunter selectionWe contacted a random sample of hunters from

the 2001 Minnesota Harvest Information Program(HIP) database who resided in Minnesota countiesthat ranked highest for mallard and total duck har-vests during years 1995-2000 (Szymanski 2004).We also directly contacted hunters encountered atboat landings, cafes, gas stations, public huntingareas, and sporting-goods stores in these and near-by counties. Additionally, we posted informational

Spinning-wing decovs in Minnesota • S7\manski and Afton 995

Figure 1. Locations of Minnesota counties (shaded) whereexperimental hunts were conducted to evaluate hunting vul-nerability of mallards to spinning-wing decoys, 28September-26 November 2002.

flyers and handed out business cards to recruithunters as volunteers; thus, some hunters contact-ed us directly to participate in experimental hunts.

Experimental huntsWe compared 2 SWD treatments within each

experimental hunt: 1) SWDs turned OFF (control),and 2) SWDs turned ON (experimental). We ran-domized the start order of SWD treatments for eachexperimental hunt and alternated treatments dur-ing 15-minute (minimum) sampling periods withineach hunt (Olsen and Afton 2000, Caswell et al.2003). We extended the duration of some sampleperiods so that flocks still under observation at theend of a period could be scored with regard totheir response to decoy sets. Each experimentalhunt consisted of 4 (minimum) to 10 (maximum)sampling periods (i.e., 2 to 5 pairs of control andexperimental periods). Some hunts were limited to4 sampling periods as per our scientific collectingpermits (see below) or scheduling difficulties withvolunteer hunters. We used 5-minute buffer peri-ods between sampling periods to ensure that duckswere not responding to stimuli from previous sam-pling periods and so that all ducks killed during asample period were retrieved and marked.Additionally, we did not allow calling or shooting

during buffer periods. We flushed ducks thatresponded to decoy sets during buffer periods andexcluded them from analyses.

We attempted to conduct experimental huntseach day from 28 September-26 November 2002with a different group of 2 volunteer hunters foreach hunt. However, due to logistical and schedul-ing difficulties, 1-4 hunters volunteered per huntand a few hunters participated in multiple hunts.Totals of 73 (33%), 106 (48%), 39 (18%), and 1(<1%) of our experimental hunts were comprisedof 1,2, 3, and 4 volunteers hunters, respectively. Atotal of 326 hunters participated once, 36 huntersparticipated twice, and 5 hunters participated 3times. We conducted experimental hunts twice aday (as could be scheduled) at locations that vol-unteer hunters had selected and were open tohunting. We asked volunteer hunters to selectexact locations of their hunting blinds and decoys.We then placed 1 drake and 1 hen Mojo MallardSWDs (HuntWise, Bastrop, La.) within 15 m ofhunters at locations and directions of their choice.We then began experimental hunts when volun-teer hunters indicated that they were ready tobegin hunting.

We prohibited hunters from altering decoy setsor blind placement after each experimental huntbegan and encouraged them to hunt as they typi-cally would under ordinary hunting conditions.Furthermore, we asked hunters to follow all stateand federal duck hunting regulations with excep-tions provided by our state and federal scientificcollecting permits. Our permits allowed volunteerhunters under our supervision to 1) hunt overSWDs during the period of prohibition on publicwaters in Minnesota (28 September-5 October2002), and 2) shoot up to 1 daily bag limit ofducks/hunter/15-minute sampling period (i.e., 4daily bag limits/hunter/hunt maximum); weallowed each hunter to retain only 1 daily bag limitat the conclusion of each day. Additionally, we werepermitted to use remote controls to turn SWDseither ON or OFF for ensuing sample periods.

