Ornithological Monographs

ORNITHOLOGICAL MONOGRAPHS

No. 3. The Birds of Kentucky. R. M. Mengel. 1965. $25.00.

ORNITHOLOGICAL MONOGRAPHS NO. 54 • SPOTTED OWL POPULATION DYNAMICS • FRANKLIN ET AL. Ornithological No. 6. Adaptations for Locomotion and Feeding in the Anhinga and the Double-crested Cormorant. O. T. Owre. 1967. $10.00. No. 7. A Distributional Survey of the Birds of Honduras. B. L. Monroe, Jr. 1968. $25.00. No. 10. The Behavior of Spotted Antbirds. E. O. Willis. 1972. $10.00. No. 11. Behavior, Mimetic Songs and Song Dialects, and Relationships of the Parasitic Indigobirds (Vidua) of Africa. R. B. Payne. 1973. $10.00. Monographs No. 12. Intra-island Variation in the Mascarene White-eye Zosterops borbonica. F. B. Gill. 1973. $10.00. No. 13. Evolutionary Trends in the Neotropical Ovenbirds and Woodhewers. A. Feduccia. 1973. $10.00. No. 54 No. 14. A Symposium on the House Sparrow (Passer domesticus) and European Tree Sparrow (P. montanus) in North America. S. C. Kendeigh, Ed. 1973. $10.00. No. 15. Functional Anatomy and Adaptive Evolution of the Feeding Apparatus in the Hawaiian Honeycreeper Genus Loxops (Drepanididae). L. P. Richards. and W. J. Bock. 1973. $10.00. No. 16. The Red-tailed Tropicbird on Kure Atoll. R. R. Fleet. 1974. $6.00. No. 17. Comparative Behavior of the American Avocet and the Black-necked Stilt (Recurvirostridae). R. B. Hamilton. 1975. $10.00. No. 18. Breeding Biology and Behavior of the Oldsquaw (Clangula hyemalis L.). R. M. Alison. 1975. $6.00. No. 19. Bird Populations of Aspen Forests in Western North America. J. A. D. Flack. 1976. $10.00. No. 21. Social Organization and Behavior of the Acorn Woodpecker in Central Coastal California. M. H. MacRoberts and B. R. MacRoberts. 1976. $10.00. No. 22. Maintenance Behavior and Communication in the Brown Pelican. R. W. Schreiber. 1977. $6.00. No. 23. Species Relationships in the Avian Genus Aimophila. L. L. Wolf. 1977. $12.00. No. 24. Land Bird Communities of Grand Bahama Island: The Structure and Dynamics of an Avifauna. J. T. Emlen. 1977. $10.00. No. 25. Systematics of Smaller Asian Night Birds Based on Voice. J. T. Marshall. 1978. $10.00. Population Dynamics of the No. 26. Ecology and Behavior of the Prairie Warbler, Dendroica discolor. V. Nolan, Jr. 1978. $45.00. California Spott ed Owl (Strix occidentalis No. 27. Ecology and Evolution of Lek Mating Behavior in the Long-tailed Hermit Hummingbird. F. G. Stiles and L. L. Wolf. 1979. $10.00. occidentalis): A Meta-Analysis No. 28. The Foraging Behavior of Mountain Bluebirds with Emphasis on Sexual Foraging Differences. H. W. Power. 1980. $10.00. No. 29. The Molt of Scrub Jays and Blue Jays in Florida. G. T. Bancroft and G. E. Woolfenden. 1982. $10.00. Alan B. Franklin, R. J. Gutiérrez, James D. Nichols, Mark E. No. 30. Avian Incubation: Egg Temperature, Nest Humidity, and Behavioral Thermoregulation Seamans, Gary C. White, Guthrie S. Zimmerman, James E. Hines, in a Hot Environment. G. S. Grant. 1982. $10.00. Thomas E. Munton, William S. LaHaye, Jennifer A. Blakesley, No. 31. The Native Forest Birds of Guam. J. M. Jenkins. 1983. $15.00. George N. Steger, Barry R. Noon, Daniel W. H. Shaw, John J. Keane, No. 32. The Marine Ecology of Birds in the Ross Sea, Antarctica. D. G. Ainley, E. F. OȇConnor and R. F. Boekelheide. x + 97 pp. 1984. $15.00. Trent L. McDonald, and Susan Britting No. 33. Sexual Selection, Lek and Arena Behavior, and Sexual Size Dimorphism in Birds. R. B. Payne. viii + 52 pp. 1984. $15.00. No. 34. Pattern, Mechanism, and Adaptive Significance of Territoriality in Herring Gulls (Larus argentatus). J. Burger. xii + 92 pp. 1984. $12.50. PUBLISHED BY

(Continued on inside back cover) THE AMERICAN ORNITHOLOGISTS’ UNION POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL (STRIX OCCIDENTALIS OCCIDENTALIS): A META-ANALYSIS ORNITHOLOGICAL MONOGRAPHS

Editor: John Faaborg

224 Tucker Hall Division of Biological Sciences University of Missouri Columbia, MO 65211

Managing Editor: Bradley R. Plummer Proof Editors: Mark C. Penrose, Richard D. Earles

AOU Publications Oﬃ ce 622 Science Engineering Department of Biological Sciences University of Arkansas Fayett eville, Arkansas 72701

Ornithological Monographs, published by the American Ornithologists’ Union, has been established for major papers too long for inclusion in the Union’s journal, The Auk. Publication has been made possible through the generosity of the late Mrs. Carll Tucker and Marcia Brady Tucker Foundation, Inc.

Copies of Ornithological Monographs may be ordered from Buteo Books, 3130 Laurel Road, Shipman, VA 22971. Price of Ornithological Monographs 54: $10.00 ($9.00 for AOU members). Add $4.00 for handling and shipping charges in U.S., and $5.00 for all other countries. Make checks payable to Buteo Books.

Corresponding author of this issue, R. J. Gutiérrez.

The cost of this Ornithological Monograph was defrayed by the U.S. Department of Agriculture, Forest Service.

Library of Congress Control Number 2004100002

Printed by Cadmus Communications, Ephrata, PA 17522

Issued 30 January 2004

Ornithological Monographs, No. 54 vi + 54 pp.

ISBN: 0-943610-58-3

Cover: Spott ed Owl (Strix occidentalis), ink sketch by Viktor Bakhtin. POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL (STRIX OCCIDENTALIS OCCIDENTALIS): A META-ANALYSIS

By:

Alan B. Franklin,1,2 R. J. Gutiérrez,3 James D. Nichols,4 Mark E. Seamans,3 Gary C. White,2 Guthrie S. Zimmerman,3 James E. Hines,4 Thomas E. Munton,5 William S. LaHaye,3 Jennifer A. Blakesley,2 George N. Steger,5 Barry R. Noon,2 Daniel W. H. Shaw,5 John J. Keane,6 Trent L. McDonald,7 and Susan Britting8

1Colorado Cooperative Fish and Wildlife Unit, Colorado State University, Fort Collins, Colorado 80523, USA; 2Department of Fishery and Wildlife Biology, Colorado State University, Fort Collins, Colorado 80523, USA; 3Department of Fisheries, Wildlife, and Conservation Biology, University of Minnesota, St. Paul, Minnesota 55108, USA; 4U.S. Geological Survey, Patuxent Wildlife Research Center, 11510 American Holly Drive, Laurel, Maryland 20708, USA; 5U.S. Department of Agriculture, Forest Service, Paciﬁ c Southwest Research Station, 2081 East Sierra Avenue, Fresno, California 93710, USA; 6U.S. Department of Agriculture, Forest Service, Sierra Nevada Research Center, Paciﬁ c Southwest Research Station, 2121 Second Street, Suite A101, Davis, California 95616, USA; 7WEST, Inc., 2003 Central Avenue, Cheyenne, Wyoming 82001, USA; and 8P.O. Box 377, Coloma, California 95613, USA

ORNITHOLOGICAL MONOGRAPHS NO. 54 PUBLISHED BY THE AMERICAN ORNITHOLOGISTS’ UNION WASHINGTON, D.C. 2004 TABLE OF CONTENTS

ABSTRACT ------1

STUDY AREAS ------5

LASSEN STUDY AREA ------7

ELDORADO STUDY AREA ------8

SIERRA STUDY AREA ------8

SEQUOIA AND KINGS CANYON NATIONAL PARKS STUDY AREA ------9

SAN BERNARDINO STUDY AREA ------9

METHODS ------10

FIELD METHODS ------10

Surveys ------10

Estimation of reproductive effort ------11

Capture, banding, sex and age identification, and resighting of owls ------11

Pre-analysis data screening ------11

Meta-analysis workshop format ------12

DATA ANALYSIS ------12

Changes in analytical methodology from previous Spotted Owl studies ------12

Estimating adult survival ------13

Estimating fecundity ------14

Estimating rates of population change ------16

Comparison of Sierra and Sequoia and Kings Canyon national parks study areas ------19

RESULTS ------19

ADULT SURVIVAL ------19

FECUNDITY ------21

RATES OF POPULATION CHANGE ------23

Meta-analysis across study areas ------27

COMPARISON OF SIERRA AND SEQUOIA AND KINGS CANYON NATIONAL PARKS STUDY AREAS ----- 30

DISCUSSION ------30

GENERAL INFERENCES ------31

Apparent survival ------31

Fecundity ------32

Population trend ------33

STUDY-AREA-SPECIFIC INFERENCES ------35

Lassen study area ------35

Eldorado study area ------36

Sierra and Sequoia and Kings Canyon national parks study areas ------37

San Bernardino study area ------39

CONCLUSION AND RECOMMENDATIONS ------40

ACKNOWLEDGMENTS ------41

LITERATURE CITED ------42

APPENDICES ------48 From the Editor

With Ornithological Monographs #54, the American Ornithologists' Union implements a new philosophy in the production of its monograph series. Since the series began in 1964, Ornithological Monographs have been published sporadically, primarily to present articles that were too large to appear in The Auk. Some of those monographs have been enormous (over 1,000 pages), although many were in the 50–100 page range. They were sold as separate issues, with press runs of at most a few thousand copies. This and subsequent monographs will be provided to all AOU members on a regular basis, pack- aged with The Auk. There are many ornithological research eff orts that the AOU and I feel deserve to be published in one sett ing, without being separated into two or three manuscripts that appear in diff erent journals. If you have a dissertation, major research project, or even a small symposium longer than the 50 pages allowed by The Auk, we hope you will consider publishing in Ornithological Monographs. Ornithological Monographs is open to all aspects of ornithology. All that we ask is that the research involve good science, have reasonably broad ornithological interest, and can truly justify the need for monographic treatment. Financial support for publication is not a requirement, although it can certainly help the AOU and may be necessary for larger volumes. We begin the "new" Ornithological Monographs with an analysis of the demography of the California Spott ed Owl. Although the Spott ed Owl has become the focal species for both sides in arguments about forestry practices in the western United States, most of the national publicity has involved the Northern Spott ed Owl of northwestern California, Oregon, and Washington. Similar controversy now surrounds the California Spott ed Owl. An att empt to have it listed as an endangered species ended up in the courts, which forced the U.S. Fish and Wildlife Service to conduct a status review. The U.S. Fish and Wildlife Service chose not to list the owl, in part because the U.S. Forest Service had developed a management plan (the Sierra Framework) designed to protect the owl and many other resources of the Sierra Nevada. However, on the day that the U.S. Fish and Wildlife Service announced that it would not list the owl, the U.S. Forest Service announced its desire to "modify" the Sierra Framework. The modifi ed framework will be completed in early 2004, and it is unclear at this time whether this will be potentially harmful to the owl. We will undoubt- edly hear more about this situation in the future. As with most threatened or endangered species, we need solid data on demographic patt erns across the species' range. This monograph provides such data, combining modern methods such as meta-analysis with sophisticated capture–recapture models across a variety of California sites. The topic is critical for conservation purposes, and the approach will introduce readers to the state- of-the-art in the conducting demographic studies. With an interesting and important bird, 16 well- qualifi ed authors, and pioneering methods of analysis, we believe this study sets a high standard for the new Ornithological Monographs. Any scientifi c editor will admit that outside review is critical to the scientifi c publishing process. Finding reviewers for the long manuscripts that are potential Ornithological Monographs will perhaps be a challenge, but we hope readers will be as excited about the concept as we are and will be willing to volunteer time when necessary. For this monograph, Jeff rey R. Walters, Evan Cooch, and Kenneth H. Pollock of the AOU Conservation Committ ee did an exceptionally detailed review of an early draft . Katie Dugger, Jeff rey R. Walters, and a reviewer who wishes to remain anonymous made comments on what became the fi nal product. We thank these reviewers for the considerable time they contributed toward making this new monograph a high-quality, interesting, and important piece of science. We also want to thank Kimberly Smith, Brad Plummer, Mark Penrose, and Richard Earles of the AOU Publications Offi ce for helping train this new editor in the art of producing scientifi c publications.

John Faaborg Ornithological Monographs Volume (2004), No. 54, 1–54 POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL (STRIX OCCIDENTALIS OCCIDENTALIS): A META-ANALYSIS Alan B. Franklin,1,2,9 R. J. Gutiérrez,3,10 James D. Nichols,4 Mark E. Seamans,3 Gary C. White,2 Guthrie S. Zimmerman,3 James E. Hines,4 Thomas E. Munton,5 William S. LaHaye,3,11 Jennifer A. Blakesley,2 George N. Steger,5 Barry R. Noon,2 Daniel W. H. Shaw,5 John J. Keane,6 Trent L. McDonald,7 and Susan Britting8 1Colorado Cooperative Fish and Wildlife Unit, Colorado State University, Fort Collins, Colorado 80523, USA; 2Department of Fishery and Wildlife Biology, Colorado State University, Fort Collins, Colorado 80523, USA; 3Department of Fisheries, Wildlife, and Conservation Biology, University of Minnesota, St. Paul, Minnesota 55108, USA; 4U.S. Geological Survey, Patuxent Wildlife Research Center, 11510 American Holly Drive, Laurel, Maryland 20708, USA; 5U.S. Department of Agriculture, Forest Service, Paciﬁ c Southwest Research Station, 2081 East Sierra Avenue, Fresno, California 93710, USA; 6U.S. Department of Agriculture, Forest Service, Sierra Nevada Research Center, Paciﬁ c Southwest Research Station, 2121 Second Street, Suite A101, Davis, California 95616, USA; 7WEST, Inc., 2003 Central Avenue, Cheyenne, Wyoming 82001, USA; and 8P.O. Box 377, Coloma, California 95613, USA

Abstract.—We conducted a meta-analysis to provide a current assessment of the population characteristics of California Spott ed Owls (Strix occidentalis occidentalis) resident on four study areas in the Sierra Nevada and one study area in southern California. Our meta-analysis followed rigorous a priori analysis protocols, which we derived through extensive discussion during a week-long analysis workshop. Because there is great interest in the owl’s population status, we used state-of-the-art analytical methods to obtain results as precise as possible. Our meta-analysis included data from fi ve California study areas located on the Lassen National Forest (1990–2000), Eldorado National Forest (1986–2000), Sierra National Forest (1990–2000), Sequoia and Kings Canyon national parks (1990–2000), and San Bernardino National Forest (1987–1998). Four of the fi ve study areas spanned the length of the Sierra Nevada, whereas the fi ft h study area encompassed the San Bernardino Mountains in southern California. Study areas ranged in size from 343 km2 (Sequoia and Kings Canyon) to 2,200 km2 (Lassen). All studies were designed to use capture–recapture methods and analysis. We used survival in a meta-analysis because fi eld methods were very similar among studies. However, we did not use reproduction in a meta-analysis because it was not clear if variation among individual study-area protocols used to assess reproductive output of owls would confound results. Thus, we analyzed fecundity only by individual study area. We examined population

trend using the reparameterized Jolly-Seber capture–recapture estimator (λt). We did not estimate juvenile survival rates because of estimation problems and potential bias because of juvenile emigration from study areas. We used mark–recapture estimators under an information theoretic framework to assess apparent survival rates of adult owls. The pooled estimate for adult apparent survival for the fi ve study areas was 0.833, which was lower than pooled adult survival rates (0.850) from 15 Northern Spott ed Owl (S. o. caurina) studies. Estimates of survival from the best model on the Lassen ( = 0.829, 95% confi dence intervals [CI] = 0.798 to 0.857), Eldorado ( = 0.815, 95% CI = 0.772 to 0.851), Sierra ( = 0.818, 95% CI = 0.781 to 0.850), and San Bernardino ( = 0.813, 95% CI = 0.782 to 0.841) were not diff erent. However, the Sequoia and Kings Canyon population had a higher survival rate ( = 0.877, 95% CI = 0.842 to 0.905) than the other study areas. Management history and forest structure (e.g. presence of giant sequoia [Sequoiadendron giganteum]) on the Sequoia and Kings Canyon study area diff ered from all other study areas. There appears to be litt le or no evidence for temporal variation in adult apparent survival on any of the study areas. Although we did not directly compare fecundity, estimates were highly variable among years within all study areas (CV of temporal process variation = 0.672–0.817). Estimates for fecundity

9E-mail: [email protected] 10E-mail: [email protected] 11Present address: P.O. Box 523, Big Bear City, California 92314, USA. 1 2 ORNITHOLOGICAL MONOGRAPHS NO. 54

among the study populations were Lassen ( = 0.336, SE = 0.083), Eldorado ( = 0.409, SE = 0.087), Sierra ( = 0.284, SE = 0.073), Sequoia and Kings Canyon ( = 0.289, SE = 0.074), and San Bernardino ( = 0.362, SE = 0.038). During most years, the Sierra Nevada populations showed either moderate or poor fecundity. However, 1992 appeared to be an exceptional reproductive year for owls in the Sierra Nevada. In contrast, the San Bernardino population had less variable reproduction (CV of temporal process variation = 0.217), but experienced neither the exceptional reproduction of 1992 nor the extremely poor years that characterized all of the Sierra Nevada study areas. Because fecundity may be infl uenced by weather patt erns, it was possible that the diff erent weather patt erns between southern California and the Sierra Nevada accounted for that diff erence.

Except for Eldorado, all estimates for λt were <1.0, but none was diﬀ erent from λ = 1.0 given the 95% conﬁ dence intervals (Lassen [ = 0.985, SE = 0.026]; Eldorado [ = 1.042, SE = 0.047]; Sierra [ = 0.961, SE = 0.024]; Sequoia and Kings Canyon [ = 0.984, SE = 0.047]; San Bernardino [ = 0.978, SE = 0.025]). However, additional evidence (in the form of realized population change based on ) strongly suggested that the Sierra population declined during the study

period. Estimated trends in λt for the Eldorado and Sierra study areas were negative. Thus, we could not distinguish deﬁ nitively between alternatives that the populations were stationary or

that the estimates of λt were not suﬃ ciently precise to detect declines on four of the study areas (Eldorado, Lassen, San Bernardino, and Sequoia and Kings Canyon). Results of the trend analyses do not allow strong inference about the decline of the popula-

tions. Because λt reﬂ ects changes in owl numbers on the study areas (i.e. it integrates emigration, immigration, birth, and death rates), it does not allow inference about the larger populations in which those local populations are imbedded. That is, it is possible that local populations could be producing fewer young but are enhanced by immigration from surrounding areas. It also is possible that the conditions within the study areas may have been bett er, in terms of habitat loss, than surrounding areas because a Spott ed Owl conservation strategy was imposed in the national forest study areas. That may be particularly true where high amounts of private land surrounded study areas (e.g. Lassen, Eldorado). The relatively low survival rates coupled with trend estimates that were either declining or <1.0 suggest a cautious approach to developing conservation strategies for the California Spott ed Owl until further analyses can be conducted that couple climatic and habitat conditions with population parameters, such as adult survival and fecundity.

Resumen.—Realizamos un meta-análisis para presentar una evaluación actualizada de las características de las poblaciones de la lechuza Strix occidentalis occidentalis residentes en cuatro áreas de estudio ubicadas en la Sierra Nevada y en un área de estudio ubicada en el sur de California. Nuestro meta-análisis siguió unos protocolos a priori rigurosos, los cuales desarrolla- mos a través de discusiones extensivas durante un taller de análisis de una semana de duración. Debido a que existe gran interés en el estatus poblacional de esta lechuza, nos esforzamos para utilizar métodos analíticos modernos que arrojarían resultados cientíﬁ camente defendibles. Nuestro meta-análisis incluyó datos de cinco sitios de estudio de California, localizados en el Bosque Nacional Lassen (1990–2000), el Bosque Nacional Eldorado (1986–2000), el Bosque Nacional Sierra (1990–2000), los parques nacionales Sequoia y Kings Canyon (1990–2000), y el Bosque Nacional San Bernardino (1987–1998). Cuatro de estos sitios se extendieron a lo largo de la Sierra Nevada, mientras que el quinto comprendió las montañas de San Bernardino, en el sur de California. Las áreas de estudio variaron en extensión entre 343 km2 (Sequoia y Kings Canyon) y 2,200 km2 (Lassen). Todos los estudios fueron diseñados para usar métodos y análisis de marcaje y recaptura. Utilizamos datos de supervivencia en un meta-análisis porque los métodos de Weld fueron muy similares entre estudios. Sin embargo, no utilizamos la reproducción en un meta-análisis porque no fue claro si la variación entre los protocolos empleados para determinar el rendimiento reproductivo en las diferentes áreas podría afectar los resultados. Por lo tanto, analizamos la fecundidad de cada área de estudio separadamente. Examinamos las tendencias

de la población usando el estimador re-parametrizado de captura–recaptura de Jolly-Seber (λt). No estimamos las tasas de supervivencia de los juveniles debido a problemas de estimación y a sesgos potencialmente causados por la emigración de los juveniles desde las áreas de estudio, pero empleamos estimadores de marcaje-recaptura bajo un marco teórico de información para determinar las tasas aparentes de supervivencia de las lechuzas adultas. El valor estimado de la supervivencia aparente combinado para las cinco áreas fue de 0.833, lo que es menor que 0.850, el valor estimado de supervivencia de adultos combinado para 15 estudios sobre POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 3

la subespecie del norte (S. o. caurina). Las estimaciones de los mejores modelos para la supervivencia en Lassen ( = 0.829, intervalo de confi anza del 95% [CI] = 0.798 a 0.857), Eldorado ( = 0.815, CI = 0.772 a 0.851), Sierra ( = 0.818, CI = 0.781 a 0.850) y San Bernardino ( = 0.813, CI = 0.782 a 0.841) no fueron diferentes. Sin embargo, la población de Sequoia y Kings Canyon tuvo una tasa de supervivencia mayor ( = 0.877, CI = 0.842 a 0.905) que las de las otras áreas. La his- toria de manejo y la estructura del bosque (e.g. presencia de sequoias gigantes [Sequoiadendron giganteum]) en Sequioa y Kings Canyon fueron diferentes de las de los otros sitios. En ninguna de las áreas de estudio parece existir evidencia que sugiera la existencia de variación temporal en la supervivencia aparente de los adultos. Aunque no comparamos la fecundidad directamente, los valores estimados fueron alta- mente variables entre años en cada uno de los sitios de estudio (coefi ciente de variación [CV] del proceso temporal = 0.672–0.817). Los valores estimados de la fecundidad en las distintas poblaciones fueron: Lassen ( = 0.336, EE = 0.083), Eldorado ( = 0.409, EE = 0.087), Sierra ( = 0.284, EE = 0.073), Sequoia y Kings Canyon ( = 0.289, EE = 0.074) y San Bernardino ( = 0.362, EE = 0.038). Durante la mayor parte de los años, las poblaciones de la Sierra Nevada exhibieron fecundidades bajas o moderadas, mientras que 1992 pareció ser un año excepcional para la reproducción de las lechuzas en esta región. En contraste, la reproducción en la población de San Bernardino fue menos variable (CV del proceso temporal = 0.217), y no presentó la excepcional reproducción de 1992, ni los años extremadamente malos que caracterizaron a todas las áreas ubicadas en la Sierra Nevada. Debido a que la fecundidad puede ser infl uenciada por los patrones climáticos, es posible que las diferencias encontradas puedan explicarse por las diferentes condiciones de clima del sur de California y la Sierra Nevada.

A excepción de Eldorado, todos los valores estimados de λt fueron menores a 1.0, pero ninguno fue diﬀ erente de λ = 1.0 dados los intervalos de conﬁ anza del 95% (Lassen [ = 0.985, EE = 0.026], Eldorado [ = 1.042, EE = 0.047], Sierra [ = 0.961, EE = 0.024], Sequoia y Kings Canyon [ = 0.984, EE = 0.047], San Bernardino [ = 0.978, EE = 0.025]). Sin embargo, evidencia adicional en forma del cambio poblacional real basado en sugirió fuertemente que

la población de Sierra declinó durante el período de estudio. Las tendencias estimadas de λt para las áreas de estudio de Eldorado y Sierra fueron negativas. Por lo tanto, no pudimos saber

de forma deﬁ nitiva si las poblaciones están estables, o si los valores estimados de λt no fueron lo suﬁ cientemente precisos como para detectar disminuciones en cuatro de las áreas de estudio (Eldorado, Lassen, San Bernardino, y Sequoia y Kings Canyon). Los resultados de los análisis de las tendencias no permiten hacer inferencias fuertes sobre

el declive de las poblaciones. Debido a que λt reﬂ eja los cambios en el número de lechuzas en las áreas de estudio (i.e. integra las tasas de inmigración, emigración, natalidad y mortalidad), este parámetro no permite hacer inferencias sobre las poblaciones más grandes en las que están inmersas las poblaciones locales. Esto quiere decir que es posible que las poblaciones locales estén produciendo muy pocas crías, pero que estén siendo suplementadas mediante inmigración desde las áreas circundantes. También es posible que las condiciones dentro de las áreas de estudio hayan sido mejores en términos de pérdida de hábitat que las de las áreas circundantes, debido a que en los bosques nacionales de estudio se implementó una estrategia de conservación para S. occidentalis. Esto podría ser particularmente cierto en sitios rodeados por gran cantidad de tierras privadas (e.g. Lassen, Eldorado). Las tasas de supervivencia relativamente bajas, en combinación con las estimaciones de las tendencias que estuvieron en declive o fueron menores a 1.0, sugieren que se debe actuar con cautela al desarrollar estrategias de conservación para S. o. occidentalis hasta que se realicen análisis que acoplen condiciones climáticas y de hábitat con parámetros poblacionales, como la supervivencia y fecundidad de los adultos.

