I chose the red pill ...I searched for an interesting and relevant topic, and then found and refined data myself. As I was learning this semester, always in the back of my mind was the question of whether I'd leave the course with the tools and knowledge to build a "real map" in support of my job and my research. It's pretty clear in my mind now that I can do this....albeit quite slowly, and with still a lot to learn. But I can do it. Pretty cool.
I found a journal article that I felt was interesting. The authors were looking for spatial correlation between tornado formation and land use areas. They found a promising spatial relationship between tornados and land use transition zones (e.g.: forested area to farmland, or farmland to urban land use). No great scientific paper ever survives first contact with a journalist....a reporter from Chicago put facts together, noticing that mobile homes tended to be located on the outer edge of the urban land use zones (where tornados show the highest frequency), so could the two be related? The term 'tornado magnet' comes to mind; in popular lore, mobile home parks attract tornados. In reality, of course, mobile homes show the most damage due to construction materials and methods, so only appear to 'attract' tornados. So my map--which compared the spatial distribution of tornados and density of mobile home, at the county level, showed pretty conclusively that there is only coincidental correlation between the two variables.
The hardest part of this assignment, by far, was getting the data into a useable format. My bivariate map used two data sets. The first was mobile homes as a percentage of total homes, by county, in the state of Indiana. I found that data set in the National Weather Service's Storm Prediction Center (SPC) GIS website. I checked the data, and deternined that it was 'good to go'. My second variable was a bit more complicated--I needed data on the frequency of strong tornados (F2/EF2 or stronger), by county, over a 30-year period of record (the standard used in climatology), for the state of Indiana. I could not find the data that I needed pre-built into a spreadsheet. So, roll up sleeves and start extracting the data. National Oceanographic and Atmospheric Administration (NOAA) has a searchable database on tornado occurrence, but it restricts the user to very small samples. In order to keep my manual data extraction within the constraints, I tailored my search to keep the data size manageable (about 500 records). I then manually counted tornado occurrence by county, entered the data into a spreadsheet; found county size and population data on a National Association of Counties web site, and manually added that data for each of Indiana's 89 counties. Then I could import the data into ArcGIS, and join the data table to a layer. I then imported the final map in AI and hacked around for hours, just trying not to destroy my map. I added drop shadow effects to the state and to the graduated symbology. And then the map was finalized.