All mallard flocks (>1 duck) observed within 100m of hunters were included in the experiment todetermine flock response to decoy sets (i.e., flewwithin 40 m; Yes or No). Additionally, we recordedflock sizes, numbers of mallards killed by hunters(shot and retrieved), and numbers of mallards crip-pled by hunters (visibly hit by shot and notretrieved). We estimated distances from hunters toflocks using known distances from landmarks

9 9 6 W ttU/ifi- Jiullctfn JOUS \*.( ^:99

measured by Nikon Laser 400 rangefinders (NikonVision Company, Limited;Tokyo, Japan).

Following each experimental hunt, we deter-mined age and sex of mallards killed using presenceor absence of notched tail feathers, cloacal charac-teristics (Hochbaum 1942), and wing plumage(Carney 1992). We removed and retained 1 wingfrom each duck so that age and sex could be con-firmed at the Mississippi Flyway wingbee.

Statistical analysisFlock response. We used a mixed linear model

analysis using a binomial error term and logit linkfunction (GlimMix Macro; Littell et al. 1996) to testwhether relative proportions of mallard flocksresponding ([number of flocks approaching within40 m]/[number of flocks observed within 100m]/hunt) differed between SWD treatments (cate-gorical; SWDs ON or SWDs OFF), time of season(categorical; early [days 1-30] or late [days 31-60]),and the 2-way interaction. We used backward selec-tion procedures to eliminate all nonsignificant (P>0.05) terms from the full model, beginning with the2-way interaction. We compared mallard flockresponses during 132 hunts; 87 hunts lacked mal-lard flock observations for 1 of the SWD treatmentsand thus were excluded from analysis.

Size of responding flocks. We used a 2-way analy-sis of variance (ANOVA; PROC MIXED; Littell et al.1996) to test whether average sizes of respondingmallard flocks differed between SWD treatments,time of season, and the 2-way interaction. We log-transformed flock size to meet assumptions of nor-mality; least-squares means (95% CI) presented areback-transformed values. Model selection proce-dures were similar to those described for the analy-sis of flock response.

Kill rates. We used a mixed linear model analysiswith a poisson error term and log-linear link func-tion (GlimMix Macro; Littell et al. 1996) to testwhether mallard kill rates differed between SWDtreatments, time of season, and the 2-way interac-tion. We calculated mallards killed/hour/hunter foreach hunt when SWDs were turned ON and OFF.Model selection procedures were similar to thosedescribed for the analysis of flock response.

Ages. We used 4 separate chi-square tests of inde-pendence (Agresti 1996) to determine whethernumbers of AHY and HY mallards killed differedbetween SWD treatments and by time of season.We used logistic regression analysis (PROC LOGIS-TIC; SAS Institute 1999) to determine whether rela-

tive proportions of AHY and HY mallards differedbetween SWD treatments, time of season, sexes,and all 2-way interactions. For this analysis wescored AHY mallards as "1" and HY mallards as "0".We used backward selection procedures to elimi-nate all nonsignificant (P>0.05) terms from the fullmodel, beginning with the 2-way interactions.

Sexes. We used logistic regression analysis (PROCLOGISTIC; SAS Institute 1999) to determinewhether relative proportions of male and femalemallards killed differed between SWD treatments,time of season, ages, and all 2-way interactions. Forthis analysis we scored males as" 1" and females as"0". Model selection procedures were similar tothose described for the age analysis.

Crippling rates and proportions. We used amixed linear model analysis with a poisson errorterm and log-linear link function (GlimMix Macro;Littell et al. 1996) to test whether mallard cripplingrates differed between SWD treatments, time of sea-son, and the 2-way interaction. For this analysis wecalculated mallards crippled/hour/hunter for eachhunt when SWDs were turned ON and OFF. Modelselection procedures were similar to thosedescribed for the analysis of flock response.

We used another mixed linear model analysiswith a binomial error term and logit link function(GlimMix Macro; Littell et al. 1996) to test whethermallard crippling proportions ([total mallards crip-pled]/[total mallards hit by shot] /hunt) differedbetween SWD treatments, time of season, and the 2-way interaction. For this analysis we comparedmallard crippling proportions during 29 experi-mental hunts; 190 hunts lacked observations of mal-lards hit by shot for 1 of the SWD treatments andthus were excluded from analysis. Model selectionprocedures were similar to those described for theanalysis of flock response.

