Selecting from amongst non–nested conditional variance models: information criteria and portfolio determination

We consider the finite sample properties of model selection by information criteria in conditionally heteroscedastic models. Recent theoretical results show that certain popular criteria are consistent in that they will select the true model asymptotically with probability 1. To examine the empirical relevance of this property, Monte Carlo simulations are conducted for a set of non–nested data generating processes (DGPs) with the set of candidate models consisting of all types of model used as DGPs. In addition, not only is the best model considered but also those with similar values of the information criterion, called close competitors, thus forming a portfolio of eligible models. To supplement the simulations, the criteria are applied to a set of economic and financial series. In the simulations, the criteria are largely ineffective at identifying the correct model, either as best or a close competitor, the parsimonious GARCH(1, 1) model being preferred for most DGPs. In contrast, asymmetric models are generally selected to represent actual data. This leads to the conjecture that the properties of parameterizations of processes commonly used to model heteroscedastic data are more similar than may be imagined and that more attention needs to be paid to the behaviour of the standardized disturbances of such models, both in simulation exercises and in empirical modelling.

Hydrological cycle in the Danube basin in present-day and XXII century simulations by IPCCAR4 global climate models

We present an intercomparison and verification analysis of 20 GCMs (Global Circulation Models) included in the 4th IPCC assessment report regarding their representation of the hydrological cycle on the Danube river basin for 1961–2000 and for the 2161–2200 SRESA1B scenario runs. The basin-scale properties of the hydrological cycle are computed by spatially integrating the precipitation, evaporation, and runoff fields using the Voronoi-Thiessen tessellation formalism. The span of the model- simulated mean annual water balances is of the same order of magnitude of the observed Danube discharge of the Delta; the true value is within the range simulated by the models. Some land components seem to have deficiencies since there are cases of violation of water conservation when annual means are considered. The overall performance and the degree of agreement of the GCMs are comparable to those of the RCMs (Regional Climate Models) analyzed in a previous work, in spite of the much higher resolution and common nesting of the RCMs. The reanalyses are shown to feature several inconsistencies and cannot be used as a verification benchmark for the hydrological cycle in the Danubian region. In the scenario runs, for basically all models the water balance decreases, whereas its interannual variability increases. Changes in the strength of the hydrological cycle are not consistent among models: it is confirmed that capturing the impact of climate change on the hydrological cycle is not an easy task over land areas. Moreover, in several cases we find that qualitatively different behaviors emerge among the models: the ensemble mean does not represent any sort of average model, and often it falls between the models’ clusters.

Potential Predictability and AMIP Implications of Midlatitude Climate Variability in Two General Circulation Models

Ensembles of extended Atmospheric Model Intercomparison Project (AMIP) runs from the general circulation models of the National Centers for Environmental Prediction (formerly the National Meteorological Center) and the Max-Planck Institute (Hamburg, Germany) are used to estimate the potential predictability (PP) of an index of the Pacific–North America (PNA) mode of climate change. The PP of this pattern in “perfect” prediction experiments is 20%–25% of the index’s variance. The models, particularly that from MPI, capture virtually all of this variance in their hindcasts of the winter PNA for the period 1970–93. The high levels of internally generated model noise in the PNA simulations reconfirm the need for an ensemble averaging approach to climate prediction. This means that the forecasts ought to be expressed in a probabilistic manner. It is shown that the models’ skills are higher by about 50% during strong SST events in the tropical Pacific, so the probabilistic forecasts need to be conditional on the tropical SST. Taken together with earlier studies, the present results suggest that the original set of AMIP integrations (single 10-yr runs) is not adequate to reliably test the participating models’ simulations of interannual climate variability in the midlatitudes.

Northern Hemisphere atmospheric blocking as simulated by 15 atmospheric general circulation models in the period 1979-1988

As a part of the Atmospheric Model Intercomparison Project (AMIP), the behaviour of 15 general circulation models has been analysed in order to diagnose and compare the ability of the different models in simulating Northern Hemisphere midlatitude atmospheric blocking. In accordance with the established AMIP procedure, the 10-year model integrations were performed using prescribed, time-evolving monthly mean observed SSTs spanning the period January 1979–December 1988. Atmospheric observational data (ECMWF analyses) over the same period have been also used to verify the models results. The models involved in this comparison represent a wide spectrum of model complexity, with different horizontal and vertical resolution, numerical techniques and physical parametrizations, and exhibit large differences in blocking behaviour. Nevertheless, a few common features can be found, such as the general tendency to underestimate both blocking frequency and the average duration of blocks. The problem of the possible relationship between model blocking and model systematic errors has also been assessed, although without resorting to ad-hoc numerical experimentation it is impossible to relate with certainty particular model deficiencies in representing blocking to precise parts of the model formulation.

