Resumo: Considerando a importância da compreensão da dinâmica espaço temporal dos componentes do balanço de energia (BE) em escala regional para o gerenciamento de recursos hídrico e o manejo agrícola, o objetivo principal desta tese foi construir e analisar uma série temporal dos componentes do BE adequada às condições de clima subtropical úmido do Estado do Rio Grande do Sul. Para tanto, inicialmente foi avaliada a adequação de modelos de estimativa de BE para o Estado. Nesta etapa foram utilizados produtos MODIS e dados de referência medidos em uma torre micrometeorológica instalada em Cruz Alta – RS, usando valores instantâneos para um período de estudo de 2009 a 2011. Na sequência foi avaliada a adequação dos modelos em representar a variabilidade espacial dos componentes do BE. Nesta etapa foram usados produtos MODIS, dados de reanálise ERA Interim, dados de referência da torre micrometeorológica e dados de estações meteorológicas do INMET, para o mesmo período de estudo. Na última etapa do trabalho foi construída a série temporal dos componentes do BE usando o modelo METRIC, a qual abrangeu um período de 14 anos, de 2002 a 2016. Os resultados demonstraram que os três modelos analisados apresentam coerência com as medidas de referência, sendo as maiores limitações apresentadas pelo modelo SEBAL, as quais se atribui principalmente às condições ecoclimáticas do Estado e a baixa resolução espacial das imagens. Na análise da variabilidade espacial, o modelo METRIC apresentou maior consistência nos resultados e proporcionou maior número de dias com resultados válidos, sendo assim apontado como o mais apto para realização do restante do estudo. A série temporal construída possibilitou a compreensão dos padrões de distribuição espaço temporal dos componentes do BE no estado do Rio Grande do Sul. Há uma marcada sazonalidade nos componentes do BE, com maiores valores no verão e menores no inverno. G (fluxo de calor no solo) é o componente de menor magnitude e sua distribuição espacial e temporal é determinada pela distribuição de Rn (saldo de radiação). Já os componentes LE (fluxo de calor latente) e H (fluxo de calor sensível), são os que mostram magnitude maior e apresentam padrões de distribuição espacial e temporal coerentes com as condições climáticas e com os tipos de uso e cobertura na área de estudo. Observase um padrão inverso, com um gradiente de LE no sentido noroeste para sudeste e para o componente H, no sentido sudeste para noroeste. Sendo estas informações de grande importância para gerenciamento de recursos hídricos em escala regional, para estudos de zoneamento agrícola. / Abstract: Given the importance of understanding the temporal and spatial dynamics of of the energy balance (EB) components in a regional scale for the management of water resources and agricultural, the main objective of this thesis was to construct and analyze a time series of the components of BE appropriate to the subtropical humid climate conditions of the State of Rio Grande do Sul. In order to reach the objective initially, the adequacy of the models for the humid climate conditions was evaluated, in this step we used MODIS data and reference data measured in a micrometeorological tower installed in Cruz Alta - RS. The analyzes performed with instantaneous values and the study period was from 2009 to 2011. The next step evaluate the spatial variability of the BE components, the data used were the MODIS products, ERA Interim reanalysis data, reference data of the micrometeorological tower and INMET meteorological stations, for the same study period. In the last stage the time series of the BE components was constructed from the METRIC model. The period series was 14 years from 2002 to 2016.The results showed that the three models analyzed were consistent with the reference measurements, with the greatest limitations presented by the SEBAL model, which are mainly attributed to the state's eco-climatic conditions and the low spatial resolution of the images In the analysis of the spatial variability, the METRIC model presented greater consistency in the results and provided greater number of days with valid results, this model thus indicated as the most suitable for the rest of the study. The time series constructed allowed us to understand the temporal distribution patterns of BE components in the state of Rio Grande do Sul. There is a marked seasonality in the BE components, with higher values in summer and lower in winter. G is the smallest magnitude component and its spatial and temporal distribution is determined by the Rn distribution. On the other hand, the LE and H components are those that show higher magnitude and present spatial and temporal distribution patterns consistent with the climatic conditions and the types of use and coverage in the study area. An inverse pattern is observed, with a LE gradient from north-west to south-east and for H-component, from southeast to northwest.

