Highway Fatality Database Created in ArcView 8 For Seven Counties in Southwest South Dakota From South Dakota Department of Transportation Data, 1983-2001

Here are some charts I made in ArcView and Excel analyzing the fatality data supplied by the South Dakota Department of Transportation.

Part of the work of creating this GIS database was accomplished during a GIS Research 393 course during Spring 2003 at Oglala Lakota College. Students who participated in that course were Michael Catches, Kyanne Dillabaugh, Adam Heriba, Justin Hooper, Cedar Lonehill, Dusty Mousseaux, Elvin Returns, and William Spoonhunter.

This GIS databasae has been formatted using descriptive terminology that will allow anyone to perform queries. I’ll give an example.

This screen capture shows how I select the accidents involving a pedestrian from the layer with all fatal accident data from 7 counties (Pennington is not complete and is not included in the database yet). The 7-county layer is called ACC_7CNTY. The ACC_TYPE field contains the entries shown on the right, BICYCLE DRIVER, PEDESTRIAN, AND VEH/DRIVER. The map shows the location of all the pedestrian accidents. I create the expression ACC_TYPE = PEDESTRIAN and perform the query.

Note the cluster between Pine Ridge and White Clay; this probably represents people walking down to White Clay to buy alcohol and coming back to Pine Ridge and being hit on the road. There is another cluster around Sturgis, too, which may represent inebriated people walking around during the August rally. It would be easy to chart those by month, too, to confirm that suspicion.

Doing the search selects all entries from the ACC_7CNTY layer that had PEDESTRIAN in the ACC_TYPE field. As shown at the bottom, that was 44 out of the 430 total fatal accidents entered here.

I can click the “Selected” button to view only these Pedestrian accidents, then I can save this out to a new layer.

I can do many analyses on this new Pedestrian layer since it contains all the data available for each of those 44 accidents. Here I’m going to summarize the accidents by county.

I get this *.dbf table showing the number of accidents by county which I import to Excel and graph.

Here is the pedestrian fatality data. Shannon leads the pack.

Here is the percentage of all 430 accidents in the seven counties just looking at the different possible road conditions and summarizing by that field. I would have thought that bad road conditions would have accounted for more of the accidents, but 86% of them occurred on dry pavement.

And here are all accidents summarized by collision type, mostly head-on accidents on the 2-lane roads common in this area.

Here I summarized the alcohol and drug related accidents by percent of total among all counties (blue) and then by percent of the total accidents in each county (purple). The “res” counties, Bennett and Shannon, have the highest percentage of accidents with alcohol and Shannon accounts for the highest percentage of all accidents with alcohol as a factor among the seven counties, with Lawrence County (Sturgis) second. Fall River also shows an abundance of alcohol-related fatalities, >50%. The percentage of accidents within Jackson County is small for alcohol/drug involvement. That’s an interesting result.

Summarizing the alcohol-related accidents by month and county is also interesting. The percentages are of total accidents in a given month among all counties. It’s a busy chart, though, so I broke it out below for comparisons with a couple of counties at a time.

Here are the same data but just showing Shannon and Lawrence counties. Shannon has high percentage of alcohol accidents in Nov. and Dec. and also April and July. The winter/holiday drinking may be predictable, but what about April and July, Easter and the 4^th holidays? I don’t know.

Lawrence shows the highest percentage in August at rally time, but not much greater than other months. That’s kind of surprising to me.

Here I looked at overturn fatalities. Again the percentages are of total accidents during a given month among the seven counties. Most counties don’t show any trend here probably due to the fact that the data are not extensive here for all 12 months and 7 counties, only 123 accidents. But the chart does show a pretty high rate of overturn fatalities in Shannon during the early winter. But it’s low in the hardest winter months of Jan. and Feb. It’s interesting when analyses raise more questions than they answer. This kind of combined graph is best for showing anomalies or outliers among all counties.

Here I looked at time of day, dividing the day into 2-hour increments. These data are sufficient over the 18 years to create a meaningful analysis. Here I’m looking at actual numbers of fatal accidents in each county spread over the 24 hour day to indicate if there’s a trend in what time of day accidents in that county occur.

