A Survey has been developed to document and compare GIS utilization in the workplace. This survey assesses GIS availability and utilization in both academic and non-academic work settings. The purpose is to document the need for GIS experience as an occupational skill. GIS is currently being underutilized by most companies. Spatial Technician and Analyst activities and a few managerial activities requiring GIS are reviewed.
This survey, which takes about 20-25 mins to complete (@25 questions), can be accessed at
OREGON CANCER CASES
[Note: All photomaps are produced using Google Map. The full report filed for the grant-funded portion of this work, prior to developing this grid technique, appears on this site at: http://brianaltonenmph.com/3-gis-environmental-health/report-for-grant-funded-research-2002/. This page is devoted to square grid cells. For hexagonal cells go to http://brianaltonenmph.com/6-gis-ecology-and-natural-history/hexagonal-grid-analysis/]
The application of grid analysis to chemical exposure has several applications relavent to chemical release sites studies. Whereas traditional epidemiology likes to make use of various still and moving spatial methods for analysis, the basemaps utilized for this method of spatially analyzing disease does have substantial limitations. It is traditional to relate case information to area features, in particular population density related features. As a part of this spatial epidemiology research method we rely quite heavily upon the assumtpion that areal spatial data can be related in a trustworthy fashion to point data. Such is not the case unfortunately.
If we take for example a theoretical situation in which a fairly heavily populated retirement center is located next to a toxic waste site and analyzed in the traditional areal-point fashion, we have the problem of first dealing with age-related features, such as the increased likelihood that cancer will develop in people residing in the retirement center due to their age and other non-local exposure and lifestyle related features. Secondly, just because the rate of something common to this age group, like the incidence of prostate cancer or colon cancer, is comparably high for these people, we cannot determine with any certainty whether or not the incidence and prevalence of these cancer had anything to do with their average ages distribution relative to the state averages. There is no way of knowing for sure if this statistically-tested relationship between the population and the state average for other populations of this age range and size can at all be linked to exposure. For this reason, the traditional areal analysis methods of reviewing incidence and prevalence in the form of case point data has to be considered just one tool in the many that are available for epidemiological researchers.
In most of my current projects involving statistical analysis, I find it reasonable to expect to see multiple methods of analysis be used to test and retest your outcomes. Statistical measures that tend to be good at measuring change of one type of fashion can fail when measuring another form of change considered equally important by other research teams. My use of ANOVA and its qualitative equivalents for example are always complimented by the application of other measurement technique to verify that a change did in fact happen the way we surmise from the first outcome. Typically, if you show an improvement using one method of calculating the outcomes, you should see more evidence for this change using another if the techniques used for these analyses are performed correctly. There is no single way to test for outcomes that may be considered the best way to accomplish this, and there is no single way that is worth completely avoiding due to its complexity or questionable misunderstanding. (Not all statisticians truly understand their formulas, just the way to perform them using their various types of PC tools and software.)
Because these studies focus on cases in relation to chemical release sites, it is the spatial association between the two that is of concern for this analysis, not the validity of the statistics underlying these rates utilizing non-spatial methods of performing these tests. In the case of the retirement center located next to a toxic release site, we have just the people in the retirement center and the point/area relationship of this population in the retirement center to consider, not the incidence/prevalence rate of the illness in relation to population density over a much larger census block area.
The techiniques recommended for carrying out this process are simply to first determine if there is some sort of spatial relationship between cause and effect, and then relate ares of seeming high risk to the underlying population density features, utilizing a fairly accurate way of defining these features, building by building, house by house (i.e. recoded, reclassified aerial photos and landsat images for starts).
Moving window analysis is one way in which spatial features are related to each other. The application of moving windows in relation to population density has been attempted for this work and resulted in potentially positive testing areas that upon post hoc observation had very few cases much less very few adequate exposure sites. With an analysis of thousands of cases over 120,000 square miles, it is no surprise that somewhere during this analysis the bell curve feature took over. Three sigmas to the right or left enabled the spatial analysis tool to produce a positive outcome, but not anywhere where it was to be expected. The best follow-through with this type of result is simply to re-run the tool and methodology used and see if once again the same outcome takes place.
The follow large town regions were analyzed for various forms of cancer in relation to toxic release sites.
The city of Astoria at the northwest tip of Oregon is one of the most interesting sites to review with regards to chemical spills and release reports history. This urban setting is fairly small in size both areally and population-wise, and has several very unique industries that make up much of its economic history.