The eddy covariance (EC) technique was used to measure half hourly fluxes of energy and evaporation from 15 December 2007 to 30 November 2008 at the Scott Research Farm, located 7 km east of Hamilton. Many other supporting measurements of climate and soil variables were also made. The research addressed three objectives: 1. To examine the accuracy of the eddy covariance measurement technique. 2. Understand the surface partitioning of energy and water vapour on a diurnal to annual timescale. 3. Compare measurements of evaporation to methods of estimation. Average energy balance closure at Scott Farm was deficient by 24%, comparable to published studies of up to 30%. Three lysimeter studies were carried out to help verify eddy covariance data. These resulted in the conclusions that; 1) lysimeter pots needed to be deeper to allow for vegetation rooting depths to be encompassed adequately; 2) forcing energy balance closure was not supported by two of the studies (summer and winter); 3) latent heat flux (λE) gap filling of night time EC data during winter over estimated values by about 10 W m-2; and 4) the spring lysimeter study verified eddy covariance measurements including the closure forcing method. Some uncertainty still exists as to the accuracy of both lysimeter and EC methods of evaporation measurement because both methods still have potential biases, however for the purpose of this study, it would appear data are sufficiently accurate to have confidence in results. Energy and water vapour fluxes varied on both a diurnal and seasonal timescale. Diurnally, fluxes were small or negative at night and were highest during the day, usually at solar noon. Seasonally, spring and summer had the highest energy and evaporation fluxes and winter rates were small but tended to exceed available energy supply. Evaporation was constrained by soil moisture availability during summer and by energy availability during winter. Estimated annual evaporation at Scott Farm was 755 mm, 72% of precipitation. Two evaporation models were compared to eddy covariance evaporation (EEC) measurements; the FAO56 Penman-Monteith model (Eo) and the Priestley-Taylor model (EPT). Both models over estimated evaporation during dry conditions and slightly under estimated during winter. The α coefficient that is applied to EPT was not constant and a seasonally adjusted value would be most appropriate. A crop coefficient of 1.13 is needed for Eo measurements during moist conditions. Eo began over estimating evaporation when soil moisture contents dropped below ~44%. A water stress adjustment was applied to both models which improved evaporation estimates, however early onset of drying was not able to be adjusted for. The adjusted Eo model is the most accurate overall, when compared to EEC.

La technique micro météorologique de covariance de turbulences est utilisée pour estimer les échanges de CO2 entre les écosystèmes et latmosphère. Des centaines de sites instrumentés, répartis dans le monde entier, lutilisent désormais pour étudier une grande variété décosystèmes. Cette technique est cependant entachée dune erreur systématique lorsquelle est appliquée sur des couverts hauts comme des forêts, en conditions atmosphériques stables, cest-à-dire essentiellement pendant les nuits peu venteuses et sans couverture nuageuse. Pendant ces périodes, en effet, le transport turbulent serait concurrencé par un autre mécanisme de transport qui est ladvection. Dans ce travail, la présence dadvection a été testée sur le site expérimental forestier de Vielsalm (Belgique) et son importance a été évaluée. A cette fin, un dispositif expérimental permettant des mesures de vitesse de vent, de concentration de CO2 et de température de lair à lintérieur de la forêt a été installé. Il a permis la mise en évidence, pendant les périodes stables, découlements gravitationnels se réalisant près du sol suite au refroidissement des surfaces et à la présence dune faible pente. Il a été montré que ces écoulements étaient responsables du transport advectif de CO2. Une analyse dincertitude a été menée à laide de campagnes de mesures spécifiques. Elle a conclu à la faisabilité des mesures de gradients verticaux et surtout horizontaux de CO2 sur le site mais a montré que le plus grand facteur dincertitude portait sur les estimations de la composante verticale de la vitesse au-dessus de et dans la forêt. Malgré ces incertitudes, une analyse fine des épisodes gravitationnels a permis de mettre en évidence un mécanisme cohérent liant les écoulements dair et le champ des concentrations de CO2 et permettant de mieux comprendre comment le CO2 pouvait être transporté latéralement et verticalement par les écoulements gravitationnels. Finalement, la faisabilité dune correction basée sur lestimation des termes advectifs a été évaluée. Il a été montré que les incertitudes portant sur ladvection étaient trop importantes pour permettre daméliorer sensiblement le bilan nocturne des flux de CO2 au moyen de cette méthode. Celle-ci savère toutefois intéressante pour mieux comprendre les processus de transport à loeuvre dans un couvert forestier./The micrometeorological technique of eddy-covariance is used for the estimation of the CO2 exchange