KGS, Open-file Report 1999-15

Optimizing the High Plains Aquifer Water-level Observation Network

by Ricardo A. Olea and John C. Davis
Kansas Geological Survey

KGS Open File Report 1999-15
May 1999

Summary

The Kansas Geological Survey (KGS) and the Division of Water Resources (DWR) monitor a network of observation wells in central and western Kansas for the purpose of determining annual changes in water level in the High Plains aquifer and other aquifers. The observation network in the High Plains aquifer is intended to have a maximum uncertainty of ±10 ft in water-level elevation anywhere within the interior of the network. There are locations where this standard is not met, either because of loss of observation wells from the network over time or defects in the original network design.

We recommend that 28 new observation wells be added to the High Plains aquifer network to compensate for locally inadequate sampling. Four more observation wells should added to verify data collected from wells that exhibit aberrant changes in year-to-year water-level measurements. Readings from these wells may be the result of unusual but real fluctuations in water level in the aquifer, or they may be caused by poor mechanical conditions in the observation wells that make the measurements unreliable.

Locations where additional observation wells and verification wells should be located are shown on maps and listed in tables. One new observation well should be designated as close as possible (within a radius of 2.5 mi) to each set of specified coordinates.

List of Plates

1. Observation wells eliminated from reliability analysis, 1999 (PDF file, 100 k)
2. Kriging standard deviation for dependable observation wells, 1999 (PDF file, 600 k)
3. Kriging standard deviation for proposed modified observation network (PDF file, 500 k)

Introduction

The Kansas Geological Survey (KGS) and the Division of Water Resources (DWR) measure water levels in observation wells in central and western Kansas to monitor the water content of underground aquifers used for irrigation, domestic, and commercial water supply. Most of the observation wells penetrate the High Plains aquifer in a fairly systematic spatial pattern, forming a "network" that is intended to provide a reliable picture of the water level within the aquifer.

In theory, the best possible configuration for a water-level observation network would consist of wells located at the centers of uniform hexagons arranged in a regular pattern (Olea, 1984). Because the present observation well network was originally constructed using wells that already were being monitored, and the practical limitations on the choice of additional wells that can be used for monitoring purposes, the Kansas High Plains aquifer observation network does not conform to the ideal regular hexagonal pattern. However, over the last two decades there has been systematic progress in modifying the network to approximate the ideal sampling pattern.

This study reviews the reliability of the observation well network for estimating water levels in the High Plains aquifer and proposes modifications to improve collection of data and reduce uncertainty in the network.

The Observation Network in 1999

There were 1,272 High Plains aquifer observation wells to be measured during the January 1999 data collection program. Of these, 1,204 were successfully measured (Olea and Davis, 1999). According to Miller and Davis (1999, p. 7-8), there are 37 wells that should be removed from the network and not measured in the future because these wells have serious mechanical problems that make them difficult or impossible to measure. Of those wells, 31 tap the High Plains aquifer. These are listed in Table 1.

Table 1. Observation wells in the High Plains aquifer that should not be measured in the year 2000 for mechanical reasons.

Miller and Davis (1999, p. 8) have identified an additional 24 wells that have erratic year-to year fluctuations. Most of these wells tap aquifers other than the High Plains aquifer. The 13 erratic wells that tap the High Plains aquifer are listed in Table 2.

Table 2. Observation wells in the High Plains aquifer that are erratic in their year-to-year behavior.

If these 44 wells are removed, there will be 1,228 dependable wells in the High Plains observation network. Plate 1 is a posting showing the locations of all wells listed in Tables 1 and 2.

Kriging Standard Deviation for Dependable Wells

The kriging standard deviation is used in geostatistics as a measure of reliability in spatial estimation (Olea, 1999, ch. 10). The larger the kriging standard deviation, the greater the uncertainty in the associated estimate, or stated in another way, the less reliable is the estimate.

In 1997, when the KGS assumed responsibility for measuring observation wells previously measured by the U.S. Geological Survey, it was agreed that a desirable and achievable level of reliability for the High Plains aquifer network could be set at a maximum kriging standard deviation of 10 ft in those areas of the major aquifer that were free from border effects (Olea, 1997a). Plate 2 is a map showing the kriging standard deviation that would be obtained if the 1,228 observations wells that are considered measurable and reliable were measured.

Proposed Modifications to the Network

Based on the standard that the maximum kriging standard deviation inside the central portions of the High Plains aquifer should not exceed 10 ft, all of the isolated areas shown in orange or green on Plate 2 are undersampled. Each year since 1997, locations where additional observation wells are necessary have been proposed (Olea, 1997a; Olea, 1997b; Olea, 1998). Selecting these well locations is straightforward--it has been theoretically determined and confirmed in practice that, for the semivariogram of the residuals of the water-level elevation, it is sufficient to have one well in each hexagon of radius 4.5 km, arranged in a regular hexagonal pattern, to achieve a kriging standard deviation of less than 10 ft. Figure 1 shows the appropriate hexagonal network drawn at the same scale as the maps in Plates 1-3.

