Variable Subset Selection 




The selection of variable subsets which are considered adequate as a surrogate for the full data set is a problem of interest  in exploratory data analysis for users of data sets with many variables. 
Work in this area involves the definition and testing of indicators measuring subset quality; the identification and testing of  algorithms to search for subsets that are optimal under those  indicators; and the expansion of the (already existing) "subselect" software module to cater for the new indicators and algorithms.  


Evapotranspiration in very sparse stands of cork and holm oak


Climate in Mediterranean Europe is characterised by dry and hot summers,with rainfall  concentrated in autumn and winter. In these conditions, evapotranspiration plays an important role affecting water resources availability. However, the influence of evapotranspiration is dependent upon the type of soil cover, being recognised that, on an annual basis and for the same climatic conditions, forests evaporate more water than other types of shorter vegetation.
To contribute to estimate the effects of the sparse agroforestry systems of Quercus suber (cork oak) and Q. rotundifolia  (holm oak) of Southern Portugal on water resources, we will measure and model the evapotranspiration of this open mixed oak woodland. To fulfil these objectives  there will be: 1) a continuous measurement of total evapotranspiration and rainfall interception; 2) the estimation of the aerodynamic parameters for the studied stand; 3) the development of an  interception model suitable for the savannah type ecosystems and its inclusion in a global  evapotranspiration model; 4) the modelling of the spatial distribution of throughfall beneath  isolated trees, based on the collection of geographical data, the development of a representation
methodology for the study object and the use of spatial analysis functions. Software prototype to support the proposed models will also be developed.



Connectivity in priority area selection for conservation 


The spatial relations of sites within networks of priority areas for conservation is critical to the long-term maintenance of key genetic, population and ecosystem processes. However, these relations have received relatively little attention in the development of mathematical methods for objectively identifying such networks. The aim is to incorporate connectivity explicitly as part of the models constraints in reserve selection problems. We intend to develop the mathematical ingredients to build a bridge between covers and connectivity in graphs which permit to design practical algorithms for the establishment and maintenance of networks of protected areas toward the long-term persistence of biodiversity.



Landscape diversity measures and landscape desertification processes



The desertification processes are actually occurring in some wide regions and typically are correlated with a loss of diversity of landscape elements and a loss of internal diversity within the habitats. Diversity in Ecology is evaluated with mathematical information theory methods (such as Shannon index and Simpson index) and global biodiversity at the landscape scale involves the two levels above refered and their integration asks for new measures such as generalizations of the previous indices.
We will try to correlate the loss of diversity and desertification level of different landscapes and provide measures of desertification severity processes either in areas of southern Portugal and East Timor.