Nest habitat selection by grassland birds : the role of vegetation structure and floristics

Transcription

1 University f Mntana SchlarWrks at University f Mntana Graduate Student Theses, Dissertatins, & Prfessinal Papers Graduate Schl 1999 Nest habitat selectin by grassland birds : the rle f vegetatin structure and flristics Steven T. Hekman The University f Mntana Let us knw hw access t this dcument benefits yu. Fllw this and additinal wrks at: Recmmended Citatin Hekman, Steven T., "Nest habitat selectin by grassland birds : the rle f vegetatin structure and flristics" (1999). Graduate Student Theses, Dissertatins, & Prfessinal Papers This Thesis is brught t yu fr free and pen access by the Graduate Schl at SchlarWrks at University f Mntana. It has been accepted fr inclusin in Graduate Student Theses, Dissertatins, & Prfessinal Papers by an authrized administratr f SchlarWrks at University f Mntana. Fr mre infrmatin, please cntact schlarwrks@ms.umt.edu.

2 I Maureen and Mike MANSFIELD LIBRARY Tlie University f IVIONXANA Pennissin is granted by the authr t reprduce this material in its entirety, prvided that this material is used fr schlarly purpses and is prperly cited in published wrks and reprts. * * Please check ''Yes" r "N" and prvide signature * * Yes, I grant permissin N, I d nt grant permissin Authr s Signature Date Any cpying fr cmmercial purpses r financial gain may be undertaken nly with the authr s explicit cnsent.

3

4 Nest Habitat Selectin by Grassland Birds: The Rle f Vegetatin Structure and Flristics By Steven T. Hekman B.S., Davidsn Cllege, 1989 Presented in partial fulfillment f the requirements fr a degree f Master f Science THE UNIVERSITY OF MONTANA 1999 Apprved by: Chairman, Bard f Examiners Dean, Graduate Schl Date 12- s-yr

5 UMI Number: EP776 All rights reserved INFORMATION TO ALL USERS The quality f this reprductin is dependent upn the quality f the cpy submitted. In the unlikely event that the authr did nt send a cmplete manuscript and there are missing pages, these will be nted. Als, if material had t be remved, a nte will indicate the deletin. UMT UMI EP776 Published by PrQuest LLO (201). Cpyright in the Dissertatin held by the Authr. Micrfrm Editin PrQuest LLO. All rights reserved. This wrk is prtected against unauthrized cpying under Title 17, United States Cde uest* PrQuest LLC. 789 East Eisenhwer Parkway P.O. Bx 146 Ann Arbr, Ml

6 ABSTRACT Hekman, Steven T., M.S., Fall 1999 Wildlife Bilgy Nest Habitat Selectin by Grassland Birds: The Rle f Vegetatin Structure and Flristics (49pp.). Directr: Dr. IJ. Ball I studied nest habitat selectin f a grassland bird cmmunity in fields f ungrazed, cl seasn grasses in westcentral Mntana. Bird species included Cinnamn Teal {Anas cyanplera\ Gadwall (A. strepera). Mallard (A. platyrhynchs). Nrthern Shveler {A. clypeata). Shrt-eared Owl {Asi flammeus). Savannah Sparrw {Passerculus sandwichensis), and Western Meadwlark {Stumella neglectd). Vegetatin characteristics (structure, structural hetergeneity, and flristics) were sampled at the nest site (<50 cm radius arund nest) and nest patch (10 m radius). I cmpared nests f each species t available vegetatin and cmpared selectin amng species. Each species selected nest sites nnrandmly {P <0.00 in all cases). Nest site chice differed amng species: 9 f 10 pairs f species were significantly different {P <0.10) in 1996, and 20 f 21 in Canpy cver, hetergeneity f plant grwth frm, and vegetatin density at varying heights best discriminated amng species. At nest patches, selectin by each species and differences amng species were relatively weak and generally reflected patterns at nest sites. Fr mst ducks and Shrt-eared Owls, vegetatin vlume immediately arund nests was significantly lwer {P <0.10) t the sutheast than t the nrthwest r suthwest. I cnclude that nest habitat selectin was perating primarily at a fine scale (the nest site). Nest site selectin was nnrandm and was expressed thrugh selectin f plant grwth frm, placement f the nest relative t vegetatin, and active manipulatin f vegetatin by birds. I suggest that managers need t prvide fine scale diversity f vegetatin characteristics within fields because f increased need t manage grasslands fr mutual benefits t a diversity f breeding birds. 11

7 ni ACKNOWLEDGMENTS Funding and lgistic supprt fr this study was prvided by the U. S. Fish and Wildlife Service; Mntana Department f Fish, Wildlife, and Parks; U. S. Gelgical Service - Nrthern Prairie Science Center; U. S. Gelgical Service - Cperative Units Prgram; Cnfederated Salish and Ktenai Tribes; Prairie Pthle Jint Venture; and the University f Mntana. The U. S. Fish and Wildlife Service and Mntana Department f Fish, Wildlife, and Parks prvided access t federal and state land. I wuld like t acknwledge the guidance f my graduate cmmittee: Dr. I. Je Ball, Dr. Tm E. Martin, and Dr. L. Sctt Mills. Fr their dedicatin and tireless assistance, I thank the many peple wh aided with my field wrk: Thmas F. Fndell, Shannn H. Gamer, Erik M. Hess, Kristpher E. Jnes, Andrew K. Knapp, Elisabeth S. Leaf, Braun E. Lwry, David A. Miller, Gregry E. Parkhurst, Stephanie L. Schmitz. Fr helpful discussin and cmments n drafts f this thesis I thank Dr. Ragan M. Callaway Dr. David A. Pattersn, and numerus graduate students

8 IV TABLE OF CONTENTS ABSTRACT...ii ACKNOWLEDGMENTS... iii LIST OF TABLES... v LIST OF FIGURES... vi INTRODUCTION... 1 STUDY AREA AND METHODS... Study Area and Species... Methds... 4 Statistical Analysis...7 RESULTS... 9 Nest Vegetatin Preferences...9 Differences in Preferences Amng Species...17 DISCUSSION Nest Site Preferences...26 Differences in Preferences Amng Species Ptential Adaptive Significance f Preferences at Nest Sites... 1 Management Implicatins...2 LITERATURE CITED... 5 APPENDIX A APPENDIX B... 46