Northern Hemisphere atmospheric blocking as simulated by 15 atmospheric general circulation models in the period 1979-1988 (Results from an AMIP diagnostic subproject)

As a part of the Atmospheric Model Intercomparison Project (AMIP), the behaviour of 15 general circulation models has been analysed in order to diagnose and compare the ability of the different models in simulating Northern Hemisphere midlatitude atmospheric blocking. In accordance with the established AMIP procedure, the 10-year model integrations were performed using prescribed, time-evolving monthly mean observed SSTs spanning the period January 1979–December 1988. Atmospheric observational data (ECMWF analyses) over the same period have been also used to verify the models results. The models involved in this comparison represent a wide spectrum of model complexity, with different horizontal and vertical resolution, numerical techniques and physical parametrizations, and exhibit large differences in blocking behaviour. Nevertheless, a few common features can be found, such as the general tendency to underestimate both blocking frequency and the average duration of blocks. The problem of the possible relationship between model blocking and model systematic errors has also been assessed, although without resorting to ad-hoc numerical experimentation it is impossible to relate with certainty particular model deficiencies in representing blocking to precise parts of the model formulation.

Forecasting exchange rate volatility using conditional variance models selected by information criteria

This paper uses appropriately modified information criteria to select models from the GARCH family, which are subsequently used for predicting US dollar exchange rate return volatility. The out of sample forecast accuracy of models chosen in this manner compares favourably on mean absolute error grounds, although less favourably on mean squared error grounds, with those generated by the commonly used GARCH(1, 1) model. An examination of the orders of models selected by the criteria reveals that (1, 1) models are typically selected less than 20% of the time.

Understanding El Niño in ocean-atmosphere general circulation models: progress and challenges

Determining how El Niño and its impacts may change over the next 10 to 100 years remains a difficult scientific challenge. Ocean–atmosphere coupled general circulation models (CGCMs) are routinely used both to analyze El Niño mechanisms and teleconnections and to predict its evolution on a broad range of time scales, from seasonal to centennial. The ability to simulate El Niño as an emergent property of these models has largely improved over the last few years. Nevertheless, the diversity of model simulations of present-day El Niño indicates current limitations in our ability to model this climate phenomenon and to anticipate changes in its characteristics. A review of the several factors that contribute to this diversity, as well as potential means to improve the simulation of El Niño, is presented.

A review of currently available in-stream water-quality models and their applicability for simulating dissolved oxygen in lowland rivers

In this paper, a review is undertaken of the major models currently in use for describing water quality in freshwater river systems. The number of existing models is large because the various studies of water quality in rivers around the world have often resulted in the construction of new 'bespoke' models designed for the particular situation of that study. However, it is worth considering models that are already available, since an existing model, suitable for the purposes of the study, will save a great deal of work and may already have been established within regulatory and legal frameworks. The models chosen here are SIMCAT, TOMCAT, QUAL2E, QUASAR, MIKE-11 and ISIS, and the potential for each model is examined in relation to the issue of simulating dissolved oxygen (DO) in lowland rivers. These models have been developed for particular purposes and this review shows that no one model can provide all of the functionality required. Furthermore, all of the models contain assumptions and limitations that need to be understood if meaningful interpretations of the model simulations are to. be made. The work is concluded with the view that it is unfair to set one model against another in terms of broad applicability, but that a model of intermediate complexity, such as QUASAR, is generally well suited to simulate DO in river systems. (C) 2003 Elsevier Science B.V. All rights reserved.

Spatial analysis of the error in a model of soil nitrogen

Models of the dynamics of nitrogen in soil (soil-N) can be used to aid the fertilizer management of a crop. The predictions of soil-N models can be validated by comparison with observed data. Validation generally involves calculating non-spatial statistics of the observations and predictions, such as their means, their mean squared-difference, and their correlation. However, when the model predictions are spatially distributed across a landscape the model requires validation with spatial statistics. There are three reasons for this: (i) the model may be more or less successful at reproducing the variance of the observations at different spatial scales; (ii) the correlation of the predictions with the observations may be different at different spatial scales; (iii) the spatial pattern of model error may be informative. In this study we used a model, parameterized with spatially variable input information about the soil, to predict the mineral-N content of soil in an arable field, and compared the results with observed data. We validated the performance of the N model spatially with a linear mixed model of the observations and model predictions, estimated by residual maximum likelihood. This novel approach allowed us to describe the joint variation of the observations and predictions as: (i) independent random variation that occurred at a fine spatial scale; (ii) correlated random variation that occurred at a coarse spatial scale; (iii) systematic variation associated with a spatial trend. The linear mixed model revealed that, in general, the performance of the N model changed depending on the spatial scale of interest. At the scales associated with random variation, the N model underestimated the variance of the observations, and the predictions were correlated poorly with the observations. At the scale of the trend, the predictions and observations shared a common surface. The spatial pattern of the error of the N model suggested that the observations were affected by the local soil condition, but this was not accounted for by the N model. In summary, the N model would be well-suited to field-scale management of soil nitrogen, but suited poorly to management at finer spatial scales. This information was not apparent with a non-spatial validation. (c),2007 Elsevier B.V. All rights reserved.

