FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W156929291> ?p ?o ?g. }

• W156929291 abstract "Changes to the way in which environmental model outcomes are reported are overdue.Increasingly, government agencies and stakeholder groups are demanding that model predictionsof system behaviour be accompanied by robust estimates of the potential errors associated withthose predictions. However, in many instances, these demands are not being met. The principalreason for this is the lack of a suitable, cost-effective methodology for implementation of rapidand robust model predictive error analysis as an adjunct to routine model-based management.In this thesis, the environmental models of interest are drawn from both the surface water andgroundwater hydrology disciplines. In calibrating such models, a set of parameters is assigned tothe model which will be employed for the making of all future predictions. If these parametersare estimated through solution of an inverse problem, formulated to be properly-posed througheither implicit or explicit parsimonising, then solution of this inverse problem produces asimplified parameter set that omits the details of reality, while achieving acceptable levels ofmodel-to-measurement misfit.There are a number of uncertainty methods which are applicable to a calibrated hydrologicmodel. This thesis initially explores three of the traditional approaches to parameter andpredictive uncertainty assessment by comparing their performance when applied to a lumpedparametermodel for surface-water flow in a large watershed; an equivalent model featuring asynthetic “measurement” dataset with noise of known stochastic characteristics added to these“measurements” was also utilised for robustness. The methods investigated are linear and nonlinearstatistical inference and Markov chain Monte Carlo. The most salient outcome of thiscomponent of the research is that each of these traditional methods of uncertainty analysis weredemonstrated to have limited application in circumstances of hydrologic extremes, a contexttypical of that required of models of this type.In most instances of model calibration, the majority of the noise associated with fieldmeasurements is, in fact, so-called “structural noise”, this being a general term that is used todescribe a model’s failure to represent every nuance of system behaviour. “Structural noise”arises from a number of sources. One of these is the imperfect nature of the model itself in itsnumerical simulation of real-world environmental processes. The other is the pre-calibrationmodel simplification required to achieve unique values for calibrated parameters. It seemsreasonable that reduction of this “structural noise” would ensure that less information is lost tothe parameter estimation process, thus instilling the calibrated model with a smaller predictiveerror variance than it would otherwise possess.Deployment of mathematical regularisation in the inversion process is one means of reducing theamount of parsimonising required to achieve uniqueness, due to the access it grants the model toan increased dimensionality of parameter space. A recently developed approach involving linearpredictive error analysis as an adjunct to regularised inversion, which accommodates the effectson potential predictive error of details that are beyond the ability of the parameter estimationprocess to capture, is utilised for the analysis of predictive error associated with a real-world,water resource groundwater management model. The analysis offers many challenges, includingthe fact that the model is a complex one that was partly calibrated by hand. Methods aredeveloped to allow additional parameters to be included in an already calibrated model in orderimprove computation of the contribution to predictive error variance made by hydraulic propertyand boundary condition heterogeneity uncaptured by the calibration process. For variouspredictions of interest, the error variance in every active cell of the model grid, or thecontributions made by different parameter types to the overall error variance, are able to becomputed expeditiously using this linear-based methodology.The focus of this research returns to a surface water model to see what could be derived fromregularised inversion principles previously applied to the groundwater problem, the aim being toexplore some of the problems that beset the more traditional calibration and uncertaintytechniques described in the first part of this study. Using linear regularisation concepts, amathematical expression is derived for lumping-induced parameter and predictive error variance,which although not easily reproducible, allows a vital issue to be addressed; namely, whetherobservation transformation and weighting methods which have been historically applied cangenerate observation noise that is characterised by a known statistical distribution and whichproduces maximum likelihood estimates of parameter values. The issue of just what “parametervalues” mean in this context is also explored. It is shown that parameter averaging introduces asignificant amount of potential error and that the parameters of a lumped model do notnecessarily represent the hydraulic properties which they purport to represent. Thus there is aprice to be paid in simplifying a model for the purpose of “living within our means” whencalibrated it. In the necessity to lump model parameters it is felt through the “structural noise”which lumping generates, and the parameter contamination that the lumping process incurs.As a result, it is posited that parameter bounds should not be too rigidly constrained at theexpense of improved fit to observation data. However, this issue is as yet unresolved. It is alsosuggested that while a model may not be as well calibrated in terms of model-to-measurement fit,use of parameter values which are in most accord with hydraulic expectations (in the event theyare known) may yield a reduced propensity for predictive error." @default.