ResultsWe conducted 219 experimental hunts with

equal numbers of SWD treatments for a total of1,556 sampling periods. A total of 367 volunteerhunters participated in our experimental hunts. Wecontacted 70 (19%) of these hunters randomly fromHIP lists and 269 (73%) directly in the field; 28 (8%)hunters contacted us directly.

Flock responseTotals of 386 (43%) and 158 (22%) mallard flocks

approached within 40 m of hunters when SWDs

Spinning-wing decoys in Minnesota • Szyniiinski and Alton 997

Table 1. Numbers of mallard flocks observed within 100 m and subsequently approached with-in 40 m (%) of hunters by time of season with spinning-wing decoys turned ON and OFF, 28September-26 November 2002 in Minnesota.

Seasona

EarlyLateCombined

100 m

372530902

ON

40

177209386

m

(48%)(39%)(43%)

100 m

257474731

OFF

40

5999158

m

(23%)(21%)(22%)

100 m

6291,0041,633

Total

40

236308544

m

(38%)(31%)(33%)

a Early = 28 September-27 October 2002; Late = 28 October-26 November 2002.

were turned ON and OFF, respectively (Table 1). Ourfinal model indicated that mallard flock responsesdiffered between SWD treatments. The odds ratioindicated that mallard flocks were 2.91 times morelikely to respond when SWDs were turned ON thanOFF (F1 131 = 37.48, P< 0.001). However, flockresponses did not differ by time of season ^ 131 =0.62, .P=0.44), and the 2-way interaction also wasnot significant (Fh 13O=O.33,/)=O.57).

Size of responding flocksSizes of responding mallard flocks (n = 544)

ranged from 1-380 with a mean±SE, median, andmode of 6.43±0.93,2,and 1 individuals, respective-ly. Our final model indicated that sizes of respond-ing mallard flocks differed between SWD treat-ments (Flt 54l =4.90, P=0.027) and time of season(Fj 54 l = 11.18,P=0.001); the 2-way interaction wasnot significant (/?

1)540=0.21,/>=0.65). Respondingmallard flocks averaged 1.25 times larger in sizeduring periods when SWDs were turned ON (x =2.63,95% CI=2.36-2.93) than OFF (x=2.10,95% CI= 1.77-2.28). Responding flocks averaged 1.37times larger in size early (x = 2.75, 95% CI =2.37-3.18) than late in the season (:v = 2.01,95% CI= 1.77-2.28).

Kill ratesVolunteer hunters killed 221 mallards during

experimental hunts for an average of 0.53 mal-lards/hunter/hunt. Only 21 (5%) hunters killed a

Table 2. Numbers (%) of mallards killed and crippled (hit and not retrieved) by time of seasonwith spinning-wing decoys turned ON and OFF, 28 September-26 November 2002 in Minnesota.

Seasona

EarlyLateCombined

ON

78 (76%)98 (82%)

1 76 (80%)

Killed

OFF

24 (24%)21 (18%)45 (20%)

Total

102 (46%)119 (54%)221 (100%)

394786

ON

(76%)(82%)(80%)

Crippled

OFF

12 (24%)10(18%)22 (20%)

Total

51 (47%)57 (53%)

108 (100%)

Early = 28 September-27 October 2002; Late = 28 October-26 November 2002

daily bag limit duringexperimental hunts(i.e., 4 mallards, nomore than 2 hens).Mallards comprised43% of the total duckkill during experimen-tal hunts (Szymanski2004); 176 (80%) and45 (20%) of these werekilled when SWDs

were turned ON and OFF, respectively (Table 2).Our final model indicated that kill rates of mallardsdiffered between SWD treatments (7^ 218= 154.84,P< 0.001). Kill rates averaged 4.71 times higherwhen SWDs were turned ON (x = 0.227, 95% CI =0.176-0.293) than OFF (x - 0.048, 95% CI =0.035-0.067). Kill rates did not differ by time of sea-son (T7! 218= 1.20,P=0.28) and the 2-way interactionalso was not significant (F1>217=1.22,P=0.28).