Elimination of the Greenland Ice Sheet in a High CO2 Climate

Projections of future global sea level depend on reliable estimates of changes in the size of polar ice sheets. Calculating this directly from global general circulation models (GCMs) is unreliable because the coarse resolution of 100 km or more is unable to capture narrow ablation zones, and ice dynamics is not usually taken into account in GCMs. To overcome these problems a high-resolution (20 km) dynamic ice sheet model has been coupled to the third Hadley Centre Coupled Ocean-Atmosphere GCM (HadCM3). A novel feature is the use of two-way coupling, so that climate changes in the GCM drive ice mass changes in the ice sheet model that, in turn, can alter the future climate through changes in orography, surface albedo, and freshwater input to the model ocean. At the start of the main experiment the atmospheric carbon dioxide concentration was increased to 4 times the preindustrial level and held constant for 3000 yr. By the end of this period the Greenland ice sheet is almost completely ablated and has made a direct contribution of approximately 7 m to global average sea level, causing a peak rate of sea level rise of 5 mm yr-1 early in the simulation. The effect of ice sheet depletion on global and regional climate has been examined and it was found that apart from the sea level rise, the long-term effect on global climate is small. However, there are some significant regional climate changes that appear to have reduced the rate at which the ice sheet ablates.

The physical properties of the atmosphere in the new Hadley Centre Global Environmental Model (HadGEM1). Part II: Aspects of variability and regional climate

The performance of the atmospheric component of the new Hadley Centre Global Environmental Model (HadGEM1) is assessed in terms of its ability to represent a selection of key aspects of variability in the Tropics and extratropics. These include midlatitude storm tracks and blocking activity, synoptic variability over Europe, and the North Atlantic Oscillation together with tropical convection, the Madden-Julian oscillation, and the Asian summer monsoon. Comparisons with the previous model, the Third Hadley Centre Coupled Ocean-Atmosphere GCM (HadCM3), demonstrate that there has been a considerable increase in the transient eddy kinetic energy (EKE), bringing HadGEM1 into closer agreement with current reanalyses. This increase in EKE results from the increased horizontal resolution and, in combination with the improved physical parameterizations, leads to improvements in the representation of Northern Hemisphere storm tracks and blocking. The simulation of synoptic weather regimes over Europe is also greatly improved compared to HadCM3, again due to both increased resolution and other model developments. The variability of convection in the equatorial region is generally stronger and closer to observations than in HadCM3. There is, however, still limited convective variance coincident with several of the observed equatorial wave modes. Simulation of the Madden-Julian oscillation is improved in HadGEM1: both the activity and interannual variability are increased and the eastward propagation, although slower than observed, is much better simulated. While some aspects of the climatology of the Asian summer monsoon are improved in HadGEM1, the upper-level winds are too weak and the simulation of precipitation deteriorates. The dominant modes of monsoon interannual variability are similar in the two models, although in HadCM3 this is linked to SST forcing, while in HadGEM1 internal variability dominates. Overall, analysis of the phenomena considered here indicates that HadGEM1 performs well and, in many important respects, improves upon HadCM3. Together with the improved representation of the mean climate, this improvement in the simulation of atmospheric variability suggests that HadGEM1 provides a sound basis for future studies of climate and climate change.

Factors influencing anthropogenic carbon dioxide uptake in the North Atlantic in models of the ocean carbon cycle

