Metodos estatisticos para localização otima de estações de mensuração de poluição em rede fixa

AUTOR(ES)
DATA DE PUBLICAÇÃO

1996

RESUMO

Optimal experiments designs are used in this work to define the optimal siting of a network of new pollution-measuring stations, to be contracted or expanded. Firstly we consider a regression model (called Model A) to represent the relationship between measured concentrations and geographic coordinates. Model A shows itself as inadequate for environmental data, which exhibit spatial and time correlations. A random regression coefficient model (called Model B) was next considered to describe the statistical dependence between pollutant concentrations. Parameter estimation and prediction are discussed for this model under three different knowledge levels on the parameter distribution. Next, the D-optimality. criterion is introduced to locate pollution - measuring networks in both models. Whittles s (1973) general equivalence theorem is applied to maximize growth rates provides corresponding to the determinant of the precision matrices for the three knowledge levels. Minimization of these growth rates provides D-optimal plans under the partial knowledge and 110 knowledge hypotheses on the random parameter distribution. The conditions of the theorem are not fulfilled under the full knowledge hypotheses. Mitchell s (1974) DETMAX algorithm is used to generate D-optimal designs. Data by Cressie et ai. (1990) and particulate matter data for the city of Fortaleza are used to illustrate the results. We conclude that the random coefficient mode! seems to be more adequate; the utility of the optimum siting depends however on data determining knowledge levels on the random parameter distribution.

ASSUNTO(S)

otimização matematica poluição - medição estatistica matematica

ACESSO AO ARTIGO

Documentos Relacionados