In the present work, a study is reported of the potential of coupled-column liquid chromatography (LC) applied to the determination of triazine residues in environmental water samples. For this purpose, two different techniques have been compared: on-line trace enrichment followed by LC (SPE-LC) and coupled-column liquid chromatography (LC-LC). First, a completely automated liquid chromatographic method based on on-line trace enrichment in a prepacked precolumn and using diode array detection has been developed for the simultaneous trace-level determination of six triazine herbicides (simazine, cyanazine, atrazine, terbumeton, terbuthylazine, and terbutryn) and the main atrazine metabolites (desisopropylatrazine, desethylatrazine, and hydroxyatrazine). After preconcentration parameters were optimized by testing two different sorbents (C18 and PRP-1) in three cartridges with different dimensions, a sample volume of 100 mL was selected in order to achieve maximal solute preconcentration. Detection limits lower than 0.1 microgram.L-1 were obtained even for the most polar analyte (desisopropylatrazine), which presented recoveries of around 30%. The method was validated by means of recovery experiments in groundwater and surface water samples spiked with the analytes at different levels (0.2-2 micrograms.L-1). Afterward, the procedure was successfully applied in a program for monitoring of triazine residues in surface water carried out in a wet area of Castellón, Spain. Different triazine herbicides such as simazine, terbumeton, terbuthylazine, and terbutryn were identified and quantified. The identity of these compounds was confirmed by their absorption UV spectra and by GC/MS analysis. Finally, two rapid, sensitive, and selective procedures, previously developed in our laboratory for the trace-level determination of triazine compounds, both based on LC-LC, were compared with the former procedure. The SPE-LC approach showed a considerable improvement in the global sensitivity at the expense of a decrease in selectivity as well as in sample throughput.