LES ZONES HUMIDES DE FONDS DE VALLEE ET LA REGULATION DES POLLUTIONS AZOTEES DIFFUSES

Depuis une vingtaine d'années, les zones humides de bas-fonds, véritables interfaces entre les bassins versants agricoles et les cours d'eau, sont connues pour être capables de réguler les pollutions azotées diffuses. Cependant, cette capacité est fonction de bon nombre de paramètres qui jouent sur l'efficacité réelle de ces zones tampons. De nombreuses tentatives d'aménagements se sont d'ailleurs soldées par des échecs en partie à raison de la méconnaissance de ces différents facteurs. En Bretagne, les teneurs en nitrate des eaux de surface et de subsurface font l'objet de nombreuses préoccupations (Potabilité, Eutrophisation…), et c'est dans ce contexte que les ripisylves peuvent être une aide précieuse pour la reconquête de la qualité de l'eau. Néanmoins, il reste encore beaucoup de questions quant à l'organisation des paramètres qui contrôlent la régulation de pollutions azotées au sein de ces milieux. Ainsi, cette thèse s'est attachée à comprendre la variabilité in situ des processus responsables de l'élimination des nitrates dans une ripisylve située au nord de l'Ille-&-Vilaine dans un bassin versant agricole, en bordure d'un affluent du Couesnon. A partir d'une approche spatiale et temporelle des principaux compartiments de l'azote (Eau, Sol, Végétation) sur 3 zones de la ripisylve possédant chacune une végétation différente, les travaux de cette thèse ont tenté d'analyser: (1) l'organisation de l'hydrologie du site, (2) la répartition spatio-temporelle de la dénitrification, (3) les relations végétation riveraine – dénitrification dans l'utilisation des nitrates, et (4) l'impact du couvert végétal sur les capacités tampons.La complexité de l'hydrologie de la plaine inondable a été démontrée grâce à l'identification des sources d'eau, des phénomènes de dilution et de l'impact de la géomorphologie de la ripisylve sur les écoulements. Ces paramètres hydrologiques sont variables dans le temps. La rétention des nitrates est cependant observée toute l'année dès l'entrée de la zone riveraine, et des taux de rétention ont été estimés. Toutes les études menées se rejoignent pour définir une zone active de rétention des pollutions azotées diffuses. Celle–ci est large de quelques mètres et est localisée juste après le talus de ceinture séparant la ripisylve et le bassin versant agricole. Sa position semble relativement stable à l'échelle d'une année. Cependant, son action s'inscrit dans un volume d'échange entre le sol et la nappe qui est fonction de l'engorgement. La dénitrification et l'assimilation végétale présentent une complémentarité d'action. Malgré son caractère de rétention temporaire, la végétation riveraine permet la fixation importante de l'azote et l'élongation de son temps de résidence. En effet, l'immobilisation et la minéralisation lente des composés azotés de la litière limitent leur lessivage et sont favorables à la dénitrification. L'utilisation de l'abondance naturelle des isotopes stables de l'azote a permis de mettre en évidence d'autres caractéristiques. La dénitrification est plus active en Été et permet donc une élimination des nitrates contenus dans l'eau de la nappe même lorsque celle-ci est basse. Ainsi, le sol peut être vu comme l'assemblage de 2 horizons: l'horizon supérieur contenant les racines et les horizons plus profonds. Lorsque la nappe submerge le site les 2 compartiments recyclent les nitrates par la végétation et la dénitrification, et lorsque la nappe est basse seule la dénitrification élimine les nitrates. L'important enrichissement en 15N de la végétation riveraine suggère une future utilisation de cette mesure pour détecter rapidement les zones où la dénitrification est réelle des zones où elle est seulement potentielle. Enfin, le type de couverture végétale ne semble pas avoir un impact significatif sur la rétention de l'azote dans les 3 zones testées. Le même contexte hydrogéomorphologique qui caractérise l'ensemble de la ripisylve impose son contrôle, et le type de végétation n'est plus influant à partir du moment où il assure une fourniture de carbone suffisante en quantité et qualité aux bactéries dénitrifiantes.Riparian zones are known to capture and transform nitrogen compounds from shallow through-flowing groundwater. In this way riparian ecosystems can function as buffer zones to diminish diffuse pollution of surface waters due to agricultural activities. Plant uptake anddenitrification can be considered the most important processes responsible for N retention. In many “natural” riparian zones, however, nutrients taken up by plants remain in the system only temporarily and may be released by mineralisation later. The rate of nutrient recycling inecosystems is determined by the longevity of plant parts, by the refractibility of litter components and by the suitability of environmental conditions for decomposition. Consequently, plants do not remove nutrients from the ecosystem but increase the residencetime of nutrients by a reduction in the mobility of N compounds. However, high nutrient inputs into riparian buffer zones are expected to lead to an increased rate of nutrient cycling because of a more rapid plant growth, shorter longevity of plant parts, higher nutrient contents(decreased retranslocation) and lower contents of secondary metabolites in litter. In this way, the increased nutrient availability would be expected to lead to a shorter retention time of nutrients in the riparian buffer zone as a whole. The aim of this study was to quantify the nutrient uptake in the vegetation, the nutrient losses from plants through leaching and litter production, the decomposition and N mineralisation / immobilisation in the three riparian zones with varying vegetation cover in order to evaluate the importance of vegetation in the effective retention of nutrients in buffer zones.Considering all sites together, the vegetation growth stores a significant amount of nitrogen. In the whole riparian strip, we found that above-ground biomass production was as high