Link to USGS home page
Prepared in cooperation with the Miami-Dade Department of Environmental Resources Management
Correlation Analysis of a Ground-Water Level Monitoring Network, Miami-Dade County, Florida

By Scott T. Prinos

The topic is Coastal Erosion. Open-File Report 2004-1412
Abstract
Introduction
Correlation Analysis of a Ground-Water Level Monitoring Network
Summary
References Cited
Appendixes I & II
image of Duval County, Florida

SUMMARY

The U.S. Geological Survey cooperative continuous ground-water level monitoring program in Miami-Dade County, Florida, expanded from 4 wells in 1939 to 98 wells in 2001. Historically, network design was coordinated with five different cooperative agencies to address the monitoring needs of these agencies. A correlation analysis of water-level data from wells in the existing network was performed to aid in the assessment of network redundancy by indicating which wells provide similar data. Because the water-supply and water-management systems in Miami-Dade County increased in complexity during the period that ground-water data were collected, temporal variation in the degree of correlation had to be considered. It was also necessary to consider the spatial variation in correlation because the water-supply and water-management systems as well as natural differences in precipitation, evapotranspiration, and hydrogeology can act spatially to affect the degree to which water-level data correlate.

Using S-PLUS 2000 statistical analysis software, correlation analyses were performed for daily maximum water-level data from 98 monitoring wells for the period November 1, 1973, to October 31, 2000 (water years 1974-2000). Water-level data from each well were subdivided by year and season (wet and dry seasons). The analysis allowed the comparison of data that were not totally complete, but rejected from consideration any seasons that were missing more than 5 percent of the data. For each year and season, correlation analyses were performed on the data from all wells where data were available. Final correlation matrices for the wet and dry seasons were created. For the comparison of water-level data for each well pair, the correlation matrix provides: (1) the average of the correlation coefficients determined, using all noncensored seasons for the period analyzed; (2) the number of noncensored seasons; and (3) the standard deviation of the noncensored seasonal correlation coefficients. Graphs showing the variation of correlation through time were prepared for selected wells where water-level data were correlated with that of numerous other wells using an average coefficient of 0.95 or greater for the comparison.

The average wet- and dry-season correlation coefficients were plotted spatially using geographic information system software. For each monitoring well, this software could readily show the location of any other monitoring wells that provided water-level data that correlated with a coefficient of 0.95 or greater for the comparison. Using this approach, wells with water-level data that correlated to this extent were almost always found to be located in relatively close proximity to each other.

The data from the majority of the wells in the network (65 wells) generally were not correlated with that of other wells during the wet and dry seasons with an average coefficient of 0.95 or greater. Five areas were identified, however, where in both seasons the water-level data from wells within the area correlated with that of other wells within the area with a coefficient of 0.95 or greater, but not with data from wells outside the area. These areas were located in the C-1 and C-102 basins (2 wells), in or near the C-6 and C-7 basins (2 wells), near the Florida Keys Aqueduct Authority Well Field (2 wells), near the Hialeah-Miami Springs Well Field (6 wells), and near the West Well Field (21 wells).

Highly correlated water-level data in two or three ground-water level monitoring wells could generally be beneficial rather than overly redundant. The added water-level data can aid in quality assurance and provide data for those periods when one of the other monitoring wells has been damaged or destroyed. Therefore, the level of redundancy in or near the C-1 basin, in the C-6 and C-7 basins, and near the Florida Keys Aqueduct Authority Well Field may not be considered excessive.

Although the greatest number of potentially redundant wells were near the West Well Field (21 wells), the short period of record available for analysis for most wells in this area generally resulted in determination of averaged correlation coefficients using only a few seasons of data. If the relations that were established for a few seasons of data remain consistent during future droughts, floods periods, and water-management changes in this area, then the case for redundancy would be stronger.

