Passerine communities and bird-habitat relationships on prescribe-burned, mixed grass prairie in North Dakota by Elizabeth Marie Madden A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Biological Sciences Montana State University © Copyright by Elizabeth Marie Madden (1996) Abstract: To more effectively manage remaining native grasslands and declining populations of prairie passerine birds, linkages between disturbance regimes, vegetation, and bird abundance need to be more fully understood. Therefore, I examined bird-habitat relationships on northern mixed-grass prairie at Lostwood National Wildlife Refuge in northwestern North Dakota, where prescribed fire has been used as a habitat management tool since the 1970’s. I sampled bird abundance on upland prairie at 310 point count locations during 1993 and 1994 breeding seasons. I then measured vegetation structure and composition at each location. Complete fire histories were available for each point, with over 80% being burned 1 to 4 times in the last 15 years. Striking differences in bird species abundance were apparent among areas with different fire histories. Baird's, grasshopper, and Le Conte's sparrows, Sprague's pipits, bobolinks, and western meadowlarks were absent from unburned prairie, but were among the most common birds seen overall. In contrast, common yelIowthroats and clay-colored sparrows both reached highest abundance on unburned prairie. These data emphasize the importance of disturbance in maintaining grassland communities, and indicate that periodic defoliations by disturbances such as fire are crucial to the conservation of endemic grassland bird populations. Bird species examined were well-distributed over gradients of vegetation structure and composition. Sprague's pipits used the shortest and sparsest cover available. Baird's sparrows, grasshopper sparrows, and western meadowlarks used more moderate amounts of vegetation cover. Bobolinks and Le Conte's sparrows preferred taller and denser cover, especially of exotic grasses. Savannah sparrows were ubiquitous, clay-colored sparrows and common yellowthroats were distinctly shrub-associated, and brown-headed cowbirds were habitat generalists. Sprague's pipits and Baird's sparrows showed preferences for native grasses. The results indicate that a mosaic of vegetation successional stages maximizes avian biodiversity.  PASSERINE COMMUNITIES AND, BIRD-HABITAT RELATIONSHIPS ON PRESCRIBE-BURNEb1 MIXED-GRASS PRAIRIE IN NORTH DAKOTA by Elizabeth Marie Madden A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Biological Sciences MONTANA STATE UNIVERSITY-BOZEMAN: Bozeman, Montana May 199,6 ^31? ii APPROVAL of a thesis submitted by Elizabeth Marie Madden This thesis has been read by each member of the thesis committee and has been found to be satisfactory regarding content, English usage, format, citations, bibliographic style, and consistency, and is ready for submission to the College of Graduate Studies. Andrew J. Hansen . LLi (Signature) Date Approved for the Department of Biology Ernest Vyse ^ (Signature) / Date Approved for the College of Graduate Studies Robert L. Brown (Signature) Date iii STATEMENT OF PERMISSION TO USE In presenting this thesis in partial fulfillment of the requirements for a Master’s degree at Montana State University-Bozeman, I agree that the Library shall make it available to borrowers under the rules of the Library. If I have indicated my intention to copyright this thesis by including a copyright notice page, copying is allowable only for scholarly purposes, consistent with “fair use” as prescribed in the U S. Copyright Law. Requests for permission for extended quotation from or reproduction of this thesis in whole or in parts may be granted only by the copyright holder. Signature Date iv ACKNOWLEDGMENTS , As major advisor, Andy Hansen shared ideas and provided guidance in many aspects of this research. Jay Rotella improved my work with countless suggestions and advice, and Pat Munholland educated me on the methods of proper sampling. Bob Murphy was instrumental in all stages of this work and provided patient counsel and friendship throughout. I thank Karen Smith, Lostwood Refuge manager, for making this project a reality; for a wealth of technical, biological, and moral support; and for sharing how she “does it.” Her courage and pioneering spirit continue to inspire me. Karen and Bob generously shared their home and many meals with, me. I am grateful to my outstanding assistants, Brian Johnson and Lisa DeMoss, for large sample sizes and unwavering dedication in the field. Maiken W inter was a cherished field companion in 1993, and Marriah Sondreal and Natalie Fahler assisted with vegetation sampling in 1994. U S. Fish and Wildlife Service-Region 6 provided financial support for this project, and I thank John Comely, Wayne King, and Stephanie Jones for strong support beyond dollars. The staffs of Des Lacs Refuge Complex and MSU Biology Department provided much-appreciated clerical, technical, and maintenance expertise. Leigh Murray, New Mexico State University, conducted logistic regression modeling. For much love and moral support I am indebted to my friends, fellow graduate students, and family. Thank you all. TABLE OF CONTENTS Page LIST OF TABLES............................................................................................. viii LIST OF FIGURES........................................................................................... xi ABSTRACT....................... xii CHAPTER 1: THESIS INTRODUCTION....................................................... 1 CHAPTER 2: EFFECTS OF PRESCRIBED FIRE ON MIXED-GRASS PRAIRIE PASSERINE COMMUNITIES IN NORTHWESTERN NORTH Dakota...........:.......... :.......................... .............................. 5 INTRODUCTION...................................;........:.................... ................ 6 STUDY AREA..................................;.............. i ..................................... 8, METHODS............................................................................................ 15 Study Design................................................ .15 Plot Selection............................................................................. 15 1993 Field Season........................................................ 15 1994 Field Season........................................................ 17 Fire History................................... 18 Bird Abundance Sampling....................................................... 19 Vegetation. Sampling............................................ 22 Data Analyses........................................................................... 24 Vegetation Data Reduction.......................................... 24 Bird Data Summarization............................................. 27 Statistical Tests.............................................................. 29 Principle Components Analysis.................................. 31 RESULTS................................................... :......................................... 31 Vegetation and Fire.................................................................. 31 Vegetation Along the Fire Index Gradient................... 32 Vegetation Variables Summarized - PCA................... 33 V TABLE OF CONTENTS--Continued. Page Bird Sampling........................................................................... 38 Birds and Fire................................................................ 39 Species Abundance in Relation to Years Since Last Fire............................ 39 Species Abundance in Relation to Number of Burns.............................................................................. 40 Species Abundance Along Fire Index Gradient,....... 45 DISCUSSION........................................................................................ 46 Vegetation and Fire.................................................................. 46 Birds and Fire............................................................................ 47 Pre- and Post-Settlement Endemic Bird Populations.......... 53 Mechanisms Underlying Bird Response to Fire................. 54 Scope and Limitations.............................................................. 56 CONCLUSIONS AND MANAGEMENT IMPLICATIONS................. 58 LITERATURE CITED......................... 60 CHAPTER 3: BIRD-HABITAT RELATIONSHIPS ON MIXED GRASS PRAIRIE IN NORTHWESTERN NORTH DAKOTA.................................... 66 INTRODUCTION....................................................................... 67 STUDY AREA......................... 69 METHODS............................................................................................. 70 Data Analyses............................................................................ 71 Data Summarization...................................................... 71 Principle Components Analysis.................................. 72 Bird-Vegetation Associations...................................... 72 Comparisons of Occupied/Unoccupied Habitats....... 73 Predictive Habitat Functions........................................ 74 RESULTS.............................................................................................. 75 Bird Species Occurrence in PC Space.................................. 75 Bird-Vegetation Associations................................................... 77 vi TABLE OF CONTENTS-Continued. Page Comparisons of Occupied/Unoccupied Habitats.................. 83 Logistic Regression Models..................................................... 85 Baird’s sparrow.................................................... 86 Predictive Models for Other Bird Species.................. 90 DISCUSSION....................................................................... !............... 95 Bird-Habitat Relationships..................................................................95 Baird’s sparrow....................................................... 95 Brown-headed cowbird................................................. 97 Bobolink.......................................................................... 98 Clay-colored sparrow.................................................... 99 Commonyellowthroat........................................................ 100 , Grasshopper sparrow....................................................... 100 Le Conte's sparrow............................................................ 101 Savannah sparrow............................................................. 102 Sprague's pipit................................................................... 103 Western meadowlark............................................ 104 General Bird Community Associations.................................. 105 Native and Exotic Grass Habitats............................................... 106 Scope and Limitations.................................................................. 109 CONCLUSIONS AND MANAGEMENT IMPLICATIONS................. 111 LITERATURE CITED................................................................................ 113 CHAPTER 4: THESIS SUMMARIZING CONCLUSION.............................. 119 APPENDICES....................................................................................................... 120 A ................................................................................................................... 121 B................................................................................................................... 124 C............................................................................-.................................. 127 D............ :...................................................................................:............ 138 E................................................................................................................... 143 vii LIST OF TABLES 1 Prescribed burn units sampled, LNWR1 1993 and 1994................... 19 2 Description of vegetation variables measured at point count locations................................................................................................... 