Effect of aggregating spatial parameters on modelling forest carbon and water fluxes

Estimating spatial variability of carbon and water fluxes is an essential task in ecological modelling. In this article, the sensitivity of carbon and water fluxes to the spatial variability of biochemical and structural properties of canopies is assessed in beech forests using a process-based model (CASTANEA). Firstly, a sensitivity analysis was carried out by varying simultaneously a combination of six key parameters within a realistic range: the above ground wood biomass (B), the soil water reserve (SWR), the canopy clumping factor (CF), the leaf area index (L), the leaf mass per area of sunlit leaves (Msun) and the leaf nitrogen content (N). Secondly, three spatial scales of variability were considered using three study sites whose areas ranged from 0.8 to 1000 ha. The first area studied was a heterogeneous stand located in old-growth forest in Fontainebleau (south of Paris, France). The spatial variability of the biophysical and biochemical ecosystem characteristics in 80 m2 out of 100 m2 was determined. For the two other case studies, we selected a sample of nine plots in which the key input parameters were measured. Sensitivity analysis indicated that photosynthesis and ecosystem respiration show a moderate non-linear response to L, SWR and B. In spite of these non-linear responses, the three case studies revealed that using parameters averaged over the whole area, induces only a slight bias in the estimation of carbon fluxes and almost no bias in the estimation of water fluxes. The implication of the low sensitivity of carbon and water fluxes to parameter aggregation is discussed in relation to the general problem of the scaling up fluxes from ecosystems to large forest regions.