• W156929291 created "2016-06-24" @default.
• W156929291 creator A5017810911 @default.
• W156929291 date "2008-10-01" @default.
• W156929291 modified "2023-09-27" @default.
• W156929291 title "Parameter and Predictive Uncertainty Analysis for Surface Water and Groundwater Flow Models" @default.
• W156929291 hasPublicationYear "2008" @default.
• W156929291 type Work @default.
• W156929291 sameAs 156929291 @default.
• W156929291 citedByCount "0" @default.
• W156929291 crossrefType "journal-article" @default.
• W156929291 hasAuthorship W156929291A5017810911 @default.
• W156929291 hasConcept C104317684 @default.
• W156929291 hasConcept C105795698 @default.
• W156929291 hasConcept C111350023 @default.
• W156929291 hasConcept C126255220 @default.
• W156929291 hasConcept C149782125 @default.
• W156929291 hasConcept C177803969 @default.
• W156929291 hasConcept C185592680 @default.
• W156929291 hasConcept C19499675 @default.
• W156929291 hasConcept C33923547 @default.
• W156929291 hasConcept C41008148 @default.
• W156929291 hasConcept C42475967 @default.
• W156929291 hasConcept C44154836 @default.
• W156929291 hasConcept C55493867 @default.
• W156929291 hasConcept C63479239 @default.
• W156929291 hasConceptScore W156929291C104317684 @default.
• W156929291 hasConceptScore W156929291C105795698 @default.
• W156929291 hasConceptScore W156929291C111350023 @default.
• W156929291 hasConceptScore W156929291C126255220 @default.
• W156929291 hasConceptScore W156929291C149782125 @default.
• W156929291 hasConceptScore W156929291C177803969 @default.
• W156929291 hasConceptScore W156929291C185592680 @default.
• W156929291 hasConceptScore W156929291C19499675 @default.
• W156929291 hasConceptScore W156929291C33923547 @default.
• W156929291 hasConceptScore W156929291C41008148 @default.
• W156929291 hasConceptScore W156929291C42475967 @default.
• W156929291 hasConceptScore W156929291C44154836 @default.
• W156929291 hasConceptScore W156929291C55493867 @default.
• W156929291 hasConceptScore W156929291C63479239 @default.
• W156929291 hasLocation W1569292911 @default.
• W156929291 hasOpenAccess W156929291 @default.
• W156929291 hasPrimaryLocation W1569292911 @default.
• W156929291 hasRelatedWork W1491564816 @default.
• W156929291 hasRelatedWork W2029073518 @default.
• W156929291 hasRelatedWork W2042553494 @default.
• W156929291 hasRelatedWork W2042647242 @default.
• W156929291 hasRelatedWork W2112444088 @default.
• W156929291 hasRelatedWork W2152932346 @default.
• W156929291 hasRelatedWork W2308452093 @default.
• W156929291 hasRelatedWork W2471732318 @default.
• W156929291 hasRelatedWork W2556558172 @default.
• W156929291 hasRelatedWork W2561272572 @default.
• W156929291 hasRelatedWork W2593353420 @default.
• W156929291 hasRelatedWork W2749152803 @default.
• W156929291 hasRelatedWork W2889837614 @default.
• W156929291 hasRelatedWork W2998618342 @default.
• W156929291 hasRelatedWork W3024946510 @default.
• W156929291 hasRelatedWork W3039601613 @default.
• W156929291 hasRelatedWork W610476797 @default.
• W156929291 hasRelatedWork W8534870 @default.
• W156929291 hasRelatedWork W87866747 @default.
• W156929291 hasRelatedWork W2328068301 @default.
• W156929291 isParatext "false" @default.
• W156929291 isRetracted "false" @default.
• W156929291 magId "156929291" @default.
• W156929291 workType "article" @default.