Keywords
entity relationship model, optimization, logging, parameterization
Start Date
1-7-2012 12:00 AM
Abstract
The Catchment Water Yield Estimation Toolset (CWYET) is a softwaretoolset for estimating daily catchment water yield and runoff characteristics inregulated and unregulated catchments. It is used to estimate water yield over up tohundreds of catchments, featuring capabilities for calibration, catchment cross verification, ensembles of models, and scenario modelling such as impact ofclimate change. Due to the combinatorial nature of the matrix of these ensembles,using simplistic text files to store model parameterization can become at the veryleast logistically tedious. Of more concern, this is a brittle storage system that isinadequate to underpin provenance tracking and reproducibility. The issue is notunique to CWYET, and there are substantial efforts in modelling software productsto use state of the art Object Relational Model (ORM) tools such as NHibernate topersist model structure and parameterisation. In this paper we present how weused the Microsoft Entity Framework version 4.1 to implement a database schemato store and manage a large number of model parameterizations. We summarisethe main use cases for these model parameterisations. Importantly, we strive for adata store that is decoupled from a particular modelling framework or tool, and notlimited to CWYET. We derive the schema of the database from the characteristicsof the results of optimization tools, and the information that is determined asnecessary from the use cases. We illustrate how the library of optimization resultsis accessed to assess visually the performance of model calibration on a largenumber of catchments. We demonstrate the use of this repository of parametersets from IronPython and from the scientific workflow Hydrologist’s Workbench.
A data model for a sustainable management of parameter sets and optimization results
The Catchment Water Yield Estimation Toolset (CWYET) is a softwaretoolset for estimating daily catchment water yield and runoff characteristics inregulated and unregulated catchments. It is used to estimate water yield over up tohundreds of catchments, featuring capabilities for calibration, catchment cross verification, ensembles of models, and scenario modelling such as impact ofclimate change. Due to the combinatorial nature of the matrix of these ensembles,using simplistic text files to store model parameterization can become at the veryleast logistically tedious. Of more concern, this is a brittle storage system that isinadequate to underpin provenance tracking and reproducibility. The issue is notunique to CWYET, and there are substantial efforts in modelling software productsto use state of the art Object Relational Model (ORM) tools such as NHibernate topersist model structure and parameterisation. In this paper we present how weused the Microsoft Entity Framework version 4.1 to implement a database schemato store and manage a large number of model parameterizations. We summarisethe main use cases for these model parameterisations. Importantly, we strive for adata store that is decoupled from a particular modelling framework or tool, and notlimited to CWYET. We derive the schema of the database from the characteristicsof the results of optimization tools, and the information that is determined asnecessary from the use cases. We illustrate how the library of optimization resultsis accessed to assess visually the performance of model calibration on a largenumber of catchments. We demonstrate the use of this repository of parametersets from IronPython and from the scientific workflow Hydrologist’s Workbench.
