The present work is about the development of analytical strategies for the determination of mercury in environmental samples using atomic absorption spectrometry with cold vapor generation (CV AAS). Mercury is a metal known to be toxic to the health of humans and ecosystems, standing out to have this toxicity even in low concentrations. Several studies reported in the literature show the need for the development of new, more sensitive methodologies for determination and analysis of mercury speciation, since this analyte is characterized by trace concentrations and is one of the main current concerns when it comes to contamination environmental. Within this context, two papers were developed. The first one consisted in the development of a method of determination of mercury in sediment samples by optimization of a simple procedure of extraction of the analyte of the samples for later determination by CV AAS. Such optimization was performed using multivariate optimization techniques, evaluating the influence of the factors in the analyte extraction step. The factors evaluated were: hydrochloric acid concentration, thiourea concentration, sonication temperature and sonication time. After optimization, the accuracy of the method was confirmed by the analysis of two certified reference materials, Channel Sediment (BCR-320R) and San Joaquin Soil (NIST 2709a). The detection and quantification limits were 1.04 and 3.46 ng g-1, respectively, and the precision, in terms of RSD%, was 4.31%. The method was applied for the determination of mercury in sediment samples collected in the São Paulo Estuary, located in the state of Bahia, Brazil, whose concentrations ranged from 13.8 to 38.5 ng g-1. The second work involved the development of an automated system for the determination of mercury in natural water samples employing multiserial flow injection analysis (MSFIA) and cold vapor generation atomic absorption spectroscopy (CV AAS). The experimental conditions were investigated and optimized using multivariate optimization methodology, such as complete factorial design (2k) and Doehlert Matrix. The optimized variables were: hydrochloric acid concentration, sodium borohydride concentration, sample flow and entrainment gas flow. The proposed method allowed the direct determination of mercury in natural water samples, that is, without preconcentration stage, with limits of detection and quantification of 43.0 and 145.0 ng L-1, respectively. The accuracy of the method was confirmed by analysis of certified water reference material, Trace elements in waste water, (CASS-4). The matrix effect was evaluated by addition and recovery tests in saline water samples, as the method was optimized using aqueous standards, obtaining recoveries above 82%. The method was applied in a sample of sea water and lake of the city of Salvador, Bahia, Brazil, whose concentrations ranged from 0.36 to 1.14 μg L-1.