The uptake and storage of anthropogenic carbon in the North Atlantic is investigated using different configurations of ocean general circulation/carbon cycle models. We investigate how different representations of the ocean physics in the models, which represent the range of models currently in use, affect the evolution of CO2 uptake in the North Atlantic. The buffer effect of the ocean carbon system would be expected to reduce ocean CO2 uptake as the ocean absorbs increasing amounts of CO2. We find that the strength of the buffer effect is very dependent on the model ocean state, as it affects both the magnitude and timing of the changes in uptake. The timescale over which uptake of CO2 in the North Atlantic drops to below preindustrial levels is particularly sensitive to the ocean state which sets the degree of buffering; it is less sensitive to the choice of atmospheric CO2 forcing scenario. Neglecting physical climate change effects, North Atlantic CO2 uptake drops below preindustrial levels between 50 and 300 years after stabilisation of atmospheric CO2 in different model configurations. Storage of anthropogenic carbon in the North Atlantic varies much less among the different model configurations, as differences in ocean transport of dissolved inorganic carbon and uptake of CO2 compensate each other. This supports the idea that measured inventories of anthropogenic carbon in the real ocean cannot be used to constrain the surface uptake. Including physical climate change effects reduces anthropogenic CO2 uptake and storage in the North Atlantic further, due to the combined effects of surface warming, increased freshwater input, and a slowdown of the meridional overturning circulation. The timescale over which North Atlantic CO2 uptake drops to below preindustrial levels is reduced by about one-third, leading to an estimate of this timescale for the real world of about 50 years after the stabilisation of atmospheric CO2. In the climate change experiment, a shallowing of the mixed layer depths in the North Atlantic results in a significant reduction in primary production, reducing the potential role for biology in drawing down anthropogenic CO2.

Quantifying mesoscale variability in ocean transient tracer fields

A powerful way to test the realism of ocean general circulation models is to systematically compare observations of passive tracer concentration with model predictions. The general circulation models used in this way cannot resolve a full range of vigorous mesoscale activity (on length scales between 10–100 km). In the real ocean, however, this activity causes important variability in tracer fields. Thus, in order to rationally compare tracer observations with model predictions these unresolved fluctuations (the model variability error) must be estimated. We have analyzed this variability using an eddy‐resolving reduced‐gravity model in a simple midlatitude double‐gyre configuration. We find that the wave number spectrum of tracer variance is only weakly sensitive to the distribution of (large scale slowly varying) tracer sources and sinks. This suggests that a universal passive tracer spectrum may exist in the ocean. We estimate the spectral shape using high‐resolution measurements of potential temperature on an isopycnal in the upper northeast Atlantic Ocean, finding a slope near k −1.7 between 10 and 500 km. The typical magnitude of the variance is estimated by comparing tracer simulations using different resolutions. For CFC‐ and tritium‐type transient tracers the peak magnitude of the model variability saturation error may reach 0.20 for scales shorter than 100 km. This is of the same order as the time mean saturation itself and well over an order of magnitude greater than the instrumental uncertainty.

Identification of nonlinear systems using generalized kernel models

Nonlinear system identification is considered using a generalized kernel regression model. Unlike the standard kernel model, which employs a fixed common variance for all the kernel regressors, each kernel regressor in the generalized kernel model has an individually tuned diagonal covariance matrix that is determined by maximizing the correlation between the training data and the regressor using a repeated guided random search based on boosting optimization. An efficient construction algorithm based on orthogonal forward regression with leave-one-out (LOO) test statistic and local regularization (LR) is then used to select a parsimonious generalized kernel regression model from the resulting full regression matrix. The proposed modeling algorithm is fully automatic and the user is not required to specify any criterion to terminate the construction procedure. Experimental results involving two real data sets demonstrate the effectiveness of the proposed nonlinear system identification approach.

Baroclinic stationary waves in aquaplanet models

An aquaplanet model is used to study the nature of the highly persistent low-frequency waves that have been observed in models forced by zonally symmetric boundary conditions. Using the Hayashi spectral analysis of the extratropical waves, the authors find that a quasi-stationary wave 5 belongs to a wave packet obeying a well-defined dispersion relation with eastward group velocity. The components of the dispersion relation with k ≥ 5 baroclinically convert eddy available potential energy into eddy kinetic energy, whereas those with k < 5 are baroclinically neutral. In agreement with Green’s model of baroclinic instability, wave 5 is weakly unstable, and the inverse energy cascade, which had been previously proposed as a main forcing for this type of wave, only acts as a positive feedback on its predominantly baroclinic energetics. The quasi-stationary wave is reinforced by a phase lock to an analogous pattern in the tropical convection, which provides further amplification to the wave. It is also found that the Pedlosky bounds on the phase speed of unstable waves provide guidance in explaining the latitudinal structure of the energy conversion, which is shown to be more enhanced where the zonal westerly surface wind is weaker. The wave’s energy is then trapped in the waveguide created by the upper tropospheric jet stream. In agreement with Green’s theory, as the equator-to-pole SST difference is reduced, the stationary marginally stable component shifts toward higher wavenumbers, while wave 5 becomes neutral and westward propagating. Some properties of the aquaplanet quasi-stationary waves are found to be in interesting agreement with a low frequency wave observed by Salby during December–February in the Southern Hemisphere so that this perspective on low frequency variability, apart from its value in terms of basic geophysical fluid dynamics, might be of specific interest for studying the earth’s atmosphere.