A new individual-based forest model for the species beech (Fagus sylvatica L.) was developed and implemented. The model called BEEch Plasticity (BEEP) describes tree crown plasticity phenomenologically and is able to model aboveground competition for PAR on a process basis. The current debate about the tree interactions in near-natural beech forests and their role in emergent forest structures and dynamics led to the research questions if (1) observed patterns can be modeled and reproduced by only describing the aboveground tree interactions, (2) what effects tree crown plasticity has on the structure and dynamics of near-natural beech forests, and (3) what effects selective thinning has on the structure and dynamics of near-natural beech forests. The BEEP model was developed, parameterized, calibrated, and validated according to data from the unmanaged forest `Schattiner Zuschlag' near Lübeck, North-Germany, while additional data from the sites Langula (Thuringia) and Fabrikschleichach (Bavaria) was used for model parameterization and calibration. Three simulation experiments were conducted. In the first experiment, the BEEP model was run 10 times for 2000 time steps with plastic tree crowns and the emergent forest structure was analyzed using structural indices. In the second experiment, the BEEP model was run again 10 times for 2000 time steps but with a modified crown model that only uses rotation-symmetric tree crowns. In the third experiment, the BEEP model was enhanced with a selective thinning procedure that uses target trees with specific diameter and heights as thinning objects. Forest structure was analyzed through the application of structural indices that capture different aspects of forest structure and by means of characterization of forest development phases. Analysis was accomplished only for the time steps 1000-2000 in order to allow transient oscillation in forest dynamics to develop. The results showed that the focus on aboveground competition and tree interactions sufficed to model beech forests and reproduced a wide range of patterns observed in near-natural and old-growth beech forest. In particular, the BEEP model was able to simulate a multi-layered forest structure with a mosaic structure of several developmental stages on a relatively small area of 0.5 ha. The simulated forest had wide diameter and age distributions. The diameter distribution was reversed-J-shaped. The age range of canopy trees exceeded 200 years. The comparison between simulations with plastic and rotation-symmetric tree crowns revealed that crown plasticity reduced tree competition for crown space and PAR and enhanced the forest structure and heterogeneity in the long term by allowing more tree cohorts of dfferent developmental stages to coexist. This supports the notion that crown plasticity drives beech forest dynamics in near-natural forests. The comparison between simulations with plastic tree crowns and with additional selective thinning showed that thinning does not affect the forest structural heterogeneity and reduces tree crown competition, while spatial patterns of tree positions remained unaltered. However, crown centroids were more regularly distributed. Model assumptions in the submodel routines, especially in the radiation and mortality submodel, question the reliability of the model results, because of the high sensitivity that these routines evoke on model outcomes. Therefore, revised versions of the submodels and a thoroughly validated crown growth model, may produce different results. Thus, the results presented in this study should be treated with care and cannot be used for generalizations about tree interactions in near-natural beech forests.

Les modèles structure-fonction de la croissance des plantes (FSPMs) combinent la description du fonctionnement biophysique et du développement architectural des plantes. On peut distinguer deux grandes familles de FSPM : d'une part les modèles décrivant finement la structure de la plante au niveau de l'organe et d'autre part les modèles à plus grande échelle qui s'intéressent directement à la forme du houppier. La paramétrisation du premier type de modèle est souvent difficile car elle nécessite des données expérimentales très riches. A l'inverse, les modèles à plus grande échelle mettent généralement en œuvre des lois empiriques qui ne permettent pas de décrire la plasticité de la croissance, et l'adaptation de la plante à des conditions environnementales différentes.Pour répondre à ces problématiques, nous nous tournons vers un nouveau paradigme : Motivé par le succès du concept de la densité spatiale dans les modèles en écologie des populations, cette thèse caractérise la distribution spatiale de feuillage dans les plantes par la densité de surface foliaire , ce qui permet une description locale ouvrant la voie à une prise en compte de la plasticité des plantes, tout en ne décrivant pas chaque feuille individuellement, ce qui permet de modéliser des vieux et grands arbres, dont le nombre de feuilles est sinon trop lourd à gérer du point de vue des calculs. Cette thèse présente des modèles dynamiques de croissance développés spécifiquement pour les plantes agricoles et les arbres. Nous explorons des approches mathématiques différentes en temps discrète et continue, tout en examinant d'un œil critique leurs aptitudes conceptuelles ainsi que des possibilités de simplifications et de solutions analytiques dans l'optique de l'accélération des simulations.La densité foliaire permet le calcul de l'interception de lumière par la loi de Beer-Lambert et la production de biomasse grâce au concept d'efficience d'utilisation de la lumière. Le mécanisme central qui est considéré pour les différentes approches