Home Archived October 29, 2018
(i)

South Florida Information Access (SOFIA)


publications > poster > automation of the estimation of missing water-level data for the everglades depth estimation network (eden)

Automation of the Estimation of Missing Water-Level Data for the Everglades Depth Estimation Network (EDEN)

Poster presented July 2010, at the Greater Everglades Ecosystem Restoration Conference

Matthew D. Petkewich1, Paul A. Conrads1, and Brian D. Reece2

1USGS South Carolina Water Science Center, Columbia, SC, mdpetkew@usgs.gov
2USGS Texas Water Science Center, Austin, TX

Background

The Everglades Depth Estimation Network (EDEN) is an integrated network of over 250 real-time water-level gaging stations, ground-elevation models, and water-surface models designed to provide scientists, engineers, and water-resource managers with current (2000-present) water-depth information for the entire freshwater portion of the greater Everglades (fig. 1) (Telis, 2006). A spatially-continuous interpolated water surface across the greater Everglades is generated from daily median water-level values (fig. 2). However, missing or erroneous data can compromise the quality of the modeled water surfaces. To increase the accuracy of the daily water-surface model, a database application was developed to estimate water levels to fill data gaps.

map of southern Florida showing the location of water-level gages in Everglades Depth Estimation Network
Figure 1. Location of water-level gages in Everglades Depth Estimation Network (EDEN). Water Conservation Areas (WCA) 2 and 3 are subdivided by canals. WCA3A is further subdivided into a northern (WCA3AN) and a southern (WCA3AS) region (from Pearlstine and others, 2007). [larger image]

example of Everglades Depth Estimation Network water-surface map for a wet season day and dry season day
Figure 2. Example of Everglades Depth Estimation Network (EDEN) water-surface map for a (A) wet season day and (B) dry season day (from Pearlstine and others, 2007). Vertical datum is North American Vertical Datum of 1988. [larger image]

EDEN Data Gap Estimation Program

To increase the accuracy of the daily water-surface elevation model, linear regression equations to estimate missing data for each gaging station in EDEN were developed (Conrads and Petkewich, 2009). To minimize the inability to estimate data due to missing data from a single input site, three or four regression equations were developed for each site using different input sites. For each site, an order was established for the regression equation to be used to fill a data gap. The order that the equations will be used was based on performance statistics, visual inspection of equation predictions and measured data, and proximity of input and output stations.

Over 740 equations were incorporated into a database application that automatically estimates missing water-level records (EDEN GAP). A Microsoft Access database® of the EDEN water-level data and estimation equations was developed to automate the filling of missing data. The performance statistics computed for each equation provides documentation of the "goodness-of-fit" of the equations (table 1). In addition, although the majority of the equations provide satisfactory estimations of water levels, the performance statistics provide a prioritization for identifying stations where improved equations are needed to provide more satisfactory water-level estimates.

Table 1. Minimum, median, and maximum values for the summary statistics for the estimation equations.

[R2, coefficient of determination; RMSE, root mean square error]

Statistic Minimum Median Maximum
R2 0.01 0.94 1
Mean error –0.19 0 0.25
RMSE 0.02 0.17 1.24
Standard error 0.02 0.16 1.04
Nash-Sutcliffe 0.01 0.94 1
Percent model error 0.40% 4.70% 21.10%
Percent model bias –38.3% 0.00% 32.30%

Once data gaps are estimated, hydrographs of each station are evaluated to delineate potential errant data and to evaluate the fit of any estimates. For some stations, estimated data are a good representation of the missing data and can be used without any adjustment (fig. 3A). For other stations, however, to improve the quality of the water-level estimates, shifting techniques are applied to the estimates similar to gage-height correction techniques used for computing continuous water-level data records (Rantz and others, 1982). The most straight-forward type of data correction is one where a uniform (+/-) correction value is applied to the estimates (fig. 3B). Occasionally prorated corrections are applied over time to improve the quality of the water-level estimates.

Planned improvements to the EDEN GAP application include moving the processing to SQL Server to improve performance and evaluate the possibility of estimating missing data on a daily basis in addition to the current quarterly or annual reviews. A second improvement to EDEN GAP is auto-filling missing data gaps of less than 8 hours by using simple-linear interpolation. Another item is to allow the user of the application to choose which estimates to use on a gap-by-gap basis. For some stations, one site is a better estimator for missing data during the wet season while another station is a better estimator during the dry season.

plot of measured and estimated water-level elevation for Site 9, October 1, 2008 through September 30, 2009
plot of measured, estimated, and shifted water-level elevation for station NTS18, October 1, 2008 through September 30, 2009
Figure 3. Measured and estimated water-level elevation for SITE 9 (A), and measured, estimated, and shifted water-level elevation for station NTS18 (B) during the October 1, 2008 through September 30, 2009. [click on the images above to view larger versions]

Summary

Data-quality evaluation and estimation of missing data can be a time-consuming process, especially for a network as large as EDEN with over 250 gaging stations. To increase the accuracy of the daily water-surface elevation model, an application was developed to address data-quality issues from the network. EDEN GAP estimates water levels to fill data gaps. This program effectively and efficiently addresses data-quality issues by automating many of the processes for data estimation and data validation and will improve the consistency and utility of the EDEN data.

References

Conrads, P.A., and Petkewich, M.D., 2009, Estimation of missing water-level data for the Everglades Depth Estimation Network (EDEN): U.S. Geological Survey Open-File Report 2009-1120, 53 p.

Pearlstine, L., Higer, A., Palaseanu, M., Fujisaki, I., and Mazzotti, F., 2007, Spatially continuous interpolation of water stage and water depths using the Everglades Depth Estimation Network (EDEN): Gainesville, Fl, Institute of Food and Agricultural, University of Florida, CIR1521, 18 p., 2 apps.

Rantz, S.E., and others, 1982, Measurement and Computation of Streamflow: U.S. Geological Survey Water-Supply Paper 2175, 631 p.

Telis, Pamela A., 2006, The Everglades Depth Estimation Network (EDEN) for Support of Ecological and Biological Assessments: U.S. Geological Survey Fact Sheet 2006-3087, 4 p.

http:// sofia.usgs.gov/eden



| Disclaimer | Privacy Statement | Accessibility |

U.S. Department of the Interior, U.S. Geological Survey
This page is: http://sofia.usgs.gov/publications/posters/auto_est_level_eden/index.html
Comments and suggestions? Contact: Heather Henkel - Webmaster
Last updated: 04 September, 2013 @ 02:04 PM(TJE)