AgesHunters killed 61 (69%) and 28 (31%) AHY mal-

lards with SWDs turned ON and OFF, respectively;115 (87%) and 17 (13%) HY mallards were killedwith SWDs turned ON and OFF, respectively (Table3). For both AHYs (xf=6.12,P=0.0l4) and HYs (%\= 36.38, P< 0.001), more ducks were killed withSWDs turned ON than OFF. Hunters killed 45 (51%)and 44 (49%) AHY mallards during the first and sec-ond halves of the season, respectively; 57 (43%) and75 (57%) HY mallards were killed during the firstand second halves of the season, respectively (Table3). For both AHYs (%\=O.OO5,P=O.94) and HYs (x\= 1.23,P=0.28), numbers of ducks killed did not dif-fer between the first and second halves of the sea-son.

The overall age ratio (HY:AHY) for mallards killedduring experimental hunts was 1.48; age ratioswere 1.89 and 0.61 with SWDs turned ON and OFF,respectively (Table 3)- Our final model indicatedthat relative proportions of AHYs and HYs killed

during experimentalhunts differed betweenSWD treatments. Theodds ratio indicatedthat when comparedto HYs, AHYs wererelatively less likelyto be killed withSWDs turned ON (oddsratio = 0.322;Wald %\ =

998 Wildlife Society Bulletin 2005.33(3):993-1001

Table 3. Numbers of HY, AHY, and age ratios (HY/AHY) of mallards killed by time of season withspinning-wing decoys turned ON and OFF, 28 September-26 November 2002 in Minnesota.

Seasona

EarlyLateCombined

HY

4966

115

ON

AHY

293261

Ageratio

1.692.061.89

HY

89

17

OFF

AHY

161228

Ageratio

0.500.750.61

HY

5775

132

Total

AHY

454489

Ageratio

1.271.701.48

Early = 28 September-27 October 2002; Late = 28 October-26 November 2002.

action also was not sig-nificant (Ft 27=0.64, P=0.43).

DiscussionVulnerability ofmallards tohunters usingSWDs

Waterfowl managers10.73, .P= 0.001). Relative proportions of AHYs andHYs killed did not differ between time of seasons(Wald %j=0.74,P=039) or sexes (Wald %\ =0.02, P= 0.92); none of the 2-way interactions were signifi-cant (all P> 0.23).

SexesThe overall sex ratio (male:female) for mallards

killed was 1.63; sex ratios were 1.71 and 1.37 whenSWDs were turned ON and OFF, respectively.Relative proportions of male and females killed didnot differ between SWD treatments (Wald %\ =0.30,P=0.58),time of season (Wald xf=2.04,P=0.15),orages (Wald %\ =0.11, P=0.74); none of the 2-wayinteractions were significant (all />>0.30).

Crippling ratesOverall, 86 (80%) and 22 (20%) mallards were

crippled when SWDs were turned ON and OFF,respectively (Table 2). Our final model indicatedthat crippling rates of mallards differed betweenSWD treatments (Fx 218 = 130.30, P< 0.001).Crippling rates averaged 5.10 times higher whenSWDs were turned ON (x = 0.102, 95% CI =0.077-0.135) than OFF (x = 0.020, 95% CI =0.013-0.028). However, crippling rates did not dif-fer by time of season (Fx 218 = 0.37,P=0.55); the 2-way interaction also was not significant (F1 217=

0.20, P= 0.66).

