Orthophotos with more and more frequent updates have become a common part of any GIS. This article describes a project in which 9,700 aerial images from 13 epochs ranging from 1929 to 1996 were catalogued, orthorectified and mosaicked for use by the Los Angeles Department of Water and Power (LADWP).
Like many agencies, LADWP has an extensive archive of historical imagery dating back many decades. As is often the case with organizations whose business is not aerial photography, this film had been accumulated over a period of years and stored casually in boxes and office filing cabinets without a written archive. In 2009 LADWP contracted with the Geomatics team at the Tetra Tech R&D office in Lafayette, CA to prepare for use in their GIS an estimated 9,700 aerial images taken in 13 epochs covering the Owens Valley and the Mono Lake Basin. The tapping of the water resources of Owens Valley and the Mono Lake basin by the City of Los Angeles, begun more than 100 years ago, is one of the epic tales of California history. The motivation for this study was the Department’s effort to understand and document the impact that the development has on local conditions.
Our plan from the start was to identify blocks of stereo photography and tie them tightly together by analytical aerotriangulation before attempting to build a mosaic from the individual images. Several inputs are required when generating an orthophoto mosaic from a set of aerial photographic images. Besides the block of overlapping aerial imagery, the rigorous process requires a calibration model of the camera used for the photography and surveyed ground control points (GCPs); in most cases GCPs have been targeted before the flight mission so as to be visible in the images and a good digital terrain model. For modern projects, location coordinates and rotational angles for each exposure are usually available. When we began work with the historical imagery of the Eastern Sierra we enjoyed few of these advantages.
The film we worked with came from several sources; most of it was in the possession of LADWP. It came to us as standard 9” color aerial film rolls with a few epochs of their holdings being individual cut negatives on clear film base. All but one epoch of photography had been flown by IK Curtis, a venerable Los Angeles area aerial photography film, a fact that proved useful in chasing down camera calibration reports. The three oldest epochs of photography had been flown by the Fairchild Aerial Mapping Company, a pioneering aerial photography firm. These photographs were held in the Fairchild archive at Whittier College (now closed) and at the Map Information Library at UC Santa Barbara. The Whittier photos, taken in 1929 and 1930 of Mono Lake and the surrounding basin, used an unusual 7”x9” film format. The photography from UCSB was an extensive set of 854 images taken in 1944 extending from Little Lake in the southern end of the Owens Valley to Crowley Lake in Long Valley.
The first step in creating a mosaic was to develop an inventory of the film resource. We were fortunate that the film had been labeled by the original photographer. Mounting the rolls of film on a light table with a reel to reel transport mechanism allowed us to build a spreadsheet with a record of 16 fields for each exposure including date and time, flight line, project area, camera ID and any instructions for scanning. Within the 13 epochs of photography we identified 50 separate blocks of stereo images, some with as many as 1350 exposures, some with as few as 4.
The next task was to convert the analog film images to digital format. With so many exposures to convert we turned to the scanning specialists at Geoscans of Tucson, AZ. Their work would be critical to our process and was not without its challenges. The film we sent them had been stored for decades under less than optimal conditions and had suffered discoloration and even physical damage as a result. Nevertheless, we sent them the film with notes detailing a name and scan density for each exposure and they returned the scans to us on hard drives, ready for us to introduce into our aerotriangulation adjustment.
Aerotriangulation is the process whereby overlapping exposures are tied to one another through the measurement of tie points in the overlapping parts of the images. The procedure also makes it possible to register the adjusted set of photographs to a map coordinate system using a relatively sparse set of ground points. Knitting the exposures together before orthorectification was a key to ensuring that the orthophotos generated from adjacent exposures would mosaic accurately. Since we were working with digital images we used softcopy methods to do the aerotriangulation. We began by developing fairly accurate estimates for the location of each exposure in a block of photography. With over 9000 images to locate we developed software that allowed us to at least partly automate the procedure by locating the first and last exposure in a flight line and then interpolate estimates for the intermediate exposures on the flight line. Another required input for aerotriangulation is camera calibration data. We were fortunate to be able to obtain calibration reports for photography back to 1981. For the older imagery the situation was more challenging. We had to construct our own camera data from what little information we could gather from the sources of the imagery and our own Internet searches.