The California Spotted Owl (Strix occiden- the other two subspecies, the Northern (S. o. talis occidentalis) is one of three Spott ed Owl caurina) and Mexican (S. o. lucida) spott ed owls subspecies. It occurs as a contiguous popula- (Barrowclough et al. 1999). Unlike the Northern tion in the Sierra Nevada of California and as and Mexican subspecies, the California Spott ed insular populations in central coastal California, Owl has not been listed as a threatened spe- southern California, and Baja California Norte, cies under the U.S. Endangered Species Act. Mexico (Gutiérrez et al. 1995). The California Nevertheless, the status, trends, and basic Spott ed Owl is genetically diﬀ erentiated from natural history (e.g. habitat selection) of the 4 ORNITHOLOGICAL MONOGRAPHS NO. 54

California Spott ed Owl have been the center issues across the Sierra Nevada (USDA Forest of controversy for more than a decade (Verner Service 1998a). An accompanying document that et al. 1992b; U.S. Department of Agriculture summarized current management direction spe- [USDA], Forest Service 1998a, b). cifi c to each of the high-priority issues was pre- Verner et al. (1992a [California Spott ed Owl pared by the USFS (USDA Forest Service 1998b). Report, hereaft er CASPO]) evaluated the sta- Information provided in those documents tus and trends of, and the state of ecological determined the scope and focus of subsequent knowledge about, the California Spott ed Owl. land-management planning that was conducted Two fundamental fi ndings of CASPO were un- as part of the Sierra Nevada Framework Project. certainty in population trends of the owl because Those eff orts culminated in a fi nal EIS and of the short duration of extant owl-demographic Record of Decision (ROD) that identifi ed new studies, and the probable decline throughout the management direction for California Spott ed Sierra Nevada of forest att ributes (e.g. very large- Owls and the other high-priority issues on USFS diameter trees) associated with Spott ed Owls. lands across the Sierra Nevada (USDA Forest The CASPO also recommended a set of interim Service 2001a, b). The Sierra Nevada Framework guidelines to the USDA Forest Service (USFS) for Project team formally requested that a Spott ed the management of Spott ed Owl habitat. On the Owl population meta-analysis be conducted us- basis of the strategy recommended by CASPO, ing the information on owl population dynamics the Forest Service implemented new guidelines gathered before and aft er the CASPO to address (USDA Forest Service 1993), with the intention the ongoing controversy regarding the con- of moving beyond the interim guidelines when servation status of California Spott ed Owls. In credible scientifi c information was gathered particular, they requested that the meta-analysis from fi eld studies to justify a change from the examine trends in California Spott ed Owl popu- interim guidelines. The USFS then embarked lations using data from fi ve existing studies that on a series of Environmental Impact Statements had collected data on demographic characteris- (EIS) (USDA Forest Service 1995, 1996), the last tics of California Spott ed Owls. of which was unfavorably reviewed (Federal Meta-analysis has been employed as an ana- Advisory Committ ee 1997). The Federal lytical tool to evaluate the status and trends of Advisory Committ ee (1997) concluded, among Northern Spott ed Owls since 1993 (Burnham other things, that there was litt le new informa- et al. 1996, Franklin et al. 1999). Meta-analyses tion on the owl’s biology that could justify the allow synthesis of data from independent stud- changes proposed in the EIS documents, citing ies where studies are considered the sampling in particular the uncertainty regarding potential units (Wolf 1986, Hunter and Schmidt 1990). impact of timber harvest strategies on owl habi- They can be performed on statistics collected tat and owl population dynamics. Concomitant from published peer-reviewed papers (e.g. to the EIS eff orts, the Sierra Nevada Ecosystem Vanderwerf 1992) or from the raw data them- Project (1996) was established, which att empted selves (Franklin and Shenk 1995). The power to identify multiple concerns relative to conser- of a meta-analysis for California Spott ed Owls vation and management of the Sierra Nevada is the ability to combine information from sev- ecosystem. The report of the Sierra Nevada eral studies to achieve greater sample size and Ecosystem Project did not explicitly discuss the perhaps investigate sources of variation and ecology or management of the California Spott ed examine potential correlations in population Owl, but it correctly framed issues facing the dynamics, otherwise unavailable from a single future of the Sierra Nevada as ecosystem-wide. study. For example, Burnham et al. (1996) used Thus, the USFS abandoned earlier EIS eff orts and raw data gathered from 13 Northern Spott ed initiated a new strategy to evaluate not only the Owl population studies in the Pacifi c Northwest California Spott ed Owl but also other sensitive to demonstrate an accelerating decline in female wildlife species and habitat conditions while con- survival over the period of study, which sup- sidering such issues as the threat of wildfi re and ported the inference of a population decline in timber harvest practices. In 1998, the USFS Pacifi c Northern Spott ed Owls. Southwest Research Station published a scien- Coincident with the meta-analysis, a Pacifi c tifi c review identifying and synthesizing current Northwest Forest Plan was developed for the knowledge on the highest-priority conservation Northern Spott ed Owl and other old-forest POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 5 species (Thomas et al. 1993) at the request of al. (1999). That rigorous process was critical to (U.S.) President W. J. Clinton. That plan pro- the success of the coordinated analysis and was posed a conservation reserve design to protect greatly facilitated by the presence and interac- Spott ed Owl habitat, which was based on earlier tion of that diverse group. conservation strategies for the Northern Spott ed Our intent in analyzing the data was to exam- Owl (Thomas et al. 1990, U.S. Department of the ine trends in demographic parameters as well as Interior [USDI] 1992). That reserve design ef- rates of population change (λ) because changes fectively removed large areas from timber har- in demographic parameters (the integral com- vest consideration, much of which was within ponents of λ) can provide bett er understanding existing Spott ed Owl demographic study areas. of population dynamics. For example, Franklin Essentially, the reserve design provided an ad et al. (2000) suggested that adult survival de- hoc test of the eff ect of habitat loss on Spott ed fi ned the magnitude of λ whereas reproduction Owl population trends. In 1998, a second and recruitment determined variation in λ over meta-analysis on the Northern Spott ed Owl time. Thus, we did not want to rely solely on population data was conducted (Franklin et a single measure of population trend. We did al. 1999). In those analyses, declining trends in not explicitly evaluate changes in population owl populations and adult female survival were numbers. Although data were available to es- either reduced or stabilized. Further, in the 1998 timate numbers of owls on several of the study meta-analysis, a new analytical method was areas, there were potential biases (see review introduced: direct estimation of λ (the annual in Pollock et al. 1990) in estimating numbers fi nite rate of population change, referred to here of owls using capture–recapture. Rather, we as λt) that was based on capture–recapture data relied on the reverse-time Jolly-Seber estima- and refl ected changes in numbers of territorial tor (herein referred to as λt) to estimate an- owls on study areas. That analysis (discussed nual changes in numbers of owls on the study below) avoided the eff ect of potentially biased areas (Pradel 1996). We were then able to re- estimates of juvenile survival that caused un- express those estimates as realized proportional certainty in estimating λ with the Leslie pro- changes in numbers of owls without having to jection matrix used in previous Spott ed Owl rely on estimation of population abundance. studies. However, a stationary population (i.e. λ = 1.0) using that newer analysis still could Study Areas not demonstrate demographic stability because stationary populations could be maintained Demographic data from fi ve study areas were solely by immigration from other populations. used in the analyses (Table 1, Fig. 1). Specifi c at- Nevertheless, it was a clear att empt to incor- tributes of each study area are described in the porate the most modern population analytical following sections, presented in latitudinal or- methods in the meta-analysis. der from north to south. Most of the study areas The purpose of our paper is to present the were considered, or included, density study results of a meta-analysis conducted using data areas, which were geographically defi ned areas generated from fi ve California Spott ed Owl that were surveyed entirely for Spott ed Owls. population studies (four within the contiguous The Sierra Nevada was the dominant physi- owl range of the Sierra Nevada and one from cal feature infl uencing the climate on four of the an insular population in southern California) study areas. That mountain range had cold, wet to assess status and trends of some California winters and hot, dry summers. Winter Pacifi c Spott ed Owl populations. In our study, owl storm systems were the main source of precipi- researchers from the fi ve demography studies, tation for the range. Those storm systems could timber industry consultants, and stakeholders be either cold or warm depending upon their met with experts in analysis of population dy- origin (e.g. Gulf of Alaska or tropical Pacifi c, re- namics from 9 to 13 July 2001 at Colorado State spectively). Sierra Nevada vegetation was heav- University in Fort Collins, Colorado to conduct ily infl uenced by climate, elevation, aspect, and a formal meta-analysis of all known California edaphic conditions (see below), which resulted Spott ed Owl population data. Participants in diverse forest types. agreed to adhere to a rigid and formal protocol Fire has been a primary force shaping the for analytical sessions proposed by Anderson et distribution and structure of vegetation in the 6 ORNITHOLOGICAL MONOGRAPHS NO. 54

Fig. 1. Relative locations of California Spotted Owl demography studies in relation to forested habitat (shaded gray) throughout California. Cross-hatched area within the Lassen inset was used to estimate λt. Dark-colored circles within the Eldorado are owl sites external to the density study area (dark shaded area within inset). POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 7 =

& Sierra Nevada because natural fi re regimes were characterized by relatively frequent fi re return intervals (Skinner and Chang 1996). Fire return intervals and fi re behavior have changed as a result of governmental fi re-suppression policies and other vegetation-management activities that followed European sett lement (McKelvey et al. 1996). Sierran vegetation also has been aff ected by logging and livestock grazing (McKelvey and Johnston 1992). Logging began in the mid- to late 1800s and harvest techniques varied from clear felling to individual tree selection. Livestock grazing was in- tense during the 1800s but has been reduced to relatively low levels today. The net eff ect of both natural and anthropogenic disturbance has led to a complex mosaic of vegetation types, seral stages, and stand structures. This has resulted in a variety of forest types used by Spott ed Owls in the Sierra Nevada. Each of the Sierra Nevada ).*8=<&8=43=9*1*2*97>==47=).++*7*3(*8=.3=57494(418=&3)=8:7;*>=*++479=<.9-=1&99*7= demography study areas has slightly diff erent histories of land use and vegetation histories, which may have infl uenced the population dynamics of Spott ed Owls (see below). In contrast to the Sierra Nevada study areas, the San Bernardino Mountains study area was located in a relatively isolated mountain range in southern California (Fig. 1). Owls oc- cupying that mountain range were the largest population of a presumed owl metapopulation found throughout the disjunct ranges of the region (Noon and McKelvey 1992, LaHaye = (43):(9.3,=89:)>= 3:2'*7=4+=>*&78= ,7&5-.(=&3&1>8*8

, et al. 1994). The climatic environment in southern California was more benign than the Sierra Nevada because the majority of winter storms pass to the north of the region (Karhl 1979). Logging occurred in the San Bernardino Mountains from the late 1800s through the mid-1980s (Robinson 1989, McKelvey and Johnston 1992). Commercial logging occurred infrequently (McKelvey and Johnston 1992). The historic fi re regime in the San Bernardino Mountains included frequent low-intensity fi res that played a major role in shaping vegetation mosaics (Minnich 1988). However, modern fi re suppression and historic logging have resulted in signifi cant changes in vegetation structure and composition in the wett er portions of the mountain range (Minnich et al. 1995). Mining,

=+_=-&7&(9*7.89.(8=4+=+.;*=&1.+473.&=5499*)=<1=)*24,7&5-.(=89:)>=&7*&8=.3=&1.+473.&_= urban expansion, and numerous other human

&71>=>*&78=+742=842*=89:)>=&7*&8=<*7*=349=.3(1:)*)=.3=8:7;.;&1`=+*(:3).9>`=&3)=545:1&9.43=,74<9-=&3&1>8*8='*(&:8*=+4(:8=4+=89: activities also have impacted owl habitat to &= = =9:)>=&7*&=$1)47&)4=%&9.43&1=&47*89='&88*3=%&9.43&1=&47*89=&3=3*73&7).34=%&9.43&1=&47*89=.*77&=%&9.43&1=&47*89=*6:4.&=&3)=9.3,8=&3>43=3&9.43&1=5&708= 9= 3= $'(= -.-= ' = ,`+.*= (743>2= 3,/= ,__=(*5&792*39=4+= ,7.(:19:7*`=&47*89=*7;.(*= 6$= ,`,**= 02 ,3.;*78.9>=4+=-.33*849&= +33*:,***=++= ,3.;*78.9>=4+=-.33*849&= 4147&)4=9&9*=,3.;*78.9>= = 03-= +33+:,***= ,__=(*5&792*39=4+= ,7.(:19:7*`=&47*89=*7;.(*= .?*= +33*:,***=++= 7,&3.?&9.43= +321:+332=+,= +320:,***=+/= +33*:,***=++= +33*:,***= +33+:+332= +33*:,***= +33*:,***= 9:)>=5*7.4)== *&78=:8*)=.3=)*248= >*&78= `= !_= some degree (Verner et al. 1992b). 8 ORNITHOLOGICAL MONOGRAPHS NO. 54

Lassen Study Area private land, accounting for 63% and 37% of the EDSA, respectively. The Lassen study area (LAS) was located We began demographic research on the EDSA in northeastern California, primarily on the in 1986. We initiated the RSA in 1997 to encom- Lassen National Forest (LNF). The greater pass more owl territories and to locate owls that study area encompassed 2,200 km2 and was may have emigrated from the EDSA. All EDSA analogous to the Regional Study Area of the owl territories were on the Eldorado National Eldorado study (see below). A subset of LAS Forest. Thirty-eight percent of the RSA terri- 2 (∼1,270 km ) was selected for estimation of λt tories were located on the Eldorado National (see below) during the meta-analysis, based on Forest, 38% on the Tahoe National Forest, and portions of the study area surveyed consistently 24% on the Tahoe Basin Management Area. during 1992–2000. Most private land within the All RSA territories were on public land. We study area boundaries was not surveyed, al- did not use owls banded in the Tahoe Basin though several owl sites on private timber land Management Area for this analysis. adjacent to LNF were included. In addition, a The study area was typical of the mid- few sites overlapped Lassen Volcanic National elevation Sierra Nevada with mountainous ter- Park, the Plumas National Forest, and Bureau of rain bisected by steep river drainages. Elevations Land Management land. ranged from 366 to 2,401 m. From 1962 to 1995, Elevations on the study area ranged from average annual precipitation at the Blodgett 1,200 to 2,100 m. Annual precipitation at 1,250– Experimental Forest (part of the RSA; 1,340 m 1,500 m averaged 141 cm in the west, 86 cm elevation) was 158 cm (Olson and Helms 1996). in the center, and 36 cm just east of the study Thirty-fi ve percent of precipitation fell as snow, area. Most of the precipitation fell as snow averaging 254 cm year–1. Average minimum from November through April. Average high temperature in January was 1°C and average temperatures at the center of the study area maximum temperature in July was 28°C. (1,380 m) ranged from 6°C in January to 29°C The EDSA and RSA were typical of Sierran in July. Average low temperatures ranged from Montane Forest (SMF; Küchler 1977). From 600 –7°C in January to 7°C in July. to 1,500 m, the SMF was dominated by pondero- Majority of forest types on the study area were sa pine on more xeric sites and white fi r on more mixed conifer, with additional stands classifi ed mesic sites. A transition zone above 1,500 m was as true fi r. Dominant tree species included white dominated by red fi r (Rundel et al. 1977). Other fi r (Abies concolor), sugar pine (Pinus lambertiana), common tree species that occurred within the ponderosa pine (P. ponderosa), incense cedar study area included sugar pine, Douglas-fi r, (Calocedrus decurrens), Douglas-fi r (Pseudotsuga incense cedar, canyon live oak (Q. chrysolepis), menziesii), red fi r (A. magnifi ca), and Jeff rey pine California black oak, Pacifi c dogwood (Cornus (P. jeff reyi). California black oak (Quercus kelloggii) nutt allii), and tan oak (Lithocarpus densifl orus). was present in the understory in some stands. Sierra Study Area Eldorado Study Area The Sierra study area (SIE) was located The Eldorado study area (ELD) consisted of ∼83 km east of Fresno, California in the south- a 355 km2 Eldorado Density Study Area (EDSA) ern Sierra Nevada within the watersheds of and a 570 km2 Regional Study Area (RSA) lo- the San Joaquin River and the North Fork of cated between Georgetown and Lake Tahoe in the Kings River. Study was initiated in 1990 on the central Sierra Nevada, El Dorado and Placer 419 km2 and then expanded in 1994 to 693 km2. counties, California. Boundaries for the EDSA The Sierra National Forest administered 92% of were defi ned by the Rubicon River, South Fork the lands within the study area. of the Rubicon River, and Middle Fork of the The SIE was mountainous with steep drain- American River to the south, north, and west; ages and elevations ranged from 304 m at the Chipmunk Ridge and Bunker Hill to the north southwestern corner to 2,924 m on the eastern and east; and Forest Road 33 to the east. The edge. Boundaries of the SIE were defi ned by ELD was characterized by a “checkerboard” USFS administrative units and major topo- distribution of alternating public (USFS) and graphic features such as ridges and drainages. POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 9

Annual precipitation from 1961 to 1990 aver- southern edge of SKC and 105 cm at Grant Grove aged 94 cm at Huntington Lake, ∼16 km north of (2,013 m) near the northern border of the study the study area at 2,139 m in elevation (National area (NOAA 1999). During winter, precipitation Oceanic and Atmospheric Administration fell primarily as snow above 1,220 m. Average [NOAA] 1998). Most precipitation occurred daily temperatures for July at Ash Mountain and during the winter and fell mainly as snow above Grant Grove were 28 and 17°C, respectively. 1,220 m. Summer temperatures averaged ∼16°C Several vegetation types (Verner and Boss at Huntington Lake (NOAA 1998) but could be 1980) were present on the study area in three >38°C at lower elevations. distinct zones. Low-elevation oak woodlands The SIE had three general vegetation types: (24% of SKC below 1,220 m elevation) included oak woodlands, mid-elevation mixed conifer low-elevation pine–oak woodlands, blue oak forests, and high-elevation conifer forests. savannas, and dense riparian deciduous for- Oak-woodland zone, at the lowest elevation ests. Tree species included blue oak, gray pine, (304–1,220 m), encompassed 26% of the study interior live oak, canyon live oak, California area and was dominated by blue oak (Q. doug- sycamore (Platanus racemosa), California buck- lasii), interior live oak (Q. wislizenii), canyon eye (Aesculus californica), and Fremont cott on- live oak, and gray pine (P. sabiniana). Various wood (Populus fremontii). Large areas adjacent foothill chaparral species were abundant. Mid- to low-elevation oak woodland consisted of elevation mixed conifer forest (1,220–2,438 m) chaparral (primarily chamise [Adenostoma fas- occupied 61% of the study area and was domi- ciculatum]). Mid-elevation conifer forests (67% nated by ponderosa pine, white fi r, incense of SKC; 1,220–2,440 m elevation) included a cedar, California black oak, Jeff rey pine, red fi r, ponderosa pine type at lower elevations, a mid- and sugar pine. A small (2 km2) grove of giant elevation riparian deciduous type that occurred sequoia (Sequoiadendron giganteum) was within throughout the zone, and a mixed conifer type that forest. High-elevation conifer forest (2,439– that was dominant in that zone. Tree species in- 2,924 m) covered 13% of the study area and was cluded ponderosa pine, Jeff rey pine, sugar pine, dominated by red fi r, lodgepole pine (P. contor- white fi r, red fi r, incense cedar, and California ta), and western white pine (P. monticola). black oak. Mixed conifer forests included 10 giant sequoia groves. Based on areas estimated Sequoia and Kings Canyon National Parks from Parsons (1994), those sequoia groves Study Area covered 7% of the study area. Sequoia groves were mixed conifer forests that contained gi- The Sequoia and Kings Canyon national ant sequoia trees and other conifer species (e.g. parks study area (SKC) was 35 km northeast white fi r and sugar pine), which were oft en of Visalia and 19 km southeast of the SNF and more numerous (Rundel 1971). High-elevation covered 343 km2 in Fresno and Tulare counties, coniferous forests (9% of the SKC; above 2,440 California. The SKC was managed primarily m elevation) consisted primarily of a red fi r type by Sequoia and Kings Canyon national parks, and a lodgepole pine type. Trees included red but the study area also included the adjacent fi r, lodgepole pine, and western white pine. Whitaker Forest (1.3 km2) that was managed by the University of California. Most of the study San Bernardino Study Area area was part of the Kaweah River watershed (primarily the North, Marble, and Middle The San Bernardino study area (SAB) was forks); however, the northern 14 km2 was part located in the San Bernardino Mountains, of the Kings River watershed. The terrain was ∼140 km east of Los Angeles, California (Fig. 1). mountainous with steep drainages; elevations In 1987, the 535 km2 Big Bear Study Area (BBSA) ranged from 427 to 3,050 m. Boundaries of the was established, centered on the majority of the study area were defi ned by U.S. Park Service Spott ed Owl locations known at that time. In administrative boundaries and topography 1989, the size of the study area was expanded (ridges and elevation). Demographic studies to encompass all forest habitat within the entire were initiated on SKC in 1990. mountain range (2,140 km2). The San Bernardino From 1961 to 1990, annual precipitation av- Mountains were one of a series of mountain eraged 66 cm on Ash Mountain (521 m) at the ranges that rise above extensive desert (Vasek 10 ORNITHOLOGICAL MONOGRAPHS NO. 54 and Barbour 1988) and semidesert (Mooney present an explicit description of departures from the 1988) vegetation types in southern California standard techniques (see also Appendix 1). (Noon and McKelvey 1992, LaHaye et al. 1994). Surveys.—Spott ed Owls were surveyed accord- Forests in those mountain ranges were isolates ing to the methods of Forsman (1983) and Franklin et al. (1996a). Surveys were performed by imitating because they occurred at elevations higher than either Spott ed Owl territorial vocalizations or playing surrounding desert and chaparral vegetation prerecorded calls from a tape player. We employed (Noon and McKelvey 1992, LaHaye et al. 1994). three types of survey: point, cruise, and walk-in sur- They occupied ∼2% of the southern California veys. We conducted those surveys from 1 April to 31 landscape (Scott et al. 1993). August, 1986–2000, except in the SIE and SKC study Elevations on the study area ranged from 800 areas where surveys were initiated on 1 March and to 3,500 m. Climate was Mediterranean with ended on 30 September and the SAB where surveys most precipitation falling during the winter ended on 30 September. Night surveys consisted of months (Fujioka et al. 1998). Annual precipita- calling at points (locations) for 10 min (15 min on tion ranged from 50 to 100 cm depending on SAB) to determine if Spott ed Owls were present in a new or historically used area. At each survey point, re- location, elevation, and topography (Minnich searchers imitated Spott ed Owl vocalizations to elicit 1988). Those mountain ranges were mesic a response. We strategically placed survey points to compared to the surrounding lowlands, which obtain complete survey coverage of individual study allowed them to support a diverse assemblage areas. Night surveys were generally conducted from of shrub and forest vegetation types (Minnich dusk to 2400 h. If an owl was detected during a night 1998). Vegetation types most commonly used survey, we conducted a walk-in survey in the same by Spott ed Owls in southern California were general location to fi nd its roosts and nest (if nesting), mixed evergreen (Sawyer et al. 1988) and mon- locate a possible mate, assess reproductive status, and tane forests (Thorne 1988). Mixed evergreen identify individuals. We att empted to resight all owls forests occurred below ∼1,600 m elevation located during each walk-in and capture and band all unbanded owls (see below). We conducted additional and the dominant tree species were canyon walk-in surveys to capture and mark unbanded owls live oak and big-cone Douglas-fi r (Pseudotsuga if those birds were not captured in the initial survey. macrocarpa). Other tree species associated with We located owls during walk-in surveys by imitating those lower elevation sites included Coulter Spott ed Owl vocalizations to elicit a response and by pine (Pinus coulteri), white alder (Alnus rhom- visually searching the area where the owl was de- bifolia), California sycamore, and big-leaf maple tected during the previous night survey. If no Spott ed (Acer macrophyllum). Montane forests occurred Owls were detected during a walk-in survey, it was above 1,600 m elevation and were dominated by termed a cruise survey. Walk-in or cruise surveys Jeff rey pine and white fi r. Other tree species oc- lasted until the objectives were met (i.e. reproductive curring in montane forests included sugar pine, status and identity of individuals were determined) or the observers deemed that further eff ort would incense cedar, California black oak, ponderosa not help accomplish the objectives (e.g. an owl could pine, western juniper (Juniperus occidentalis), not be located within the fi rst few hours of survey). pinyon pine (P. monophylla), and limber pine (P. Therefore, survey eff ort was a function of the actual fl exilis). time allott ed to surveys of various types. We performed multiple complete surveys of each Methods entire study area during the course of a fi eld season. In a study of Northern Spott ed Owls, Reid et al. (1999) Field Methods detected all eight radiomarked nonjuvenile males in their study within three 10-min vocal surveys, which Methods among Spott ed Owl population dynam- were spaced one week apart. All studies reported ics studies have been similar for some time; that is herein conducted ≥3 surveys at multiple point loca- particularly true of Northern Spott ed Owl studies tions within owl territories. In addition, if an owl was (Forsman 1983, Franklin et al. 1996a). However, there detected at night, a walk-in survey was conducted to is some variation among studies of California Spott ed locate any mates. Thus, survey eff ort should have been Owls because of local diff erences in owl behavior, dif- suffi cient to detect nearly all territorial Spott ed Owls ferent environmental conditions, and diff erent initial on the study areas. Within each study area boundary, study objectives that required modifi cations of stan- we surveyed all areas regardless of land ownership or dard protocols used in Northern Spott ed Owl stud- habitats present, with the exception of the greater LAS ies. Therefore, we present here the most consistent where only known Spott ed Owl habitat or previously methodology used among the studies, but we also occupied habitats were surveyed, and the RSA of the POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 11

ELD where owl territories were selected from historic captured most owls with noose poles, snare poles, locations provided by the USFS or from territories or mist nets. Once captured, we fi tt ed all owls with that we located during surveys conducted in 1997. We a locking federal (U.S. Fish and Wildlife Service) alu- surveyed territories on the greater LAS and RSA of the minum band on the tarso-metatarsus of one leg. On ELD individually. Because those individual territories the opposite leg, we marked adult and subadult owls were not within the bounded density study areas, we with a unique combination of color band and color did not use them in estimation of λt (see below for tab (Forsman et al. 1996), whereas we fi tt ed juvenile assumptions of λt). Prior to 1990 on the ELD, funding owls with a plastic band having a color unique to was not suffi cient to adequately survey the density their cohort. We refi tt ed juveniles with unique color study area; thus, abundance estimates before 1990 bands and tabs when recaptured as territory holders were not comparable to estimates from later years in later years. and were not used in estimation of λt. In SIE and SKC, We determined sex of nonjuvenile owls by their study areas were divided into sites of the approximate calls and behavior. Males have a lower-pitched call size of owl territories. We included only those sites than females and only females were known to incu- that were surveyed consistently each year in that bate or brood young (Forsman et al. 1984). We did analysis. In 1990, survey eff ort on SKC was less than not sex juveniles, except on the SIE and SKC (Steger during the following years and those data were not 1995). We identifi ed four age-classes on the basis of included in our analysis (Table 1). Finally, in the SAB, plumage characteristics (Forsman 1981, Moen et al. we only surveyed forest habitat because owls only 1991): juvenile; one year old (fi rst-year subadult); two occupied forested habitat in that mountain range and years old (second-year subadult); and three or more there were extensive areas of nonforest habitat (e.g. years old (adult). chaparral) throughout the range. Aft er initial capture, we identifi ed adult and sub- Estimation of reproductive eff ort.—We estimated owl adult owls as individuals by resighting their unique reproductive activity by feeding live mice to owls color bands and tabs. We resighted band colors us- during walk-in surveys (Forsman 1983, Franklin et al. ing binoculars. When possible, two biologists made 1996a). Reproducing owls usually take off ered prey independent observations of the same bird’s color to their nest or young, whereas nonreproducing owls band–tab combination. We recaptured birds and re- usually eat or cache the mice. We estimated fecundity placed bands when a color tab became frayed through (i.e. the number of female young fl edged per female; wear. When color bands were changed, we recorded Caughley 1977) from number of females checked for the metal band number. Band loss was minimal reproductive status and number of young observed. (Forsman et al. 1996). We assumed a 1:1 sex ratio of juveniles for fecundity Pre-analysis data screening.—Our basic philosophy estimates (Steger 1995, Franklin et al. 1996b). and framework for the meta-analysis workshop fol- Criteria for inferring nonreproduction for a pair of lowed Anderson et al. (1999). All groups conducting owls varied slightly among study areas (Appendix 1). California Spott ed Owl demographic research and However, we used a data screening process to evalu- experts in demographic analysis and parameter ate the internal consistency of data collected given the estimation were invited to att end the meta-analysis methods used by an individual study (see below). In workshop. Further, representatives from the timber addition, we discussed the effi cacy of each study area’s industry and environmental groups were invited to criteria at length during the workshop to assess if the att end. To our knowledge, the data analyzed repre- data were suffi ciently consistent and rigorous to use sented the extent of current data on population dy- in a collective meta-analysis. Researchers generally namics of California Spott ed Owls. The demographic agreed their data provided unbiased estimates of fe- parameters of interest for the meta-analysis were sex- cundity within their respective study areas. However, specifi c survival, female fecundity, and population we felt that, because diff erent protocols were used, it rate of change. Therefore, in the interest of data was not appropriate to analyze the study areas jointly consistency, researchers from each study area were without further investigation, such as comparing requested by the organizers to summarize their data estimated fecundity for each study using protocols in (1) a data fi le with a capture-history matrix that employed by the other studies. Such an analysis described the capture–recapture history of each in- would have required writing programs to subsample dividual owl, its federal band number, its age at fi rst complete data sets of survey data (each study area has capture (juvenile, fi rst-year subadult, second-year conducted thousands of surveys and each study area subadult, or adult), and its sex; (2) a data fi le with stores those records diff erently), which would have annual number of young fl edged (0, 1, 2, or 3) for taken more time than allott ed for the workshop. individual territorial owls, their territory, social status Capture, banding, sex and age identifi cation, and re- (paired or single owl), age of the male, and age of the sighting of owls.—We att empted to capture and band female; and (3) a data fi le with a capture-history ma- all detected Spott ed Owls following the methods trix that documented the capture–recapture history of of Forsman (1983) and Franklin et al. (1996a). We all individuals encountered as territory holders (i.e. if 12 ORNITHOLOGICAL MONOGRAPHS NO. 54 an individual was fi rst banded as a juvenile only the viewed (see Anderson et al. 1999). All researchers territorial portion of the history was included), its age signed the certifi cation. at fi rst capture, its sex, and its federal band number. Meta-analysis workshop format.—We devoted the The latt er database was created aft er the survival da- fi rst day-and-a-half of the workshop to a discussion of tabase had been checked for errors (see below) and the methods used to infer reproductive output of owls was used to estimate population rate of change. because methods varied somewhat between study ar- Although most studies did not use DNA tests to eas (Appendix 1, see above). Consensus was reached ascertain the sex of juveniles, we assumed a 1:1 sex among researchers that the methods used by the ratio at fl edging in each year for the capture-history respective studies, despite their diff erences, were ap- matrix (Steger 1995, Franklin et al. 1996b). We only propriate given the study objectives, location, and be- used owl reproductive data for each study area that havior of owls within the study area. In addition, we was consistent with the protocols developed for each engaged in considerable discussion and debate about study (see Appendix 1). However, number of young the nature of the analyses to be performed, appropri- found in each territory each year had to be based on ate inferences to be drawn given a particular analysis, a minimum of two visits within a year. An exception and advantages and disadvantages of diff erent ap- could be made if statistical justifi cation was provided proaches. We reached agreement on the structure and that indicated single visits had high accuracy (>85%) nature of analyses and who would perform a specifi c for counting young. Regardless, we did not compare analysis. Our discussion also led to a departure from estimates of fecundity across study areas in a meta- past approaches for Spott ed Owl population analyses analysis because of diff erences in protocol used by (see discussion of λ estimation below). Because we each study. knew our eff ort would be closely followed by many Most researchers had to modify their database interested parties, we developed and recorded a pro- structure to conform to the meta-analysis specifi ca- tocol during the workshop (Appendix 2). tions for a compatible database structure (see above). We devoted the remainder of the workshop to selec- There also was some variation in research protocols tion of relevant a priori models for reproduction, recap- among the study areas. Therefore, prior to att ending ture, and survival modeling (see Appendix 2); then to the one-week workshop and conducting data analysis, executing the a priori models. Because researchers were all research groups agreed to undergo a formal data at diff erent stages of their studies, we agreed that covari- screening process to ensure quality control, to ensure ates (e.g. precipitation, habitat) would not be included in that the original fi eld data matched the data in the the modeling process. However, we also agreed unani- computer fi les, and to ensure that the specifi c criteria mously that that was a worthwhile endeavor to be pur- used by a study was actually followed by that study sued in a future meta-analysis (see below). Researchers (i.e. data collection was internally consistent within a were convened as a group to discuss particular issues, as study). A Spott ed Owl researcher not involved in the they arose, that might aff ect the analysis or to maintain California Spott ed Owl meta-analysis was tasked with consistency in the analytical process. Thus, we agreed randomly selecting information from the databases that the results of our analysis would be a fi rst step supplied by the respective study-area researchers. in sett ing the basis for subsequent workshops, which Ten records were randomly drawn from the capture- would allow more inclusive analyses. history database from each study area and 10 from the reproductive database; individual researchers Data Analysis were then required to provide paper copies of the associated original data forms or fi eld notes. At least one Direct inferences from our results are limited to male and one female were drawn from each age-class study populations analyzed and time periods during to check the survival database. The randomly selected which data were collected. Inferences beyond the study information was then compared with the actual fi eld populations (e.g. to the California Spott ed Owl through- data recorded on original fi eld notes. If errors were out the Sierra Nevada) are not possible with those data found, an additional 10 were randomly selected for because the study areas represent only a small fraction checking. If errors were found aft er the second check, of the total area of the Sierra Nevada range, and those the entire database was checked for errors. study areas were not selected randomly from a sam- Aft er the fi rst day of the workshop and prior to pling frame encompassing the Sierra Nevada. any data analysis, all researchers were required to Changes in analytical methodology from previous sign a certifi cation lett er stating that their data were Spott ed Owl studies.—There has been considerable correct, had been checked and rechecked, and were debate over the most appropriate measure of the ready for fi nal analysis. Failure to sign the certifi ca- fi nite rate of population change (λ) in Spott ed Owl tion would have meant exclusion of their data from populations. Historically, Spott ed Owl researchers the fi nal analysis. Further, by signing the certifi cation, have estimated λ using a Leslie projection matrix researchers explicitly agreed that their data could not (λPM), which was based on estimates of age- or be withdrawn from the analysis aft er results were stage-specifi c survival and fecundity (Franklin et al. POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 13

