"Perhaps the most valuable accomplishment thus far is the development of the GIS database....

This recorded archive of drinking water quality, contaminant plumes, and sources of various other potential environmental impact factors will be a most valuable resource for a variety of investigations for years to come.

It is very important that the GIS be maintained and supplemented as necessary."

This paper will describe the GIS system which Applied Geographics, Inc. (AGI) designed and set up for Silent Spring Institute(SSI). Then we will describe some of the data sets acquired and integrated into the system. Finally we will describe one of the applications used for analysis in the study.

In 1994 the Massachusetts Department of Public Health issued an RFP to look at the possible relationships between environmental exposures and the elevated incidence of breast cancer on the Cape. In response to the RFP, Silent Spring Institute formed a multi-discipline team: Applied Geographics, Inc., Boston University School of Public Health, the Slone Epidemiology Unit of the Boston University School of Medicine, Focus Group, and Tufts University Medical School. Silent Spring was awarded a three-year contract to investigate the environmental risk factors for breast cancer.

Figure 1. Breast Cancer on Cape Cod

AGI staff met with SSI staff to discuss their needs and how they would use the system. All the researchers at SSI used Macintosh machines. The GIS system was designed with this in mind. Because of the nature of the study a number of public meetings were planned in communities on the Cape. It would be necessary to have understandable output, so good graphics were essential. The Sun workstation is connected to the Macintosh network that is used by all the researchers at the Newton office. The Sun Microsystems SPARC 20 workstation is equipped with ArcInfo and ArcView software with multi-user licenses. This is networked to a Calcomp Drawing Board III digitizer, a Hewlett Packard Design Jet 650C large scale plotter and PaintJet XL300 color printer. ArcPress is used to generate plots efficiently. ArcInfo is need for manipulating geographic information and streamlining queries and data entry. ArcView allows SSI staff to access the data in the GIS with a point and click interface. ArcView is used to visualize, query, analyze and create plots of data.

All the attribute data in the GIS can be exported to the Macintosh network for analysis using Macintosh software such as Excel, Datadesk, Stata and Foxpro. In the opposite direction, data from the Macintosh system can be imported into the GIS and incorporated into maps. AGI staff have worked on data integration, query and data development and analysis, system support and training. All ArcInfo programming has been done by AGI.

Data available at the beginning of the study indicating that the characteristics of the women on Cape Cod were similar to those who live in the rest of the state, offered some confirmation that intensive study of individual risk factors was not the most promising avenue to explain elevated breast cancer incidence on Cape Cod. The environment of the Cape, on the other hand, is clearly different from that of the rest of the state. A single aquifer supplied nearly all of the region’s drinking water; and the porous, sandy geology of the Cape makes ground water vulnerable to environmental impact from septic systems, municipal wastewater, and synthetic chemicals applied to the land, such as pesticides.

Figure 2. Data Sources

AGI knew there was existing data at two important sources: the MassGIS and the Cape Cod Commission. We were particularly interested in parcel data because we knew we would be submitting a request to the Cancer Registry for addresses of women diagnosed with breast cancer from 1982-1992. The Cape Cod Commission had parcel data for all of the Cape communities with the exception of Barnstable. The parcel data was obtained from the Commission and separately from the Town of Barnstable. In addition to the parcel data, the Commission also provided data on: water distribution pipes, plumes on the Cape, and the Lower Cape Water Management Task Force private well data.

Figure 3. Parcels

In the Massachusetts Executive Office of Environmental Affairs, the MassGIS was a great source of data. This data included: Capewide basemap, Sole source aquifer, Public water supplies, Well zones of contribution, Permitted waste facilities and sites, US Census Gis 1990 data set, Basins, Areas of Critical Environmental Concern,Capewide roads, Surface water, Wetlands, MacConnell Land use for 1971, 1985 and 1990, National Pollution Discharge Elimination System Permit Sites, Outstanding resource waterways, Groundwater Discharge Permit Sites, Underground storage tanks, State hazardous waste sites (21E).

Data sets were also obtained from: USGS, Barnstable County Department of Health and Environment, Cape Cod Water Districts, MA Dept. of Environmental Management, MISER, Datamocracy, US EPA Region I, and Boston University School of Public Health.

The MacConnell land use data is a particularly useful data set. As part of this study, AGI arranged for researchers at UMass Amherst to scan the 1951 aerial photographs, and incorporate them into an ArcInfo coverage and interpreted the land use using the 1990 classification. As with many of the data sets there were issues. The minimum mapping unit for the 1951 photographs is 3 acres and for all the other years, the minimum acreage is 1 acre.