9 LIST OF TABLES Table 1. Variables used in statistical analysis f nest site and nest patch characteristics. 8 Table 2. Summary results f DFA in separating nest site frm available vegetatin fr grassland birds in westcentral Mntana in 1996 and Table. Structure crrelatins frm DFA separating nest sites and available vegetatin fr grassland birds in westcentral Mntana in Table 4. Structure crrelatins frm DFA separating nest sites and available vegetatin fr grassland birds in westcentral Mntana in Table 5. Summary results f DFA in separating nest site frm available vegetatin fr grassland birds in westcentral Mntana in 1996 and Table 6. Structure crrelatins frm DFA separating nest patches and available vegetatin fr grassland birds in westcentral Mntana in Table 7. Structure crrelatins frm DFA separating nest patches and available vegetatin fr grassland birds in westcentral Mntana in Table 8. Cn^arisns f nest site characteristics between pairwise cmbinatins f grassland birds in westcentral Mntana in Table 9. Rate f crrect classificatin f nest sites frm DFA f 7 grassland birds in 1996 and 1997 based n separate cvariance matrices...19 Table 10. Cmparisns f nest site characteristics between pairwise cmbinatins f grassland birds in westcentral Mntana in Table 11. Cmparisns f nest patch characteristics between pairwise cmbinatins f grassland birds in westcentral Mntana in Table 12. Rate f crrect classificatin f nest patches frm DFA f grassland birds in 1996 and 1997 based n separate cvariance matrices...24 Table 1 Cmparisns f nest patch characteristics between pairwise cmbinatins f grassland birds in westcentral Mntana in Table Al. Mean vegetatin characteristics and univariate /^statistics fr nest sites f 5 grassland birds in westcentral Mntana, Table A2. Mean vegetatin characteristics and univariate /^statistics fr nest sites f 7 grassland birds in westcentral Mntana, Table A. Mean vegetatin characteristics and univariate F-statistics fr nest patches f 5 grassland birds in westcentral Mntana, Table A4. Mean vegetatin characteristics and univariate F-statistics fr nest patches f 7 grassland birds in westcentral Mntana, Table Bl. Mean vegetatin characteristics relative t nearby available habitat fr nest habitat f 5 grassland birds in westcentral Mntana, Table B2 Mean vegetatin characteristics relative t nearby available habitat fr nest habitat f 5 grassland birds in westcentral Mntana,

10 VI LIST OF FIGURES Figure 1. Discriminant functin analysis f nest sites f grassland birds in westcentral Mntana, 1996 and Figure 2 Discriminant functin analysis f nest patches f grassland birds in westcentral Mntana, 1996 and

11 INTRODUCTION Cnsiderable cnservatin effrt t manage upland vegetatin fr birds in grassland-wetland systems f the nrthern prairies has fcused n prviding tall, dense, mntypic stands f vegetatin This gal has largely been pursued by limiting disturbance n managed grasslands r thrugh restratin prgrams planting grass r grass/legume mixtures, cmmnly referred t as Dense Nesting Cver, r DNC (Higgins and Barker 1982, Kirby et al. 1992, Hartley 1994, Reynlds et al. 1994). This emphasis has ccurred largely because ducks were thught t benefit fi*m such cnditins (Duebbert 1969, Duebbert and Lkemen 1976), but site-specific benefits expected fi*m this apprach ften have nt accrued (Klett et al. 1988, Clark and Nudds 1991, Sargeant 1996). Additinal reasns t suggest that fundamental re-evaluatin f grassland cnservatin practices is in rder include recent evidence f substantial ppulatin declines in many species f grassland birds (Jhnsn and Schwartz 199, Knpf 1995) and grwing interest in prviding suitable habitat fr a diversity f bird species in grasslands (Tme et al. 1994, Hartley 1994, Reynlds et al. 1994). Despite the imprtance f detailed knwledge f habitat preferences by individual species fr develping unambiguus targets fr grassland management, much f the available infrmatin n this tpic is difficult t synthesize meaningfully. Mst studies which that measured vegetatin characteristics address nly ne (r a few clsely related) species, and methds used t measure and analyze vegetatin usually differs amng studies. Differing eclgical scales (Wiens et al. 1987) further cmplicate attempts t cmpare amng studies, and hence amng species. Finally, habitat preferences f 1

12 grassland birds ften have been inferred frm density r ccupancy f breeding birds in relatin t habitat manipulatins such as grazing, haying, r burning (Kantrud 1981, Pylypec 1991, Gilbert et al. 1995, Kruse and Bwen 1996, Dale et al 1997). Apparent respnses by individual species t specific manipulatins ften differ, hwever, prbably because birds respnd t changes in vegetatin structure rather than t the manipulatin per se (Wiens 197, Kantrud 1981, Kirby et al. 1992, Saab et al. 1995, but see Bwen and Kruse 199). Differences in the timing and intensity f a manipulatin may alter the effect (Bwen and Kruse 199, Kruse and Bwen 1996), and manipulatin may have different effects different habitats r misture regimes (Wiens 197, Bindini et al. 1998) Habitat preferences may be reflected at multiple eclgical scales (Jhnsn 1980, Ktliar and Wiens 1990). At intermediate scales, crrelating presence r density f breeding birds t sme measure f average structure within a field is cmmnly used t demnstrate general habitat preferences by grups f species with similar preferences (Rtenberry and Wiens 1980, Renken and Dinsmre 1987, Delisle and Savidge 1997). Hwever, wide verlap in preferences is cmmn at such brad scales; preferences f individual species can nt be discerned, and respnses t specific manipulatins are difficult t predict (Rtenberry and Wiens 1980). I therefre chse the mre pwerful apprach f measuring vegetatin at sites used by individual birds (James 1971, Larsn and Bck 1986, Sedgwick and Knpf 1992) t investigate habitat preferences f nesting birds. Because selectin f nest habitat may invlve the area immediately arund the nest site r a much larger area (Martin and Rper 1988, Knpf and Sedgwick 1994, Badyaev 1995), I sampled vegetatin at the scale f the nest site (<50 cm arund nest) and nest