between the ecosystems and the atmosphere. Hundreds of instrumented sites, spread all over the world, use it henceforth to study a great variety of ecosystems. This technique is however affected by a systematic error when applied above tall canopies like forests, in stable atmospheric conditions, i.e. primarily during non windy nights without cloud cover. Indeed, during these periods the turbulent transport would be competed with by another transport mechanism which is called advection. In this work, the presence of advection has been tested on the experimental forested site of Vielsalm (Belgium) and its importance has been evaluated. For this purpose, an experimental set-up allowing the measurements of wind velocity, CO2 concentration and temperature of the air inside the forest has been installed. It has allowed the description, for the stable periods, of gravitational flows being carried out close to the ground due to the cooling of surfaces and the presence of a weak slope. These flows were shown to be responsible for advective CO2 transport. An uncertainty analysis has been carried out using dedicated measurement campaigns. It has conclude with the feasibility of measurements of vertical and especially horizontal CO2 gradients on the site but has shown that the greatest factor of uncertainty related to the estimate of the vertical velocity component above and in the forest. In spite of these uncertainties, a fine analysis of the gravitational episodes has made it possible to highlight a coherent mechanism linking the flow field and the CO2 concentration field and making it possible to better understand how CO2 could be transported laterally and vertically by the gravitational flows. Finally, the feasibility of a correction based on the estimate of the advective terms has been evaluated. It has been shown that uncertainties relating to advection were too important to make it possible to appreciably improve the night assessment of CO2 fluxes by means of this method. This one proves however interesting for better understanding the processes of transport at work in a forest cover.

The boreal forest is one of the worlds largest forest biomes and comprises a major portion of the terrestrial carbon (C) sink. Quantifying the net C change in forest ecosystems is an important step in understanding and modeling the global C cycle. The goals of this project were: to estimate and compare the total change in ecosystem C over a 10-year period in two boreal forest stands using biometric and eddy-covariance approaches, and to evaluate the year-to-year changes in C uptake. This study utilized 10 years of eddy-covariance data and ecosys model data from the Old Aspen (OA) and Old Jack Pine (OJP) sites in central Saskatchewan, part of the Boreal Ecosystem Research and Monitoring Sites (BERMS). According to the eddy-covariance and C stock approaches, between 1994 and 2004 the net change in C storage at OA was 15.6 ± 4.0 and 18.2 ± 8.0 Mg C ha-1, respectively. At OJP, the 10-year net change in C storage from eddy-covariance was 5.8 ± 2.0 Mg C ha-1 in comparison to 6.9 ± 1.6 Mg C ha-1 from the carbon stock approach. While both sites were sinks of C between 1994 and 2004, the greatest increase in C occurred in different components - the forest floor at OA (14.6 Mg C ha-1) and in the living vegetation at OJP (8.0 Mg C ha-1). In 2004, total ecosystem C content was greater at OA (180.6 Mg C ha-1) than OJP (78.9 Mg C ha-1), with 50% (OA) and 39% (OJP) of the C in the detritus and mineral soil pools. During the 10-year period of eddy-covariance measurements, there was a positive correlation between both annual and growing season gross ecosystem photosynthesis (GEP) and live stem C biomass increment at OA, whereas no significant relationships were found at OJP. Stem C increment accounted for 30% of total net primary productivity (NPP) at both sites, and NPP/GEP ratios were 0.36 and 0.32 at OA and OJP, respectively. Overall, this study found good agreement between eddy-covariance and biometric estimates of ecosystem C change at OA and OJP between 1994 and 2004. Over that period at OA, eddy-covariance estimates of photosynthesis captured the inter-annual variability in C uptake based on the growth of tree rings.

Methane fluxes were calculated from measurements carried out at a temperate upland forest in Central Ontario using the eddy covariance method over five months in the summer and fall seasons of 2011. Measurements were made by an off-axis integrated cavity output spectrometer Fast Greenhouse Gas Analyzer (FGGA) which simultaneously measured methane (CH4), carbon dioxide (CO2), and water at 10 Hz sampling rates. Observed methane fluxes showed net uptake of methane over the measurement period with an average uptake flux value (±standard deviation of the mean) of -2.7±0.13 nmol m-2 s-1. Methane fluxes showed a diurnal pattern of increased uptake during the day and increasing uptake with seasonal progression. There was also a significant correlation in methane fluxes with soil water content and wind speed. Comparison of the FGGA measurements to those using a static chamber method and canister sampling showed close agreement in flux and mixing ratio values respectively.