Look at the difference in the number of fatal accidents in Shannon and Custer counties. Shannon has most of its accidents between 4 and 8 pm, the after work driving time. Custer has its highest number 12-2 pm, the noon lunch hour driving time, with Shannon also showing a secondary peak there, too. Shannon county also has a third peak of fatal accidents in the early morning, from midnight to 6 AM.

Lawrence, Meade, and Shannon show pretty similar trends in accidents by time of day, the going home from work and school time being the worst. That supports having and enforcing those lower speed signs in school zones.

Comparing the Shannon data with Jackson is interesting. Jackson has a low point during the same hours of the day when Shannon’s accidents are highest. Jackson’s peak is 8-10 am.

I looked at sex and age next. Lawrence County had more total accidents with males 20-40 years old than Shannon, which was second.

Shannon has the most females causing accidents, again in the 20-40 year age range.

To make a clearer picture of male and female participation, this graph shows the 20-40 year old males are the main ones involved in fatal accidents, with males 40-60 next, and females 20-40 third.

Here I looked at male drivers under 20 in all counties. The highest percentage of male teenage driver accidents was in Lawrence, with Shannon second.

These data don’t show any real difference by county for female teenager drivers.

Looking at percentages of accidents caused by male and female drivers of all ages combined (in vehicle 1, the primary one causing most accidents) shows males in Lawrence to be the worst drivers, with Shannon second, Meade third.

Comparing the number of accidents with fatal and non-fatal outcomes when the driver was not wearing a seatbelt, there were significantly more fatality accidents than non-fatal ones (Chi-Square = 3.28, p < 0.001). When seatbelts were worn, the number of fatal accidents was significantly less than the number of non-fatal accidents (Chi-Square = 10.52, p < 0.001).

Lawrence, Meade, and Shannon counties lead the way in accidents in which seatbelts were not used.

Below I analyze all fatal accidents that occurred on a curve in the roadway to see if there were any outstanding factors associated with this type of accident.

Most of the accidents occurring on road curves took place in clear weather with dry road conditions. Where wet, icy, snowy road conditions were a factor in the accident, Lawrence had the greatest occurrence of fatal accidents on road curves.

With regard to the first harmful event, DOT terminology for the main cause of injury in an accident, collision with a fixed object (such as a tree, power pole, or fence post) and overturns were the main causes of injury. These would result from leaving the roadway, an obvious hazard on a curve.

By county, Lawrence had far more fatal accidents on curves than any other. As a percentage of total accidents in the county, accidents on curves were more frequent in Lawrence, Custer, and Fall River, in that order. The percentage of total accidents by county are shown above the bars and the total number of fatal accidents are shown below the bars. Since there were more fatal accidents in some counties and fewer in others, the percentage of accidents compared to the total in a county is a more accurate way to look at the effect of curve accidents by county. Shannon county actually had a rather low percentage of curve-related fatal accidents, only Jackson and Bennett had fewer as a percentage of total accidents in the county.

Looking at all fatal accidents occurring on curves by time interval throughout the day, the raw numbers don’t show much because of different numbers of total accidents occurring during each time interval. But looking at the fatal accidents on curves as a percentage of total accidents during a particular time interval, we see that the interval 10 PM to midnight has the most curve fatalities, 63% of the fatal accidents during that time period were related to curves in the roadway. The next most frequent accidents on curves occurred during the intervals 2-4 AM, 8-10 AM, and midnight until 2 AM in decreasing order.

Many more types of analyses are possible using this database. We welcome the comments of any interested persons who might want to do research using this highway fatality database. I also encourage any student wishing to gain experience building this type of database in a GIS to contact me. The work remaining to incorporate fatality data from Pennington County into this database would make a great learning experience and would serve to complete a required three semester hour Research 393 or 493 course at Oglala Lakota College.

Jim Taulman, Conservation Biology instructor

Oglala Lakota College

jtaulman@olc.edu