Forecasting climate variables using a mixed-effect state-space model

Abstract

This paper demonstrates the potential advantage of using a linear, mixed-effect state-space model for statistical downscaling of climate variables compared to the frequently used approach of linear regression. This comparison leads to the development of a method for estimation of model parameters using the EM algorithm approach. The model is applied to the prediction of temperature and rainfall statistics at both a sub-tropical and temperate location in Australia. The results indicate that for lead times of 1-10 years this state-space approach is able to predict observed seasonal temperature and rainfall means with substantially greater precision than climatology, multivariate linear regression (MLR) or a standard linear state-space (LSS) approach. The model is seen as a first step in the development of a short-term climate change projection system that will utilise both historical climate data as well as dynamically derived mean climate change projection information obtained from global climate models (GCMs).

Climate projections for Australia: a first glance at CMIP5

“…In fact, for the updated national projections a number of more sophisticated analysis methods are being considered, such as the 'Representative Climate Futures' approach (Whetton et al 2012), multimodel stratification based on an index of tropical ocean warming (Watterson 2012), pattern scaling (Mitchell 2003) and dynamical and statistical downscaling techniques (e.g. Grose et al in press;Kokic et al 2011). Consideration will also be given to issues including model independence (Jun et al 2008) and drift (Sen Gupta et al 2012), which were not considered here.…”

A Spatiodynamic Model for Assessing Frost Risk in South-Eastern Australia

“…(), page 1550). This has led some researchers to consider non‐linear approaches to represent an evolving climate system such as hidden Markov chain models (Charles et al ., ; Greene et al ., ) and state space models (Berliner et al ., ; Kokic et al ., ). In this paper, the development of a spatiodynamic model was deliberate to allow some of the coefficients to vary over both time and space, as dictated by the data.…”

Resilience of Cambodian lowland rice farming systems to future climate uncertainty

“…Evaluation of climate scenarios requires a reference 'baseline' period, that encompasses both temporal variation over the comparison climate period and is representative of present day climatology of the study region (IPCC, 1994). The baseline climate file for the region of Svay Rieng was generated from locally available observations from 1978 to 2011 for Tan Ninh in Vietnam (11.32 N, 106.1 E), using Linear Mixed-Effect State-Space (LMESS) statistical downscaling (Kokic et al, 2011). The LMESS model combines historic climate observations with downscaled climate projections from GCMs to produce effective point scale projections suitable as baseline and near-future (2021-2040) climates for use in biophysical modelling applications (Kokic et al, 2013).…”