Nous avons étudié la variabilité saisonnière de la relation entre les fluctuations des concentrations en carbone organique dissous (COD) dans le ruisseau de l'Hermine (Québec, Canada) et les changements du débit (Q). Un total de 93 événements hydrologiques échantillonnés de 1994 à 2003 et regroupés sur une base saisonnière (hiver-printemps, été, automne) a été analysé. Le modèle de régression linéaire est utilisé afin de déterminer, pour chaque événement, la pente de la relation entre la concentration en COD dans le ruisseau et le débit. Ces pentes sont regroupées par saison et selon un seuil arbitraire de un qui permet de contraster les conditions hydrologiques et climatiques initiales des événements répertoriés. Les résultats du test de Kruskal-Wallis, visant la comparaison entre les événements de pentes supérieures et inférieures à un, montrent clairement la saisonnalité de la relation entre le COD et le débit. La saisonnalité de la relation COD/Q est ensuite mise en relation avec des variables climatiques et hydrologiques susceptibles de conditionner le transport du COD dans le bassin de l'Hermine. Les résultats montrent que les changements saisonniers des conditions climatiques et hydrologiques dans le bassin versant ont un impact significatif sur la relation entre le COD et le débit. Ainsi, le volume de précipitation tombé durant l'événement, la température moyenne de l'air et la température du sol régissent significativement (p =0,041; 0,001 et 0,009 respectivement) le transport du COD pour la période hiver-printemps. Les basses températures du sol et l'apport élevé en eau via les précipitations et la fonte favorisent le lessivage intense du COD soluble déjà limité par les basses températures. Au cours de l'été, l'état initial d'humidité du bassin est le principal facteur contrôlant l'évolution des concentrations de COD lors d'une crue; les fortes relations avec le pourcentage d'humidité des sols et le débit total 24 h avant l'événement le prouvent (p =0,039 et 0,0003 respectivement). Les changements les plus prononcés du COD surviennent, au cours de l'été, suite à une période prolongée de sécheresse. À l'automne, le transport du COD est influencé par le volume de précipitation tombé durant l'événement (p =0,031) et la température du sol (p =0,042). La modélisation de la relation COD/Q par les variables hydro-climatiques montre que 40% de la relation COD/Q s'explique par la température du sol durant la période d'hiver-printemps. Durant l'été, les conditions initiales d'humidité du bassin, traduites par le débit 24 h avant l'événement, expliquent à 51% la relation COD/Q. À l'automne, la relation COD/Q est gouvernée à 50% à la fois par le volume de précipitation tombé durant l'événement et la température du sol. L'analyse de ces données établit clairement la saisonnalité de la relation COD/Q et que des variables climatiques et hydrologiques permettent de quantifier ces fluctuations saisonnières., The terrestrial organic carbon (C) pool, estimated to 1.5 x 1015 kg C for the first meter of soil (Amundson, 2001), represents a major terrestrial elemental stock for which the recycling rate and the response to perturbations are still unknown. Under the present changing climatic conditions, C fluxes in terrestrial ecosystems could be significantly disturbed during the next decades. Indeed, the multi-annual changes in temperature and precipitation are likely to have a major impact on the net primary production and on organic matter decomposition in soils. This situation influences the production of the dissolved organic carbon (DOC) in soils, its transport to surface waters and hence, water quality. In this context, a better knowledge of the climatic and hydrologic factors influencing seasonal variations in DOC export is crucial to improve our understanding of the potential transformation of carbon stocks and fluxes in terrestrial ecosystems.The objectives of the present study were 1) to evaluate the seasonality in the relationship between dissolved organic carbon (DOC) concentrations in the stream and streamflow (Q) and 2) to quantify the impact of seasonal changes in climatic and hydrological conditions in the watershed on the DOC/Q relationship.The Hermine catchment is located about 80 km north of Montréal, Québec, Canada. An intermittent first-order stream drains the 5.1 ha catchment. Soils are Orthic and Gleyed Humo-Ferric and Ferro-Humic Podzols. The stream water was sampled daily, from 1994 to 2003, with an automatic sampler. The stream discharge was calculated from the water level above a 90º V-notch weir using a Global level sensor bubbler. Soil organic C content was analysed by the modified Walkley-Black method. Because of the high cost of DOC analysis for numerous samples, the DOC content was estimated by the relationship obtained between eight stream water samples analysed with a Shimadzu TOC analyser (Shimadzu, Kyoto, Japan) and the corresponding absorbance measured at 254 nm. From the initial year of the project, 1994, the regression used was Y=-0.05 + 32.60 X with an r2 value of 0.58 and a precision of 0.05 mg·L-1.The relationship between the DOC concentration and Q at the Hermine was