Proceedings
4th Microcomputer Applications in 
Fish & Wildlife Conference
October 24-27, 1999
Stateline, Nevada

Statistical Analysis Software, Techniques, And Applications for Spatial Data

Martha H. Cavit, Alaska Biological Science Center, US Geological Survey, 1011 E. Tudor Rd., Anchorage AK 99503

The Map Oriented Tracking Editor (MOTE) uses ARC/INFO for UNIX to allow investigators to edit raw satellite location data. Our data comes from Collecte, Localisation, Satellites (CLS) and Service Argos, Inc. whose satellites and computer systems process satellite transmitter signals into location and sensor data. Satellite transmitters or Platform Transmitter Terminals (PTT’s) have become common tools in the study of ocean and air currents and in studies of movements of animals living in remote areas. After overcoming the challenge of attaching the PTT to the animal, the researcher needs to reduce the voluminous data streams coming from each PTT to meaningful data. MOTE is a two stage editing process designed to allow a user to choose the locations that are most reliably correct and to remove those that are erroneous or suspect.

Three problems are addressed by the MOTE approach. . First, the many variables in the system used to acquire satellite data can and often do generate erroneous locations. In fact if an animal were to traverse all of the points in the Argos data, it would have to move at speeds of several thousand kilometers per hour. Clearly, the erroneous data needs to be removed. Second, in about 6% of the data, we have found that unreasonable rates of speed result from using the wrong mathematical solution to the location calculations. Because these solutions consist of a pair of possible locations, simply choosing the other solution will salvage data that would otherwise have been deleted. Finally, turning the transmitter on and off to increase battery life, results in clumps of highly correlated data. For statistical reasons, it is often desirable to reduce these clumps to a single location for each transmission period or "duty-cycle".

MOTE1.aml solves the problem of using the wrong location-solution by displaying the locations, one pair at a time on a background map, allowing the user to choose the correct one. MOTE2.aml solves the other two problems. The locations are displayed on a map, one duty cycle at a time. The researcher deletes the erroneous data and selects the best or most representative location for each duty cycle using up to 11 attributes as well as rate and distance to guide the choice.

There are several advantages to this approach not found in other satellite data editing techniques. It is extremely flexible. The researcher can use all of his/her knowledge of the animal to determine the validity of the locations even if the rule used is too complex to automate. More data is retained. Some behaviors worth studying inherently cause low quality location data, creating a dilemma if location quality alone is to be used for editing. Questions involving shorter time spans can be addressed. Movements during a duty cycle as well as between duty cycles can be studied from the resulting database. Finally, the MOTE interactive editing process provides the user the opportunity to develop insight into the animal’s behavior that might not be obvious from a more automated process.

Application of the Fast Fourier Transform for Bootstrap Confidence Intervals on Kernel Density Home Range Estimators: Methods and Computer Software

John W. Kern* and Trent L. McDonald, Western EcoSystems Technology Inc., 2003 Central Ave., Cheyenne WY 82001

Steven. C. Amstrup, USGS/BRD 1011 Tudor Road, Anchorage AK

Recognition that polar bears are shared by Canada and Alaska prompted development of the "Polar Bear Management Agreement for the Southern Beaufort Sea." This agreement mandated that the harvests of polar bears from the southern Beaufort Sea (SBS) would be split between Inupiat hunters of Alaska and Inuvialuit hunters of Canada. It also dictated that the quota for each jurisdiction would be reviewed annually in light of the best available scientific information. Based on radio-telemetered bears it is known that the utilization ranges of bears overlap jurisdictional boundaries. In order to maintain accurate harvest estimates from each population, a technique is needed to estimate the probability that a bear comes from a particular jurisdiction given the location from which the bear was harvested.

In this paper, we develop new statistical/computational methods, and software to estimate the probability that a bear sited (harvested) at a particular location belongs to a particular population using radio telemetry data. The proposed statistical method relies upon a Kernel smoothing technique analogous to kernel density home range estimation, although the experimental unit is based on the number of radio-collared individuals as opposed to the number of relocations. Confidence intervals are developed using a bootstrapping approach where the telemetered bears are resampled with replacement and probability of membership in a particular population is recalculated using kernel smoothing over a set of regularly spaced pixels for each resampling event. Kernel density estimators are computationally expensive to calculate, and conventional computational methods would make bootstrapping prohibitive. To avoid this computational bottleneck, we applied the Fast Fourier Transform to efficiently perform the kernel smoothing for each bootstrap sample. The technique is also applicable to individual animals where bootstrap samples are developed by resampling individual telemetry relocations with replacement. This technique can be used to develop confidence intervals for the utilization range.

Kevin Sallee, Ecological Software Solutions, 3154 53rd Street, Sacramento, CA 95820, Tel 916-456-5617 FAX 916-456-5616, ess.mail@ecostats.com

Biotas designed to facilitate the descriptive and statistical analysis of biogeographical data on a PC using Windows®. Biotas offers a extensive set of graphical, charting, and statistical techniques for exploratory analysis and extracting statistical inferences from spatial data quickly and easily. Possible analyses with Biotas include creating home range contours, testing for habitat use, and spatial and temporal statistics for spatial data and geographical coverages. Presentations of possible home range analyses using Biotas will include both traditional estimators such as the minimum convex polygon and highlight more recent methods such as the adaptive kernel and Voronoi polygons. Preliminary home range data analysis such as utilization curves will also be presented.