The introduction of a standard projection was critical to preparing the photography for use in GIS. To accomplish this, ground control points had to be identified and measured on the photography. Not having the convenience of paneled ground control points, we had to pick natural features such as intersections of dirt roads that were found to exist in the historical imagery and were still identifiable in today’s reference imagery of known projection. We were often surprised at how little some features had changed over the years while other locations were barely recognizable as the same place. To derive the horizontal location we used NAIP, DOQQ and rectified Ikonos imagery provided by the client. Elevations were interpolated from an IFSAR DTM provided by the client or from USGS NED where the photography extended beyond the IFSAR data.
With estimates, camera models, and a selection of ground control points we tied the individual blocks of photographs together with image measurements of automatically generated tie points. To overcome deficiencies in the photography including gaps between flight lines, inadequate endlap between sequential photos in a flight line, flight lines apparently deflected by strong winds, poor image quality or even the quite unconventional circular flight lines flown around Mono Lake in 1930, we found it necessary to stabilize a block by the introduction of manual points.
When we were satisfied with the quality of the adjustment of a block of photos we orthorectified the imagery based on either a DTM from USGS or, where the terrain had changed, an autocorrelated DTM. The images were then color balanced, mosaicked with automatic and manual seamlines, and cut into rectangular tiles for delivery. At the start of the project we had developed a tiling scheme with the client that was based on the USGS 7.5’ quadrangles for the area each divided into nine tiles
The overall accuracy of our results was related directly to the quality of the control points we used in aerotriangulation. Because we did not have the benefit of surveyed and targeted points and had been forced to use photo-identified points we had results on the order of a meter or two for most epochs of photography. As expected we had more difficulty identifying common points on the older epochs of photography, but we and our client were pleased with the way that the orthophotos from different years overlaid one another.
Along the way, several other technologies were used during the project. A Microsoft ‘Sharepoint site’ allowed us to update and share documents with the client so they could keep current with the progress being made. Also, during the course of the project Tetra Tech’s ‘GeoManager’ was used to provide the client with access to thumbnails of the high resolution scans. GeoManager is a Google Maps API that provides a mashup between Google Maps and the client’s project data. Before the triangulation and orthorectification was complete and the results delivered, the client could locate any photograph in their archive by year and by geographic location and download a 300 dpi scan of the negative for reference.
Through this project some valuable historic imagery has been preserved for the future. Its historic interest certainly stretches beyond water resources as the area includes historic sites such as the Cerro Gordo mine and other ghost towns, the Manzanar camp and historic railroads.
Our client for this project, the LADWP, has a long and storied history in the development of water resources in the Owens Valley. Over 100 years ago much of these resources were acquired to provide a reliable water supply for a burgeoning metropolis 250 miles to the south. Aqueduct engineering on such a grand scale had not yet been attempted in the American West. To conceive of such an effort took people of remarkable vision, and to plan and complete the construction of the necessary works took enormous energy and drive. But the economic consequences of exporting most the local water and the impact on the ecology of the Owens Valley were profound and long lasting. As a result, the relations between the valley residents and the LADWP were tense for many years. But in recent decades the Department has changed its approach to water management and a new relationship with the residents of the Owens Valley has emerged. In particular, LADWP are seeking to better understand the consequences their actions have in the Valley and to mitigate environmental concerns that may have been caused by water management practices. Tetra Tech is pleased to consider that our orthorectification of their archival photos will help them in their efforts to improve the stewardship of the land whose resources they share.
BAAMA, a chapter of the Urban and Regional Information Systems Association (URISA), is a non-profit, professional organization that organizes bi-monthly educational forums, the annual California GIS Conference, and periodic technical tours on a broad range of geographic information systems (GIS) and automated mapping topics. BAAMA is the vital organization of GIS professionals in the San Francisco Bay Region that promotes partnerships and teamwork with users of GIS technology to improve our environment and community.