1996a, Caswell 2000). That method was the best avail- would not support a projection matrix approach for able at the time for estimating rates of population some of those demographic studies. Thus, we de- change. Nevertheless, the debate on rates of popula- cided to rely on λt, which estimates λ directly from tion change in Spott ed Owls using λPM has centered the capture–recapture data, to estimate changes in on the central issue of whether λPM is biased because owl numbers within study areas (Pradel 1996, Nichols the populations are not geographically closed (e.g. and Hines 2002). Inferences and assumptions relevant there are unknown rates of juvenile emigration from to that technique are explained more fully below and the study areas). If banded juvenile owls leave the in Appendix 3. study area, live, and remain undetected, an estimate Estimating adult survival.—The meta-analysis of of juvenile survival using mark–recapture estimators adult apparent survival was based on adult female will be negatively biased. For example, estimates of and adult male capture histories for the fi ve study ar- juvenile survival probabilities on three study areas eas, where captures were either initial captures, recap- for the Northern Spott ed Owl increased 42–137% tures, or resightings of color-banded individuals. We when they were adjusted, using radiotelemetry data, defi ned apparent survival (φ) as the probability that for emigration from those study areas (Franklin et an owl alive in a particular year t survived to the same al. 1999). Conversely, reproductively active owls are time next year (t + 1) and remained on the study area more likely to be detected than nonreproductively and, hence, was available for recapture. The reciprocal active owls, which could result in an overestimate of apparent survival was a function of both death and (i.e. positive bias) of reproductive output. If biased emigration. We assumed that permanent emigration survival or reproductive output estimates are used of adult Spott ed Owls from study areas was very low, on the basis of data on Northern Spott ed Owls (e.g. in the projection matrix, estimates of λPM would be biased as well. Thus, an important issue concerns the Franklin et al. 1996b, Forsman et al. 2002). Hence, we considered apparent survival for California Spott ed correct inference to be taken from λPM (Raphael et al. 1996). With the exception of the SAB (essentially, a Owls to be an approximate estimate of true survival, geographically closed population for which there was the reciprocal of which was death only. a good estimate of juvenile survival), we could not be From capture histories of individuals fi rst captured certain that we did not have a biased estimate of juve- as juveniles or subadults, we removed encounters at nile survival because of the likelihood of undetected the younger ages, leaving only captures at the adult age. Estimates of apparent survival and recapture juvenile emigration from the study areas. Previously, probability (p, probability that an animal alive in year studies in the Sierra Nevada (with the exception of t is captured, recaptured, or resighted) were obtained the LAS, where juvenile survival was estimated from with the Cormack-Jolly-Seber model (Lebreton et recapture data) used a projection matrix based on al. 1992) using Program MARK. The global model estimates of territorial owl survival and fecundity, considered was (φ , p ), where φ was apparent and a “surrogate” estimate of juvenile survival which g*t*s g*t*s survival probability, p was recapture probability, g was “borrowed” from the SAB. The use of a surrogate was study area, t was time (year), and s was sex. We estimate of juvenile survival probably introduced an assessed goodness-of-fi t of this model with program unknown bias into the estimates of λ because of po- PM RELEASE (Burnham et al. 1987). Assumptions under- tential geographic variation in survival rates. lying use of mark–recapture data for Spott ed Owls Despite those potential problems, we decided, and use of goodness-of-fi t to evaluate those assump- during the early stages of the workshop, to estimate tions was discussed in greater detail by Franklin et al. λ because there was suffi cient disagreement from PM (1996a). In general, studies on California Spott ed Owls some participants in the workshop concerning com- were very similar in design to those for the Northern plete exclusion of λPM from the analysis. Thus, our Spott ed Owl. We estimated overdispersion in the initial approach was to estimate rates of population data using ĉ = χ2/df using the combined chi-square 2 change using both λPM (see Appendix 2) and a re- (χ ) values and degrees of freedom (df) from TEST 2 cently developed analytical technique for estimation and TEST 3 from program RELEASE (Lebreton et al. of λ (referred to here as λt; see Pradel 1996). That new 1992). Estimates of ĉ were used to correct estimated method was employed in a Northern Spott ed Owl meta- standard errors and Akaike's Information Criterion analysis (Franklin et al. 1999). We spent considerable (AICc) values (see below). Twenty-seven models were time att empting to estimate juvenile survival from the initially fi t to the data from the fi ve study areas with capture–recapture data. However, we encountered three structures (φg*t*s, φg*t+s, and φg*t) on apparent sur- problems in estimability of parameters for juvenile vival, and nine structures (pg+t, pg+t+s, p[g+t]*s, pg*t, p[g*t]+s, survival. In att empting to solve those problems, other pg*t*s, pr, pr+s, and pr*s) on recapture, where r was annual issues concerning bias in estimates of juvenile surviv- reproductive output estimated from the fi ve study al became apparent (see also Appendix 3). Eventually, areas. Program MARK (White and Burnham 1999) we collectively decided that estimates of λPM would generated the log-likelihood function value, degrees have some unknown bias because of those problems, of freedom, and the small-sample bias-corrected and analysts and most researchers agreed the data quasi-likelihood AIC (QAICc) (Sakamoto et al. 1986, 14 ORNITHOLOGICAL MONOGRAPHS NO. 54

Burnham and Anderson 1998) for each model evalu- separately because estimates were not comparable ated. That criterion was computed as follows: across study areas (except for the SIE and SKC study areas) because of diff erences in the fi eld protocols used to estimate reproductive output (number of QAIC = + 2K + c fl edged young per pair). Raw data used in the analysis for each study area consisted of number of young where K was the number of parameters estimated, fl edged on a particular site, site (territory) where the ĉ was the estimate of overdispersion, and n was the young were detected, year, and age of the female eff ective sample size (i.e. number of binomial trials (fi rst-year subadult, second-year subadult, or adult), included in the likelihood following Burnham et al. for each female monitored. Prior to analysis, we di- 1987). The smaller the QAIC value for a given model, c vided the estimates of reproductive output for each the bett er an approximation the model was for the site within each year by 2 to estimate fecundity, as- information in the observed data, given the set of suming a 1:1 sex ratio. We used mixed models because models examined. (1) individuals and territories were confounded over Using the minimum QAIC model for p from the c time because the same females oft en bred on the same initial 27 models, we fi t the following 10 additional territory for >1 year; that lack of independence would models for φ: φ , φ , φ , φ , φ , φ , φ , φ , φ , and g g+t g+T g*T g+TT g*TT t T TT underestimate standard errors if methods assum- φ , where T denoted a linear time trend, TT denoted a · ing independence were used (Franklin et al. 1999); quadratic time trend, and “.” denoted a means only (2) modeling could be conducted in a maximum- model. Using the minimum QAIC model from those c likelihood framework; (3) inference was made to sites 10 models that included a study area eff ect, we fi t rather than to separate outcomes–year by adjustments additional models: φ , φ , φ , and φ latitude SAB vs. rest SKC vs. rest SAB, of the standard errors; (4) the error covariance matrix , where the term rest denoted a single survival SKC vs. rest could be structured appropriately; and (5) models al- parameter estimated for the remaining study areas lowed for unbalanced designs (e.g. missing data). combined. Thus, we considered a total of 41 models Raw data used in the analysis were integer data (0, (Table 2). 1, 2, and rarely 3 or 4). Analysis of Northern Spott ed The Akaike weights (Burnham and Anderson 1998) Owl fecundity data showed that variation in number were computed as fl edged within a year was proportional to the mean, which suggests a Poisson distribution (Evans et al. 1993), although data were not distributed as Poisson (Franklin et al. 1999, 2000). When we analyzed the California Spott ed Owl data using mixed-model analysis of variance, we relied on sample sizes that were suffi ciently large to justify normal distributional where ∆i was the diff erence in QAICc value for the assumptions. On the basis of simulations, analysis of minimum QAICc model and model i, and R was the variance models were robust to severe departures number of models (41 in our case) in the set. Model from normality (White and Bennett s 1996). In addi- likelihood was computed as the QAICc weight for the tion, analysis of variance models were more robust to model of interest divided by the QAICc weight of the data from discrete distributions, such as the negative best model. Temporal and spatial process variation of binomial, than was Poisson regression, even when it apparent survival were estimated with the variance was corrected for overdispersion (White and Bennett s components module of Program MARK (White et al. 1996). Therefore, we decided to rely on the robustness 2002, Burnham and White 2002). We distinguished of analysis of variance to non normally distributed 2 between process variation (σ process)—the variation in a data, rather than relying on Poisson regression, to given parameter, such as φ—and sampling variation analyze the fecundity data, which have properties —the variation att ributable to estimating similar to a Poisson but are not distributed as Poisson. a parameter from sample data (White 2000). Process The mixed-model procedures also allowed us to ac- variation in population parameters can be decom- count for that dependence of sampling variation on posed into temporal process variation (variation in the mean (see below). As with Northern Spott ed Owl a parameter over time) and spatial process variation analyses (Franklin et al. 1999), we did not separate (variation in a parameter among diff erent locations), individual bird eff ects from territory eff ects because which requires that sampling variation be “removed” of the longevity of most individual females on from the total variation in the annual or spatial esti- territories. mates of interest. We used PROC MIXED (SAS Institute 1997) to fi t Estimating fecundity.—We analyzed fecundity various models to estimate fecundity for each study data for each study area with mixed analysis of vari- area. Initially, we followed the protocol developed ance models (Rao 1997). We analyzed study areas during the workshop. However, several complications POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 15

=,_=*8(7.59.438=4+=2&70:7*(&59:7*=24)*18=*=&2.3*)=.3=9-*=2*9&8&3&1>8.8=4+=&):19=&55&7*39=8:7;.;&1=φ= 4+=2&1*=&3)=+*2&1*=&1.+473.&= 5499*)= <18=+742=+.;*=89:)>=&7*&8=.3=&1.+473.&_= = "4)*1= *8(7.59.43=4+=φ=897:(9:7*= *8(7.59.43=4+=5=897:(9:7*=

φ_`=5,$9 == &4=*++*(98= 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= φ,`=5,$9 == 9:)>=&7*&=*++*(9= 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438=

φ,(9`=5,$9 == 9:)>=&7*&=*++*(9=<.9-=&)).9.;*=>*&7=*++*(98= 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438=

φ,(`=5,$9 == 9:)>=&7*&=*++*(9=<.9-=&)).9.;*=1.3*&7=9.2*= 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= = = *++*(9=

φ,(`=5,$9 == 9:)>=&7*&=*++*(9=<.9-=&)).9.;*=6:&)7&9.(= 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= = = 9.2*=*++*(9=

φ,$9`=5,$9$8 == 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= 9:)>=&7*&`=>*&7`=&3)=,*3)*7=*++*(98=<.9-=&11== = = = .39*7&(9.438=

φ,$9`=5,$9(8 == 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438== = = = &3)=&3=&)).9.;*=,*3)*7=*++*(9=

φ,$9`=5,(9(8 == 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= ))).9.;*=89:)>=&7*&`=>*&7`=&3)=,*3)*7=*++*(98=

φ,$9`=5,$9 == 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= φ,$9`=57 == 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= )33:&1=7*574):(9.;*=7&9*=*++*(9=

φ,$9`=57$8 == 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= )33:&1=7*574):(9.;*=7&9*=&3)=,*3)*7=*++*(98== = = = <.9-=.39*7&(9.438=

φ,$9`=5,(9 == 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= ))).9.;*=89:)>=&7*&=&3)=>*&7=*++*(98=

φ,$9`=57(8 == 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= ))).9.;*=&33:&1=7*574):(9.;*=7&9*=&3)== = = = ,*3)*7=*++*(98=

φ,$9`=5,(9$8 == 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= ))).9.;*=89:)>=&7*&=&3)=>*&7=*++*(98== = = = .39*7&(9.3,=<.9-=,*3)*7=

φ,$`=5,$9 == 9:)>=&7*&=&3)=1.3*&7=9.2*=*++*(98= 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= = = <.9-=.39*7&(9.438=

φ,$9(8`=57 == 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= )33:&1=7*574):(9.;*=4:95:9= = = &3)=&3=&)).9.;*=,*3)*7=*++*(9=

φ,$9(8`=57(8 == 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= )33:&1=7*574):(9.;*=4:95:9=<.9-=&3=&)).9.;*== = = &3)=&3=&)).9.;*=,*3)*7=*++*(9= = ,*3)*7=*++*(9=

φ,$9(8`=5,$9$8 == 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= 9:)>=&7*&`=>*&7`=&3)=,*3)*7=*++*(98=<.9-=&11== = = &3)=&3=&)).9.;*=,*3)*7=*++*(9== = .39*7&(9.438=

φ,$9(8`=5,$9(8 == 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438== = = &3)=&3=&)).9.;*=,*3)*7=*++*(9= = &3)=&3=&)).9.;*=,*3)*7=*++*(9=

φ,$9(8`=5,(9$8 == 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= ))).9.;*=89:)>=&7*&=&3)=>*&7=*++*(98== = = &3)=&3=&)).9.;*=,*3)*7=*++*(9= = .39*7&(9.3,=<.9-=,*3)*7=

φ,$9(8`=5,(9 == 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= ))).9.;*=89:)>=&7*&=&3)=>*&7=*++*(98= = = &3)=&3=&)).9.;*=,*3)*7=*++*(9=

φ,$9(8`=5,(9(8 == 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= ))).9.;*=89:)>=&7*&`=>*&7`=&3)=,*3)*7=*++*(98= = = &3)=&3=&)).9.;*=,*3)*7=*++*(9=

φ,$9(8`=5,$9 == 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= = = &3)=&3=&)).9.;*=,*3)*7=*++*(9=

φ,$9(8`=57$8 == 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= )33:&1=7*574):(9.;*=7&9*=&3)=,*3)*7=*++*(98== = = &3)=&3=&)).9.;*=,*3)*7=*++*(9= = <.9-=.39*7&(9.438=

φ,$9$8`=57(8 == 9:)>=&7*&`=>*&7`=&3)=,*3)*7=*++*(98=<.9-= ))).9.;*=&33:&1=7*574):(9.;*=7&9*=&3)== = = &11=.39*7&(9.438== = ,*3)*7=*++*(98=

φ,$9$8`=57$8 == 9:)>=&7*&`=>*&7`=&3)=,*3)*7=*++*(98=<.9-= )33:&1=7*574):(9.;*=7&9*=&3)=,*3)*7=*++*(98== = = &11=.39*7&(9.438== = <.9-=.39*7&(9.438=

φ,$9$8`=5,(9$8 == 9:)>=&7*&`=>*&7`=&3)=,*3)*7=*++*(98=<.9-= ))).9.;*=89:)>=&7*&=&3)=>*&7=*++*(98== = = &11=.39*7&(9.438== = .39*7&(9.3,=<.9-=,*3)*7=

φ,$9$8`=5,(9 == 9:)>=&7*&`=>*&7`=&3)=,*3)*7=*++*(98=<.9-= ))).9.;*=89:)>=&7*&=&3)=>*&7=*++*(98= = = &11=.39*7&(9.438=

φ,$9$8`=57 == 9:)>=&7*&`=>*&7`=&3)=,*3)*7=*++*(98=<.9-= )33:&1=7*574):(9.;*=7&9*=*++*(98= = = &11=.39*7&(9.438=

φ,$9$8`=5,$9 == 9:)>=&7*&`=>*&7`=&3)=,*3)*7=*++*(98=<.9-= 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= = = &11=.39*7&(9.438= 16 ORNITHOLOGICAL MONOGRAPHS NO. 54

=,_=439.3:*)_= = "4)*1= *8(7.59.43=4+=φ=897:(9:7*= *8(7.59.43=4+=5=897:(9:7*=

φ,$9$8`=5,$9$8 == 9:)>=&7*&`=>*&7`=&3)=,*3)*7=*++*(98=<.9-= 9:)>=&7*&`=>*&7`=&3)=,*3)*7=*++*(98=<.9-=&11== = = &11=.39*7&(9.438== = .39*7&(9.438=

φ,$9$8`=5,(9(8 == 9:)>=&7*&`=>*&7`=&3)=,*3)*7=*++*(98=<.9-= ))).9.;*=89:)>=&7*&`=>*&7`=&3)=,*3)*7=*++*(98= = = &11=.39*7&(9.438=

φ,$9$8`=5,$9(8 == 9:)>=&7*&`=>*&7`=&3)=,*3)*7=*++*(98=<.9-= 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438== = = &11=.39*7&(9.438== = &3)=&3=&)).9.;*=,*3)*7=*++*(9=

φ,$`=5,$9 == 9:)>=&7*&=&3)=6:&)7&9.(=9.2*=*++*(9=<.9-= 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= = = .39*7&(9.438=

φ &9.9:)*`=5,$9 == *&9.9:)*=*++*(9= 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438=

φ`=*89`=5,$9 == +74:5=*++*(9=4+= ),=89:)>=&7*&=;*78:8=49-*7= 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= = = 89:)>=&7*&8=

φ`=`=*89`=5,$9 == +74:5=*++*(9=4+= ),=89:)>=&7*&=;*78:8= -= 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= = = 89:)>=&7*&=;*78:8=49-*7=89:)>=&7*&8=

φ`=*89`=5,$9 == +74:5=*++*(9=4+= -=89:)>=&7*&=;*78:8=49-*7= 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= = = 89:)>=&7*&8=

φ9`=5,$9 == .*&7=*++*(9= 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438=

φ`=5,$9 == *.3*&7=9.2*=*++*(9= 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438=

φ`=5,$9 == /:&)7&9.(=9.2*=*++*(9= 9:)>=&7*&=&3)=>*&7=*++*(98=<.9-=.39*7&(9.438= arose, including (1) mis-specifi cation of the SAS code apparent that some of the covariance structures had used to run the models and (2) nonconvergence of been incorrectly coded. In addition, some models and some of the models because there were too few in- covariance structures analyzed during the workshop dividuals in the subadult age classes on most study failed to converge: the log-linear covariance structure areas. In both the initial analysis and the subsequent for the LAS, SIE, and SKC study areas; and fi xed- re-analysis, we used a two-stage approach to fi t eff ect models with age and year interactions for the models to the data for each study area (see Wolfi nger ELD study area. That failure to converge was because 1993). First, we used restricted maximum likelihood there were ≤3 subadults (fi rst- and second-year sub- estimation with model {age + T; fecundity = β0 + β1 adult age-classes combined) for six, four, four, and

(age) + β2 (year)} for each study area, where age of the eight years on the ELD, LAS, SIE, and SKC study ar- bird was a categorical variable (either fi rst-year sub- eas, respectively. Therefore, we re-analyzed the data adult, second-year subadult, or adult) and year was with correctly specifi ed covariance structures and we a continuous variable. That model was executed with used only adult females for all the time-trend models each of four candidate variance structures: log-linear on all fi ve study areas. Using data from only ≥3 year variance (LOCAL = EXP(AGE YEAR), compound old females, the SAS code for each of the covariance symmetric (CS), fi rst-order autoregressive (AR1), structures examined using restricted maximum likeli- and heterogeneous fi rst-order autoregressive (ARH1) hood estimation was: (SAS Institute 1997). We selected the most appropriate PROC MIXED METHOD = REML; covariance structure using AICc with only the covari- RANDOM SITE YR; ance parameters as the number of parameters used in REPEATED / LOCAL = EXP(YR) SUB = SITE; calculating AICc because restricted maximum likeli- for log-linear variance; hood estimation ignores the fi xed eff ects (Wolfi nger PROC MIXED METHOD = REML; 1993). We used this step to select the most appropriate RANDOM YR; covariance structure for inclusion in the following REPEATED YR / TYPE = CS SUB = SITE; fi xed-eff ects models with fecundity as the response for the compound symmetric; variable: quadratic time trend (TT), linear time trend PROC MIXED METHOD = REML; (T), even–odd years (EO), linear time trend with an RANDOM SITE YR; additive even–odd year eff ect (T + EO), and no time REPEATED YR / TYPE = AR(1) SUB = SITE; trend (intercept, a means or intercepts-only model). for the fi rst-order autoregressive; and We used full maximum-likelihood estimation (rather PROC MIXED METHOD = REML; than restricted maximum-likelihood estimation) to RANDOM SITE YR; analyze those models; the number of parameters in REPEATED YR / TYPE = ARH(1) SUB = SITE; that case were the number of covariance parameters for the heterogeneous fi rst-order autoregressive. plus the number of fi xed eff ect parameters. The model Again, we did not compute a meta-analysis across that best explained the data for each study area was the fi ve study areas because of the diff erence in proto- selected using AICc. Aft er the workshop, it became cols used to estimate reproductive output in the fi eld. POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 17

However, we made comparisons between the SIE and approximate true survival (e.g. previous Spott ed Owl SKC study areas because those two areas used similar analyses in Burnham et al. 1996, Franklin et al. 1999). fi eld protocols. In addition, we estimated temporal Those eff orts reduce the movement asymmetry and process variation in fecundity using an intercept-only its eff ect on . However, those adjustments require model for each study area. additional information on emigration, such as infor- Estimating rates of population change.—We estimation from radiomarked birds. Second, the com- mated the rate of population increase (λ) using the puted is an asymptotic value expected to result temporal symmetry capture–recapture model of Pradel from the complete absence of temporal variation in (1996), which was denoted as where RJS was the vital rates, whereas there is likely to be evidence “Reparameterized Jolly-Seber.” For notational ease, we of temporal variation in the data. Thus, is a con- denoted a year-specifi c RJS estimator as for year t. stant value over a specifi ed time period whereas Prior analyses of Spott ed Owl data have used a provides annual estimates that capture the temporal modifi ed Leslie projection matrix (Franklin et al. variability in rates of population change. Third, val- 1996a) which can be denoted as where PM ues of fecundity may be positively biased if nonbreed- denotes “projection matrix” and refers to a stage- ing birds are not detected or if unsuccessful birds are structured projection matrix approach (Caswell 2000). not detected as readily as successful birds. The fourth The two types of λ diff er in their defi nitions and inter- reason is related to the fi rst and involves the fact that pretations, as well as in their methods of computation. estimates of juvenile survival are probably nega- Here, we present a brief discussion of those distinc- tively biased when they are obtained using capture– tions and diff erences. recapture methods (Franklin et al. 1999). That is of concern with California Spott ed Owls because of Variable estimates λt, the rate of change in population size between years t and t + 1: the paucity of data for estimating juvenile survival. In summary, the should provide reasonable estimates of annual rates of change in abundance of territorial birds on the study area. The is perhaps best viewed as an abstraction to the extent that (1) it is an asymptotic quantity that assumes no temporal where Nt is abundance at year t. In the case of the variation in vital rates, and (2) it includes all losses California Spott ed Owl analysis, abundances and from, yet not all gains to, the population (no recruit- apply to subadult and adult territorial owls on the ment from outside the study area is included in that study areas. That rate of change in abundance is a quantity). function of the four fundamental demographic vari- There are several assumptions underlying estima- ables: survival rate, reproductive rate, emigration, tion of λt that need to be considered (see Hines and and immigration. Thus, demographic explanations Nichols 2002 and Franklin 2002 for more complete for specifi c values of require additional informa- details). First, interpretation of such estimates is most tion on those fundamental demographic variables. straightforward when study area size and boundary Variable is computed from projection ma- confi gurations remain unchanged through time. If trices parameterized with means of time-specifi c study areas expand or contract over time, the result- estimates, or constant-parameter model estimates, ing will refl ect that the population to which infer- for stage-specifi c survival and fecundity for juvenile, ences are being made is also expanding or contracting. subadult, and adult survival. The resulting from Second, all animals within the study area must have those computations represents the asymptotic growth some probability of being recaptured throughout the rate for a population exposed to the projection matrix study. If portions of the study area are inaccessible vital rates year aft er year. That value can be viewed as during some years of the study, but then become a function of the average vital rates but is not neces- accessible for trapping in subsequent years, individu- sarily a good estimate of the average rate of change als captured in the inaccessible portion of the study in number of birds on the study area for at least four area will suddenly become “new recruits” to the reasons. First, there is an asymmetry in the way move- population even though they had been present, but ment is treated in vital rates representing gains and not available for sampling, in previous years. Third, losses. New individuals are added to the projected permanent trap response in capture probability can population only via in situ reproduction, as refl ected bias estimates of λt (Hines and Nichols 2002). When in the fecundity estimates. However, Spott ed Owl animals respond positively or negatively to being cap- survival estimates represent apparent survival in that tured (Seber 1982), a diff erence in capture probability their complements include both death and perma- occurs between animals that have, and have not, been nent emigration from study areas. Thus, losses from captured previously and marked. Permanent trap the population occur via both death and permanent response in the standard Cormack-Jolly-Seber mod- emigration. Note that sometimes eff orts are made to els induces no bias in survival estimates (Pollock et adjust apparent survival estimates so that they bett er al. 1990), but estimates of population size under the 18 ORNITHOLOGICAL MONOGRAPHS NO. 54

Cormack-Jolly-Seber model are biased in the face of oft en include the speciﬁ cation that all emigration is permanent trap response because the diﬀ erence in permanent. As noted by Kendall et al. (1997), random capture probability between marked and unmarked temporary emigration led to unbiased estimates of the animals causes predictable problems (Nichols and size of the “superpopulation,” consisting of birds hav- Hines 1984). That same bias also applies to estimates of ing some chance of being in the sampled area during

λt. Hines and Nichols (2002) found that bias is positive the sampling period. In the case of random temporary in the presence of a trap-happy response and negative emigration, we expected estimates of λt to be unbiased in the presence of trap-shy response. That bias is not for changes in superpopulation size. However, non- substantial for small levels of trap responses but could random (e.g. Markovian) forms of temporary emigra- be if levels of trap response are high. If trap response tion could lead to biased estimates of λt, and we were changes over time, then misleading trends in could not aware of investigations on the consequences of result. Fourth, estimates of λt are biased in the presence such temporary emigration to estimation. of heterogeneous capture probabilities among indi- The primary inference from regarding popula- viduals or unidentifi ed classes of individuals. Hines tion rate of change was to the territorial owls on the and Nichols (2002) show that heterogeneous capture study area. Previously, the primary inference from probabilities do not bias estimates of λt when popula- had been phrased as “did the territorial owls tion growth rate is modeled as a single constant over on the study area replace themselves?” (Burnham all time periods. Small bias does occur when estimat- et al. 1996, Franklin et al. 1999). However, that infer- ing time-specifi c λt. However, that bias was not as sub- ence provided no information on where replacement stantial a problem as that resulting from permanent owls might go with respect to study area boundar- trap response. ies. In other words, the inference applied if all of the Violation of the fi rst two assumptions do not pro- young produced on a study area remained on that duce bias, in that the estimator of λt is not performing area and then exhibited similar survival and fecun- as it was intended (Hines and Nichols 2002). When the dity rates of adults on that area. Thus, we rephrased study area changes, the estimated population change that inference as “would the territorial owls on the is the result of two conceptually distinct processes. study area replace themselves if the system was geo- The fi rst process involves expansion of the study graphically closed?” In contrast, the inference from area and increase in number of animals exposed to included recognition that the system may not be sampling that result from that expansion. The second geographically closed and was phrased as “were the process involves changes in the number of animals on territorial owls on the study area being replaced?” the sampled area; that is the change of interest and the (Franklin et al. 1999). That inference was about the one to which we would like estimates of λt to apply. owl populations residing within a specifi c study area.