25 3 Relationships between individual vegetation variables and:amount of fire. Simple linear regressions were performed on individual vegetation variables for each of 8 different burn treatments rated by fire index.............................................................................................. 33 4 Eigenvector loadings (based on correlation matrix) of principle components (PCI and 2) of vegetation structure variables in 1993 and 1994................................................................................................... 34 5 Bird species abundance and standard error by category of years since last fire, LNWR, 1994................................................................... 41 6 Bird species abundance and standard error by category of years since last fire, LNWR, 1993................................................................... 42 7 Bird species abundance and standard errors by category of number of burns in previous 15 years, 1994..................................... 43 8 Bird species abundance and standard errors by category of number of burns in previous 15 years, 1993....................................... 44 9 Relationships between bird abundance and amount of fire based on linear regressions of abundance on the fire index.......... .............. 45 10 Spearman rank correlations between bird abundances and vegetation structure variables, LNWR, 1993..................................... 78 11 Spearman rank correlations between bird abundances and plant species associations, LNWR, 1993...................................................... 79 12 Spearman rank correlations between bird abundances and vegetation structure variables, LNWR, 1994..................................... 80 viii Table Page A LIST OF TABLES—Continued. Table Page 13 Spearman rank correlations between bird abundances and plant species associations, LNWR1 1994.......................................................... 81 14 Logistic regression models that best predict occurrence of Baird’s sparrow..................................................................................................... 87 15 Logistic regression models that best predict occurrence of grassland birds. Variables were selected using a backward- elimination routine.................................................................................. 91 16 Common and scientific names of passerine and upland shorebird species observed during point counts at LNWR1 1993 and 1994.... 122 17 Range of values for vegetation variables, LNWR1 1993 and 1994... 128 18 Vegetation values and standard errors among categories of years i since last fire, LNWR1 1993...................................................................... 130 19 Vegetation values and standard errors among categories of years since last fire, LNWR1 1994...................................................................... 132 20 Vegetation values and standard errors among categories of number of burns in last 15 years, LNWR1 1993.................................. 134 21 Vegetation values and standard errors among categories of number of burns in last 15 years, LNWR1 1994.................................. 136 22 Relative abundance indexes and standard errors of singing male passerines from point counts, 1993 and 1994, LNWR...................... 139 r ■" 23 Bird species abundance and standard error (SE) over years since last fire, LNWR1 1993 trends........................................... 141 24 Bird species abundance and standard error (SE) over years since last fire, LNWR11994 trends.........................................................................142 25 Comparison of vegetation values for point count locations occupied and unoccupied by Baird’s sparrow in 1994, LNWR......... 144 ix X LIST OF TABLES—Continued. 26 Comparison of vegetation values for point count locations occupied and unoccupied by brown-headed cowbird in 1994, LNWR.......:.........................f..................................................................... 145 27 Comparison of vegetation values for point count locations occupied and unoccupied by bobolink in 1994, LNWR..................... 146 28 Comparison of vegetation values for point count locations , occupied and unoccupied by clay-colored sparrow in 1994, LNWR.............. 147 29 Comparison of vegetation values for point count locations occupied and unoccupied by common yellowthroat in 1994, LNWR........................................................................................................ 148 30 Comparison of vegetation values for point count locations occupied and unoccupied by grasshopper sparrow in 1994, LNWR.................................................................. 149 31 Comparison of vegetation values for point count locations occupied and unoccupied by Le Conte’s sparrow in 1994, LNWR... 150 32 Comparison of vegetation values for point count locations occupied and unoccupied by savannah sparrow in 1994, LNWR.... 151 33 Comparison of vegetation values for point count locations occupied and unoccupied by Sprague’s pipit in 1994, LNWR.......... 152 Table Page 34 Comparison of vegetation values for point count locations occupied and unoccupied by western meadowlark in 1994, LNWR. i 53 LISTOF FIGURES Figure Page 1 Factors driving grassland bird populations on the Great Plains.... 2 2 Lostwood National Wildlife Refuge, North Dakota, with area of study and burn units delimited............................................................ 9 3 Description of first 2 Principle Components based on vegetation structure................................................................................................. 35 4 Composite confidence ellipses for areas with different burn histories...................................... 37 5 Composite 95% confidence ellipses based on the mean for plots at which individual bird species occurred in 1994, with reference to Principle Component 1 and 2 ......................................................... 76 6 Incidence of Baird’s sparrow as predicted by logistic regression equations for 4 vegetation attributes.................................................. 88 7 Incidence of 3 grassland birds as predicted by logistic regression, equations for Robel visual obstruction measures of vegetation.... 89 8 Incidence of grasshopper sparrow and bobolink as predicted by logistic regression equations for frequency of native and exotic grasses.................................................. 93 9 Incidence of clay-colored sparrow as predicted by logistic regression equations for shrub cover..................................... 94 10 Incidence of western meadowlark as predicted by logistic regression equations for forb cover.....................'................ .......... 94 xi xii ABSTRACT To more effectively manage remaining native grasslands and declining populations of prairie passerine birds, linkages between disturbance regimes, vegetation, and bird abundance need to be more fully understood. Therefore, I examined bird-habitat relationships on northern mixed-grass prairie at Lostwood National W ildlife Refuge in northwestern North Dakota, where prescribed fire has been used as a habitat management tool since the 1970’s. I sampled bird abundance on upland prairie at 310 point count locations during 1993 and 1994 breeding seasons. I then measured vegetation structure and composition at each location. Complete fire histories were available for each point, with over 80% being burned 1 to 4 times in the last 15 years. Striking differences in bird species abundance were apparent among areas with different fire histories. Baird's, grasshopper, and Le Conte's sparrows, Sprague's pipits, bobolinks, and western meadowlarks were absent from unburned prairie, but were among the most common birds seen overall. In contrast, common yelIowthroats and clay- colored sparrows both reached highest abundance on unburned prairie. These data emphasize the importance of disturbance in maintaining grassland communities, and indicate that periodic defoliations by disturbances such as fire are crucial to the conservation of endemic grassland bird populations. Bird species examined were well-distributed over gradients of vegetation structure and composition. Sprague's pipits used the shortest and sparsest cover available. Baird's sparrows, grasshopper sparrows, and western meadowlarks used more moderate amounts of vegetation cover. Bobolinks and Le Conte's sparrows preferred taller and denser cover, especially of exotic grasses. Savannah sparrows were ubiquitous, clay-colored sparrows and common yellowthroats were distinctly shrub-associated, and brown-headed cowbirds were habitat generalists. Sprague's pipits and Baird's sparrows showed preferences for native grasses. The results indicate that a mosaic of vegetation successional stages maximizes avian biodiversity. 1 CHAPTER 1: THESIS INTRODUCTION Grasslands of the North American Great Plains historically were maintained as treeless, herbaceous communities by climate and periodic disturbances, especially fire and grazing (Sauer 1950, Wells 1970, Kucera . 1981). European settlement of the Great Plains disrupted these natural processes. Large tracts of prairie were lost completely upon conversion to cropland. Fire suppression and replacement of native herbivores (e.g., bison [Bison bison! and prairie dogs [Cynomys IudovicianusD with domestic livestock altered natural disturbance patterns on remaining prairie. In response to the dynamic environment of the Great Plains, grasslands birds show considerable fluctuations in seasonal ranges and densities, and bird diversity and density are low in grasslands compared with most other habitats (Cody 1985). Climate, especially climatic instability, is believed to be the main driver of these bird populations (Wiens 1974, Zimmerman 1992). It affects birds both directly (i.e., effects of drought/flooding), and indirectly, through its influence on vegetation. The role of vegetation structure and composition in determining grassland bird habitat use has been well-documented (Cody 1968, Wiens 1969, Rotenberry and Wiens 1980). Grassland vegetation is patterned largely by climate, and locally by disturbance and land use practices (e.g., fire, herbivory, agriculture), and area geomorphology. These processes, in turn, all drive grassland bird abundance and distribution (Figure 1). 2 insect foods nesting cover DISTURBANCE fire grazing CLIMATE rainfall evapotranspiration GEOMORPHOLOGY soil type topography BIRD COMMUNITY abundance distribution diversity VEGETATION PATTERN structure composition patchiness FUNCTION productivity succession Figure 1. Factors driving grassland bird populations on the Great Plains. This thesis explores the role of one type of grassland disturbance, fire, in defining vegetation patterns and bird communities in mixed-grass prairie. Fire is integral to grassland ecology; without it, most grasslands would ultimately succeed to forest or shrubland (Sauer 1950, Wells 1970). Fire reduces woody vegetation, increases productivity by removing dead plant material and releasing soil nutrients bound up in litter, and increases native vegetation growth and reproduction (Kucera 1981, Wright and Bailey 1982). Although fire has always played a large role in maintaining Great Plains grasslands, it is difficult to estimate historic grassland fire patterns without trees to carry records of burning in fire scars. Extrapolating from fire histories in grasslands under pine forests, Wright and Bailey (1982) estimated a 5-10 year 3 fire frequency for grasslands. Assessing rates of fuel accumulation and woody plant invasion has produced estimates of 3-4 year fire return intervals (FRI) in tallgrass prairie, 4 years in sandhill prairie, 6 years in northern mixed prairie (with up to 25 years in the dry, western mixed prairie), and 5-10 years in short- grass prairie (Bragg 1995). Grasslands receiving more moisture have higher productivity and thus are more fire-dependent (Kucera 1981). The distinction between drier and mesic mixed prairies is important, as fire effects appear to differ greatly with varying moisture conditions between these areas (Higgins et at. 1989, W right and Bailey 1982). Resource managers on the Great Plains often incorporate prescribed fire into habitat management, but information on fire effects is sparse, and managers often must rely on personal experience as a gauge for management decisions. Objectives of this thesis were to define relationships among bird abundance, habitat features, and fire history on northern mixed-grass prairie. The ultimate goal was to provide land managers with predictive models relating vegetation characteristics and bird occurrence. My general approach was to sample vegetation and birds over the stages of fire succession and explore relationships among them. Chapter 2 examines the effects of prescribed fire on prairie bird communities, and secondarily on prairie vegetation. Chapter 3 defines grassland birci-habitat associations. A final summary links conclusions and implications of both parts. 