Methane fluxes were calculated from measurements carried out at a temperate upland forest in Central Ontario using the eddy covariance method over five months in the summer and fall seasons of 2011. Measurements were made by an off-axis integrated cavity output spectrometer Fast Greenhouse Gas Analyzer (FGGA) which simultaneously measured methane (CH4), carbon dioxide (CO2), and water at 10 Hz sampling rates. Observed methane fluxes showed net uptake of methane over the measurement period with an average uptake flux value (±standard deviation of the mean) of -2.7±0.13 nmol m-2 s-1. Methane fluxes showed a diurnal pattern of increased uptake during the day and increasing uptake with seasonal progression. There was also a significant correlation in methane fluxes with soil water content and wind speed. Comparison of the FGGA measurements to those using a static chamber method and canister sampling showed close agreement in flux and mixing ratio values respectively.

Large-scale mining practices have disturbed many natural watersheds in northern Alberta, Canada. To restore disturbed landscapes and ecosystems functions, reconstruction strategies have been adopted with the aim of establishing sustainable reclaimed lands. The success of the reconstruction process depends on the design of reconstruction strategies, which can be optimized by improving the understanding of the controlling hydrological processes in the reconstructed watersheds. Evapotranspiration is one of the important components of the hydrological cycle; its estimation and analysis are crucial for better assessment of the reconstructed landscape hydrology, and for more efficient design. The complexity of the evapotranspiration process and its variability in time and space has imposed some limitations on previously developed evapotranspiration estimation models. The vast majority of the available models estimate the rate of potential evapotranspiration, which occurs under unlimited water supply condition. However, the rate of actual evapotranspiration (AET) depends on the available soil moisture, which makes its physical modeling more complicated than the potential evapotranspiration. The main objective of this study is to estimate and analyze the AET process in a reconstructed landscape.<p> Data driven techniques can model the process without having a complete understanding of its physics. In this study, three data driven models; genetic programming (GP), artificial neural networks (ANNs), and multilinear regression (MLR), were developed and compared for estimating the hourly eddy covariance (EC)-measured AET using meteorological variables. The AET was modeled as a function of five meteorological variables: net radiation (Rn), ground temperature (Tg), air temperature (Ta), relative humidity (RH), and wind speed (Ws) in a reconstructed landscape located in northern Alberta, Canada. Several ANN models were evaluated using two training algorithms of Levenberg-Marquardt and Bayesian regularization. The GP technique was employed to generate mathematical equations correlating AET to the five meteorological variables. Furthermore, the available data were statistically analyzed to obtain MLR models and to identify the meteorological variables that have significant effect on the evapotranspiration process. The utility of the investigated data driven models was also compared with that of HYDRUS-1D model, which is a physically based model that makes use of conventional Penman-Monteith (PM) method for the prediction of AET. HYDRUS-1D model was examined for estimating AET using meteorological variables, leaf area index, and soil moisture information. Furthermore, Wavelet analysis (WA), as a multiresolution signal processing tool, was examined to improve the understanding of the available time series temporal variations, through identifying the significant cyclic features, and to explore the possible correlation between AET and the meteorological signals. WA was used with the purpose of input determination of AET models, a priori.<p> The results of this study indicated that all three proposed data driven models were able to approximate the AET reasonably well; however, GP and MLR models had better generalization ability than the ANN model. GP models demonstrated that the complex process of hourly AET can be efficiently modeled as simple semi-linear functions of few meteorological variables. The results of HYDRUS-1D model exhibited that a physically based model, such as HYDRUS-1D, might perform on par or even inferior to the data driven models in terms of the overall prediction accuracy. The developed equation-based models; GP and MLR, revealed the larger contribution of net radiation and ground temperature, compared to other variables, to the estimation of AET. It was also found that the interaction effects of meteorological variables are important for the AET modeling. The results of wavelet analysis demonstrated the presence of both small-scale (2 to 8 hours) and larger-scale (e.g. diurnal) cyclic features in most of the investigated time series. Larger-scale cyclic features were found to be the dominant source of temporal variations in the AET and most of the meteorological variables. The results of cross wavelet analysis indicated that the cause and effect relationship between AET and the meteorological variables might vary based on the time-scale of variation under consideration. At small time-scales, significant linear correlations were observed between AET and Rn, RH, and Ws time series, while at larger time-scales significant linear correlations were observed between AET and Rn, RH, Tg, and Ta time series.

