Catálogo de publicaciones - libros

Ce chapitre a pour but de faire le point sur l’état des connaissances et d’identifier les lacunes relatives à la problématique de l’émission de gaz à effet de serre (GES) par les réservoirs hydroélectriques et les écosystèmes naturels. Il est devenu essentiel d’intégrer nos connaissances du cycle du carbone à des échelles temporelles et spatiales plus vastes de façon à mieux définir l’ampleur des flux de GES associés aux réservoirs1 et aux écosystèmes naturels. Les données disponibles proviennent d’études à petite échelle et de courte durée (1 à 10 ans), effectuées surtout en région boréale, mais aussi en régions semi-aride et tropicale. La variabilité naturelle des flux de GES due à des variations climatiques régionales et leurs impacts sur la production biologique globale est plus importante que celle des méthodes de mesures. Il faut donc garder à l’esprit que les incertitudes concernant les flux de GES sont avant tout le résultat de variations spatiales et temporelles naturelles des flux, et non pas des techniques de mesure disponibles. La présente synthèse se base sur les résultats de plus de dix ans de suivis obtenus par différentes équipes de recherche de plusieurs universités, institutions gouvernementales et compagnies d’électricité.

In this chapter, we have addressed several issues regarding the use of network-based mathematical programming techniques for solving various problems arising in the broad area of data mining. We have pointed out that applying these approaches often proved to be effective in many applications, including biomedicine, finance, telecommunications, etc. In particular, if a real-world massive dataset can be appropriately represented as a network structure, its analysis using standard graph-theoretical techniques often yields important practical results.

However, one should clearly understand that the success or failure of applying a certain methodology essentially depends on the structure of the considered dataset, and there is no “universal recipe” that would allow one to obtain useful information from any type of data. This indicates that despite the availability of a great variety of data mining techniques and software packages, choosing an appropriate method of the analysis of a certain dataset is a non-trivial task.

Moreover, as technological progress continues, new types of datasets may emerge in different practical fields, which would lead to further research in the field of data mining algorithms. Therefore, developing and modifying mathematical programming approaches in data mining is an exciting and challenging research area for years to come.

Hydro-Québec and its partners have been measuring greenhouse gases (GHG) gross fluxes from hydroelectric reservoirs and natural water bodies since 1993. Over the years the methods have changed with a constant aim for improvement. The methods used are: the thin boundary layer, the use of floating chambers with or laboratory analysis and the use of floating chambers coupled to an automated instrument (NDIR or FTIR). All these methods have their pros and their cons. Over the years many tests were done to compare the methods.

There is no significant difference in the results obtained with the or laboratory analysis. For CO fluxes, the number of results rejected is similar for the NDIR and the laboratory analysis methods. For CH fluxes, the number of results rejected is three times lower with the floating chamber with laboratory analysis than with the other methods. The precision for duplicate measurements of fluxes is similar for all methods with floating chambers. In general, the thin boundary layer method tends to measure lower fluxes than the floating chamber method with laboratory analysis and there is no good correlation between the two methods. Fluxes measured with automated instruments (specially for fluxes > 5000 mg·m·d) tend to be higher compared with the laboratory analysis method but the correlation between the two methods is very good (R=0.92 for CO). The method with the less logistical constraints is the floating chamber coupled to an automated instrument. This method enables the sampling of about 5 times more sites in the same amount of time as the method with laboratory analysis. The floating chamber coupled to an automated instrument has therefore been retained as the method of choice by Hydro-Québec for GHG gross flux measurements over water bodies.

Diode laser second derivate modulation spectroscopy combined with an open atmospheric path is a well suited technique for trace gas monitoring above wide areas. This paper presents the development of a portable long optical path near infrared spectrometer based on telecommunication laser diodes in order to provide a powerful tool for real time simultaneous measurements of greenhouse gas (GHG) concentrations above lakes and hydro reservoirs. CO and CH are respectively monitored at 1572 and 1653 nm along optical paths of several hundreds of meters above the area of interest. Simultaneous measurements of the target gases at two different heights above the water surface allows to detect concentration gradients on a continuous basis. A simplified turbulent diffusion model involving both the measured concentration gradients and local wind data has been used to estimate average GHG fluxes emitted by lakes and hydro reservoirs in different regions. Recent optimizations of the developed prototype allow quick on site set-up and operation of the laser device, even in remote areas without local facilities, as well as the continuous measurement of low GHG concentration gradients during long time periods with a minimum of local surveillance. Further improvements of this laser system would allow simultaneous detection of other trace gases such as NO emitted by agricultural soils and other gases present in various environments.

