Arthur M. Greene

A verification framework for interannual-to-decadal predictions experiments

Decadal predictions have a high profile in the climate science community and beyond, yet very little is known about their skill. Nor is there any agreed protocol for estimating their skill. This paper proposes a sound and coordinated framework for verification of decadal hindcast experiments. The framework is illustrated for decadal hindcasts tailored to meet the requirements and specifications of CMIP5 (Coupled Model Intercomparison Project phase 5). The chosen metrics address key questions about the information content in initialized decadal hindcasts. These questions are: (1) Do the initial conditions in the hindcasts lead to more accurate predictions of the climate, compared to un-initialized climate change projections? and (2) Is the prediction model’s ensemble spread an appropriate representation of forecast uncertainty on average? The first question is addressed through deterministic metrics that compare the initialized and uninitialized hindcasts. The second question is addressed through a probabilistic metric applied to the initialized hindcasts and comparing different ways to ascribe forecast uncertainty. Verification is advocated at smoothed regional scales that can illuminate broad areas of predictability, as well as at the grid scale, since many users of the decadal prediction experiments who feed the climate data into applications or decision models will use the data at grid scale, or downscale it to even higher resolution. An overall statement on skill of CMIP5 decadal hindcasts is not the aim of this paper. The results presented are only illustrative of the framework, which would enable such studies. However, broad conclusions that are beginning to emerge from the CMIP5 results include (1) Most predictability at the interannual-to-decadal scale, relative to climatological averages, comes from external forcing, particularly for temperature; (2) though moderate, additional skill is added by the initial conditions over what is imparted by external forcing alone; however, the impact of initialization may result in overall worse predictions in some regions than provided by uninitialized climate change projections; (3) limited hindcast records and the dearth of climate-quality observational data impede our ability to quantify expected skill as well as model biases; and (4) as is common to seasonal-to-interannual model predictions, the spread of the ensemble members is not necessarily a good representation of forecast uncertainty. The authors recommend that this framework be adopted to serve as a starting point to compare prediction quality across prediction systems. The framework can provide a baseline against which future improvements can be quantified. The framework also provides guidance on the use of these model predictions, which differ in fundamental ways from the climate change projections that much of the community has become familiar with, including adjustment of mean and conditional biases, and consideration of how to best approach forecast uncertainty.

Distinguishing the Roles of Natural and Anthropogenically Forced Decadal Climate Variability: Implications for Prediction

Abstract Given that over the course of the next 10–30 years the magnitude of natural decadal variations may rival that of anthropogenically forced climate change on regional scales, it is envisioned that initialized decadal predictions will provide important information for climate-related management and adaptation decisions. Such predictions are presently one of the grand challenges for the climate community. This requires identifying those physical phenomena—and their model equivalents—that may provide additional predictability on decadal time scales, including an assessment of the physical processes through which anthropogenic forcing may interact with or project upon natural variability. Such a physical framework is necessary to provide a consistent assessment (and insight into potential improvement) of the decadal prediction experiments planned to be assessed as part of the IPCCs Fifth Assessment Report.

Probabilistic Multimodel Regional Temperature Change Projections

Abstract Regional temperature change projections for the twenty-first century are generated using a multimodel ensemble of atmosphere–ocean general circulation models. The models are assigned coefficients jointly, using a Bayesian linear model fitted to regional observations and simulations of the climate of the twentieth century. Probability models with varying degrees of complexity are explored, and a selection is made based on Bayesian deviance statistics, coefficient properties, and a classical cross-validation measure utilizing temporally averaged data. The model selected is shown to be superior in predictive skill to a naive model consisting of the unweighted mean of the underlying atmosphere–ocean GCM (AOGCM) simulations, although the skill differential varies regionally. Temperature projections for the A2 and B1 scenarios from the Intergovernmental Panel on Climate Change (IPCC) Special Report on Emissions Scenarios are presented.

Web tool deconstructs variability in twentieth‐century climate

Climate, as experienced over time in a particular locality or region, exhibits a wide spectrum of variability, ranging from daily weather to century-scale trends and beyond. As a result, climate-related risks will tend to vary as well, with slower variations modulating the likelihood of adverse or beneficial events that play out on shorter time scales. For example, the risk of crop loss due to insufficient seasonal rainfall will generally be increased during decades that themselves are drier than normal, and vice versa. Short-term planning for such risk might involve securing adequate supplies of a drought-resistant cultivar, while on longer time scales such infrastructure decisions as the design of irrigation systems or reservoirs may be at issue. An understanding of how variations on different time scales have combined to produce observed climate histories can inform adaptation or risk mitigation strategies in light of such multi-time-scale risk variability.