développées dans cette thèse est celui de l'expansion locale de la surface foliaire dans la direction du gradient de lumière. Par ce concept téléonomique, nous faisons l'hypothèse que la plante cherche par son développement à optimiser la productivité de la surface foliaire pour la production de biomasse. Ce principe induit ainsi un développement horizontal et vertical du feuillage vers l'extérieur du houppier. Le développement horizontal cesse quand on s'approche trop de plantes voisines, leur ombrage diminuant le gradient de lumière et donc l'expansion de densité de surface foliaire dans ces directions. Le modèle de production de biomasse est également généralisé pour une prise en compte explicite de la teneur en eau du sol en introduisant une composante hydraulique permettant de décrire l'équilibre mécaniste entre le potentiel hydrique dans les feuilles et la transpiration par la régulation stomatale. Finalement, nous prenons en compte l'allocation de biomasse produite à d'autres compartiments de la plante tels que les racines et le bois selon la théorie du « pipe model ».Les résultats des modèles sont comparés à un large jeu de données expérimentales sur des plantations à différentes densités et conditions environnementales. Celui-ci montre de remarquables capacités d'une part à prévoir les variables biométriques importantes (hauteur, diamètre du tronc) ainsi que certaines relations d'allométrie, et d'autre part à générer des formes de houppier en accord avec les formes observées, ceci pour les différents scénarios de compétition et comme propriété émergente du modèle. Ainsi, cette thèse démontre le potentiel du concept de densité de surface foliaire en modélisation de la croissance des plantes, par sa capacité à reproduire les comportements locaux et l'adaptation à des conditions environnementales variées sans compromettre l'efficacité et la robustesse. / Functional-structural plant growth models (FSPMs) have emerged as the synthesis of mechanistic process-based models, and geometry-focussed architectural models. In terms of spatial scale, these models can essentially be divided into small-scale models featuring a topologic­al architecture – often facing data-demanding parametrisations, parameter sensitivity, as well as computational heaviness, which imposes problematic limits to the age and size of in­dividuals than can be simulated – and large-scale models based on a description of crown shape in terms of rigid structures such as empirical crown envelopes – commonly strug­gling to allow for spatial variability and plasticity in crown structure and shape in response to local biotic or abiotic growth conditions.In response to these limitations, and motivated not least by the success-story of spatial density approaches in theoretical populations ecology, the spatial distribution of foliage in plants in this thesis is characterised in terms of spatial leaf density, which allows for a com­pletely local description that is a priori unrestricted in terms of plasticity, while being robust and computationally efficient. The thesis presents dynamic growth models specific­ally developed for crops and trees, exploring different mathematical frameworks in continu­ous and discrete time, while critically discussing their conceptual suitability and exploring analyt­ical simplifications and solutions to accelerate simulations.The law of Beer-Lambert on the passing of light though an absorbing medium allows to infer the local light conditions based on which local biomass production can be computed via a radiation use efficiency. A key unifying mechanism of the different models is the local ex­pansion of leaf density in the direction of the light gradient, which coincides with the direc­tion most promising with regard to future biomass productivity. This aspect falls into the line of teleonomic and optimization-oriented plant growth models, and allows to set aside the otherwise complex modelling of branching processes. The principle induces an expans­ive horizontal and upward-directed motion of foliage. Moreover, it mechanistically accounts for a slow-down of the horizontal expansion as soon as a neighbouring competitor's crown is reached, since the appropriate region is already shaded, implying a corresponding adapta­tion of the light gradient. This automatically results in narrower crowns in scenarios of in­creased competition, ultimately decreasing biomass production and future growth due to lesser amount of intercepted light. In an extension, the impact of water availability is incor­porated into the previously light-only dependency of biomass production by means of a novel hydraulic model describing the mechanistic balancing of leaf water potential and tran­spiration in the context of stomatal control. The allocation of produced biomass to other plant compartments such as roots and above-ground wood, e.g. by means of the pipe model theory, is readily coupled to leaf density dynamics.Simulation results are compared against a variety of empirical observations, ranging from long-term forest inventory data to laser-recorded spatial data, covering multiple abi­otic environmental conditions and growth resources as well as stand densities and thus de­grees of competition. The models generate a series of complex emergent properties includ­ing the realistic prediction of biometric growth parameters, the spontaneous adaptability and plasticity of crown morphologies in different competitive scenarios, the empirically documented insensitivity of height to stand density, the accurate deceleration of height growth, as well as popular allometric relationships – altogether demonstrating the potential of leaf density based approaches for efficient and robust plant growth modelling.