Some sections of the city, being on the Columbia River as it empties into the Pacific Ocean, were important lumber product manufacturing sites and important fishery canning sites. Closer to the west coast of the city, there are a number of international and national shipping ports, and numerous storage facilities for incoming solid and liquid products, ranging from simple factory-made automobiles and store products to industrial chemicals and petroleum crude and processed chemicals. There are a number of federal sites along the shoreline in this area, including coast guard sites, with the most important coast guard school in the United States situated just north of this city geographically in the state of Washington.
This is a popular tourism site for the state of Oregon and has a number of protected lands situated east, south, and southeast of Astoria in and around Clatsop County. The Columbia River inlet provides for heavy transaportation needs by major industries. Inland from the bay-inlet area of Columbia River is a large number of oil storage facilities, a nuclear reactor site, and much further inland a storage site for chemical (and biochemical?) weapons. The transportation needs of the valley resulted in the establishment of numerous major shipping ports and train depots. A large number of major manufactories are found in this region due to water-related cooling requirements. These riveredge manufactories and various industrial sites contrast with the types of site situated along the coastal shores beginning from the western land border just west of the city of Astorian southward. In this corner of the state, the northern boundary toxic release sites are greatly different from the western shoreline sites. The Astoria region studied for this part of the GIS work details the mixed nature of these two types of sites and how they relate to people residing in one of the oldest urban settings for the state of Oregon.
There are three parts of the Astoria urban setting of interest due to their clustering of cases and close proximity to important chemical release sites. There is the inland shoreline setting along the Columbia river, the tip of Astoria itself, and the southern shorebanks of this part of the state close to Astoria. The cases clustered in these three areas are a result of the population clustering seen in these three town-like settings. On the maps below depicting these cases, they are defined using circles (two color and circle crosses). The remaining point data (from small to largest in size) represent various forms of chemical release sites (TRIs, CRIs, SFA, SF and High Risk), with the larger symbols represnting more toxic sites. The boundaries map is the Census block-group map.
A ~1 mile by ~1 mile grid was then produced and data from the point files linked, or in some cases copied, to the grid data. Grid cells with cases in close proximity to release sites were then identified. The greyness of each grid cell is determined in relation to the relative amount of toxicity associated with that one square mile area.
Each type of site can be scored a value meant to imply level of risk. For example the highest risk Superfund Applicant sites could be given a value of 10, the High Risk sites a value of 7, the CRI sites a value of 5 and the remaining TRIs a value of 3. Cartographically speaking, each grid cell has a centroid related to it (some grid extensions require the use of another extension to do this). To this centroid you can attached the values of all release sites added up per cell, calculate in the number of reports and degree or toxicity/carcinogenicity per report filed, and then use this centroid data to produce a contour map depicting risk as a combined areal-point feature. The contour lines were then cleaned up (unnecessary lines removed) in order to focus on the area of greatest concern–the northern shore. This final product depicts the spatial distribution of cases in relation to potential exposure sites defined by risk and combined risk, at a per square mile basis.
Since a fairly large grid cell was used for this study of potential for exposure to chemical release in Astoria, the coarseness of the outcomes and possibility for inaccuracy and misleading risk depictions are at issue regarding the use of a one square mile square grid cell. The next attempt at employing this methodology made use of smaller cells and involved a smaller urban setting in Oregon with a greater dispersal of suburban population features.
Coos Bay, Oregon
For the Coos Bay study, a similar method was used, but in order to deal with the fairly non-connectivity noted due to the use grid cells that were perhaps too large, the same grid applied to the Astoria project was overlain by a grid containing quarter-size cells. Again a similar analysis was made and possible relationships between cases and sites deduced. Upon further review, one could see that there were still problems due to a corner effect, and so the development of a method for applying hexagonal grid cells began.
There are three research areas for the Coos Bay area study.
At the north end is Winchester Bay. The central area is Coos Bay. Between Winchester Bay and Coos Bay is Lakeside area with a number of dammed up streams located inland from the town center that look like two “squiggly” inlets on the google map above.
Between Coos Bay and Bandon there is the small town of Charleston and an inlet called the South Slough. Further to the south is Bandon, an area with a very small inlet. The town of Coqulle is just to the north and well inland to the east from Bandon.
From north to south the major townships go Reedsport (Winchester Bay), Lakeside, North Bend (Coos Bay), and Bandon, with Coquille further inland.
The reason I bring this up is that the grid maps printed out from ArcView are numbered in the order of review. For this presentation, the northmost area is reviewed first (but called Coos Bay – 3) , followed by the southmost area (Coos Bay – 2), with the more complex Coos Bay area saved for last (just Coos Bay). [Note the very round bend in the inlet heading to the ocean following Reedsport; this contrasts with the more irregular pattern of the same part of the stream near North Bend emptying into Coos Bay.]