13 patch (10 m frm nest). My gal was t describe nesting habitat preferences f a diverse cmmunity f grund-nesting birds in grassland, using bjective measures f vegetatin characteristics (structure, structural hetergeneity, and flristics). Specifically, I asked; d species select vegetatin characteristics at nest sites and nest patches nnrandmly relative t availability?; d preferences fr vegetatin characteristics at nest sites and nest patches differ amng species?; and des directinal rientatin f vegetatin vlume immediately surrunding nest sites differ frm randm? STUDY AREA AND METHODS Stucfy Area and Species I cnducted research during the 1996 and 1997 breeding seasns n 227 ha f ungrazed grassland habitat n the Flathead Indian Reservatin in the lwer Missin Valley f westcentral Mntana, 80 km nrth f Missula Glacial tpgraphy characterizes the area, which exhibits lw relief and high densities f wetlands (Lkemen 1962). Management f the study area and ther lcal state and federal land fcused primarily n prviding breeding habitat fr upland nesting ducks and Ring-necked pheasant (Phasianus clchicus). I searched fr nests and sampled vegetatin in seven cntiguus fields supprting varying cmbinatins f tame, cl seasn grasses, primarily Intermediate Wheatgrass (Agrpyrn intermedium). Smth Brme (Brmus inermis), Kentucky Bluegrass (Pa pratensis), Quackgrass {Agrpyrn repens), and Orchard Grass {Dactylis glmerata)', differing cmbinatins f plant species created a wide diversity f vegetatin structure within and amng fields. N shrubs r trees existed n the study area, but

14 lcally cmmn extic (rbs ccurred at lw densities in all fields. The breeding bird cmmunity included bth grassland species and wetland species. I chse study species that were cmmn n the study area, but als attempted t maximize diversity f taxa and life histry traits. I sampled nests f Mallard, Gadwall, Nrthern Shveler (hereafter Shveler ), Cinnamn Teal (hereafter Teal ), Shrt-eared Owl (hereafter Owl ), Savannah Sparrw (hereafter Sparrw ) and Western Meadwlark (hereafter Meadwlark ). Observers ften culd nt distinguish between female Cinnamn and Blue-winged Teal (Anas discrs), which are clsely related and behavirally similar (Cnnelly and Ball 1984). I assumed that ur sample f teal nests reflected the cnsistent 7:1 rati f Cinnamn t Blue-winged Teal bserved in the area (Frman 199; U. S. Fish and Wildlife Service, unpublished data) In 1996,1 searched 5 fields (162 ha) and sampled nests f ducks and Owls Tw adjacent fields were acquired fr cnservatin purpses in 1997, allwing us t enlarge the search area t 227 ha. Cnsequently, I was able t btain sufficient samples f ducks. Owls, Sparrws, and Meadwlarks in Methds I cnducted nest searches n the study area three times each breeding seasn (at day intervals) using a cable-chain device (Higgins et al. 1969). This technique was effective fr finding duck and Owl nests but was less efficient fr finding passerine nests. Cnsequently, I bserved parental behavir (Martin and Geupel 199) frm twer blinds (Fndell et al. In prep) t find mst Sparrw and sme Meadwlark nests. I mapped all Sparrw territries early in the breeding seasn, and searched each territry fr nests thrughut the breeding seasn.

15 Vegetatin sampling ccurred when nests were n lnger active because I suspected that intensive sampling prvided visual and lfactry cues that culd increase nest predatin. Vegetatin sampling plts were centered n nest sites and n available sites randmly lcated within the study area. T lcate available sites, I generated randm crdinates fr a grid superimpsed n an aerial pht f the study area. I used tpgraphical features t get t the indicated lcatins. Then, the plt center was lcated by tssing a stick behind me in a randmly selected directin. Nest and available plts were sampled within days t accunt fr tempral change in vegetatin, which can be rapid in grassland. 1 assumed that rates f change in vegetatin characteristics were similar at nest and available sites. Furthermre, I sampled nly under suitable cnditins relative t wind (<25 km/h), light (>I h frm sunrise r sunset), and misture (i.e. I avided perids when misture caused vegetatin t drp). Transects extended frm each plt center at intercardinal directins with sampling pints at 0 and 10 m. Sampling pints at 0 m characterized the nest site r available site, and sampling pints at 10 m characterized the nest patch r available patch. I estimated vegetatin characteristics at each sampling pint. A ple marked in 2 cm intervals (mdified frm Rbel et al. 1970) was used t estimate vegetatin vlume Frm a height f 1 m and a distance f 2 m, I recrded the lwest interval nt cmpletely bscured by vegetatin. The ple was placed at each sampling pint, and readings were taken facing the plt center A rd marked in 10 cm intervals (Wiens 1969) was used t estimate vertical structure, a measure f vegetatin density, in At each sampling pint, I lwered the rd vertically t the grund and recrded the number f vegetatin

16 hits n the ple in each 10 cm interval between 0-50 cm, hits abve 50 cm, and the highest vegetatin hit. At plt center, estimates f vertical structure were taken 10 cm frm plt center t measure structure immediately arund the nest bwl Canpy cver at the nest was estimated at the time f nest lcatin as the percentage f the nest bwl bscured by vegetatin frm 1 m abve the nest. I defined structural hetergeneity as variatin in vlume and in vertical structure amng sampling pints: hetergeneity variables were calculated as the standard deviatin f vlume and f vertical structure variables after transfrmatin t remve psitive crrelatins between means and variances at randm plts. Flristics were recrded nly n a single transect. Frm 1 m abve each sampling pint, the bserver estimated percent cver f the grund in a 0.5 m diameter circle by each plant species independent f thers. Cver was estimated by categry (0-5%, 5-25%, 25-50%, 50-75%, 75-95%, and %; Mueller-Dumbis and Ellenburg 1974), then transfrmed int mid-pint percentages (Huberty 1994). Litter cver and litter depth were nt measured because Fndell (1997) had demnstrated that litter was abundant in all f the fields and that litter characteristics f nest sites did nt differ frm available sites fr mst species. I used lgical and statistical screening prcedures fr variable selectin and reductin (Huberty 1994). I drpped frm cnsideratin plant species with <5% mean cver and ccurring at <5% f nest plts because these species prbably cntributed little t analysis f nest site preferences. I als fund that subjectivity in estimating vertical structure increased with height (due mainly mvement f vegetatin caused by wind). Therefre, I drpped estimates f highest vegetatin hit and restricted analysis f

