Neural network applications in Ecology and Oceanography. (c) Michele Scardi, 2000

Visit the web site of the 4th Conference of the International Society for Ecological Informatics that was held in Korea (October 24-28, 2004)!

If you want some background info, you can have a look at the previous Conference web site too.

The next Conference has been announced and will be held in Cancun (Mexico) in fall 2006 (more info here ASAP).

Have a look at the web page of Ecological Informatics, a new journal published by Elsevier.

What is a neural network?

Neural networks are powerful computational tools that can be used for classification, pattern recognition, empirical modeling and for many other tasks. Neural networks (or artificial neural networks - a longer but more correct definition) can be "trained" to provide the right output (binary, fuzzy, quantitative) if enough input-output patterns are available and if these patterns effectively describe the system that is to be modeled.

A very quick introduction to the most common neural networks, i.e. error back-propagation neural networks, is available here, as well as an example of calculations with a trained neural network. However, if you really want to peek under the neural network hood, check out the neural network primer by Herve Abdi or the textbook by Ben Kröse and Patrick van der Smagt (both of them are freely downloadable).

If you're interested in Self-Organizing Maps (aka Kohonen networks), a nice introduction can be found here (unfortunately most links in this page are out of date), whereas the SOM Toolbox home page, which is managed by Kohonen's group, is the first stop for people who want a very quick start.

For Italian people who don't like reading English docs, I recently added two PDF files:

"Applicazione delle reti neuronali alla valutazione della produzione primaria e della biomassa fitoplanctonica" (some notes about NN for a remote sensing course, delle note sulle reti neuronali per un corso di telerilevamento);

"Calibrazione di dati telerilevati di biomassa e produttività fitoplanctonica nell'Oceano Australe" (a poster about NN applications in the Southern Ocean, un poster sulle applicazioni di reti neuronali nell'Oceano Australe).

My own neural networks applications.

My first neural network application dealt with phytoplankton primary production modeling. At present, I'm developing different neural network applications and some information about them can be found (or will be found, as lazy people are used to say) in this page. Here are some of them:

Phytoplankton primary production, a global model that is participating in the Primary Production Algorithm Round Robin for the selection of a "consensus" algorithm for SeaWiFS data processing.
Phytoplankton primary production, a comparison with other models and an example of application in the Mediterranean Sea (see the poster I presented at "Progress in Oceanography of the Mediterranean Sea", Rome 17-19/11/97).
Phytoplankton primary production, applications of NN models at different spatial scales.
Phytoplankton primary production, a Chesapeake Bay model (in collaboration with Larry Harding, University of Maryland). A case history about this model was presented at the "International Workshop on Applications of Artificial Neural Networks to Ecological Modelling", Toulouse, 14-17 December 1998 (see the abstract). A Java implementation of the model was also developed (click here to play with it).
Adult and juvenile eel catches in the River Tevere (in collaboration with Stefano Cataudella and coworkers, University of Rome "Tor Vergata").
Calibration-validation of oxygen data collected by CTD probes (in collaboration with Maurizio Ribera d'Alcalà and Fabio Conversano, Stazione Zoologica "A. Dohrn" of Naples). I already used this method for some data sets and a full paper about it is almost ready.
Recovering information from old scientific paper: recognition in modern data sets of the zoocoenoes of the Adriatic Sea as defined by an Italian marine biologist (A. Vatova) in the thirties.
PAEQANN - Predicting Aquatic Ecosystem Quality using Artificial Neural Networks: impact of environmental characteristics on the structure of aquatic communities (algae, benthic and fish Fauna): this is the title of a 5th Framework Research Project that I will carry out during the next three years together with other european partners. The coordinator of the Project is Sovan Lek (France) and a dedicated web page is already available.
Prediction of fish assemblage structure and assessment of the Ecological Status of streams (as defined in the European Water Framework Directive), in a dedicated page in this site (Italian only, sorry!).
Several other will be soon available (later in 2005?)...

My contributions to past events.

Scardi M. - Neural networks: a new approach to phytoplankton production modeling. International Conference "Progress in Oceanography of the Mediterranean sea", Rome, Italy, 17-19/11/1997

Scardi M. - Empirical models in ecology: some applications of artificial neural networks. VII International Congress of Ecology, Florence (Italy), 19-25 July 1998 (Symposium 5.6, 21 July 1998, 8:00-16:30)

Tozzi S., D'Ortenzio F., Santoleri R. & Scardi M. - Comparazione e validazione dei dati SeaWiFS con misure in situ nel Canale di Sicilia. XIV Congresso A.I.O.L., Ancona, Italy, 1998

Scardi M., Tozzi S., Di Dato P. & Fresi E. - Calibration of remotely sensed phytoplankton biomass and productivity data in the Southern Ocean. Atti 1.o Convengo Nazionale delle Scienze del Mare, Ischia, Italy, 11-14/11/98

Scardi M. - Neural network models of phytoplankton primary production. International Workshop on Applications of Artificial Neural Networks to Ecological Modelling, Toulouse (France), 14-17 December 1998 (see reference)

Scardi M. & L.W. Harding, Jr. - Developing an empirical model of phytoplankton primary production: a neural network case study. International Workshop on Applications of Artificial Neural Networks to Ecological Modelling, Toulouse (France), 14-17 December 1998 (see reference)

Harding, Jr., L. W, Hood R.W. & Scardi M. - Enhanced phytoplankton biomass and productivity associated with persistent mesoscale eddies in the lower Chesapeake Bay. ASLO Aquatic Sciences Meeting, Santa Fe, New Mexico (USA), February 1-5, 1999.