Ocean and bayside industries are quite different from those in the more inland sections of the state. Whereas a simple trip east of the location on the above maps would reveal a large number of lumber and wood product and milling industries (no creosote industries, these are more northward), and industries and toxic release sites typical of heavily urbanized settings (offices, dry cleaners, auto repairs, gas spills, etc. etc), the trip along the coastline of Oregon exposes one instead to sites that service the shipping industry, the shipping ports, and are engaged in tourism and recreational boat use. The two most important features of these settings pertain to fuel products, boat rentals, managing the use and upkeep of recreational boats, and the demands for ship repair and hull surface refinishing industries. Fuel products form two of the major categories evaluated using the reclass method for toxic release site chemistry–petrols from gas stations and fuel storage facilities, and PAHs (a generic entry allowed by EPA) for sites defined as such using the standard classification system in use for evaluating these types of sites. The ship maintenance and repair industries bear a number of very unique chemical differences from the petrol/PAH sites (see Chemical Signatures/Chemical Profiling work on another page). Both have unique and different relationships to carcinogenesis, and may even involve different tissues in terms of their carcinogenicity.
Coos Bay itself is the most actice region socially, economikcally amd probably in terms of exposure to toxic chemicals. The following is a close-up of the Primary Urban setting in and around Coos Bay.
.In this map, two industrial areas (not sites) are noted using the map and the distribution of toxic release sites. These two areas had a natural boundary formed spatially by the large void between the oceanshore and bay-inlet shoreline facilities. These two sets of sites were evaluated and a point location identified to represent each of these two pairs of sites. Chemical types were then mapped for the toxic release of each area. .As already mentioned, geography has everything to do with allocated land use rights. This means that sites along the oceanshore are expected to be different from those in the bay-inlet shores. The major difference between oceanside sites and bay-inlet sites is the type of industries established in each. Along the oceanshore we tend to see marinas and ports where boats are fueled, stored over the winter, repaired, and cleaned and recoated each year with a special paint. Major repairs and major refurbishing of watercrafts takes place at these facilities. Marina chemical histories tend to be more prone to certain chemical types, including heavy metal paints and chrome finishes. The PAH versus heavy metal differences in the two bar charts above demonstrate this major difference..Depending on the type of cancer one is researching, the cases found in the heavily populated section just north of these two centroids could be caused by either site in theory. But if we are more concerned about Chromium induced cancers, we would probably be focused more on the marinas. If the cancers are petrol-induced, then both sites would have to be further reviewed due to the Petrol and PAH bars for each, with emphasis on the bay-inlet sites. Another way to review the above data, if the cancer points are for multiple forms of cancer, would be to ascertain what types of cancer are being portrayed, and to determine if there were significant differents in one form of cancer versus another between the oceanside and bay-inlet sites or centroids. We ould also assign a centroid to the bulk of the human cases and use this in developing some sort of analytic approach to determining where the cancer clusters best relate spatially in termsp of chemical toxicity, carcinogenicity and proximity to the two centroids produced for oceanside and bay-inlet sites..By applying grid cell scoring techniques to this methodology, we can evaluate the region using a different point of view. In this second method of evaluation, the distributions of the sites are retain, and the centroids for each collection of ocean-inland or east-west sites is not used. Instead, each site is assigned a risk score based on the types of chemicals related to each site, the amounts release, etc. These values assigned to each site point (using an avenues tool) are then added up on a per grid cell basis, and the grid cell mapped using a color pattern to depict risk. Using this method, we see that based on types of chemicals release, the few sites along the oceanshore are much more toxic than the many sited randomly dispersed throughout the bay-inlet area. There are a number of very important sites in the bay-inlet area that could draw local attention, for example the site with numerous CRI reports (depicted by the red pentagon). This is an important site to watch, but not necessarily one in need of further review and clean up. The most toxic sites are the two along the oceanside, situated very close to one another, with each bearing a significant number of reports. Applying the bar charts previously reviewed to this, we can tell that one or both of these sites may have a link to heavy metal release, due to the industry type located in this type of setting. Rather than blame the oil companies and storage tanks for local cancer, this suggests that the dispersal of sand-blasted paint and ship metals produced through annual cleanings ande repaints of the ships hulls could be the cause for the local jump in cases, assuming incidence/prevalence rates show there to be an increase in local cancer cases…
Again, notice the displacement of sites relative to the grid cell centroid. The two white cells to the far east best represent this problem. In a one-square mile cell this would result in a spatial displacement of the data by about 0.5-0.72 miles (sqrt2/2). For the 0.25 sq-mile grid cell (0.5 mi x 0.5 mi sides), this displacement from the cell centroid is greatly reduced, but still at about 0.25-0.36 miles or nearly 2000 feet. For this reason, the hexagonal grid cell method was developed to eliminate this corner effect.