17 structurai hetergeneity t hits belw 20 cm. I used crrelatin analysis t indicate variables that measured similar characteristics and therefre culd be cmbined. Adjacent intervals f vertical structure that were highly crrelated were lumped. Grass species with similar grwth frms shwed similar patterns f preference amng bird species, s I lumped cver estimates frm thse species in statistical analysis (Table 1). Statistical Analysis I tested fr differences between nest and available habitat fr each species and amng nest habitat preferences f all species using univariate Analysis f Variance and descriptive Discriminant Functin Analysis (DFA; Huberty 1994, Nrusis/SPSS 1997). Cmparisn f habitat preferences amng species was based n difference variables created by subtracting available habitat variables frm nest habitat variables. Univariate analysis f variance shwed which variables were significantly different amng grups. I presented univariate results and descriptive statistics fr cmparisns amng habitat preferences f species t shw effect sizes f individual variables fr each species. I did nt present cmparisns f nest and available habitat fr each species because qualitatively similar interpretatins can be reached frm cmparisns f preferences amng species. I used DFA t derive the linear cmbinatin f variables that best separated grups. Althugh I used a stepwise prcedure that maximized Mahalnbis distance between grup centrids as the basis f variable selectin, I als utilized best subsets analysis and judgement t select parsimnius and bilgically interpretable final mdels (Huberty 1994, Nrusis/SPSS 1997). F-statistics were used t test if Mahalnbis lcatins in discriminant space differed between each pair f species. Experimentwise Type I errr

18 D OQ. C gq. C/) ' O 8 ( O ' " OQ. C a O Q. Table 1. Variables used in statistical analysis f nest Measurement Type Variable Descriptin Structure Vlume Mean vegetatin vlume Surface Structure Lw Structure Intermediate Structure High Structure Canpy* Mean Wiens hits frm 0-10 cm Mean Wiens hits frm cm Mean Wiens hits frm cm Mean Wiens hits abve 40 cm Percent f nest bwl bscured frm abve Hetergeneity SD Vlume Standard deviatin f squarert {Vlume) SD Surface Structure SD Lw Structure Standard deviatin f lg {Surface Structure) Standard deviatin f lg {Lw Structure) Flristic Bunchgrass Cver * Percent cver by Intermediate Wheatgrass and Orchard Grass Bradleaf Grass Cver* Bluegrass Cver Estimated nly at nest sites. ^ Cespitse grasses with rbust, erect culms. Rhizmatus grasses with lng, brad leaves. Percent cver by Smth Brme and Quackgrass Percent cver by Kentucky Bluegrass C/) C/) 00

19 rates f 10% were maintained fr univariate and pairwise tests n a tablewide basis. Classificatin f grups was presented as an index t effect size (Huberty 1994). Cvariance matrices were tested fr hmgeneity using Bx s M criterin. Because sme matrices shwed significant heterscedasticity, separate grup cvariance matrices were emplyed fr classificatin. Althugh sme f the DFA analyses vilated assumptins f nrmality and hmscedasticity {P < 0.05), parallel analyses using Cannical Crrespndence Analysis, which assumes nly unimdal distributins (ter Braak and Verdnscht 1986, Palmer 199), yielded virtually identical results. Structure crrelatins shwed the mst imprtant cmpnent variables fr each discriminant functin. Analysis was separated by year because f differences between years in study area size, study species, and habitat measurements. Fr species sampled in bth years, I pltted 1997 samples n the 1996 discriminant functin t facilitate cmparisns between years. I tested if Vlume arund nest sites f ducks and Owls (fr which n >20) r randm sites varied by directin using ne-way Analysis f Variance (Nrusis/SPSS 1997). I cmbined years fr this analysis because patterns were similar between years. Where significant differences existed, I tested fr pairwise differences using Tukey s methd t maintain a Type I errr rate f 10% fr each species (Day and Quinn 1989). RESULTS Nest Vegetatin Preferences Differences betw^n nest site and available vegetatin fr each species (Table 2) were highly significant in 1996 {P ^ ) and 1997 (P ^ ). High mean rates fr crrect classificatin fi*m DFA fr all species in 1996 (86% relative t 50% expected by

20 10 Table 2. Summary results f Discriminant Functin Analysis in separating nest site (<50 cm arund nest) frm available vegetatin fr grassland birds in westcentral Mntana in 1996 and Species n Wilks Lambda P % crrect classificatin Cinnamn Teal Gadwall < Mallard < Nrthern Shveler < Shrt-eared Owl Overall * 1997 Cinnamn Teal 0 0. < Gadwall Mallard < Nrthern Shveler Shrt-eared Owl < Savannah Sparrw < Western Meadwlark Overall ' *72% reductin in errr relative t crrect classificatin expected by chance (50%). **78% reductin in errr relative t crrect classificatin expected by chance (50%). chance) and 1997 (89% relative t 50% expected by chance) demnstrated lw verlap between nest site and available vegetatin. Structure crrelatins shwed strength f assciatin f vegetatin variables with each discriminant frinctin in 1996 (Table ) and

21 (Table 4). Fr species sampled in bth years, patterns f selectin generally were bilgically similar between years. Differences between nest patch and available vegetatin fr each species in 1996 and 1997 were weaker in all cases (Table 5) than between nest site and available vegetatin, but nest patch preferences in 1996 (Table 6) and 1997 (Table 7) generally reflected nest site preferences. Lw mean rates fr crrect classificatin frm DFA fr all species in 1996 (66% relative t 50% expected by chance) and 1997 (67% relative t 50% expected by chance) indicated high verlap between nest patch and available vegetatin. Because Canpy was estimated nly at nests (1996, Table Al; 1997, Table A2), it was nt included in DFA. Fr species sampled in bth years. Canpy was similar between years. Averaging acrss years. Canpy was very high fr Sparrws (84%) and Meadwlarks (87%), high fr Teals (65%) and Shvelers (58%), mderate fr Gadwalls (5%) and MaHards (9%), and lw fr Owls (12%). Vlume at nest sites varied by directin fr Teals (F = 4.2; df =,122; P = 0.007), Gadwalls {F =.7; df =,15; P ), Shvelers (F= 2.9; df =,16; f = 0.04), and Owls (F= 1.5; df =,84; P < ), but nt fr Mallards (F= 1.4; df =,19; P = 0.24) r at randm pints {F = 2.0; df =,764; P = 0.11). Vlume was lwest t the sutheast and highest t the suthwest r nrthwest fr all species. Fr Teals, Vlume t the sutheast was.4 cm less than t the suthwest (P = 0.10) and 5.1 cm less than t the nrthwest {P = 0.004). Vlume t the sutheast f Gadwall sites was 5.8 cm less than t the nrthwest (P = 0.008) and 4.4 cm less than t the nrtheast (P = 0.075). Fr Shvelers, Vlume t the sutheast was.1 cm less than t the suthwest (P = 0.019).