Harding L.W., Jr., Itsweire E.C., Esaias W.E. & Scardi M. - An aircraft observing system in Chesapeake Bay: Remote sensing of chlorophyll 1989-1999. Estuarine Research Federation, New Orleans, Louisiana, September, 1999 (Invited talk, chaired session for T.C. Malone).

Scardi M., Lek S., Lim P., Di Dato P. & Oberdorff T. - Artificial neural networks as a tool for predicting fish community composition in rivers. The Third World Congress of Nonlinear Analysts (WCNA-2000), Catania (Italy), July 19-26, 2000 (PDF manuscript)

Scardi M. - Advances in neural network modeling of phytoplankton primary productivity. 2nd International Conference on Applications of Machine Learning to Ecological Modelling, Adelaide (Australia), 27 November - 1 December 2000 (click here for the conference web site or PDF reprint)

Lek S., Jorgensen S.E., Scardi M., Descy J.P., Coste M., Ector L., Verdonschot P. & Knoflacher M. - PAEQANN project: Predicting Aquatic Ecosystem Quality using Artificial Neural Networks. Impact of Environmental characteristics on the Structure of Aquatic Communities (Algae, Benthic and Fish Fauna).
2nd Symposium for European Freshwater Sciences (SEFS-2), July 8-12, 2001, Toulouse, France

Scardi M., Cataudella S., Di Dato P., Maio G., Marconato E., Salviati S., Tancioni L., Turin P. & Zanetti M. - Selecting predictive variables in artificial neural network modeling of fish community composition: an ecologist's perspective. Workshop "Parameter selection in modelling aquatic community structure", Namur (Belgium), September 15 - 16, 2001.

Mancini L., Formichetti P., Tancioni L., Di Dato P. & Scardi M. - Analysis of the role of some predictive variables in artificial neural network modeling of macrobenthic fauna composition in the Tevere river basin. Workshop "Parameter selection in modelling aquatic community structure", Namur (Belgium), September 15 - 16, 2001.

Lek S., Park Y.S., Gevrey M. , Scardi M. & Oberdorff Th. - Predicting riverine fish assemblages using artificial neural network models. Workshop "Parameter selection in modelling aquatic community structure", Namur (Belgium), September 15 - 16, 2001

Scardi M., Cataudella S., Di Dato P., Maio G., Marconato E., Salviati S., Tancioni L., Turin P. & Zanetti M. - Previsione della struttura della fauna ittica mediante reti neurali artificiali. IX Convegno AIIAD, Acquapartita (Fo), Italy, 11-13/6/2002

Several other applications have been presented since 2002 and will be available ASAP (hopefully!)...

Some other ecological applications.

Information about some other interesting ecological applications of neural networks can be found in my (incomplete) reference list and at the following sites:

Assessing benthic impacts of fish farming with an expert system based on neural networks, a preprint by M.J. Baptist (Delft Hydraulics) and W. Silvert (Bedford Institute of Oceanography). A nice slide show on the same topic is also available.
Chlorophyll a level prediction in freshwater basins by Szilvester Juhos (Balaton Limnological Research Institute).
Neural network applications in landscape ecology (Geographic Modeling Systems Lab).
F. Recknagel's ecological applications (at the Aquatic Ecology Lab, Dept. of Soil and Water, University of Adelaide, Australia).

General information and interesting NN sites.

If you need more information about neural networks or if you are looking for papers, software, etc., take a look at:

The comp.ai.neural-nets newsgroup FAQ, where can also be found a lot of information about literature and links to freely available neural network software (difficult to log in, it's a very busy FTP site, but this is the best start for neural network newbies).
A very quick introduction to neural networks at the Geographic Modeling Systems Lab.
The home page of the The Pacific Northwest Laboratory, where several links to neural network resources and a Gateway to the World of neural networks can be found.
Neural network for remote sensing problems: a slide presentation by V. Krasnopolsky.
Do you want to see how a simple NN works? Check the interactive demo applet by Fred Corbett at the University of Manitoba.
An outstanding NN site developed by Jochen Fröhlich.
The Don Tveter's web site is a vary handy reference for backpropagation and other AI topics.
Are you a Matlab addicted? Check the NNSYSID toolbox (it's free and very nice).
Do you need a complete package for serious applications? Take a look at SNNS, the Stuttgart Neural Network Simulator. A caveat: even though it's free and well documented, SNNS needs an X-Window environment, so it could be a little tricky to set up in a Windows95/NT system.

Last update: 7/12/2000

Back | Home