ASA-ENVR
Session Slot: 8:30-10:20 Thursday
Estimated Audience Size: 40-50
AudioVisual Request: overhead & slide projector
Session Title: Statistical Multiscale Assessment of Landscapes and
Watersheds with Satellite and Synoptic Data
Theme Session: No
Applied Session: No
Session Organizer: Patil, G.P. Pennsylvania State University
Address: Center for Statistical Ecology and Environmental Statistics Department of Statistics The Pennsylvania State University 421-D Thomas Bldg. University Park, PA 16802
Phone: 814-865-9442
Fax: 814-865-1278
Email: gpp@stat.psu.edu
Session Timing: 110 minutes total (Sorry about format):
Opening Remarks by Chair - 5 minutes First Speaker - 25 minutes Second Speaker - 25 minutes Third Speaker - 25 minutes Fourth Speaker - 25 minutes Floor Discussion - 5 minutes
Session Chair: Patil, G.P. Pennsylvania State University
Address: Center for Statistical Ecology and Environmental Statistics Department of Statistics The Pennsylvania State University 421-D Thomas Bldg. University Park, PA 16802
Phone: 814-865-9442
Fax: 814-865-1278
Email: gpp@stat.psu.edu
1. Quantitative Characterization of Hierarchical Scaled Landscape Patterns
Taillie, C., Pennsylvania State University
Address: Center for Statistical Ecology and Environmental Statistics Department of Statistics The Pennsylvania State University 421-C Thomas Bldg. University Park, PA 16802
Phone: 814-865-5212
Fax: 814-865-1278
Email: taillie@stat.psu.edu
Patil, G.P., Pennsylvania State University
Abstract: When a natural landscape is cast as a categorical raster map, a multiresolution characterization of spatial pattern can be obtained whereby the entropy is computed for a finer resolution map, conditioned on the values of a coarser resolution map. After application to a sequence of rescaled maps which have increasingly degraded resolution, the conditional entropy is plotted as a function of measurement scale (resolution), thus resulting in a multiresolution profile of fragmentation patterns. For neutral landscapes that are simulated by multiresolution stochastic generating models, we present a method to directly compute conditional entropy profiles. Such profiles can provide benchmarks for comparing results obtained from raster maps of actual landscapes that are classified from satellite images. Results show that characteristic landscape types give rise to characteristic features of these fragmentation (conditional entropy) profiles.
2. Guided Transect Sampling for Assessing Sparse Populations
Ringvall, Anna, Swedish University of Agricultural Sciences
Address: Department of Forest Resource Management Swedish University of Agricultural Sciences S-901 83 Umea, SWEDEN
Phone: 46-90-7866838
Fax: 46-90-778116
Email: anna.ringvall@resgeom.slu.se
Stahl, Goran, Swedish University of Agricultural Sciences
Abstract: In guided transect sampling, prior information is used to guide the field survey within randomly laid out broad strips. Generally, the prior information consists of remote sensing data. The strips are divided into grid-cells, and a covariate value is estimated for each such cell. The covariate data could be, e.g., estimates of the biomass of hardwoods in case the intention is to survey some population which prefers such stand and site conditions. Different strategies can be used for guiding the surveyor through the strips, based on the covariate values in the grid-cells. The different strategies lead to different properties of the estimators of population size, etc. In this paper, the general principles of guided transect sampling are outlined, and results from theoretical and practical tests of the method are reported.
3. Spatial and Spectral Classification of Compressed Image Data for Landscape Analysis
Filipponi, Danila, University of Chieti, Italy
Address: until 6/1/98 Visiting Faculity Center for Statistical Ecology and Environmental Statistics Department of Statistics The Pennsylvania State University 421-B Thomas Bldg. University Park, PA 16802; after 6/1/98 Department of Quantitatiave Methods and Economic Theory Universita degli Studi ``G. D'Annunzio'', Viale Pindaro, 42, 65127 Pescara, ITALY
Phone: 814-863-8126
Fax: 814-865-1287
Email: danila@stat.psu.edu
Abstract: Remote sensing data are becoming increasingly important in environmental studies. Two limitations have been the high cost of acquiring these proprietary data sets and the disk storage and processing demands imposed by their large sizes. Data compression is one approach for addressing these limitations. Here, pixels are clustered into a fixed number of categories on the basis of their spectral response values. Quantitative response information is partially retained by recording a set of summary statistics for each cluster.Central issues are the development of methodology for analyzing the compressed data. The paper examines these issues in the specific context of supervised classification into eight landcover types and both spectral and spatial considerations enter into the final classification. A spectral classification using the Kullback-Liebler distance between clusters is proposed and is compared with a more ad hoc approach using only the information in the diagonals of the summarizing variance-covariance matrices. Spatial coherence of the resulting landcover map has been assessed by an iterative method involving the use of indicator kriging.
4. Wavelet-based Multiscale Analysis in Landscape Ecology
Li, Bai-Lian, University of New Mexico
Address: Department of Biology University of New Mexico Albuquerque, NM 87131-1091
Phone: 505-277-5140
Fax: 505-277-0304
Email: blli@unm.edu
Abstract: Wavelet analysis is relatively a new mathematical theory and computational method. Because of its characteristics of time/space-frequency localization and multiresolution, the wavelet transform of a signal can provide detailed information of underlying ecological processes in time and/or spatial scale. In wavelet representation, a signal is decomposed into a sum of elementary building blocks describing its local frequency content. This paper describes the use of wavelet analysis as an innovative hierarchical data analysis technique for investigating multiscale spatio-temporal relationships in several long-term ecological research data sets. We use wavelet transforms, wavelet variance, cross-scale correlation analysis of transform coefficients, wavelet-based change point detection, and wavelet-based local self similarity index to characterize and quantify heterogeneous landscape patterns on a wide range of temporal and spatial scales. Preliminary results demonstrate that these methods are suitable to characterize multiscale spatio-temporal data and the scaling problem. The scaling information and the proposed approach provide a new way to look at the long-term dynamics of the coupling between the terrestrial ecosystem, the hydrological cycle, and the physical climatic system, and has potential to give a detailed extrapolation across temporal or spatial scales.
List of speakers who are nonmembers: 3, Goran Stahl (replaced by Anna Ringvall 3-4), Danila Filipponi, and Bai-lian Li