240 likes | 637 Views
Counting Animals from Space:. Chapter Two Transitions from Captivity to Wild Places. Scott Bergen & Eric Sanderson. Why Count Wildlife?. Fundamental to Conservation Foundational to Population Studies Federal Programs Spend Millions of Dollars Annual to Count Animals. Nov. 10, 2004.
E N D
Counting Animals from Space: Chapter Two Transitions from Captivity to Wild Places Scott Bergen & Eric Sanderson
Why Count Wildlife? • Fundamental to Conservation • Foundational to Population Studies • Federal Programs Spend Millions of Dollars Annual to Count Animals
Nov. 10, 2004 10:52:45 am 35 people involved 21 keepers 15 Volunteers 28 Enclosures mapped for individual animal locations 300 Faux fur targets placed in 4 ‘habitats’ Digital Globe Inc ©
Ground –vs- Sky Digital Globe Inc ©
Information Shadow Digital Globe Inc ©
Tallying Identification by Species Logit (identified targets) = -3.666 + 0.019(Color) + 0.970(Size) - 0.230(VegHt) - 0.421(Shade).
Counting Animals • Most reliable estimate use transect with repeat measures • Population estimates w/ standard deviation • Findings usually extrapolated from small area to available habitat or other limiting feature • Costly to count animals on ground • Remote sensing rarely used (aerial imagery) • Time and scale rarely match satellite scale & time • Mismatch in terms of time and location in reference to identifying- verifying high spatial resolution satellite imagery
Why the National Elk Refuge? • Reliable elk & bison congregations during winter • Logistic regression equation shows good fit for size, color, vegetation and shadow • Annual census of both elk and bison
Animal Count Comparisons • Refuge level, elk (weekly), bison (annual) • Ground census estimate @ time of satellite acquisition • Panoramic photo estimate @ time of satellite acquisition • Heads up digitizing estimate • Object oriented estimate
Jackson Wyoming • Access limited • Freakin’ cold -20f • Snow bleaching histogram of sensor Digital Globe Inc ©
Ground Census of Elk Group • High Ground limited • Limited by distance • 1360 individuals • 60/40 female- male ratio
Thick In Elk Digital Globe Inc ©
Panaramic Resesults • Verified over 1,000 elk sex, position, direction position in less than 10 seconds • Estimated 1070 individuals • 679 females, 299 males, 89 ? • Knew there were more but individuals > 1km were not identifiable as well as those totally blocked by other elk
Heads up • 1503 individuals Digital Globe Inc ©
Object Oriented Approach • Scale based segmentation • > classification • > revision • >classification • Hierarchical strutured • Means both smaller and larger Digital Globe Inc ©
Segmentation • Adds new dimensions to data • Area, spectra, variability within polygons • Adjacency • Contextual • Generates data Important to distinguish animals and differentiate types of animals
Initial Classification • Good Results • Identified 1540 individuals • Misidentification within riparian areas • Grouped elk in close proximity
Classification • Refined with an area classifier • 1482 individuals • Further refinement, standing – sitting, elk vs bison, sexes in bison
Summary of Animal Counts • Park Estimate: 4,900 elk, 951 bison • Ground Estimate: 1,360 elk, 60/40 f/m • Panoramic: 1,071 elk, 69/31 f/m • Heads up: 1,503 elk • 1st Object Oriented Class: 1,540 • Ob. Orient w/ Area: 1,480