A comment by Henk Meij
Compliments on the very nice layout of this population information resource. Some general questions I have are: - can a page be added describing what software components are used to handle map display, graph/plots, and information retrival? - what is the reason for the lack of the 1990 data point? 1993 is an estimate (?) and if so, should be flagged as such. - How are new/old and changing boudarry conditions handled, what are the boundary data sources? Finally, it is ofcourse fairly misleading to look at population density at the county level, the latter not being very homogeneous in composition at all. Densely populated counties may have extremely remote, isolated areas and vice versa. The primary source of increases in pop density in the west is in migration. The graph would be extremely useful if it showed the rate of in-migration at each time interval showing it's tremendous contribution to population increases of each county. -Henk SEDAC Project Scientist
Explanation and some details
A comment by Peter N. Schweitzer
I got the data from Martin Chourre and Stewart Wright, who describe them in good detail in their paper for this conference. Martin has indicated to me that the data I call "1993" is in reality 1990 census data plotted on more recent (1993) county boundaries. So the program should refer to the latest data as 1990 rather than 1993. We (the workshop administrators) decided that we shouldn't fix things like this in "mid-stream" because the fixes will make the comments confusing. I agree that county-level population data aren't good for understanding all of the processes that affect local population density for the reasons you mention. This is intended as a broad summary only. I suspect, however, that it would be a lot harder to depict the data at a finer spatial scale because census tracts have likely changed considerably in the last 100 years, so it might not be possible to create a credible time- series for smaller geographic units. Just guessing. Anyway, the boundaries change with every decade (see the discussion by Chourre and Stewart on this). Some counties are split in two or otherwise change their total areas. Where the area changes I've got the CGI outputting the area as a column in the table, but if the area didn't change, I leave that column out. The plots don't show (or correct for) changes in area, so some plots of total population have dips that can be attributed to loss of area. The data are stored in a database on the server using mSQL (mSQL is used for the conference's users, documents, and document threading as well). The county boundaries from the Chourre and Stewart data were converted to imagemap areas by my intern Yew Yuan using Arc/Info's ungenerate command followed by some C programming. When you click on a county, the CGI program (written by me in C) gets the county's data from the SQL database and creates the plot image if the plot image has not already been created (since they don't change, it makes sense to save them rather than generate them each time). Then the data are expressed numerically in a table and the plot file is included inline (<img src="...">). The plots are created as GIF images using Boutell's gifDraw library, a set of rudimentary graphics routines callable from C that create reasonable (if primitive) GIF images. So aside from the first page, all of the rest are generated through the CGI. Most of the code is fairly specific to our system here, but if you're really interested, I'd be happy to show it. Please let me know if I can provide additional information or assistance.