Progress Report (September)

Initial Goal : At the beginning of the semester the group defined our goal for our thesis project as:

?to simulatethe red tide phenomenon using Neural Networks.We opt to use artificial neural networks because this can fit the complexity and the nonlinearity of this ecological phenomena.The software will have the capability of taking into consideration the various dynamic environmental parameters as salinity, temperature, nutrients and growth factors, and hydrologic and meteorologic factors.If time permits, we opt to develop this software into one which can predict the occurrence of red tides.

What we did?

The group did a comprehensive reading on the biological part of their topic.Upon reading we learned about the many factors affecting the growth of phytoplanktons.Until now, even the marine biologists and oceanographers themselves are having a difficulty in defining the red tide phenomenon.One scientist even claimed that:"?our understanding of the interaction of these conditions and microalgal ecology is imperfect, and we are far from being able to predict the consequences of perturbation in these conditions.Furthermore, all these conditions are subject to natural fluctuation?disentangling cause and effect becomes virtually impossible.?(Dr. Rhodora Azansa).

The readings on Red Tide were far from being encouraging so the group channeled their attention on the study of Neural Networks.In this study, we were enlightened by the fact that even if the biological scientists, with their expertiseare doubting the feasibility of modeling the red tide phenomenon, the neural network provided that it be fed with the necessary data can learn to do the job.

To be able to come up with a neural network model of the red tide one reading suggested that 8 years of time series data for training plus two years for testing are needed.

Alongside reading, the group regularly corresponded with experts from the Marine Science Institute(MSI), Bureau of Fisheries and Aquatic Research(BFAR), and the University of Maine (Donald Anderson).In addition, we have also been able to attend several seminars on red tide modeling provided by MSI.

Through extensive study of the research topic the group has found out that their initial goal is too broad given the time and data constraints.We have therefore decided to narrow down our subject.

The group has its main sources of data, the Marine Science Institute and the Bureau of Fisheries and Aquatic Resources.

BFAR has agreed to give us initially, 5 years of the Shellfish Toxicity Data they?ve gathered (1987 to 1992) from Manila de Bay provided we do a project for them.This is one of the factors of our narrowing down since the data available to us is limited.To fully create a model for the Red Tide will require more data than just Shellfish Toxicity (needed will be wind direction, and other meteorological data which unfortunately are not available).

Currently BFAR has 27 monitoring stations along the coast of Manila Bay.The bureau's main concern now is to cut back on expenses by pruning out of these stations ones that are redundant.

In addition, doable still is to identify which of the stations are primary and secondary and if ever some areas along the coastline are unprotected identify those gaps.

What Neural Planktons can do?

The goal is therefore to reduce the total cost in resources that these institutions are using without giving away the reliability and consistency of their data or without paralyzing the capability of their stations to successfully monitor the starting occurrences of possible red-tide phenomenon. We are then faced with a 3-fold problem of (1) reducing the number of stations that are required to monitor a certain area and reveal deficiencies with respect to the number of stations and their respective positions, (2) identifying primary and secondary monitoring stations to limit monitoring overhead to the primary stations first and then to the secondary stations upon occurrence of red-tide, and (3) given some new data gathered from a number of stations, to be able to identify stations that should be taken into consideration in monitoring a possible red-tide outbreak. 

The project may sound simple enough just as how computer chess and robot soccer (in AI) were treated as mere games but are now used in vast fields in society.But we see other future useful applications of this research.This may be just one of the few steps into red tide modeling and prediction.

Home