In 2001, the Antarctic Biology and Medicine Program, National Science Foundation (NSF), Arlington, Va., joined the U.S. Geological Survey (USGS), Reston, Va., to investigate orbital sensor data for acquiring enumeration data of Emperor penguins in the Ross Sea area of Antarctica. The project’s ultimate goal is to monitor the effect of regional climate change on Emperor penguin habitats.

High-resolution sensor data collected by commercial satellites like DigitalGlobe’s QuickBird (www.digitalglobe.com) offer fast and continuous wide-area coverage in Antarctica. QuickBird imagery was collected over selected monitoring sites in the Ross Sea area from 2001 through 2003. To supplement the satellite data, researchers used resources from the USGS National Civil Applications Program (NCAP).

Previous attempts to develop population counts of Antarctic wildlife are well documented. In 1988, Graham Robertson, as part of the 35th ANARE expedition to Mawson, counted Emperor penguins using stepladders and cameras hung from a helium-filled balloon. Comparing counts made by observers on the ground with photographs from the remotely operated camera, Robertson and his colleagues calculated that on cold days as many as 10 Emperor penguins could huddle within one square meter of space. Huddling by male Emperors is the most efficient way of minimizing space between themselves and keeping warm during the long incubation fast, because the warmer they are the longer their fat lasts.

For the Ross Sea project, lack of low sunlight prevented the collection of satellite images of these huddle formations during the winter months of July and early August. In addition, two attempts to collect geodetic ground control data of Cape Crozier, the primary project site, were unsuccessful. Consequently, the researchers aborted orthobase development and an automated method for counting Emperors in huddles. Instead they used October and September satellite images of disbanded huddles and applied a geographic information system (GIS) point data entry and retrieval operation.

Project Test Area and Source Data
Cape Crozier is located in the vicinity of the Ross Sea. The project’s home base, McMurdo Station, is a year-round U.S. Antarctic base operated by Raytheon Corp. under contract with the NSF. Cape Crozier, as well as most sites in the Ross Sea vicinity, typically is accessible only by helicopter.

The QuickBird satellite data (sub-meter-resolution panchromatic and 4-meter-resolution multispectral images) were collected for this project during October 2002 and September 2003. Supplemental data were acquired from NCAP and the U.S. Antarctic Resource Center (US-ARC). The primary image processing and mapping components were ERDAS Imagine 8.5 from Leica Geosystems (www.gis.leica-geosystems.com), ArcGIS 8.1 from ESRI Inc. (www.esri.com) and SOCET SET 4.4 photogrammetric software from BAE Systems (www.socetgxp.com).

Data Preparation and Entry
The researchers developed several image datasets from the source material to generate integrated color raster and vector files. The overall data preparation consisted of image processing, point data entry and retrieval phases.

Image Processing
To merge the QuickBird 60-centimeter panchromatic image files, each 4-meter QuickBird multispectral image was interpolated to 60-centimeter pixels. The geometric registration routine converted the 4-band multispectral data to the same pixel size as the panchromatic image. This routine required the registration of the multispectral data using image-to-image x and y control points. Using the appropriate panchromatic image as the base file, the routine developed a planimetric transformation by selecting several well-distributed control points in the base and correlating their UTM coordinates to corresponding x and y positions in each multispectral image. These points then were applied in a least-square regression analysis to derive a second-order polynomial transformation of a representative multispectral band with less than a 2-pixel root mean square error.

Following the transformation to 60- by 60-centimeter output pixels, the researchers used a high-pass box filter to enhance higher spatial frequency components of each output color image. Due to an absence of usable geodetic ground control data, digital terrain elevation data needed to orthorectify the QuickBird images weren’t available for the project.

Point Data Entry and Retrieval
To count the Emperor penguins the researchers created a point file or overlay over each raster image backdrop. Each penguin was collected as a point that was logged and tallied automatically in a feature and attribute file. Point files were displayed individually in contrasting colors over the image backdrop and systematically inspected for proper fit. The use of “flicker” and “swipe” toggle switches to adjust the color display provided a quick visual check of the registration between the point file and the raster image. The researchers performed a systematic sampling by reviewing each point file over the image backdrop in subsections and visually comparing graphic plots of the overlays with the corresponding raster images. Any point feature in the image that appeared to be missing, altered or unidentifiable was marked for further inspection.

Visual comparison of the processed imagery with field data helped verify spatial inconsistencies such as mismatching, overcounts and undercounts. After inspecting the data collection results, the data were exported to ArcGIS for display and further verification of the results. Once compiled, the spatial database was stored on compact discs.

Review and Analysis
The research team compared each habitat between images derived from existing civilian sensors and other sources. The results varied from negligible to moderate differences in informational content and interpretability. Although QuickBird imagery has sufficient spatial data to detect Emperor penguin groups, it needs to be supplemented with NCAP data to accurately validate penguin counts. Applying both sources of imagery, counts of group size from imagery are comparable to counts made by observers on the ground.

As for positional accuracy, due to insufficient horizontal ground control in Antarctica, QuickBird can only guarantee horizontal accuracy of 200 meters from true position, and standard vertical positioning coordinates or digital terrain elevation data simply aren’t available. The team examined absolute orientation of photoidentifiable features using horizontal positional coordinates in the imagery and ground reference points. The team also compared differences between feature positions and the corresponding ground reference points. The measured horizontal difference was greater than 200 meters from the ground-level positions of the existing reference points—insufficient horizontal positional accuracy to meet large-scale map accuracy requirements.

Thus, the study’s findings show that the spatial data collection assets of the commercial satellites may not meet the stringent requirements for mapping detailed changes in penguin habitats. Achieving this goal may require low-altitude aerial photography—particularly airborne, integrated digital mapping GPS/inertial systems for direct geopositioning. However, the study does demonstrate the possibilities of commercial satellite data and NCAP resources for long-term archiving and environmental monitoring. The potential benefits of adding these data to mapping wildlife habitats in Antarctica are far reaching and include continuous year-round coverage. It is important for researchers to realize the limits of these data sources and to balance the criteria for their use against practical considerations of doing short- or long-term environmental studies in the Antarctic. A wide array of NCAP resources data increases the opportunity for NSF researchers to make better-informed decisions.

Authors’ Note: The authors would like to thank NSF’s Polly Penhale and Scott Borg for supporting this study. Much of the material in this article originally appeared in USGS Open-File Report 2004-1379.

 

  See  Featured Images
 
  Subscribe to Earth Imaging Journal

  Go to Home Page