Highly parameterized model calibration with cloud computing: an example of regional flow model calibration in northeast Alberta, Canada

Expanding groundwater datasets collected by automated sensors, and improved groundwater databases, have caused a rapid increase in calibration data available for groundwater modeling projects. Improved methods of subsurface characterization have increased the need for model complexity to represent g...

Full description

Saved in:
Bibliographic Details
Published in:Hydrogeology journal 2014-05, Vol.22 (3), p.729-737
Main Authors: Hayley, Kevin, Schumacher, J., MacMillan, G. J., Boutin, L. C.
Format: Article
Language:English
Subjects:
Aquatic Pollution
Aquifers
Boundary conditions
Calibration
Cloud computing
Computation
Earth and Environmental Science
Earth Sciences
Fluid flow
Geology
Geophysics/Geodesy
Groundwater
Groundwater availability
Groundwater data
Groundwater flow
Hydraulics
Hydrogeology
Hydrology
Hydrology/Water Resources
Mathematical models
Parametrization
Piezometric head
Simulation
Technical Note
Waste Water Technology
Water Management
Water Pollution Control
Water Quality/Water Pollution
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Expanding groundwater datasets collected by automated sensors, and improved groundwater databases, have caused a rapid increase in calibration data available for groundwater modeling projects. Improved methods of subsurface characterization have increased the need for model complexity to represent geological and hydrogeological interpretations. The larger calibration datasets and the need for meaningful predictive uncertainty analysis have both increased the degree of parameterization necessary during model calibration. Due to these competing demands, modern groundwater modeling efforts require a massive degree of parallelization in order to remain computationally tractable. A methodology for the calibration of highly parameterized, computationally expensive models using the Amazon EC2 cloud computing service is presented. The calibration of a regional-scale model of groundwater flow in Alberta, Canada, is provided as an example. The model covers a 30,865-km 2 domain and includes 28 hydrostratigraphic units. Aquifer properties were calibrated to more than 1,500 static hydraulic head measurements and 10 years of measurements during industrial groundwater use. Three regionally extensive aquifers were parameterized (with spatially variable hydraulic conductivity fields), as was the aerial recharge boundary condition, leading to 450 adjustable parameters in total. The PEST-based model calibration was parallelized on up to 250 computing nodes located on Amazon’s EC2 servers.
ISSN:1431-2174
1435-0157
DOI:10.1007/s10040-014-1110-8