The Geomatics and Landscape Ecology Research Laboratory at Carleton University Doug King, Geography and Environmental Studies Lenore Fahrig, Biology Kathryn Lindsay, Canadian Wildlife Service Presented to the Canadian Remote Sensing Society, Ontario Chapter, April 5, 2005.
Presentation Structure Overview of the lab Development context Backgrounds of the co-directors Research issues and questions in landscape ecology related to habitat and species conservation Initial projects Research directions
The Geomatics and Landscape Ecology Lab
The Geomatics and Landscape Ecology Lab Objectives Remote Sensing Equipment and Data 1. Develop methods for Multispectral digital camera characterizing critical habitat required by species of concern Spectroradiometer and targets (i.e., identify where critical E. Ontario Landsat data habitat exists) High resolution satellite data 2. Develop models relating species abundance and Field Equipment persistence to habitat Truck and car availability across landscapes Hemispheric photography (i.e., estimate how much Trac habitat is required). Real-time differential GPS Laser rangefinders Other ecology
Backgrounds of the Co-directors
D. King: Research Program Develop remote sensing methods for evaluation and monitoring of forest and wetland composition, structure, and health at local to regional scales
Sensor Research 1983-90: Video sensor 1990: Kodak Megaplus 1991-00: Kodak DCS CIR 1993-99: Digital Camera 1300 x 1024 pixels 8-bands, 10nm bandwidth 2002+ : Duncantech MS3100 and 4100
Evaluation of Forest Structure Condition at an Acid Mine Site 1993-2003 with: J. Levesque, N. Walsworth, E. Seed, I. Olthof, C. Butson,Y. Cosmopoulos In boreal forest adjacent to 350ha of acid tailings Runoff pH = 1.4 - 2 High metal concentrations Landsat
1. Growth-mortality analysis of individual trees using archived air photos (1961 – 1991) Trend of increased forest opening within 200m of tailings, with poor regeneration Beyond 200m from tailings, forest is more stable; gradual infilling of conifers
Digital camera with filter wheel: 10 nm bandwidths
2. Multiple regression modelling of forest structure and health Best image predictors of forest variables are spatial • Semivariance range of image brightness and of image fractions (vegetation, shadow, wood) • Texture, particularly for individual crown health Additional spectral information in models • NDVI • Shadow and sunlit crown fractions • Brightness of shadow fraction more robust with view angle than shadow fraction itself for modelling LAI e
3. Development of a Multivariate Health Index Canonical Correlation showed: Decreased LAI and Cover % + increased blow down and standing dead forest volume significantly associated with Increased image texture, texture variation, and crown edge shadow + decreased deep shadow, sunlit crown and NIR brightness. Used 1997 index to predict 1999 condition Stable and improved condition well predicted Deteriorated forest close to tailings (more open) not well predicted
Ice Storm Damage Modelling and Mapping 40 - 110mm of freezing rain across E. Ontario, S. Quebec, and NE. U.S.