Other assumptions underlying open capture– An advantage of λt was that estimates of juvenile sur- recapture models have not been specifi cally investi- vival were not required because both immigration and gated with respect to eff ects on . For example, we emigration were accounted for by changes in number assumed no tag loss and no tag-induced mortality. of owls over time. Thus, the potential bias from impre- Because we had no reason to suspect that those were cise and inaccurate estimates of juvenile survival was important problems, we did not investigate conse- avoided. However, a primary limitation of inferences quences of their violation. However, loss of the same concerning was that it was not possible to estimate type of bands used on 3,788 Northern Spott ed Owls the relative contributions of the diff erent components was only 0.1% (Forsman et al. 1996). In addition, to population growth (e.g. reproduction, immigration, we recaptured owls when color band combinations death, and emigration) without additional data. For became diffi cult to read. Similarly, homogeneity of example, immigration could sustain a demographi- demographic rate parameters (e.g. survival) among cally (based on survival and reproduction) declining individuals is assumed in open population cap- population (i.e. the population could be a sink; sensu ture–recapture models. Our focus on territorial birds Pulliam 1988). It should be noted that most Spott ed eliminated the potentially large variation between Owl populations, as defi ned by the usual scale of territorial and “fl oater” birds, and we did not know study, were likely maintained by immigration (behav- the consequences of remaining variation in param- ior commonly att ributed to sink populations), while eters among individual territorial birds. As with also supplying recruits to other populations (behavior most explorations of heterogeneous rate parameters, commonly att ributed to source populations). Thus, we suspected that substantial variation could lead to the source–sink dichotomy may not be as useful with important bias, whereas relatively minor variation Spott ed Owl populations as with some other animal would be less of a problem. The high annual survival populations. Thus, was an important tool but will estimates for Spott ed Owls did not permit substantial not suffi ce as a single assessment of the health of a heterogeneity (i.e. it would not be possible to have Spott ed Owl population. Consequently, all relevant in- such high mean survival if many individuals exhibited formation should be used to draw inferences about the greatly reduced survival). Open model assumptions stability of California Spott ed Owl populations (see POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 19

below). For example, estimates and trends in survival population size was computed as Nx = nx/px (where nx and fecundity rates should also be evaluated to assess is the number of birds caught in initial year x) and the health of a Spott ed Owl population. the ﬁ rst recruitment rate as ,

To estimate λt for each study area, we employed assuming . That latt er assumption was random eﬀ ects models in MARK that used the needed because we required an initial abundance for from model {φt, pt, λt} as the basis for the analysis. demographic simulations but were unable to estimate

Model {φt, pt, λt}, which was used as the basis for the the initial capture probability. That particular solution random-eff ects models, allowed φ, λ, and p to vary (equating the fi rst and second capture probabilities) by year; none of the parameters were otherwise con- seemed reasonable to us, although other solutions strained. We considered the following random-eff ects to that same problem have been used by others (e.g. models: linear trend in λ (T), quadratic trend in λ (TT), Jolly 1965; p1 = 1, Schwarz and Arnason and mean λ across time (.). Those three models were 1996). For each simulation, we used model φt, pt, λt to considered with and without elimination of the fi rst compute estimates of λt and, from those, . We ran estimable λt because of potential biases due to trap re- 1,000 simulations; from that distribution of simula- sponse, perhaps exacerbated by a “learning curve” on tions, we computed 95% confi dence intervals from the the part of observers. Because of the possibility of dif- ith and jth values of , where i = (0.05)(1000) and j = ferent capture probabilities for marked and unmarked (0.95)(1000). birds, we tended to disregard the fi rst estimable λt and We also conducted a meta-analysis using only the to focus on models that did not include that param- study areas in the Sierra Nevada (LAS, ELD, SIE, and eter. Prior to conducting the analysis on λt for each SKC) to examine potential correlations in annual study area, we adjusted some of the study area defi - variation among those study areas. We considered nitions to meet the assumption of a geographically fi xed-eff ects models λt, λg*t, and λg+t, where g was consistent area where birds had some probability study area and with φ and p structured as g * t in all of being surveyed. For the ELD, only the capture– three models. We also considered λg+t without the fi rst recapture data from the smaller density study area and second estimable λt; without the fi rst, second, and were used; for the SIE and LAS study areas, a subset last λt; without the fi rst and last λt; and without the of the capture–recapture data from a smaller geo- fi rst λt. In the analysis of the individual study areas graphic area were used. Estimates of overdispersion and the meta-analyses across the four study areas, we (ĉ) were recomputed for the capture–recapture data corrected for overdispersion, and we used the same using program RELEASE and using a global model of model selection approach that was used to estimate

{φs*t, ps*t, λs*t} for each study area, with an interaction adult survival. between sex and year for φ, p, and λ. Comparison of Sierra and Sequoia and Kings Canyon

To make the annual estimates of λt more interpre- national parks study areas.—The SIE and SKC represent- table, we translated those estimates into estimates of ed paired study areas where SIE had been managed realized change of the populations on each study area. for timber production and SKC had been managed Annual realized changes were estimated as the propor- as a national park. They covered similar ranges in tion of the initial population (i.e. in the initial year used elevation and were in close proximity (19 km), mini- for analysis) remaining in year t (i.e. ∆t = Nt/Nx where x mizing diff erences in weather patt erns. Also, investi- is the initial year). Therefore, realized change provided gators used the same fi eld protocols for reproductive the estimated trajectory of the population over the time output on the two study areas, which allowed direct period for which λt was estimated, without requiring comparison of fecundity rates. We used eff ect sizes to estimation of numbers of owls on each of the study compare the diff erences between the two study areas areas. Realized change (∆t) was estimated as: in terms of the three demographic parameters: adult fecundity, adult survival, and mean .

Results

Adult Survival where x was the fi rst estimated λt. For example, if λt was 0.9, 1.2, and 0.7 for years 1993, 1994, and 1995, We used 975 marked adults in the analysis respectively, then for t = 1996 was (0.9)(1.2)(0.7) = of apparent survival (171 from the ELD, 223 0.756 indicating that 75.6% of the starting population from the LAS, 307 from the SAB, 168 from the remained aft er three years. SIE, and 106 from the SKC). From program To compute 95% confi dence intervals for , we used a bootstrap algorithm. Specifi cally, we computed RELEASE, the overall goodness-of-fi t for the 2 time-specifi c recruitment rates (f ) for each study area global model was χ = 184.74 with df = 167 t (P = 0.165), indicating that the global model fi t using model φt, pt, ft in Program MARK. Using those estimates, we simulated data in MARK where the initial the capture–recapture data. The results of that 20 ORNITHOLOGICAL MONOGRAPHS NO. 54

test yielded ĉ = 1.11. The minimum QAICc model {φSKC vs. rest, pg*t}, the apparent survival estimate for for p combined with the three models φg*t*s, φg*t+s, the ELD, LAS, SAB, and SIE study areas com- and φg*t for apparent survival was pg*t. Model bined was = 0.819 (SE = 0.008, 95% CI = 0.802 selection for all 41 models is shown in Table to 0.835), whereas the separate estimate for SKC

3. On the basis of minimum QAICc, the best φ was = 0.877 (SE = 0.016, 95% CI = 0.842 to 0.905). model containing separate survival parameters Conﬁ dence intervals for the estimate from the for all ﬁ ve study areas was {φg, pg*t}, and the four study areas combined and the estimate best overall model was {φSKC vs. rest, pg*t} (Table 3). for SKC did not overlap. The two top-ranked

Estimates of apparent survival for the {φg, pg*t} models (Table 3) included separate estimates of model are shown in Table 4. Under model φ for SKC; those two models comprised >86% of

= -_= 4)*1= 8*1*(9.43= 7*8:198= +742= 574,7&2= = +47= &55&7*39= &):19= 8:7;.;&1= φ= 4+= 2&1*= &3)= +*2&1*= &1.+473.&=5499*)=<18=43=+.;*=89:)>=&7*&8=.3=&1.+473.&_= = = = = 0&.0*= 4)*1=

4)*1= "#(= ∆"#(= <*.,-98= 1.0*1.-44)= ==== %*;.&3(*=

φ=;8_=7*89`=5,$9 == .*-._-+-= *= *_0+1= +_***= .2= ++00_.-1=

φ`=`=;8_=7*89`=5,$9 == .*-0_+..= +_2-+= *_,.1= *_.**= .3= ++00_,*/=

φ,`=5,$9 == .*-3_2..= /_/-+= *_*-3= *_*0-= /+= ++0/_11/= φ_`=5,$9 == .*.+_*/3= 0_1.0= *_*,+= *_*-.= .1= ++1/_,.0=

φ=;8_=7*89`=5,$9 == .*.+_0.,= 1_-,3= *_*+0= *_*,0= .2= ++1-_101=

φ,(`=5,$9 == .*.+_1-1= 1_.,.= *_*+/= *_*,.= /,= ++0/_0**= φ`=5,$9 == .*.,_+-1= 1_2,.= *_*+,= *_*,*= .2= ++1._,0,=

φ`=5,$9 == .*.,_230= 2_/2-= *_**2= *_*+.= .3= ++1,_3/2=

φ&9.9:)*`=5,$9 == .*.-_*-2= 2_1,/= *_**2= *_*+-= .2= ++1/_+00= φ,(`=5,$9 == .*.-_+*-= 2_13*= *_**2= *_*+,= /-= ++0._232=

φ9`=5,$9 == .*.._+20= 3_21-= *_**.= *_**2= //= ++0+_2-3=

φ,(9`=5,$9 == .*.._2*2= +*_.3/= *_**-= *_**/= /3= ++/._+0,= φ,$`=5,$9 == .*.2_-**= +-_321= *_**+= *_**+= /0= ++0-_22*=

φ,$`=5,$9 == .*/,_/,*= +2_,*1= *_***= *_***= 0+= ++/1_1+1=

φ,$9`=5,(9$8 == .*2*_/+/= .0_,*,= *_***= *_***= 1-= ++0*_0//= φ,$9`=5,(9 == .*2+_*2-= .0_11*= *_***= *_***= 0-= ++2,_++2=

φ,$9`=5,$9 == .*2,_,+3= .1_3*0= *_***= *_***= 21= ++-,_212=

φ,$9`=5,(9(8 == .*2,_-.0= .2_*--= *_***= *_***= 0.= ++2+_,31= φ,$9(8`=5,(9$8 == .*2,_-31= .2_*2.= *_***= *_***= 1.= ++0*_..*=

φ,$9(8`=5,(9 == .*2,_130= .2_.2-= *_***= *_***= 0.= ++2+_1.1=

φ,$9`=5,$9(8 == .*2,_3*-= .2_/3*= *_***= *_***= 22= ++-+_..0= φ,$9(8`=5,$9 == .*2-_211= .3_/0.= *_***= *_***= 22= ++-,_.,*=

φ,$9(8`=5,(9(8 == .*2-_3+1= .3_0*.= *_***= *_***= 0/= ++2*_12-=

φ,$9(8`=5,$9(8 == .*2._1/1= /*_...= *_***= *_***= 23= ++-+_+2,=

φ,$9`=57 == .++1_020= 2-_-1-= *_***= *_***= .2= +,.3_2+,=

φ,$9`=57(8 == .++3_*30= 2._12-= *_***= *_***= .3= +,.3_+/2=

φ,$9$8`=5,(9 == .++3_--.= 2/_*,+= *_***= *_***= +*+= ++.*_,.-=

φ,$9(8`=57 == .++3_-.-= 2/_*-*= *_***= *_***= .3= +,.3_.*.=

φ,$9$8`=5,$9 == .++3_/.*= 2/_,,1= *_***= *_***= +,.= +*3*_32*=

φ,$9(8`=57(8 == .+,*_0,*= 20_-*1= *_***= *_***= /*= +,.2_0+0=

φ,$9$8`=5,(9(8 == .+,*_1*0= 20_-3-= *_***= *_***= +*,= ++-3_.2*=

φ,$9$8`=5,$9(8 == .+,*_232= 20_/2/= *_***= *_***= +,/= +*3*_+1*=

φ,$9`=57$8 == .+,+_*/3= 20_1.0= *_***= *_***= /*= +,.3_*//=

φ,$9(8`=57$8 == .+,,_/12= 22_,0/= *_***= *_***= /+= +,.2_/*3=

φ,$9$8`=5,(9$8 == .+,-_-30= 23_*2-= *_***= *_***= ++-= +++2_/22=

φ,$9`=5,$9$8 == .+,1_-1/= 3-_*0,= *_***= *_***= +--= +*13_,/-= φ,$9(8`=5,$9$8 == .+,3_*.+= 3._1,2= *_***= *_***= +-.= +*12_1-3=

φ,$9$8`=57 == .+00_/.2= +-,_,-/= *_***= *_***= 3+= +,*2_1-/=

φ,$9$8`=57(8 == .+02_*-.= +--_1,+= *_***= *_***= 3,= +,*2_+**= φ,$9$8`=57$8 == .+1*_+*3= +-/_130= *_***= *_***= 3-= +,*2_*/+=

φ,$9$8`=5,$9$8 == .+11_/,3= +.-_,+0= *_***= *_***= +1.= +*-2_2*,= POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 21

= ._= 89.2&9*8= 4+= &55&7*39= &):19= 8:7;.;&1= +742= 0.831 (SE = 0.012), with a spatial process stan- 24)*1= φ,`=5,$9 =+47=&1.+473.&=5499*)=<18=43=+.;*= dard deviation = 0.0212 (95% CI = 0.0000 to 89:)>=&7*&8=.3=&1.+473.&_= 0.0752) among the apparent survival estimates. 9:)>=&7*&= φ =φ = 3/= = Estimates of temporal variation in apparent survival were computed for each of the ﬁ ve study !"= *_2+/= *_*,*= *_11,:*_2/+= !)= *_2,3= *_*+/= *_132:*_2/1= areas from model {φg*t, pg*t} (Table 6). On the )*= *_2+-= *_*+/= *_12,:*_2.+= basis of coeﬃ cients of process variation (CV), = *_2+2= *_*+1= *_12+:*_2/*= both spatial process variation (CV = 0.025) and ,= *_211= *_*+0= *_2.,:*_3*/= temporal variation (CVs = 0.011 to 0.033; Table ''7*;.&9.438a==-=1)47&)4=89:)>=&7*&`==-=&88*3=89:)>=&7*&`= 6) were relatively low. =-=&3=*73&7).34=89:)>=&7*&`==-=.*77&=89:)>=&7*&`=&3)==-= *6:4.&=&3)=.3,8=&3>43=3&9.43&1=5&708=89:)>=&7*&_= Fecundity the Akaike weights. The second-ranked model We analyzed fecundity data from 1,557 re-

{φSAB,SKC vs. rest, pg*t} suggested that both the SAB productive outcomes for adult female Spott ed and SKC had separate estimates of φ from each Owls on all ﬁ ve study areas (see Table 7 for other as well as from the other three study areas study area sample sizes). There was uncertainty combined. However, that model had a nearly in selecting a smooth trend model for fecundity identical deviance to the top-ranked model and in all of the study areas (Table 7). In all study the ∆QAICc ≈ 2 was a result of adding another areas, multiple highly ranked models had parameter for the SAB that did not explain any similar Akaike weights, especially for the ELD, additional variation in the data; the top two SAB, and SIE study areas. In the case of four of models essentially explained the same amount the study areas (ELD, LAS, SAB, and SKC), the of variation. Thus, we concluded that there uncertainty in selecting an appropriate model was strong evidence for model {φSKC vs. rest, pg*t}. for fecundity suggested diﬀ erent trends. For The annual recapture probabilities from model example, an intercept-only model was almost as

{φSKC vs. rest, pg*t} were generally high, with 87% of likely as a model suggesting a linear trend, and the annual estimates >0.8 and 54% of the annual an intercept-only model was almost as likely estimates >0.9 (Table 5). The combination of as a model with even–odd year variation for high survival probabilities coupled with high the SAB. However, models that were similarly recapture probabilities reduced any bias that likely for the SIE study area all included a linear may be caused by assumption violations, such trend, which suggests that a linear trend may be as heterogeneous capture probabilities (Pollock supported for that study area. et al. 1990). The issue of model uncertainty was further Using a random eﬀ ects model with estimates illustrated by the parameter estimates for the ef- from model {φg, pg*t}, mean apparent survival fects of interest (Table 8). Two of the study areas across the study areas was estimated to be = (ELD and SAB) had estimates that were not

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a==-=1)47&)4=89:)>=&7*&`==-=&88*3=89:)>=&7*&`==-=&3=*73&7).34=89:)>=&7*&`==-=.*77&=89:)>=&7*&`=&3)= =-=*6:4.&=&3)=.3,8=&3>43=3&9.43&1=5&708=89:)>=&7*&_= 22 ORNITHOLOGICAL MONOGRAPHS NO. 54

= 0_= 89.2&9*8= 4+= 2*&3= &55&7*39= 8:7;.;&1 φ &3)= diff erent from zero in the top-ranked mod- 9*2547&1= 574(*88= 89&3)&7)= )*;.&9.43 σ .3= els, based on 95% CI. Parameter estimates for &55&7*39= 8:7;.;&1= +47= 9-*= &1.+473.&= 5499*)= <1= 43= the fi xed eff ects were diff erent from zero for +.;*= 89:)>= &7*&8= .3= &1.+473.&_= 89.2&9*8= <*7*= '&8*)= the LAS, SIE, and SKC study areas; and there 43= ;&7.&3(*= (42543*398= &3&1>8*8= +742= &33:&1= was evidence of a negative linear trend in *89.2&9*8=4+ φ +742=24)*1= φ,$9`=5,$9 _= = fecundity on the SIE study area. 9:)>= = = 3/="= Although fecundity estimates were not & =&7*&== φ =φ = σ =# 9*2547&1 ==σ === comparable across study areas (except for SIE and SKC), we felt that estimates of tem- $%= *_2,-= *_*+3= *_**3+= *_*++= *_****:*_+-,/= $,= *_2-1= *_*+1= *_*,1+= *_*-,= *_****:*_*3.-= poral process standard deviations and their ,/= *_2+.= *_*+/= *_****= *_***= *_****:*_*.//= coeffi cients of variation were comparable, as- "= *_2,.= *_*+1= *_****= *_***= *_****:*_*/11= suming that biases within study areas were 0= *_23+= *_*+3= *_*,3-= *_*--= *_****:*_+**/= constant among years. Except for the SAB study area, there was considerable temporal σ &=89.2&9*)=&8a== = φ variation in fecundity in four study areas ''7*;.&9.438a==-=1)47&)4=89:)>=&7*&`= =-=&88*3=89:)>=&7*&`= =-= with coeffi cients of process variation ranging &3= *73&7).34= 89:)>= &7*&`= = -= .*77&= 89:)>= &7*&`= &3)= = -= *6:4.&= &3)= from 67.2% to 81.7% (Table 9, Fig. 2). The SAB .3,8=&3>43=3&9.43&1=5&708=89:)>=&7*&_= study area had much lower annual variation

TABLE 7. Ranking and weighting of mixed models used to fit trends in adult fecundity for California Spotted Owls from five study areas in California.

a Model –2loge K AICc 'AICc Akaike weight Eldorado study area (ARH1 covariance structure; n = 252b) T 219.3 16 253.6 0.00 0.28 Intercept 221.9 15 253.9 0.32 0.24 T + EO 217.4 17 254.0 0.40 0.23 EO 220.5 16 254.8 1.20 0.16 TT 219.3 17 255.9 2.30 0.09 Lassen study area (ARH1 covariance structure; n = 418) EO 379.7 16 413.0 0.00 0.44 T + EO 378.8 17 414.3 1.27 0.23 Intercept 383.4 15 414.5 1.54 0.20 T 382.8 16 416.1 3.10 0.09 TT 382.4 17 417.9 4.87 0.04 San Bernardino study area (AR1 covariance structure; n = 323) EO 389.4 6 401.6 0.00 0.31 Intercept 391.6 5 401.8 0.12 0.29 T + EO 388.2 7 402.6 0.89 0.20 T 391.2 6 403.5 1.80 0.13 TT 390.0 7 404.4 2.69 0.08 Sierra study area (ARH1 covariance structure; n = 312) T 178.3 16 212.1 0.00 0.39 TT 176.7 17 212.8 0.64 0.28 T + EO 177.1 17 213.2 1.04 0.23 Intercept 183.9 15 215.5 3.38 0.07 EO 183.3 16 217.1 5.00 0.03 Sequoia and Kings Canyon study area (EXP[YR] covariance structure; n = 252) Intercept 190.7 13 218.3 0.00 0.47 EO 189.8 14 219.6 1.35 0.24 T 190.7 14 220.5 2.25 0.15 T + EO 189.7 15 221.8 3.53 0.08 TT 190.0 15 222.1 3.83 0.07 a Number of estimable covariance and fixed-effect parameters. b Number of reproductive outcomes for adult females. Abbreviations: T = linear time trend, TT = quadratic time trend, EO = even–odd years, T + EO = linear time trend with an additive even–odd year effect. POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 23

= 2_= 89.2&9*8= 4+= +.=*)= *++*(9= 5&7&2*9*78= +47= 9-*= 94587&30*)= 2.=*)= 24)*1= :8*)= .3= *89.2&9.3,= 97*3)8=+47=&):19=+*(:3).9>=4+=&1.+473.&=5499*)=<18=43=+.;*=89:)>=&7*&8=.3=&1.+473.&_= = 9:)>=&7*&= ++*(9======β = = = 3/=$= %&= = :*_*.,= =*_*,.= *_/2/= :*_*3*=94=:*_**0= %,= = *_,30= =*_+.+= *_.1/= :*_*,*=94=:*_/1+= ,/= = *_+++= =*_*1+= *_0.1= :*_*-*=94=:*_,/+= $= = :*_*.3= =*_*+2= *_-1*= :*_*2/=94=:*_*+-= 1= $39*7(*59= *_,23= =*_*1.= *_,/0= :*_+..=94=:*_.-.= &=89.2&9*8=4+=.39*7(*598=<*7*=*_001==-=*_+1*=+47=9-*=`=*_+10==-=*_+*-=+47=9-*=`=*_-*2==-=*_*.3=+47=9-*="`=&3)=*_/2+= =-=*_+,0=+47=9-*=&_= ''7*;.&9.438a==-=1)47&)4=89:)>=&7*&`==-=&88*3=89:)>=&7*&`="=-=&3="*73&7).34=89:)>=&7*&`=&=-=.*77&=89:)>=&7*&`=&3)= -.=-=*6:4.&=&3)=-.3,8=.&3>43=3&9.43&1=5&708=89:)>=&7*&_=

=3_=89.2&9*8=4+=2*&3=+*(:3).9> &(7488=>*&78`= =+*_=*8(7.59.438=4+=)*38.9>=89:)>=&7*&=5479.438=

9*2547&1= 574(*88= 89&3)&7)= )*;.&9.43 σ `= 4+= 949&1= 89:)>= &7*&8= &3)= 9.2*= 5*7.4)8= :8*)= 94= &3)=&24:39=4+=574(*88=;&7.&9.43=*=51&.3*)='>=9-*= *89.2&9*= λ9= +47= +*2&1*= &3)= 2&1*= 9*77.947.&1= 2.=*)= 24)*1= +47= 97*3)8= .3= +*(:3).9>= 8*1*(9*)= 43= &1.+473.&=5499*)=<18=.3=&1.+473.&_= = 9-*= '&8.8= 4+= 2.3.2:2= (= +47= &):19= &1.+473.&= 5499*)= <18=43=+.;*=89:)>=&7*&8=.3=&1.+473.&_== = = 5574=.2&9*== = 9:)>=&7*&= 9:)>=5*7.4)= 8.?*=02,======!*7(*39&,*=4+= !"= +33*:,***= -//= 9:)>= = = = = ;&7.&9.43== "= +33,:,***= +`,1*= ( ) &7*&= ==σ =#= *=51&.3*) = *= +33+:+332= ,`+.*= $%= *_.*3= *_*21= *_,1/= *_01,= ,._,= -!= +33*:,***= .+3= $= *_--0= *_*2-= *_,01= *_13/= -*_/= .= +33+:,***= -.-= .= *_-0,= *_*-2= *_*12= *_,+1= ,3_3= ''7*;.&9.438a=8**=&'1*=2_= = *_,2.= *_*1-= *_,-,= *_2+1= .._.= 0= *_,23= *_*1.= *_,,,= *_101= :=*= Rates of Population Change &=*&3=*89.2&9*=&(7488=>*&78='&8*)=43=.39*7(*5988431>=2.=*)=24)*1_= '=9&3)&7)=)*;.&9.43=4+=9*2547&1=574(*88=;&7.&9.43_= We used 1,019 marked subadult and adult σ (=89.2&9*)=&8= _= individuals in the analysis of λt (144 from the ELD, 191 from the LAS, 401 from the SAB, 163 )= *7(*39&,*= 4+= 9-*= 9*2547&1= ;&7.&9.43= *=51&.3*)= '>= 9-*= +.=*)8*++*(9= from the SIE, and 120 from the SKC). Those 97*3)=24)*1=8*1*(9*)=:8.3,=2.3.2:2= (`=(&1(:1&9*)=&8a= , , σ − σ numbers diﬀ ered from those used to estimate = adult apparent survival because of inclusion of σ , <-*7* σ .8= 9-*= ;&7.&9.43= 7*2&.3.3,= +742= 9-&9= *=51&.3*)= '>= 9-*= the subadult age-classes, changes in study area 8*1*(9*)=24)*1_= sizes, and time periods examined. *=#49=&551.(&'1*='*(&:8*=8*1*(9*)=24)*1=<&8=&3=.39*7(*598431>=24)*1_= ''7*;.&9.438a=8**=$&'1*=2_= Individual study areas.—Study area sizes and

time periods used in estimating λt are shown in fecundity than the other study areas. Except in Table 10. Overdispersion was evident in the for the SIE, the percent variation explained by data for all study areas except SKC (Table 11). the top-ranked trend models for the study areas We corrected for that in subsequent analyses. was low, indicating that those models were not Two of the study areas, the ELD and SKC, had very useful for explaining temporal variation in strong evidence for trends in λt. The lowest fecundity (Table 9, Fig. 2). Those three indica- QAICc model for the ELD suggested a linear tors (model selection uncertainty, imprecise pa- trend in λt (Table 12, Fig. 3). The second-ranked rameter estimation, and low amounts of process model was also a linear trend model but with variation explained) suggested that there was the fi rst estimable λt eliminated. Together, those litt le support for a smooth trend in fecundity in two models accounted for 70% of the Akaike four of the fi ve study areas. The single excep- weights. The estimated slope for the linear trend tion was the SIE study area, where there was in λt on the ELD was negative and diff erent from some evidence for a negative linear trend in zero on the basis of 95% CI (Table 13). The best fecundity. However, the top-ranked model that model for the SKC suggested a quadratic trend suggested that trend still only explained 44% of in λt with the fi rst estimable λt eliminated (Table the temporal process variation in fecundity. 12, Fig. 3). The second-ranked model was also a 24 ORNITHOLOGICAL MONOGRAPHS NO. 54

Fig. 2. Annual estimates of fecundity of California Spotted Owls on five study areas in California. Annual estimates (dots) and 95% confidence intervals (bars) are least-squares means from mixed intercept-only model, whereas dashed lines represent estimates from the top-ranked mixed trend model. Abbreviations: ELD = Eldorado study area, LAS = Lassen study area, SAB = San Bernardino study area, SIE = Sierra study area, and SKC = Sequoia and Kings Canyon national parks study area. POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 25

=++_=89.2&9*8=4+=4;*7).85*78.43=M=+47=(&59:7*: quadratic trend model but included the fi rst esti- 7*(&59:7*= )&9&= 8*98= :8*)= 94= *89.2&9*= λ = +742= +.;*= 9 mable λ . Those two models accounted for 45.5% &1.+473.&= 5499*)= <1= 89:)>= &7*&8= .3= &1.+473.&_= t of the Akaike weights. Slope parameters for that &1:*8=+47=(-.886:&7*=χ,`=)*,7**8=4+=+7**)42=)+`= &3)=574'&'.1.9>==&7*=+742==,=&3)==-=.3= quadratic trend were diff erent from zero on the 574,7&2='()_= basis of 95% CI (Table 13). The best model for = the SIE study area also suggested a negative lin- 9:)>=&7*&= χ,=)+= ======M= ear trend in λt (Table 12, Fig. 3), but the second- (*= /+_-.= -2= *_*1,*= +_-/,0= ranked model suggested a quadratic trend. In ()= //_*2= -/= *_*+01= +_/1-0= addition, the estimated slope for the linear trend )1= .+_/2= -,= *_++30= +_,33.= was not diff erent from zero (Table 13), although 3= .1_+3= -,= *_*.*2= +_.1.0= 95% CI barely overlapped 1.0. There was a weak 4= +._//= ,1= *_31/-= *_/-22&= negative linear trend for the SAB study area &=*9==94=+_*=+47=8:'8*6:*39=&3&1>8*8_= ''7*;.&9.438a==-=1)47&)4=89:)>=&7*&`==-=&88*3=89:)>=&7*&`= (Table 12, Fig. 3). However, there was uncer- =-=&3= *73&7).34=89:)>=&7*&`= =-=.*77&=89:)>=&7*&`=&3)==-= tainty as to whether a linear or means model *6:4.&=&3)=.3,8=&3>43=3&9.43&1=5&708=89:)>=&7*&_= best explained the data because Akaike weights

Fig. 3. Trends in λt for California Spotted Owls on five study areas in California. Trend lines are from random effects models selected by minimum QAICc. Dots, with 95% confidence intervals, are annual estimates of λt from model {φt, pt, λt} used to develop the random effects models. Abbreviations: see Table 11. 26 ORNITHOLOGICAL MONOGRAPHS NO. 54

=+,_=4)*1=8*1*(9.43=7*8:198=+742=λ9=&3&1>8*8=+47=&1.+473.&=5499*)=<18=43=+.;*=89:)>=&7*&8=.3= &1.+473.&_= =

