Home Reports Start

1998 Annual Water Level Raw Data Report for Kansas

Prev Page--Spatial Quality Control || Next Page--Conclusions


IV: Hexagonal Basis of the Observation Well Network

John C. Davis and Ricardo Olea, Mathematical Geology Section

The original observation well network was intended to be a regular hexagonal pattern, with observation wells located at or near the centers of the hexagons (Figure 1). This pattern is based on theoretical work done by R.A. Olea (1982, 1984), who demonstrated that such a sampling pattern was the most efficient for characterizing a spatially distributed property such as the water table elevation. It is "most efficient" in the sense that the hexagonal pattern requires the fewest number of observations to estimate a surface to within a specified level of uncertainty. Any other pattern of observations will either have greater uncertainty or will require more points. Based on the rate of lateral change in water table elevation as a function of the distance between observation wells, Olea (1982) determined the cell size for a regular hexag-onal network over the High Plains Aquifer whose average uncertainty (standard error of estimate) would be less than 12 feet; each hexagon would encompass approximately 16 square miles.

Figure 1. Regular face-centered hexagonal pattern of observation wells.

diagram of hexagonal pattern, wells in center

However, the High Plains Aquifer network was built using observation wells that already existed, since it was undesirable to abandon the historic record of measurements that previously had been made. The pre-existing observation wells did not form a spatial pattern that could be incorporated into an ideal set of regular hexagons. The irregular arrangement of points that has the highest efficiency of any non-uniform pattern is a hexagonal stratification, in which the area is divided into regular hexagons and an observation is located at random within each hexagon. This is the pattern chosen for the High Plains Aquifer network because it could be established by deleting excess wells from the previous network and adding a few wells in areas where no observation wells were present in the hexagonal cells (Figure 2).

Figure 2. Stratified hexagonal sampling network for observation wells.

diagram of hexagonal pattern, wells somewhere in hexagon

The uncertainty associated with a stratified hexagonal network differs from place to place depending upon the placement of wells within adjacent hexagons. If all of the observation wells are located near the centers of their hexagons, the reliability of the network will be quite uniform and close to the theoretical maximum. If two or more wells are close to each other near the edges or corners of adjacent hexagons, the network will be unnecessarily dense and locally inefficient. If the observation wells in adjacent hexagons are located near opposite sides, there may be a gap or hole in the network. In addition, observation wells are constantly being lost from the network, which may lead to new holes or increase the seriousness of existing gaps.

One way of analyzing the seriousness of gaps that are the result of unfortunate local placement of observation wells within hexagons is to consider each well as though it were located at the center of an irregular polygon. These irregular polygons (called Voronoi polyhedrons or Teissen polygons) are calculated so that every point within a polygon is closer to the observation well at its center than the point is to any other observation well. The wells from the stratified hexagonal network of Figure 2 are shown within a set of such polygons in Figure 3. The two shaded regular hexagons represent empty cells that would fit within the irregular network, demonstrating that the network contains an inadequately sampled gap.

Figure 3. Irregular polygonal network of observation wells from Figure 2, showing unsampled gap (shaded hexagons).

diagram of irregular polygons, showing gap in network

The Kansas Geological Survey intends to restore the integrity of the network, at least in areas where the Survey is responsible for the measurement of observation wells, by including new wells which are in optimal locations to fill in the existing gaps. Determining the very best places for replacement wells requires a geostatistical study, which assesses the standard error of estimate of the water table elevation across the network in a continuous fashion. This is done elsewhere in this report, and the results shown as maps on which the gaps in the network are revealed as areas of excessive standard error. An assortment of candidate additions to the well network can be evaluated in an iterative process to determine the best wells to include in the revised network.

Unfortunately, such geostatistical studies take time and may delay the measurement of replacement wells. Sometimes field crews discover that an observation well is no longer suitable for measurement and must be dropped from the network. If a replacement well could be chosen and measured immediately, while the crew is still in the field, the cost of sending people back into the field could be avoided. However, simply choosing the nearest well as the replacement for an abandoned observation well may make a gap in the network even worse. A good rule of thumb for field workers to use for selecting a replacement for a lost observation well is to choose a new well that is as close as possible to the center of the hexagon containing the lost well. Choosing a replacement well near the center of the hexagon will result in better network coverage than simply replacing the lost well with a nearby well.

References

Olea, R.A., 1982, Optimization of the High Plains aquifer observation network, Kansas: Kansas Geological Survey, Ground Water Series No. 7, Lawrence, KS, 73 p.

Olea, R.A., 1984, Sampling design optimization for spatial functions: Mathematical Geology, v. 16, no. 4, p. 369-392.

Prev Page--Spatial Quality Control || Next Page--Conclusions


Kansas Geological Survey, 1999 Water Level CD-ROM
Send comments and/or suggestions to webadmin@kgs.ku.edu
Updated Mar. 8, 1998
Available online at URL = http://www.kgs.ku.edu/Magellan/WaterLevels/CD/Reports/OFR987/rep05.htm