Welcome Home to the edge of Nowhere.

GIS Final Project

Using Archived Aerial Photographs for Coastline Erosion Analysis at Double Bluff, Whidbey Island, Washington

This project was an attempt to map areas of coastline erosion on the Double Bluff area of Whidbey Island, Washington, using a series of aerial and satellite photos. It was also a case study to determine whether archived physical raster data alone could be used accurately and effectively in a vector-based analysis of a dynamic process. This kind of analysis is an exploration of concepts in physical geography, specifically, the evolution of glacially-sculpted landscapes. I’ve been fortunate enough to be introduced to these concepts through ESS 209, an Exploration Seminar to Greenland, and have studied the glacial history of Puget Sound and especially Whidbey Island in ESS 302 – “The Great Ice Age” through lectures, maps, lab exercises, and even a field trip to the area in question. This sort of process analysis is quickly becoming more important as global climate change and sea level rise begin to melt and erode polar coastlines and threaten low-lying coastal communities worldwide.

For this project, remotely sensed data needed to be of a small enough scale to run detailed analysis on an area only a few miles long – a satellite image of the entire United States would not be effective. I discovered that Washington State-level agencies generally had plenty of images at the scale of subsections of counties. For this project, I used photos from the Washington State Dept. of Transportation, NASA, the U.S. Agricultural Stabilization and Conservation Service, and the Washington Dept. of Natural Resources, all via the UW Map Collection. The photos were no larger than a normal sheet of paper, and were scanned into .TIF format at 600dpi. Unfortunately, with some of the aerial photos only a few inches in length and width, the average raster resolution was not particularly sharp. Luckily, the contrast between beach sand and trees on the island was high enough that resolution did not prove to be a significant obstacle for marking coastlines.

In order to make a consistent determination of coastline position, all raster data had to be georeferenced to account for photographic distortion. I downloaded an ESRI Census 2000/TIGER roads vector layer and georeferenced rasters based on road intersections, taking care to give preference to intersections near to coastlines to minimize the distortion in this key project area.

Georeferencing the aerial photo using the Roads vector layer as the primary reference

I created a vector line layer for each of my raster images, using ArcMap’s edit mode to manually place nodes. Each aerial photo and its vector coastline are visible in the map “Double Bluff, Whidbey Island Coastlines: 1966-1995” I graduated the color scheme for each coastline based on the date of the raster; choosing a readable and clear color scheme was one of the more difficult components of displaying the vector layers, as visibility was important for not only the map reader’s comprehension, but also for my own ability to use the layers to spot change over time.

Creating coastline vector layers in Edit mode

Displaying coastlines from all four aerial photos

I then displayed all of the vector layers on top of one another in order to determine consistent areas of erosion. The result is visible in the “Double Bluff, Whidbey Island Coastline Composite, 1966-1995” map. At this point, it became clear that the dimensions of distortion due to georeferencing had almost completely overwhelmed the dimensions of possible erosion, meaning that coastline vectors frequently crossed and rarely displayed a smooth, logical progression inwards over time. It was at this point that Tim and I decided that a mathematical analysis would be practically ineffective, and that visual determination of erosion would be the only analysis method that is even remotely reputable with this data. For the sake of accuracy, I did not calculate the total area lost to erosion (this number would be pointless). I simply did my best to identify areas where erosion appeared to be significantly inward and consistent over time, despite distortion. These results are displayed in the “Likely Erosion Areas” map.

The composite of all four coastline vectors. Note the overlapping and inconsistent change between years.

Estimating erosion areas for future study

The apparent failure of the originally proposed techniques in this project highlights a number of crucial data theory, geographic, and scientific concepts. First, no matter how much cleaning up is done after collection, some data is simply too old or low-resolution to enable meaningful analysis on any processes not already blindingly apparent. Next, distortion due to projection can, in fact, be another significant obstacle to detailed analysis. Despite these obstacles, the compilation of this data did result in the identification of two areas that may have experienced erosion and warrant further study. If a subsequent project were to be launched with more refined and accurate data, the results of the first project would certainly give those two areas precedence.

This project did manage to create several data sets of the Double Bluff area and its coastline, albeit with a few imperfections. Even in the GIS-rich State of Washington, it’s surprising how much historical and raster data has yet to be digitized. The project implemented several methods for digitizing this archived data for later analysis, and brought in outside questions to try to visualize a real-world process with real implications. The biggest improvement that could have been done would have been consistent standards (resolution, scale, projection) for remotely sensed data from all involved agencies since the beginning of its collection in this area in the 1940s – a tall order, and something we certainly can’t change now.

If I were in charge of future analysis in this area, I would recommend data collection by GPS in the field done continuously, over consistent intervals. Measuring coastline is by far more accurate with points taken on-site, rather than from 10,000 feet up in the air, 40 years later. This sort of continuous data, combined with already-existing topographic data (DEM, LIDAR, USGS topo) regarding slope and elevation, would be useful in modeling future erosion in the area as well. This project has shown that, in this case, an armchair analysis of data not specifically designed and intended for coastline analysis is not the most accurate, useful, or effective means for answering the proposed question, though it provides a good starting framework for future work.