Crippling proportionsOverall, 262 (80%) and 67 (20%) mallards were hit

by shot when SWDs were turned ON and OFF,respectively; 86 (33%) and 22 (33%) of those hit byshot were crippled when SWDs were turned ONand OFF, respectively (Table 2). Overall, 33% (n =329) of mallards hit by shot were crippled (Table 2).Proportions of mallards crippled did not differbetween SWD treatments (i?x 28 = 0.76, P= 0.39) ortime of season (Fx 28=0.29,P=0.60); the 2-way inter-

in Minnesota and other states are concerned thatincreased kill rates associated with the use of SWDsmay negatively affect local breeding populations ofmallards (Szymanski 2004). Hunters may havegreater harvest opportunity if flocks of mallards aremore likely to respond to decoy sets containingSWDs. Our results and those of others (Eadie et al.2001, Caswell and Caswell 2004) strongly supportthe hypothesis that mallards are more vulnerable tohunters using SWDs.

As predicted, we found that mallard flocks weremore likely to respond to decoy sets when SWDswere turned ON than OFF, with size of respondingmallard flocks 1.25 times larger when SWDs wereturned ON. Also as predicted, we found that mal-lard kill rates averaged 4.71 times higher whenSWDs were turned ON than OFF. However, only 5%of volunteer hunters actually killed a daily bag limitof mallards during our experimental hunts.Furthermore, volunteer hunters, on average, killedonly 0.53 mallards/hunter/hunt, despite the poten-tial to exceed daily bag limits as allowed by our sci-entific collecting permits. Thus, despite increasedkill rates, use of SWDs in Minnesota did not guaran-tee achievement of a daily bag limit of mallards.However, given the large differential in kill ratesbetween SWD treatments and the large number ofMinnesota waterfowl hunters, the percentage ofhunters using SWDs could greatly influence mal-lard harvests in Minnesota.

HY ducks generally are more vulnerable tohunters than are AHY ducks (Anderson 1975, Coxet al. 1998, Pace and Afton 1999). We found thatmore AHY and HY mallards were killed whenSWDs were turned ON than OFF; however, AHYmallards were relatively less likely than were HYmallards to be killed with SWDs turned ON. Thus,HY mallards that survive their initial hunting seasonmay learn to avoid hunters using SWDs in subse-quent years.

Spinning-wing decoys in Minnesota • Szymanski and Afton 999

Potential effects of SWDs on mallardharvests in Minnesota

We modeled potential increases in mallard har-vests for various percentages of Minnesota huntersusing SWDs. Our simple predictive model wasbased on 1) the observed kill rate differential, 2)actual mallard harvests for Minnesota in 2000 and2002 (E. M. Martin and P. I. Padding, United StatesFish and Wildlife Service, unpublished HIP report),and 3) estimated percentages of hunters usingSWDs in Minnesota during 2000 and 2002 (Fultonet al. 2002, Schroeder et al. 2004). We first estimat-ed mallard harvests in 2000 and 2002 without theuse of SWDs (i.e., 197,740 and 141,705 mallards,respectively) and then predicted harvests, assuminga linear relationship, for various percentages ofhunters using SWDs for those years. Based on ourcalculations, 47% and 79% of hunters using SWDswould have been sufficient to double the 2000 and2002 Minnesota mallard harvests, respectively(Figure 2).

Given a general lack of information and a desireto present a worst-case scenario, we made theassumption that the relationship between SWD useand increases in mallard harvests was linear.However, we suspect that subsequent research will

ooo

1"38o

800 -

600 -

400 -

200 -

0

2002

—' ^

20 40 60

Percent of Minnesota duck hunters using SWDs

Figure 2. Predicted Minnesota mallard harvests with increasing use of spinning-wing decoys(SWDs) by Minnesota duck hunters using a worst-case-scenario model. Drop lines indicateactual Minnesota mallard harvests and percentages of hunters using SWDs in 2000 and 2002,and projected percentages of hunters using SWDs that would double harvest in those years(bold lines).

detect a curvilinear relationship between thesevariables because of 1) a possible negative relation-ship between mallard flock responses and percent-ages of hunters using SWDs (cf. Eadie et al. 2001),and 2) a possible decline in numbers of ducks avail-able to harvest as use of SWDs increases.Consequently, our predictions of percentages ofhunters using SWDs required to double Minnesotamallard harvests probably are biased low and dou-bling of the harvest may not be achievable even ifall duck hunters used SWDs in Minnesota.

