Cluster-based GSA: Global sensitivity analysis of models with temporal or spatial outputs using clustering

Roux, Sébastiên; Buis, Samuel; Lafolie, François; Lamboni, Matieyendou

doi:10.1016/j.envsoft.2021.105046

Cited by 12 publications

(14 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It would also be necessary to investigate sensitivity of some time series to get a more comprehensive vision of the model functioning. To do so, the temporal series can be analyzed as a multivariate output for example with clustering-based GSA (Roux et al, 2021) or using the principal components of the model's functional outputs. The definition and the use of adequate hydrological signatures such as proposed in Branger and McMillan (2020) and Horner (2020) may also be of interest to understand space-time variability and to capture a broader range of physical processes.…”

Section: Discussionmentioning

confidence: 99%

How to perform global sensitivity analysis of a catchment-scale, distributed pesticide transfer model? Application to the PESHMELBA model

Rouzies

Lauvernet

Sudret

et al. 2021

Preprint

View full text Add to dashboard Cite

Abstract. Pesticide transfers in agricultural catchments are responsible for diffuse but major risks to water quality. Spatialized pesticide transfer models are useful tools to assess the impact of the structure of the landscape on water quality. Before considering using these tools in operational contexts, quantifying their uncertainties is a preliminary necessary step. In this study, we explored how global sensitivity analysis can be applied to the recent PESHMELBA pesticide transfer model to quantify uncertainties on transfer simulations. We set up a virtual catchment based on a real one and we compared different approaches for sensitivity analysis that could handle the specificities of the model: high number of input parameters, limited size of sample due to computational cost and spatialized output. We compared Sobol' indices obtained from Polynomial Chaos Expansion, HSIC dependence measures and feature importance measures obtained from Random Forest surrogate model. Results showed the consistency of the different methods and they highlighted the relevance of Sobol' indices to capture interactions between parameters. Sensitivity indices were first computed for each landscape element (site sensitivity indices). Second, we proposed to aggregate them at the hillslope and the catchment scale in order to get a summary of the model sensitivity and a valuable insight into the model hydrodynamical behaviour. The methodology proposed in this paper may be extended to other modular and distributed hydrological models as there has been a growing interest in these methods in recent years.

show abstract

Section: Discussionmentioning

confidence: 99%

How to perform global sensitivity analysis of a catchment-scale, distributed pesticide transfer model? Application to the PESHMELBA model

Rouzies

Lauvernet

Sudret

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…4) also seem in line with the stress-gradient hypothesis (Maestre et al, 2009), an empirically observed pattern that states that in stressful environments, positive interactions should occur more often than in benign environments (e.g., Callaway, 2007). For the ecosystem that we consider, we interpret increasing temperature as increasing stress (e.g., Ruiz-Pérez and Vico, 2020) and structure as the best indicator for competitive interactions as the structure dictates resource allocation (e.g., bigger crown but identical stem diameter leads to more photosynthesis, or more sapwood to heartwood turnover requires less NPP). With this interpreta-tion, one would conclude that under increasing stress the importance of competition-related parameters decreases in the model, as expected from the stress-gradient hypothesis.…”

Section: Geographical and Environmental Patterns In Sensitivities And...mentioning

confidence: 99%

“…As the model is sensitive to parameters and environmental drivers and these elements influence each other, we treated them in a combined sensitivity and uncertainty analysis (Saltelli et al, 2019); however, when interpreting it should be kept in mind that the one group relates to uncertainties in the model, while the other is external, meaning that the two are conceptually very different (see also Dietze, 2017b). A certain ambiguity also arises from the definition of the indicators: here, we calculated sensitivities and uncertainties by capturing only linear components and second-order interactions, and we may therefore miss highly nonlinear (and in particular hump-shaped) responses in LPJ-GUESS (Roux et al, 2021). However, our comparison to uncertainties calculated with random forest variable importance, a method that would also capture nonlinearities, did not reveal any qualitative differences in the ranking of parameter importance (Appendix A1.3).…”

