Combining sampling, profiling, remote sensing and modelling to efficiently monitoring estuarine and coastal water quality

Monitoring of coastal and estuarine water quality has been traditionally performed by sampling with subsequent laboratory analysis. This has the disadvantages of low spatial and temporal resolution and high cost. In the last decades two alternative techniques have emerged to overcome this drawback: profiling and remote sensing. Profiling using multi-parameter sensors is now in a commercial stage. It can be used, tied to a boat, to obtain a quick “picture” of the system. The spatial resolution thus increases from single points to a line coincident with the boat track. The temporal resolution however remains unchanged since campaigns and resources involved are basically the same. The need for laboratory analysis was reduced but not eliminated because parameters like nutrients, microbiology or metals are still difficult to obtain with sensors and validation measurements are still needed. In the last years the improvement in satellite resolution has enabled its use for coastal and estuarine water monitoring. Although spatial coverage and resolution of satellite images in the present is already suitable to coastal and estuarine monitoring, temporal resolution is naturally limited to satellite passages and cloud cover. With this panorama the best approach to water monitoring is to integrate and combine data from all these sources. The natural tools to perform this integration are numerical models. Models benefit from the different sources of data to obtain a better calibration. After calibration they can be used to extend spatially and temporally the methods resolution. In Algarve (South of Portugal) a monitoring effort using this approach is being undertaken. The monitoring effort comprises five different locations including coastal waters, estuaries and coastal lagoons. The objective is to establish the base line situation to evaluate the impact of Waste Water Treatment Plants design and retrofitting. The field campaigns include monthly synoptic profiling, using an YSI 6600 multi-parameter system, laboratory analysis and fixed stations. The remote sensing uses ENVISAT\MERIS Level 2 Full Resolution data. This data is combined and used with the MOHID modelling system to obtain an integrate description of the systems. The results show the limitations of each method and the ability of the modelling system to integrate the results and to produce a comprehensive picture of the system.

A hybrid training method for B-spline neural networks

Current and past research has brought up new views related to the optimization of neural networks. For a fixed structure, second order methods are seen as the most promising. From previous works we have shown how second order methods are of easy applicability to a neural network. Namely, we have proved how the Levenberg-Marquard possesses not only better convergence but how it can assure the convergence to a local minima. However, as any gradient-based method, the results obtained depend on the startup point. In this work, a reformulated Evolutionary algorithm - the Bacterial Programming for Levenberg-Marquardt is proposed, as an heuristic which can be used to determine the most suitable starting points, therefore achieving, in most cases, the global optimum.

Modelling health state preference data using fixed and random effects models

The use of preference-based measures of health in the measurement of Health Related Quality of Life has become widely used in health economics. Hence, the development of preference-based measures of health has been a major concern for researchers throughout the world. This study aims to model health state preference data using a new preference-based measure of health (the SF- 6D) and to suggest alternative models for predicting health state utilities using fixed and random effects models. It also seeks to investigate the problems found in the SF-6D and to suggest eventual changes to it.