La télédétection spatiale permet un suivi à large échelle spatio-temporelle de la concentration en Chlorophylle-a (Chl-a) comme proxy des microalgues dans l’océan mais son application à la zone littorale est un défi en raison de la variabilité et complexité de l’atmosphère, de la turbidité, et de l’hétérogénéité des constituants colorés en suspension dans les eaux intertidales. Ce travail de thèse a pour objectif d’améliorer la télédétection de la Chl-a dans trois sites intertidaux turbides de la façade atlantique Française: les baies de Marennes-Oléron et Bourgneuf, et l’estuaire de la Loire. En premier lieu, une approche originale basée sur l’utilisation de la télédétection hyperspectrale aéroportée a été proposée pour valider la correction atmosphérique des données satellites MERIS. Pour les eaux très turbides (concentration en matière en suspension > 50 g m-3), la méthode FLAASH s’est avérée être la plus performante. En second lieu, des algorithmes d’inversion de la Chl-a ont été régionalisés à partir de données de réflectance acquises in situ dans les trois sites. Plusieurs modèles basés sur la combinaison des bandes rouge et proche-infrarouge de la réflectance marine ont donné de bons résultats, mais une variabilité spatiale a été mise en évidence d’un site à l’autre. Cette observation suggère le développement d'un algorithme multi-conditions qui s'adapterait à la diversité optique des eaux littorales. Pour les eaux les plus turbides, une méthode robuste a été développée pour la détection de la Chl-a. L’algorithme est applicable aux données MERIS (2002-2012) et OLCI (2016-présent), permettant le suivi des variations de Chl-a sur plusieurs décennies. / Spatial remote sensing makes it possible to monitor the variation of Chlorophyll-a concentration (a proxy of microalgae suspended in seawater) at large spatiotemporal scale in the ocean, but its applicability to the coastal zone is a challenge due to atmospheric variability, seawater turbidity, and heterogeneity of suspended colored constituents. The objective of the present study is to improve Chl-a remote sensing in three turbid intertidal sites of the French Atlantic coast: Marennes-Oléron Bay, Bourgneuf Bay, and the Loire estuary. First, an original method using hyperspectral airborne remote sensing was proposed to validate the atmospheric correction of MERIS satellite data. For very turbid waters (suspended particulate matter concentration > 50 g m-3), the FLAASH algorithm appears as the most efficient method. Then, several regional Chl-a algorithms were developed using in situ reflectance measurements acquired in the three study sites. Bio-optical models using a combination of the red and near-infrared spectral bands of the marine reflectance led to satisfactory results, but were variable from one site to another. The implementation of a multiconditional algorithm would therefore be recommended in order to better take into account the optical diversity of nearshore waters. For the most turbid waters, a robust method was validated for the detection of Chl-a. The algorithm is applicable to MERIS (2002-2012) and OLCI (2016-present) data, thus allowing the monitoring of Chl-a during several decades in turbid intertidal waters.

