INLA-MRA: A Bayesian method for large spatiotemporal datasets
share
Large spatiotemporal datasets are a challenge for conventional Bayesian models, because of the cubic computational complexity of the algorithms for obtaining the Cholesky decomposition of the covariance matrix in the multivariate normal density. Moreover, standard numerical algorithms for posterior estimation, such as Markov Chain Monte Carlo (MCMC), are intractable in this context, as they require thousands, if not millions, of costly likelihood evaluations. To overcome those limitations, we propose INLA-MRA, a method that mixes an estimation algorithm inspired by INLA (Integrated Nested Laplace Approximation), and a model taking advantage of the sparse covariance structure produced by the Multi-Resolution Approximation (MRA) approach. INLA-MRA extends MRA to spatiotemporal data, while also facilitating the approximation of the hyperparameter marginal posterior distributions. We apply INLA-MRA to large MODIS Level 3 Land Surface Temperature (LST) datasets, sampled between May 18 and May 31, 2012 in the western part of the state of Maharashtra, India. We find that INLA-MRA can produce realistic prediction surfaces over regions where concentrated missingness, caused by sizable cloud cover, is observed. Through a validation analysis, we also find that predictions tend to be very accurate.
READ FULL TEXT
