Assessing tree seedling vitality tests using sensitivity analysis of a process-based growth model

A process-based model of tree seedling growth was used to assess the predictive capability of various plant vitality tests. Eight model parameters which could be directly equated with specific plant vitality tests were varied in a sensitivity analysis, using a Monte Carlo approach with Latin hypercube sampling. Probability distributions of these parameters were estimated using data from two research projects in which relevant parameters were measured in a range of experiments across Europe. The sensitivity of predicted growth to these parameters was analysed at four European sites chosen to have contrasting climates, ranging from cool oceanic to Mediterranean (Greece, Portugal, Germany and Scotland). In general, those parameters to which the predictions are most sensitive will correspond to the best vitality tests. The objective was to identify vitality tests which will be useful predictive tools in the context of forest establishment, and to what extent their usefulness varies with climate. The results were related to experimental evaluations of vitality tests reported in the literature. The pattern of sensitivity to the eight parameters was relatively consistent at the four sites. In all cases, initial mass was the most sensitive parameter, followed by maximum relative growth rate of roots. Initial soluble nitrogen content and root fraction generally ranked third and fourth, though the latter was rather variable. The remaining four parameters (maximum electron transport rate, maximum carboxylation rate, initial starch content and initial soluble sugar content) were generally poor indicators of growth. The results suggest that commonly-used assessments such as stem diameter and root growth potential (RGP) should be reliable indicators of field performance. The value of our approach is in quantifying the importance of observed variation in a parameter in relation to variation in other model parameters and inputs. The approach permits seedling vitality tests to be linked to a process-based modelling framework.