positive and significant (p < 0.01) when all data were considered (n=1960). Because of the weakness of this relation (r2 =0.12), the stream samples, from 1994 to 2003, were seasonally split into 93 distinct hydrological events: 33 for winter-spring, 34 for summer and 26 for fall. A linear regression model was used to determine, for each event, the slope of the relationship between the DOC concentrations in the stream and Q. To contrast the antecedent conditions of the Hermine watershed, the events from a given season were divided into two groups. The Kruskal-Wallis test was then used to establish the link between the slope of the DOC/Q relationships on the one hand, and the environmental watershed conditions on the other hand: the climatic variables (volume of precipitation during event, mean air and soil temperatures) and the hydrological variables (stream discharge 24 h before the event, soil moisture, and ground water level).The DOC concentrations in the stream varied on an annual, a seasonal and an event basis. For the period 1994 to 2003, the annual mean concentrations, calculated from daily samples, varied from 2.0 to 2.5 mg DOC·L-1. On a seasonal basis, mean daily DOC was higher during the summer and the fall (2.9 and 2.8 mg DOC·L-1 respectively), and lower in the winter-spring (2.1 mg DOC·L-1). The relation between DOC concentrations and Q fluctuated as a function of the seasonal evolution of climatic and hydrological conditions in the Hermine catchment. For winter-spring events, 79% of the events had a DOC/Q slope lower than one. This period was characterised by high streamflow levels and high total DOC fluxes even though the daily mean DOC concentrations were low (2.1 mg DOC·L-1). The volume of precipitation during the event (p =0.041), the mean air temperature (p =0.001) and the soil temperature (p =0.009) were significantly related to the difference between events with slopes lower and higher than one. Indeed the slope of the relation increases when soil temperatures are elevated. When the temperatures are higher, DOC export increases and subsurface flow in soil horizon is enriched in DOC. Under colder temperature, the DOC production is limited and the soluble organic substances stored in soils are leached out the catchment with the high volume of precipitation and with the water coming from the snowmelt. For the summer period, there were 20 events with slopes greater than one against 14 with slopes lower than one. The soil humidity (p =0.039) and the total streamflow 24 h before the event (p=0.0003), were the two variables that significantly distinguished both slope groups. Rapid changes in DOC concentration occur during hydrological events following a long drought period. Under dry conditions, the subsurface flow in soil horizons rich in organic matter, the re-hydration of bed sediments and the hydrophobic behaviour of soil particles can all contribute to the export of very high DOC concentrations, even during small events. The relationships between DOC and Q, for the fall season, were significantly influenced by the volume of precipitation during the event (p =0.031) and the mean soil temperature (p =0.042). The events with the lower slopes showed the highest volume of precipitation during event and the lowest soil temperature. For these events occurring under wet conditions, the water originates essentially from the B and C horizons, and DOC fluctuations are then limited because of the low concentrations of the DOC in these horizons (anionic sorption of soluble organic substances by iron oxides).Best-fit from multiple regressions indicated that 40% of the link between DOC and Q was explained by the soil temperature during the winter-spring period (p =0.0001). For summer, the streamflow 24 h before events accounted for 51% of the variation in DOC/Q relationships (p =0.00001). For the fall period, the volume of precipitation during event and the soil temperature both contributed equally to the DOC/Q relationships (p =0.001). From these results, obtained from a multi-year project, it is clear that the relation between DOC and Q is a function of the variability in the climatic and hydrological watershed conditions. In a context of global warming, it is possible that warmer air temperatures have an effect on soil temperature. Thus, during winter-spring and fall periods, the duration and the intensity of the DOC production in soils will increase and the export of DOC from the watershed to other surface water system could become more important under equivalent or higher streamflow. Higher air temperature also means higher evapotranspiration by the forest during the summer period, and consequently dryer watershed conditions. A low streamflow and a low soil humidity level could be expected and then, brief rain events will sporadically flush the soluble organic carbon accumulated in the soil. The DOC export would be insignificant for that period, but the DOC would reach the highest annual level. The new knowledge on the DOC/Q relationships, at the hydrological event scale, will be added to the accumulated data on the possible effects of global warming on the carbon cycle in forested ecosystems.