Site 1: Local Damage and Structure Modelling 1998 + with: E. Seed, P. Pellikka (PDF), R. Bemrose Kodak DCS 460 CIR Conducted similar modelling of structure and health as at mine site
Current Studies Temporal analysis of vertical forest structure in relation to damage Up to 12m from hemispheric photos Evaluation of the impact on biophysical models of ortho-rectification, topographic brightness, BRDF and optical corrections
Site 2: Regional Damage Mapping 1998-2002 with I. Olthof and OMNR Damage 0-25% + + 25-50% >50% • Compared 4 classifiers: Plot Satellite NN performed best • Used pre and post storm model Landsat, distance to + forest edge, elevation, ice accumulation. Other environmental Damage map
Forest Regeneration Assessment 1996+ with: D. Pouliot, K. Haddow, OMNR, CFS Includes Automated conifer (crop) tree detection, delineation and measurement Modelling the amount of competing deciduous vegetation Species classification
Study Site 1 Kodak DCS 420 CIR 2.5cm pixels OMNR/CFS regeneration experiment Sault Ste. Marie, Ontario
Tree Detection-Delineation Algorithm First developed using this well controlled experimental set-up Apex Spruce Detection Automated delineation Pine (yellow) vs manual (blue)
Site 2: Operational Cutovers Apply detection-delineation algorithm to operational regeneration Three sites of varying competition Develop method for leaf-off competition assessment Based on thresholds of woody stem density used in silvicultural treatment decisions
Imagery and Tree Delineation Camera mounted at end of boom Duncantech MS3100 images of field plots with conifers delineated
Site 3: Fallingsnow Control: no treatment OMNR led, with D. Pouliot Herbicide (Vision) treatment
L. Fahrig Research Background Pioneered spatial simulation modelling of population dynamics Includes intensive field study for model development and validation Conducts multi-landscape, multi-scale empirical analysis to assess hypotheses of species response to landscape structure and composition Studies abundance, distribution and persistence of organisms Co-founder of the J. Conservation Ecology
L. Fahrig (cont’d) Current research interests: Impacts on population distribution or species richness of: Landscape fragmentation Increased road and traffic density • Includes determining which species are most vulnerable to roads, and what road patterns are least damaging to wildlife populations Other disturbance types Relations between organism dispersal behaviour and landscape structure
K. Lindsay Research Background Pioneered landscape level studies in agriculture of effects of herbicides on bird abundance and species richness Previous team leader in Terrestrial Ecology Research at EPA Member of SARA advisory teams for various species Adjunct professor in Biology and Geography-Env. Studies Conducts research in Population viability analysis Population modelling using environmental and other variables Biodiversity conservation network design Includes research on socio-economic role of landscape ecology in conservation and resource management policy
General Research Approach
Research Issues Over-arching Research Question How do species respond to landscape structure and disturbance? Example responses: Abundance, or presence/absence • of a given species, its predators, or a pathogen that affects it. Small mammal movement success Each response needs to be analyzed spatially, and often temporally Need geospatial information on landscape composition and structure Examples: Forest and open habitat composition and cover Road and traffic density Landscape heterogeneity and fragmentation Density of forest/agriculture edge
Typical Landscape Ecology Design Species response variable(s) (e.g., abundance) measured at focal points (e.g., ponds, forest sites) Landscape predictor variable(s) (e.g., traffic density) measured at multiple scales in landscapes around each focal point Leopard Frog Abundance at Focal Ponds vs. Traffic landscapes Density in Landscapes within 1.5 km of Focal Ponds 4 focal point 3 Residual Standardized Leopard Frog 2 Population Abundance landscape 1 scales 0 -1 -2 -3 0 1 2 2.8 3.2 3.6 4 4.4 4.8 Kilometers Log (Traffic Density)
Example Landscape Ecology Study Effect of Landscape To test hypothesis, need geospatial data: number of large dead trees within 1 km of Structure on Raccoon each raccoon sampling point Occurrence landscape maps that differentiate cover by potential raccoon food Peak occurrence found in • (grain fields) vs. other open habitats (forage moderate forest cover within 1 crops, grazed lands) km of raccoon sampling sites raccoon occurrence Hypothesis: due to balance between availability of den sites (e.g. hollow dead trees) and availability of food (e.g. corn fields). forest cover in landscapes within 1 km of raccoon sampling sites
Initial Collaborative Projects in the Lab
Hooded Warbler Habitat Mapping with J. Pasher, K. Lindsay Hooded Warbler found only in SW Ontario Field measured association of forest gaps with nesting sites Non nest site Nest site
Hooded Warbler Habitat Mapping Landsat images for June and early September Non-forest masked out Landsat 5 Relatively calibrated Significant differences in image: spectral brightness textures fractions of shadow, soil and coniferous Classification of forests of SW Ontario Habitat Patterns of known nest and non nest areas well classified in focus forest Validation nest areas throughout SW Nests Ontario well classified (> 80% accuracy) Validation non-habitat areas not well Non nest areas classified
Wetland Classification and Productivity Mapping with K. Dillabaugh, K. Lindsay, L. Fahrig Riparian marsh classification and biomass modelling using Ikonos
Butterfly Occurrence Prediction using Environmental Modelling with T. Nguyen, K. Lindsay Model occurrence of butterfly species in the A2A corridor using climate, topographic, and land cover data.
Recommend
More recommend
