PhD Abstract and Outline: Falk Hüttmann (Ph.D., Diplom-Forstwirt univ.) PostDoc Fellow, Simon Fraser University

Falk Hüttmann
Centre for Wildlife Ecology, Biology Dept.
8888 University Drive, Simon Fraser University (SFU), Burnaby B.C., Canada V5A 1S6
Tel: 604 291 5618 Fax: 604 291 3496 Email: huettman@sfu.ca

ABSTRACT

This thesis investigates major aspects of seabird distribution. The general topic of seabird distribution is first reviewed and presented. I then introduce the study area by stratifying the Seascape of the Northwest Atlantic using 20 environmental data sets, a Geographic Information System (SPANS-GIS), and Clustering evaluated by Classification and Regression Trees (Cart). For the first time the physical environment of the study area is described, laying the groundwork for the ‘Seascape Ecology’ approach developed in the thesis.

I evaluated existing knowledge of seabird migration with data from the PIROP (Programme Intégré de recherches sur les oiseaux pélagiques) database using monthly observations of immature and moulting seabirds to track migration. The results allow, for the first time, the identification of flyways using at-sea observations throughout the year.

I then used the 20 environmental data sets, and GIS-overlaid them with standard seabird at-sea observations for winter, including results from Principal Component Analysis to characterize seabird assemblages. Logistic regression identified significant predictors, and Cart developed descriptive models. Unknown to previous knowledge, aspect of sea floor and distance to coast were the most frequent significant predictors of seabird distributions.

For the breeding season I hypothesized that the seabird colony location is the driving force for seabird distribution. The same methods as for the investigation of winter seabird distribution were used. Specific seabird model predictions using Cart were evaluated by the nearest distance to seabird colonies, and by geo-referenced observation residuals. Results showed that breeding seabirds are attached to colony locations. Non-breeders can stay far away from colonies, and colonies can also be located in non-optimal foraging habitat. Distribution modelling allows for filling gaps in previous information on seabird distribution and colony locations.

I investigated scale issues, a major methodological question in seabird distribution research, using again the PIROP data set. The results indicated that scale had no major effect on previous PIROP analyses, but was useful for describing seabird distribution and for seabird management questions.

I have demonstrated that seabird distributions need to be examined throughout the year, and that new understanding can be achieved using abiotic parameters alone. Modern remotely-sensed data sets and statistical approaches offer unparalleled opportunities to track changing responses of seabirds to environmental change.

TABLE OF CONTENTS with page numbers

IN CASE SOME MORE BACKGROUND INFORMATION ON THE THESIS TOPIC IS NEEDED,
HERE AN EXCERPT FROM THE ORIGINAL PHD THESIS PROPOSAL:

INTRODUCTION
The question: "why is an animal at this location and not another one ?" is fundamental to ecology and has intrigued human beings for a long time. The question can be asked for one individual, a metapopulation, a single species, a (trophic) guild (e.g. sharing a foraging method) and for a community or an entire biomass.

Only 300 from app. 9000 bird species on the world are classified as seabirds, among them is one of the most abundant birds of the world , Wilsons' storm-petrel (Oceanites oceanicus). Detailed information about seabirds at sea is lacking for most of the species. Seabirds are extremely mobile animals and to answer questions on their distribution, techniques such as tagging, banding, telemetry, remote sensing, geographic information systems (GIS) are used in conjunction with modeling, statistics, linear and non-linear mathematics. Understanding distribution is fundamental to conservation biology and species management. If one wants to conserve seabirds at sea in a seabird management plan, one has to understand why they are located where they are, in order to predict where they will be located in the future and which approach has to be chosen. Explaining bird distribution is as well a key question for their ecology and can lead to a better understanding of seabird foraging strategies at sea and their role and position in the food web.

The aim of this PhD research is to characterize and to explain the distribution of seabirds at sea in terms of the characteristics of their marine environment. The study area is the Atlantic region of Eastern Canada, where seabird distribution data have been maintained by various agencies and institutes.

Due to the high number of parameters that might potentially influence distribution - the literature has identified more than 20 parameters thought to be important directly or indirectly for the distribution of seabirds - this is likely a demanding question to answer.

Which reasons are important for the distribution of species on a more or less 2-dimensional surface such as the sea ? This research should bring more light to the questions: What are the reasons and factors ? How are they linked together ? And depending on the resolution scale, which are of the most importance for objectives of study ? The use of a GIS in this research has to be considered as a tool to improve knowledge about seabirds, but not only creating colorful maps. The PhD thesis will investigate whether there is a hierarchy of parameters determining distribution or if the distribution is even randomly determined (which does not necessarily mean that there are no reasons to be randomly distributed).