22 CO II (/) ' 8 C5-5 Û. C g Table. Structure crrelatins frm Discriminant Functin Analysis separating nest sites (<50 cm arund nest) and available vegetatin fr grassland birds in westcentral Mntana in Structure crrelatins are presented fr vmiables included by stepwise prcedures (r with structure crrelatins >0.25). Variable Species Cinnamn Teal Gadwall Mallard Nrthern Shveler Shrt-eared Owl Vlume SD Vlume Bunchgrass Cver (-0.51) Bradleaf Grass Cver Bluegrass Cver T Û. OC "O w

23 D OÛ. C gq. 00W ' O 8 ci' O. " O Q. C a O "O O Û. "O C/) C/) Table 4. Structure crrelatins frm Discriminant Functin Analysis separating nest sites (<50 cm arund nest) and available vegetatin fr grassland birds in westcentr^ Mntana in Structure crrelatins are presented fr variables included by stepwise prcedures (r with structure crrelatins >0.25), Variable Cinnamn Teal Species Gadwall Mallard Nrthern Shveler Shrt-eared Owl SavannWi Sparrw Vlume (0.4) Western Meadwlark Surface Structure (0.28) Lw Structure (0.8) (0.50) (0.8) Intermediate Structure 0.59 (0.0) 0.44 (0.1) High Structure (0.9) 0.72 SD Vlume SD Surface Structure 0.6 SD Lw Structure 0.1 (0.4) Bunchgrass Cver (0.48) 0.68 (-0.25) Bradleaf Grass Cver Bluegrass Cver (0.52) u>

24 14 Table 5. Summary results f Discriminant Functin Analysis in separating nest patch (10 m frm nest) frm available vegetatin fr grassland birds in westcentral Mntana in 1996 and Species n Wilks Lambda P % crrect classificatin Cinnamn Teal Gadwall Mallard Nrthern Shveler Shrt-eared Owl Overall * 1997 Cinnamn Teal Gadwall Mallard Nrthern Shveler Shrt-eared Owl Savannah Sparrw Western Meadwlark Overall * *4% reductin in errr relative t crrect classificatin expected by chance (50%). Mallards shwed a pattern and effect size similar t ther ducks, but high variance resulted in lw pwer. Directinal rientatin f vegetatin was strngest fr Owls: Vlume t the sutheast was 12.4 cm less than t the suthwest {P < ) and 6.6 cm less than t the nrthwest {P ).

25 D OQ. C gq. C/) C/) 8 Table 6 Structure crrelatins frm Discriminant Functin Analysis separating nest patches (10 m frm nest) and av^afle vegetatin fr grassland birds in westcentral Mntana in Structure crrelatins are presented fr variables included by stepwise prcedures (r with structure crrelatins >0.25). Species Variable Cinnamn Teal Gadwall Mallard Nrthern Shrt-eared Owl Shveler ^ Vlume (0.45) 1.00 SD Vlume 0.. " O Q. C Q Bunchgrass Cver Bradleaf Gmss Cver (0.44) Bluegrass Cver O Q. "O C/)CO LA

26 D OO. C g Q. C/) C/) Table 7. Structure crrelatins frm Discriminant Functin Analysis separating nest patches (10 m frm nest) and available vegetatin fr grassland birds in westcentral Mntana in Structure crrelatins are presented fr vmiables included by stq?wise prcedures (r vdth structure crrelatins >0.25). Species (D Variable Cinnamn Gadwall Mallard Nrthern Shrt-eared Savannah Western I Teal Shveler Owl Sparrw Meadwlark Vlume 1.00 (0.26) i 5 Surface Structure. " O Q. C g O "O O Q. C/) " Lw Structure 0.77 Intermediate Structure (0.46) (0.56) High Structure SD Vlume SD Surface Structure SD Lw Structure Bunchgrass Cver 1.00 Bradleaf Grass Cver 1.00 Bluegrass Cver Os

27 17 Differences in Preferences Amng Species Species preferences at nest sites differed fr all variables cllected in 1996 (Table Al). DFA included all variables except Bradleaf Grass Cver (Figs. 1A and IB) and significantly discriminated amng species (A = 0.21, P < ). Three significant discriminant axes (Axis 1: 125.6, P < ; Axis 2: 29.1, P = 0.008; Axis ; 12., P = 0.056) discriminated between all pairs f species except Mallards and Gadwalls (Table 8). The 68% mean rate f crrect classificatin frm DFA (Table 9) fr all species, relative t 20% expected by chance, demnstrated lw verlap amng species. Mallards and Gadwalls had similar scres n all axes, but all ther species ccupied unique lcatins in discriminant space. Lw scres n axis 1 shwed a preference fr increased Canpy and decreased Vlume that separated Teals and Shvelers frm Gadwalls, Mallards, and Owls. Teals and Shvelers differed n axes 2 and, reflecting preference Table 8. Cmparisns f nest site (<50 cm arund nest) characteristics between pairwise cmbinatins f grassland birds in westcentral Mntana in F-statistics and P-values shw the significance f Mahalanbis distances between nest sites in discriminant space. Of 10 pairs, nly Mallard and Gadwall nest sites were nt significantly different at the a = 0.10 level. Fr all tests, df = 5, 78. Species Gadwall 12.8 < Species Cinnamn Teal Gadwall Mallard Nrthern Shveler F P F P F P F P Mallard 17.7 < Nrthern Shveler < < Shrt-eared Owl 17. < <0.0001

28 i i ilî" M A L I^^L [GADW Ts e w T * NSHO-*- A <N O) O P!î. # %g I -1 h- MALLi GADW SEOW CITE I f-h J- NSHO^ Canpy (0.68) > Canpy (0.68) - -Vlume (-0.2) Vlume (-0.2) - Discriminant Axis 1 Discriminant Axis UJ 5 S Species CM < g II S I ü -1 T s e w Tmall W E M ^ S A V ^ T ^ D Canpy (0.70)... Surface Structure (0.44) Discriminant Axis 1 î il ,5!.. c JZ -0.5 CITE. KSH -- WEME SAV 'S Lj i i i ) 1 1 (r-l MALL GAW SE:w E Int. (0.65) and High Structure (0.79), Bluegrass Cver (0.6) Discriminant Axis Figure t. Discriminant functin analysis f nest sites (< 50cm arund nests) f grassland birds in westcentral Mntana, 1996 (A and B) and 1997 (D and E). T facilitate cmparisns between species sampled in btti years, 1997 samples were pltted n the first tw discriminant functins frm 1996 (C). Variables mst strngly cn^elated with each axis are shwn with structure crrelatins. Vectrs emph asize the directin f crrelatin. Centrids shw the mean lcatin n each axis with standard enrs. Abbreviatins; CITE=Cinnamn Teal, GADW=Gadwall, MALL=Mallard, NSHO= Nrthern Shveler, SEOW=Shrt-eared Owl, SAVS=Savannah Sparrw, and WEME=Western Meadwlark.

