No abstract

No abstract

In this study we present the results of the comparison of three different algorithms: the Genetic Algorithm coupled with Sequential Quadratic Programming (GA-SQP), the Pattern Search also coupled with SQP (PS-SQP) and the Shuffled Complex Evolution (SCE-UA). The analyses were conducted using a conceptual rainfall-runoff model applied both to a single basin and to a complex basin. For both types of basin, a theoretical case without model and data errors was considered, in which the true values of the parameters are known a priori, and several real-world cases where model and data errors exist. With reference to the single basin, the SCE-UA algorithm was the most reliable while the other two algorithms gave solutions equivalent to those of the SCE-UA in the theoretical case, but in the realworld cases they showed an increasing tendency (particularly the PS-SQP) to be trapped in local minima. With reference to the complex basin, none of the three algorithms identified the exact solution in the theoretical case. However, the SCE-UA was the one which systematically approximated it better than the others. In the real-world case its solutions were stable but characterized by many parameter values set at the boundary of their own range. The other two algorithms produced a very unstable set of parameters.Techniques d'optimisation globales pour le calage de modèles conceptuels pluie-débit Résumé Cette étude présente les résultats de la comparaison entre trois algorithmes différents: l'Algorithme Génétique couplé à la Programmation Quadratique Séquentientelle (GA-SQP), le "Pattern Search" lui-aussi couplé à la SQP (PS-SQP) et le "Shuffled Complex Evolution" (SCE-UA). Les analyses ont été effectuées à partir d'un modèle conceptuel pluie-débit appliqué d'une part à un bassin simple et d'autre part à un bassin complexe. Pour les deux bassins, nous avons considéré un cas théorique, sans erreurs ni de modèle ni de données, où les valeurs véritables des paramètres sont connues a priori et plusieurs cas réels où existent des erreurs de modèle et de données. Dans le cas du bassin simple, l'algorithme SCE-UA s'est révélé le plus fiable tandis que les deux autres algorithmes donnaient des solutions équivalentes à celles de SCE-UA dans le cas théorique; ils présentaient toutefois dans les cas réels une tendance croissante (notamment le PS-SQP) à se limiter à des minima locaux. Dans le cas du bassin complexe, aucun des trois algorithmes n'a identifié la solution exacte du cas théorique. Néanmoins, le SCE-UA s'est systématiquement mieux approché de la solution que les autres. Dans les cas réels ses solutions étaient stables mais caractérisées par la tendance de nombreux paramètres à se fixer à une limite de leur intervalle. Les deux autres algorithmes ont produit des ensembles de paramètres instables. Open for discussion until 1 December 1998 444Marco Franchini et al.

The problem of density-dependent transport of salt in unconfined coastal aquifers is solved numerically by means of an implicit Eulerian-Lagrangian finite element formulation. Such a formulation leads to symmetric positive definite finite element matrices which are ideally suited for efficient solution by preconditioned conjugate gradient methods. Additional known advantages of the formulation are unconditional stability, reduced numerical dispersion and suitability for parallel computation. The method has been used to study the effect of dewatering on seawater intrusion within a vertical cross section through an aquifer in southern Italy, related to the construction of a thermoelectric power plant. To investigate the extent to which the dependence of fluid density on salt concentration affects the numerical solution, the flow and advection-dispersion equations were solved in coupled (iterative), partially coupled (noniterative) and completely decoupled modes. Partial coupling was found to yield results very close to those obtained by full coupling but at great savings in computer time; the less rigorous decoupled approach led to results substantially different from those obtained through coupling and partial coupling. Effects of aquifer heterogeneity and the construction of a cutoff wall on seawater intrusion are discussed.

The paper evaluates the performance of a non-parametric scheme, the nearest neighbour (NN) method, to predict the daily mean discharge in a mountain basin supplying a hydroelectric reservoir in northeastern Italy. The results are compared with those of an autoregressive model with exogenous input (ARX), coupled with a previously developed snow cover evolution model. Both methods give good performances, but the NN prediction requires a much simpler simulation structure. In the case investigated, for example, the snowpack accumulationmelting model can be completely eliminated.