FLORIDA PROJECTS

• Aquifer Characterizations

• Availability of Water

• Climate Change and Hydrologic Extreme Events

• Ecosystems

• Hydrologic Modeling
• Analysis of Groundwater and Lake Level Data in the CFCA
• Coupled Groundwater and Hydrodynamic Surface-Water Modeling
• Deep-well injectate in Miami-Dade County
• Freshwater-Saltwater Interface in Broward County
• Groundwater Flow and Solute Transport Models
• Integrated Model of Surface-Water and Groundwater Flow
• Modeling Effects of Deepening the SJR Shipping Channel
• Simulation of Transient Groundwater Flow in East-Central FL
• Surface-Water and Groundwater Flows to Lake Tsala Apopka
• Surface-Water influx to the Biscayne aquifer in response to wellfields

• Surface Water

• Water Quality Conditions

• Wetlands

ABOUT FLORIDA SCIENCE

• Office Information:
  • Florida Water Science Center office locations
  • Information Requests

• National Information:
  • USGS Phonebook
  • USGS Science Strategy
  • Employment

Ask Florida USGS

Quantification of natural and anthropogenic effects in groundwater and lake-level data in the Central Florida Coordination Area (CFCA)

Project Chief: Andrew M. O’Reilly, and Paul A. Conrads (South Carolina WSC)
Cooperator: St. John’s River Water Management District; Southwest Florida Water
Period of Project: October 2008 – December 2012

Problem Statement

Figure 1. Map showing Central Florida Coordination Area and site locations of historical hydrologic data. Hydrologic data comprise 963 sites in total:  470 wells, 22 springs, 307 lakes, 143 rain gages, and 21 air temperature sites. Artificial neural network models are being developed for the 51 sites shown in blue.

As a result of increasing development in central Florida, there is growing concern about the adequacy of available groundwater resources to meet future demand for water supply. Water management decisions must be made on the allocation of groundwater resources for both regulatory and planning purposes. Considerable historical groundwater and lake-level data exist, yet a comprehensive compilation and systematic analysis of these data in the Central Florida Coordination Area (CFCA) has not been performed to ascertain the effects of both natural and anthropogenic stresses on the hydrologic system (fig. 1).

Objectives

1) Identify and quantify both natural and anthropogenic effects in historical groundwater-level, spring flow, and lake-level data by using artificial neural networks (ANN) and other data-mining techniques

2) Develop a decision support system (DSS) composed of multiple site-specific ANN models to predict groundwater levels, spring flows, and lake levels that may be used to evaluate scenarios of interest to water managers

3) Compare groundwater system behavior simulated by physics-based groundwater flow models with the behavior exhibited in the historical record as quantified by the data-mining analysis

Figure 2. Example of predicted groundwater level for a site-specific artificial neural network developed for an Upper Floridan aquifer well in Orange County using rainfall inputs only. Due to the potential correlation between rainfall and groundwater pumpage, final artificial neural network model results incorporating both rainfall and water-use data may yield a different relation between rainfall and groundwater level.

Approach

Objectives 1 and 2 were addressed using a 5-step methodology.  Steps 1 and 2 involved acquiring and compiling the data for data mining. Signal processing in Step 3 and data mining analyses performed in Step 4 were performed iteratively to obtain the broadest possible understanding of the process and how to best construct the predictive models. Using the results of Steps 1-4, Step 5 produced the "production quality" ANN models in a form that can be evaluated by water managers and other stakeholders. Each step is described in more detail as follows:

Step 1 – Data Compilation.
Available hydrologic data (groundwater levels, spring flows, lake levels, rainfall, air temperature, and potential evapotranspiration) were obtained from the responsible agencies (USGS, water management districts, NOAA, and local governments) (fig. 1). Groundwater withdrawal data was provided by the water management districts.

Step 2 – Data Quality Assurance.
The data obtained from the responsible  agencies were evaluated for quality-assurance purposes.

Step 3 – Data Preparation.
Methods were applied to maximize the information content in the raw data, while diminishing the influence of poor or missing measurements. Signal (time series) processing methods included clustering, filtering, spectral decomposition, estimation of data characteristics and time delays, and synthesizing missing data.  A subset of 51 sites was selected from the results for detailed ANN modeling.

Step 4 – Correlation and Sensitivity Analysis.
Correlation analysis was performed to quantify the strength of association, or covariation, between pairs of variables, providing a detailed understanding of the data.

Step 5 – Predictive Modeling.
Site-specific ANN models (for example, see fig. 2) were developed, trained, and incorporated into the DSS for final delivery.

Objective 3 was addressed by comparing predictions for the 51 site-specific ANN models with a physics-based groundwater flow model. Simulated values from the USGS ECFT MODFLOW model were used (see USGS Project Number DHW00). Comparisons were made using model performance statistics (R2, RMSE, ME, Nash-Sutcliffe) and graphical analysis (scatter plots, residual error time series, frequency distribution curves, cumulative Z-scores).

Results

Not yet available.

Information Product

A Microsoft AccessTM , database of hydrologic data for the sites shown in figure 1 will be created, a DSS developed as a Microsoft Excel™/Visual Basic for Applications (VBA) program will be completed, and a USGS Scientific Investigations Report (SIR) will summarize the results of the project.