Predicting Regolith Depth Using Bayesian Hierarchical Spatial Models

• Activity Number: 496
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Transportation Statistics Interest Group
• Author(s): Wen-Hsi Yang* and David Clifford and Ross Searle and John Wilford
• Companies: CSIRO and The Climate Corporation and CSIRO Land and Water Flagship and Geoscience Australia
• Keywords: Bayesian hierarchical spatial model ; regolith depth ; large spatial dataset ; censored data

Abstract

Regolith is a layer from the Earth's surface down to unweathered bedrock at depth. The regolith zone plays an important role in controlling hydrological and landscape processes which are essential for plant growth, aquifer recharge and landscape salinity. As such, understanding regolith depth is important for activities such as agriculture, forestry and mining. In Australia, regolith depth has been measured at thousands of locations in the state of Queensland. These measurements are typically derived from estimates of bedrock depth recorded when bores are dug in search of groundwater resources. Hence, some measurements may be right censored as groundwater may be encountered before bedrock is reached. In this study, we propose Bayesian hierarchical spatial models for large spatial and censored datasets. Importantly, our model extracts features of geographic environmental covariates to improve prediction rather than to explain variation of regolith depth. Finally, we illustrate the effectiveness of our approach through simulation and by applying our model to the Queensland dataset.