The results will be discussed with regard to seabird management, assessment of impacts on marine environments and selection of marine protected areas for an international frame; and furthermore will help to evaluate data sets and fill existing gaps in research on seabirds at sea.

RESEARCH APPROACH
The following research areas are thought to be important for this study:

- Seabird biology

- Ecology

- Oceanography

- Fisheries

- Biodiversity

- (Spatial) Statistics

- GIS & Databases

1. At the largest spatial scale (study area off the Canadian east coast: Gulf of Maine to the Canadian Arctic), distribution can be explained by mechanisms caused by physical features such as bathymetry,wind, water temperature, salinity, weather and pack-ice.
All transects in the PIROP data set for the whole study area of the North Atlantic will be analysed by overlaying the available (time and spatial) information, such as profiles from bathymetry, sea surface temperature, salinity, windspeed and ice cover.
Bathymetry profiles from the ETOPO5 seafloor map can be received (depths per latitude/longitude coordinates as ASCII file) and allow further processing. The depth will be classified into three groups:shallow (<50m), medium (50m-100m) and deep (>100m).
The sea surface temperature and salinity information per transect will be achieved by analysing the pixel values from georeferenced satellite images for the transects from the National Oceanographic Data Centre (NODC). Temperature information will be classified into steps of one degree, salinity will be classified in parts 1/1000. The weather information, based on a query of the weather/sea condition column of the PIROP watch information, will be classified in wind, rain,fog, sun, snow. Additional data have to be evaluated, such as windspeed data achieved by analysing pixel values of ERS-1 satellite images.
The ice cover, derived from databases and satellite images, can be grouped into ice, no ice and ice edge with a 2 km bufferzone.
The approach taken will consider Multiple Regression, but mainly using the Classification and Regression Tree (CART) functions in Splus and Factor Analysis, if enough overlap of data can be found. For the CART technique tree pruning has to be done before each test. The level of deviance determines the number of nodes per tree, allowing for appropriate conclusions. A random selected subdata set will be used to test the results for significance, using standard techniques, such as Chi-square tests. Results from each node in the tree can be plotted with SPANS GIS, and due to the relational database system, they make also other related factors obvious. This can lead to modelling, new hypotheses and findings.

2. At the medium temporal scale (20 years), distribution and habitat use is associated with the predictable annual cycle in a seabird's life (e.g. migration, breeding, wintering), reflecting a hierarchy of distribution factors. Foraging for food is mainly following traditional knowledge of food sources.
The general biological patterns of seabirds in the North Atlantic are known and will be taken from the scientific literature. For instance, shearwaters arrive in the North Atlantic in July and disappear in late October. Phalaropes (Phalaropus spec.) pass through the North Atlantic region in spring and fall, Skuas (Catharacta skua) disperse in winter along the coast of the eastern Atlantic. Black-legged kittiwakes (Rissa tridactyla) perform a transatlantic migration to the wintering grounds off Newfoundland, Puffins (Fratercula arctica) disperse widely in the southern North Atlantic. These patterns will become very obvious by analysing the abundance of seabirds per location (latitude and longitude coordinates) in the PIROP data set for transects through the fully covered time by PIROP (1966-1992). The behaviour column in PIROP, gives further information about the activity of the observed seabirds, in particular travelling in certain directions. Analysing transects from north to south (and west to east for transatlantic migration) will show the migration fronts, by the presence of the birds and their behaviour (migrating) per transects.
Database queries of the PIROP database, using standard SQL techniques in dBASE, FoxPro and other appropriate programs will be carried out, looking for large flocks/number of seabirds resting at sea, sometimes provided with information about feather moult (following the PIROP definitions for moulting) and wintering (winter plumage, flock size, resting). Database queries in the behavioural (=> carrying food) and age (=> breeder and non-breeder) column of the PIROP data set will be done to answer these questions.
Before a breeding season starts, resting seabirds will be observed near colonies, sitting at sea and waiting to start breeding. This will be checked by looking for resting flocks in relation to colony distance. Resting areas at sea will be found, a higher number of observations per area represents migration corridors at sea.
Processed maps, plotted observations per months/season and location will make these observations quite obvious. Therefore, weighed information per location about the first and the last migratory bird (grouped) and time of biggest observation per day will be used. Overlaying the general breeding time cycle per guild will probably show, that young birds will be observed with a higher abundance after the season. Floaters, nonbreeders widely dispersed among the colonies, will be observed through the whole breeding season. Since they are not involved in the breeding activities they show a different behaviour and foraging strategy and can be found farther away from the colonies.
An approach, using a "GIS bufferzone" of foraging distance of app.100 km around a colony for breeding birds from June - August will be used (dates depend on the breeding season and location) to determine the distance to the colony. The closest distance to the nearest shore can be achieved by using standard distance functions in the GIS.
To obtain the results, this approach will use the CART technique, following the technique described previously. Since it is believed that these seabird observations are independent from oceanographical processes, no homogenous picture will be found with environmental features. Plotting the results over oceanographical features is expected to show a high variance.