Conselho Nacional de Desenvolvimento Científico e Tecnológico / Este trabalho teve como objetivo principal implementar um algoritmo empírico para o monitoramento do processo de eutrofização da Baía de Guanabara (BG), Rio de Janeiro (RJ), utilizando dados de clorofila-a coletados in situ e imagens de satélite coletadas pelo sensor MERIS, a bordo do satélite ENVISAT, da Agência Espacial Européia (ESA). Para a elaboração do algoritmo foi utilizada uma série histórica de clorofila-a (Out/2002 a Jan/2012) fornecida pelo Laboratório de Biologia Marinha da UFRJ, que, acoplada aos dados radiométricos coletados pelo sensor MERIS em datas concomitantes com as coletas in situ de clorofila-a, permitiu a determinação das curvas de regressão que deram origem aos algorítmos. Diversas combinações de bandas foram utilizadas, com ênfase nos comprimentos de onda do verde, vermelho e infra-vermelho próximo. O algoritmo escolhido (R = 0,66 e MRE = 77,5%) fez uso dos comprimentos de onda entre o verde e o vermelho (665, 680, 560 e 620 nm) e apresentou resultado satisfatório, apesar das limitações devido à complexidade da área de estudo e problemas no algoritmo de correção atmosférica . Algorítmos típicos de água do Caso I (OC3 e OC4) também foram testados, assim como os algoritmos FLH e MCI, aconselhados para águas com concentrações elevadas de Chl-a, todos com resultados insatisfatório. Como observado por estudos pretéritos, a Baia de Guanabara possui alta variabilidade espacial e temporal de concentrações de clorofila-a, com as maiores concentrações no período úmido (meses: 01, 02, 03, 10, 11 12) e nas porções marginais (~ 100 mg.m-3), particularmente na borda Oeste da baia, e menores concentrações no período seco e no canal principal de circulação (~ 20 mg.m-3). O presente trabalho é pioneiro na construção e aplicação de algoritmos bio-óptico para a região da BG utilizando imagens MERIS. Apesar dos bons resultados, o presente algorítmo não deve ser considerado definitivo, e recomenda-se para trabalhos futuros testar os diferentes modelos de correção atmosférico para as imagens MERIS. / This work aimed to implement an empirical algorithm for monitoring the process of eutrophication at Guanabara Bay (BG), Rio de Janeiro (RJ), using in situ chlorophyll-a data and satellite images by MERIS sensor, onboard ENVISAT satellite, from European Space Agency (ESA). A time series of chlorophyll-a (Dec / Jan 2002/2012) provided by Marine Biological Laboratory from UFRJ, was used to elaborate the algorithm, coupled with the radiometric data collected by MERIS sensor on concurrent dates with the collections, what allowed the determination of the regression curves that gave rise to algorithms. Several band combinations were used, with emphasis on wavelengths of green, red and near infrared. The algorithm chosen (R = 0.66 and SRM = 77.5%) made use of wavelengths between green and red (665, 680, 560 and 620 nm) and showed satisfactory results, despite the limitations, due to the complexity of the study area and problems in atmospheric correction algorithm. Typical algorithms water Case I (OC3 and OC4) were also tested, as well as FLH MCI and algorithms suggested for water with high concentrations of Chl-a, all with unsatisfactory results. As noted by past studies, Guanabara Bay has high spatial and temporal variability of chlorophyll-a concentrations, with the highest concentrations in the rainy seasons (months: 01, 02, 03, 10, 11, 12) and in the marginal portions (~ 100 mg.m-3), particularly in the western edge of the bay, and lower concentrations in the dry season and in the main circulation channel (~ 20 mg.m-3). This study is a pioneer in the construction and application of bio-optical algorithms for the region of BG using MERIS images. Despite the good results, the algorithm should not be considered definitive, and it is recommended for future work to test different models of atmospheric correction for MERIS images.