4 Bragg, T. B. 1995. Climate, soils and fire: The physical environment of North American grasslands, in The Changing Prairie, K. Keeler and A. Joerris, editors, in Press Oxford University Press, New York. Cody, M. L. 1968. On the methods of resource division in grassland bird communities. American Naturalist 102:107-147. Cody, M. L. 1985. Habitat selection in grassland and open-country birds. Pp. 191- 226, in M. L. Cody, ed. Habitat selection in birds. Academic Press, Orlando, Florida. <' Higgins, K. F., A. D. Kruse, and J. L. Piehl. 1989. Effects of fire in the northern Great Plains. South Dakota State University Extension Circular 760. 47 pp. Kucera, C. L. 1981. Grasslands and fire. Pp. 90-111 in Fire regimes and ecosystem properties: proceedings of the conference. USDA Forest Service General Technical Report WO-26, , Washington, D C. 594 pp. Rotenberry, J. T., and J. A. Wiens. 1980. Habitat structure, patchiness, and avian communities in North American steppe vegetation: a multivariate analysis. Ecology 61:1228-1250. Sauer, C. O. 1950. Grassland climax, fire, and man. Journal of Range Management 3:16-21. Wells, P. V. 1970. Postglacial vegetational history of the Great Plains. Science 167:1574-1582. Wiens, J. A. 1969. An approach to the study of ecological relationships among grassland birds. Ornithological Monographs 8:1-93. Wiens, J. A. 1974. Climatic instability and the “ecological saturation” of bird communities in North American grasslands. Condor 76:385-400. Wright, H. A., and A. W. Bailey. 1982. Fire Ecology. John W iley and Sons, NY. 501 pp. Zimmerman, J. A. 1992. Density-dependent factors affecting the avian diversity of the tallgrass prairie community. Wilson Bulletin 104:85-94. LITERATURE CITED I 5 CHAPTER 2 EFFECTS OF PRESCRIBED FIRE ON MIXED-GRASS PRAIRIE PASSERINE COMMUNITES IN NORTHWESTERN NORTH DAKOTA 6 INTRODUCTION In the past 30 years, grassland birds have exhibited the most significant, widespread declines of any group of North American birds, including neotropical migrants (Knopf 1994). Declines in nonmigratory grassland birds (Finch 1,991, Dobkin 1992) indicate that declines are at least partly due to problems on the breeding grounds. Grasslands represent the largest vegetative province in North America (Knopf 1994). Cumulative effects of overgrazing, suppression of fire, and conversion of prairie to cropland have severely reduced and fragmented grassland habitats throughout the Great Plains. Remaining tracts of native prairie thus become increasingly valuable, and their proper management critical. Individual bird species do not respond alike to common grassland management practices, with various species preferring specific habitat conditions along a continuum of prairie succession (reviews in Kirsch et al. 1978, Ryan 1990, Dobkin 1992). Species that prefer vegetation communities , promoted by human activities such as mowing, season-long grazing, and fire suppression generally have benefited, while those requiring more natural disturbance regimes (i.e., periodic fire and grazing) have suffered serious declines in recent years (Dobkin 1992). To improve the effectiveness of managing both grassland birds and remaining native prairies, linkages between disturbance regimes, vegetation, 7 and bird communities need to be more clearly understood. Given that avian demographic data are expensive and difficult to collect, habitat studies based on bird abundance provide a practical alternative for studying multi-species communities (Hansen et al. 1993). Although such studies cannot determine causal mechanisms for the observed patterns or be assumed to measure actual habitat quality or selection (Van Horne 1983), they are useful in determining and predicting bird occupancy patterns for management purposes. Few studies have addressed effects of fire on prairie passerines. These typically include only a single burn and.control (Huber and Steuter 1984, Pylypec 1991) and most follow responses for only 1 -3 years postfire (Forde et al. 1984, Herkert 1994). My goal was to document longer-term effects of fire by examining multiple, independent burns done over an extended period (15 years) in mixed-grass prairie. The objective of this study was to quantitatively define relationships between fire history and bird abundance in northern mixed-grass prairie, and secondarily, to assess associated vegetation. Key questions were: 1) How do songbird abundance and species richness vary over stages of fire succession in northern mixed-grass prairie? 2 2) What are the characteristics of vegetation associated with these successional stages? 8 STUDY AREA Lostwood National Wildlife Refuge (LNWR) covers 109 km2 of rolling to hilly, mixed-grass prairie in Mountrail and Burke counties, northwestern North Dakota (48°37’N; 102°27’W) (Figure 2). Its large tracts of grassland are interspersed with more than 4100 wetland basins and many clumps of quaking aspen (Populus tremuloides). Habitat composition is 55% native prairie, 21% previously-cropped prairie (revegetated with tame and native prairie plants), 20% wetland, 2% trees, and 2% tall shrubs (Murphy 1993). Major vegetation is a needlegrass-wheatgrass CStipa spp,- Agropyron spp.) association (Coupland 1950), with diverse forbs and scattered shrubs. When delimiting the study area, I excluded the extreme southern and western portions of the refuge because they differed significantly from the rest of the refuge in topography, soils, and vegetation. Several other areas, including 6 western-most sections of the Lostwood Wilderness Area (Figure 2), were excluded because many parts had been tilled and cropped into the 1950s and were dominated by tame grasses. LNWR lies within the Great Plains physiographic region known as the Missouri Coteau (Bluemle 1980), This 20-30 km wide strip of dead-ice moraine deposited by the Wisconsin glacier is characterized by knob-and-kettle 9 Lostwood National Wildlife Refuge Refuge Boundary Burn Boundaries j j Study Area location in North Dakota Figure 2. Lostwood National Wildlife Refuge, North Dakota, with area of study and burn units delimited. 10 topography (685 - 747m) with non-integrated drainage (i.e., precipitation collects in wetland basins via surface runoff and seepage) and prairie pothole wetlands of varied sizes and hydrologic types (see Stewart and Kantrud 1971). The climate in this region is semi-arid, with individual years ranging from humid to arid (Weaver and Albertson 1956). Drought occurs often, and dry years tend to be grouped together, alternating with years of above-average precipitation. Temperatures range widely, with average monthly temperatures of 20°C in July and -15°C in January (Jensen 1972). Mean annual precipitation on the refuge is 42 cm, with a range of 22-74 cm (1936-89; USFWS unp'ubl. refuge files). Most (>75%) precipitation falls as rain between April and September, with May and June normally the wettest months. This study incorporated one extremely wet season (1993) and one drier season (1994). It was initiated at the end of 6 years of drought: most wetlands except major lakes were completely dry in early spring 1993. This drought ended during the 1993 breeding season, when 38 cm of rain fell in May, June, and July (54-year average for this period is 20 cm; USFWS unpubl. refuge files). Wetland basins began filling over the summer, and heavy snowfall over winter further filled basins. The 1994 field season was much drier, with only 14 cm of rain falling in May, June, and July. LNWR is within the northern region of the mixed-grass prairie (Singh et al. 1983), supporting a mix of approximately 60% cool-season (C3 photosynthetic pathway) and 40% warm-season (C4) grasses. Although upland vegetation is 11 dominated by needlegrasses and wheatgrasses, the plant community is extremely diverse; tall- and short-grass prairie types are intermixed. LNWR’s undulating topography provide diverse vegetation microsites based on soil type, moisture, and slope and aspect. Native mesic communities include big bluestem (Androoooon gerardiO, switchgrass (Panicum yirgatum), prairie dropseed (Sporobolus herterolepisl porcupine grass (Stipa spartea). and little bluestem (Schizachvrium scoparium). Native xeric sites (usually hilltops) are comprised mainly of blue grama (Bouteloua gracilis), threadleaf sedge (Carex filafgjja), prairie junegrass (Koeleria gracilis), and plains muhly (Muhlenbergia cuspidata). Family Asteraceae dominates the forb community, but over 40 other families are represented (LNWR Herbarium List 1991, unpubl. data). Western snowberry fSvmohoricarpos occidentalism is the dominant low (<1m) shrub, with silverberry (Elaeagnus commutata) and western wild rose (Rosa woodsii) also common. Tall (1-4m) shrub thickets of chokecherry (Prunus virginiana), serviceberry (Amelanchier alnifolia), and round-leaved hawthorn (Crataegus chrysocarpa) are scattered over the landscape. Exotic grasses (Kentucky bluegrass FPoa pratensisl, smooth brome (Bromus inermisl and quack grass (Agropyron regens]) have become established on large areas of the refuge. Before European settlement, the landscape of LNWR was a treeless expanse of mixed-grass prairie, maintained in a shorter grass, or even barren state due to frequent fire and bison impacts (Murphy 1993). Reconstruction of 12 pre-settlement vegetation patterns for the area indicate western snowberry probably covered <5% of upland areas, and almost no trees existed in the late 1800’s (Murphy 1993). Aspen was limited to aspen suckers and saplings around wetland borders. Murphy’s (1993) review of early historical accounts also indicated that the area had periodic rest from disturbance, during which vegetation would recover. Best estimates are that this region supports a- 5 to10- year fire return interval (Wright and Bailey 1982:81, Murphy 1:993, Bragg 1995). Although some of LNWR was tilled and farmed during the early 1900s, most (70%) upland areas remained unbroken and were, either rested or grazed season-long during 1930s-1970s. Settlers suppressed lightning fires, resulting in a loss of early successional vegetation (i.e., grassland). Western snowberry, quaking aspen, and exotic grasses have proliferated and now dominate the mixed-grass community. By 1985 western snowberry covered >50% of upland areas, and aspen had increased from no tree clumps in 1910 to 518 clumps covering 184 ha: most of LNWR could now be classified as aspen parkland instead of mixed-grass prairie (Murphy 1993). Several studies have made reference to the idea of present-day LNWR being better classified as aspen parkland. In a floristic study of northwestern North Dakota, Hegstad (1973) noted that the area of Missouri Coteau around LNWR is slightly higher and cooler, and has higher rainfall than surrounding areas of the Coteau, creating a unique floristic composition that “ ...has the 13 appearance of an aspen parkland...” (Hegstad 1973:26). Murphy (1993) also refers to the uniqueness of this area, delineating approximately 900 km2 of the Missouri Coteau (including LNWR) as being more suitable for development of woody vegetation. This localized area of mesic and forested prairie stands out as a small island on the range map of aspen in western United States (DeByIe and W inokur 1985:9). LNWR was created in 1935 "... as a refuge and breeding ground for migratory birds and other wildlife...” (Executive Order 7171-A 1935). The refuge mission is: To restore and preserve the indigenous biological communities of the mid- to late 1800s on a representative sample of the physiographic region known as the Missouri Coteau of the Northern Great Plains’ mixed-grass pra irie .. Since the 1970s, USFWS has used prescribed fire to reduce woody vegetation and restore a more natural diversity of successional stages to LNWR. The refuge is divided into approximately 20 prescribed burn units, ranging from 5 to 2265 ha, with an average size of 310 ha. Between 1978 and 1993, 5-35% of the refuge was burned annually (n = 63 burns), with most prescribed burns (75%) conducted in summer (mid-July through August), and the remainder in late spring (late April through early May). Burns typically were conducted in 10-30 kmh wind, 20-40% relative humidity, and 10-30°C, and generally removed 80- 95% of above-ground biomass. Short-duration grazing has also been used since 1989 to help accomplish habitat management objectives. 