Section: Limitationsmentioning

confidence: 99%

Climate and parameter sensitivity and induced uncertainties in carbon stock projections for European forests (using LPJ-GUESS 4.0)

et al. 2022

View full text Add to dashboard Cite

Abstract. Understanding uncertainties and sensitivities of projected ecosystem dynamics under environmental change is of immense value for research and climate change policy. Here, we analyze sensitivities (change in model outputs per unit change in inputs) and uncertainties (changes in model outputs scaled to uncertainty in inputs) of vegetation dynamics under climate change, projected by a state-of-the-art dynamic vegetation model (LPJ-GUESS v4.0) across European forests (the species Picea abies, Fagus sylvatica and Pinus sylvestris), considering uncertainties of both model parameters and environmental drivers. We find that projected forest carbon fluxes are most sensitive to photosynthesis-, water-, and mortality-related parameters, while predictive uncertainties are dominantly induced by environmental drivers and parameters related to water and mortality. The importance of environmental drivers for predictive uncertainty increases with increasing temperature. Moreover, most of the interactions of model inputs (environmental drivers and parameters) are between environmental drivers themselves or between parameters and environmental drivers. In conclusion, our study highlights the importance of environmental drivers not only as contributors to predictive uncertainty in their own right but also as modifiers of sensitivities and thus uncertainties in other ecosystem processes. Reducing uncertainty in mortality-related processes and accounting for environmental influence on processes should therefore be a focus in further model development.

show abstract

“…Firstly, the class of objective functions (e.g., Saltelli & Tarantola, 2002;Saltelli et al, 2004) consists of all transformations of the model outputs that do not modify the initial distribution of the model inputs. It includes transformations done by i) projecting the model outputs onto a given basis (Campbell et al, 2006;Lamboni et al, 2011;Xiao et al, 2018), using the kernel-based principal components, using feature maps of the outputs (Aronszajn, 1950;Schölkopf & Smola, 2002;Berlinet et al, 2004); ii) considering the probabilities of the stochastic and dynamic outputs to exceed a given threshold (Lamboni et al, 2014), iii) using a regression-based classifier (Lamboni et al, 2016), and iv) considering the membership functions from either a crisp (a.k.a binary) or fuzzy clustering (Roux et al, 2021).…”

Section: Introductionmentioning

confidence: 99%

“…Performing global SA (e.g., Sobol, 1993;Borgonovo et al, 2014;Owen, 2014;Gamboa et al, 2014;Mara et al, 2015;Lamboni & Kucherenko, 2021;Fort et al, 2021) on MFs allows for identifying the input variables that drive the model output(s) into a given domain of interest or a specific model behavior (Lamboni et al, 2014;Roux et al, 2021). However, such analysis does not provide the drivers of specific model behaviors.…”

Section: Introductionmentioning

confidence: 99%

Kernel-based sensitivity indices for any model behavior and screening

Lamboni

2023

SESMO

View full text Add to dashboard Cite

Complex models are often used to understand interactions and drivers of human-induced and/or natural phenomena. It is worth identifying the input variables that drive the model output(s) in a given domain and/or govern specific model behaviors such as contextual indicators based on socioenvironmental models. Using the theory of multivariate weighted distributions to characterize specific model behaviors, we propose new measures of association between inputs and such behaviors. Our measures rely on sensitivity functionals (SFs) and kernel methods, including variance-based sensitivity analysis. The proposed ℓ1-based kernel indices account for interactions among inputs, higher-order moments of SFs, and their upper bounds are somehow equivalent to the Morris-type screening measures, including dependent elementary effects. Empirical kernel-based indices are derived, including their statistical properties for the computational issues, and numerical results are provided.

show abstract

Cluster-based GSA: Global sensitivity analysis of models with temporal or spatial outputs using clustering

Cited by 12 publications

References 27 publications

How to perform global sensitivity analysis of a catchment-scale, distributed pesticide transfer model? Application to the PESHMELBA model

How to perform global sensitivity analysis of a catchment-scale, distributed pesticide transfer model? Application to the PESHMELBA model

Climate and parameter sensitivity and induced uncertainties in carbon stock projections for European forests (using LPJ-GUESS 4.0)

Kernel-based sensitivity indices for any model behavior and screening

Contact Info

Product

Resources

About