Conselho Nacional de Desenvolvimento Científico e Tecnológico / Este trabalho teve como objetivo principal implementar um algoritmo empírico para o monitoramento do processo de eutrofização da Baía de Guanabara (BG), Rio de Janeiro (RJ), utilizando dados de clorofila-a coletados in situ e imagens de satélite coletadas pelo sensor MERIS, a bordo do satélite ENVISAT, da Agência Espacial Européia (ESA). Para a elaboração do algoritmo foi utilizada uma série histórica de clorofila-a (Out/2002 a Jan/2012) fornecida pelo Laboratório de Biologia Marinha da UFRJ, que, acoplada aos dados radiométricos coletados pelo sensor MERIS em datas concomitantes com as coletas in situ de clorofila-a, permitiu a determinação das curvas de regressão que deram origem aos algorítmos. Diversas combinações de bandas foram utilizadas, com ênfase nos comprimentos de onda do verde, vermelho e infra-vermelho próximo. O algoritmo escolhido (R = 0,66 e MRE = 77,5%) fez uso dos comprimentos de onda entre o verde e o vermelho (665, 680, 560 e 620 nm) e apresentou resultado satisfatório, apesar das limitações devido à complexidade da área de estudo e problemas no algoritmo de correção atmosférica . Algorítmos típicos de água do Caso I (OC3 e OC4) também foram testados, assim como os algoritmos FLH e MCI, aconselhados para águas com concentrações elevadas de Chl-a, todos com resultados insatisfatório. Como observado por estudos pretéritos, a Baia de Guanabara possui alta variabilidade espacial e temporal de concentrações de clorofila-a, com as maiores concentrações no período úmido (meses: 01, 02, 03, 10, 11 12) e nas porções marginais (~ 100 mg.m-3), particularmente na borda Oeste da baia, e menores concentrações no período seco e no canal principal de circulação (~ 20 mg.m-3). O presente trabalho é pioneiro na construção e aplicação de algoritmos bio-óptico para a região da BG utilizando imagens MERIS. Apesar dos bons resultados, o presente algorítmo não deve ser considerado definitivo, e recomenda-se para trabalhos futuros testar os diferentes modelos de correção atmosférico para as imagens MERIS. / This work aimed to implement an empirical algorithm for monitoring the process of eutrophication at Guanabara Bay (BG), Rio de Janeiro (RJ), using in situ chlorophyll-a data and satellite images by MERIS sensor, onboard ENVISAT satellite, from European Space Agency (ESA). A time series of chlorophyll-a (Dec / Jan 2002/2012) provided by Marine Biological Laboratory from UFRJ, was used to elaborate the algorithm, coupled with the radiometric data collected by MERIS sensor on concurrent dates with the collections, what allowed the determination of the regression curves that gave rise to algorithms. Several band combinations were used, with emphasis on wavelengths of green, red and near infrared. The algorithm chosen (R = 0.66 and SRM = 77.5%) made use of wavelengths between green and red (665, 680, 560 and 620 nm) and showed satisfactory results, despite the limitations, due to the complexity of the study area and problems in atmospheric correction algorithm. Typical algorithms water Case I (OC3 and OC4) were also tested, as well as FLH MCI and algorithms suggested for water with high concentrations of Chl-a, all with unsatisfactory results. As noted by past studies, Guanabara Bay has high spatial and temporal variability of chlorophyll-a concentrations, with the highest concentrations in the rainy seasons (months: 01, 02, 03, 10, 11, 12) and in the marginal portions (~ 100 mg.m-3), particularly in the western edge of the bay, and lower concentrations in the dry season and in the main circulation channel (~ 20 mg.m-3). This study is a pioneer in the construction and application of bio-optical algorithms for the region of BG using MERIS images. Despite the good results, the algorithm should not be considered definitive, and it is recommended for future work to test different models of atmospheric correction for MERIS images.

Total suspended matter (TSM) concentrations were derived from ocean colour imagery (MERIS data) in Minas Basin. Analysis of time series of TSM in 1-km2 pixel boxes throughout the Basin revealed an annual cycle in TSM in most parts of the Basin. Higher TSM of up to 85 g/m3 was observed in late-winter (February - March), and lower TSM of 5-10 g/m3 characterized late-summer (July - August). The largest annual variation occurred in the centre of Basin, and the smallest variation occurred in shallow areas. Satellite-derived TSM were compared to predictions using the Delft3D model. Increasing model erosion rate in winter relative to summer was necessary to improve agreement between model and satellite-derived TSM. In comparison with the satellite-derived estimates, the model underestimated TSM in shallow areas in summer and overestimated it in winter. This discrepancy is likely due to inaccurate satellite-derived TSM in shallow, high concentration areas of the Basin.