14 Management geared toward restoring or maintaining indigenous plant and wildlife species at LNWR involves three phases (U S. Fish and W ildlife Service 1994): 1) Renovation, 2) Renovation-Maintenance, and 3) Maintenance. The first phase, Renovation, utilizes burning 3-5 times over 7-10 years. Currently 50% of the refuge is in this phase. The second stage, Renovation-Maintenance, involves about 3 years of grazing, 2-3 years of rest, and 1-2 burns over a 7-year period. 15% of the refuge is in this phase. In the third stage, Maintenance, burning and grazing would be alternated with rest periods of 2-3 years. None of the refuge is yet in this phase. Thirty percent of the refuge has not yet received any treatment except rest or season-long grazing from 1940-1970, and another 5% (all previously cropped) is in a re-seeding program for native grasses and forbs. At least 226 species of birds occur on LNWR1 with 104 species documented as breeding on the refuge (Murphy 1990). These include many wetland-dependent species (waterfowl, rails, shorebirds, passerines), as well as upland breeders (raptors, shorebirds, sharp-tailed grouse ITvmpanuchus ohasianellusl. passerines). Common migratory songbirds breeding on upland areas include eastern kingbird, house wren, Sprague's pipit, common yellowthroat, a variety of grassland ,sparrows (grasshopper, Baird's, Le Conte's, vesper, savannah, song, clay-colored), bobolink, western meadowlark, Brewer’s 15 blackbird, brown-headed cowbird, and American goldfinch (See Appendix A for scientific names of bird species). METHODS Study Design This study was designed in conjunction with an assessment of vegetation correlates of prairie bird abundance (Chapter 3). I measured bird abundance and vegetation characteristics on 160 (1993) and 150 (1994) sample points distributed over 9 independent burns. Sample point selection was done systematically in 1993, and was then modified to include a randomized, systematic design in 1994. Sample points were > 250 m apart to provide statistical independence in terms of birds and vegetation (Hutto et al. 1986, Ralph et al. 1993) (see Chapter 3). Sample points within a burn unit were not independent in terms of fire history. Therefore data within a burn were averaged when comparing birds and vegetation among units of different burn history. Plot Selection 1993 Field Season In 1993 sample points were located systematically over upland prairie across the study area. On a map of the study area I gridded each square mile section into 268- by 268-m blocks and used the grid intersections as potential 16 point count locations. This assured that all points were >250 m apart. To be included as sample points, points had to be: (1) located in "upland prairie" as delineated by the National Wetland Inventory (NWI) map of cover types of LNWR (NWI Project 1989), (2) >200 m from any aspen grove, (3) >50 m from roads or firebreaks, and (4) currently ungrazed. These restrictions limited confounding effects from surrounding habitats and alterations, allowing me to concentrate on upland prairie treated with fire. Large areas of cropland that had been tilled through the 1950s also were omitted, but pre-1935 cropland areas, which were revegetated with native plants and appeared similar to existing native prairie, were included for sampling (about 12% of points were in pre-1935 cropland). Although these points were not chosen using a random sampling probability design, the grid pattern placed points in essentially a random pattern relative to landscape features (i.e., the grid pattern did not follow any obvious environmental gradients or biases). In all likelihood selected points were independent of vegetation and environmental patterns. Of 185 selected sampling points, the 10 most remote were omitted to improve sampling efficiency, and an additional 15 were omitted after ground- truthing revealed unexpected problems (e.g., chemical spraying/mowing for leafy spurge [Euphorbia esulal). One.hundred and sixty points ultimately were selected and sampled. A fie ld assistant and I located points in the field by 17 pacing on a compass bearing from marked section corners and then ground- truthing with aerial photographs (1:7920). Each point was marked with a flagged, I -m tall wooden stake painted white at the top. A 50-m radius was measured and flagged at a conspicuous spot. Points could be easily relocated by an observer travelling north-south or east-west from point to point. 1994 Field Season In 1994 I used the same study area as 1993, but plot selection incorporated a random sampling design. I generated a random, systematic grid for each square mile section by randomly picking 2 numbers between 0-250 m to serve as the X and Y coordinates in the southwest corner of the section. From this random starting point, I gridded points every 250 m north and south across the grid. W ith this grid overlaid on it, each square mile section had 49 possible point count locations (each grid intersection was a possible point count location). I then excluded any points not meeting the 1993 selection criteria, although I reduced the, buffer from roads and aspen from 200 m to 100 m based on 1993 sampling observations, and added a 50 m buffer from any seasonally flooded wetland zone because of high water levels. To improve sampling efficiency, the 10 most remotely-situated points in each burn unit were omitted from consideration. This left a total of 265 potential points, from which I randomly selected 150 sample points. 18 A field assistant and I located points in the field as in 1993, and measured and flagged 30-, 50- and 75-m radii with 0.2-m wire surveying flags to facilitate accurate distance estimation to birds., If ground-truthing revealed unexpected problems, e.g., chemical spraying for leafy spurge, we paced an additional 50 m on the same orientation and placed the point. Fire History Fire history for each plot was compiled from LNWR maps and narratives (Table 1). Fire history was described as: 1) number of years since the plot was burned (0.5-8 or >80 years), and 2) number of times the plot was burned in last 15 years (0-4 times). Fire variables were later divided into categories to test for differences in bird abundance and vegetation among them. . Number of years since last burn and number of burns were correlated (r = -0.62), i.e., areas burned recently tended to have been burned many times. This could potentially confound results, making it difficult to tease out relative impacts of each variable. As both variables played a role in defining fire history, I combined them into an index to describe the amount of fire an area had experienced: Fire Index (FI) = Number of Burns/Years Since Last Fire Using this formula, sampled burn units were scaled from 0 (no fire) to 6 (many burns, recently). Although there may be problems extrapolating this 19 precise index to all fire histories, it appeared to aptly describe the 9 burns sampled in this study, extending them along a reasonable gradient of amount of fire experienced. Table 1. Prescribed burn units sampled, ,LNWR1 1993 and 1994. Burn name Size (ha) 1993 1994 na No. burns YSFb Fire Index0 na No. bums YSFb Fire •Index0 Unburned - West 526 17 0 >80 0.00 12 0 >80 0.00 Unburned - Cd 75 3 0 >80 0.00 — — — — Unburned - S.C.Core 445 15 0! >80 0.00 18 0 >80 0.00 Green Needle 398 11 1 6 0.17 6 1 7 0.14 Wilderness 2265 34 2 5 0.44 25 2 6 0.33 N. Dead Dog Slough 89 5 4 7 0.57 9 4 8 0.50. Kruse 645 23 2 3 0.67 20 3 0.5 6.00 Aspen 518 10 2 0.5 4.00 20 2 2 1.00 . Thompson Lake 372 14 4 1 4.00 20 4 2 2.00 Teal Slough 494 28 4 1 4.00 20 4 2 2.00 a Number of sample points within burn unit b YSF = Years Since last Fire 0 Fire Index = Number of burns/years since last fire d Not used in fire analyses in 1993 and not sampled in 1994 Bird Abundance Sampling To estimate bird abundance, I used the fixed-radius method for point counts (Hutto et al. 1986) in 1993. In 1994 I modified methods to include distance sampling (Buckland et al. 1993). This involved estimating the distance 20 to each bird in 5 broad distance categories rather than merely estimating each bird as "in or out" of a fixed, 50-m radius. To meet the assumptions of distance sampling, observers must strive to ensure that accurate measurements are made of all birds on or near the point, prior to any undetected movement. Results of bird sampling.can be influenced by varying skill levels and training of observers (Reynolds et al. 1980, Kepler and Scott 1981). To minimize observer differences, I allowed 2 weeks of training for observers to practice point counts and standardize techniques while plots were being located and marked. We (1 part-time and 2 full-time observers) spent each morning bird-watching to become familiar with local birds by sight and sound. Observers also estimated distances to individual birds or songs and then measured the true distances with a measuring tape or by pacing. Point counts were conducted only on mornings when weather conditions did not impede detection of birds (i.e., no rain, fog, or wind >15 kmh). Counts began one-half hour before sunrise and continued until 0830-0900 hours when the wind generally began to increase. Each point was surveyed 3 times during the breeding season. Observers and the order in which points were visited were rotated to minimize sampling bias. Surveying was conducted during 29 May -7 July in 1993, and 26 May - 24 June in 1994. Observers approached each point quietly, being alert for any birds that might flush from the area. It was critical to record distances from the point to 21 birds before they were disturbed by the observer's approach., Such birds were included in the tally for that point at the distance at which they were first observed. Observers then stood at point center for 10 minutes, recording all birds seen or heard and estimating their respective distances from point center. In 1993 observers merely distinguished between birds within a 50-m radius and those farther away (out to an unlimited distance). In 1994 observers placed birds in the following distance categories: 0-14.9 m, 15-29.9 m, 30-49.9 m, 50-. 