Hunter selectivity and effectivenessusing SWDs

Hunters prefer to shoot drakes over hens (Metzand Ankney 1991), and many believe that SWDsenable them to better select drakes over hens byattracting ducks closer (Szymanski 2004).However, we found no evidence that drakes wererelatively more likely than were hens to be killed byvolunteer hunters in Minnesota when SWDs wereturned ON. Thus, we conclude that use of SWDsdid not allow hunters to better select drakes overhens. Furthermore, many hunters believe thatSWDs increase their effectiveness by decreasingcrippling (Szymanski 2004). However, we found

that mallard cripplingrates (cripples/hunter/hr/hunt) were higherwhen SWDs were turnedON than OFF, whichprobably was related torelatively greater shoot-ing opportunities whenSWDs were turned ON.In contrast, we found noevidence that mallardcrippling proportions dif-fered between «WD treat-ments. However, ouranalysis of crippling pro-portions was limited(power = 0.10; Universityof California, Los Angeles2002) by a relatively smallnumber of hunts (w = 29).Accordingly, we tentative-ly conclude that use ofSWDs did not increasehunter effectiveness inMinnesota.

80 100

1000 linlh-rin 2

Caswell and Caswell (2004) reported that mal-lard crippling proportions were lower when SWDswere turned ON than OFF during experimentalhunts in Manitoba. However, they did not analyzeindividual hunts as the experimental unit, and thusdifferent groups of hunters, possibly with differentshooting abilities and/or hunting situations, mayhave greatly influenced their results.

Time of seasonEadie et al. (2001) suggested that effectiveness of

SWDs declined later in the hunting season becausenaive ducks were harvested early in the season.Based on their results, the California Fish and GameCommission (2001) prohibited the use of SWDsuntil 30 November in California. In contrast, ourresults and those of Caswell and Caswell (2004)indicated that kill-rate differentials between SWDtreatments remain high throughout the huntingseason.

Local HY mallards frequently are located neartheir natal areas at the beginning of the hunting sea-son and comprise a large proportion of the killearly in the season in Minnesota (Gilmer et al. 1977,Kirby et al. 1989). We found that numbers of HYmallards killed during the first and second halves ofthe season did not differ; however, further researchis needed to determine whether harvest of localand migrant HY mallards differs by time of season.Our logistic regression analysis indicated that rela-tive proportions of AHY and HY mallards killed alsowere similar during the first and second halves ofthe season.

Management implicationsIf mallard harvests were to increase in other

states, as projected by our worst-case scenariomodel in Minnesota, then promulgation of restric-tive regulatory packages probably would increaseunder Adaptive Harvest Management models(Williams and Johnson 1995). Increasing use ofSWDs by duck hunters in Minnesota and othernorthern states could result in a partial redistribu-tion of annual mallard harvests if naive ducks areharvested upon initial exposures to SWDs andthose ducks that survive become habituated toSWDs (cf. Eadie et al. 2001). Indeed, our results sug-gest that AHY mallards have learned to avoid SWDs.However, our study was confined to a single hunt-ing season in Minnesota and thus did not assesswhether vulnerability of mallards to hunters using

SWDs differs among years or geographically. Amulti-year, flyway-wide study is needed to makestronger and more rigorous inferences regardingpotential changes in harvest distribution and annu-al harvest rates of mallards due to increasing use ofSWDs by hunters in North America.