Proliferation of cyanobacteria is a growing problem in lacustrine environment that results in rapid degradation of water quality. Moreover, certain cyanobacteria species produce harmful toxins. Phycocyanin (PC) is a photosynthetic pigment typical of cyanobacteria and affects the water color: it is therefore possible to study them using remote sensing. At least three algorithms to estimate PC concentration ([PC]) have been published, but their relative errors are important, especially for lower concentration. In this study, we are presenting the results of a new algorithm that uses the second order variability (anomalies) of water's reflectance spectrum to estimate [PC]. This method has never been used in lacustrine environment. The dataset used to develop and validate the algorithm was obtained between 2001 and 2005 in 57 different lakes and reservoirs of the Netherlands and Spain. The performance of the second order algorithm is equivalent or better than the three previously published algorithms. For the subset were [PC] > 32 mg m[superscript -3], the contribution of the second order term (R[superscript 2] =0.68 and RMSE=0.25) seems to improve considerably the first order algorithm (R[superscript 2] =0.50 and RMSE=0.35). The accuracy of the second order algorithm for [PC] > 32 mg m[superscript -3] is superior to the one calculated for the whole dataset (R[superscript 2] =0.69 and RMSE=0.44). The algorithm can also be adapted to the bands of satellite sensor MERIS for the study of cyanobacteria. The application of this algorithm to a MERIS image acquired the 29 August 2010 taken over the Missisquoi Bay (Quebec, Canada) demonstrates the potential of this new algorithm for a future cyanobacteria' monitoring system. Note that all the statistical results presented above are for the logarithm of [PC] and the units of the RMSE are log(mg/m[superscript 3]).

The application of high temporal satellite image data for designation of the spectral characteristic of vegetation Abstract The objektive of this paper is to evaluate possibilities of high temporal satellite data assimilation for continuous monitoring of the spectral characteristic of vegetation. There is also given the suggestion of metodology for processing MERIS data and for continuous monitoring of spectral characteristic of landscape objects. Finally, vegetation cover database for the Czech Republic in the year 2009 is created from sectorial analysis. In the paper there is used the LSU classification and thresholding of vegetation indicies histograms. The universal decision algorithm for classification of vegetation landscape component are described and particular thresholding values for the year 2009 given. The finally product of this paper is Czech vegetation cover database for the year 2009 with overall accuracy of 63,35 %. Accuracy for forest is even over 80 %. Keywords: remote sensing, MERIS, vegetation, spectral reflectance, LSU, BEAM

Nous présentons un outil logiciel qui permet de simuler une mission spatiale dans son intégralité. Ceci constitue une approche nouvelle en télédétection. Les concepteurs de missions et les scientifiques peuvent ainsi mieux appréhender la complexité toujours croissante des futures missions. Ce simulateur permet, bien avant le lancement de la mission, de considérer l'ensemble dit système, qui comprend évidemment le capteur, mais aussi les caractéristiques de l'orbite de la plate-forme qui l'emportera, ainsi que les algorithmes de traitement des données qui seront effectués au sol. Ce simulateur apporte une aide précieuse dans le cadre du dimensionnement et de l'analyse d'une mission spatiale ; il permet de comprendre comment la performance de tout ou partie du système contribue à la performance globale de la mission. La mission spatiale simulée dans le cadre de cette thèse est la mission du capteur MERIS (Medium Resolution Imaging Spectrometer) de l'Agence Spatiale Européenne, dont l'objectif principal est la mesure de la couleur de l'eau. Nous utilisons le simulateur pour évaluer la dynamique du signal en entrée du capteur, pour calculer la résolution radiométrique nécessaire au capteur pour satisfaire les objectifs de la mission et pour étudier la sensibilité du signal à de petites variations en attitude de la plateforme. Nous étudions ensuite l'influence de la qualité de la modélisation de l'état de la surface océanique sur la qualité de la mesure de la couleur de l'eau. Nous montrons que les incertitudes, liées à une modélisation trop imprécise des effets de surface et à une méconnaissance de la vitesse du vent, induisent des erreurs sur le signal marin pouvant être supérieures à la précision attendue par la mission.

