Calibration of a stochastic model for riparian vegetation dynamics from LiDAR acquisitions. The case study of Cinca River.

Carbon cycle has recently been revised in view of the understanding of the relevant contribution of inland waters to the definition of the total C budget. Particularly, the role of rivers has been revaluated, recognizing their relevance as active components of this cycle, and not merely as passive pipes that convey carbon from lands to oceans, and thus driving numerous academic institutions to point their researches to the so-called River Carbon Cycle. As the contribution of Politecnico di Torino in this direction is the investigation of the carbon sequestration in riparian corridors, this dissertation was inserted in this frame by the definition of its two main objectives: the evaluation of the use of LiDAR to estimate riparian biomass and the calibration of a stochastic model, which describes riparian vegetation growth according to topographic and hydrological constraints, on the basis of real LiDAR data measurements. Although numerous LiDAR-based models for the estimation of above-ground biomass in forest stands have been regressed, no studies concerned riparian corridors, where tree population is generally younger and sparser. Thus, the reliability of these forest models in riparian environments was tested. Firstly, an insight about the state of the art of LiDAR technology and its airborne implementation for forestry was provided. Secondly, a selection of literature models was applied to the study area, which was a reach of Cinca River (Spain), in order to choose the most reliable one, after the processing of LiDAR data with FUSION/LDV, a free software package released by USDA. Finally, the influence of grid discretization on results and the sensibility of models to record vegetation growth over time were evaluated. The achieved results were the definition of a procedure that combines the use of GIS and FUSION/LDV for determining vegetation statistics and the choice of the model proposed by Means et al (2000) for the conversion of these statistics in above-ground biomass. This model demonstrated to be implementable also with coarser grid discretization, thus reducing computation time, and to be sensible enough to record vegetation growth. This defined procedure was applied to the study site, in order to return the pdf of dimensional biomass and its first moments. These results were used to calibrate the stochastic model provided by Camporeale and Ridolfi in 2006 and that describes the distribution of phreatophyte riparian vegetation by modelling the impact that the randomness of hydrological fluctuations has on its growth. Indeed, despite literature values for most of its parameters were already available, the parameter that represents the ratio between the rates of, respectively, decay of vegetation during inundations and growth during exposure, still needed to be properly set. To this purpose, hydrometric data were processed with HEC-RAS and MATLAB to obtain the pdf and integral scale of water levels, while the geometry was set by the realization of a DTM with FUSION/LDV. The stochastic model was calibrated by realizing a MATLAB script that minimizes the deviation among computed values of dimensionless biomass, derived by the combination of geometric and hydrological data with parameters linked to tree species, and real data provided by LiDAR acquisitions. The reliability this calibration was successively assessed and returned in a cartographic representation.