3. At intermediate spatial scales such as currents, tide waves, fronts (including eddies/gyres), ice-edges, shelf areas, shelf edges, deep sea, shallow areas, closest distance to the nearest shore, distribution is explained through processes leading to easier prey (plankton, fish) catch ability, summarized by the term "trophic relations for food.
The abundance of fish and plankton per unit (currents, tide waves, fronts `including eddies/gyres`, ice-edges, shelf areas, shelf edges, deep sea, shallow areas, closest distance to the nearest shore) will be compared with the seabird abundance per unit in the spatial and time window of PIROP. The spatial and temporal constraint for superimposing the data sets will be limited by the fish- and plankton data. Techniques to allow comparisons such as averaging, blocking or Principal Component Analysis (PCA) for several satellite images through a time series will be used. Catching results from North American Fisheries Organisation (NAFO) are available for the ICES/NAFO management blocks and the PIROP time window . Towing results from St.Andrews and the Bedford Institute for Oceanography database will also be implemented.
The stratified seabird data will be overlayed with the bathymetry information from ETOPO5 seafloor maps. Location of currents will be achieved by analysing SST gradients from satellite images/NODC. The plankton distribution will be achieved by analysing CZCS (Coastal Zone Colour Scanner) satellite images. These information will be analysed with the help of a Classification and Regression Tree (CART, explained previously), so that relations of this kind become quite obvious.

4. At the smallest spatial scale (transect length and 10 min observation period), seabird distribution is reflecting the biology (plumage colour, moulting stage, predator avoidance strategy), the foraging methods (surface feeders, divers, plunge divers, parasitic feeding, offshore feeders) and habitat attraction (ice bergs, fishing vessels and floating debris).
Only the transects fulfilling the standard 10min watch length requirement for PIROP will be analyzed by the number of observed guilds (surface feeders, divers, plunge divers, parasitic feeders, offshore feeders), individuals and behavior. The information about weather, waves, fishing vessels, floating debris during the survey will be achieved with a database query from PIROP. The transect information will be split up in full transects and also in the 10min transect watches, allowing for a stratification their comparisons. Each of these results will also be compared by their expected abundance in this area from the literature, when sufficient data are available. The results, achieved by using Classification and Regression Trees (CART) will allow conclusions about the reliability of observations for different guilds and for different places and observation conditions so that a correction factor can be calculated and the "true" number of birds can be achieved for future surveys in the study area. These results allow as well for an evaluation of the PIROP data set and the counting scheme.

Since PIROP covers a coarse resolution a comparison with a smaller resolution will be carried out in the Bay of Fundy, where fieldwork in the form of (ferry) transects is conducted, following the PIROP scheme. Areas of interest are the biological rich upwellings of Brier Island, Grand Manan, the banks (Georges Banks and Brown Banks) and a transect between St.John and Digby. These data will be analyzed in addition with the tide information, available from tidal prediction software. The results will be compared with the results derived from the analysis of the PIROP database for the eastern Canadian coast and the Manomet Bird Observatory data set. The approaches for the selected tests in the other data sets will be repeated. An evaluation of these results will then be carried out. This approach will determine whether matching problems are occurring, how they could be explained biologically and how they could be avoided in an experimental design for seabird counting schemes in the future, so that the best results can be achieved.

SELECTED REFERENCES
Ashmole, N. P. (1971). Seabird ecology and the marine environment. D. S. Farner, & J. R. King (eds), Avian Biology (Vol. 1pp. 223-286). London and New York: Academic Press.

Bourne, W. R. P. (1963). A review of oceanic studies of the biology of seabirds, The Proceedings XIIIth International Ornithological Congress, p. 831-854.

Brown, R.G.B. (1977). Atlas of eastern Canadian seabirds Supplement 1 Halifax-Bermuda transects. CWS.

Brown, R.G.B. (1986). Revised Atlas of Eastern Canadian Seabirds. Beford Institute of Oceanography: CWS.

Brown, R. G. B., Nettleship, D. N., Germain, P., Tull, C. E., & Davis, T. (1975). Atlas of eastern Canadian seabirds. CWS.

Cairns, D. K. (1992). Bridging the gap between ornithology and fisheries science; use of seabird data in stock assessment models. Condor, 94(4): 811-824.

Diamond, A. W., Gaston, A. J., & Brown, R. G. B. (1986). Converting PIROP counts of seabirds at sea to absolute densities. Progress Notes No 164.

