Several problems arose throughout the progression of this project. Initially, I had problems with finding a suitable DEM that allowed a detailed enough view of Prince William Sound. Other problems arose with overlaying layers that were in the same projection, but refused to work together. For example, I had wanted to use a detailed soil record to compare patterns of sand deposits with the site layers, however, I wasn't able to get the layer into the same projection. Other data problems included the significant loss of positional accuracy, for the sites, because of digitizing from hand drawn maps in journal articles and, in some cases, verbal descriptions of locations. Such a limited number of sites also posed a problem because some layers only had three sites, which is a pretty limited sample to compare spatial accuracy. Also, in the records, Deirdre and I noted a few locations that had thick layers of till around the same time of tsunami activity in other sites. We hypothesized, in conjunction with notes by some of the researchers, that these may have been the result of glaciation and the glaciers may have prevented tsunamis from depositing thin layers of sand on those sites. Landslide induced tsunamis may have also been the result of some of the site records and this would be difficult to determine without looking at geomorphic/geological evidence of historical landslides in regard to size, direction, and time period. In dating the layers of records it was difficult to be very accurate as the dates given to cultural artifacts, or volcanic tephra layers weren't noted at the exact depth. It made it somewhat difficult to translate these dates to other layers and we ended up grouping those that seemed to be roughly within the same time frame to try and compensate for this.
The models themselves, I realize, are somewhat simplistic as they failed to take into account a more accurate frictional surface, the effects of shoreline processes over time, and the effect of wave dynamics on tsunami movement. Further investigation into other, more advanced, models of shoreline processes and wave dynamics could certainly improve the quality of the models developed here. A better frictional surface could certainly be built within IDRISI by taking advantage of the VARCOST and multi-criteria evaluation properties. I did attempt to employ the VARCOST function with the following cartographic model, however, SLOPE refused to accept my files despite numerous conversions of data type and constant refering to the program manuals. I even threatened violence, but the computer responded with the "Blue Screen of Death" and after that we, grudgingly, got along. (Until it came to creating this web page, but that's a different story.)
The other option, of using FUZZY to weight elevation greater then distance for a weighted linear combinationfrictional surface, also failed. I was able to use FUZZY, but the WEIGHT command refused my files. It may have been due to incorrect data types, but all efforts proved to be ineffectual. In writing this it dawned on me that I could've used image calculator to weight the elevation and then multiplied it with elevation to give me the weighted cost surface I was trying to produce. Maybe next time.
Given the amounts of time put into the project it turned out not quite what I'd hoped at the start, though I am fairly pleased with the product. In retrospect, I should have hammered out a more detailed process before wading in and having to do each step three or four times before getting it right and more time should have been given to the building of the web page (especially for a first timer like myself).