74.9 m, and >75 m (out to an unlimited distance). Again, each bird was placed in the distance category at which it was first observed, i.e., if a bird was observed flying in from >75 m and perching in the 50-75 m category, it was recorded as >75 m because that is where it was first observed (each count was essentially instantaneous [Buckland et al. 1993]). Observers recorded individuals as ( I) singing males or (2) those detected by sight only or call only. Males that sang in flight over a point (within a cylinder) were recorded as singing males. This applied mostly to Sprague's pipits and bobolinks. Brown-headed cowbirds were counted differently, as they do not sing and defend territories in the same manner as most passerines. Observers recorded all individual cowbirds seen or heard by their sex: male, female, or unknown sex. Unidentified birds were pursued at the end of the count for definite identification. If there was any question as to the distance category 22 of an individual, observers confirmed the distance by pacing to the observed location at the end of the 10-minute count period. In 1993 all bird species seen or heard were recorded, but only passerine species were included in the final data set. Most other birds were poorly surveyed with these point count methods, with few individuals recorded within 50 m of points. In the open habitat of LNWR, observers heard many wetland- and aspen-associated birds from great distances (> 400 m). Beyond providing a species list for an area, observations from such distances are of little value, especially in a study assessing habitat associations, and often distract observers. In 1994 surveys, observers recorded only passerines and upland­ nesting shorebirds. These shorebirds (marbled godwit, upland sandpiper, willet, and killdeer) were of particular interest because they utilize prairie habitats and thus were included in counts. None was adequately surveyed, however, and they were omitted from subsequent analyses. Vegetation Sampling Vegetation structure and composition were measured at each bird sampling point using the same methods each breeding season. Sampling was conducted during 28 June - 7 August in 1993, and 28 June - 28 July in 1994. Twenty subsample points were located along 2 transects within each.bird plot. Both transects were positioned on the same random compass bearing (i.e., 23 parallel), each a different random number of paces from plot center. Ten subsample points were then located systematically every 10 paces along each transect. The following measurements were taken at each of the 20 subsample points and averaged for the plot: 1) Visual obstruction - Using a Robel pole (Rebel 1970), visual obstruction measurements were taken in each of the 4 cardinal directions from a height o f I m and a distance of 4 m. Each measurement was taken at the mark oh the pole where vegetation began to hide the pole 100%. Measurements were rounded to the next lowest half-decimeter, except in the first dm height interval where they were rounded to the next lowest quarter-dm. The 4 measurements were averaged for each subsample. 2) Litter depth - This was measured directly (cm) by lowering a thin (6-7 mm diameter), metal rod vertically into the litter layer 30 cm east of the Robel pole. Dead vegetation from previous years that was standing but no longer vertical was considered litter when that layer became roughly continuous down to the ground, forming a mat-like layer. 3) Vertical vegetation density and percentage live vegetation - These were measured along with litter depth, using the same thin metal rod, following the methods of Rotenberry and Wiens (1980). For each dm height interval, the total number qf "hits" or contacts of vegetation were counted. Each hit was recorded as either live (current year's growth) or dead (not current year's 24 growth). Total, number of hits represented vertical density. The maximum dm height interval with vegetation contacts was considered a measure of maximum height. Percentage of hits represented by live vegetation also was calculated. 4) Coverage estimation - 1 visually estimated percentage areal cover of shrubs, forbs, grasses, litter, clubmoss, and bare ground within a 1-m diameter circular quadrat centered on each Rebel pole, using coverage classes of Daubenmire (1959). 5) Plant species associations - Each subsample was assigned to 1 of 28 plant species associations, based on a modification of mixed-grass prairie community types developed for northwestern North Dakota by Hegstad (1973). See Appendix B for descriptions of plant associations. Data analyses Vegetation Data Reduction For each vegetation structure variable, I calculated the mean and coefficient of variation (CV) for the 20 subsamples at each point. Coefficient of variation can be visualized as a measure of horizontal “patchiness” or heterogeneity of that vegetation attribute (Roth 1976). To reduce overlap in characteristics measured and high correlations among variables, I omitted some variables from analyses, as described below. Fourteen vegetation structure variables and 9 plant associations ultimately were retained (Table 2). 25 Table 2. Description of vegetation variables measured at point count locations. Values are mean or Coefficient of Variation (CV) of 20 subsamples at each point. Plant associations are expressed as the mean frequency of the association in the 20 subsamples. Vegetation code Description Structural Variables MEANSHRB Average percent areal cover of shrubs MEANFORB Average percent areal cover of forbs MEANGRAS Average percent areal cover of grass CVSHRB Coefficient of variation of MEANSHRB CVFORB Coefficient of variation of MEANFORB CVGRAS Coefficient of variation of MEANGRAS MEANROB Average visual obstruction (dm) - Robel height/density CVROB Coefficient of variation of Robel readings MEANLITD Average litter depth (cm) CVLITD Coefficient of variation of litter depth, TOTHITS Total number of vegetation contacts (“hits”) CVTOTH IT Coefficient of variation of TOTHITS PCTLIV Percentage of total hits that were live CVPCTUV Coefficient of variation of PGTLIV Plant associations NATIVE Upland native arasses (Stioa. Aoroovron. Koeleria, and Bouteloua spp.) POPR Kentucky bluearass (P. Pratensis) POPR/NTVS BROAD Codominance of Kentucky bluegrass and native grasses Broad-leaved, exotic arasses (B. inermis. A. reoens)BROAD 26 Table 2. Continued. Vegetation code SYOC/EL SYOC/EXTC SHRB/POPR SHRB/NTVS WETLAND Description W. snowberry (Si occidentalism with scattered Silverberry shrub (E. commutatal Codominance of SYOCEL and broad-leaved, exotic grasses Codominance of SYOCEL and Kentucky bluegrass Codominance of SYOCEL and native grasses Low prairie, wet meadow, or central marsh zones of wetland Percent cover of bare ground and clubmoss were omitted from analyses because they occurred infrequently (coverage < 5%) and seldom varied among plots. Percent cover of litter was dropped in favor of litter depth. Measuring “areal” cover did not adequately capture the amount of litter present below the shrub overstory; litter depth better reflected amounts of litter. Robel visual obstruction and mean maximum height (highest dm on vertical rod that contained hits) were highly positively correlated (r=0.901 in 1993 and r=0.856 in 1994), and had identical correlations with bird species. They essentially measured the same characteristics, and because maximum height was a more nebulous measurement, it was dropped in favor of the more standard Rebel reading. Plant associations were summarized as frequencies for each point. Of the 28 original plant associations, 6 (ANGE, FESC1 ST/SC/MU, AMPRCR, 27 POTR/SALX, and MELI) (see Appendix B for codes) were omitted because they occurred on less than 8 plots or had less than 21 total observations and could not be meaningfully combined with another association. Based on a more thorough knowledge of the plant communities, and on similarities in correlations with bird species, remaining associations were combined into 9 types (Table 2). Unfortunately, native mesic grasses (e.g., big and little bluestem, porcupine grass) were not represented in these plant associations because most mesic sites at LNWR are dominated with invasive smooth brome and quack grass, and thus were classified under the association of broad-leaved, exotic grasses. Frequency of these plant associations was calculated for each sample point. Each plant association had a relative frequency at each point between 0 and 20. For example, if Kentucky bluegrass was the plant association at 8 of 20 subsamples, then it had a frequency of 8 for that point. If not recorded, it had a frequency of 0. Bird Data Summarization Program DISTANCE (Buckland et al. 1993) was used to construct detectability profiles for 10 individual bird species in 1994 and to assess suitability of 50- and 75-m radii for estimating relative abundance (Rotella et al. In review). Based on these results, I used a 75-m radius for 1994 point counts. Abundance data for a given species were averaged from 3 visits conducted at a sampling point during a breeding season. Several abundance I measures were calculated, both using only singing males. (Singing males provide the best indices of abundance because female grassland birds are secretive and not reliably surveyed with these methods. For several species I never observed a non-singing individual.) First, I calculated the mean number of singing males of each species per 50- (1993) or 75-m (1994) radius point count (average of 3 visits). Then I calculated the percentage of points at which a species was detected, i.e., if a species was detected at 20 of the 150 points, its percent occurrence was 20/150 = 0.133 x 100 = 13.3%. Abundance for brown- headed cowbirds was calculated differently: I divided total number of cowbirds seen or heard by two, for comparability with singing male data. Total passerine abundance was simply the total number of singing male passerines at each point averaged over 3 visits. Species richness was calculated for singing males using the total number of species observed within a 50- (1993) or 75-m (1994) radius of each point over 3 visits (i.e., cumulative richness). To avoid bias due to different number of points sampled in each burn unit, I chose a random subsample of 6 points (lowest common denominator) from each burn unit and calculated richness based on those data. Abundance of several bird species differed significantly between the 2 breeding seasons. Because of this, and other strong seasonal differences (e.g., \ weather extremes), data for 1993 and 1994 were analyzed separately. 28 i J 29 Statistical Tests Only bird species detected at >10% of points were included in statistical analyses. Lack of normality and constant variance of many bird and vegetation data required the use of nonparametric techniques to test for differences in vegetation and bird abundance among categories of burns. I used Kruskal- Wallis 1 -way Analysis of Variance (SAS Institute 1988) to test the hypotheses that there were no differences in vegetation or bird abundance among categories of burns. Post-hoc multiple comparisons were conducted according to Conover (1980:231). To test these hypotheses, I categorized 2 measures of fire history: 1) Years since last fire (1-3, 5-7, or >80 years in 1993 and 2, 6-8, or >80 years in 1994) 2) Number of burns in last 15 years (0, 1-2, or 4). To reduce confounding effects between years since last fire and number of burns when comparing vegetation or birds among categories of years, since last fire, I left out the one area (Green Needle Burn) that had received only one burn in the previous 15 years. By using only areas having 2 or 4, or no burns, I thus controlled for some of the variability due to number of burns. Because of reduced power to detect differences due to small sample sizes, differences were considered significant when P < 0.10. 30 I used simple linear regression (Chatterjee and Price 1991) to quantify relationships between vegetation variables or bird abundance and the fire index. Regressions were performed on vegetation variables or mean bird species abundance for each of 8 burn units rated by fire index. Examination of normality and residual versus predicted value plots did not show dramatic departures from normality or constant variance for bird species in 1994, but 1993 data diagnostics indicated violation of regression assumptions for most species. Data transformations were attempted, but did not remedy these problems, so only 1994 bird data were used in these analyses. Several vegetation variables did not meet regression assumptions and were not used: frequency of snowberry/exotic grasses and shrubs/Kentucky bluegrass in 1993, and CV litter depth and frequency of snowberry/siIverberry in both years. In all statistical tests, areas that had not had a full growing season of vegetation recovery (i.e., years since last fire = 0.5) were left out of analyses. Burned in the previous August, these areas generally had extreme vegetation and bird responses, unreflective of longer-term fire effects I sought to document. These type of burns could not reasonably be grouped with others, and were not replicated sufficiently for statistical comparisons with other fire categories. But, bird abundances for them are reported in Appendix D. Areas burned in the spring previous (i.e., years since last fire = I), having had 1 full growing season for vegetation to recover, were included in all analyses. 