The Kansas Observation Well Network

Most don't know it but the Kansas Observation Well Network was redesigned in 1984 based on some pretty sophisticated statistical foundations.  The former network consisted of 1749 wells measured at least every Winter - with some being measured quarterly and a few being automatically recorded.  These 1749 wells had more or less been haphazardly added over time since the early 1940s - being added where there were holes in the network, but to no specific pattern.

Turns out, the pattern of well locations does impact the data to some degree, and haphazard is not good.  The original network of 1749 wells statistically gave us a standard average error of about 12 feet.  This was from the beginning considered good enough to at least spot trends in water table attitudes in decadal time frames. 

Kansas found that converting the haphazard pattern to a a hexagonal stratified pattern, the state could achieve the same standard average error with 1135 wells measured annually.  This allowed the state to choose between either operating the network at a reduced cost without sacrificing accuracy, or, improving the network accuracy considerably for the same cost.  Turns out they did both when they added wells very selectively (in previously scarce areas) and reduced the total measured wells.

This is not the whole story though. The size of the hexagonal cells is critical and dependent on the variability of the bedrock and water table data.  The more variable the data, the smaller the hex cells must be to achieve any desired accuracy.  In Kansas the bedrock is far more variable than the water table, so each had its own semivariogram generated.  The final cell size was a 16 square-mile hexagon.

Of course, today's debate is whether the 12 feet of standard average error is sufficient to describe our smaller enhanced management areas in shorter than decadal time frames. Of course it isn't, so we may soon be going back to the drawing board in places.  But it's nice to know we can design exactly what we need in terms of accuracy and be assured that it costs as little as possible.