Figure 4. Landuse

The land use data was used to identify potential sources of exposure such as cranberry bogs, golf courses and locations of potential receptors as indicated by residential areas.

Most of this data was incorporated into the system prior to obtaining the breast cancer data from the Massachusetts Cancer Registry. This data was received in January of 1996. All staff at AGI associated with the data signed a confidentiality agreement prior to working with the data. This data was screened for duplicate entries, non-Cape residents diagnosed at Cape hospitals. The final number of cases was 2,205. This case file was saved as a dBase4 file. In ArcInfo, AGI personnel use the dbaseinfo command to create a lookup table with a field for the original address provided by the MCR data and comments filed to document changes to the original data and the reasons for changes. The lookup table was then divided into 15 tables representing the 15 towns. AGI then reviewed the addresses in the lookup table to correct an obvious typographical errors and alternate spellings or abbreviations of street names. AGI incorporated additional address information provided by the MCR (entire Cape) and MDPH Bureau of Environmental Health Assessment (Upper Cape towns) into the data base. This refined case list was then ready for geocoding.

In addition to the parcel coverage, AGI acquired the GDT Barnstable County Landbase Road Centerline database. AGI then considered the advantages of geocoding using the centerline database versus using the parcel coverage. AGI and SSI designed and conducted a pilot study to decided which of the two available databases, the GDT centerline or the parcel coverage, would provide a better geocoding for the breast cancer incidents. AGI established a baseline by attempting to match the entire data file as obtained from MCR to street addresses and zip codes using the GDT Centerline database. The result was a 52% match rate, Cape-wide.

Figure 5. Geocode Comparison

Two towns were selected for the pilot study: Sandwich, because it was expected to have high quality parcel-level data and Dennis, because it had less complete parcel data. The pilot study demonstrated a higher rate of successful matches using the parcel data. SSI decided that the parcel coverage would yield a higher number of cases mapped ad the accuracy and precision of the assigned location. Final results for the geocoding of the breast cancer incidents ranged from 85% in Truro to 100 % in Brewster. The Cape has a high summer population, which results in many addresses being post office boxes. The other cultural factor, is that women change surnames with changes in marital status and women’s heirs may have different surnames again.

Pesticides

Cranberry Bogs

The MacConnell Land Use, the MassGIS hydrography, and parcel Department of Revenue codes are three sources of information on locations of cranberry bogs. AGI investigated how well these data sources related to each other. The MacConnell land use and the MassGIS hydrography data corresponded well. The Department of Revenue did not correspond well to the other two sources. AGI created a coverage, which includes the MacConnell land use bogs along with the MassGIS hydrography. This coverage was used for the pesticide exposure assessment.

Figure 6. Cranberry Bogs and Pesticides

Bogs were divided into three categories:

1. Old, former bogs that include bogs that include bogs that appear on the 1951 maps
2. Old, continuing bogs which appear on the 1951 maps and on the 1971, 1984 or 1990 map
3. New bogs that first appear on the 1971 map or a later map.

 Once the bogs were identified, it was necessary to identify the presence of forested areas between a residential area and a source of pesticide spraying since the forested area will reduce the exposure to pesticides in the residential areas. ArcView was used to select residential land use polygons that intersect or are adjacent to the pesticide sources. ArcInfo was then used to calculate which parcels in the corresponding point coverage intersected these residential land use polygons and were also within 1300 feet of the source. This analysis result in a very rough approximation of parcels separated by a forest buffer, but it is better than not being able to consider the forest buffer at all.

 Figure 7. Cranberry Bogs and Pesticides with Forested Area

The GIS has been a powerful tool and helped to improve exposure assessments. GIS enabled us to use land use, parcel data and pesticide source data together. There are now coverages for each town that identify for each parcel, the age of housing based on the MacConnell land use data, the census block group, the distance to the nearest pesticide source and an estimate of whether there is a forest buffer between the parcel and the source.

Through the use of GIS SSI was able to refine its cancer surveillance workand identify subregions of the Cape where breast cancer is elevated. . If you compare Graphic # 1 with Graphic # 2 you will clearly see the refinement.

Figure 8.  Breast Cancer Incidence by Sub-Region

The Massachusetts Department of Public Health has funded Silent Spring Institute with Applied Geographics, Inc. for a next phase of the study. Clearly the success for the GIS has been an essential part of a groundbreaking study in the search for clues to the cause of breast cancer.