WRIR 01-4275

Design of a Real-Time Ground-Water Level Monitoring Network and Portrayal of Hydrologic Data in Southern Florida

By Scott T. Prinos, A.C. Lietz, and R.B. Irvin

This report has been reformatted for presentation on the World Wide Web. The official text of WRIR 01-4275 (6.3 MB download) is available in PDF format. The Adobe PDF Reader program is available, at no cost, from Adobe.

Summary and Conclusions

The USGS collects and computes data from over 200 continuous ground-water level recorders and performs salinity monitoring at 138 wells in southern Florida. These data are critical to decisions concerning water management in southern Florida. As the population in southern Florida has increased, ground-water withdrawals have increased, and in some instances have resulted in saltwater intrusion and loss of pumpage at privately owned water-supply wells.

Estimates of population for the 1980-98 period indicate increases of 34 percent in Miami-Dade and Broward Counties; 109 percent in Collier, Hendry, and Lee Counties; and 82 percent in Martin, Palm Beach, and St. Lucie Counties. These increases correspond to increases in ground-water usage of 26, 99, and 91 percent, respectively. In Miami-Dade and Broward Counties, the population increases have been sustained with minimal effect on the potentiometric surface of the Biscayne aquifer. This partially is because the Biscayne aquifer contains areas with transmissivities on the order of 1,000,000 ft²/d. Aquifers in southwestern Florida, however, have transmissivities that are generally less than 134,000 ft²/d. As a result, increased water withdrawals have formed large cones of depression in the mid-Hawthorn aquifer, lower Hawthorn producing zone, and sandstone aquifer, and areas of every aquifer in southwestern Florida have occasionally been drawn down below sea level.

Water managers need to constantly assess ground-water levels and salinity so that they can evaluate ground-water conditions and take measures to minimize adverse effects. Because of the extensive time required to process, quality assure, and publish water-level and chloride concentration data, the entirety of these data is often not available when timely water-management decisions must be made. A real-time ground-water level monitoring network has been designed and a prototype website (http://www.sflorida.er.usgs.gov/ddn_data/index.html) has been built to provide data in a timely manner for evaluation of ground-water resources, particularly during periods of decreased recharge to the aquifer during meteorologic droughts.

This network is designed to consist of ground-water monitoring wells equipped with satellite telemetry that can transmit data to the USGS every 4 hours. Two of the most important factors considered in the design of this network were period of continuous ground-water level record available for each well considered and well construction. These two primary factors determined whether results of the statistical analyses used to design the network were significant or not. Short-term agreement between two indicator wells or between a monitoring well and precipitation may appear to be "statistically significant." These statistically significant agreements, however, are far more relevant if they occur during periods when the aquifer is under the most stress, such as during meteorologic droughts, which both cause a reduction in recharge and an increase in withdrawals from the aquifer for irrigation.

Four methods of statistical analysis were used to examine the water-level data monitoring wells included in the real-time ground-water level monitoring network. These methods are described below.

The Seasonal Kendall trend test was used to examine long term trends in each aquifer. These tests showed areas where each of the aquifers was changing the most. In many cases, the trend tests that were conducted with data spanning 26 years showed statistically significant increases in aquifer water levels over time. Most of the wells with data for this period are completed in the Biscayne aquifer. Only 14 wells in aquifers in southwestern Florida and 9 wells in the surficial aquifer system in Martin and Palm Beach Counties had data for the period of record. Thus, several shorter periods were evaluated to compensate for many monitoring wells that did not have data for the entire period. The trend tests revealed small but statistically significant upward trends in most aquifers; however, large and localized downward trends occurred in the sandstone and mid-Hawthorn aquifers.
Summary statistics were computed for all wells examined. These statistics provided information regarding the minimum monthly levels recorded and the measures of the amount of variation in water levels.
Linear and polynomial regressions, LOWESS smoothing, and frequency analyses were combined to establish the relation between each candidate ground-water monitoring well and precipitation over the region. Focus was on the timing of extreme lows in precipitation and water level and determining which candidate monitoring wells most strongly showed aquifer-wide changes caused by reductions in recharge during meteorologic droughts. Analyses were performed on water-level data from 246 wells. Of this total, 104 wells had periods of record greater than 20 years and could be compared against several drought periods. After lag, seasonal cyclicity, and cumulative functions were considered, the timing of minimum values of water level from 15 wells and average minimum rainfall values were in agreement 57 to 62 percent of the time over 20 to 26 years. On average, the timing of water-level and rainfall minimums were in agreement about 52 percent of the time (and only about 29 percent of the time in some instances).
A regression analysis was used to evaluate daily water levels from 203 monitoring wells that are currently, or recently had been, part of the network. The mean R² value was determined from a series of stepwise regressions between each candidate monitoring well and every other candidate monitoring well in the aquifer. In each case, the explanatory variables included the water-level data of the candidate well, the water level squared, and time. The mean R² value provided a good indicator of how representative of the other wells in the network the candidate well is. Additionally, the individual R² values from each step-wise regression were used to determine the coverage that the single candidate well would provide. Coverage was defined using R² values of 0.64 or greater, which correspond to a correlation coefficient of 0.80. This degree of correlation was considered to be sufficient for an interim assessment of aquifer conditions. Results indicated that 35 wells had 10 years or more of data and could be used to directly monitor water levels or to estimate water levels at 180 of 203 wells (89 percent of the network).