Earth observation satellites cover large areas with frequent temporal repetition and provide us with new insight into ocean and coastal processes. Ocean colour measurements from satellite remote sensing are linked to the bio-optics, which refers to the light interactions with living organisms and dissolved and suspended constituents in the aquatic environment. Human pressures have changed the aquatic ecosystems, by, for example, the increased input of nutrient and organic matter leading to eutrophication. This thesis aims to study and develop the link between bio-optical data and the remote sensing method to the monitoring and management of the Baltic Sea. The results are applied to the European Union’s Water Directives, and the Baltic Sea Action Plan from the Helsinki commission. In paper I indicators for eutrophication, chlorophyll-a concentration and Secchi depth were evaluated as a link to remote sensing observations. Chlorophyll-a measurements from an operational satellite service (paper I) were compared to conventional ship-based monitoring in paper II and showed high correlations to the in situ data. The results in paper I, II and IV show that the use of remote sensing can improve both the spatial and temporal monitoring of water quality. The number of observations increased when also using satellite data, thus facilitating the assessment of the ecological and environmental status within the European Union’s water directives. The spatial patterns make it possible to study the changes of e.g. algae blooms and terrestrial input on larger scales. Furthermore, the water quality products from satellites can offer a more holistic and easily accessible view of the information to decision makers and end-users. In paper III variable relationships between in situ bio-optical parameters, such as coloured dissolved organic matter (CDOM), dissolved organic carbon, salinity and Secchi depth, were found in different parts of the Baltic Sea. In paper IV an in situ empirical model to retrieve suspended particulate matter (SPM) from turbidity was developed and applied to remote sensing data. The use of Secchi depth as an indicator for eutrophication linked to the concentrations of chlorophyll-a and SPM and CDOM absorption was investigated in paper V. The variations in Secchi depth were affected differently by the mentioned parameters in the different regions. Therefore, one must also consider those when evaluating changes in Secchi depth and for setting target levels for water bodies. This thesis shows good examples on the benefits of incorporating bio-optical and remote sensing data to a higher extent within monitoring and management of the Baltic Sea. / <p>At the time of the doctoral defense, the following papers were unpublished and had a status as follows: Paper 4: Manuscript. Paper 5: Manuscript.</p>

This PhD study focusses on the use of MEdium Resolution Imaging Spectrometer (MERIS) data for reliable and quantitative water-quality assessment of optically-complex waters (lake, brackish and coastal waters). The thesis is divided into two parts: A. intercalibration of reflectance measurements in different optically-complex water bodies (Paper I), and validation of various satellite processing algorithms for the coastal zone (Paper II). B. Applications: the use of MERIS data in integrated coastal zone management mostly using Himmerfjärden bay as an example. Himmerfjärden bay is one of the most frequently monitored coastal areas in the world and it is also the recipient of a large urban sewage treatment plant, where a number of full-scale nutrient management experiments have been conducted to evaluate the ecological changes due to changes in nutrient schemes in the sewage plant. Paper I describes the development and assessment of a new hyperspectral handheld radiometer for in situ sampling and validation of remote sensing reflectance.  The instrument is assessed in comparison with readily available radiometers that are commonly used in validation. Paper II has a focus on the validation of level 2 reflectance and water products derived from MERIS data. It highlights the importance of calibration and validation activities, and the current accuracy and limitations of satellite products in the coastal zone.  