Using Artificial Neural Network Models to Integrate Hydrologic and Ecological Studies of the Snail Kite Falcon in the Everglades, USA

Paul A. Conrads¹, Ruby Daamen², Edwin A. Roehl², Wiley M. Kitchens³ and Christa Zweig^3
1USGS South Carolina Water Science Center, Columbia, S.C. USA
²Advanced Data Mining, Greenville, S.C., USA
³USGS Florida Cooperative Fish and Wildlife Research Unit, Gainesville, FL., USA

Hydrologists and ecologists have been working on integrating a long-term hydrologic data network and a short-term ecological database to support ecological models of the habitat of the snail kite, a threatened and endangered bird. Hydroperiods of water depths have a significant affect on the nesting and foraging of the snail kite. Data mining techniques, including artificial neural network (ANN) models, were applied to simulate the hydrology of snail kite habitat in the Water Conservation Area 3A. Seventeen water-depth recorders are co-located at transects where extensive plant sampling is ongoing. These continuous recorders were established in 2002. A long-term network of three water-level recorders has been maintained since 1991 by the USGS. Using inputs representing the three long-term gages, very accurate ANN models were developed to predict the water levels at the 17 short-term sites. The models were then used to hindcast water levels at the 17 short-term sites back to 1991. The result was extended water-level records to help scientists better learn how the snail kite's habitat is affected by changing hydrology.

A Decision Support System (DSS) was developed to disseminate the models in an easily used package. The DSS is a MS Excel^TM/VBA application that integrates the models and database with interactive controls and streaming graphics to run long-term simulations. As part of the Everglades restoration Interim Operating Plan (IOP), a regional hydrologic model is used to generate water levels for alternative flow regulation schedules. The alternative IOP water levels are input to the DSS to predict the hydrology of the snail kite habitat. The application demonstrates how very accurate empirical models can be built directly from data and readily deployed to end-users to support interdisciplinary studies.

Contact Information: Paul A. Conrads, USGS South Carolina Water Science Center, Stephenson Center – Suite 129, 720 Gracern Road, Columbia SC, 29210 USA, Phone: 803.750.6140, Fax: 803.750.6181, Email: pconrads@usgs.gov

(This abstract is from the 2006 Greater Everglades Ecosystem Restoration Conference.)

Back to Project Homepage