The relation between water levels and chloride concentrations in 114 ground-water wells was examined using Spearman's r and Pearson's r correlation coefficients. Statistically significant results included both positive and negative relations. These analyses, however, have limitations in that they do not portray a regional assessment of saltwater intrusion, but indicate the relation between chloride concentrations and water levels at specific well locations. Regional assessments of saltwater intrusion have been made in the past, and these chloride/water-level analyses may be used to select wells that may be appropriate for drought monitoring. There were no statistically significant correlations between chloride concentrations and water levels in the water-table aquifer (west coast). Only three wells in the lower Tamiami aquifer showed significant correlation between chloride concentrations and water levels. No correlation existed between chloride concentration and water levels in the sandstone aquifer. Chloride concentrations and water levels were positively correlated in two wells in the mid-Hawthorn aquifer. There were no wells in the surficial aquifer system for which correlation between chloride concentrations and water levels existed; however, seven wells in the Biscayne aquifer showed statistically significant correlations between chloride concentrations and water levels.

The Seasonal Kendall trend test was also used to examine trends in chloride concentration in 113 wells. Of these, a total of 61 wells showed statistically significant trends. Fifty-four percent of the observed trends in chloride concentration were upward (33 of 61 wells), and 46 percent (28 of 61 wells) were downward. Of the 61 wells analyzed, data were from 4 wells in the water-table aquifer (west coast), 11 wells in the lower Tamiami aquifer, 2 wells in the sandstone aquifer, 7 wells in the mid-Hawthorn aquifer, 34 wells in the Biscayne aquifer, and 3 wells in the surficial aquifer system. Upward trends were found in data from 3 wells in the water-table aquifer (west coast), 7 wells in the lower Tamiami aquifer, 1 well in the sandstone aquifer, 3 wells in the mid-Hawthorn aquifer, and 19 wells in the Biscayne aquifer. Downward trends were found in data from 1 well in the water-table aquifer (west coast), 4 wells in the lower Tamiami aquifer, 1 well in the sandstone aquifer, 4 wells in the mid-Hawthorn aquifer, 15 wells in the Biscayne aquifer, and 3 wells in the surficial aquifer system.

Statistical analyses of water levels and chloride concentrations combined with consideration of period of water-level record, completeness of water-level data, well construction, and prior existence of satellite telemetry indicated that a total of 33 water-level monitoring wells (17 already instrumented with satellite telemetry) would provide good coverage of ground-water conditions in southern Florida during times when the data from the full ground-water level monitoring network are unavailable. These 33 wells are not intended to replace the existing continuous monitoring wells because statistical analyses conducted by other investigators have previously indicated that a much more extensive network would be needed to fully assess conditions. However, these 33 wells can provide very useful real-time updates for changes in each of the principal aquifers in southern Florida.

The prototype website (http://www.sflorida.er.usgs.gov/ddn_data/index.html), developed to portray the data from the real-time ground-water level monitoring network, utilizes many of the same concepts that were originally used to design the network, including use of the Seasonal Kendall trend test, removal of long term trends using linear or polynomial regressions, and duration analysis of data (which originated from the frequency analysis used in network design). This prototype website also aids in the examination of chloride data by presenting long-term plots as well as the water-level data whenever possible.

Selected References

Return to Table of Contents || Portrayal of Real-Time Ground-Water Level Data

Funding for the USGS to design and maintain this site has been provided through a cooperative agreement with the South Florida Water Management District (SFWMD). Water-level conditions are monitored by the USGS with support from Federal, State, and local cooperators.

Real-time conditions
(de-trended data)

End-of-month conditions
(de-trended data)

FISC-WRS, home page

Page Contact Information:: WWW Administration: GS-W-FL_Webmasters@usgs.gov; WWW maintenance and data requests: R.B. Irvin - rbirvin@usgs.gov; Statistical methods: S.T. Prinos - stprinos@usgs.gov

Page Last Modified: Thursday, 27-May-2004 16:03:13 EDT