Bio-optical in situ data is highlighted as one of the key components for assessing the reliability of current and future satellite missions. Besides suspended particulate matter (SPM), the standard MERIS products have shown to be insufficient to assure data quality retrieval for Baltic Sea waters. Alternative processors and methods such as those assessed and developed in this thesis therefore will have to be put in place in order to secure the success of future operational missions, such as Sentinel-3. The two presented manuscripts in the applied part B of the thesis (paper III and IV), showed examples on the combined use of in situ measurements with optical remote sensing to support water quality monitoring programs by using turbidity and suspended particulate matter as coastal indicators (manuscript III). The article also provides  a new turbidity algorithm for the Baltic Sea and a robust and cost-efficient method for research and management.  A novel approach to improve the quality of the satellite-derived products in the coastal zone was demonstrated in manuscript IV. The analysis included, the correction for adjacency effects from land and an improved pixel quality screening.  The thesis provides the first detailed spatio-temporal description of the evolution of phytoplankton blooms in Himmerfjärden bay  using quality-assured MERIS data, thus forwarding our understanding of ecological processes in in Swedish coastal waters. It must be noted that monitoring from space is not a trivial matter in these optically-complex waters dominated by the absorption of coloured dissolved organic matter (CDOM). These types of coastal waters are especially challenging for quantitative assessment from space due to their low reflectance.  Papers III and IV thus also provide tools for a more versatile use in other coastal waters that are not as optically-complex as the highly absorbing Baltic Sea waters. The benefits of the increased spatial-temporal data coverage by optical remote sensing were presented, and also compared to in situ sampling methods (using chlorophyll-a as indicator). / <p>Research funders:</p><p>European Space Agency (ESA, contract no.21524/08/I-OL)</p><p>NordForsk funding: Nord AquaRemS Ref. no. 80106</p><p>NordForsk funding:  NordBaltRemS Ref.no. 42041</p><p>At the time of the doctoral defense, the following papers were unpublished and had a status as follows: Paper 3: Manuscript. Paper 4: Manuscript.</p>

Résumé : Au-delà de tous débats scientifiques actuels, un constat unanime est certainement la réduction du couvert de glace dans l’océan Arctique, associé au réchauffement planétaire. La réduction du couvert de glace aura sans doute des impacts encore imprévisibles sur le milieu marin. Nous avons, dans ce contexte, traité des données satellitaires et des données de mesures de réalité de terrain de campagnes océanographiques dans la portion sud-est de la mer de Beaufort afin d’étudier les variabilités spatiale et temporelle de la biomasse phytoplanctonique et tenter de les relier aux processus physiques existants dans ce milieu. La mer de Beaufort étant fortement influencée par les eaux douces du fleuve Mackenzie, il était probable que les algorithmes de couleur de l’eau opérationnels actuels ne permettaient pas une estimation juste de la concentration de la chlorophylle-a (chl-a) et, par conséquent, de la production primaire qui est à la base de la chaîne alimentaire marine. L’analyse des données bio-optiques a confirmé cette hypothèse montrant une surestimation de la chl-a in situ par un facteur variant entre 3 et 5. La forte contribution de la matière organique colorée dissoute et des particules non-algales à l’absorption de la lumière apparaît comme la source principale de cette surestimation. Nous avons donc proposé des algorithmes adaptés ainsi que de nouveaux algorithmes utilisant deux rapports de bandes spectrales permettant une estimation plus précise de la chl-a dans le sud-est de la mer de Beaufort. Une comparaison entre des données de réalité de terrain et des images satellitaires a aussi montré que la réflectance normalisée à la surface de l’eau, de même que le rapport bleu-vert, étaient plus précis à l’aide des données du capteur SeaWiFS que de celles des capteurs MODIS et MERIS. Nous avons procédé à une analyse des patrons de chl-a et de température de surface pour cinq sous-régions géographiques dans la mer de Beaufort à l’aide de sept années de données satellitaires SeaWiFS et AVHRR (1998-2004). Les résultats ont montré que les variabilités spatiale, temporelle et interannuelle de la biomasse phytoplanctonique sont régies par plusieurs facteurs environnementaux affectant la stratification de la colonne d’eau, soit le forçage du vent, la dynamique de la glace, la température de l’air, l’ensoleillement et les courants