|Mission with Wichita Wildlife Refuge|
|Mission with Wichita Wildlife Refuge|
Two GISCorps volunteers assisted the Wichita Wildlife Refuge in Oklahoma
By: Lacey Mason & Gericke Cook (GISCorps Volunteers)
The Wichita Mountains Wildlife Refuge (WMWR) encompasses 59,020 acres in
Southwest Oklahoma was opened to settlement in 1901. During that time, the Refuge was the site of a gold rush; hundreds of abandoned mines dot the landscape. In the 1930s, there were three Civilian Conservation Corps (CCC) camps and many Work Progress Administration (WPA) projects on the Refuge. In the 1930 and 40s, the Refuge was used for Army field maneuvers. Of particular importance is the impact of these human activities on soil disturbance. Soil disturbance can allow invasive plant species to gain a foothold that if left unchecked could change the structure and function of habitats. The purpose of the Early Detection Rapid Response (EDRR) Invasive program is to find high risk invasive species on the Refuge and eradicate them before they become established. Areas of historic soil disturbance are targets of interest in the search for invasive species on the Refuge. The staff and volunteers of the Refuge are currently doing a grid search on the ground for evidence of human activities. A Refuge volunteer requested assistance in mosaicking and orthorectifying the 1933-1934 aerial photos, creating digitized features, and possibly create a change detection classification to assist in the land disturbance analyses.
The first phase of the WMWR project was to georectify 37 historical aerial images taken during 1933-1934. It is not known who took the imagery, or the purpose of the original effort to collect the imagery. The 1933-1934 imagery was scanned from prints by a volunteer with the US Fish and Wildlife Service at the WMWR. The imagery was then sent to the two GIS Corps volunteers, Gericke Cook and Lacey Mason. It is an ambitious project to turn historical aerial imagery into a reliable reference source for digitizing features with a reasonable level of accuracy. Typical georectification in ArcGIS or other software rectifies images one at a time, with a minimum of 3 ground control points per image. For 37 images, that would require 111 ground control points. Also, each photo would rectify differently meaning that alignment of features during the mosaicking process would be very difficult. Considering the age and condition of the imagery and the work request, it was determined that ERDAS LPS software would be the best choice for orthorectification of the imagery.
Figure 1: The image depicts the coverage of the 37 aerial photos after georectification, according to the refuge management boundaries. Dots in green are the ground control points manually identified as common features in both the aerial and reference imagery (NAIP 2010, shown in background).
A bit more detective work had to be done in order to determine the camera information and flight footprints of each of the images. Given the unique negative size of the images, 7" x 9", the age of the imagery, and the few companies and cameras used to collect imagery at that time, it was determined from the Manual of Photogrammetry (1944) that the camera mostly likely used was a Goertz with an 8.25" focal length. The flight height was determined by calculating the scale of the photo. The “bundle block adjustment” method available in ERDAS LPS software was used to rectify the imagery. This method establishes “tie points” or common features in overlapping areas of the photos to create the proper alignment of the photos in relation to each other. Also, fewer ground control points are needed (approximately one every other photo) to georectify or triangulate the entire block of photos. Finally, error is minimized and distributed across the block making the overall product higher quality compared to an individually pieced product.
The interior orientation was calculated from the 4 fiducial marks on each photo, ground control points were placed with the 2010 National Agricultural Imagery Program (NAIP) imagery and auto-generate tie points between the images. The exterior orientation was completed on the 37 images. The final RMSE for the triangulation was 3.8 pixels (or less than 2 meters). The images were also corrected for feature distortion caused by terrain relief, by applying a DEM correction from the 10m resolution USGS National Elevation Dataset (NED). This process is known as orthorectification, the product of which is scale accurate and more appropriate for digital feature capture or measurement of distances. The individual orthophotos were color balanced and mosaicked into a final product. The mosaic product was exported to both MrSID and to compressed GeoTIFF formats. An informal visual assessment of the completed mosaic compared to the 2010 NAIP reference imagery showed amazingly accurate results, despite all the challenges.
Phase 2 of the project involved digitizing five feature types from the photo mosaic including: roads and trails, manmade structures, water features, plowed and/or harvested areas, and anomalies by heads-up digitizing. As the features were digitized, the GIS Corps volunteers noticed the imagery was taken during a time of great change at the Refuge, and also digitized the locations of dams newly installed or under construction. Quality control of the feature digitization was completed and Federal Geographic Data Committee (FGDC) compliant metadata was included.
Figure 2: Results of the GISCorps project for the Wichita Mountains Wildlife Refuge, focused on the current location of the headquarters grounds and
Phase 3 of the project requested a land use analysis of the 1933-1934 imagery to be used for comparing to other land use analysis classification from later years. Review of the documentation on the previous land use analysis revealed use of an image segmentation software package (e.g., eCognition), neither volunteer had access to this type of software. However, unsupervised classification was also a process used in former classifications, which the volunteers used to determine tree canopy on the refuge.
A tree canopy classification was completed by the volunteers using ESRI ArcGIS for Desktop 10 Image Classification extension. Given the condition of the imagery, achieving accurate color balance across the image was difficult and inconsistent in the final mosaic. This caused difficulties when testing unsupervised and supervised classifications. Instead of trying to classify all land use on the image the volunteers focused on classifying only the tree canopy. During the process of mosaicking the 37 images using ERDAS LPS, each image was orthorectified before mosaicking. By classifying each individual image, this allowed the volunteers to bypass the issue of color balancing the completed mosaic, and thus produce a composite tree canopy layer. An issue became apparent during this process; water bodies and plowed areas were being classified as possible tree canopy. This was remedied by removing the features from the tree canopy by using the digitized features as a mask. The completed tree canopy coverage will allow the Refuge staff to compare the tree canopy from multiple years and meet their objectives of finding areas of greatest disturbance and change, especially the effects of fire suppression, canopy closure, and woody encroachment, within refuge boundaries.
The final products for this project include a historical orthophoto image mosaic, digitized features, and tree canopy classification, which have been shared with the Refuge staff and volunteers. The final mosaic of the 1933-1934 imagery is already in daily use at the Refuge and the digitized features are being used to guide field work. The final products will also be used in the future for identifying locations of invasives. The tree canopy classification will be used to determine woodland canopy closure and encroachment of woody vegetation into grassland areas. This will assist the Refuge staff in determining fire history, or lack thereof, on the Refuge."GISCorps volunteers accomplished each goal stated and were instrumental in defining what could be accomplished from our antique imagery. The product from this project greatly exceeded our original expectations."
"I initially applied to the GISCorps for two reasons: 1) as a way to contribute my professional skills on a volunteer basis and 2) to earn the final credits for GISP certification. I didn’t anticipate the amount of emotional investment I would have in this project. I feel like I gained so much in return for my time, both professionally and personally. The gratitude from the recipients was honest and heartfelt, and more rewarding than I have ever received in a professional setting. You may find yourself working harder than you ever did, simply because you know it’s worth your time."