The boreal forest is one of the worlds largest forest biomes and comprises a major portion of the terrestrial carbon (C) sink. Quantifying the net C change in forest ecosystems is an important step in understanding and modeling the global C cycle. The goals of this project were: to estimate and compare the total change in ecosystem C over a 10-year period in two boreal forest stands using biometric and eddy-covariance approaches, and to evaluate the year-to-year changes in C uptake. This study utilized 10 years of eddy-covariance data and ecosys model data from the Old Aspen (OA) and Old Jack Pine (OJP) sites in central Saskatchewan, part of the Boreal Ecosystem Research and Monitoring Sites (BERMS). According to the eddy-covariance and C stock approaches, between 1994 and 2004 the net change in C storage at OA was 15.6 ± 4.0 and 18.2 ± 8.0 Mg C ha-1, respectively. At OJP, the 10-year net change in C storage from eddy-covariance was 5.8 ± 2.0 Mg C ha-1 in comparison to 6.9 ± 1.6 Mg C ha-1 from the carbon stock approach. While both sites were sinks of C between 1994 and 2004, the greatest increase in C occurred in different components - the forest floor at OA (14.6 Mg C ha-1) and in the living vegetation at OJP (8.0 Mg C ha-1). In 2004, total ecosystem C content was greater at OA (180.6 Mg C ha-1) than OJP (78.9 Mg C ha-1), with 50% (OA) and 39% (OJP) of the C in the detritus and mineral soil pools. During the 10-year period of eddy-covariance measurements, there was a positive correlation between both annual and growing season gross ecosystem photosynthesis (GEP) and live stem C biomass increment at OA, whereas no significant relationships were found at OJP. Stem C increment accounted for 30% of total net primary productivity (NPP) at both sites, and NPP/GEP ratios were 0.36 and 0.32 at OA and OJP, respectively. Overall, this study found good agreement between eddy-covariance and biometric estimates of ecosystem C change at OA and OJP between 1994 and 2004. Over that period at OA, eddy-covariance estimates of photosynthesis captured the inter-annual variability in C uptake based on the growth of tree rings.

Large-scale mining practices have disturbed many natural watersheds in northern Alberta, Canada. To restore disturbed landscapes and ecosystems functions, reconstruction strategies have been adopted with the aim of establishing sustainable reclaimed lands. The success of the reconstruction process depends on the design of reconstruction strategies, which can be optimized by improving the understanding of the controlling hydrological processes in the reconstructed watersheds. Evapotranspiration is one of the important components of the hydrological cycle; its estimation and analysis are crucial for better assessment of the reconstructed landscape hydrology, and for more efficient design. The complexity of the evapotranspiration process and its variability in time and space has imposed some limitations on previously developed evapotranspiration estimation models. The vast majority of the available models estimate the rate of potential evapotranspiration, which occurs under unlimited water supply condition. However, the rate of actual evapotranspiration (AET) depends on the available soil moisture, which makes its physical modeling more complicated than the potential evapotranspiration. The main objective of this study is to estimate and analyze the AET process in a reconstructed landscape.<p> Data driven techniques can model the process without having a complete understanding of its physics. In this study, three data driven models; genetic programming (GP), artificial neural networks (ANNs), and multilinear regression (MLR), were developed and compared for estimating the hourly eddy covariance (EC)-measured AET using meteorological variables. The AET was modeled as a function of five meteorological variables: net radiation (Rn), ground temperature (Tg), air temperature (Ta), relative humidity (RH), and wind speed (Ws) in a reconstructed landscape located in northern Alberta, Canada. Several ANN models were evaluated using two training algorithms of Levenberg-Marquardt and Bayesian regularization. The GP technique was employed to generate mathematical equations correlating AET to the five meteorological variables. Furthermore, the available data were statistically analyzed to obtain MLR models and to identify the meteorological variables that have significant effect on the evapotranspiration process. The utility of the investigated data driven models was also compared with that of HYDRUS-1D model, which is a physically based model that makes use of conventional Penman-Monteith (PM) method for the prediction of AET. HYDRUS-1D model was examined for estimating AET using meteorological variables, leaf area index, and soil moisture information. Furthermore, Wavelet analysis (WA), as a multiresolution signal processing tool, was examined to improve the understanding of the available time series temporal variations, through identifying the significant cyclic features, and to explore the possible correlation between AET and the meteorological signals. WA was used with the purpose of input determination of AET models, a priori.