31 Principle Components Analysis Because many vegetation structure variables were correlated, I used Principle Components Analysis (PCA) to identify major gradients in vegetation structure (SAS Institute 1988). Although most ecological data violate some fundamental assumptions of PCA (e.g., normality) for purposes of hypothesis testing, it is useful for synthesizing and describing patterns in data (Gauch 1982:143). To synthesize these results conceptually, I plotted the 1994 sample points by fire history on a plot of PC 1 versus PC 2 to describe each burn unit’s location in PC vegetation space. I then plotted a 95% confidence ellipse (Wilkinson 1990) based on the mean for each fire interval and superimposed the ellipses onto a single graph for purposes of comparison. RESULTS Vegetation and Fire General trends in vegetation variables among categories of burn history were similar in each of 2 years analyzed, whether differences were significant both years or not (Appendix C, Tables 18-21). Most significant differences in vegetation were between areas burned 1-3 or 2 years previously and areas ranging from 5-8 years to >80 years postfire. Shrub cover, visual obstruction, litter depth, and vegetation density (either means or CVs or both) were highest in unburned areas and lowest 1-3 years postfire, while grass cover and percentage 32 of live vegetation were lowest in unburned areas and highest 1-3 years postfire. Shrub cover and visual obstruction were lowest and grass cover was highest in areas burned 4 times. Areas with 4 burns also had the least variability in visual obstruction of vegetation, and the highest frequencies of broad-leaved, exotic grasses. To facilitate general comparisons with other geographic areas, ranges of vegetation values across all sample plots are given in Appendix C, Table 17. Vegetation Along the Fire Index Gradient When simple linear regressions were performed on means and CVs of vegetation variables for each of 8 burn units rated by fire index (Table 3), trends for individual variables were similar between years, even if not significant in both. Five variables had significant regressions in 1993 only, and 4 variables had significant regressions in 1994 only. Grass cover increased with fire in both years; while shrub cover, CV grass cover, and vegetation density decreased in both years. Kentucky bluegrass/native grasses and broad-leaved, exotic grasses both increased in frequency with amount of fire, and variances in vegetation variables (CVs) generally decreased with amount of fire. 33 Table 3. Relationships between individual vegetation variables and fire index (see text for explanation of fire index) based on simple linear regression. Only results for vegetation variables with significant (P < 0.05) regression relationships in at least 1 year are reported. 1993 1994 Response to fire3 Vegetation Variable R z F P R z F P Shrub cover 0.56 7.63 0.03 0.53 6.73 0.04 - Grass cover 0.86 36.24 0.00 0.83 28.49 0.00 + CV forb cover 0.11 0.75 0.42 0.53 6.73 0.04 - CV grass cover 0.49 5.69 0.05 0.55 7.28 0.04 - CV visual obstruction 0.22 1.67 0.24 0.68 12.95 0.01 - Litter depth 0.90 57.15 0.00 0.15 1.05 0.35 - Density (# hits) 0.65 10.92 0.02 0.66 11.84 0.01 - % live vegetation 0.90 54.62 0.00 0.38 3.69 . 0.10 + CV % live veg. 0.67 12.00 0.01 0.26 2.10 0.20 - K.bluegrass/natives 0.48 5.51 0.06 0.60 9.09 0.00 + Br-Ieaf exotic grass 0.73 16.40 0.01 0.34 3.15 0.13 + Shrub/K. bluegrass b b b 0.50 5.98 0.05 - Shrub/natives 0.50 5.97 0.05 0.40 4.07 0.09 - aFire response based on slope (+ or-) of regression line b Variable did not meet regression assumptions in 1993 Vegetation Variables Summarized - PCA For each year of vegetation data (1993 and 1994), 4 principle components had eigenvalues greater than 1, accounting for 81% and 83% of the variation in vegetation, respectively. Overall the eigenvectors and gradients for 34 the 2 years were remarkably similar for the first 2 components (Table 4 and Figure 3). Table 4. Eigenvector loadings (based on correlation matrix) of principle components (PC1 and 2) of vegetation structure variables in 1993 and 1994. Vegetation Variable 1993 1994 PC .1 PC 2 PC 1 PC 2 Shrub Cover 0.33 0.29 0.33 0.34 Forb Cover -0.17 -0.39 -0.15 -0,28 Grass Cover -0.27 0.27 -0.28 -0.20 CV Shrub Cover -0.20 -0.32 . -0.19 -0.30 CV Forb Cover 0.20 0.42 0.22 0.24 CV Grass Cover 0.29 0.12 0.29 0.26 Visual Obstruction 0.33 0.24 0.31 0.23 CV Visual Obstruction 0.08 -0.12 0.02 0.16 Litter Depth 0.33 -0.23 . 0.32 -0.29 CV Litter Depth -0.32 0.28 -0.30 0.34 Density (total # hits) 0.34 -0.02 0.37 -0.16 CV Density -0.08 -0.11 • -0.24 0.25 % Live Vegetation -0.33 0.33 -0.31 0.36 CV % Live Vegetation 0:24 -0.29 0.20 -0.22 Eigenvalue 5.51 2.46 5.33 2.82 % Variance explained 39.32 17.6 38.10 20.11 Total Variance explained 39.32 56.93 38.10 58.22 35 -2a " D C CUI C 01 r CD " D C 0 3 j§ CO CN O CL Short, sparse, grass- and shrub- dominated; low litter Tall, dense, shrub-dominated; low litter Short, sparse, herbaceous with litter Tall, dense, shrub- and forb- dominated, with litter Short, sparse, PC 1 Tall, dense, grass-dominated * shrub-dominated Figure 3. Description of first 2 Principle Components based on vegetation structure. The first axis (PC1) accounted for 39% and 38% of the variation in 1993 and 1994 respectively, and represented a gradient from short, sparse, grass- dominated vegetation to tall, dense, shrub-dominated vegetation (Table 4 and Figure 3). This gradient was identical for the 2 years. The second axis (PC2) accounted for 18% and 20% of the variation, in 1993 and 1994 respectively, and represented a gradient from litter and forb-dominated vegetation to mostly live, shrub-dominated vegetation with few forbs (Table 4 and Figure 3). There was more variability between years on PC 2, but it was generally similar. A third axis (not shown) accounted for only 9% and 12% of the variation and represented a 36 gradient of increasing horizontal patchiness in terms of height and density (structural complexity) of the vegetation. Because this third axis explained only a small portion of the variation, and 3 dimensions are often difficult to visualize, . it was not examined further. Confidence ellipses for 1994 burn units in PC vegetation space indicated that the unit with the most distinct vegetation was the area burned the previous August (Figure 4). Its confidence ellipse was located at the extreme negative end of PC1, defined by very short, sparse grassy cover, and the extreme positive end of PC2, indicating low litter with shrubs and few forbs. An area burned 2 times, the last being 2 years previous, was also distinct, but closer to the plot origin (i.e., “average” vegetation characteristics). This area fell to the short, sparse, grassy (-) side of PC1 and was centered on zero on PC 2, indicating moderate amounts of litter, forbs, and shrubs. Areas burned 2 years previous, but 4 times in the last 15 years, fell wholly in the short and sparse, litter-, grass- and forb-dominated quadrat of the plot and overlapped slightly with an area burned 1 time, 7 years previous. This I -burn area had the largest confidence ellipse, indicating high variability in terms of vegetation. It fell mostly in the quadrat defined by tall, dense, shrub- and forb- dominated vegetation with much litter. Complete overlap with this ellipse occurred for areas burned 6 years ago (2 times) and 8 years ago (4 times). This illustrates great similarity in vegetation 6, 7, and 8 years postfire. 37 Shrubby, few forbs, low litter 3 burns / 0.5 yrs postfire 0 bums / >80 yrs postfire 2 burns / 2 yrs postfire 4 burns / 8 yrs postfire 4 burns / 2 yrs postfire 2 burns / 6 years postfire I burn / 7 yrs postfireHigh litter and forbs, few shrubs short, sparse ' 0 1 tall, dense grassy -------------------------------------------------------------------------------------- > shrubby Figure 4. Composite confidence ellipses for areas with different burn histories. The ellipses represent 95% confidence intervals around the mean position of sample plots with similar burn histories, located with reference to Principle Components 1 and 2. Based on 1994 data. The confidence ellipse for unburned plots overlapped no others and was located on the tall, dense, shrub-dominated side of PC1, and the shrubby, low forb/low litter side of PC2. Although lower amounts of litter may seem surprising for unburned plots, a portion of the unburned plots were decadent snowberry 38 , shrub with little understory except lush, green growth of smooth brome. This shrub/exotic grass combination probably accounts for the higher ranking toward the “shrub-dominated, low litter” side of PC 2. Bird Sampling I observed 43 and 41 passerine bird species during point counts in 1993 and 1994, respectively. In 1993, 744 singing males of 19 species were recorded on 50-m radius plots. Using the 75rm radius cutoff in 1994, we observed 1658 singing males of 24 passerine species (see Appendix D for list of species and relative abundances). Nine and ten species were detected at >17 points (10%) in 1993 and: 1994, respectively: Baird’s sparrow, bobolink, brown-headed cowbird, clay- colored sparrow, common yellowthroat, grasshopper sparrow, Le Conte’s sparrow (1994 only), savannah sparrow, Sprague’s pipit, and western meadowlark. Clay-colored sparrow and savannah sparrow were by far the most abundant species sampled; both had mean abundances well over twice that of any other species during both seasons. All further analyses consider these 10 species, total passerine abundance, and species richness. 39 Birds and Fire Several species were completely absent from unburned areas in both years. Baird’s sparrows, grasshopper sparrows, Le Conte’s sparrows, Sprague’s pipits, or western meadowlarks were never detected during 3 visits to 65 different point count locations in unburned areas. Bobolink were absent from unburned areas except for 2 detections of a singing male in 1993. In contrast, none of 10 species examined was completely absent from areas treated with fire. Species Abundance in Relation to Years Since Last Fire Statistical comparisons of bird abundance in relation to years since fire were not made because replicates of many burn histories were lacking, but general trends for the 2 years of data were apparent (Appendix D, Tables 23 and 24). Assessment of these trends revealed Baird's sparrows, bobolinks, Le Conte's sparrows, and western meadowlarks typically reaching highest abundances 1 -3 years postfire, and species richness peaking 1 and 2 years postfire. Sprague's pipits and grasshopper sparrows were most abundant about 2-3 years postfire, but sometimes up to 7 years postfire, and savannah sparrows reached maximum abundance 6-7 years postfire. Clay-colored sparrows and common yellowthroats were most abundant > 80 years postfire (Appendix D, Tables 23 and 24). 