4)*1= (= = *;.&3(*= ∆(= 0&.0*=<*.,-9= 1)47&)4=89:)>=&7*&== #.3*&7=7&3)42=*++*(98= +*./_,3/= ,-_**= 1*+_.*3= *_***= *_/*0= #.3*&7=7&3)42=*++*(98&= +*.1_,++= ,._**= 1*+_++0= +_3+0= *_+3.= :&)7&9.(=7&3)42=*++*(98= +*.1_,0.= ,._**= 1*+_+1*= +_303= *_+23= :&)7&9.(=7&3)42=*++*(98&= +*.3_.*,= ,/_**= 1*+_*23= ._+*1= *_*0/= *&3=7&3)42=*++*(98= +*/+_.0+= ,0_/-= 033_1.-= 0_+00= *_*,-= *&3=7&3)42=*++*(98&= +*/+_110= ,0_--= 1**_/*2= 0_.2+= *_*,*= .2*885*(.+.(=+.=*)=*++*(98= +*//_3.*= ,3_**= 032_0/3= +*_0./= *_**,= .2*=×=8*==+.=*)=*++*(98= ++*0_,1,= /3_**= 010_.20= 0*_311= *_***= &88*3=89:)>=&7*&== *&3=7&3)42=*++*(98= +*.,_*./= +3_**= 0,0_,23= *_***= *_.3.= #.3*&7=7&3)42=*++*(98= +*.._+03= ,*_**= 0,0_,11= ,_+,.= *_+1+= *&3=7&3)42=*++*(98&= +*.._+1+= ,*_**= 0,0_,13= ,_+,0= *_+1+= :&)7&9.(=7&3)42=*++*(98= +*./_/3/= ,+_**= 0,/_/0+= -_//*= *_*2.= #.3*&7=7&3)42=*++*(98&= +*.1_,,.= ,+_32= 0,/_*2,= /_+13= *_*-1= :&)7&9.(=7&3)42=*++*(98&= +*.1_,.-= ,,_**= 0,/_*/3= /_+32= *_*-1= .2*885*(.+.(=+.=*)=*++*(98= +*/*_/..= ,._**= 0,._*-3= 2_.33= *_**1= .2*=×=8*==+.=*)=*++*(98= +*20_/3/= .1_**= 0*2_,-*= .._//*= *_***= &3=*73&7).34=89:)>=&7*&== #.3*&7=7&3)42=*++*(98= ,,.-_333= +0_**= ++*+_/,2= *_***= *_-3/= *&3=7&3)42=*++*(98&= ,,./_*12= +0_**= ++*,_0*0= +_*13= *_,-*= #.3*&7=7&3)42=*++*(98&= ,,.0_*/2= +1_**= ++*+_/,-= ,_*/3= *_+.+= *&3=7&3)42=*++*(98= ,,.0_/*.= +0_**= ++*._*--= ,_/*/= *_++-= :&)7&9.(=7&3)42=*++*(98&= ,,.2_++2= +2_**= ++*+_/+0= ._++3= *_*/*= :&)7&9.(=7&3)42=*++*(98= ,,.2_++3= +2_**= ++*+_/+1= ._+,*= *_*/*= .2*885*(.+.(=+.=*)=*++*(98= ,,.3_3,0= +3_**= ++*+_,/-= /_3,1= *_*,*= .2*=×=8*==+.=*)=*++*(98= ,,2*_.--= .*_**= +*21_-0,= -0_.-.= *_***= .*77&=89:)>=&7*&== #.3*&7=7&3)42=*++*(98= +*.._,00= ,-_**= 013_./0= *_***= *_-/2= :&)7&9.(=7&3)42=*++*(98= +*./_-0+= ,._**= 012_-10= +_*3/= *_,*1= #.3*&7=7&3)42=*++*(98&= +*.0_,+/= ,._**= 013_,-*= +_3.3= *_+-/= *&3=7&3)42=*++*(98&= +*.0_,02= ,-_32= 013_--,= ,_**,= *_+-,= *&3=7&3)42=*++*(98= +*.0_3,2= ,._1/= 012_-*,= ,_00,= *_*3/= :&)7&9.(=7&3)42=*++*(98&= +*.1_/*+= ,/_**= 012_--,= -_,-/= *_*1+= .2*885*(.+.(=+.=*)=*++*(98= +*/._*1+= ,3_**= 010_*3*= 3_2*/= *_**-= .2*=×=8*==+.=*)=*++*(98= +*31_,22= /2_**= 0/+_-3,= /-_*,,= *_***= *6:4.&=&3)=.3,8=&3>43=89:)>=&7*&='= :&)7&9.(=7&3)42=*++*(98&= +**0_/--= ,._*-= 0,*_.1+= *_***= *_,0,= :&)7&9.(=7&3)42=*++*(98= +**1_+.+= ,._-0= 0,*_-0,= *_0*2= *_+3-= #.3*&7=7&3)42=*++*(98= +**1_012= ,._3+= 0+3_02.= +_+./= *_+.2= *&3=7&3)42=*++*(98= +**2_**.= ,/_*-= 0+3_1/.= +_.1+= *_+,/= *&3=7&3)42=*++*(98&= +**2_,/2= ,/_*3= 0+3_2/3= +_1,/= *_++*= #.3*&7=7&3)42=*++*(98&= +**2_.+,= ,/_-+= 0+3_/-*= +_213= *_+*,= .2*885*(.+.(=+.=*)=*++*(98= +**3_.33= ,0_**= 0+3_*30= ,_300= *_*/3= .2*=×=8*==+.=*)=*++*(98= +*.,_.1-= /,_**= /3+_.1+= -/_3.*= *_***=

& =.789=*89.2&'1*=1&2')&=+742=24)*1= 9`=59`= 9 =<&8=42.99*)=+742=7&3)42=*++*(98=24)*1_= '= 89.2&9*8='&8*)=43=(=7&9-*7=9-&3=(='*(&:8*=34=4;*7).85*78.43=<&8=*;.)*39_=

were similar (Table 12) and the estimated slope Estimates of realized change (∆t) represented for the linear model was not diﬀ erent from zero the trajectory (or trend in numbers) of each

(Table 13). There was no evidence of a trend in λt study population (Fig. 4). Those estimates were for the LAS study area (Table 12, Fig. 3); a means based solely on the estimates of λt and did not model best supported the data for that area. require estimating annual population size (Nt) POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 27

= +-_= &7&2*9*7= *89.2&9*8= +47= '*89= 7&3)42= *++*(98= 24)*1= 4+= λ9= +47= &1.+473.&= 5499*)=

<18=43=+.;*=89:)>=&7*&8=.3=&1.+473.&_=*89=24)*18=<*7*=8*1*(9*)=:8.3,=9-*=14<*89= !"(_= = = = ======#4)*1=5&7&2*9*78= 9:)>=&7*&= *89=24)*1= &7&2*9*7===%89.2&9*= %= 3/="=

%)*= ).3*&7= β*= =+_,2*= *_*32= =+_*23== 94=:+_.1,=

== β+= :*_*//= *_*,+= :*_*3.= 94=:*_*+/=

)!= #*&3= β*= =*_32/= *_*,0= =*_3-.= 94=:+_*-0= != ).3*&7= β*= =+_*/.= *_*12= =*_3*+= 94== +_,*1=

== β+= :*_*,3= *_*,2= :*_*2.= 94=:*_*,0=

"%= ).3*&7= β*= =+_*..= *_*00= =*_3+/= 94=:+_+1.= == β+= :*_*,*= *_*+/= :*_*/*= 94=:*_**3=

2= :&)7&9.(= β*= =+_.-3= *_+01= =+_++,= 94=:+_10/=

== β+= :*_,1/= *_*3.= :*_./3= 94=:*_*3+=

== β,= =*_*-.= *_*+,= =*_*++= 94=:*_*/2= ''7*;.&9.438a==-=1)47&)4=89:)>=&7*&`==-=&88*3=89:)>=&7*&`==-=&3=*73&7).34=89:)>=&7*&`==-=.*77&=89:)>= &7*&`=&3)==-=*6:4.&=&3)=.3,8=&3>43=3&9.43&1=5&708=89:)>=&7*&_= for each study area. Trends in Figure 4 repre- were <1 for four of the study areas (LAS, SAB, sent the proportion of the population remaining SIE, and SKC). The SIE had the lowest estimate each year, given the initial population in the of with 95% CI that barely overlapped λ = 1, ﬁ rst year. For example, if there were 100 owls which suggests that the owl population on the on the SIE study area in 1993, there were 71 SIE was declining. Temporal process variation owls in 1999, based on the estimates of λt and, in λt was low (CV for ELD = 8.4%, for SIE = 1.5%, hence, ∆t for that study area. Populations on the for SKC = 10.1%, and 0% for LAS and SAB). The

LAS, SAB, and SKC study areas remained fairly model with the negative linear trend in λt for stationary over the course of the study. The ELD the ELD accounted for an estimated 100% of the population increased and then decreased to a temporal process variation in λt, and the model level that was higher, but not diff erent than the with the quadratic trend for the SKC accounted initial population, on the basis of 95% CI. The for 75.7% of the temporal process variation in λt, SIE population decreased steadily from 1993, which suggests that those were useful models resulting in a population that was substantially for explaining the temporal process variation in lower and diff erent from the initial population, λt (Franklin et al. 2001). on the basis of 95% CI. In 1999, 70.9% of the SIE Meta-analysis across study areas.—There was population remained from the initial popula- strong support for an additive study area eff ect tion in 1993 (Fig. 4). over time (model λg+t; Table 15); model λg+t in- On the basis of a random eff ects intercepts- cluded an additive eff ect and was weighted only model, using annual estimates of λ from much more heavily (99.4% of the Akaike model {φt, pt, λt}, 95% CI of the estimated mean λ weights) than models with a study area × year ( ) across years for each of the study areas was interaction, just a year eff ect, or just a study- not diff erent from a stationary population (λ = area eff ect. That suggested λt changed simi- 1; Table 14, Fig. 5), although point estimates larly among the four Sierran study areas, even

=+._=89.2&9*8=4+=2*&3 λ λ &(7488=9.2*=&3)=9*2547&1=574(*88=89&3)&7)=)*;.&9.43 σ =+47= &1.+473.&=5499*)=<18=43=+.;*=89:)>=&7*&8=.3=&1.+473.&_=89.2&9*8=&7*='&8*)=43=2*&38=.39*7(*598 431>=7&3)42=*++*(98=24)*18=:8.3,=9.2*885*(.+.(=*89.2&9*8=4+=`=5`=&3)=λ_= = 9:)>=&7*&= λ =λ = 3/=&=+47= λ = σ = 3/=&=+47= σ = '(= +_*.,= *_*.1= *_3/*=94=+_+--= *_*22= *_***=94=*_-,-= '.= *_32/= *_*,0= *_3-.=94=+_*-0= *_***= *_***=94=*_,.+= .0= *_312= *_*,/= *_3,3=94=+_*,0= *_***= *_***=94=*_+/1= &= *_30+= *_*,.= *_3+/=94=+_**2= *_*+/= *_***=94=*_+23= 1= *_32.= *_*.1= *_23,=94=+_*10= *_+**= *_**+=94=*_-+-= ''7*;.&9.438a=8**=&'1*=+-_= 28 ORNITHOLOGICAL MONOGRAPHS NO. 54

Fig. 4. Trends in populations of California Spotted Owls from five study areas in California. Trends are expressed as realized change (∆t) based on estimates of λt, which represent the proportion of the initial population remaining for each year. Bars around estimates are 95% confidence intervals. Note that the graph for the ELD study area is scaled differently than the other study areas. Abbreviations: ELD = Eldorado study area, LAS = Lassen study area, SAB = San Bernardino study area, SIE = Sierra study area, and SKC = Sequoia and Kings Canyon national parks study area. POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 29

though there were diﬀ erences in magnitude of

the λt (Fig. 6). In the best model, study area effects (expressed as the diﬀ erence from SIE on a logit scale) were 0.036 (95% CI = –0.030 to 0.101) for the SKC, 0.083 (95% CI = –0.070 to 0.173) for the ELD, and 0.050 (95% CI = –0.031 to 0.132) for the LAS, which suggests that the ELD was slightly higher than the other study areas in terms of magnitude of during the time period (1994–1999) examined. Because we were interested in temporal covariation, we did not consider any smooth time trend models as we did with the individual study areas. On the basis of the results for the individual study areas, Fig. 5. Estimates of mean λt with 95% confidence intervals for California Spotted Owls on five study areas three of the studies had negative linear trends. in California. Abbreviations: ELD = Eldorado study However, the additive eﬀ ect of time suggests area, LAS = Lassen study area, SAB = San Bernardino that the study areas tracked each other in terms study area, SIE = Sierra study area, and SKC = Sequoia of changes in λt over time. and Kings Canyon national parks study area.

Fig. 6. Trends in λt based on model {λg+t} from meta-analysis on California Spotted Owls from four study areas (ELD, LAS, SIE, and SKC) in the Sierra Nevada, California. Abbreviations: see Fig. 5.

= +/_= 4)*1= 8*1*(9.43= 7*8:198= &)/:89*)= <.9-= M= -= +_,1-3= +47= 2*9&8&3&1>8.8= 4+= λ9= +47= !&1.+473.&= "5499*)=$<18=43=+4:7=89:)>=&7*&8=%&'`=&)"`="*%`=&3)="+!=.3=!&1.+473.&_= =

4)*1= = '*;.&3(*= )*!(= ∆)*!(= )0&.0*=<*.,-9=

φ,$9`=5,$9`=λ,(9= 1-= -00*_20= 0..0_,3= ====*_**= *_33.=

φ,$9`=5,$9`=λ9= 1/= -000_2+= 0./0_-3= ==+*_+*= *_**0= φ,$9`=5,$9`=λ,$9= 3,= -0.+_1+= 0.00_01= ==,*_-2= *_***=

φ,$9`=5,$9`=λ,= 1,= -1*1_1,= 0.3+_*2= ==.._13= *_***=

φ,`=5,`=λ,= +,= .*21_./= 01.2_-2= -*,_*3= *_***= ''7*;.&9.438a=8**=.,_=/_= 30 ORNITHOLOGICAL MONOGRAPHS NO. 54

Using the estimates from model {φg*t, pg*t, λg*t}, In general, population studies of California we used random eff ects intercepts-only models Spott ed Owls began later than those of to estimate for each Sierran study area for the Northern Spott ed Owls because there was a common time period, 1994–1999. In general, perceived diff erence in habitat use patt erns estimates were lower than those computed in and prior anthropogenic disturbance of habitat Table 14: 0.942 (95% CI = 0.878 to 1.005) for the (Verner et al. 1992b). The fi rst study to estimate SIE, 0.958 (95% CI = 0.866 to 1.049) for the SKC, population characteristics and assess Spott ed 1.001 (95% CI = 0.913 to 1.090) for the ELD, and Owl population trends began in 1986 in the 0.985 (95% CI = 0.927 to 1.042) for the LAS. central Sierra Nevada on the Eldorado National Forest (Noon et al. 1992, Seamans et al. 2001a). Comparison of Sierra and Sequoia and Kings The ELD study was followed in 1987 by a study Canyon National Parks Study Areas (SAB) in the San Bernardino Mountains of southern California. The SAB study area con- On the basis of the 95% CI for the eff ect sizes, tained the largest population of Spott ed Owls only adult survival was diff erent between the in southern California (LaHaye et al. 1994). A two study areas (95% CI for the diff erence did third study on a small population occurred not overlap zero) with SKC having higher adult between 1988 and 1992 on Mount San Jacinto, survival than SIE (Table 16). However, point southern California (Gutiérrez and Pritchard estimates for adult fecundity and were also 1990). That latt er study was terminated because higher for the SKC than for the SIE study area. of lack of funding. Finally, three studies were In addition, population trends in the two com- begun in 1990 in the northern (Lassen National panion studies (SIE representing lands managed Forest, LAS) and southern Sierra Nevada for timber production and SKC representing (Sierra National Forest, SIE, and Sequoia and National Park Service lands) had diff erent Kings Canyon national parks, SKC). trajectories in ; the SIE had a negative linear Our studies were well suited to a meta- trend in , whereas the SKC had a quadratic analysis because the techniques were very trend over the same time period; initially similar among studies. Our ability to detect the declined and then increased on the SKC. That owls with vocal lures and color band them fa- resulted in a decline in the population on the cilitated a strong mark–recapture study design. SIE but not on the SKC (Fig. 4). However, there However, Milligan et al. (2003) suggested that was less strength of evidence for the linear trend observers misread band combinations, with 5% in on the SIE than for the quadratic trend in and 16% error rates for trained and untrained on the SKC. observers, respectively. We did not believe such high error rates pertained to our studies Discussion for several reasons. First, part of the high error rate observed by Milligan et al. (2003) was The Spott ed Owl has been the focus of both from switching color combinations between extensive and intensive population studies left and right legs. In our study, birds carried for nearly two decades (Gutiérrez et al. 1995). color bands on only one leg; the other leg was

= +0_= 425&7.843= 4+= *++*(98= .3= )*24,7&5-.(= 5&7&2*9*78= +47= &1.+473.&= 5499*)= <18= '*9<**3= 9-*= .*77&= 89:)>=&7*&= =&3)=*6:4.&=&3)=".3,8=&3>43=3&9.43&1=5&708=89:)>=&7*&="=.3=&1.+473.&_======"=89:)>=&7*&= ====== =89:)>=&7*&= = ======++*(9=8.?*= &&7&2*9*7= 89.2&9*= = 89.2&9*= = 89.2&9*&= '= 3/== *):19=+*(:3).9>(= *_,23= *_*1.= *_,2.= *_*1-= *_**/= *_+*.= :*_+33=94=*_,*3= *):19=8:7;.;&1=)= *_211= *_*+0= *_2+2= *_*+1= *_*/3= *_*,-= ==*_*+-=94=*_+*/= λ = *_32.= *_*.1= *_30+= *_*,.= *_*,-= *_*/-= :*_*2*=94=*_+,0= &=89.2&9*=+47= =:=*89.2&9*=+47= _= ) ) , , '= + = (=89.2&9*8=+742=&'1*=3_= )=89.2&9*8=+742=&'1*=._= *=89.2&9*8=+742=&'1*=+._= POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 31 banded with an aluminum federal band. Color conditions on the study areas may have been combinations were unique to each bird. Second, diff erent from those on the surrounding areas nonexistent color combinations that might have (i.e. study areas were of higher quality because erroneously been observed could be corrected there were more owls). Therefore, our infer- to the appropriate combination, or else discard- ences applied to the populations of owls within ed and the bird re-observed to obtain the correct the study areas that participated in the meta- combination. Finally, the geographic location of analysis. Further, those population studies did an observation oft en uniquely identifi ed the not encompass the range of the California sub- band combination, in contrast to the Milligan et species because a signifi cant population exists al. (2003) protocol, where multiple birds were in central-coastal California and other popula- observed at the same location. For example, if tions exist in southern California and northern observers, upon returning from the fi eld in our Baja California, Mexico (Gutiérrez et al. 1995). study, noted that a bird with an unexpected Nevertheless, the extant population studies band combination was recorded in a territory, spanned a major latitudinal gradient over the other observers would return to confi rm that range of that subspecies. Each of the fi ve study observation. Although it was possible that mis- areas had unique characteristics that capture taken band identifi cations occurred in our stud- much of the inherent environmental variation ies, we do not believe that the error rate was within the California Spott ed Owl range. nearly as high as that reported by Milligan et al. Therefore, in the following discussion, we (2003). Further, the high resighting probabilities fi rst present a set of general inferences regard- and survival rates estimated in our study sug- ing population characteristics and trends for the gested that the eff ect of an erroneous identi- California Spott ed Owl populations represented fi cation was not of major consequence in the in our meta-analysis. Those general inferences Cormack-Jolly-Seber models. That is, mistak- are followed by study- area-specifi c inferences, enly labelling bird A as bird B means that bird which capture some of the unique environmen- B will get a 1 in its encounter history, and bird tal characteristics or owl population dynamics A will be recorded as not being observed. With for the individual areas. In this latt er section, the high resighting probability, bird B will likely we include a discussion on the comparisons be- be observed anyway, given that it is alive, which tween the SIE and SKC, which represent paired is also likely because of the high survival rate. studies using the same fi eld techniques, but However, frequent occurrences of that scenario having diff erent management regimes. Lastly, would have lowered the estimated resighting we synthesize the information from the general probability. The high resighting probabilities and area-specifi c inferences and present recom- we reported here suggest that the hypothesized mendations for managers and for future meta- scenario of mistaken band identifi cations oc- analyses of the California Spott ed Owl. curred infrequently. Although the study areas covered a large lati- General Inferences tudinal gradient, results of our meta-analysis cannot be considered representative of owl Apparent survival.—Noon and Biles (1990) demographic trends throughout the Sierra and Blakesley et al. (2001) demonstrated that Nevada. If, at the inception of those studies, survival rate of breeding adults was a key de- habitat management on the study areas was ei- mographic parameter of Spott ed Owl popula- ther diff erent than that of the surrounding areas tion dynamics. Our best model for apparent or changed as a result of the study location (i.e. survival of adults indicated that rates were study areas were preferentially protected from similar among the ELD, LAS, SAB, and SIE management activities), then general inference study areas (0.819), but higher for the SKC beyond the study areas cannot be made. A study study area (0.877). The pooled estimate of φ for comparing habitat quality within the study ar- ELD, LAS, SAB, and SIE was comparable to esti- eas to habitat quality off the study areas in the mates of two Mexican Spott ed Owl populations Sierra Nevada could elucidate that question. (Seamans et al. 1999), but slightly lower than If initial placement of the demography stud- the mean estimate for adult Northern Spott ed ies was based on a history of owl occupancy Owls among 15 study areas (0.850; Franklin et (i.e. they were not randomly located), then the al. 1999); whereas apparent survival on the SKC 32 ORNITHOLOGICAL MONOGRAPHS NO. 54 was higher than average rates for Northern and groves promoted higher survival, the causative Mexican spott ed owls. Average adult apparent mechanism was unknown. survival estimates across all three subspecies of An important biological issue with open Spott ed Owl ranged from 0.8 to 0.9 (Gutiérrez capture–recapture models is that mortality and et al. 1995). emigration are confounded (Schwarz and Seber In our meta-analysis, temporal variation 1999). Therefore, those models provide esti- in adult apparent survival on all study areas mates of apparent survival, which is defi ned as was very low (CV = 0–3%). That inference the probability of surviving and staying within was expected for an animal with high adult the sampled population (Franklin et al. 2000). survival rates (e.g. Gaillard et al. 1998, Pfi ster However, territorial Spott ed Owls infrequently 1998). Estimates of temporal variation in adult switch territories, which indicates that perma- survival for Mexican Spott ed Owl populations nent emigration is low (Raphael et al. 1996). were slightly higher (Seamans et al. 2002; CV = For example, within the insular SAB, territorial 7–13%), but estimates from a study of Northern male and female owls dispersed an average of Spott ed Owls were comparable (Franklin et al. 2.95 and 4.28 km, respectively, from their initial 2000; CV = 4%). Weather explained >50% of the location. Breeding dispersal was detected in 7% temporal process variation in apparent survival of between-year observations (territory change; in Mexican Spott ed Owls (Seamans et al. 2002) n = 54 of 743 observations) in the LAS. The me- and all of the temporal process variation in a dian breeding dispersal distance within the LAS Northern Spott ed Owl population (Franklin et was 7 km (range 1–33, n = 54; Blakesley 2003). al. 2000). Thus, weather may have accounted for Because dispersal distances were short relative most of the temporal variability in Spott ed Owl to the size of the LAS, most dispersing owls survival. However, because our data did not probably remained in the study area. Further, support time-based apparent survival models we estimated emigration from the ELD density and because temporal process variation in ap- study area using observations of banded owls parent survival was very low, we suspected on the regional study area and documented that weather did not have a large eff ect on survival. 1.4% (5/354) of territorial owls moved from the Spatial process variation in adult survival density to the regional study area over a six- from this meta-analysis (CV = 3%) was similar year period. Thus, adult emigration was prob- to estimates from a meta-analysis on Northern ably low in those studies. Spott ed Owls (Franklin et al. 1999; CV = 2%). Fecundity.—Except for the SKC and SIE, we The small amount of spatial process variation did not directly compare fecundity estimates in our meta-analysis was largely due to diff er- among study areas because we were not sure ences between SKC and the other study areas. if diff erences in protocol were confounded with The four study areas with similar apparent sur- other diff erences among study areas. However, vival were on national forests (i.e. managed for because protocol was consistent within study multiple use) whereas the SKC was in national areas, we believed we could safely compare tem- parks (i.e. managed as a preserve). Therefore, poral process variation and general patt erns. diff erences in forest management were one Although our point estimates of mean fecun- possible cause of the diff erences in adult appar- dity for the California Spott ed Owl were within ent survival between SKC and the other Sierra the range of those for the Northern (Franklin et Nevada study areas (see individual study area al. 1999) and Mexican spott ed owls (Seamans et discussions for local habitat and management al. 2002), it was possible that diff erences were details). Another major diff erence between SKC the result of unique protocols among studies. and the other study areas was the presence of Except for the SIE, there was litt le support for giant sequoia groves on the SKC. Almost half a smooth trend in fecundity for any of the other of the historic roost sites of pairs using mixed- study areas. For Sierra Nevada studies, fecun- conifer forests in SKC were in, or within 1 km of dity varied substantially among years within all sequoia groves. In contrast, there was only one study areas (range in CV of temporal process sequoia grove on the SIE and one on the RSA variation = 67.2–81.7%); 1992 appeared to be an of the ELD. It was possible that those groves exceptional year for reproduction, whereas all promoted higher survival of Spott ed Owls other years were either moderately successful compared to other conifer forest types. If those or poor reproductive years. That large temporal POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 33 variation in fecundity was consistent with re- drawn by Noon et al. (1992) when they estimat- sults of a Northern Spott ed Owl meta-analysis ed California Spott ed Owl trends using a Leslie (Franklin et al. 1999) as well as with general projection matrix (see below). Three of the study predictions for long-lived animals (Gaillard et areas (LAS, SKC, SAB) had litt le evidence sup- al. 1998, 2000; Pfi ster 1998). porting a declining trend in λt. A negative trend

Climate models explained all of the estimable in λt was observed on both the ELD and SIE. temporal process variation in fecundity in a pop- That was cause for concern on the SIE because ulation of Northern Spott ed Owls (Franklin et it suggested an accelerated rate of decline in the al. 2000). Other studies also have linked climatic owl population on the SIE during most of the factors with reproductive output in Northern study period. However, the high reproduction (Wagner et al. 1996, Zabel et al. 1996), California in 1992 on the four Sierra Nevada studies may (North et al. 2000), and Mexican spott ed owls have been responsible for driving those trends (Seamans et al. 2002). In addition, weather pat- under the following hypothetical scenario. The terns were highly variable throughout the range estimate of λt for 1992–1993 was a function of of the California Spott ed Owl (Kahrl 1979) and both numbers of adults surviving from 1992 may have accounted for the large temporal and number of new recruits entering the territo- variation in fecundity. Factors that may have rial population in 1993. If juvenile survival was contributed to the occurrence of exceptional ∼30%, then a relatively large number of new years in Spott ed Owl reproduction included recruits would have been available to enter the mild spring weather (Franklin et al. 2000) and territorial population in 1993 (and later years) high prey abundance. Prey abundance has been from the large cohort produced in 1992. Thus, positively associated with reproduction in other high estimates of λt in the interval 1992–1993 owl species (Verner et al. 1992a). The exception- (and subsequent years) may have resulted from ally high fecundity in 1992, which occurred near the increased recruitment from that cohort. If the beginning of the Sierra Nevada studies, may density increased because owls occupied lower- have driven the negative linear trend seen on quality habitats, a decline would be expected the SIE. Additional study and analyses using because of lower vital rates of owls in the less weather covariates will be needed to elucidate productive habitats subsequent to that breeding those trends (e.g. correlation of fecundity and pulse. That prediction could be tested on some weather patt erns). study areas by examining territory occupancy San Bernardino study area exhibited neither patt erns to determine if territories originally the extremely successful reproduction in 1992 occupied aft er 1992 were the fi rst to become nor any of the very poor reproductive years vacant. Alternatively, populations prior to 1993 seen in all of the Sierra Nevada studies and thus could have been at relatively high or low den- was less variable (CV of temporal process varia- sities, thus confounding subsequent territory tion = 21.7%) than the Sierra Nevada study ar- occupancy patt erns. That breeding–recruitment eas. Weather patt erns appeared to have a strong pulse could have aff ected estimates of mean λt; link to owl fecundity in the SAB (W. S. LaHaye earlier high estimates would have increased the et al. unpubl. data). Therefore, diff erent weather mean estimate and also contributed to a larger patt erns were one possible explanation for the standard error. diff erences in temporal variability. Although estimates of temporal process

Population trend.—Mean estimates of λt from variation appeared higher for the ELD and SKC, all study areas were <1.0 except for the ELD. 95% CI for all study areas overlapped consider- However, 95% CI for all estimates included ably, and four of the fi ve confi dence intervals 1, indicating that all of the populations were included zero. Therefore, rate of population stationary or the estimates of λt were not suf- change varied only slightly among years on the fi ciently precise to detect declines if they oc- individual study areas. However, the best mod- curred. That latt er point was important because el from the meta-analysis of the Sierra Nevada point estimates for four of the fi ve study areas study areas indicated that λt was diff erent for indicated annual population declines of 2–4%, each of the study areas, that the estimates varied but the estimates were not suffi ciently precise among years, and that diff erences among study to diff erentiate those estimates from stationary areas were additive across years. The added populations. That inference was similar to one power of the meta-analysis may have allowed 34 ORNITHOLOGICAL MONOGRAPHS NO. 54 us to detect time eff ects in the Sierra Nevada area (i.e. recruitment of immigrants born outside study areas. the study area).

