NCEAS INTERIM REPORT - MARCH 19, 1998

ACTIVITY TITLE: Resident Fellow - Application of Geostatistics in Ecology

PRINCIPAL INVESTIGATOR: Andrew Liebhold

Northeastern Forest Experiment Station
USDA Forest Service
180 Canfield St.
Morgantown, WV 26505
304-285-1609
304-285-1505 (FAX)
sandy@gypsy.fsl.wvnet.edu

SUMMARY: Geostatistics is a field of statistics that focuses on measuring and modeling variation through space and/or time. Geostatistical methods were originally developed for applications in the earth sciences. Recently there has been growing interest in the use of these methods for studying landscape-level ecological problems. While many ecologists have recognized the potential uses of geostsatistics in ecology, many of these applications have not occurred because there are no texts that are written in a way that ecologists can understand. Virtually all of the existing geoststatistics texts focus on geological problems. Thus, the purpose of the proposed Resident Fellow is to spend 3 months writing a text that introduces and explains geostatistics and how it can be applied to ecological problems. The text will rely heavily upon the use of examples for illustrating statistical procedures.

PROGRESS TO DATE: Liebhold was a resident fellow from June 7 - August 5, 1997. During that time he completed an outline of the geostatistics text, compiled data and information for the book, and completed a rough draft of two chapters. The PI plans to return to NCEAS as a resident fellow to continue working on the book from Nov. 15, 1998, - Jan. 1, 1999. The current outline is given below:

Introduction to spatial data in ecology
Types of data
The origins of geoststistics
Geoststistics relative to other spatial analyses and computer software
Computing questions

Exploring spatial patterns in gridded data: satellite imagery:
Introduction to AVHRR data
h-scatterplots
correlograms
variograms
nugget effect
sill
range
Anisotropy

introduction to chrimstmas bird count data
transformations
variograms
cross-correlation
trend (non-stationarity)
variogram models
types of models
fitting models

Sampling for variogram estimation
numbers of samples
clustering of samples
cell declustering

kriging
ordinary kriging
point kriging vs block kriging
indicator kriging
indicator kriging with "soft" data

simulation
unconditional simulation
conditional simulation
gaussian simulation
multiple indicator simulation
simulated annealing

other topics
how to deal with autocorrelation in tests of association
geostatistics in 1, 2, or 3 dimensions