Session Information
| Session | Poster Session | | Date | Monday (2008-04-07) | | Time | 5:00 PM - 7:00 PM | | Room | Grand Terrace |
Presentation Information
| Presenter | Peter Wolter | | Title | A multi-scale approach to remote determination of forest canopy size in Minnesota using SPOT, ASTER, and Landsat | | Affiliation | University of Wisconsin - Madison | | Authors | Peter Wolter, Philip Townsend | | Keywords | Canopy size, Geostatistics, Minnesota, Multi-resolution, Satellite data | | Presentation Type | Poster | Abstract:
Determining tree canopy size, percent cover, and other forest structural measures from satellite image data has been a persistent goal of remote sensing scientists. From a landscape management perspective, the ability to remotely and efficiently gather such basic inventory data on a regional scale carries with it the potential to revolutionize ecosystem management and decision making. From an ecological perspective, automated large-scale measurement and mapping of canopy attributes will facilitate efficient validation of both regional and global scale models of terrestrial carbon cycling and forest response to environmental change. LIDAR imagery is generally considered ideal for measuring forest structure and subsequently estimating biomass, but it does not provide direct measures of canopy size, nor is it currently practical for frequent, regional monitoring. Thus, satellite-based solutions are still being sought.
Past exploration of canopy texture measures using 20 and 30m imagery have failed to yield usable data for basic modeling or mapping applications in the Upper Midwest due tree sizes that are typically only a fraction of the pixel resolution. However, recent theoretical studies have shown that the ratio of forest canopy sill variance measured at two different spatial resolutions should be indicative of canopy size. However, the sole test of this technique using actual imagery for a southern hardwood forest yielded poor results due to both homogeneous canopy structure and high crown closure limiting image contrast. In general, 100% canopy closure in northern Minnesota is uncommon. Thus, these forests presented an excellent test of a geostatistical method to map forest structure for a diverse range of sub-boreal to temperate forest types. Results using simulated 5 and 10m imagery generated from panchromatic Quickbird data (0.61 m pixels) indicate that canopy semivariance measured at the sill for varying tree crown sizes (2.2 to 10.2 m) are spatially distinct. At 14m image resolution, the two smallest canopy sizes were not directly distinguishable from each other, but were still distinct from the larger canopies.
Here we present test results of this geospatial ratio technique for northern Minnesota forest conditions using satellite data collected at 5, 10, 15, and 30m spatial resolution. By calculating a proxy for actual stand sill semivariance at each resolution, the specific stand-wise determination of sill semivariance is bypassed, which greatly reduces the need for extensive ground-based, parameterization data that is a requirement of the original method.
|
|