A Space-Time Emulator for Generating Hydrological Measures for Australian Catchments

• Activity Number: 182
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Daniel Gladish* and Petra Kuhnert and Dan Pagendam
• Companies: CSIRO and CSIRO and CSIRO
• Keywords: Hydrology ; Physical-statistical ; Computer models ; Spatio-temporal ; Emulator

Abstract

Modeling hydrological processes across varying landscapes and temporal resolutions can be challenging to quantify with any confidence. Previous attempts have resulted in the development of deterministic models that attempt to capture the spatial characteristics of the landscapes modeled as well as the temporal dynamics of the hydrological systems moving through the catchments. These models can be quite complex, require input from a large number of unknown parameters that are difficult to estimate and by nature are computationally intensive, requiring considerable computational time. The challenge therefore, is constructing a methodology that can estimate these parameters in addition to quantifying the associated uncertainty to determine which parameters in the model are important.

Statistical emulators have become more prominent in the literature as a mechanism for approximating large, complex and computationally intensive computer models. We develop a spatio-temporal statistical emulator to approximate a computer model that accounts for fluxes in water balance across Australian catchments and provide advice to hydrological modelers on which parameters of the model are important.