Diamond, A. W., Gaston, A. J., & Brown, R. G. B. (1993). Studies of high-latitude 3.A model of the energy demands of the seabirds of eastern Arctic Canada. W. A. Montevecchi (ed.), Occasional Paper Number 77 (p. 39). CWS.

do Carmo Gomes Manuel (1993). Predictions under Uncertainty: St.John's Newfoundland, Institute of Social and Economic Research.

ESRI. (1995). Understanding GIS, The ArcInfo Method. (Vers7,3.Edition ed.). Redlands,CA: Environmental Research Institute (ESRI) Inc.

Evans, P. G. H., & Nettleship, D. N. (1985). Conservation of the Atlantic Alcidae. The Atlantic Alcidae (pp. 427-88). London: Academic Press.

Furness, R. W., & Monaghan, P. (1987). Seabird ecology. Glasgow: Blackie.

Gaston, A. J., Collins, B. T., & Diamond, A. W. (1987). Estimating densities of birds at sea and the proportion in flight from counts made on transects of indefinte width. Occasional Paper. Ottawa: CWS.

Haney,J.C. (1989). Remote Characterization of Marine Bird Habitats with Satellite Imagery: Colonial Waterbirds, 12(1):67-77.

Hannon, Bruce, & Ruth, Matthias (1994). Dynamic Modeling: New York,Berlin, Springer-Verlag.

Hunt, G. L., Piatt, J. F., & Erikstad, K. E. (1991). How do foraging seabirds sample their environment ? Seabirds as monitors of changing marine environments: Symposium 41. Seabirds as monitors of changing marine environments (pp. 2272-2280).

INTERA TYDAC. (1992). SPANS GIS. Ontario: INTERA TYDAC Technologies Inc.

Levinton, J. S. (1982). Marine Ecology. Englewood Cliffs: Prentice-Hall inc.

Lock, A. R., Brown, R. G. B., & Gerriets.S.H. (1994). Gazetteer of Marine Birds in Atlantic Canada. CWS.

Mann, K. H., & Lazier, J. R. N. (1991). Dynamics of Marine Ecosystems. Boston: Blackwell Scientific Publications.

Martinez, N. D., & Lawton, J. H. (1995). Scale and food-web structure - from local to global. Oikos, 73: 148-154.

Murphy, R. C. (1936). Oceanic Birds of South America. Am.Nat.Hist.Mus. New York

Nettleship, D. N. (1994). CWS Seabird Colony Registry: A Computerized Data Management System for Acces to Seabird Colony Data. Science Review 1992 & '93 (p. 147). Bedford Institute of Oceanography, Dartmouth: Department of Fisheries and Oceans.

Pimm, S. L. (1982). Food webs. New York, USA: Chapman and Hall, New York.

Pinet, R. P. (1992). Oceanography, An Introduction to the Planet Oceanus. St.Paul,MN: West Publishing Company.

Platt, T., Sathyendranath, S., & Longhurst, A.(1995). Remote sensing of primary production in the ocean: promise and fulfillment: Phil.Trans.R.Soc.Lond., 348:191-202.

Schneider, D. C. (1994). Quantitative Ecology, Spatial and Temporal Scaling. Academic Press.

Shaw, D. M., & Atkinson, S. F. (1990). An introduction to the use of geographic information systems for ornithological research. Condor, 92: 564-570.

Sherman, K.(1991). The Large Marine Ecosystem Concept: Research And Management Strategy For Living Marine Ressources: Ecological Applications, 1(4), p. 349-360.

Shuntov, V. P. (1972). Seabirds and the biological structure of the ocean. Dal'izdat. Vladivostok, Russia: National Technical Information Service translation 74055032 (1974) Springfield, Virginia, USA.

Sokal, R., & Rohlf, F. J. (1995). Biometry. New York: W.H.Freeman & Co.

StatSci Division (1996). S-Plus,Guide to Statistical & Mathematical Analysis: Seatle, Washington, MathSoft, Inc.

Tasker, M. L., Hope Jones, P., Dixon, T., & Blake, B. F. (1984). Counting birds at sea from ships: a review of methods employed and a suggestion for a standardized approach. Auk, 101: 567-577.

Wiens, J. A. (1989). Spatial scaling in ecology. Functional Ecol., 3:385-397.

Wynne-Edwards, V. C. (1935). On the habits and distribution of birds on the North Atlantic. Proc.Boston Soc. Nat.Hist., 40: 233-346.

Zar, J. H. (1984). Biostatistical Analysis. Englewood Cliffs, NJ: Prentice Hall.

© 1997, 1998, 1999, 2000, 2001 Falk Hüttmann
Last updated: February 28th 2001 by Falk Hüttmann
URL: http://www.geocities.com/CollegePark/Quad/5377/webphdsm.html