Our estimates of λt apply only to the years Burnham et al. (1996) recognized the poten- when the studies occurred (Raphael et al. 1996). tial for underestimating juvenile survival from Thus, predictions regarding the trajectory of mark–recapture methods using study areas California Spott ed Owl populations beyond with fi nite size (see also Barrowclough 1978). those time frames are not appropriate. Estimates They corrected that bias by using estimates of λt also should not be interpreted as numbers of emigration rates derived from studies of of birds; those are annual estimates of rates of radiomarked juvenile Spott ed Owls (Forsman change in the number of birds. For example, et al. 1996). Following Northern Spott ed Owl periodic estimates of λt that are <1.0 in the SIE meta-analysis (Burnham et al. 1996), research- and LAS study areas represent a decrease in the ers on the ELD proposed a study of juvenile number of birds. Intervening values >1.0 do not survival using radiotelemetry. Unfortunately, indicate that the population was restored to the funding constraints did not allow for such a original numbers at the beginning of the study; study. California Spott ed Owl researchers on they only indicate that numbers increased rela- the ELD, SIE, and SKC recognized the limita- tive to numbers in the preceding year. Thus, a tions of their juvenile survival estimates and cyclic patt ern in λt can exist that ultimately re- how those limitations potentially biased analy- sults in losses of birds over time. However, that ses using a Leslie projection matrix. Therefore, should be somewhat balanced in the estimates they used a “surrogate” estimate of juvenile of mean λt over time. We att empted to under- survival derived from studies of the SAB popu- stand how λt relates to changes in abundance by lation, which was believed to accurately refl ect estimating realized changes in populations on juvenile survival in that population (Steger et each study area, on the basis of estimates of λt. al. 2000, Seamans et al. 2001b). That approach On the basis of those population trajectories, the was unsatisfactory because the conditions on SIE was the only study area exhibiting evidence the SAB were not similar to those on the ELD, of a signifi cant reduction of the original popula- SIE, and SKC. Consequently, an alternative ap- tion over the course of the study. proach was to examine the juvenile survival rate Historically, stage-based Leslie projection needed to achieve an estimate of λ = 1.0, given matrices were used to estimate λPM in Spott ed the other vital rates (Seamans et al. 2001b). Owl populations (e.g. Forsman et al. 1996, Noon The LAS study was designed to use a Leslie et al. 1992, Blakesley et al. 2001). Initially, we did projection matrix approach; thus, they used an not use the Leslie projection matrix approach estimate of juvenile survival derived from their because of problems in estimability of param- data (Blakesley et al. 2001). The percentage of eters for juvenile survival. However, a number of juveniles born on the ELD and subsequently workshop participants felt that the larger prob- recruited into the ELD population (9%; 13 of lem was that we could not calculate an unbiased 153) was substantially less than in other stud- estimate of juvenile survival or juvenile emi- ies, which suggests that either the dispersal gration, parameters necessary for an unbiased patt ern for that population was very diff erent estimate of λPM. The possible exception to that from other owl populations (i.e. more juveniles problem was the SAB study area. Using the alter- disperse further) or that juvenile survival was native estimator for population rates of change much lower. Thus, because we were uncertain

(λt), the demographic components of λt (apparent if juvenile survival rates from all studies were survival and recruitment) in our analyses were biased by emigration, we did not use λPM for the confounded; apparent survival was a function meta-analysis of owl trends. of death and emigration, whereas recruitment Summary of general results.—Metrics estimated was a function of owls born on the study area in this meta-analysis were linked because λt and recruited into the study population plus im- is the sum of apparent survival (survival migration and subsequent recruitment of owls and emigration) and recruitment (births plus from outside the study area. Consequently, λt did immigration). For example, survival of ter- not separate population growth within the study ritorial owls probably established the baseline area (i.e. recruitment of young born on the study for λ and fecundity and recruitment aﬀ ected area) from immigration from outside the study temporal variability in λ for a Northern Spott ed POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 35

Owl population (Franklin et al. 2000). Similarly, low estimates of and low adult apparent sur- low process variation in our apparent survival vival. Those analyses indicated that four of the estimates suggested that variation in λt will like- study populations (ELD, LAS, SAB, and SKC) ly result from variation in recruitment (e.g. the appeared to be stationary or slightly declining. boom reproductive year and the hypothesized The SIE owl population probably experienced subsequent lag eff ect in λt, described above). a decline during the period of study. Evidence All of the study areas in our analyses dem- that the SIE population declined included (1) a onstrated similar patt erns in vital rates, with linear decline in fecundity, (2) a 95% CI on two main exceptions: (1) apparent survival that barely encompassed 1, and (3) a linear de- was higher for SKC than the other study areas, cline in that suggests an accelerated decline in and (2) SAB did not experience the boom and the owl population, which was best illustrated bust years in fecundity observed in the Sierra with the realized change estimates. Our results Nevada. Two diff erences between those study supported the third scenario, that Spott ed Owls areas and the other three were that the SKC was declined within a portion of the studies and ap- in a national park and the SAB was an insular peared stationary in some of the studies. Thus, population disjunct from the Sierra Nevada. there was uncertainty across all studies whether However, the cause of demographic diff erences California Spott ed Owls were declining or were among the SAB, SKC, and other study areas was stationary. For example, point estimates of for unknown. four out of fi ve study areas were <1.0 but were We att empted to address three possible sce- not suffi ciently precise to diff erentiate them as narios with our meta-analysis: (1) California slight declines or as stationary populations. Spott ed Owls declined across the studies, (2) Further, estimates of apparent survival were California Spott ed Owls were stationary across less than those reported in a meta-analysis of the studies, and (3) California Spott ed Owls de- the declining Northern Spott ed Owl (Franklin clined within a portion of the studies. When the et al. 1999). Thus, additional data will be needed estimates and trends in the population param- to resolve those uncertainties. eters were examined as a whole, there appeared to be two ends of a spectrum (Table 17). In the Study-area-specific Inferences Sierran province, the SKC population on national park lands seemed to be the most viable Lassen study area.—Estimated apparent sur- population with the highest adult apparent survival probability of adult Spott ed Owls was vival, a positive trend in , and no evidence of nearly identical to a previously published es- a trend in fecundity. At the opposite end of that timate for the LAS, 1990–1999 ( = 0.827, SE = spectrum was the SIE study, which had the low- 0.015; Blakesley et al. 2001). Additional research est estimate of , low adult apparent survival, indicated that apparent survival was positively and declining trends in both and fecundity. correlated with the amount of specifi c cover The ELD and LAS studies were between those types selected by the owl at the landscape scale two endpoints and were diffi cult to classify in for nesting, roosting, and foraging within terms of their status. The SAB study area was owl core and nest areas (814 and 203 ha areas also diffi cult to classify, although it had both around nest areas; Blakesley 2003).

TABLE 17. Summary of estimates and trends in population parameters for California Spotted Owls from five demographic study areas in California. ˆ Study area Land owner Province O Trend in Ot I Trend in fecundity SKC National park Sierran 0.984a Declining then 0.89 No trend increasing LAS National forest Sierran 0.985a Stable 0.84 Even–odd year ELD National forest Sierran 1.042a Declining 0.82 No trend SIE National forest Sierran 0.961b Declining 0.82 Declining SAB National forest S. Californian 0.978a Stable? 0.81 No trend a Not different than O = 1.0 (see Fig. 5). b Considered different than O = 1.0 (see Fig. 5). Abbreviations: SKC = Sequoia and Kings Canyon national parks study area, LAS = Lassen study area, ELD = Eldorado study area, SIE = Sierra study area, SAB = San Bernardino study area. 36 ORNITHOLOGICAL MONOGRAPHS NO. 54

Our estimate of mean fecundity was slightly smaller LAS study area was designated post hoc higher than a previously published estimate during this analysis (i.e. to be comparable to for the LAS, 1990–1999 ( = 0.291, SE = 0.065; the other studies in the meta-analysis, contigu- Blakesley et al. 2001). Our higher fecundity es- ous segments of LAS were chosen that would timate resulted from the additional year of data serve as a de facto density study area by virtue of (2000), during which the owls had relatively their complete survey each year), the assump- high fecundity. Blakesley (2003) found that re- tion of equal capture probability of banded productive output in the LAS was negatively and unbanded birds may not have been met. correlated with elevation and amount of non- In addition to the behavioral response of ani- forested habitat or habitat dominated by trees mals to being trapped (discussed above), “trap <30 cm diameter at breast height (DBH) within response” may have been present in the data if 203 ha of the nest area. fi eld personnel learned where to fi nd individual The even–odd year (high–low) eff ect on fecun- owls over time, increasing the probability that dity in the LAS may be a general trait of Spott ed banded owls were recaptured. Although that Owls in the Cascade geographic province. For may have occurred to a small extent in all of example, all of the study areas in the Oregon the Spott ed Owl demography studies, it may and Washington Cascades showed that same have been more pronounced in the LAS study eff ect in a meta-analysis of Northern Spott ed area because the study was originally designed Owls (Franklin et al. 1999). However, the even– to follow the fates of individual owls through odd year relationship was more regular in many time, rather than to locate and capture every of the Northern Spott ed Owl Cascade study owl within a predefi ned area. If trap response areas than in the LAS. Forces driving that phe- was present, it would have positively biased nomenon were unknown, but the patt ern may estimates of λt. have been infl uenced by a combination of vari- Eldorado study area.—Seamans et al. (2001b) able prey densities and weather. The very high reported that from 1990 to 1999 the apparent owl reproductive output in 1992 coincided with survival rate of males ( = 0.844, SE = 0.015) was a mild winter, and a peak in Peromyscus density higher than that of females ( = 0.819, SE = 0.018), which immediately followed an unusually large and that the apparent survival rate of both sexes sugar pine cone crop (J. A. Blakesley pers. obs.). followed a log-linear patt ern over time. Here, the The second best model of fecundity for LAS in- meta-analysis of apparent survival examined the cluded a weak negative time trend in addition study areas jointly, which precluded modeling to the even–odd year eff ect of the best model. sex eff ects or time trends separately for the ELD. That weak downward trend may have been an Thus, it was unclear if those patt erns would still artifact of the extremely high reproductive year be supported with the additional year of data. occurring in year 3 of the 11-year study. Regardless, estimates of temporal process varia- Our estimated population trend for LAS tion indicated that apparent survival varied litt le

1994–1999 using λt was higher than that esti- over time. That suggested that annual changes in mated for LAS 1990–1999 using the Leslie pro- biotic and abiotic factors, such as prey availabil- jection matrix ( = 0.910, SE = 0.025; Blakesley ity, habitat, and weather, had only a small eff ect et al. 2001). If both estimates were correct, that on annual apparent survival. implied that the population of owls in the LAS Our current estimate of mean fecundity was was being sustained by immigration or that the similar to that from a previous published esti- population declined more steeply from 1990 to mate for the ELD ( = 0.400, SE = 0.010; Seamans 1994 than from 1994 to 1999. Immigrants to the et al. 2001b). Although a negative linear decline LAS would most likely come from the Plumas best fi t the data, this model explained only 24% of National Forest, to the south, because there was the inter-annual variation in fecundity estimates. litt le suitable Spott ed Owl habitat to the west, Selection of the linear model may have been a north, and east of the LAS. In addition to the consequence of the relatively high estimate of diff erences in methods used to estimate rates fecundity in 1992, which was early in the study. of population change, and in the inferences Further, the large amount of process variation to be drawn from the two methods (discussed left unexplained by the linear time trend sug- above), the discrepancy in λ estimates may have gested that annual variation in biotic and abiotic resulted from the following factor. Because the factors probably aff ected owl reproduction. POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 37

The estimated population trend in the ELD, late 1800s (Beesley 1996). Most of that land has

1992–1999, using λt ( = 1.042, SE = 0.047; this since passed to individuals or corporations study) was higher than that estimated for the and has been managed primarily for timber ELD, 1990–1999, using the Leslie projection ma- production. Although both public and private trix ( = 0.948, SE = 0.026; Seamans et al. 2001b). land were harvested on the ELD, private land Although the confi dence intervals overlapped for harvest rates were probably higher than on those two estimates of λ, they indicated diff erent adjacent public land during the study because population trajectories; λt a stable or increasing of application of CASPO guidelines (Verner population, and λPM a declining population. It et al. 1992a). Timber harvest on the ELD may was unlikely that was due to the diff erent time have aff ected territorial owls by reducing suit- periods because abundance of territorial owls in- able habitat for roosting, foraging, or nesting creased from 1990 to 1992 (Seamans et al. 2001b). (Bias and Gutiérrez 1992, Moen and Gutiérrez Therefore, the diff erence occurred because either 1997). Conversely, forest succession may have the two methods did not share the same infer- increased the amount of mature forests during ences (particularly because λPM assumes constant the study; that is, forests with characteristics vital rates over the study period), or, that one or that were associated with Spott ed Owls on both were biased (see above). the ELD (Bias and Gutiérrez 1992, Moen and We do not know if the negative linear trend Gutiérrez 1997). Although we did not know the in λt, and the point estimates of λt < 1.0 in the net eff ect of the above factors on habitat quality, latt er part of the study were reason for concern. they likely will infl uence the long-term popula- Further, understanding the mechanisms behind tion trajectory of Spott ed Owls on the ELD. the increase and subsequent decline in the pop- Sierra and Sequoia and Kings Canyon national ulation will be of ultimate interest. Vital rates parks study areas.—Adult apparent survival responsible for changes in abundance exhibit may have been diff erent between the SIE and varying degrees of temporal variability. During SKC because of natural or management-in- our study, inter-annual variation in λt was prob- duced diff erences in vegetation. We propose ably most closely related to vital rates with three hypotheses that could relate diff erences large temporal variability, such as fecundity and in vegetation to diff erences in apparent sur- recruitment (Franklin et al. 2000, Seamans et al. vival between the study areas: (1) natural diff er- 2002). Thus, factors that exhibited temporal ences in amount of oak-woodland and sequoia variability, such as weather and prey availabil- groves, (2) diff erences in timber management, ity, may have been responsible for the observed and (3) diff erences in the application of pre- variation in λt. scribed burning. Twenty-nine percent (13 of 45) Changes in habitat quality on the ELD may of owl sites on the SIE were in the low-eleva- also have aff ected λt and were of greater concern tion oak woodlands, whereas 21% (8 of 39) on because they may have long-term eff ects on vital the SKC were in that vegetation type. Owl sites rates. Two apparent causes of habitat loss during in oak woodland had less live-tree basal area the study were logging and wildfi re. There were and canopy cover than sites in conifer forests two catastrophic wildfi res that occurred dur- (Verner et al. 1992a). If sites in oak woodland ing our study, the 1992 Cleveland Fire (99 km2) were of lower quality, then owls in those sites and the 2001 Star Fire (65 km2), which burned may have had lower survival rates. Another all or part of six owl territories. Although the diff erence in vegetation was that there were Cleveland Fire was adjacent to the ELD in what substantially more giant sequoia groves on would later become part of the Regional Study the SKC (see above). Giant sequoia groves may Area, that large wildfi re may have displaced in- have provided high-quality Spott ed Owl habi- dividual owls that subsequently immigrated to tat because they represented large blocks of old the ELD. It was too early to determine the eff ect growth forest, which were positively correlated of the Star Fire on the population. to Northern Spott ed Owl survival (Bart 1995, The ELD was unique among the study areas Franklin et al. 2000). Hunsaker et al. (2002) also because it was a mosaic (“checkerboard”) of found that owl productivity scores, based on public and private land ownership (Bias and presence of owls and their reproduction, were Gutiérrez 1992). That patt ern was a relic of land positively correlated with the proportion of the allocation to the Central Pacifi c Railroad in the area having ≥50% canopy cover on the SIE. 38 ORNITHOLOGICAL MONOGRAPHS NO. 54

Changes in vegetation resulting from timber typically occurred in summer or fall and were management may have lowered adult survival of moderate intensity with smaller areas of high in the SIE. Timber harvest on the SIE began intensity. In contrast, the Sierra National Forest in the late 1800s (Johnston 1968) and has con- did not begin a prescribed burning program in tinued until the present, although at a much the SIE until 1995. Prescribed fi re in the SIE was reduced rate since 1930. Nearly all logging was designed to be multi-entry, with the fi rst entry in the form of selective cutt ing with only scat- in the cool season (November to May); actual tered small (<8 ha) clearcuts. Selective cutt ing burn dates are dictated by weather conditions. removed the largest, most valuable trees and The emphasis of the fi rst burn was the removal in some areas reduced large-diameter stands of fi ne fuels, shrubs, and small-diameter (1 to to small remnant populations of large trees 5 cm) conifers. There were 16 cool-season burns (Verner et al. 1992a). During the course of our in the SIE that covered ∼4,850 ha (7% of study study, there were at least 11 timber sales within area). Thus, prescribed fi res in the SKC were of SIE, several of them occurring within owl higher intensity based on season of burn (obser- roost or nest areas. In 1993, CASPO guidelines vations of resulting tree mortality and reduction (Verner et al. 1992a) resulted in restricted timber of large-diameter logs), had diff erent objectives, harvest on national forest land within the study and have been used as a management tool for a area. However, in 1997 and 1998, ∼600 ha in the longer time than in the SIE. Kings River Administration Study were given Mean fecundity was not statistically dif- an exemption from the CASPO guidelines to ferent between the SIE and SKC. However, facilitate an experiment using more intensive fecundity appeared to decline linearly on the forest management strategies. Other timber SIE, but remained relatively constant on the harvest activities within SIE included “haz- SKC. Those diff erent patt erns in fecundity may ard” tree removal along roads and salvage of have been real or may have been an artifact of insect- infested trees. In contrast, relatively litt le having an additional year of data (1990) for the timber harvesting occurred on the SKC. Some SIE. Annual fecundity rates from 1995 through SKC areas were harvested in the late 1800s 1998 were the lowest estimates for the SIE and and early 1900s. In 1890, the Sequoia National were among the lowest for the SKC. That may Park and General Grant National Park (which have resulted from variation in prey abun- became part of Kings Canyon National Park) dance, weather conditions, habitat quality, or were established (Dilsaver and Tweed 1990). all three during that portion of the study. We Park designation protected ∼80% of SKC from did not monitor prey populations in the study commercial logging activities. By the end of area so their eff ect on fecundity rates was un- 1940, only two small areas totaling <8 km2 were known. North et al. (2000) found that fl edgling not under National Park Service administra- production was negatively correlated with pre- tion. Management activities, such as hazard tree cipitation during the nesting period and posi- removal and the construction of new visitor fa- tively correlated with minimum temperatures cilities, have removed relatively few trees. Thus, in April on the SIE and SKC. From 1990 to 1993, there was probably more mature or old-growth minimum temperatures in April on the SIE were coniferous forest in the SKC than the SIE. higher than for the period 1994 to 1998, and pre- The history of fi re management was diff er- cipitation was higher in years 1991, 1995, 1996, ent between the SKC and SIE. Prescribed fi re and 1998 (North et al. 2000). Thus, weather may appeared to have litt le immediate eff ect on have reduced mean annual fecundity from 1995 apparent survival of adult owls on the SIE, or to 1998. in other studies (Bond et al. 2002). However, Selection of diff erent time-trend models for long-term eff ects of prescribed fi re (e.g. for the SIE and SKC suggested that popula- changes in vegetation structure and commu- tion dynamics may have been diff erent between nity composition, and reduction of the risk of the two study areas. Even though linear and catastrophic wildfi re) on Spott ed Owl survival quadratic models were similarly weighted for were unknown. Prescribed burning on the SKC each of the study areas, realized changes based occurred prior to and throughout the study on the annual estimates (without model con- period; ∼3,724 ha (10.9% of study area) in SKC straints) suggested the two populations were were burned during the study period. Burns following diff erent trajectories. Although the POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 39

SIE and SKC were designed to provide insight populations. Southern California was at the into the eff ect of diff erent land management southern margin of the polar front jet stream practices on Spott ed Owl demographics, the (Minnich 1986). Therefore, winter storms, application of CASPO guidelines (Verner et al. which provided most of the annual precipita- 1992a) may have reduced the eff ect of the dif- tion in the San Bernardino Mountains (Minnich ferences in management. Nonetheless, point 1986), occurred less frequently and tended to estimates of apparent survival, fecundity, and be of shorter duration than storms in northern were higher on the SKC. However, only and central California. Thus, milder winter apparent survival was signifi cantly diff erent weather may have been one explanation for between the study areas. lower temporal variability in fecundity for the San Bernardino study area.—Estimates of ap- SAB. Alternatively, low temporal variability in parent survival were likely unbiased estimates SAB fecundity could have been infl uenced by of true survival for the SAB. Insular populations less variable prey dynamics or intrinsically low of Spott ed Owls in southern California were re- variability in fecundity with infrequent pulses stricted to the higher elevations of the larger in reproduction (e.g. Simmons 1996). However, mountain ranges because desert and shrub en- rare events have been diffi cult to document in vironments dominated lower elevations in the studies of short duration (Weatherhead 1986). region (Noon and McKelvey 1992, LaHaye et Some data from the SAB study area were not al. 1994). Consequently, those owl populations included in the fecundity analysis because pro- were relatively isolated from one another, and tocols for assessing reproduction diff ered from intermountain movements (i.e. inter-population the other study areas. Most deviations from movements) were rare (LaHaye et al. 2001). standard protocols occurred when owls failed That isolation may have been exacerbated by to fl edge young, because owls without young urban expansion and associated vegetation were less aggressive and less likely to take mice changes that have occurred during the last (W. S. LaHaye and R. J. Gutiérrez pers. obs.). century (Noon and McKelvey 1992, LaHaye et Thus, our fecundity estimates may have been al. 2001). biased high. That conjecture was supported by Although research has not demonstrated a the low number of unbanded owls detected on causative infl uence of habitat on apparent sur- the study area each year (i.e. if we were failing vival, landscape structure (Franklin et al. 2000) to detect and band fl edglings, we would have and forest management (Franklin et al. 1999) expected to see more unbanded recruits to the have been correlated with apparent survival territorial population). in previous studies. Logging was limited in the Although estimates of population change

San Bernardino Mountains during the study, so based on λPM contained an unknown amount timber harvest probably did not inﬂ uence ap- of bias in many Spott ed Owl studies, λPM may parent survival estimates in the SAB. However, have been appropriate for the SAB because of other factors that could have inﬂ uenced Spott ed the insular nature of the study area (LaHaye et

Owl habitat and apparent survival in the SAB al. 2001). A previous estimate of λPM from the included long-term drought (LaHaye et al. SAB study area, using all 12 years of data, was 1994), habitat fragmentation and loss because of signifi cantly less than 1.0 ( = 0.91, SE = 0.01; urbanization (LaHaye et al. 2001), and air pol- LaHaye et al. 1999). The diff erent population lution (e.g. ozone levels were particularly high trajectories indicated by λPM from LaHaye et in some forested areas of the San Bernardino al. (1999) and λt in this meta-analysis may have Mountains; Miller et al. 1997). Ozone could been due to (1) a decline in the SAB population have aff ected owls directly through potential between 1988 and 1992 that was not captured damage to lung tissue (Rombout et al. 1991), by estimation of λt in our meta-analysis; (2) or indirectly by reducing forest productivity recruitment of fl oaters that were present prior (Miller et al. 1997). to the study’s initiation (i.e. individuals fl edged Fecundity was more variable than survival prior to the beginning of the study), such that in the SAB. Temporal variability in fecundity territorial birds were being replaced, but fe- correlated with weather in Northern (Franklin males were not replacing themselves; (3) bias et al. 2000), California (North et al. 2000), and in one or more of the components of λPM; or (4) Mexican (Seamans et al. 2002) spott ed owl a violation of the assumption of constant vital 40 ORNITHOLOGICAL MONOGRAPHS NO. 54 rates during the study period when estimating Further, diff erent habitat structures and con-

λPM. Regardless, concern may still be warranted fi gurations may have aff ected individual vital for that population because (1) the estimate of rates diff erently (Franklin et al. 2000). Thus, apparent survival was lower than reported for understanding how diff erent landscape charac- declining Northern Spott ed Owl populations teristics and management strategies aff ect vital (Franklin et al. 1999), (2) the point estimate of rates, especially survival, will be essential for

λt was <1.0, and (3) the model for λt indicated a conservation of the Spott ed Owl in the Sierra negative linear trend. Nevada. Although the data in our analyses spanned Conclusion and Recommendations 7–10 years, the study periods were still relatively short for capturing some of the dynam- Although sequoia groves may have provided ics of California Spott ed Owl populations. For a positive infl uence on apparent survival rates, example, the high reproductive output observed we suspected that the disparity in apparent sur- in 1992 may have aff ected rates of population vival rates between SKC and the other study ar- change for several years following that event. eas in the Sierra Nevada was also infl uenced by If California Spott ed Owl population dynamics diff erent rates of timber harvesting. Apparent were largely driven by such events, then con- survival varied litt le among years for any of tinued monitoring of those populations will be the study areas. Conversely, fecundity exhib- necessary to capture those relatively rare events. ited relatively large temporal variation, which In addition, comparisons between the SAB and is most likely att ributable to weather or prey the Sierran studies and between the SIE and SKC availability. That relationship was suggested will yield further understanding on the eff ects of by estimates from the SAB, which experienced weather patt erns and habitat conditions on de- diff erent weather patt erns and exhibited much mographic parameters and rates of population lower temporal variability in fecundity. In addi- change. For those reasons, we believed that the tion, the SAB did not experience the peak year currently ongoing studies should be continued in fecundity, 1992, as did all the Sierra studies. and the SAB study should be reinstated. The The role habitat played in the above is unclear. SAB was unique among those studies because Therefore, we were unable to ascribe a causative of its diff erent dynamics, its closed population eff ect on estimates of apparent survival or fe- characteristic, and its representation as the only cundity to land management without incorpo- segment of the southern California Spott ed Owl rating habitat, management, or both covariates metapopulation under study. in the anaysis. Our studies are observational in their design, Although four of the fi ve study areas had and not experimental. However, they will be the point estimates of λt < 1.0, there was uncertainty only means by which we can understand the regarding the trajectory of those populations eff ects of environmental variation on Spott ed because, if there were small declines, we could Owl population dynamics (and the interactions not have statistically detected them given the with habitat) and will assist in defi ning specifi c precision of our estimates. With the exception treatments for large-scale experiments (Noon of SKC, estimates of apparent survival for the and Franklin 2002). Thus, those long-term study areas were lower than those reported by observational studies should form the backbone Franklin et al. (1999) for the Northern Spott ed for a larger research program that also includes Owl, whose numbers were probably declin- experiments examining current and alternative ing. That was of concern because survival of silvicultural treatments. We also recommend territorial owls may establish the baseline for that researchers from the fi ve study areas work λ; whereas other population parameters, such closely together to develop more consistent as fecundity and recruitment, may be respon- protocols for conducting surveys and estimat- sible for most of the temporal variability in ing reproductive output in the fi eld. In that way, λ (Franklin et al. 2000). The populations we more meaningful comparisons could be made studied may have exhibited slightly diff erent in future analyses. In general, we feel that the dynamics than the northern subspecies but following short-term recommendations would still followed a similar “bet-hedging” life his- clarify the results from our initial meta-analysis tory strategy (Boyce 1988, Franklin et al. 2000). of the data: POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 41

• Analyze those data with covariates— • Design landscape-scale experiments to as- such as climate, rates of timber harvest, sess the eff ects of silvicultural treatments presence of sequoia groves, and territory- designed to reduce fi re risks and the owl’s specifi c habitat confi gurations—combined response to controlled logging and silvi- with appropriate biologically based hy- cultural treatments. potheses that include both positive and negative infl uences of the covariates. Thus, we recommend that those study popu- In that manner, some of the processes lations continue to be monitored because (1)

inﬂ uencing the initial patt erns observed uncertainties exist in interpreting λt with respect during this meta-analysis can be bett er to source and sink population dynamics, (2) most

understood. of the point estimates of λt were <1.0, (3) apparent survival rates on four study areas were relatively • Continue the existing studies to capture low compared to those for Northern Spott ed the “infrequent” reproductive pulses (and Owls, and (4) they provide the best opportu- subsequent recruitment) observed in 1992 nity for long-term monitoring and future test- that may have important inﬂ uences on ing of hypotheses regarding the eﬀ ect climate population dynamics of the owls. and habitat have on Spott ed Owl population dynamics. Advances continue to be developed in

• Encourage the reﬁ nement of estimation partitioning the components of λt, such as local

models using λt that have particular appli- recruitment from outside immigration (Nichols cation to those studies on California Spott ed and Hines 2002), and those should be used in Owls. Such refi nements should include the conjunction with continued data collection on all separate estimation of local births from im- of the study areas. migration to bett er understand the role and In summary, analyses we present in this spatial scale of recruitment in population meta-analysis represent the best available data dynamics of California Spott ed Owls. on demography of the California Spott ed Owl. Therefore, they will be essential for conservation In 1992, the CASPO outlined an interim plan planning. Indeed, the California Spott ed Owl for the conservation of the California Spott ed was petitioned to be listed as an endangered Owl and its habitat in the Sierra Nevada (Verner species under the ESA. The USFWS issued a 12- et al. 1992a). One recommendation of that month fi nding on 14 February 2003, not to list report was to expand the basic demographic the owl (USDI 2003). We believe it is important studies to answer questions of critical manage- that such conservation assessments on the status ment concern regarding the California Spott ed of the owl be based on all the relevant data. We Owl. A decade later, we are still asking many also believe that all the demographic evidence of the same questions because a comprehensive available—such as estimated vital rates, rates research program, beyond the demographic of population change, and diff erences between studies, was never implemented. Therefore, we paired studies —suggest substantial caution in reiterate the following recommendations based owl conservation and management eff orts. on those in the CASPO: Acknowledgments • Develop comprehensive, accurate vegetation maps on the demographic study areas We thank P. Carlson for organizing and executing to evaluate the infl uence of landscape habi- the data verifi cation process. M. Conner, L. McDonald, tat characteristics on variation and trends B. Manly, T. Shenk, and J. Verner participated in the in demographic parameters of California discussion sessions during the workshop and pro- Spott ed Owls. vided many insights to the analysis and its biological– management implications. D. Mitchell read several draft s of the paper and provided helpful comments • Coordinate the existing demographic and corrections. J. Fox provided additional organi- studies with forest management activi- zational assistance. P. Stine and W. Yung provided ties to develop quasi-experiments on the logistical support. J. Baldwin provided assistance and eff ects of those activities on demographic very helpful suggestions on the fecundity analysis. parameters. Funding for this meta-analysis was provided by the 42 ORNITHOLOGICAL MONOGRAPHS NO. 54