Current estimates of GHG budgets of forests indicate that these ecosystems act, over all, as C sinks, regardless of climatic region. The sequestering of GHGs by forests is the result of the balance between a substantial uptake of CO by vegetation and CO release through soil respiration, a weak consumption of CH by methanotrophic bacteria, and a more or less significant emission of NO from soil, a by-product of nitrification and denitrification reactions. According to the estimated GHG budgets, the forest C sink is weaker in the boreal region (mean −236 mgC·m·d) than in the temperate and tropical regions (−398 and −632 mgC·m·d, respectively), in agreement with the NEE (Net Ecosystem Exchange in CO) pattern observed with latitude. The increase in solar radiation and growing season length from north to south could be responsible for such a pattern. The C sinks typical of tropical forests can be offset by nearly 30% by NO emissions from the soils. In northern forests, some sites have been found to be net sources of CO, particularly during warm and dry years.

In wetlands, water-saturated soils are conducive to anaerobic decomposition of organic matter and methane production. CH largely dominates the GHG budget of wetlands. Overall, northern peatlands are sources of GHGs (mostly as CH) (173 gCO-eq.·m·yr), while accumulating small quantities of CO as peat (−22 gC·m·yr). Our GHG budgets for peatlands do not generally take into account the high CO and CH emissions from ponds. Tropical wetlands, dominated by marshes and swamps, emit large amounts of CH into the atmosphere (mean 71 mgCH·m·d) compared to northern peatlands (34 mgCH·m·d).

The estimated GHG budget values for forests depends on a) the error in NEE measurements, which dominate the forest GHG budgets and whose percent correction for calm nights can represent up to 50% of their initial value; b) the small number of published data on NEE for boreal and, especially, tropical forests, for which two out of the three available NEE data sets are contested; and c) the lack of NEE data in certain stands, such as in boreal forests post-fire stands, to which a certain potential for GHG emission is attributed. In wetlands, the popular use of the chamber method for CH flux measurements, along with a sampling period of less than a year, make it difficult to estimate an annual GHG budget that would integrate these systems' considerable spatial and temporal variability. The use of flux towers in forests, and recently in peatlands, allows for estimations of annual CO fluxes (NEE) typical of the local variability of environmental conditions within these systems.

Sediments were cored from 19 lakes and 4 reservoirs and were analyzed for CO and CH. Additionally sediment concentration profiles were measured in 107 cores. Concentrations of methane in the surface (2 ± 1.7 cm, n=107) sediment porewater in 4 oligotrophic lakes was in the order of 0.04 ± 0.03 mM (0–0.14, n=25), 0.24 mM in one oligotrophic reservoir to high values of 1.4 ± 0.96 mM (0–2.8, n=18) in 5 eutrophic-hypereutrophic lakes and 1.4 ± 1.0 mM (0–3.5, n=24) in 2 eutrophic reservoirs. Concentrations of porewater CO in oligotrophic lakes were ten times higher than methane concentration with values in the order of 0.4 ± 0.36 mM but slightly lower for the same eutrophic lakes with values around 1.35 ± 0.9 mM and reservoirs [1.0 ± 0.8 (0.1–3.2, n=23)]. Surface sediment CH concentrations were low in 3 acidotrophic lakes [0.2 ± 0.3 (0–1, n=22)] and very high in geothermal lakes [2.4 ± 0.6 (1.8-3, n=3)]. The CO concentrations in acidotrophic lakes were, however, high [0.9 ± 0.7 (0.1–2, n=14)]. Diffuse flux of the two greenhouse gases - CH and CO — from the surficial sediments across the sediment-water interface (SWI) were calculated from Fick's first law of diffusion. These resulted in low methane fluxes from oligotrophic systems (lakes and reservoirs) of 0.2–0.4 mM CH·m·d (3–6 mg CH·m·d) to much higher fluxes from eutrophic systems [3.9 mM CH·m·d (62 mg CH·m·d) for lakes and 5.2 mM (83 mg) for reservoirs] to very high fluxes at the two geothermal lakes [14 mM CH·m·d (220 mg CH·m·d)]. Dissolved CO fluxes were double the methane fluxes in oligotrophic lakes, about equal for eutrophic lakes [3.8 mM CO·m·d (167 mg CO·m·d)] and slightly lower than methane fluxes in eutrophic reservoirs [4.3 mM CO·m·d (190 mg CO·m·d)]. Diffuse fluxes of CO in acidotrophic systems [3.3 mM CO·m·d (143 mg CO·m·d)] were almost the same as observed in eutrophic lakes. Even though it is unclear why there are such great differences between temperate and tropical ecosystems, CO gas fluxes at the SWI in one tropical reservoir [Lobo Broa at 16 mM CO·m·d (700 mg CO·m·d)] were much higher than temperate ecosystems while CH diffuse fluxes [9 mM CH·m·d (140 mg CH·m·d)] are only slightly higher than temperate ones. There are few data to evaluate the importance of sediment diffuse fluxes of these two greenhouse gases as related to aquatic surface emissions. One study in an 11-m deep tropical high elevation reservoir observed that surface losses represented 10% of the sediment diffuse flux of CH (thus, 90% was oxidized at the SWI or in the water column). For CO it is suspected that 20% of the surface emissions come from sediment sources. The sediments represent an important repository of carbon which contributes gases to overlying waters. These fuel the activities of microorganisms and substantially contribute to oxygen depletion in overlying waters as well as contributing to climate change.

