Predictive Models

     Predictive models for biological communities are based on the use of similarity indices that provide an indication of how a biological community from a specific site is similar to the community of other site, or even to the reference community (De Pauw et al. 2006). In this sense, it could be establish biological classifications or ordinations of sites by means of multivariate analyses, identifying the relationships between the biological communities and present potential perturbations (Norris & Georges 1993).

After the Reference Condition Approach (Reynoldson et al. 1997; Bailey et al. 2004) these models are able to obtain the capture probability of each taxon, using the biological classification obtained from a set of reference sites and its relationship to a set of selected environmental variables (Wright 2000). In this way it is possible to obtain the reference community that would be expected at no-perturbation conditions, and it could be compared with the current community of any test site. 

     These methodologies have been evolved from the pioneer approach develop for aquatic macroinvertebrates in the United Kingdom, RIVPACS (Wright et al. 1984; Moss et al. 1987). Its large international repercussion have provoked the development of a high number of predictive models: AUSRIVAS in Australia (Marchant et al. 1997; Smith et al. 1999; Simpson & Norris 2000), BEAST in Canada (Reynoldson et al. 1995, 2000, 2001), SWEPACSRI in Sweden (Johnson & Sandin 2001), PERLA in the Czech Republic (Kokeš et al. 2006), as well as different predictive models developed in USA, Luxemburg or Portugal (Hawkins et al. 2000b; Ferréol et al. 2005; Feio et al. 2007; Hargett et al. 2007).


     At the beginning of the 90s, Alba-Tercedor (1994) underlined the importance of these methodologies. Afterwards, Alba-Tercedor and Pujante (2000) considered that the idea of developing a RIVPACS-type system for the whole of Spain was completely realistic since a scientific and technical point of view, pointing out the lack of financial support to accomplish the task. In 1998, during the first steps of the WFD, the GUADALMED project was conceived. The first phase of the project finished in 2002 and during its second phase, it was considered the development of predictive models of aquatic macroinvertebrate for Iberian Mediterranean streams. Such predictive models constitute the commonly known MEDiterranean Prediction And Classification System - MEDPACS (Poquet 2007; Poquet et al. 2009), and its implementation is materialized with the current web service.
     Presently there are two types of predictive models implemented into the web service, firstly seasonal models for spring and autumn, and secondly a combined model that allows the evaluation of the ecological status for the combination of several seasons (spring, summer and autumn). In future updates, it will be include new predictive models with a wider application area, providing the versatility required to the web service to evaluate the ecological status, attending in each case to the specific characteristics of each monitoring and control program.

Web application MEDPACS - Predictive Models
needs authentication