A time constant for hemispheric glacier mass balance

The notion is developed of a mass-balance time constant applicable to the Northern Hemispheric glacier inventory taken as a whole. Ice dynamics are incorporated only implicitly in its estimation, which follows directly from a consideration of observed mass-balance and hemispheric temperature time series. While such a parameter must certainly be related to the rate at which glacier hypsometry adjusts to variations in climate, as are time constants derived via dynamic considerations, the parameter discussed herein differs with respect to its statistical character. For an ensemble of Northern Hemisphere glaciers a time-constant value on the order of a century is estimated. It is shown that such a value is consistent with the hemispheric near-equilibration of glaciers that prevailed around 1970. A ‘reference climate’ is defined, such that the mass balance in a given year is a function only of the difference between that year’s climate and the reference. This difference was small during the hemispheric near-equilibrium that prevailed around 1970, implying that the glacier wastage of the late 20th century is essentially a response to post-1970 warming. It is shown that precipitation fluctuations play a compensating role in the hemispheric net mass budget, in that they are strongly anticorrelated with fluctuations in temperature-induced melting. However, the contribution of precipitation does not override that of temperature, which remains the dominant influence on hemisphere-wide glacier fluctuations.

A Bayesian Hidden Markov Model of Daily Precipitation over South and East Asia

AbstractA Bayesian hidden Markov model (HMM) for climate downscaling of multisite daily precipitation is presented. A generalized linear model (GLM) component allows exogenous variables to directly influence the distributional characteristics of precipitation at each site over time, while the Markovian transitions between discrete states represent seasonality and subseasonal weather variability. Model performance is evaluated for station networks of summer rainfall over the Punjab region in northern India and Pakistan and the upper Yangtze River basin in south-central China. The model captures seasonality and the marginal daily distributions well in both regions. Extremes are reproduced relatively well in the Punjab region, but underestimated for the Yangtze. In terms of interannual variability, the combined GLM–HMM with spatiotemporal averages of observed rainfall as a predictor is shown to exhibit skill (in terms of reduced RMSE) at the station level, particularly for the Punjab region. The skill is lar...

Bayesian nonhomogeneous Markov models via Pólya-Gamma data augmentation with applications to rainfall modeling

Discrete-time hidden Markov models are a broadly useful class of latent-variable models with applications in areas such as speech recognition, bioinformatics, and climate data analysis. It is common in practice to introduce temporal non-homogeneity into such models by making the transition probabilities dependent on time-varying exogenous input variables via a multinomial logistic parametrization. We extend such models to introduce additional non-homogeneity into the emission distribution using a generalized linear model (GLM), with data augmentation for sampling-based inference. However, the presence of the logistic function in the state transition model significantly complicates parameter inference for the overall model, particularly in a Bayesian context. To address this we extend the recently-proposed Polya-Gamma data augmentation approach to handle non-homogeneous hidden Markov models (NHMMs), allowing the development of an efficient Markov chain Monte Carlo (MCMC) sampling scheme. We apply our model and inference scheme to 30 years of daily rainfall in India, leading to a number of insights into rainfall-related phenomena in the region. Our proposed approach allows for fully Bayesian analysis of relatively complex NHMMs on a scale that was not possible with previous methods. Software implementing the methods described in the paper is available via the R package NHMM.

A Bayesian Multivariate Nonhomogeneous Markov Model

We present a Bayesian scheme for the downscaling of daily rainfall over a network of stations. Rainfall is modeled locally as a state-dependent mixture, with the states progressing in time as a first-order Markov process. The Markovian transition matrix, as well as the local state distributions, are dependent on exogenous covariates via generalized linear models (GLMs). The methodology is applied to a large network of stations spanning the Indian subcontinent and extending into the proximal Himalaya. The combined GLM-NHMM approach offers considerable flexibility and can also be applied to maximum and minimum temperatures. The modeling framework has been made available in the NHMM package for the R programming language.