Available estimates of the soil organic carbon (SOC) density in northern and tropical forests vary between 8.5 and 13.9 kgC·m for the top first meter. Values of SOC for boreal forests are higher when considering the carbon stored as peat and in the forest floor. Overall, current SOC estimates often underestimate the total soil carbon content of boreal and tropical forests, because in many cases sampling is often limited to the soil's first meter.

The estimates of organic carbon sequestered in the vegetation of Amazonian forests (15.2 to 23.3 kgC·m) are two to five times higher compared to boreal (4.0 to 6.4 kgC·m) and temperate forests (4.8 to 5.7 kgC·m). Apart from their greater productivity, the more stable natural conditions prevailing in tropical rainforests have contributed to the accumulation of large amounts of carbon as biomass. On the other hand, the frequent recurrence of forest fires and insect outbursts in northern forests greatly limit carbon storage in the biomass.

The distribution of the total organic carbon stock between soil and vegetation varies with latitude. In northern forests, 72% of the organic carbon is found in the soil, with the remainder (28%) in the plant biomass. In tropical forests, the distribution is reversed, with 38% of the organic carbon stored in the soil and 62% in the vegetation. This difference can be explained by slower decomposition rates and a shorter growing season in relatively cold and humid boreal regions.

The boreal peatland carbon pool (98 to 335 PgC) is comparable to that of the whole boreal forest (180 to 330 PgC), but its surface area is five times smaller with an organic carbon content two to 10 times greater (39 to 134 kgC·m).

Export rates of organic carbon from terrestrial to aquatic ecosystems are small compared to their total stock but could be significant in the global forest carbon balance.

Carbon dioxide and methane emissions from estuaries are reviewed in relation with biogeochemical processes and carbon cycling. In estuaries, carbon dioxide and methane emissions show a large spatial and temporal variability, which results from a complex interaction of river carbon inputs, sedimentation and resuspension processes, microbial processes in waters and sediments, tidal exchanges with marshes and flats and gas exchange with the atmosphere. The net mineralization of land- and marsh-derived organic carbon leads to high CO atmospheric emissions (10–1000 mmol·m·d. 44–44 000 mg·m·d) from inner estuarine waters and tidal flats and marsh sediments. Estuarine plumes at sea are sites of intense primary production and show large seasonal variations of pCO from undersaturation to oversaturation; on an annual basis, some plumes behave as net sinks of atmospheric CO and some others as net sources; CO atmospheric fluxes in plumes are usually one order of magnitude lower than in inner estuaries. Methane emissions to the atmosphere are moderate in estuaries (0.02–0.5 mmol·m·d. 0.32–8 mg·m·d), except in vegetated tidal flats and marshes, particularly those at freshwater sites, where sediments may be CH-saturated. CH emissions from subtidal estuarine waters are the result of lateral inputs from river and marshes followed by physical ventilation, rather than intense in-situ production in the sediments, where oxic and suboxic conditions dominate. Microbial oxidation significantly reduces the CH emissions at low salinity (<10) only.

Carbon dioxide (CO), methane (CH) and nitrous oxide (NO) gross fluxes were measured at the air-water interface of 205 aquatic ecosystems in the Canadian boreal region from 1993 to 2003. Fluxes were obtained with a floating chamber connected to an automated NDIR or a FTIR instrument. The results show a temporary increase in CO and CH fluxes, followed by a gradual return to values comparable to those observed in natural aquatic ecosystems (lakes, rivers and estuaries). Mean values for CO and CH measured in Québec's reservoirs older than 10 years were 1508±1771 mg CO·m·d and 8.8±12 mg CH·m·d. Our results showed a strong similarity between lakes, rivers, and old reservoirs across a 5000 km transect from the west coast to the east cost of Canada. These values are comparable to those observed in Finland or in the sub-tropical semi-arid western USA. Although several limnological parameters can influence these fluxes, none showed a statistical relationship. However, levels of CO or CH fluxes are influenced by pH, wind speed, depth at sampling stations and latitude.

Carbon dioxide (CO) gross fluxes were measured at the air-water interface of 57 aquatic ecosystems in the western semi-arid region of the USA in April 2003. Fluxes were obtained with a floating chamber connected to an automated NDIR instrument. The results showed a strong similarity between lakes and reservoirs, as is also the case in boreal regions. The values ranged from −1500 to 10800 mg CO·m·d for both natural systems and reservoirs. These values are similar to those observed in boreal regions. Although several limnological parameters can influence the fluxes of CO, only the pH was significantly related with the CO gross fluxes, which decrease with increasing pH.