29 XI O Q. C g Q. Table 9. Rate f crrect classificatin (%) f nest sites (<50 cm firm nest) frm Discriminant Functin Analysis f 7 grassland birds in 1996 and 1997 based n separate cvariance matrices. Classificatin rates indicate the amunt f verlap f each species with thers. Species n Year Mean Cinnamn Gadwall Mallard Nrthern Shrt-eared Savannah Western ; Teal Shveler DM Sparrw* Meadwlark* I " IC a O " * 'Sampled nly in "60% reductin in errr relative t crrect classificatin expected by chance (20%). *65% reductin in errr relative t crrect classificatin expected by chance (14%). Û. O c "O \

30 20 by Teals relative t Shvelers fr increased Bunchgrass Cver, SD Vlume, and Bluegrass Cver. Amng species with high scres n axis 1, Owls preferred lwer Canpy and higher Vlume relative t Gadwalls and Mallards. Owls were strngly separated frm Gadwalls and Mallards n axis 2 by preference fr Bradleaf Grass Cver rather than Bunchgrass Cver and n axis by preference fr highest SD Vlume. Species preferences at nest sites differed fr 10 f 12 variables measured in 1997 (Table A2). DFA included 6 variables and significantly discriminated amng species (A = 0.15, P < ). Fur significant axes (Figs. ID and IE; Axis 1: 178.6, P < ; Axis 2: 62.8, P < ; Axis : %^= 2.1, ; Axis 4; %^= 17.7, P 0.08) discriminated between 20 f 21 pairs f species (Table 10): nly Teals and Shvelers {P = 0.12) did nt differ at the a = 0.10 level. The 70% mean rate f crrect classificatin frm DFA fr all species, relative t 14% expected by chance, demnstrated lw verlap between species (Table 9). Increased Canpy and Surface Structure segregated Meadwlarks and Sparrws frm ther species n axis 1. Meadwlarks were separated frm Sparrws by preference fr increased Bunchgrass Cver and SD Lw Structure, shwn n axis 2, and increased structure abve 20 cm and Bluegrass Cver by Sparrw, shwn n axis. Amng ducks and Owls, Teals and Shvelers had the highest preference fr Canpy and Surface Structure, Owls had the lwest, and Gadwalls and Mallards were intermediate. Differences amng these species n axis 2 reflected relative preferences fr bunchgrasses and the cnsequent increased SD Lw Structure. Mallards had higher scres n axis 2 relative t Shvelers and Owls, and Gadwalls and Teals were intermediate. Increased preference fr vertical structure abve 20 cm islated Mallards

31 ::d OQ. C 8Q. C/) ' O 8 5 ( O ' O. " "O OQ. O O Q. Table 10. Cmparisns f nest site (<50 cm arund nest) characteristics between pairwise cmbinatins f grassland birds in westcentral Mntana in F-statistics and F-values shw the significance f Mahalanbis distances between nest sites in Species Cinnamn Teal Gadwall Mallard Nrthern Shveler Shrt-eiu'ed Owl Savannah Sparrw Species F P F P F P F P F P F P Gadwall Mallard Nrthern Shveler Shrt-eared Owl Savannah Sparrw 9.6 < < < < < Western Meadwlark 6. < < < < < C/) C/)

32 22 frm ther ducks and Owls n axis. Teals were unique relative t ther ducks and Owls n axis 4 because f preference fr increased Intermediate Structure in cnjunctin with lw preference fr High Structure. Samples f ducks and Owls frm 1997 pltted n the 1996 discriminant functin (Fig. 1C) shwed that relative preferences remained similar between years. Nest patches f bird species differed significantly nly in Bunchgrass Cver in 1996 (Table A). DFA included Bunchgrass Cver and (Fig. 2a) discriminated amng species (A = 0.8, = 15.9, P = 0.001) and the single axis discriminated between 6 f 10 pairs f species (Table 11). Hwever, the 0% mean rate f crrect classificatin indicated almst cmplete verlap amng species (Table 12). Mallard preferred nest patches with increased Bunchgrass Cver and decreased Bradleaf Grass Cver relative t Teal, Shveler, and Owl; Gadwall preferred intermediate values. Species preferences at nest patches differed nly fr Bluegrass Cver in 1997 (Table A4). Hwever, DFA Table 11. Cmparisns f nest patch (10 m frm nest) characteristics between pairwise cmbinatins f grassland birds in westcentral Mntana in F-statistics and F-values shw the significance f Mahalanbis distances between nest patches in discriminant space. Six f 10 pairs were significantly different at the a = 0.10 level. Fr all tests, df= 1, 82. Cinnamn Teal Species Gadwall Mallard Nrthern Shveler Species F P F P F P F P Gadwall Mallard Nrthern Shveler Shrt-eared Owl

33 2 I I C i E 1 b CITE GADW MALL NSHO SEOW Species A I IQ CO CM 1 0 MALL NSHO WEM SAVS GADW CITE 1 SEOW Discriminant Axis 1 Lw Structure- Intermediate Structure- Figure 2. Discriminant functin analysis f nest patches (10 m frm nest) f grassland birds in westcentral Mntana, 1996 (A) and 1997 (B). Variables mst strngly crrelated with each axis are shwn with structure crrelatins. Vectrs emphasize the directin f crrelatin. Centrids shw the mean lcatin n each axis with standard errrs. Abbreviatins: CITE-Cinnamn Teal, GADW~6adwall, MALL=Mallard, NSHO=Nrthern Shveler, SEOW-Shrt-eared Owl, SAVS=Savannah Sparrw, and WEME=Western Meadwlartc.

