ANALYSIS OF BIVARIATE CENSORED LOW FLOWS

In this paper, a bivariate model is established to characterize the se ven-day, two-year (Q(7,2)) and seven-day, 10-year (4(7,10)) low flows of streams in West-Central Florida under the restriction of Q(7,10) le ss than or equal to Q(7,2). Analysis of prediction errors showed that our model describes these stream low flows well. The measurements unde r the detection limit were treated as censored data and a bivariate im putation method was developed to impute them into pseudo-complete samp les. All-subsets regression was applied to these imputed data for sele cting appropriate models, which link the low flows to their basin char acteristics. The parameters are first estimated by the least-squares m ethod in a bivariate normal regression and then adjusted to yield thei r maximum likelihood estimates. This process of imputation, model sele ction and parameter estimation is repeated iteratively until convergen ce of the selected model terms, parameter estimates, and imputed data. With the established model, predictions of low flows can be made at g auged and ungauged sites according to their basin variables for water- resources management.