Figure 1. Regular hexagonal network of observation wells in the High Plains aquifer required to reduce the maximum uncertainty in water level to ±10 ft.

An advantage of using the kriging standard deviation as the measure of uncertainty is that its value depends on the semivariogram and the spatial configuration of the observation points, and not on the individual measurements of water level (Olea, 1999, p. 24-25). Plate 3 is a map of the kriging standard deviation that would be obtained if 32 wells were added to the network at the locations specified in Table 1. Note that wells do not have to be placed precisely at the positions specified. Depending on the locations of neighboring observation wells, usually any spot within a hexagon circumscribed by a circle 2.5 mi (4 km) in radius and centered at the listed position will result in a kriging standard deviation that is less than 10 ft. Of course, the closer a well can be found to the specified location, the greater and more uniform the reduction in kriging standard deviation.

In Table 3, identification numbers less than 70 are the same well locations that were proposed previously. Identification numbers below 50 are locations originally proposed in 1997 (Olea, 1997b), and numbers between 50 and 70 represent locations first proposed in 1998 (Olea, 1998). This year for the first time we suggest two types of new observation wells. Locations with identification numbers between 70 and 90 indicate where wells are needed to compensate for original inadequate control or to replace one of the 31 wells to be deleted from the network because of serious mechanical problems.

The four locations with identification numbers in the 90s are needed to verify anomalous water-level measurements from wells where the kriging standard deviation would exceed 10 ft if the erratic wells were dropped from the network. At the present, we recommend that all 13 anomalous observation wells listed in Table 2 continue to be measured, because their unusual readings may represent true behavior of the aquifer and not the result of poor mechanical conditions in the observation wells. New observation wells at locations 91-94 will provide a check on these readings; if the new wells also behave in an unusual manner, this will confirm that the behavior is real and not a sampling artifact.

Table 3. Centroids of proposed locations for new wells required to reduce the kriging standard deviation below 10 ft in the High Plains aquifer.

The uncertainty near the five wells listed in Table 4 exceeds the minimum standards, but because these wells are located on or near the boundaries of the High Plains aquifer, this is not surprising. Accurately sampling the edges of the High Plains aquifer is not a priority at present, so no action is recommended to investigate the reliability of water level measurements in areas surrounding these wells.

Table 4. Observation wells in the High Plains aquifer that are erratic but are located near the aquifer boundary.

Conclusions and Recommendations

Field reports on the mechanical conditions of wells, combined with a year-toyear analysis of changes in water level at individual wells and geostatistical analysis, supports the following conclusions and recommendations:

1. There are 31 High Plains observation wells with mechanical problems so serious that they probably cannot be successfully measured in the future. We concur with Miller and Davis (1999), that these wells should be dropped permanently from the observation well network.
2. There are 13 observation wells with unusually variable year-to-year water levels. We propose that they continue to be measured, but that four new observation wells in their neighborhoods also be measured to determine if the behavior is real or the result of poor measurement conditions.
3. We recommend that 28 wells be added to the High Plains network to achieve a maximum kriging standard deviation of no more than ±10 ft. throughout the central part of the High Plains aquifer. The number of additional wells proposed is a 44% drop from the 50 new wells proposed last year, indicating that progress in filling the gaps in the network was made in 1999.

References

Miller, R.D., and J.C. Davis, 1999, 1999 Annual Water Level Raw Data Report for Kansas: Kansas Geological Survey Open-File Report No. 99-5, 271 p. and 1 compact disk.

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

Olea, R.A., 1997a, Sampling Analysis of the Annual Observation Water-Level Wells in Kansas: Kansas Geological Survey Open-File Report No. 97-73, 44 p.

Olea, R.A., 1997b, Modifications to the High Plains Aquifer Observation Network Expansion in Open-File report 97-73: Kansas Geological Survey Open-File Report No. 97-84, 3 p., 1 plate.

Olea, R.A., 1998, Proposed Expansion of the High Plains Aquifer Observation Network for Reliability Enhancement: Kansas Geological Survey Open-File Report No. 98-39, 3 p., 2 plates.

Olea, R.A., 1999, Geostatistics for Engineers and Earth Scientists: Kluwer Academic Publishers, New York, 303 p.

Olea, R.A., and J.C. Davis, 1999, Sampling Analysis and Mapping of Water Levels in the High Plains Aquifer of Kansas: Kansas Geological Survey Open-File Report No. 99-11, 35 p. and 9 plates.