Improvement of LPJ dynamic global vegetation model by means of numerical assimilation methods: possible implications for regional climate models

摘要

Some of the feedbacks between vegetation and climate have been studied in the Euro-Mediterranean area using both models and data. The Lund-Potsdam-Jena (LPJ) Dynamic Global Vegetation Model describes the water, carbon, and energy exchange between land surface and atmosphere by means of a given set of parameters and input variables. In order to retrieve the underling probability density function of some key-model parameters controlling water and carbon cycle as well as to improve the efficiency of LPJ to simulate water and carbon fluxes a data assimilation system has been developed; it is based on a Bayesian approach that consistently combines prior knowledge about parameters with observations. Daily values of evapotranspiration and gross primary production, measured with eddy covariance technique in ten different CarboeuropeIP sites, have been compared with modeled data in order to constrain parameter values and uncertainties. Results show how data assimilation is a useful tool to improve the ability of the model to simulate correctly water and carbon fluxes at local scale: after the inversion, in fact, LPJ successfully matches the observed seasonal cycle of the diverse fluxes, and corrects for the prior misfit to day-time GPP and ET. The impact of land cover change on regional climate have been analyzed using the mesoscale model RegCM3. Three different simulations have been performed to asses the effects of an hypothetical deforestation and afforestation on climate. Results show how land cover changes have a substantial impact on dynamic and thermodynamic, and how also area does not affected by land cover changes shows a significant variability in some climatic fields. Finally, the land cover changes have an important impact on the extreme events. This thesis highlights how vegetation dynamics and climate influence each other. For such reason to improve simulation results we should develop fully coupled models that take into account some of the most important feedbacks between land surface and atmosphere.