Sierra Nevada Research Center, Pacifi c Southwest Werner were particularly helpful. We appreciate Research Station, and U.S. Department of Agriculture, the support received from G. Gould, Jr., and the Forest Service, Pacifi c Southwest Region. California Department of Fish and Game. We thank E. Cooch, K. Pollock, and J. Walters provided the University of California, Berkeley, for allowing us extensive and insightful reviews, which greatly access to Whitaker Forest and for providing housing strengthened this paper. J. Baldwin, R. King, and for our fi eld crews. T. Max also provided a thorough statistical review. San Bernardino study area.—We thank the fi eld Additional acknowledgments, specifi c to each study crews who assisted in collecting data. We thank San area, are listed below. Bernardino National Forest and the personnel of the Lassen study area.—We are grateful for the dedi- Big Bear Ranger District for their cooperation and assis- cated work of many fi eld assistants over the years. We tance during this study. We also thank G. Gould, Jr., J. appreciate the cooperation of Collins Pine Company, Palmer, J. Verner, D. Solis, and J. Robinson for their sup- Roseburg Resources Company, Sierra Pacifi c port. R. Smith, D. Call, M. Engle, and M. McLeod ably Industries, and W. M. Beaty and Associates, Inc., for served as assistant project leaders during the course sharing information and permitt ing us access to their of this study. The California Department of Fish and properties. This study was primarily funded by U.S. Game, California Department of Natural Resources, Department of Agriculture, Forest Service Region Southern California Edison, U.S. Department of 5. Additional support was received from the Forest Agriculture, Forest Service Region 5 (contract #FS/53- Service Pacifi c Southwest Research Station, the Lassen 91S8-00-EC14 to RJG), Pacifi c Southwest Forest and National Forest, and Colorado State University. Range Experiment Station, Snow Summit Mountain Eldorado study area.—We thank our many fi eld as- Resort, Big Bear Mountain Resort, Salad King, Inc., pri- sistants over the 14-year study period. We also thank vate donors, and the University of Minnesota provided the Eldorado National Forest and the personnel of the funding for this project. Georgetown, Pacifi c, and Foresthill Ranger Districts for their help during the study. M. Bias, M. Bond, D. Lutz, C. Moen, and Z. Peery ably served as project or Literature Cited assistant project leaders at various times during the study. G. Gould, Jr., D. Solis, J. Verner, and J. Robinson Anderson, D. R., K. P. Burnham, A. B. Franklin, provided important support during the study. We R. J. Gutiérrez, E. D. Forsman, R. G. Anthony, thank Blodgett Forest Research Station, University and T. M. Shenk. 1999. A protocol for confl ict of California, Berkeley, for providing housing for our resolution in analyzing empirical data related research staff . The California Department of Fish and to natural resource controversies. Wildlife Game, California Department of Natural Resources, Society Bulletin 27:1050–1058. Pacifi c Southwest Forest and Range Experiment Armsworth, P. R. 2002. Recruitment limitation, Station, U.S. Department of Agriculture, Forest population regulation, and larval connectiv- Service Region 5 (contract #FS/53-91S8-00-EC14 to ity in reef fi sh metapopulations. Ecology 83: RJG), and the University of Minnesota provided fund- 1092–1104. ing for this project. Barrowclough, G. F. 1978. Sampling bias in Sierra study area.—We thank the many fi eld biolo- dispersal studies based on fi nite area. Bird- gists that assisted with the data gathering; it was their Banding 49:333–341. untiring eff ort that made this project possible. We also Barrowclough, G. F., R. J. Gutiérrez, and J. G. thank K. Johnson, G. Eberlein, W. Laudenslayer, Jr., and Groth. 1999. Phylogeography of Spott ed Owl J. Verner who played a substantial role in data collec- (Strix occidentalis) populations based on mito- tion, analysis, or guidance as this study progressed. chondrial DNA sequences: Gene fl ow, genetic In addition, we are grateful for the help from Region structure, and a novel biogeographic patt ern. 5 and J. Robinson, California Department of Fish and Evolution 53:919–931. Game, and G. Gould, Jr., Sierra National Forest, Kings Bart, J. 1995. Amount of suitable habitat and via- River Ranger District, and Southern California Edison. bility of Northern Spott ed Owls. Conservation Sequoia and Kings Canyon national parks study Biology 9:913–916. area.—We thank the many employees whose eff orts in Beesley, D. 1996. Reconstructing the landscape: the fi eld, oft en under diffi cult conditions, allowed the An environmental history, 1820–1960. Pages successful collection of the data. We thank P. Shaklee 3–24 in Sierra Nevada Ecosystem Project: for supervising fi eld crews, G. Eberlein for managing Final Report to Congress, vol. II, Assessments the data, and W. Laudenslayer, Jr., and J. Verner for and Scientifi c Basis for Management Options. their roles in the development and guidance of this Centers for Water and Wildland Resources, study. We thank Sequoia and Kings Canyon national University of California, Davis. parks for allowing us to conduct the study and for Bias, M. A., and R. J. Gutiérrez. 1992. Habitat providing logistical assistance. D. Graber and H. associations of California Spott ed Owls in POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 43

the central Sierra Nevada. Journal of Wildlife Dilsaver, L. M., and W. C. Tweed. 1990. Challenge Management 56:584–595. of the Big Trees: A Resource History of Sequoia Blakesley, J. A. 2003. Ecology of the California and Kings Canyon National Parks. Sequoia Spott ed Owl: Breeding dispersal and asso- Natural History Association, Three Rivers, ciations with forest stand characteristics in California. northeastern California. Ph.D. dissertation, Evans, M., N. Hastings, and B. Peacock. 1993. Colorado State University, Fort Collins. Statistical Distributions, 2nd ed. J. Wiley and Blakesley, J. A., B. R. Noon, and D. W. H. Shaw. Sons, New York. 2001. Demography of the California Spott ed Federal Advisory Committee. 1997. Final Report Owl in northeastern California. Condor 103: of the California Spott ed Owl Federal Advisory 667–677. Committ ee. U.S. Department of Agriculture, Bond, M. L., R. J. Gutiérrez, A. B. Franklin, W. Washington, D.C. S. LaHaye, C. A. May, and M. E. Seamans. Forsman, E. D. 1981. Molt of the Spott ed Owl. Auk 2002. Short-term eff ects of wildfi res on Spott ed 98:735–742. Owl survival, site fi delity, mate fi delity, and Forsman, E. D. 1983. Methods and materials for reproductive success. Wildlife Society Bulletin locating and studying Spott ed Owls. U.S. 30:1022–1028. Department of Agriculture, Forest Service Boyce, M. S. 1988. Bet hedging in avian life his- General Technical Report PNW-GTR-162. tories. Pages 3–18 in Acta XIX Congressus Forsman, E. D., R. G. Anthony, J. A. Reid, P. J. Internationalis Ornithologici (H. O. Ouellet, Loschl, S. G. Sovern, M. Taylor, B. L. Biswell, Ed.). National Museum of Natural Science, A. Ellingson, E. C. Meslow, G. S. Miller, K. Ott awa, Ontario. A. Swindle, J. A. Thrailkill, F. F. Wagner, Burnham, K. P., and D. R. Anderson. 1998. Model and D. E. Seaman. 2002. Natal and breeding Selection and Inference: A Practical Information dispersal of Northern Spott ed Owls. Wildlife Theoretic Approach. Springer, New York. Monographs, no. 149. Burnham, K. P., D. R. Anderson, G. C. White, C. Forsman, E. D., A. B. Franklin, F. M. Oliver, and Brownie, and K. H. Pollock. 1987. Design J. P. Ward. 1996. A color band for Spott ed Owls. and analysis methods for fi sh survival experi- Journal of Field Ornithology 67:507–510. ments based on release–recapture. American Forsman, E. D., E. C. Meslow, and H. M. Wight. Fisheries Society Monograph, no. 5. 1984. Distribution and biology of the Spott ed Burnham, K. P., D. R. Anderson, and G. C. Owl in Oregon. Wildlife Monographs, no. 87. White. 1996. Meta-analysis of vital rates of Franklin, A. B. 2002. Exploring ecological rela- the Northern Spott ed Owl. Studies In Avian tionships in survival and estimating rates of Biology 17:92–101. population change using program MARK. Burnham, K. P., and G. C. White. 2002. Evaluation Pages 290–296 in Wildlife, Land, and People: of some random eff ects methodology appli- Priorities for the 21st Century (R. Field, R. J. cable to bird ringing data. Journal of Applied Warren, H. Okarma, and P. R. Sievert, Eds.). Statistics 29:245–264. Wildlife Society, Bethesda, Maryland. Caswell, H. 2000. Matrix Population Models: Franklin, A. B., D. R. Anderson, E. D. Forsman, K. Construction, Analysis, and Interpretation, P. Burnham, and F. F. Wagner. 1996a. Methods 2nd ed. Sinauer Associates, Sunderland, for collecting and analyzing demographic data Massachusett s. on the Northern Spott ed Owl. Studies in Avian Caughley, G. 1977. Analysis of Vertebrate Biology 17:12–20. Populations. John Wiley and Sons, New York. Franklin, A. B., D. R. Anderson, R. J. Gutiérrez, Connor, E. F., S. H. Faeth, and D. Simberloff. and K. P. Burnham. 2000. Climate, habitat 1983. Leafminers on oak: The role of immi- quality, and fi tness in Northern Spott ed Owl gration and in situ reproductive recruitment. populations in northwestern California. Ecology 64:191–204. Ecological Monographs 70:539–590. Cooch, E., R. F. Rockwell, and S. Brault. 2001. Franklin, A. B., K. P. Burnham, G. C. White, R. Retrospective analysis of demographic re- G. Anthony, E. D. Forsman, C. J. Schwarz, J. sponses to environmental change: A Lesser D. Nichols, and J. E. Hines. 1999. Range-wide Snow Goose example. Ecological Monographs status and trends in Northern Spott ed Owl 71:377–400. populations. U.S. Fish and Wildlife Service DeSante, D. F. 1995. Suggestions for future direc- unpublished report, Portland, Oregon. tions for studies of marked migratory land- Franklin, A. B., R. J. Gutiérrez, B. R. Noon, birds from the perspective of a practitioner and J. P. Ward, Jr. 1996b. Demography of in population management and conservation. the Northern Spott ed Owl: Northwestern Journal of Applied Statistics 22:949–965. California. Studies in Avian Biology 17:83–91. 44 ORNITHOLOGICAL MONOGRAPHS NO. 54

Franklin, A. B., and T. M. Shenk. 1995. Meta- immigration-stochastic model. Biometrika 52: analysis as a tool for monitoring wildlife pop- 225–247. ulations. Pages 484–487 in Integrating People Kahrl, W. L. 1979. The California Water Atlas, 2nd and Wildlife for a Sustainable Future (J. A. ed. California Department of Water Resources, Bissonett e and P. R. Krausman, Eds.). Wildlife Sacramento, California. Society, Bethesda, Maryland. Kendall, W. L., J. D. Nichols, and J. E. Hines. Franklin, A. B., T. M. Shenk, D. R. Anderson, and 1997. Estimating temporary emigration using K. P. Burnham. 2001. Statistical model selec- capture–recapture data with Pollock’s robust tion: The alternative to null hypothesis testing. design. Ecology 78:563–578. Pages 75–90 in Modeling in Natural Resource Küchler, A. W. 1977. The map of natural vegeta- Management: Development, Interpretation, and tion of California. Pages 909–938 in Terrestrial Application (T. M. Shenk and A. B. Franklin, Vegetation of California (M. G. Barbour and J. Eds.). Island Press, Washington, D.C. Major, Eds.). John Wiley and Sons, New York. Fujioka, F. M., J. O. Roads, and S.-C. Chen. 1998. LaHaye, W. S., R. J. Gutiérrez, and H. R. Akçakaya. Climatology. Pages 28–43 in Oxidant Air 1994. Spott ed Owl metapopulation dynam- Pollution Impacts in the Montane Forests of ics in southern California. Journal of Animal Southern California: A Case Study of the San Ecology 63:775–785. Bernardino Mountains (P. R. Miller and J. R. LaHaye, W. S., R. J. Gutiérrez, and J. R. Dunk. McBride, Eds.). Springer, New York. 2001. Natal dispersal of the Spott ed Owl in Gaillard, J.-M., M. F.-B. Bianchet, and N. G. Yoccoz. southern California: Dispersal profi le of an 1998. Population dynamics of large herbivores: insular population. Condor 103:691–700. Variable recruitment with constant adult sur- LaHaye, W. S., R. J. Gutiérrez, and G. S. vival. Trends in Ecology and Evolution 13:58–63. Zimmerman. 1999. Demography of the Gaillard, J.-M., M. F.-B. Bianchet, N. G. Yoccoz, California Spott ed Owl in the San Bernardino A. Loison, and C. Toigo. 2000. Temporal varia- Mountains. Final Report to Region 5, U.S. tion in fi tness components and population dy- Department of Agriculture, Forest Service, namics of large herbivores. Annual Review of Vallejo, California. Ecology and Systematics 31:367–393. Lebreton, J.-D. 1996. Demographic models for Gutiérrez, R. J., A. B. Franklin, and W. S. LaHaye. subdivided populations: The renewal equation 1995. Spott ed Owl (Strix occidentalis). In The approach. Theoretical Population Biology 49: Birds of North America, no. 179 (A. Poole and 291–313. F. Gill, Eds.). Academy of Natural Sciences, Lebreton, J.-D., K. P. Burnham, J. Clobert, and Philadelphia, and American Ornithologists’ D. R. Anderson. 1992. Modeling survival and Union, Washington, D.C. testing biological hypotheses using marked Gutiérrez, R. J., and J. Pritchard. 1990. animals: A unifi ed approach with case studies. Distribution, density, and age structure of Ecological Monographs 62:67–118. Spott ed Owls on two southern California habi- McKelvey, K. S., and J. D. Johnston. 1992. tat islands. Condor 92:491–495. Historical perspectives on forests of the Sierra Hines, J. E., and J. D. Nichols. 2002. Investigations Nevada and the transverse ranges of southern of potential bias in the estimation of λ using California: Forest conditions at the turn of Pradel’s (1996) model for capture–recapture the century. Pages 225–246 in The California data. Journal of Applied Statistics 29:573–588. Spott ed Owl: A Technical Assessment of Its Hunsaker, C. T., B. B. Boroski, and G. N. Steger. Current Status (J. Verner, K. McKelvey, B. R. 2002. Relations between canopy cover and Noon, R. J. Gutiérrez, G. I. Gould, Jr., and T. W. the occurrence and productivity of California Beck, Eds.). U.S. Department of Agriculture, Spott ed Owls. Pages 697–700 in Predicting Forest Service General Technical Report PSW- Species Occurrences: Issues of Accuracy and GTR-133. Scale (J. M. Scott , P. J. Heglund, M. L. Morrison, McKelvey, K. S., C. N. Skinner, C. Chang, D. C. J. B. Haufl er, M. G. Raphael, W. A. Wall, and F. B. Erman, S. J. Husari, D. J. Parsons, J. W. Van Samson, Eds.). Island Press, Washington, D.C. Wagtendonk, and C. P. Weatherspoon. 1996. Hunter, J. E., and F. L. Schmidt. 1990. Methods An overview of fi re in the Sierra Nevada. Pages of Meta-analysis: Correcting Error and Bias in 1033–1040 in Sierra Nevada Ecosystem Project, Research Findings. Sage Publications, Beverly Final Report to Congress, vol. II, Assessments Hills, California. and Scientifi c Basis for Management Options. Johnston, H. 1968. Thunder in the Mountains. Centers for Water and Wildland Resources, Trans-Anglo Books, Los Angeles. University of California, Davis. Jolly, G. M. 1965. Explicit estimates from Miller, P. R., M. J. Arbaugh, and P. J. Temple. capture–recapture data with both death and 1997. Ozone and its known and potential ef- POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 45

fects on forests in the western United States. Administration (NOAA). 1999. Climatological Pages 39–68 in Forest Decline and Ozone: A Data Annual Summary California, vol. 103, no. Comparison of Controlled Chamber and Field 13. U.S. National Oceanic and Atmospheric Experiments (H. Sandermann, A. R. Wellburn, Administration, Asheville, North Carolina. and R. L. Heath, Eds.). Ecological Studies, vol. Noon, B. R., and C. M. Biles. 1990. Mathematical 134, Springer, Berlin. demography of Spott ed Owls in the Pacifi c Milligan, J. L., A. K. Davis, and S. M. Altizer. Northwest. Journal of Wildlife Management 2003. Errors associated with using colored leg 54:18–27. bands to identify wild birds. Journal of Field Noon, B. R., and A. B. Franklin. 2002. Scientifi c Ornithology 74:111–118. research and the Spott ed Owl: Opportunities Minnich, R. A. 1986. Snow levels and amounts in for major contributions to avian population the mountains of southern California. Journal ecology. Auk 199:311–320. of Hydrology 89:37–58. Noon, B. R., and K. McKelvey. 1992. Stability Minnich, R. A. 1988. The Biogeography of Fire in properties of the Spott ed Owl metapopula- the San Bernardino Mountains of California: tion in southern California. Pages 187–206 A Historical Study. University of California in The California Spott ed Owl: A Technical Press, Berkeley. Assessment of Its Current Status (J. Verner, Minnich, R. A. 1998. Vegetation, fi re regimes, and K. McKelvey, B. R. Noon, R. J. Gutiérrez, forest dynamics. Pages 44–80 in Oxidant Air G. I. Gould, Jr., and T. W. Beck, Eds.). U.S. Pollution Impacts in the Montane Forests of Department of Agriculture, Forest Service Southern California: A Case Study of the San General Technical Report PSW-GTR-133. Bernardino Mountains (P. R. Miller and J. R. Noon, B. R., K. McKelvey, D. W. Lutz, W. S. McBride, Eds.). Springer, New York. LaHaye, R. J. Gutiérrez, and C. A. Moen. Minnich, R. A., M. G. Barbour, J. H. Burk, and 1992. Estimates of demographic parameters R. F. Renau. 1995. Sixty years of change in and rates of population change. Pages 175–186 Californian conifer forests of the San Bernardino in The California Spott ed Owl: A Technical Mountains. Conservation Biology 9:902–914. Assessment of Its Current Status (J. Verner, Moen, C. A., A. B. Franklin, and R. J. Gutiérrez. K. McKelvey, B. R. Noon, R. J. Gutiérrez, 1991. Age determination of subadult Northern G. I. Gould, Jr., and T. W. Beck, Eds.). U.S. Spott ed Owls in northwest California. Wildlife Department of Agriculture, Forest Service Society Bulletin 19:489–493. General Technical Report PSW-GTR-133. Moen, C.A., and R. J. Gutiérrez. 1997. California North, M., G. N. Steger, R. Denton, G. Eberlein, Spott ed Owl habitat selection in the central T. Munton, and K. Johnson. 2000. Association Sierra Nevada. Journal of Wildlife Management of weather and nest-site structure with repro- 61:1281–1287. ductive success in California Spott ed Owls. Mooney, H. A. 1988. Southern coastal scrub. Pages Journal of Wildlife Management 64:797–807. 471–490 in Terrestrial Vegetation of California, Olson, C., and J. A. Helms. 1996. Forest growth 2nd ed. (M. G. Barbour and J. Major, Eds.). and stand structure at Blodgett Forest Research California Native Plant Society, Sacramento. Station 1933–1995. Pages 681–732 in Sierra Nichols, J. D., and J. E. Hines. 1984. Eff ects of Nevada Ecosystem Project: Final Report to permanent trap responses in capture prob- Congress, vol. III. Centers for Water and Wild- ability on Jolly-Seber capture–recapture model land Resources, University of California, Davis. estimates. Journal of Wildlife Management 48: Parsons, D. J. 1994. Objects or ecosystems? 289–294. Giant sequoia management in the national Nichols, J. D., and J. E. Hines. 2002. Approaches parks. Pages 109–115 in Proceedings of the for the direct estimation of λ and demographic Symposium on the Giant Sequoias: Their Place contributions to λ, using capture–recapture in the Ecosystem and Society (P. S. Aune, Ed.). data. Journal of Applied Statistics 29:539–568. U.S. Department of Agriculture, Forest Service, Nichols, J. D., and K. H. Pollock. 1990. Estimation General Technical Report PSW-GTR-151. of recruitment from immigration versus in situ Pfister, C. A. 1998. Patt erns of variation in stage- reproduction using Pollock’s robust design. structured populations: Evolutionary predic- Ecology 71:21–26. tions and ecological implications. Proceedings National Oceanic and Atmospheric Admini- of the National Academy of Science USA 95: stra tion (NOAA). 1998. Climatological Data 213–218. Annual Summary California, vol. 102, no. Pollock, K. H., J. D. Nichols, C. Brownie, and J. E. 13. U.S. National Oceanic and Atmospheric Hines. 1990. Statistical inference for capture– Administra tion, Asheville, North Carolina. recapture experiments. Wildlife Monographs, National Oceanic and Atmospheric no. 107. 46 ORNITHOLOGICAL MONOGRAPHS NO. 54

Pradel, R. 1996. Utilization of capture–mark– recapture experiments in open populations. recapture for the study of recruitment and Biometrics 52:860–873. population growth rate. Biometrics 52: Schwarz, C. J., and G. A. F. Seber. 1999. Estimating 703–709. animal abundance: Review III. Statistical Pulliam, H. R. 1988. Sources, sinks, and population Science 14:427–456. regulation. American Naturalist 132:652–661. Scott, J. M., F. Davis, B. Csuti, R. Noss, B. Rao, P. S. R. S. 1997. Variance Components Butterfield, C. Groves, H. Anderson, S. Caicco, Estimation: Mixed Models, Methodologies and F. D'erchia, T. C. Edwards, Jr., J. Ulliman, and Applications. Chapman and Hall, London. R. G. Wright. 1993. Gap analysis: A geographic Raphael, M. G., R. G. Anthony, S. DeStefano, E. approach to protection of biological diversity. G. Forsman, A. B. Franklin, R. Holthausen, Wildlife Monographs, no. 123. E. C. Meslow, and B. R. Noon. 1996. Use, in- Seamans, M. E., R. J. Gutiérrez, and M. Bond. terpretation, and implications of demographic 2001a. Population ecology of the California analyses of Northern Spott ed Owl populations. Spott ed Owl in the central Sierra Nevada: Studies in Avian Biology 17:102–112. Annual results, 2000. U.S. Department of Reid, J. A., R. B. Horn, and E. D. Forsman. 1999. Agriculture, Forest Service, unpublished Detection rates of Spott ed Owls based on report, Region 5, Contract No. #53-91S8-6, acoustic-lure and live-lure surveys. Wildlife Vallejo, California. Society Bulletin 27:986–990. Seamans, M. E., R. J. Gutiérrez, and C. A. May. Robinson, J. W. 1989. The San Bernardinos: The 2002. Mexican Spott ed Owl population Mountain Country from Cajon Pass to Oak dynamics: Infl uence of climatic variation on Glen, Two Centuries of Changing Use. Big Santa survival and reproduction. Auk 110:321–334. Anita Historical Society, Arcadia, California. Seamans, M. E., R. J. Gutiérrez, C. A. May, and M. Rombout, P. J. A., J. A. M. A. Dormans, L. van Bree, Z. Peery. 1999. Demography of two Mexican and M. Marra. 1991. Structural and biochemi- Spott ed Owl populations. Conservation cal eff ects in lungs of Japanese Quail following Biology 13:744–754. a 1-week exposure to ozone. Environmental Seamans, M. E., R. J. Gutiérrez, C. A. Moen, and Research 54:39–51. M. Z. Peery. 2001b. Spott ed Owl demogra- Roughgarden, J., Y. Iwasa, and C. Baxter. 1985. phy in the central Sierra Nevada. Journal of Demographic theory for an open marine popu- Wildlife Management 65:425–431. lation with space-limited recruitment. Ecology Seber, G. A. F. 1982. The Estimation of Animal 66:54–67. Abundance and Related Parameters. Charles Rundel, P. W. 1971. Community structure and Griffi n, London. stability in the giant sequoia ecosystem of the Sierra Nevada Ecosystem Project (SNEP). 1996. Sierra Nevada, California. American Midland Status of the Sierra Nevada: Assessment Naturalist 85:478–492. Summaries and Management Strategies, vol. 1. Rundel, P. W., D. J. Parsons, and D. T. Gordon. University of California, Davis. 1977. Montane and subalpine vegetation of Simmons, R. E. 1996. Population declines, viable the Sierra Nevada and Cascade ranges. Pages breeding areas, and management options for 559–599 in Terrestrial Vegetation of California fl amingos in southern Africa. Conservation (M. G. Barbour and J. Major, Eds.). John Wiley Biology 10:504–514. and Sons, New York. Skinner, C. N., and C. Chang. 1996. Fire regimes, Sakamoto, Y., M. Ishiguro, and G. Kitagawa. past and present. Pages 1041–1069 in Sierra 1986. Akaike Information Criterion Statistics. Nevada Ecosystem Project, Final Report to KTK Scientifi c Publishers, Dordrecht, The Congress, vol. II, Assessments and Scientifi c Netherlands. Basis for Management Options. Centers for SAS Institute. 1997. SAS/STAT Soft ware: Changes Water and Wildland Resources, University of and Enhancements through Release 6.12. SAS California, Davis. Institute Inc., Cary, North Carolina. Steger, G. N. 1995. Sex ratios of fl edgling and re- Sawyer, J. O., D. A. Thornburgh, and J. R. Griffin. captured subadult Spott ed Owls in the south- 1988. Mixed evergreen forest. Pages 359–382 ern Sierra Nevada. Transactions of the Western in Terrestrial Vegetation of California, 2nd ed. Section of The Wildlife Society 31:36–39. (M. G. Barbour and J. Major, Eds.). California Steger, G. N., T. E. Munton, G. P. Eberlein, K. Native Plant Society, Sacramento. D. Johnson, and P. A. Shaklee. 2000. A study Schoen, R. 1988. Modeling Multigroup Populations. of Spott ed Owl demographics in the Sierra Plenum Press, New York. National Forest and Sequoia and Kings Canyon Schwarz, C. J., and A. N. Arnason. 1996. A gen- National Parks, Annual Progress Report 2000. eral methodology for the analysis of capture– U.S. Department of Agriculture, Forest Service, POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 47

Pacifi c Southwest Research Station, Fresno, Forest Service, Pacifi c Southwest Region, California. Sacramento, California Thomas, J. W., E. D. Forsman, J. B. Lint, E. C. U.S. Department of Agriculture, Forest Service Meslow, B. R. Noon, and J. Verner. 1990. A (USDA). 1998b. Sierra Nevada framework for conservation strategy for the Northern Spott ed conservation and collaboration: Summary Owl. Report of the Interagency Scientifi c of existing management direction, report of Committ ee to address the conservation of the the Forest Service Interdisciplinary Analysis Northern Spott ed Owl. U.S. Department of Team. U.S. Department of Agriculture, Forest Agriculture, Forest Service, U.S. Department Service, Pacifi c Southwest Region, Sacramento, of Interior, Bureau of Land Management, California. Fish and Wildlife Service, and National Park U.S. Department of Agriculture, Forest Service Service. U. S. Government Printing Offi ce 791- (USDA). 2001a. Sierra Nevada Forest Plan 171/20026, Washington, D.C. Amendment Final Environmental Impact Thomas, J. W., M. G. Raphael, R. G. Anthony, E. D. Statement, vols. 1–6. U.S. Department of Forsman, A. G. Gunderson, R. S. Holthausen, Agriculture, Forest Service, Pacifi c Southwest B. G. Marcot, G. H. Reeves, J. R Sedell, and Region, Sacramento, California. D. M. Solis. 1993. Viability assessments and U.S. Department of Agriculture, Forest Service management considerations for species asso- (USDA). 2001b. Sierra Nevada Forest Plan ciated with late-successional and old-growth Amendment Final Environmental Impact forests of the Pacifi c Northwest: The report of Statement: Record of Decision. U.S. Department the Scientifi c Analysis Team. U.S. Department of Agriculture, Forest Service, Pacifi c Southwest of Agriculture, Forest Service, National Forest Region, Sacramento, California. System, Forest Service Research, Portland, U.S. Department of Interior (USDI). 1992. Final draft Oregon. recovery plan for the Northern Spott ed Owl. U.S. Thorne, R. F. 1988. Montane and subalpine forests Department of Interior, Washington, D.C. of the transverse and peninsular ranges. Pages U.S. Department of Interior (USDI). 2003. 537–558 in Terrestrial Vegetation of California, Endangered and threatened wildlife and 2nd ed. (M. G. Barbour and J. Major, Eds.). plants: 12-month fi nding for a petition to list California Native Plant Society, Sacramento, the California Spott ed Owl (Strix occidentalis California. occidentalis). Federal Register 68:7580–7608. Vanderwerf, E. 1992. Lack's clutch size hypothesis: U.S. Department of Agriculture, Forest Service An examination of the evidence using meta- (USDA). 1993. California Spott ed Owl Sierran analysis. Ecology 73:1699–1705. Province interim guidelines environmental Vasek, F. C., and M. G. Barbour. 1988. Mojave assessment, decision notice, and fi ndings of desert scrub vegetation. Pages 835–868 in no signifi cant impact. U.S. Department of Terrestrial Vegetation of California, 2nd ed. Agriculture, Forest Service, Pacifi c Southwest (M. G. Barbour and J. Major, Eds.). California Region, San Francisco, California. Native Plant Society, Sacramento, California. U.S. Department of Agriculture, Forest Service Verner, J., and A. S. Boss. 1980. California wildlife (USDA). 1995. Draft environmental impact state- and their habitats: Western Sierra Nevada. ment: Managing California Spott ed Owl habitat U.S. Department of Agriculture, Forest Service in the Sierra Nevada national forests of California, General Technical Report PSW-GTR-37. an ecosystem approach. U.S. Department of Verner, J., K. E. McKelvey, B. R. Noon, R. J. Agriculture, Forest Service, Pacifi c Southwest Gutiérrez, G. I. Gould, Jr., and T. W. Beck. Region, San Francisco, California. 1992a. The California Spott ed Owl: A tech- U.S. Department of Agriculture, Forest Service nical assessment of its current status. U.S. (USDA). 1996. Revised draft environmen- Department of Agriculture, Forest Service tal impact statement: Managing California General Technical Report PSW-GTR-133. Spott ed Owl habitat in the Sierra Nevada Verner, J., K. S. McKelvey, B. R. Noon, R. J. national forests of California, an ecosystem Gutiérrez, G. I. Gould, Jr., and T. W. Beck. approach. U.S. Department of Agriculture, 1992b. Assessment of the current status of the Forest Service, Pacifi c Southwest Region, San California Spott ed Owl, with recommendations Francisco, California. for management. Pages 3–26 in The California U.S. Department of Agriculture, Forest Service Spott ed Owl: A Technical Assessment of its (USDA). 1998a. Sierra Nevada science review: Current Status (J. Verner, K. S. McKelvey, B. R. Report of the Science Review Team charged Noon, R. J. Gutiérrez, G. I. Gould, Jr., and T. W. to synthesize new information of rangewide Beck, Eds.). U.S. Department of Agriculture, urgency to the national forests of the Sierra Forest Service General Technical Report PSW- Nevada. U.S. Department of Agriculture, GTR-133. 48 ORNITHOLOGICAL MONOGRAPHS NO. 54