40 In 1994, abundance of bobolink, clay-colored sparrow, Le Conte's sparrow, and Sprague’s pipit differed (P < 0.10 ) among years-since-fire groups (Table 5). Bobolinks, Le Conte's sparrows, and Sprague’s pipits reached maximum abundance in areas burned 2 years previous, whereas clay-colored sparrows were most abundant in unburned areas. Total passerine abundance was remarkably similar among years-since-fire groups (P = 0.90), and species richness was not different (P = .17). Categories of years since fire differed slightly in 1993: 1-3 years, 5-7 years, and >80 years. Although most 1993 trends (Table 6) were similar to 1994 trends, smaller plot sizes used in 1993 made it more difficult to detect ■ differences in bird abundance among burn categories. In 1993, abundance of grasshopper sparrow and western meadowlark was highest 1-3 years postfire, while abundance of common yellowthroat was lowest then. Species Abundance in Relation to Number of Burns In 1994, Baird's sparrow, bobolink, and clay-colored sparrow differed among categories of number of burns (Table 7). In 1993, abundances of bobolink, clay-colored sparrow, and Sprague's pipit differed over the same categories (Table 8). Trends in each of the 2 years were similar for most species. Baird's sparrow, bobolink, and Sprague's pipit reached maximum abundance in areas burned 4 times, and clay-colored sparrow was most abundant in areas with no burns. Table 5. Bird species abundance and standard error by category of years since last fire, LNWR1 1994. Abundance is mean number of singing males per 75-m radius point count. For species showing significant differences among categories, medians with similar letters do not differ. Years since last fire 2 years (n=3a) 6-8 years (n=2a) >80 years (n=2a) Bird species Median Mean SE Median Mean SE Median Mean SE p-value6 Baird’s sparrow 0.38 0.37 0.31 0.08 0.08 0.01 0.00 0.00 0.00 0.135 Brown-headed cowbird 0.08 0.09 0.08 0.11 0.11 0.15 0.02 0.02 0.01 0.915' Bobolink 0.45A 0.41 0.15 0.15B 0.15 0.01 0.00B 0.00 0.00 0.065 Clay-colored "sparrow 0.90A 0.86 0.22 1.38B 1.38 0.22 2.22B 2.22 0.24 0.069 Common yellowthroat 0.02 0.03 0.04 0.15 0.15 0.10 0.40 0.40 0.07 0.107 Grasshopper sparrow 0.57 0.44 0.27 0.15 0.15 0.21 0.00 0.00 0.00 0.143 Le Conte’s sparrow 0.08 A 0.10 0.03 0.03B 0.03 0.04 0.00B 0.00 0.00 0.075 Savannah sparrow 1.00 0.99 0.05 1.39 1.39 0.18 0.96 0.96 0.33 0.153 Sprague’s pipit 0.08 A 0.10 0.04 0.01 B 0.01 0.01 0.00B 0.00 0.00 0.079 Western meadowlark 0.10 0.11 0.03 0.05 0.05 0.08 0.00 o.oo 0.00 0.215 All passerines 3.94 3.61 0.59 3.64 3.64 0.53 3.73 3.73 0.11 0.898 Species richness 5.67 5.22 1.55 3.83 3.83 0.94 3,08 3.08 0.35 0.165 a Number of burn units in category b P-value associated with KruskaLWaIIis AN OVA. Post-hoc multiple comparisons conducted according to Conover (1980). Table 6. Bird species abundance and standard error by category of years since last fire, LNWR1 1993. Abundance is mean number of singing males per 50-m radius point count. For species showing significant differences among categories, medians with similar letters do not differ. Years since last fire 1-3 years (n=3a) 5-7 years (n=2a) >80 years (n=2a) Bird species Median Mean SE Median Mean SE Median Mean SE p-value6 Baird’s sparrow 0.13 0.12 0.02 0.05 0.05 0.08 0.00 0.00 0.00 0.143 Brown-headed cowbird 0.08 0.11 0.06 0.05 0.05 0.07 0.03 0.03- 0.02 0.331 Bobolink 0.43 0.33 0.23 0.09 0.09 0.06 0.04 0.04 0.06 0.331 Clay-colored sparrow 0.26 0.28 0.05 0.40 0.40 0.19 0.57 0.57 0.01 0.107 Common yellowthroat 0.00 A 0.01 0.02 0.12B 0.12 0.11 0.24B 0.24 0.02 0.065 Grasshopper sparrow 0.18A 0.22 0.08 0.06B 0.06 0.01 O.OOC 0.00 0.00 0.065 Savannah sparrow 0.35 0.39 0.16 0.53 0.53 0.29 0.25 0.25 0.07 0.253 Sprague’s pipit 0.10 0.09 0.03 0.11 0.11 0.13 0.00 0.00 0.00 0.148 Western meadowlark 0.1 OA 0.10 0.02 0.03B 0.03 0.05 0.00B 0.00 0.00 0.079 All passerines 1.73 1,75 0.18 1.51 1.51 0.23 1.22 1.22 0.17 0.107 Species richness 3.00 3.07 0.12 2.70 2.70 0.42 2.50 2.50 0.14 0.160 a Number of burn units in category b P-value associated with Kruskal-Wallis ANOVA. Post-hoc multiple comparisons conducted according to Conover (1980). Table 7. Bird species abundance and standard errors by category of number of burns in previous 15 years, 1994 Abundance is mean number of singing males per 75-m radius point count. For species with significant differences over groupings, medians with similar letters do not differ. Bird species Number of burns in last 15 years 0 bums (n=2a) 1-2 burns (n=3a) 4 burns (n=3a) Median Mean SE Median Mean SE Median Mean SE p-valuea Baird’s sparrow 0.00 A 0.00 0.00 0.06B 0.07 0.02 0.38B 0.37 0.30 0.066 Brown-headed cowbird 0.02 0.02 0.01 0.02 0.08 0.12 0.08 0.09 0.09 0.945 Bobolink 0.00A 0.00 0.00 0.16B 0.16 0.10 0.45B 0.38 0.20 0.093 Clay^colored sparrow 2 .2 2 k 2.22 0.24 1.50B 1.36 0.27 0.90B 0.91 0.30 0.069 Common yellowthroat 0.40 0.40 0.07 0.08 0.13 0.08 0.02 0.08 0.12 0.106 Grasshopper sparrow 0.00 0.00 0.00 0.29 0.31 0.18 0.57 0.39 0.34 0.220 Le Conte’s sparrow 0.00 0.00 0.00 0.05 0.05 0.04 0.08 0.07 0.07 ■ 0.311 Savannah sparrow 0.96 0.96 0.33 1.27 1.27 0.34 1.03 1.18 0.29 0.574 Sprague’s pipit - 0.00 0.00 0.00 0.06 0.05 0.04 0.07 0.07 0.08 0.233 Western meadowlark 0.00 0.00 0.00 0.08 0.06 0.06 0.10 0.08 0.08 0.342 All passerines 3.73 3.73 0.11 4.00 3.65 0.62 3.94 3.72 0.40 0.707 Species richness 3.08 3.08 0.35 3.67 3.89 0.54 5.67 5.11 1.73 0.236 ' a Number of burn units in category b R-value associated with Kruskal-Wallis ANOVA. Post-hoc multiple comparisons conducted according to Conover (1980). Table 8. Bird species abundance and standard errors by category of number of burns in previous 15 years, 1993. Abundance is mean number of singing males per 50-m radius point count. For species showing significant differences over groupings, medians with similar letters do not differ. Bird species Number of burns in last 15 years 0 burns (n=2a) 1-2 burns (n=3a) 4 burns (n=3a) Median Mean SE Median Mean SE Median Mean SE p-value15 Baird’s sparrow 0.00 0.00 0.00 0.11 0..08 0.07 0.10 0.08 0.07 0.365 Brown-headed cowbird 0.03 0.03 0.02 0.08 0.06 0.05 0.06 0.08 0.09 0.799 Bobolink 0.04A 0.04 0.06 0.06 A 0.06 0.01 0.43B 0.35 0.19 0.082 Clay-colored sparrow 0.57A 0.57 0.01 0.53A 0.49 0.14 0.26B 0.26 0.02 0.077 Common yellowthroat 0.24 • 0.24 0.02 0.03 0.03 0.02 0.04 0.08 0.11 0.125 Grasshopper sparrow 0.00 0.00 0.00 0.06 0.13 0.15 0.17 0.14 0.06 0.116 Savannah sparrow 0.25 0.25 0.07 0.57 0.56 0.23 0.35 0.45 0.25 0.120 Sprague’s pipit 0.00A 0.00 0.00 0.03B 0.05 0.04 0.12B 0.13 0.07 0.066 Western meadowlark 0.00 0.00 0.00 0.07 0.06 0.05 0.07 0.06 0.06 0.342 All passerines 1.22 1.22 0.16 1.61 1.56 0.20 1.67 1.73 0.19 0.120 Species richness 2.50 2.50 0.14 2.40 2.60 0.35 3.00 3.07 0.12 0.124 a Number of burn units in category b p-value associated with Kruskal-Wallis ANOVA. Post-hoc multiple comparisons conducted according to Conover (1980). 45 Bird Species Abundance Along Fire Index Gradient Eight bird species and species richness had significant (P < 0.05) relationships with amount of fire (i.e., fire index rating) in 1994 (Table 9). Six species and species richness responded positively to amount of fire, while 2 species responded negatively. Savannah sparrow, brown-headed cowbird, and total passerine abundance did not exhibit significant responses to amount of fire. Table 9. Relationships between bird abundance and amount of fife based on linear regressions of abundance on the fire index (see text for fire index). Bird species R F p Response to fire3 Baird’s sparrow 0.79 23.14 0.00 + Bobolink 0.97 202.67 0.00 + Grasshopper sparrow 0.49 5.70 0.05 + Le Conte’s sparrow 0.79 22.18 0.00 + Sprague’s pipit 0.65 11.08 0.02 + Western meadowlark 0.64 10.82 0.02 + Species richness 0.75 18.45 0.01 + Clay-colored sparrow 0.74 17.50 0.01 — Common yellowthroat 0.67 12.38 0.01 — Savannah sparrow 0.09 0.58 0.48 NS Brown-headed cowbird 0.18 1.28 0.30 NS All passerines 0.01 0.05 0.84 NS 3 Response to fire based on slope (+ or -) of regression line 46 DISCUSSION Vegetation and Fire Repeated fires reduced shrub cover, litter build-up, and vegetation height and density, while increasing percentage of live vegetation and grass cover on mixed-grass prairie. Native grasses, Kentucky bluegrass/native grasses, and broad-leaved exotic grasses increased in frequency with fire, but Kentucky bluegrass frequency generally did not differ among areas of different burn history. Fire decreased horizontal patchiness of grasses, forbs, live vegetation, and visual obstruction, and increased litter patchiness. These results are consistent with previous findings concerning the role of fire in maintaining mixed-grass prairie vegetation. Annual die-offs of grass and the resulting litter build-ups necessitate some type of defoliation in these semi- arid environments where plant decomposition is otherwise slow (Kucera 1981, Wright and Bailey 1982, Bragg 1995). Fire reduces vegetation biomass and litter and increases growth and reproduction of native vegetation (Kucera 1981, Wright and Bailey 1982). Vegetation in burned areas generally begins growing earlier, matures earlier, and produces more flowering stalks, especially of native plants (Ehrenreich 1959). This enhanced growth is thought to result from reduced litter, warmer soil temperatures, increased light for emerging shoots, and greater nutrient availability after burning (Ehrenreich 1959, Daubenhnire 47 1968, Kucera 1981). Suppression of woody vegetation is a more obvious and well-documented effect of fire on grassland habitats (review in Higgins et al. 1989). High coverages and frequencies of western snowberry were successfully reduced with fire at LNWR. Most significant differences in vegetation were between areas burned 1 -3 or 2 years previously and areas ranging from 5-8 years to >80 years postfire. The lack of differences between areas burned 5-8 years previous and areas unburned in >80 years may have been more a function of low statistical power than a lack of true differences, for examination of mean and median values revealed differences that appeared biologically meaningful. The PCA interpretation did show unburned areas as distinctly different from areas 6-8 years postfire in terms of vegetation, mostly in having more shrub cover. Birds and Fire Birds at LNWR were highly affected by amount and frequency of burning. In almost all cases trends were similar between the 2 years examined, yet 1993 trends often were not significant. Using higher bird numbers afforded by 75-m radius plots, the 1994 bird data were more powerful statistically and may better represent effects o f f ire on birds. 48 Bird species responded to fire in a variety of ways: positively, negatively, moderately, or negligibly. Baird’s sparrow, bobolink, grasshopper sparrow, Le Conte's sparrow, Sprague’s pipit, and western meadowlark responded favorably (i.e., increased abundance) to repeated fire. Brown-headed cowbird showed an inconclusive response to fire, clay-colored sparrow and common yellowthroat showed consistent, distinctly negative reactions to fire, and savannah sparrow was associated with intermediate stages of post-fire succession. Although species composition shifted, total bird abundance did not differ over burn categories in any analyses. Species richness was generally highest in areas experiencing more fire, and lowest in unburned areas and areas <1 year postfire. Other studies of prairie passerines have shown species richness highest in shrubby, late successional stages associated with lack of fire (Arnold and Higgins 1986, D. Johnson, Northern Prairie Science Center, unpubl. data). These studies have included trees and shrub thickets (and wetlands) in plots, which inflates total species richness by adding non-prairie species associated with woody habitats. Because I limited my plots to upland prairie and avoided aspen groves and wetlands, I exclusively sampled grassland habitats, and thus mainly grassland bird species. Had I included aspen groves common on unburned prairie at LNWR1 total species richness would have been highest on unburned prairie due to presence of many woodland bird species. For grassland managers charged 49 with maintaining a native bird community, emphasis on woody, later successional stages simply because they support greater avian diversity may be inappropriate. In strictly grassland habitats, my data suggest that burned (less shrubby) prairie will support maximum species richness of native grassland birds. Grassland bird species are typically delineated as either endemic (indigenous) or pandemic (more widespread). Of 10 species considered in my analyses, 2 (Baird's sparrow and Sprague’s pipit) are endemic grassland bird species (IVIengeI 1970, Johnsgard 1978), having evolved exclusively on the North American