Habitat use analysis such as compositional analysis may also be performed directly in Biotas. Biotas offers additional analytical tools for the spatial analysis of point patterns with nearest neighbor analysis, spatial association and co variation, tests of complete spatial randomness, temporal and spatial autocorrelations, permutations and simulation capabilities, and programmability. Biotas is also a GIS program and may be used to display geographical coverages or determine summary statistics on landscape polygons such as perimeter to area ratios, landscape complexity indices, and polygon summary statistics.

Joel Sartwell, Missouri Dept. of Conservation, 1110 S. College Ave., Columbia, M0 65201, (573) 882-9880 x3245, sartwj@mail.conservation.state.mo.us

We developed a software package to ease the collection, analysis, and management of triangulation and survival data for animal telemetry studies. This Window-based, highly graphical system provides an the capability of being used in real-time field data collection activities using a laptop through analysis of animal dispersal and survival in either a laptop or desktop environment. Telemetry bearings, individual sightings, and internally generated location estimates are stored graphically as dynamic data overlays on base maps. Locations from telemetry bearings can be estimated using maximum likelihood, Huber m-estimator, or Andrews-m estimators. An easy to use built-in query builder allows users to easily cull data for plotting on a map or to perform other operations, such as batch location estimation, convex polygon plotting, or even staggered entry Cox Proportional Hazard or Kaplan-Meier survival analysis. This system provides an interface between ODBC compatible databases, Geographical Information Systems (GIS), and Fortran programs for triangulation, survival, and animal dispersal analysis. The software allows for the import and incorporation of GIS map layers from a wide variety of spatial data formats including: ArcView Shape, MapInfo Vector, AutoCad DWG, TIFF Image, Windows BMP, JPEG, GIF, PCX, etc. It also can incorporate any ODBC compatible database system. The use of an open-database system, which includes practically all commercially available database software, allows the user and this software package to take full advantage of Structured Query Language (SQL) functionality and facilitates transfer of data to other software systems. This software provides a full set of database management and analysis applications that meet biologists needs as they conduct wildlife telemetry projects.

Stephen J. Brady*, USDA NRCS Natural Resources Inventory & Analysis Institute , 3825 E. Mulberry St., Ft. Collins, CO 80524, 970-498-1744; sbrady@tasc.usda.gov

Curtis H. Flather, USDA FS Rocky Mountain Research Station, Ft. Collins, CO 80524

Dean M. Thompson, USDA NRCS Natural Resources Inventory & Analysis Institute, ISU, Ames, IA

A developing theoretical basis in landscape ecology coupled with advancing inventory techniques, geographically-based information systems, and computer technology have improved our ability to conduct regional and national assessments of natural resource issues. The 1997 National Resources Inventory (NRI) included a new data element to capture the spatial configuration of habitat features at each of nearly 765,000 sample locations. It is now possible to use the NRI to quantify spatial patterns by hydrologic, physiographic or political units and associate them with extant data on related resources. The addition of spatial measures expands the NRI from an inventory of national assets to a data set useful for analysis and modeling of ecological relationships as data on composition are enhanced with configuration. We describe computer assisted survey information collection, data elements for habitat composition and configuration, and demonstrate a series of spatial measures useful for the assessment of species-habitat relationships and for modeling the effect of USDA farm and other large-scale programs.

Frederick A. Leban* and Edward O. Garton, Department of Fish & Wildlife Resources, University of Idaho, Moscow, ID 83844-1136, (208) 885-6434

Michael J. Wisdom and John G. Kie, USDA Forest Service, Forestry and Range Sciences Laboratory, 1401 Gekeler Lane, La Grande, OR 97850

We investigated the effect of sample size (number of animals and number of locations per animal) on 5 methods of analyzing resource selection (Neu, Friedman, Johnson, compositional analysis, and compositional analysis on ranks). Forty-two female elk (Cervus elaphus) were intensively monitored at the Starkey Experimental Forest and Range in northeast Oregon from April to November 1994. We separated the observations into 4 time periods (day, dusk, night, dawn) for 2 seasons (spring, summer). We systematically resampled elk locations by varying the number of animals (5, 10, 20, 30, 42) and the number of locations per animal (10, 20, 30, 50, 100, all), and calculated the percentage of correct conclusions (accuracy) for 1,000 runs for 6 resource types (aspect, distance to open roads, distance to cover, distance to forage, % canopy closure, and vegetation). Elk selection patterns were more different among periods of the day than they were between seasons. Accuracy increased with increasing number of animals and increasing observations per animal for all variables. However, accuracy was low (<60%) for few animals (5 or 10) with only 10 observations. We recommend a minimum of 20 animals with 50 observations each during a time of day to adequately determine resource selection for a population during a single season.

NOTE: * indicates presenter