Acknowledgments. We greatly appreciate thelogistical support and funding provided by J. S.Lawrence and the Minnesota Department ofNatural Resources. The United States GeologicalSurvey-Louisiana Cooperative Fish and WildlifeResearch Unit also provided funding, equipment,and vehicles. Cabela's East Grand Forks, Minnesota,provided spinning-wing decoys and chest-wadersat reduced prices. We thank managers of Sully's HillNational Game Preserve, Thief Lake WildlifeManagement Area (WMA), and Whitewater WMAfor providing housing. The United States Fish andWildlife Service (Regions 3 and 4); managers ofDetroit Lakes Wetland Management District(WMD), Fergus Falls WMD, Morris WMD, TamaracNational Wildlife Refuge and Upper MississippiNational Fish and Wildlife Refuge; and MinnesotaDepartment of Natural Resources kindly issuedaccess and scientific collection permits. Our studywas conducted under Animal Use and CareCommittee Protocol #AE02-12 at Louisiana StateUniversity. We thank P. I. Padding and the 2003Mississippi Flyway wingbee certified checkers forconfirming ages of mallard wings. We also thank C.D. Ankney, M. J. Anteau, S. D. Cordts, J. P. Geaghan, D.A. Haukos, J. S. Lawrence, D. P. Rave,V L.Wright, and2 anonymous reviewers for providing helpful com-ments on the manuscript. We indebted to C. T.Garrett, N. P.Jerstad, and B.W Meixell for their tirelessefforts in contacting hunters and conducting experi-mental hunts. Finally, we thank the many hunterswho participated in our experimental hunts.

literature citedAGRESTI, A 1996. An introduction to categorical data analysis.

Wiley and Sons, New York, New York, USA.ANDERSON, D. R. 1975. Population ecology of the mallard: V.

Temporal and geographic estimates of survival, recovery, andharvest rates. United States Fish and Wildlife Service,Resource Publication 125.

CALIFORNIA FISH AND GAME COMMISSION. 2001. 2001-02 California

waterfowl hunting regulations. California Fish and GameCommission, Sacramento, USA.

CARNEY, S. M. 1992. Species, age, and sex identification of ducksusing wing plumage. United States Department of theInterior, Fish and Wildlife Service, Washington, D.C., USA.

Spinning-wing decoys in Minnesota • SzynuiTski and Afton 1001

CASWELL, J. H., A. D. AFTON, AND F. D. CASWELL. 2003. Vulnerability of

non-target goose species to hunting with electronic snowgoose calls. Wildlife Society Bulletin 31:1117-1125.

CASWELL, J. H., AND F. D. CASWELL. 2004. Vulnerability of mallardsto hunting with a spinning-wing decoy in Manitoba. WildlifeSociety Bulletin 32:1297-1304.

Cox, R. R., JR., A. D. AFTON, AND R. M. PACE, III. 1998. Survival of

female northern pintails wintering in southwesternLouisiana. Journal of Wildlife Management 62:1512-1521.

EADIE, J. M., T. G. MOORE, AND J. T. ACKERMAN. 2001. Experimental

evaluation of the effect of a mechanical decoy (moto-duck)on hunting success and waterfowl response in California1999-2000. Final report to the California WaterfowlAssociation. University of California, Davis, USA.

FULTON,D.C.J.VLAMING,J.S.LAWRENCE,AND EW.PRICE. 2002. The

2000 waterfowl hunting season in Minnesota: a study ofhunters' opinions and activities. Final report to MinnesotaDepartment of Natural Resources. United States GeologicalSurvey-Minnesota Cooperative Fish and Wildlife ResearchUnit, University of Minnesota, St. Paul, USA.

GILMER, D. S.,R. E. KIRBY,I.J. BALL.ANDJ. H. REICHMANN. 1977. Post-

breeding activities of mallards and wood ducks in north-cen-tral Minnesota. Journal of Wildlife Management 41:345-359.