Wagner, F. F., E. C. Meslow, G. M. Bennet, C. J. populations of marked animals. Bird Study 46 Larson, S. M. Small, and S. DeStefano. 1996. (Supplement):120–138. Demography of Northern Spott ed Owls in the White, G. C., K. P. Burnham, and D. R. Anderson. southern Cascades and Siskiyou Mountains, 2002. Advanced features of program MARK. Oregon. Studies in Avian Biology 17:67–76. Pages 368–377 in Integrating People and Weatherhead, P. J. 1986. How unusual are un- Wildlife for a Sustainable Future (R. Fields, usual events? American Midland Naturalist Ed.). Wildlife Society, Bethesda, Maryland. 128:150–154. Wolf, F. M. 1986. Meta-analysis: Quantitative White, G. C. 2000. Population viability analy- Methods for Research Synthesis. Sage sis: Data requirements and essential analy- Publications, Beverly Hills, California. ses. Pages 288–331 in Research Techniques Wolﬁ nger, R. 1993. Covariance structure selection in Animal Ecology: Controversies and in general mixed models. Communications in Consequences (L. Boitani and T. K. Fuller, Statistics—Simulation and Computation 22: Eds.). Columbia University Press, New York. 1079–1106. White, G. C., and R. E. Bennetts. 1996. Analysis of Zabel, C. J., S. E. Salmons, and M. Brown. 1996. frequency count data using the negative bino- Demography of Northern Spott ed Owls mial distribution. Ecology 77:2549–2557. in southwestern Oregon. Studies in Avian White, G. C., and K. P. Burnham. 1999. Biology 17:77–82. Program MARK: Survival estimation from

=+_=&7.&9.43=.3=8:7;*>=2*9-4)8=&243,=&1.+473.&=5499*)=<1=89:).*8_= = = = = =]=!= "= 7.9*7.&=94== <18=-*&7)=&9== <18=-*&7)=&9== <18=-*&7)=.3== <18=-*&7)=.3== *89.2&9*== 8&2*=14(&9.43=43== 8&2*=14(&9.43=43== 14(&9.438=1*88=9-&3= 14(&9.438=1*88=9-&3== 4((:5&3(>== 3,=4((&8.438== 3,=4((&8.438== 43*86:&79*7=2.1*=&5&79== 43*86:&79*7=2.1*= = 8*5&7&9*)='>=31== 8*5&7&9*)='>=31=)&>8_== 43=3,=4((&8.438= &5&79=43=3,= = )&>8=47=8&2*=4<18== = 8*5&7&9*)='>=31=)&>8== 4((&8.438_= = +742=9-*=57*;.4:8=>*&7== = 47=8&2*=4<18=+742=9-*== = 4'8*7;*)=&9=9-*=8.9*_= = 57*;.4:8=>*&7=4'8*7;*)== = = = &9=9-*=8.9*_= = 7.9*7.&=+47= ,437*574):(9.43= ,437*574):(9.43== ,437*574):(9.43== ,437*574):(9.43= 7*574):(9.;*== .3+*77*)=.+=&3>=4+=9-*= .3+*77*)=.+a=+=4<1=9440== .3+*77*)=.+a=+=+*2&1*= .3+*77*)=.+a=+== )*9*72.3&9.43= +4114<.3,=4'8*7;*)=43== 3,=2.(*=&3)=(&(-*)== 74489*)=+47=3-*=2.3=43== +*2&1*=74489*)=+47== = 9<4=4((&8.438=3+=<**0= 43*=<.9-4:9=9&0.3,=94== &9=1*&89=9<4=;.8.98='*+47*== 3./=2.3=43=&9== = &5&79a=+=+*2&1*=4<1== >4:3,b=,=43*=4<1=&9*== *+=4:3*b=,=43*=4<1=9440== 1*&89=43*=;.8.9`= = 74489*)=+47=30*=2.3== 3-=2.(*b=47=-=43*=4<1== 3,=2.(*=4+=.=':9=).)== 57.47=94=+/=7&>b= = '*+47*=+=7&>b=,=4<1== &9*=9<4=2.(*=&3)== 349=)*1.;*7=94=3*89== ,=+=4<1=9440=3.= = &9*=47=(&(-*)=47='49-== .,347*)=9-.7)_= 47=>4:3,=):7.3,=3,== 2.(*=':9=).)=349= = +4:7=2.(*b=-=4<1=&9*= == 8:7;*>8=43(*=5&.7== )*1.;*7=94=3*89= = 47=(&(-*)=47='49-=3,== = 034<3=94=3*89`=;.8.9*)== ):7.3,=3+=8:7;*>= = 2.(*=&3)=7*+:8*)== = 8.9*=94=;.8:&11>=4'8*7;*== 43(*=5&.7=034<3= = &349-*7b=.=+*2&1*== = &3)=(4:39=+1*),1.3,8b== 94=3*89`=;.8.9*)== = 4'8*7;*)=.3=-&3)= = -=+*2&1*=)*9*72.3*)== 8.9*=94=;.8:&11>== = <.9-4:9='744)=5&9(-= = 94='*=8.3,1*=84(.&1= 4'8*7;*=&3)=(4:39= = +742=+/=57.1=94=-*=4:3*= = 89&9:8=8.==(4251*9*== +1*),1.3,8_= = 431>=43*=4'8*7;&9.43= = 8:7;*>8=4+=8.9*`== = 7*6:.7*)_= = 34=2&9*=)*9*(9*)_= 5hhqugu ( F5S 1 Fg q p gqg 9F8 1 9xpgp p gqg SC9 1 Suqg p gqg SE7 1 Squg gp Eus 7g gugx gw p gqg gp S56 1 Sg 6qgpu p gqg POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 49

Appendix 2 3. Data handling issues for fecundity (b): a. Unknown-age birds should be coded as Final writt en protocol developed during the work- unknown. Using unbanded birds in fecundity shop on status and trends in California Spott ed Owl fi le is acceptable. populations, held 9–13 July 2001 at Colorado State b. Delete radioed birds to be consistent with University, Fort Collins, Colorado. survival analysis. c. Delete records that do not meet the protocol 1. Study area characteristics (Table A1): for each study area. Ignore July 15 cutoff on a. San Bernardino Mountains (SAB) SIE and SKC because that was not used on b. Sierra (SIE) National Forest other study areas. c. Sequoia and Kings Canyon (SKC) d. Records of resident birds with unknown social d. Eldorado National Forest (ELD) status were not included in the SIE and SKC. e. Lassen National Forest (LAS) e. Recognize that b is biased high because 2. Data handling issues for apparent survival (φ): establishing no reproduction is more diffi cult a. Bird identifi ed by juvenile cohort band than demonstrating reproduction. but not captured to give adult individual f. Estimates are based on females of known age identifi cation bands. Those birds were not only (unknown-age females not included in recorded as being resighted until the bird was estimates). actually recaptured and banded as a uniquely 4. Estimate demographic parameters for California identifi able individual. Spott ed Owls across fi ve study areas: b. Birds recaptured off study areas that were a. To be estimated: located by someone else were treated as not i. Age-specifi c apparent survival (φ) using

recaptured (0 cases for SAB, 2 cases for LAS, 2 QAICc model selection criteria in program cases for SIE, 0 cases for ELD, 1 case for SKC) MARK. and frequency was converted from –1 to 1. 1. {φ(a * s * t) p(a * s * t)} is the global model c. One capture history for Barred Owl (Strix varia) where juvenile PIMs are structured with × Spott ed Owl hybrid on LAS was deleted. time-specifi c p values separate from S1, d. Expansion of study area on ELD—Birds that S2, or A age-classes. emigrated into regional before expansion area 2. Using {φ(a * s * t)}, fi nd best p value from was surveyed and then were found. Solution the following set: was to keep those records and, hence, ignore a. For juveniles, consider the four that source of individual heterogeneity in p models J., J3, JA, or JAA, and pick the values because of the small number of cases. best juvenile model, holding the non- e. Juvenile capture histories were split at fi rst juveniles as a dot model—total of four recapture to separate juvenile recapture models. probabilities from S1, S2, and A age-classes b. For nonjuveniles, consider the four (e.g. a juvenile CH of 1011101 with frequency models dot, sex (s), reproductive rate = 1 was split into 1010000 with frequency = –1 covariate (= r), or r * s, plus a model of and 0011101 with frequency = 1). choice based on previous experience,

=+_=:22&7>=4+=)&9&=(411*(9.43=+47=9-*=+.;*=89:)>=&7*&8_= = -&7&(9*7.89.(= = = = = = .2*=5*7.4)=949&1=&= 21:32= 3*:**= 3*:**= 20:**= 3*:**= .2*=5*7.4)=φ= 3+:32= 3*:**= 3+:**= 3*:**= 3*:**= .2*=5*7.4)='= 3+:32= 3*:**= 3+:**= 3*:**= 3*:**= &7&)*1=λ=9.2*=5*7.4)= 3+:32= 3*:**= 3+:**= 3*:**= 3,:**= *81.*=2&97.==1&2')&= 3+:32= 3*:**= 3+:**= 3*:**= 3*:**= 9:)>=&7*&=(-&3,*8= 23=3.31= 31= 31= := ,3:8:&1=(.7(:289&3(*8= 3+= -.,7&9.43= := 31= := -*9-4)=(438.89*3(>= 3+:32= 3*:**= 3+:**= 3*:**= 3*:**= *38.9>=&7*&=02,= ,+.*= .+3= -.-= -//= +./*= 9:)>=&7*&=02,= ,+.*= 020= -.-= 3,/= ,,**= :7;*>=5*7.4)=2439-= 57.1::,:89= -&7(-:*59*2'*7= -&7(-:*59*2'*7== 57.1::,:89= 57.1::,:89= &=.2*=5*7.4)=949&1=-=49&1=9.2*=5*7.4)=<-*3=)*24,7&5->=89:)>=<&8=(43):(9*)_= ''7*;.&9.438a= = -= &3= *73&7).34= 89:)>= &7*&`= = -= .*77&= 89:)>= &7*&`= != -= *6:4.&= &3)= .3,8= !&3>43= 3&9.43&1= 5&708= 89:)>= &7*&`= $%= -= 1)47&)4=89:)>=&7*&`=&3)=$=-=$&88*3=89:)>=&7*&_= 50 ORNITHOLOGICAL MONOGRAPHS NO. 54

using the best juvenile model from φ(g + TT) step (a) above—total of four models. φ(t) 3. Using best p model from the models in φ(TT) step (2) above, run the following age and φ(T) sex structure models on φ: φ(.) (J, [S1, S2 + A]) * s φ(Latitude) with best group eff ect (J, [S1, S2 + A]) + s model from above 6 (J, [S1, S2 + A]) φ(SAB, Rest) with best group (J, [S1 + S2, A]) * s eff ect model from above 6 (J, [S1 + S2, A]) + s φ(SKC, Rest) with best group (J, [S1 + S2, A]) eff ect model from above 6 (J, [S1 + S2 + A]) + s φ(SAB, SKC, Rest) with best (J, [S1 + S2 + A]) group eff ect model from above against the following time models (all 6 age-classes): φ(SKC vs. SIE) with best group + t eff ect model from above 6. + T 4. p models to be estimated (with and * T without sex eff ect, both * and +): + TT p(g * t) * TT p(g + t) dot p(r) giving a total of 48 additional models. 5. Total models = 43. 4. Take top models from within three 6. Goodness of fi t determined with

QAICc units of the best model from step RELEASE. (3) and run them with no time eff ect on iii. Age-specifi c fecundity (b) by study area juvenile φ. analyzed in PROC MIXED in SAS. 5. Run the best 2–3 φ models from steps 1. Fixed eff ects. (3) and (4) above with the best 2–3 p a. Female age (S1, S2, A) as a fi xed models for 4–9 more additional models. eff ect. Maximum number of models to be run is b. Fixed TT model. 70 for each study area. c. Fixed T model. 6. Goodness of fi t determined with d. Fixed intercept-only model (dot). RELEASE on the basis of male and e. Fixed even–odd model (denoted EO). female groups (with S1, S2, and A pooled f. Fixed even–odd model with a linear across each sex with the juvenile portion trend (T). of CHs removed). Implication is that ĉ 2. Random eff ects with ALL of the above may be high. fi xed eff ects: ii. Conduct a meta-analysis of adult females a. Territories as a random eff ect. and males (>S2, but truncate J, S1, and S2 b. Year as a random eff ect.

histories) across fi ve study areas using QAICc 3. Structure variance as proportional error model selection criteria in program MARK. distribution, LOCAL = EXP (female age, 1. {φ(g * t * s) p(g * t * s)} global model. year, or even–odd) for the on-diagonal 2. The following φ models are to be elements. Candidate variance structures estimated against the nine p models in for off -diagonal elements are compound step (4) below: symmetric (CS) or autoregressive with φ(g * t * s) lag of 1 (AR1). Choice between those two φ(g * t + s) models will be made using AIC model φ(g * t) selection with REML in an AGE + T 27 Models model. 3. Include a sex eff ect, if necessary, from the 4. Models run will be: above 27 models with the best p model of Age + TT the 27 and run the following additional Age * TT models on φ: Age + T φ(g) Age * T φ(g + t) Age φ(g * T) Age + EO φ(g + T) Age * EO φ(g * TT) Age + EO + T POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 51

Age * T + EO i. Sierra study areas should have higher Intercept only temporal variation in survival and 5. Model selection using AIC. reproduction than SAB because of weather 6. Ten models to run for each study area. patt erns. iv. λ from Pradel model. ii. Two northern Sierra study areas (LAS 1. Truncate data sets to ﬁ rst year when and ELD) should have diﬀ erent temporal “density” study area adequately variation in survival and reproduction than surveyed. the two southern Sierra study areas (SIE 2. Select either {λ(t) φ(t) p(t)} or {λ(s * t) φ(s * and SKC). Not sure about the SAB.

t) p(s * t)} based on AICc. b. Temporal process variation ( ).

3. Eliminate λ1 (confounded), λ2 (may be i. Within each study area.

biased), and λk–1 (confounded) from 1. Estimate for adult apparent

analysis. survival (φA) from random eﬀ ects means 4. Do variance components on best model model in MARK using estimates from above with the following structures, and model {φ(a * s * t) p(a * s * t)} if there is a

select the best of those three with AICc: sex eﬀ ect and {φ(a * t) p(a * t)} if there is a. T no sex eﬀ ect. b. TT 2. Estimate for juvenile apparent

c. dot survival (φJ) from random eff ects means 5. Total of fi ve models run for each study model in MARK using estimates from area. model {φ(a * s * t) p(a * s * t)} if there is a 6. Goodness of fi t determined with sex eff ect and {φ(a * t) p(a * t)} if there is RELEASE. no sex eff ect.

v. λ from Leslie matrix. 3. Estimate for λRJS from random 1. Use 4-age class projection matrices for all eﬀ ects means model in MARK using study areas. estimates from model {λ(t),φ(t), p(t)}. 2. Estimate age-speciﬁ c φ and b (and 4. Estimate for adult fecundity

their standard errors) from best model (bA) from intercepts-only model in PROC resulting from age-speciﬁ c survival MIXED in SAS.

modeling in MARK and from best 5. Do not estimate for φS1, φS2,

model in fecundity analysis from PROC bS1, or bS2 because of known sample size MIXED, respectively, as inputs for the limitations. matrix. c. Spatial process variation ( ).

3. Calculate juvenile (EJ|λ=1) and adult 1. Within each study area.

(EA|λ=1) emigration rates necessary to a. Estimate for bA based on achieve a stationary population (λ = 1.0). territories (owl sites).

4. Calculate juvenile (φJ|λ=1) and adult 2. Among study areas.

(φA|λ=1) survival rates necessary to a. Estimate for adult apparent

achieve a stationary population (λ = 1.0). survival (φA) from mean adult 5. Estimate SE(λ) using delta method. apparent survival ( ) computed from 6. If unable to estimate juvenile survival for each study area (see 4.a.v.(2)) and a given study area, then won’t att empt to using method-of-moments variance estimate λ. components outlined in Burnham et b. Not to be estimated: al. (1987). i. Percent territory occupancy; i. Estimate temporal covariance ii. Number of owls detected per unit eﬀ ort from model g + t in meta-analysis (assuming N works above); of apparent survival. iii. Juvenile emigration; b. Estimate for juvenile apparent

iv. Year-speciﬁ c N, from Jolly-Seber model. survival (φJ) from mean juvenile Would have used same data used to apparent survival ( ) computed

estimate λRJS; from each study area (see 4.a.v.2) and v. Change in age of new recruits through using method-of-moments variance time to evaluate change in age structure of components outlined in Burnham et ﬂ oater population. al. (1987). 5. Estimate temporal and spatial variation in c. Estimate for λRJS from demographic parameters for California Spott ed computed from each study area (see Owls across ﬁ ve study areas. 4.a.iv) and using method-of-moments a. Predictions: variance components outlined in 52 ORNITHOLOGICAL MONOGRAPHS NO. 54

Burnham et al. (1987). d. λPM represents the asymptotic change in the i. Estimate temporal covariance female population size given a speciﬁ c set of from model g + t in meta-analysis apparent survival and fecundity rates. of apparent survival. i. Does not include immigration.

d. Do not estimate for φS1, φS2, bS1, ii. Represents asymptotic conditions for ﬁ xed

or bS2 because of known sample size values of apparent survival and fecundity limitations and do not estimate for (i.e. emigration [part of apparent survival]

bA because of diﬀ erences in protocols is a function of study area size and edge to

among study areas. However, area ratio, so that λPM is a function of study

compare trends in bA across study area characteristics). areas. iii. Assume 50:50 sex ratio of juveniles. 6. Interpretation and reporting of results. e. General inferences. a. Apparent survival estimates: i. Inferences are confi ned to: i. Changes in φ over time represent changes in 1. Within the study areas. emigration, death, or both. 2. Within the study period. ii. The emigration component in juvenile 3. The territorial population of owls? This apparent survival is greater than for other interpretation needs to be reconsidered age-classes. carefully. iii. Bias from resighting heterogeneity in 4. Among study areas for meta-analyses. estimates of apparent survival for S1, S2, f. Reporting results. and A will be small. i. Lead responsibility for compiling fi nal iv. Inferences about apparent survival estimates report will be R. J. Gutiérrez. apply only to the marked population. ii. Draft report will be writt en (not submitt ed) b. Fecundity estimates: by end of October 2001. i. May be positively biased because 1. All participants will be authors. of methodology used to determine 2. All participants will review fi nal report reproductive output (number of fl edged before submission. young). a. Deadline for participant reviews; ii. Point estimates are not comparable among report will be submitt ed regardless study areas because of diff erences in if participant reviews not received protocol used on diff erent study areas. An within deadline to be established by exception is comparisons between the SIE R. J. Gutiérrez. and SKC study areas. 3. Final report will be reviewed (prior to iii. Diff erences in fecundity estimates within submission) by two outside reviewers each study area are comparable because familiar with analytical methods used bias due to protocol should be similar in report (may be vetoed by U.S. Forest among years within a particular study area. Service). Thus, trends in fecundity will be examined a. Outside reviewers will be paid to within study areas. ensure timely review.

c. λRJS represents change in number of territorial 4. Internal editor will arbitrate reviews and owls. content of ﬁ nal report. i. provides information on probability of a. Gary White selected. change over a speciﬁ ed time period, given iii. Final report will eventually be published in under conditions of study. a peer-reviewed outlet.

ii. Change based on λRJS can be due to local 1. Wildlife Monographs suggested outlet. birth, immigration, death, emigration, or 2. Order of authorship and inclusion of both. authors will be determined by group. POPULATION DYNAMICS OF THE CALIFORNIA SPOTTED OWL 53

Appendix 3 believe that Spott ed Owl populations are similar to many populations of passerine birds (DeSante 1995), Further Discussion of Projection Matrix insects (Connor et al. 1983), small mammals (Nichols Population Models and Pollock 1990), and marine ﬁ sh (Roughgarden et al. 1985, Armsworth 2002).

During the workshop, the group debated the util- Some of the recommendations to use λPM sug- ity of developing projection matrices and computing gest that that metric may provide insights about the asymptotic λPM. Subsequent reviews of the draft work- relevance of movement to population dynamics that shop report also raised that issue, so we present here cannot be obtained using λt. We do not believe that our reasoning for not estimating λPM more completely. that is true. Instead, we believe that the key issue is A central theme of our internal discussions and the re- not one of which λ to use, but of how to estimate rel- views was that λt estimated using capture–recapture evant quantities. Inferences about movement require data refl ects changes in numbers of birds resulting the ability to separate gains and losses in the estima- from all sources of loss from and gain to the study ar- tion process. Indeed, at the workshop, we worked eas. Although we believe that λt is relevant to popula- on expressions for decomposing recruitment into in tion change at the scale of the individual study areas, situ reproduction and immigration components us- we recognize that for some questions, it would be ing a multi-age version of the approach suggested valuable to separate losses from death and emigration by Nichols and Pollock (1990). But although the ap- and gains from in situ recruitment and immigration, propriate estimators were developed, that work was because Spott ed Owl populations at the scale of study not completed at the workshop, so we have no such appear to be geographically open. Thus, dynamics at estimates at this time. the scale of study areas are determined at least in part Thus, parameters that were well estimated and that by contributions from other areas. Here, we discuss were suited to incorporation into projection matrix geographic openness and the related source–sink models were age-specifi c reproductive rates and rates dichotomy and ask whether development of projec- of loss that include both death and emigration. As not- tion matrix models might be useful in increasing our ed here and in other Spott ed Owl reports, that asym- understanding of those open systems. metry in the treatment of movement would lead to During both internal workshop discussion and projection matrix results that were of litt le use. Thus, external reviews, it was noted that computation of λt previous eff orts have adjusted or corrected estimates was inadequate to determine whether an area was a of juvenile survival to remove permanent emigration “source” or a “sink.” It is fairly clear that the studied as a source of loss on the basis of poor estimates of Spott ed Owl populations do not correspond to the juvenile dispersal. The rationale was that removal of strict source–sink model system of Pulliam (1988). movement from projection matrix entries would yield However, we understand that most ecologists no lon- inferences about λ that were based on only reproduc- ger view those terms in a strict manner, and that the tion and mortality, thus providing inferences about term “source” has come to mean an area that supplies whether the populations would decline or increase if recruits to other locations. Similarly, we believe that they were geographically closed. Projection matrices many ecologists view a “sink” as an area in which used previously for Spott ed Owl work were based on population size is not maintained strictly by recruit- closed-population modeling, assuming that all sur- ment of locally produced young. We believe that even viving individuals produced in the population exhib- those relaxed defi nitions refl ect a conceptual frame- ited the stage-specifi c survival and reproductive rates work that may not be especially useful when consider- of the area. So even when we “correct” estimates of ing open systems such as those of Spott ed Owls at the apparent juvenile survival to remove the movement spatial scale of study. Indeed, we believe that at least component, the projection matrix is assigning sur- four of the fi ve Spott ed Owl study areas are sources in vival rates and reproductive rates of the study area to the sense that many juvenile owls emigrate and likely the surviving juveniles in subsequent years. However, recruit to the breeding population elsewhere. We also the reality of the modeled system is that juveniles are recognize that those four study areas represent sinks thought to move elsewhere and to experience the vital in the sense that many if not most birds recruited to rates of the populations into which they recruit. Thus, the breeding population come from elsewhere and are we believe that closed-population projection matrix not produced on the study areas. Thus, the source– results are much more of an abstraction of the dynam- sink dichotomy does not seem to be especially useful ics of open systems than is generally realized. for describing that system. We suggest that Spott ed The above discussion is not intended as a criti- Owl populations, as defi ned by the scale of study, are cism of the use of traditional projection matrices in bett er viewed as open-recruitment systems in which a general but simply argues that we do not expect them substantial fraction of the recruitment to the breeding to yield useful inferences for geographically open population is by birds that are not produced on the systems. One of the report reviewers thus suggested study area. In that sense of geographic openness, we that we modify the projection matrices to incorporate 54 ORNITHOLOGICAL MONOGRAPHS NO. 54 immigration (e.g. see Cooch et al. 2001). As noted, we In summary, we agree with reviewers that more did not have reliable estimates for such incorporation, detailed inferences about movement would be valu- and matrix asymptotics (e.g. projected λ) would de- able. In particular, the ability to decompose gains pend heavily on the magnitude of the immigration. A and losses to study populations has potential to yield more satisfying approach would be to use multi-site increased understanding of those systems. However, projection matrices (e.g. Schoen 1988, Lebreton 1996) we disagree with reviewers that that problem can to include not only the dynamics of the study popu- be dealt with via modeling with existing estimates, lations but also of the population(s) with which they in particular via computation of λPM as in previous are connected via movement. Still another alternative Spott ed Owl reports. If we did choose to model the is to use open-recruitment models similar to those study systems, we would select diff erent model developed for marine systems (e.g. Roughgarden et structures that were more appropriate for those open al. 1985). The central point here is that we have con- systems. More importantly, we view that problem of sidered alternative modeling approaches and have insights about movement as fundamentally a problem some ideas about how to proceed, but we believe that in estimation rather than modeling. We see no reason it makes litt le sense to implement those approaches to expect model-based asymptotics from projection now in the absence of estimates of the relevant move- matrices of poorly estimated (even guessed) vital ment parameters and possibly vital rates of connected rates to yield reliable insights. Our focus should thus populations. be on estimation of movement-related parameters. ORNITHOLOGICAL MONOGRAPHS (Continued from back cover)

No. 35. Ecogeographical Variation in Size and Proportions of Song Sparrows (Melospiza melodia). J. W. Aldrich. x + 134 pp. 1984. $15.00 ($12.00). No. 37. Avian Monogamy. P. A. Gowaty and D. W. Mock, Eds. vi + 121 pp. 1985. $15.00 ($12.00). No. 38. An Analysis of Physical, Physiological, and Optical Aspects of Avian Coloration with Emphasis on Wood-Warblers. E. H. Burtt, Jr. x + 122 pp. 1986. $15.00 ($12.50). No. 39. The Lingual Apparatus of the African Grey Parrot, Psittacus erithacus Linne (Aves: Psittacidae): Description and Theoretical Mechanical Analysis. D. G. Homberger. xii + 236 pp. 1986. $30.00. No. 40. Patterns and Evolutionary Significance of Geographic Variation in the Schistacea Group of the Fox Sparrow (Passerella iliaca). R. M. Zink. viii + 119 pp. 1986. $15.00. No. 41. Hindlimb Myology and Evolution of Old World Suboscine Passerine Birds (Acanthisittidae, Pittidae, Philepittidae, Eurylaimidae). R. J. Raikow. viii + 81 pp. 1987. $15.00. No. 42. Speciation and Geographic Variation in Black-tailed Gnatcatchers. J. L. Atwood. vii + 74 pp. 1988. $10.00. No. 43. A Distributional Survey of the Birds of the Mexican State of Oaxaca. L. C. Binford. viii + 418 pp. 1989. $20.00. No. 44. Recent Advances in the Study of Neogene Fossil Birds: I. The Birds of the Late Miocene- Early Pliocene Big Sandy Formation, Mohave County, Arizona (K. J. Bichart); II. Fossil Birds of the San Diego Formation, Late Pliocene, Blancan, San Diego County, California (R. M. Chandler). vi + 161 pp. 1990. $20.00. Nos. 45 & 46. Descriptions of Thirty-two New Species of Birds from the Hawaiian Islands: Part I. Non-Passeriformes (S. L. Olson and H. F. James), 88 pp.; Part II. Passeriformes (H. F. James and S.. L. Olson), 88 pp. 1991. Bound together (not available separately). $25.00 ($22.50). No. 47. Parent-Offspring Conflict and Its Resolution in the European Starling. E. Litovich and H. W. Power. 71 pp. 1992. $15.00 ($12.00). No. 48. Studies in Neotropical Ornithology Honoring Ted Parker. J. V. Remsen Jr., Ed. xiv + 918 pp. 1997. $49.95 ($39.95). No. 49. Avian Reproductive Tactics: Female and Male Perspectives. P. G. Parker and N. T. Burley, Eds. v + 195 pp. 1998. $20.00 ($16.00). No. 50. Avian Community, Climate, and Sea-Level Changes in the Plio-Pleistocene of the Florida Peninsula. S. D. Emslie. iii + 113 pp. 1998. $20.00 ($16.00). No. 51. A Descriptive and Phylogenetic Analysis of Plumulaceous Feather Characters in Charadriiformes. C. J. Dove. iii + 163 pp. 2000. $19.95 ($15.96). No. 52. Ornithology of Sabah: History, Gazetteer, Annotated Checklist, and Bibliography. F. H. Sheldon, R. G. Moyle, and J. Kennard. vi + 285 pp. 2001. $25.00 ($22.50). No. 53 Evolution of Flightlessness in Rails (Gruiformes: Rallidae): Phylogenetic, Ecomorphological, and Ontogenetic Perspectives. B. C. Livezey. x + 654 pp. 2003. $10.00 ($9.00).

Order from: Buteo Books, 3130 Laurel Road, Shipman, VA 22971, 1-800-722-2460; E-mail [email protected]; or www.buteobooks.com. Prices in parentheses are for AOU members.

For a complete list of Ornithological Monographs including both in-print and out-of-print books, please visit the American Ornithological Union website at www.aou.org.