Data from five research cruises performed between 1997 and 2001 were processed in order to investigate the potential for improving remote sensing algorithms in the Estuary and Gulf of St. Lawrence. Measured in situ parameters included concentration-dependent indicators of the three critical, optically-active constituents, chlorophyll, Coloured Dissolved Organic Matter (CDOM) and Suspended Particulate Matter (SPM). The radiometric dataset used to investigate different types of algorithms consisted of multi-band above-surface remote sensing reflectance (R[subscript rs]) estimates. These estimates were computed from downwelling surface irradiance and upwelling sub-surface radiance measurements acquired using a SeaWiFS Profiler Multichannel Radiometer (SPMR). The chlorophyll data varied from approximately 0.1 to 17.3 mg.m[superscript -3] in the study region which extended from stations near the Saguenay River to the outer extremes of the Gulf of St. Lawrence. The CDOM and SPM concentration indicators were lower in the Gulf compared to the Estuary. Moderate correlation between in situ measurements was found between chlorophyll and SPM, as well as between CDOM and SPM. Chlorophyll and CDOM were virtually uncorrelated. The standard SeaWiFS Case-I chlorophyll retrieval algorithm, OC4v4, was applied to SPMR data acquired over a significant number of sampling stations (N=169). Algorithm shortcomings were noted when the OC4v4 algorithm was applied directly to the study region. Specific shortcomings, the overestimation of low, and underestimation of high chlorophyll concentrations were consistent with previous findings in coastal regions and particularly with previous findings in the NW Atlantic and in high latitude regions. In addition, the algorithmic output was found to be fairly strongly correlated with CDOM and SPM. A perturbation approach, based on the analysis of residuals between OC4v4 estimates and in situ data, showed that the retrieved chlorophyll biases (overestimates) were dependent on SPM and CDOM (especially at low in situ chlorophyll concentrations). An analysis of the spectral parameters (band ratios and spectral slopes) with respect to in situ constituent concentrations showed that both band ratios and band slopes have a greater dependency on CDOM and/or SPM than on chlorophyll. This observation was supported by radiative transfer calculations which showed that the variability of the blue-to-green band ratios due to changes in CDOM and SPM concentrations could be greater than the variability due to changes in chlorophyll concentration. These findings showed that there was no adequate, single band-ratio algorithm for the remote sensing of chlorophyll in our study region. Systematic testing of a large combination of spectral parameters within the context of specific algorithmic formulations resulted in seven prescribed algorithms which provided slight to moderate improvement in the correlation coefficients and root mean square errors relative to in situ chlorophyll and significant decorrelation relative to CDOM and SPM parameters. In general, algorithms based on multiple spectral parameters were more accurate predictors of in situ chlorophyll. In addition to new algorithms, a set of previous algorithms developed by Jacques (2000) for a subregion of the Estuary were validated in the present study. This validation demonstrated a rather remarkable robustness of correlations between in situ and spectral parameters across time and for different types of instruments and measuring conditions. A relatively smaller number of matching SeaWiFS pixels (N=39) and in situ measurements were used to evaluate the performance of the SPMR-derived algorithms. The accuracy of all algorithms deteriorated when applied to satellite data (one possible reason being the shortcomings of the atmospheric correction algorithm, as underscored by the existence of negative values in the reflectance data). Nonetheless, the improvement of the two selected algorithmic formulations relative to the OC4v4 algorithm showed a certain robustness in the face of environmental influences such as atmospheric effects and sensor response variations. Model simulations showed significant shortcomings of the new algorithms in specific turbidity conditions. The selected algorithms were shown to achieve chlorophyll retrievals which were as good as or better than OC4v4 retrievals. Even though the APD<35% accuracy target of the SeaWiFS project could not be reached, new algorithms succeeded to decrease the APD of the remote estimations from 226% to 65% for SPMR data, and from 502% to 95% for SeaWiFS data. In general, our findings showed that the selected algorithmic formulations had the potential for improving chlorophyll retrieval in the St. Lawrence Estuary and Gulf."--Résumé abrégé par UMI.