HOCHBAUM, H. A. 1942. Sex and age determination in waterfowlby cloacal examination. Transactions of the North AmericanWildlife Conference 7:299-307.

HUMBURG, D. D., D. A GRABER, AND A. H. RAEDEKE. 2002. Missouri

waterfowl status, 2002. Missouri Department ofConservation, Jefferson City, USA.

KIRBY, R.E.,L.M. COWARDIN, AND J.R TESTER. 1989. Premigrational

movements and behavior of young mallards and wood ducksin north central Minnesota. United States Fish and WildlifeService, Fish and Wildlife Research Report 5.

LlTTELL, R C , G. A. MlLLIKEN, W W. STROUP, AND R D. WOLFINGER.

1996. SAS system for mixed models. SAS Institute, Cary,North Carolina, USA.

METZ, K. J., AND C. D. ANKNEY. 1991. Are brightly coloured maleducks selectively shot by duck hunters? Canadian Journal ofZoology 69:279-282.

MINNESOTA STATUTES. 2002. 2002 Regular Session. Chapter 351-Senate File 2674, Section 20, Section 97B.811, Subdivision 4a.St. Paul, Minnesota, USA.

OLSEN, R. E., AND A D. AFTON. 2000. Vulnerability of lesser snowgeese to hunting with electronic calling devices. Journal ofWildlife Management 64:983-993-

PACE, R. M., HI, AND A D. AFTON. 1999. Direct recovery rates oflesser scaup banded in northwest Minnesota: sources of het-erogeneity. Journal of Wildlife Management 63:389-395.

SAS INSTITUTE. 1999. SAS/STAT user's guide, 8.02 edition. SASInstitute, Cary, North Carolina, USA.

SCHROEDER, S., D. C. FULTON, AND J. S. LAWRENCE. 2 0 0 4 . T h e 2 0 0 2

waterfowl hunting season in Minnesota: a study of hunters'opinions and activities. Final report to MinnesotaDepartment of Natural Resources. United States GeologicalSurvey-Minnesota Cooperative Fish and Wildlife ResearchUnit, University of Minnesota, St. Paul, USA.

SZYMANSKI, M. L. 2004. Effects of spinning-wing decoys on flockbehavior and hunting vulnerability of local and migrant mal-lards and other ducks in Minnesota. Thesis, Louisiana StateUniversity, Baton Rouge, Louisiana, USA. Available online athttp://etd.lsu.edu/docs/available/etd-01262004-143729/.[Date accessed 1 February 2005.]

UNIVERSITY OF CALIFORNIA, LOS ANGELES. 2002. Power calculator.

Available online athttp://calculators.stat.ucla.edu/power-calc. [Date accessed 1 February 2005.]

WILLIAMS, B. K., AND F. A JOHNSON. 1995. Adaptive management

and the regulation of waterfowl harvests. Wildlife SocietyBulletin 23:430-436.

Michael L (Mike) Szymanski (photo) is the migratory game birdbiologist for the North Dakota Came and Fish Department. Hereceived a B.S. (2001) in fisheries and wildlife biology from theUniversity of North Dakota and an M.S. (2004) in wildlife atLouisiana State University. Mike's research interests involveissues relating to harvest management of migratory game birds,waterfowl breeding habitat, and waterfowl migration. Alan D.(Al) Afton is Assistant Leader-Wildlife of the United StatesGeological Survey-Louisiana Cooperative Fish and WildlifeResearch Unit and adjunct professor of wildlife in the School ofRenewable Natural Resources at Louisiana State University. Alholds a B.S. in wildlife biology (1973) from Kansas StateUniversity, an M.S. in wildlife ecology (1977) from theUniversity of Minnesota, and a Ph.D. in biology-behavioralecology (1983) from the University of North Dakota. He and hisgraduate students focus their research on basic and appliedquestions concerning waterfowl ecology and the conservationand management of wetland habitats.

Associate editor: Haukos