Of the six wells near the Hialeah-Miami Springs Well Field, water-level data from well G-3466 correlated at 0.95 or higher on average with that of three other wells (G-3465, S-19, and S-68). These correlations, however, were based on the data from only five wet seasons and six dry seasons after seasons missing more than 5 percent had been censored. The censoring of data substantially reduced the ability to assess the relation between the wells for recent years because missing data were more frequent after the 1992 water year. Examination of the results from the few recent seasons that had 5 percent or less missing data, as well as analysis of seasons that had more than 5 percent missing record, indicates that the average correlation coefficient for recent years would be much lower than 0.95. Therefore, these data may not be as redundant as the average correlation coefficient would indicate.

The correlation analysis of available data indicates that there are areas where the water-level data from certain wells may be redundant. In general, however, comparison of data is needed for longer periods of record than currently exist. For the few instances identified where the data have remained highly correlated on average over a lengthy period of record (specifically in wells G-757A, G-864, G-864A, and G-1362), an informed decision about redundancy would require considering the short-term reductions in correlation and the specific water-management issues in these areas.

Next:


Figures: Click on a caption to view the figure.
Figure 1. Map showing location of continuous ground-water level monitoring network wells in Miami-Dade County, Florida.

Figure 2. Map showing water-supply and water-management systems in Miami-Dade County.

Figure 3. Maps showing lines of equal rainfall in Miami-Dade County during (a) Hurricane Irene on October 14-16, 1999, and an (b) unnamed storm on October 2-3, 2000.

Figure 4. Graphs showing seasonal variation in mean water levels and variation in monthly standard deviation of mean water levels for wells G-620, G-864, G-1183, and S-18.

Figure 5. Hydrograph showing variation in water levels at wells G-3 and G-1368A along with estimated average daily pumpage based on annual pumpage totals during water years 1974-2000.

Figure 6. Hydrograph showing variation in water level at well G-1502 during water years 1974-2000.

Figure 7. Map showing grouping of wells based on average correlation of water-level data during the wet season.

Figure 8. Map showing grouping of wells based on average correlation of water-level data during the dry season.

Figure 9. Map showing grouping of wells based on average correlation of water-level data during both the wet and dry seasons.

Figure 10. Map showing grouping of wells near the West Well Field based on average correlation of water-level data during both wet and dry seasons.

Figure 11. Graph showing temporal variation in seasonal correlation between water-level data from well G-1487 and that of well G-855 during water years 1974-2000.

Figure 12. Hydrographs showing water-level elevations from wells G-855 and G-1487 during the 1986 and 1998 water years.

Figure 13. Map showing grouping of wells near the Hialeah-Miami Springs Well Field based on average correlation of water-level data during both the wet and dry seasons.

Figure 14. Graph showing temporal variation in seasonal correlation between water-level data from well G-3466 and that of wells G-3465, S-19, and S-68 during water years 1988-2000.

Figure 15. Hydrographs showing water-level elevations from wells G-3465, G-3466, S-19, and S-68 during the 1990 and 1996 water years.

Figure 16. Hydrograph showing water-level elevations from wells G-3465, G-3466, S-19, and S-68 during water years 1988-99.

Figure 17. Graph showing temporal variation in seasonal correlation between censored and uncensored water-level data from well G-3466 and that of wells G-3465, S-19, and S-68 during water years 1988-2000.

Figure 18. Graph showing temporal variation in seasonal correlation between water level data from well G-1362 and that of well G-757A during water years 1974-2000.

Figure 19. Hydrograph showing water-level elevations from wells G-757A and G-1362 during the 1989 and 1997 water years.

Figure 20. Graph showing temporal variation in seasonal correlation between water-level data from well G-864 and that of well G-864A during water years 1974-2000.

Figure 21. Hydrograph showing water-level elevations from wells G-864 and G-864A during the 1990 and 2000 water years.


FirstGov.gov U.S. Department of the Interior | U.S. Geological Survey
Water Resources Discipline, Florida

email Feedback | USGS privacy statement | Accessibility