marins. Une approche statistique basée sur le concept de provinces non statiques a permis de partitionner la mer de Beaufort en quatre provinces biophysiques distinctes, apportant un nouvel éclairage sur les propriétés biophysiques de cette mer. L'analyse des données a aussi permis de détecter une tendance à l'augmentation de la chl-a dans deux secteurs de la mer de Beaufort : le plateau du Mackenzie et la partie sud du golfe d'Amundsen. Finalement, une analyse de gradients spatiaux, effectuée à partir d’images de température de surface de l’eau a permis de détecter des fronts thermiques récurrents. Ces structures spatiales jouent un rôle majeur dans l’écosystème marin, en particulier en raison de leur impact sur le développement de la biomasse phytoplanctonique. Nous avons mis en évidence des nouvelles structures frontales sur le plateau du Mackenzie et dans la région de la polynie du cap Bathurst. Les nouveaux fronts détectés sont principalement reliés à des particularités bathymétriques de la région, à la présence du panache du fleuve Mackenzie ainsi qu’à la gyre de Beaufort. En conclusion, la réalisation de cette étude a permis de générer de nouvelles informations sur les interactions entre les processus physiques et biologiques, permettant ainsi de mieux appréhender les conséquences biogéochimiques et écologiques résultant des modifications climatiques dans la mer de Beaufort. // Abstract : The Arctic Ocean ecosystem is experiencing significant changes such as a drastic reduction in seasonal sea-ice cover linked to global warming. These changes are likely to modify the physics, biogeochemistry and ecology of this unique environment in ways that are yet to be understood. In this context, we processed satellite data and in situ measurements in the southeastern Beaufort Sea to explore the spatial and temporal variability of phytoplankton biomass and link it to existing physical processes in this region. The optical properties of the Beaufort Sea being under the influence of the Mackenzie River plume, it was likely that operational ocean color algorithms did not allow an accurate estimate of chlorophyll-a concentration (Chl-a) that is a key indicator of phytoplankton biomass and marine productivity. Analysis of bio-optical data confirmed this hypothesis showing an overestimation of Chl-a in situ by a factor of three to five. High contribution of colored dissolved organic matter and non algal particles to the blue light absorption appears as the source of that poor performance. We propose regionally adapted and new algorithms using ratio of two spectral bands allowing better accuracy estimation of Chl-a in the southeastern Beaufort Sea. A match-up analysis of coincident in situ data and satellite overpass showed that the normalized water-leaving reflectance and the blue-to-green ratio retrieval were more accurate for SeaWiFS data than for MODIS and MERIS data. We investigated temporal and spatial linkages between physical and biological parameters to infer the boundaries of biophysical areas in the Canadian Beaufort Sea. Monthly sea surface temperature (AVHRR) data and chlorophyll a data from SeaWiFS were collected over seven years in five geographical sub-regions in the Beaufort Sea (1998-2004). Results showed that the spatial, temporal and inter-annual variability of phytoplankton biomass are driven by several environmental factors affecting the stratification of the water column : wind forcing, ice dynamics, air temperature, irradiance and currents. A cluster analysis based on the concept of non-static provinces was used to define four biophysical provinces in this sea. Positive temporal trends were detected for Chl-a over two regions of the Beaufort Sea : the Mackenzie Shelf and the southern portion of Amundsen Gulf. Finally, an analysis of spatial gradients, using 11 years of sea surface temperature images, allowed the detection of recurrent thermal fronts. These spatial structures play a major role in the marine ecosystem, particularly because of their impact on the development of phytoplankton biomass. We highlighted new frontal structures on the Mackenzie Shelf and in the Cape Bathurst polynya area. These identified new fronts are mainly related to bathymetric features of the region, the presence of the Mackenzie River plume and the Beaufort Gyre. In conclusion, this study has generated new information on the interactions between physical and biological processes to better understand the biogeochemical and ecological consequences of climate change in the Beaufort Sea.