34 D OQ. C gq. C/) (/i 8. " O Q. Table 12. Rate f crrect classificatin (%) f nest patches (10 m frm nest) frm Discriminant Functin Analysis f grassland birds in 1996 and 1997 based n separate cvariance matrices. Classificatin rates indicate the amunt f verlap f each species with thers. Species Year Mean Cinnamn Gadwall Mallard Nrthern Shrt-eared Savannah Western Teal Shveler OM Sparrw' Meadwlark* 1996 (f ' 'Sampled nly in **1% reductin in errr relative t crrect classificatin expected by chance (20%). '6% reductin in errr relative t crrect classificatin expected by chance (14%ÿ C a O "O O Q. C/) C/)

35 a> Q. c gq. C/)W " O Species I Cinnamn Teal Gadwall Mallard Nrthern Shrt-ewed Owl Savannah CÛ go. " O Q. C g O O Q. Table 1. Cmparisns f nest patch (10 m frm nest) characteristics between pairwise cmbinatins f grassland birds in westcentral Mntana in F-statistics and F-values shw the significance f Mahalanbis distances between nest patches in discriminant space. Eight f 21 pairs were significantly different at the a = 0.10 level; hwever, Savannah Sparrw nest patches accunted fr 6 f the 8 significant differences. Fr all tests, df=^, Shveler Sparrw Species F P F P F P F P F F F F Gadwall Mallard Nrthern Shveler Shrt-eared Owl O il Savannah Sparrw < Western Meadwlark C/) 00

36 26 included 4 variables (Fig. 2b) which significantly discriminated amng species (A = 0.5, P = ). The first discriminant axis was highly significant (%^= 58.9, P = ), the secnd marginally significant (%^= 25.2, P = 0.047). These axes significantly discriminated between 8 f 21 pairs f species (Table 1), but separatin f Sparrws frm ther species accunted fr 6 f these differences. The 45% mean rate f crrect classificatin indicated mderate verlap between species (Table 12). Lw scres n axis 1 separated Sparrw frm all ther species. Sparrw preferred increased Lw Structure but decreased Intermediate Structure, reflecting preference fr dense but nt tall patches. High scres n axis 2 indicated Mallard preferred patches with increased structural hetergeneity, reflected in increased Bunchgrass Cver and SD Lw Structure, relative t Owl. Scres fr ther species were intermediate. Discriminatin between species was much weaker at the scale f the nest patch relative t the scale f the nest site, and nest patch differences generally reflected nest site differences. DISCUSSION Nest Site Preferences Vegetatin characteristics at nest sites f all seven species shwed little verlap with vegetatin characteristics f available vegetatin. Vegetatin characteristics at nest sites shuld reflect preferences within available habitat, expressed primarily thrugh nest placement and secndarily thrugh manipulatin f vegetatin. Nnrandm selectin f nest sites based n numerus micrhabitat characteristics (e.g. vegetatin structure, hetergeneity, flristics, relief, rientatin f vegetatin) has been demnstrated acrss

37 27 many systems (Petersn and Best 1985, Rdrigues 1994, Badyaev 1995, Martin 1998). Hwever, studies f nesting grassland birds have generally fcused n preferences fr vegetatin vlume and brad flristic grups. Mallards and Gadwalls ften have been cnsidered t prefer nest sites with tall, dense vegetatin (Duebbert and Lkemen 1976, Crabtree et al. 1989, Kruse and Bwen 1996): in ur study, this pattern was reflected in preference fr increased Vlume at nest sites by Mallards and Gadwalls and preference fr increased vertical structure abve 20 cm by Mallards. Mallards and Gadwalls typically placed nests between clumps f bunchgrasses. This patchy grwth frm resulted in mderate Canpy and high SD Vlume fr Mallards and mderate Canpy and mderate SD Vlume fr Gadwalls. Mallards and Gadwalls have als shwn high use f shrubs fr nesting, suggesting that plants with clumped grwth frms are attractive t these species (Kruse and Bwen 1996). Teals and Shvelers have been reprted t prefer mderately tall grasses at nest sites (Livezey 1981, Kruse and Bwen 1996). Teals typically placed nests beneath verhanging grass and shwed preference fr high Canpy, increased SD Lw Structure, and increased vertical structure frm 0-40 cm, but decreased vertical structure abve 40 cm. The preference fr dense but nt tall vegetatin is cnsistent with studies shwing higher use f grazed and hayed areas relative t ther upland nesting ducks (Kirsch et al. 1978). Shvelers placed nests in Bradleaf grasses and manipulated leaves t frm lse dmes (Swls 1955), resulting in high Canpy. Shvelers als shwed preference fr increased vertical structure abve 40 cm and increasw SD Surface Structure, but decreased SD Vlume. Previus bservatins f Owl nests have suggested a preference

38 28 fr sites with tall grasses (Kantrud and Higgins 1992). In ur study. Owls typically selected nest sites with lw canpy cver adjacent t a large tuft f bradleaf grasses. Increased vertical structure was preferred at cm, but nest sites had relatively little structure near the grund apart fi m the tuft, resulting in increased SD Vlume. Sparrws ften place nests under grass tufts with a tunnel entrance, but nests in dense cver may be simple pen cups (Wheelwright and Rising 199). Sparrws typically placed pen cup nests in Kentucky Bluegrass with increased Vlume. The numerus basal leaves f Kentucky Bluegrass prvided extreme high Canpy and abundant, unifrm vertical structure belw 20 cm. Meadwlarks ften cnstruct dmed nests in grass f mderate vlume with lw variatin in vlume (Lanyn 1994, Granfrs et al. 1996). I fund that Meadwlarks preferred nest sites with increased vertical structure belw 20 cm, and nests typically adjined a small tuft f grass, which increased SD Lw Structure but nt SD Vlume Meadwlarks cnstructed nearly cmplete dmes ver nests, resulting in almst cmplete Canpy. Studies f bird habitat preferences have cmmnly related habitat use t vegetatin structure (Andersn and Shugart 1974, Whitmre 1975, Rtenberry and Wiens 1980). Hwever, because vegetatin structure is largely derived frm plant grwth frm, preferences fr flristics may be driving apparent preferences fr vegetatin structure (Wiens and Rtenberry 1981, Rice et al. 1984, Rtenberry 1985). I bserved that plant grwth frm did nt dictate vegetatin structure at nests. Fr example, bth Shvelers and Owls preferred Bradleaf grasses, but Owls placed nests next t tufts f residual grass, resulting in higher SD Vlume than at Shveler nests Shvelers actively