<p> The results of this study indicated that all three proposed data driven models were able to approximate the AET reasonably well; however, GP and MLR models had better generalization ability than the ANN model. GP models demonstrated that the complex process of hourly AET can be efficiently modeled as simple semi-linear functions of few meteorological variables. The results of HYDRUS-1D model exhibited that a physically based model, such as HYDRUS-1D, might perform on par or even inferior to the data driven models in terms of the overall prediction accuracy. The developed equation-based models; GP and MLR, revealed the larger contribution of net radiation and ground temperature, compared to other variables, to the estimation of AET. It was also found that the interaction effects of meteorological variables are important for the AET modeling. The results of wavelet analysis demonstrated the presence of both small-scale (2 to 8 hours) and larger-scale (e.g. diurnal) cyclic features in most of the investigated time series. Larger-scale cyclic features were found to be the dominant source of temporal variations in the AET and most of the meteorological variables. The results of cross wavelet analysis indicated that the cause and effect relationship between AET and the meteorological variables might vary based on the time-scale of variation under consideration. At small time-scales, significant linear correlations were observed between AET and Rn, RH, and Ws time series, while at larger time-scales significant linear correlations were observed between AET and Rn, RH, Tg, and Ta time series.

<p>A large portion of the anthropogenic emissions of greenhouse gases (<italic>GHG</italic>s) are cycled through the terrestrial biosphere. Quantifying the exchange of these gases between the terrestrial biosphere and the atmosphere is critical to constraining their atmospheric budgets now and in the future. These fluxes are governed by biophysical processes like photosynthesis, transpiration, and microbial respiratory processes which are driven by factors like meteorology, disturbance regimes, and long term climate and land cover change. These complex processes occur over a broad range of temporal (seconds to decades) and spatial (millimeters to kilometers) scales, necessitating the application of simplifying models to forecast fluxes at the scales required by climate mitigation and adaptation policymakers. </p><p>Over the long history of biophysical research, much progress has been made towards developing appropriate models for the biosphere-atmosphere exchange of <italic>GHG</italic>s. Many processes are well represented in model frameworks, particularly at the leaf scale. However, some processes remain poorly understood, and models do not perform robustly over coarse spatial scales and long time frames. Indeed, model uncertainty is a major contributor to difficulties in constraining the atmospheric budgets of greenhouse gases. </p><p>The central objective of this dissertation is to reduce uncertainty in the quantification and forecasting of the biosphere-atmosphere exchange of greenhouse gases by addressing a diverse array of research questions through a combination of five unique field experiments and modeling exercises. In this first chapter, nocturnal evapotranspiration -- a physiological process which had been largely ignored until recent years -- is quantified and modeled in three unique ecosystems co-located in central North Carolina, U.S.A. In the second chapter, more long-term drivers of evapotranspiration are explored by developing and testing theoretical relationships between plant water use and hydraulic architecture that may be readily incorporated into terrestrial ecosystem models. The third chapter builds on this work by linking key parameters of carbon assimilation models to structural and climatic indices that are well-specified over much of the land surface in an effort to improve model parameterization schemes. The fourth chapter directly addresses questions about the interaction between physiological carbon cycling and disturbance regimes in current and future climates, which are generally poorly represented in terrestrial ecosystem models. And the last chapter explores effluxes of methane and nitrous oxide (which are historically understudied) in addition to CO₂ exchange in a large temperate wetland ecosystem (which is an historically understudied biome). While these five case studies are somewhat distinct investigations, they all: a) are all grounded in the principles of biophysics, b) rely on similar measurement and mathematical modeling techniques, and c) are conducted under the governing objective of reducing measurement and model uncertainty in the biosphere-atmosphere exchange of greenhouse gases.</p> / Dissertation