Great Plains. In addition, Johnsgard (1978) considered clay- colored sparrow and Le Conte’s sparrow endemic grassland species, but most authors classify them as pandemic because their ranges extend beyond the Great Plains. Clay-colored sparrow has extended its range east and north since the turn of the century as suitable early successional habitats were created by logging and agricultural activities (Knapton 1994), and Le Conte's sparrow is found in the taiga as well as grasslands (IVIenge11970). Savannah sparrow, bobolink, grasshopper sparrow, and western meadowlark are pandemic, or secondary grassland species, being widespread in grassland habitats throughout North America. Common yellowthroat is a widespread shrub- or wetland-associated species found throughout North America. Brown-headed cowbird was limited to open grasslands of the Great Plains before European 50 settlement, but expanded east and west during the 1800s and 1900s as forests were cleared and agriculture expanded (Mayfield 1965, Lowther 1993). Bird responses to fire observed in this study are consistent with trends seen with long-term songbird monitoring at LNWR and elsewhere in North Dakota. At LNWR, most grassland birds declined the first season postfire, and then increased greatly 2-3 years postfire (1978-1994; USFWS, unpubl. refuge files). Controls (unburned areas) continued to have low numbers of grassland species during the 16-year period. A 24-year study on the Missouri Coteau in south-central North Dakota (D. Johnson, Northern Prairie Science Center, unpubl. data) also documented immediate (1-year postfire), negative fire effects followed by positive responses for Baird’s sparrow, bobolink, grasshopper sparrow, savannah sparrow, and western meadowlark. Bobolink numbers appeared highest 2-3 years postfire compared with about 1 year postfire in my study, Baird’s sparrow numbers were highest 2-7 years postfire compared with 1-3 years in my study; and grasshopper sparrows peaked 3-4 years postfire compared with 2-3 years in my study. In fescue prairie in Saskatchewan, savannah sparrow and clay-colored sparrow were adversely affected by an October fire and had not recovered to control levels by 3 years postfire (Pylypec 1991). Baird’s sparrow, Sprague’s pipit, and western meadowlark numbers were adversely affected initially, but had recovered by the third year postfire. Adverse responses during the first year 51 postfire were more extreme than in this study, but an October fire would have allowed far less vegetation recovery by the next breeding season than the spring or summer burns done at LNWR. And on mixed-grass prairie in South Dakota, Western meadowlark and grasshopper sparrow also generally decreased after fire, but as in my study, bird densities 3-years postfire were often higher than pre-burn levels, pointing to long-term benefits of prescribed fire (Forde et al. 1984). Tallgrass prairie bird/fire studies show a similar, but much more accelerated, bird-species succession after fire (Herkert 19.94, Zimmerman 1993). In tallgrass prairie in Illinois, bird species recovery after fire was more rapid than in this study for many of the same bird species (Herkert 1994). Of 11 species examined, 7 recorded highest abundance 1 growing season postfire. This compares with highest abundances about 2-3 years postfire for my study, but these differences would be predicted given the higher fire frequency in tallgrass (3-4 years) versus mixed-grass prairie (5-10 years). Vegetation recovers from fire more quickly in tallgrass prairie, so bird species Succession also is accelerated. It is also instructive to look at bird communities on idle prairie in other studies (not necessarily fire studies) and compare them with communities of idle areas in this study. Most other fire studies merely point out that, for many bird species, abundance is depressed for 1 to several years postfire and then 52 recovers. My study illustrated that most grassland birds (6 of 9 species examined) were eventually excluded from mixed prairie untreated with fire (or some type of defoliation) for long periods. This was also suggested by a study of shrubby versus shrubless prairie on the Missouri Coteau in south-central North Dakota (Arnold and Higgins 1986), where many true grassland species were absent (Baird’s sparrow, savannah sparrow) or reduced (grasshopper sparrow, bobolink) on shrubby prairie such as that associated with a lack of fire. In much of mixed-grass prairie Canada, however, long-idle lands support high numbers of endemic grassland birds, especially Baird’s.sparrow and Sprague’s pipit, and frequent defoliations do not seem as critical as at LNWR. At Last Mountain Lake, Saskatchewan, native grasslands idle for 15 years supported the richest endemic avifauna, compared with invaded grasslands and annual and emergency hayfields (Dale 1992). Highest numbers of Baird’s sparrow were found on 32-year idled prairie at Matador Research Station in southwestern Saskatchewan, compared with grazed native and grazed introduced vegetation (Sutter et al. 1995). In Alberta, highest populations of Baird’s sparrows occurred in idle fescue grasslands and in “lush” mixed grasslands that had not been grazed or burned for about 20 years (Wershler et al. 1991). These mixed prairie areas are drier than those of northwestern North Dakota (Singh et al. 1983); vegetation of drier prairies exhibits slower recovery 53 after fire and creates a much longer fire return interval of up to 25 years (Bragg 1995). Pre- and Post-settlement Endemic Bird Populations After 4 renovation burns, areas at LNWR are believed to be roughly similar to pre-settlement prairie. High numbers of Baird's sparrow and Sprague's pipit in these areas are consistent with pre-settlement observations by Coues (1878). When the US. Northern Boundary Commission crossed North Dakota in 1873, Coues remarked repeatedly on the “trio of the commonest birds" encountered as the expedition passed through the Souris River region and onto the Missouri Coteau near present-day LNWR: chestnut-collared longspur, Baird's sparrow, and. Sprague's pipit. Baird's sparrows often outnumbered all other birds together in some areas, and Sprague’s pipits were sometimes “...so numerous that the air seemed full of them; young ones were caught by hand in the camp, and many might have been shot without stirring from my tent...” (Coues 1878:560). In 1972, after 100 years of settlement and agricultural development, Baird's sparrow and Sprague's pipit had declined to the point that they were no longer among even the 15 most common birds in North Dakota (Stewart and Kantrud 1972). Population declines in these species (Knopf 1994) 54 are assumed to be a result of large reductions in extent and quality of native prairie habitats. Mechanisms Underlying Bird Response To Fire Birds may ultimately respond to some combination of fire effects: changes in vegetation structure and composition, changes in functional responses such as primary production and energy transfer, and/or shifts in other organisms (e.g., insect prey) in response to these factors. Fire clearly alters habitat structure and composition, which consequently influence bird abundance and distribution. No studies to date have clearly teased out exactly why vegetation structure is important to grassland birds, but obvious roles in providing nesting cover, singing perches, protection from predators, and mobility in the grass while foraging seem important (Cody 1985), Specific bird species associations with vegetation are examined in Chapter 3 of this thesis. Although habitat structure has been shown to influences birds, it generally explains only a portion of the variability in bird abundance and distribution. Other possible influences are now being considered. Recently, levels of energy (i.e., productivity) in a system have been shown to best explain differences in species diversity (review in Wright et al. 1993). Several studies show corresponding increases in species richness and net primary productivity (NPP). 55 There is disagreement as to whether this increase is linear or if species richness increases and then drops off at highest levels of NPP (Abrams 1995). In either case, for semi-arid areas such as grasslands, which have relatively low levels of NPP, bird species diversity would be predicted to increase with elevated NPP. Fire greatly affects energy cycling and NPP in grasslands (Kucera 1981, Bragg 1995), and thus may affect bird species abundance and distribution. Studies in humid grasslands (i.e., ta l!grass) indicate highest productivity immediately after fire, and then decreasing productivity as litter accumulates with time since fire (Kucera 1981). In contrast, drier, shorter-grass prairies have lower productivity after burning. Production then increases with time since fire as litter build-up facilitates conservation of moisture, and eventually levels off. . The relationship between fire frequency and NPP has not been well- defined for mixed-grass prairie, but it would be reasonable to expect it to be intermediate between the 2 described above, with perhaps some small decline in productivity immediately following fire, but then a fairly rapid increase and period of high productivity that eventually drops off with time since fire. Although discussion here on fire and NPP is mostly speculative, it illuminates an intriguing and as yet unexamined aspect of grassland bird response to fire: that of primary productivity driving species diversity. More study is warranted. Whatever the causal mechanisms for associations among fire, vegetation, and birds, fire remains an integral component of mixed prairie ecosystems, and 56 along with climate, topography, and other natural disturbances defines prairie plant and bird communities. Scope and Limitations Before drawing conclusions from these results it is important to clearly define the scope of this study and consider design limitations, namely low statistical power, and high variability in grassland ecosystems and disturbance cycles on the Great Plains. Statistical power is a function of sample size, variance, desired size of differences to be detected, and desired level of confidence in test results (Ratti and Garton 1994). With low sample size and/or high variance, power to detect significant differences among groups is reduced. Drawing conclusions based on a lack of significance in these cases can result in Type Il error, i.e., determining that there is “no effect” when there actually is one. In this study, statistical power was low due to small samples of burn units combined with often high variances in bird and vegetation data. Fortunately, many differences in bird abundance and vegetation among burn categories were so great that they were detected even with low power. These data should not be used, however, to draw conclusions about vegetation attributes or bird species that failed to show responses to fire in this study alone. Effects of fire on grasslands both within and among sub-regions of the Great Plains are extremely variable (Wright and Bailey 1982, Bragg 1995). 57 Factors such as the pre-burn vegetation community, soil conditions, climatic conditions before, during and after burning, season and frequency of burning, and grazing history all affect postfire conditions. Caution must be also used when extrapolating among geographic areas. Singh et al. (1983) describe a significant north-south division in mixed prairie. Wright and Bailey (1982) further divide mixed-grass prairie into the dry (semi-arid) mixed prairie of eastern Montana, southeastern Alberta and southwestern Saskatchewan and more mesic areas of southeastern Saskatchewan, North Dakota, and South Dakota, and describe different fire effects for each. LNWR is an especially mesic area in the mixed-grass prairie region (see Study Area description), and is not representative of all mixed-grass prairie. Results should be interpreted with this in mind, and extrapolated only to similar mesic areas. The picture is further clouded because prescribed burns at LNWR were not merely “maintenance” burns on healthy mixed-grass prair