39 29 cnstructed dmes ver nests, resulted in much higher Canpy than at Owl nests. In additin t plant grwth frm, vegetatin structure at nest sites was influenced largely by placement f the nest in vegetatin and partly by manipulatin f vegetatin. Birds in many systems have demnstrated nnrandm selectin f nest patches (Martin and Rper 1988, Knpf and Sedgwick 1994, Badyaev 1995). Hwever, I fund little preference fr vegetatin characteristics at the scale f the nest patch, and I cnclude that the primary nest vegetatin preferences ccurring amng these 7 species in ungrazed, seeded grass cver ccurred at a scale <10 m arund the nest site. Because preferences at nest sites ften reflected placement relative t vegetatin immediately arund the nest r manipulatin f vegetatin ver the nest, selectin appeared t be mst strngly influenced by vegetatin in a diameter <50 cm arund the nest. Differences in Preferences Amng Species Species shwed strng segregatin f preferences fr vegetatin characteristics at nest sites. Species with similar preferences fr ne vegetatin characteristic usually differed in preferences fr thers. Because f differences in species and variables included, the DFA mdels I cnstructed in each year can nt be directly cmpared. Hwever, cmparisns f lcatins f 1996 and 1997 samples in 1996 discriminant space shw that althugh preferences changed smewhat between years, the relative preferences f ducks and Owls were cnsistent. In additin, the first tw discriminant axes derived frm the 1996 and 1997 data sets were bilgically similar. Canpy was strngly crrelated in bth years with the first discriminant axis, the mst imprtant in

40 0 discriminating amng species. In bth years. Canpy increased with decreasing bdy size, with the exceptin f the Owls. The secnd discriminant axis in each year was crrelated psitively with Bunchgrass Cver and described a gradient f increasing variatin in vegetatin structure. The negative crrelatin f Bradleaf Grass Cver in 1996 shwed decreasing preference fr plant species with mre unifrm structure, and the psitive crrelatin f SD Lw Structure in 1997 quantitatively reflected the increase in patchiness f structure due t the clumped grwth frm f bunchgrasses. Measures f vertical structure shwed these species differed in preferences fr vegetatin at different heights abve the grund. In particular, preference fr Surface Structure increased with decreasing bdy size. 1 fund that Vlume, which has been the primary vegetatin characteristic used t cmpare nest sites f grassland species (especially ducks) (Kantrud and Higgins 1992, Gilbert et al. 1995, Kruse and Bwen 1996) cntributed relatively little t segregating amng these species. Partitining f resurces has seemed t be largely absent in grassland systems. In frested systems, cmpetitin fr fd has traditinally been invked as the prcess driving partitining f vegetatin and, ultimately, cmmunity structure (MacArthur and MacArthur 1961, James 1971, Willsn 1974). In prairie ecsystems, cexisting species shw wide verlap in fd resurces and fraging habitat, fd resurces are nt limiting in mst years, and cmpetitin fr fd des nt appear t drive cmmunity structure (Wiens 197, Wiens 1977, Rtenberry and Wiens 1980, Wiens and Rtenberry 1980, Dubwy 1988). Hwever, nest sites are als a resurce that can influence habitat selectin and cmmunity structure. Partitining f habitat selected fr nesting based n

41 1 vegetatin characteristics has been demnstrated in frested systems (Martin 1998). In many systems the availability f r cmpetitin fr suitable nest sites has a strng rle in determining patterns f ccupancy and cmmunity structure (Parker 1985, Arnld and Higgins 1986, Martin 1988, Martin 199). Given the brad verlap in fraging preferences amng species (and because many waterfwl frage in wetlands rather than grasslands), I suggest that differences in nest site preferences may play an imprtant rle in determining habitat ccupancy and hence cmmunity structure in grasslands. Ptential Adaptive Significance f Preferences at Nest Sites Because reprductive success directly influences fitness, natural selectin shuld favr chices f nest sites that minimize reprductive failure (Martin 199). Extreme nest micrclimates can stress adults and yung, and birds appear t place nests in vegetatin t mderate nest micrclimate (Walsberg 1985, Glutney and Clark 1997). Orientatin f increased Vlume t the NW r SW relative t the SE by ducks and Owls may facilitate heat gain in the mrning but prvide shading against excessive inslatin in the afternn (Petersn and Best 1985, Facemire et al. 1990, With and Webb 199). Orientatin f Vlume als may minimize cnvective heat lss during cld perids by sheltering nests (Cannings and Threlfall 1981, Sakai and Nn 1991) frm prevailing westerly winds. Because verhead cver may prvide shading frm mid-day sun and prtectin radiative heat lss at night and because smaller birds are mre subject t thermal stress, the trend f increasing Canpy with decreasing bdy size suggests that Canpy may becme increasingly imprtant in thermal regulatin with decreasing bdy size.

42 2 Nest predatin is the primary cause f nest failure in grasslands (Ricklefs 1969) and hence shuld exert strng selective pressure n nest site preferences. If increases in cncealment f nests by vegetatin, vegetatin density, r vegetatin hetergeneity cause decreased nest predatin (Schrank 1972, Bwman and Harris 1980, Crabtree et al 1989, Clark and Nudds 1991), I wuld expect species with similar suites f nest predatrs t cnverge n similar nest sites. Hwever, nests sites f each species in this study were usually unique. Differences amng species culd be related t different micrclimate needs, fraging needs, r differences in predatin pressures; hwever, the clsely related dabbling ducks have similar life histries, nest micrclimates, and predatr cmmunities. Density dependent nest predatin has been shwn t drive partitining f nest site characteristics amng species in a frested system (Martin 1988, Martin 199). Fr nests that are clsely spaced r have similar characteristics, predatrs may cncentrate search effrts where nests have already been fund r develp search images fr cmmn nest types (Granssn et al. 1975, Dunn 1977, Knaptn 1979, Sugden and Beyersbergen 1986, Martin 199). Therefre, cexisting species wuld be expected t diverge in nest site preferences and nest initiatin dates. Mcamgement Implicatins Management f grasslands has fcused n prviding tall, high vlume, mntypic vegetatin fr upland nesting ducks (Higgins and Barker 1982), either by limiting disturbance n existing managed grasslands r thrugh restratin prgrams fr wildlife management (e.g. Waterfwl Prductin Areas) r fr limiting crp surpluses and sil