as 9800 kg dry mass ha-1 year-1 or 150 kg N ha-1 year-1. Thus, diffuse pollution of surfacewaters can be efficiently limited by vegetation uptake during the growing season; however the groundwater level does have to be in contact with the roots. When the above-ground biomass is about to die, a portion of the nitrogen contained in the green tissues is retranslocated inother plants parts for storage. In this study we found 44 kg N ha-1 year-1 of retranslocation as an average for the whole riparian area. Following this, the litter falls to the ground and the decay starts. First, there is a fast and short loss of nitrogen via leaching (22 kg N ha-1 year-1),then the immobilisation process requires nutrients from the surrounding environment (37 kg N ha-1 year-1) to balance the still high C:N ratio of the fresh litter compared to that of the decomposer community. This process can reduce the amount of available nitrogen and phosphorus in the soil environment and contribute to the short-term retention of nutrients in riparian zones. As litter production predominantly occurs in autumn, the retention through immobilization mainly occurs over the winter period when plant uptake is low and the risk of nutrient losses is high due to excessive rainfall. Hence, immobilisation may be an important process in the N retention in buffer zones. Finally, the litter is slowly mineralised and N compounds return to soil as inorganic nitrogen which is therefore available to denitrification, vegetation uptake or leaching.The difference between the absolute amount of N taken up and the amount of N retranslocated and decayed can be attributed either to a long-term storage system like wood production or to root exudates and death. In fact, the results of this study still have to bebalanced with the N losses due to the root turn-over which is certainly an important component of the riparian nitrogen cycle (Fig. 16). Yet, from the above mentioned experiments, it seems that the root litter can not be important in the nitrogen retention via the immobilisation process, since the mineralisation is instantaneous after the beginning of the litter formation (Figure 13). Given the young age of its tree canopy, primary production at the Forest site is as important as in the Meadow; this will change once the canopy is mature with lower needs in term of N assimilation. Furthermore, the shade of the tree leaves will limit the growth of herbaceous species too. One positive aspect of the tree canopy is that the N retention time is longer than in the other study sites. Indeed, only 4.5 % of the nitrogen content of the Forest litter is released in the first year of decomposition, while the Shrub and the Meadow release 13.7 and 16.5 % respectively (Table 2). This is mainly due to the low decomposition rates. In fact, the decomposition rates of the herbaceous-dominated riparian sites (Shrub and Meadow) are significantly higher than the Forested one (p = 0.01). This is probably caused by the presence of ligneous leaf veins in tree leaves and other differences in litter quality (e.g. lignin content). Initial litter N content in forest derived litter was higher than herbaceous litter, so that this aspect of litter quality cannot explain the differences in decomposition rate (Figure 8). The Shrub vegetation differs from the other study sites in the amount of nitrogen which is taken up, since the primary production results mainly from the vegetation located in Zone 3 (Upland side). However, the amount of immobilised nitrogen in zone 3 of the Shrub site is also important and must significantly restrict the loss of nitrogen via leaching in the winter period. The high N uptake in the Shrub site (25 g N m-2 year-1) is comparable to that found in natural Phragmites marshes. The other sites have values of 6-10 g N m-2 year-1, which is within the range characteristic for eutrophic wetlands. The maximum value in constructed, extremely high loaded Phragmites wetlands is about 120 g N m-2 year-1. The Meadow vegetation has the highest capacity to delay nitrogen release to soil. Indeed, the retranslocation capacity and the immobilisation rate are important and 68 % of the yearly nitrogen uptake is momentarily "blocked" via these processes (Table 2). Yet, the mineralisation rate compared to the total amount of nitrogen in the litter allows a fast liberation of inorganic nitrogen. We did not find any obvious relation between N loading and plant uptake. This does not imply that there was no variation in biomass production or plant N concentrations between sites (Tables 2 and 3). The absence of a relation of plant uptake with N loading may be explained by growth limitation by other macronutrients such as P, K, by light or by flooding. Another explanation may be that N is very efficiently removed by soil processes as denitrification, so that actual availability of N for plants is strongly reduced. Another possibility may be that the high C:N ratios of the fresh litter (from 33.9 to 34.7) make the decomposers (lower C:N ratio, about 5:1) dependant on the inorganic nitrogen which is the immobilised and not available for plant assimilation (Killham, 1994; Hodge et al., 2000). The hydromorphic gradient was mainly obscured by the differences in species composition from zone 3 to zone 1. However, we grouped the three study sites and discussed the few characteristics that come out from this averaging (Table 3). The upper zone, i.e. zone 3, is remarkable in the amount of nitrogen which is cycling through the vegetation; the immobilisation rate is significantly higher than in the other zones. The intermediate zone, i.e. zone 2, is characterized by the percentage of the nitrogen uptake which is retranslocated. This zone has low leaching and mineralisation rates, though the nitrogen turn-over is relatively fast since the amount of mineralised nitrogen is almost equivalent to the N content of the litter. Finally, the lower zone, i.e. zone 1 closest to the stream and more often submerged, is characterized by vegetation which grows mainly by using autochthonous nitrate