HR: 1340h
AN: B43B-0278 [Abstracts]
TI: Wavelet Estimation of Plant Spatial Patterns in Multi-temporal Aerial Photography
AU: * Strand, E K
EM: evas@uidaho.edu
AF: University of Idaho, Department of Rangeland Ecology and Management,
College of Natural Resources,
975 West 6th Street,
University of Idaho, Moscow, ID 83844-1142
United States
AU: Vierling, L A
EM: leev@uidaho.edu
AF: University of Idaho, Department of Rangeland Ecology and Management,
College of Natural Resources,
975 West 6th Street,
University of Idaho, Moscow, ID 83844-1135
United States
AU: Smith, A M
EM: alistair@uidaho.edu
AF: University of Idaho, Department of Forest Resources,
College of Natural Resources,
975 West 6th Street,
University of Idaho, Moscow, ID 83844-1133
United States
AU: Bunting, S C
EM: sbunting@uidaho.edu
AF: University of Idaho, Department of Rangeland Ecology and Management,
College of Natural Resources,
975 West 6th Street,
University of Idaho, Moscow, ID 83844-1135
United States
AU: Hann, D B
EM: hannd@tcd.ie
AF: Trinity College, Department of Mechanical and Manufacturing engineering,
Parsons Building, Trinity College
, Dublin, 2
Ireland
AU: Gessler, P E
EM: paulg@uidaho.edu
AF: University of Idaho, Department of Forest Resources,
College of Natural Resources,
975 West 6th Street,
University of Idaho, Moscow, ID 83844-1133
United States
AB:
Abstract Spatial wavelet analysis represents a powerful set of image processing techniques that has considerable potential to
quantify ecologically relevant patterns at multiple scales. We utilized 1-m panchromatic aerial photography from 1939 and
1998 to automatically detect the location and crown diameters of western juniper (Juniperus occidentalis) plants as they
encroach upon a sagebrush (Artemisia spp.) steppe landscape via 2-dimensional wavelet analysis. We further demonstrate how
wavelet analysis can be used to estimate plant cover, above-ground carbon stock and net C sequestration rate in above-ground
woody tissue. The juniper crown diameters derived from wavelet analysis produced a strong correlation with crown diameters
measured via comparable hand-digitizing in a geographic information system (r=0.96, n=69) with a 5% commission and an 8%
omission error. When compared to field data we found that plants with crown diameters three times larger than the image pixel
size were accurately depicted by the wavelet technique (r=0.86, n=60) with a 19% omission and 0% commission error.
Similarly, the GIS hand-digitizing yielded a 15% omission error and 0 % commission error (r =0.92, n=65). Plants with a
crown diameter smaller than 3 m were not detected by the wavelet method nor were they possible to record in a GIS, however
these small plants comprised only 4% of the woody carbon across the sampled landscape. We further compared the wavelet
estimated plant cover to field data for 20 plots with juniper cover ranging from 1.2 - 61.8 %. Cover estimates using
wavelets are accurate up to approximately 20% juniper cover (r=0.81, n=13), but plant cover is underestimated for plots with
a canopy closure > 20% due to crown clumping. Considering all of the 20 plots, the wavelet method estimated 10.7% total
plant cover compared to the field estimate of 13.9%. We attribute approximately 50% of the error in the cover estimate to
the inability to detect plants < 3 m in size, and 50 % of the error to misclassification due to clustering of trees in
plots with > 20% plant cover. Through comparison with historical photography, we estimate that juniper plant cover
increased 2.7 fold (from 2.7% to 7.3% total cover) during the period from 1939 to 1998 within a 15 ha pilot study area.
Employing allometric relationships for western juniper we calculate a 730% increase in above ground woody plant carbon stock
corresponding to a carbon sequestration rate of 3.0±0.6 g C m-2 yr-1. Because it is rapid, objective, and automatic, and
allows for consistent analysis of a range of image types this approach has considerable potential for the long-term
monitoring of biogeochemical properties and vegetation change via aerial photography and other fine scale remotely sensed
imagery.
DE: 4806 Carbon cycling (0428)
SC: Biogeosciences [B]
MN: Fall Meeting 2005