William Kleiber

Nonstationary modeling for multivariate spatial processes

We derive a class of matrix valued covariance functions where the direct and cross-covariance functions are Matern. The parameters of the Matern class are allowed to vary with location, yielding local variances, local ranges, local geometric anisotropies and local smoothnesses. We discuss inclusion of a nonconstant cross-correlation coefficient and a valid approximation. Estimation utilizes kernel smoothed empirical covariance matrices and a locally weighted minimum Frobenius distance that yields local parameter estimates at any location. We derive the asymptotic mean squared error of our kernel smoother and discuss the case when multiple field realizations are available. Finally, the model is illustrated on two datasets, one a synthetic bivariate one-dimensional spatial process, and the second a set of temperature and precipitation model output from a regional climate model.

Locally Calibrated Probabilistic Temperature Forecasting Using Geostatistical Model Averaging and Local Bayesian Model Averaging

AbstractThe authors introduce two ways to produce locally calibrated grid-based probabilistic forecasts of temperature. Both start from the Global Bayesian model averaging (Global BMA) statistical postprocessing method, which has constant predictive bias and variance across the domain, and modify it to make it local. The first local method, geostatistical model averaging (GMA), computes the predictive bias and variance at observation stations and interpolates them using a geostatistical model. The second approach, Local BMA, estimates the parameters of BMA at a grid point from stations that are close to the grid point and similar to it in elevation and land use. The results of these two methods applied to the eight-member University of Washington Mesoscale Ensemble (UWME) are given for the 2006 calendar year. GMA was calibrated and sharper than Global BMA, with prediction intervals that were 8% narrower than Global BMA on average. Examples using sparse and dense training networks of stations are shown. Th...

Geostatistical Model Averaging for Locally Calibrated Probabilistic Quantitative Precipitation Forecasting

Accurate weather benefit many key societal functions and activities, including agriculture, transportation, recreation, and basic human and infrastructural safety. Over the past two decades, ensembles of numerical weather prediction models have been developed, in which multiple estimates of the current state of the atmosphere are used to generate probabilistic forecasts for future weather events. However, ensemble systems are uncalibrated and biased, and thus need to be statistically postprocessed. Bayesian model averaging (BMA) is a preferred way of doing this. Particularly for quantitative precipitation, biases and calibration errors depend critically on local terrain features. We introduce a geostatistical approach to modeling locally varying BMA parameters, as opposed to the extant method that holds parameters constant across the forecast domain. Degeneracies caused by enduring dry periods are overcome by Bayesian regularization and Laplace approximations. The new approach, called geostatistical model averaging (GMA), was applied to 48-hour-ahead forecasts of daily precipitation accumulation over the North American Pacific Northwest, using the eight-member University of Washington Mesoscale Ensemble. GMA had better aggregate and local calibration than the extant technique, and was sharper on average.

A Bayesian kriging approach for blending satellite and ground precipitation observations

Drought and flood management practices require accurate estimates of precipitation. Gauge observations, however, are often sparse in regions with complicated terrain, clustered in valleys, and of poor quality. Consequently, the spatial extent of wet events is poorly represented. Satellite-derived precipitation data are an attractive alternative, though they tend to underestimate the magnitude of wet events due to their dependency on retrieval algorithms and the indirect relationship between satellite infrared observations and precipitation intensities. Here we offer a Bayesian kriging approach for blending precipitation gauge data and the Climate Hazards Group Infrared Precipitation satellite-derived precipitation estimates for Central America, Colombia, and Venezuela. First, the gauge observations are modeled as a linear function of satellite-derived estimates and any number of other variables—for this research we include elevation. Prior distributions are defined for all model parameters and the posterior distributions are obtained simultaneously via Markov chain Monte Carlo sampling. The posterior distributions of these parameters are required for spatial estimation, and thus are obtained prior to implementing the spatial kriging model. This functional framework is applied to model parameters obtained by sampling from the posterior distributions, and the residuals of the linear model are subject to a spatial kriging model. Consequently, the posterior distributions and uncertainties of the blended precipitation estimates are obtained. We demonstrate this method by applying it to pentadal and monthly total precipitation fields during 2009. The models performance and its inherent ability to capture wet events are investigated. We show that this blending method significantly improves upon the satellite-derived estimates and is also competitive in its ability to represent wet events. This procedure also provides a means to estimate a full conditional distribution of the “true” observed precipitation value at each grid cell.

Coupled Stochastic Weather Generation Using Spatial and Generalized Linear Models

We introduce a stochastic weather generator for the variables of minimum temperature, maximum temperature and precipitation occurrence. Temperature variables are modeled in vector autoregressive framework, conditional on precipitation occurrence. Precipitation occurrence arises via a probit model, and both temperature and occurrence are spatially correlated using spatial Gaussian processes. Additionally, local climate is included by spatially varying model coefficients, allowing spatially evolving relationships between variables. The method is illustrated on a network of stations in the Pampas region of Argentina where nonstationary relationships and historical spatial correlation challenge existing approaches.

Nonstationary matrix covariances: compact support, long range dependence and quasi-arithmetic constructions

Flexible models for multivariate processes are increasingly important for datasets in the geophysical, environmental, economics and health sciences. Modern datasets involve numerous variables observed at large numbers of space–time locations, with millions of data points being common. We develop a suite of stochastic models for nonstationary multivariate processes. The constructions break into three basic categories—quasi-arithmetic, locally stationary covariances with compact support, and locally stationary covariances with possible long-range dependence. All derived models are nonstationary, and we illustrate the flexibility of select choices through simulation.

Kriging and Local Polynomial Methods for Blending Satellite-Derived and Gauge Precipitation Estimates to Support Hydrologic Early Warning Systems

Robust estimates of precipitation in space and time are important for efficient natural resource management and for mitigating natural hazards. This is particularly true in regions with developing infrastructure and regions that are frequently exposed to extreme events. Gauge observations of rainfall are sparse but capture the precipitation process with high fidelity. Due to its high resolution and complete spatial coverage, satellite-derived rainfall data are an attractive alternative in data-sparse regions and are often used to support hydrometeorological early warning systems. Satellite-derived precipitation data, however, tend to underrepresent extreme precipitation events. Thus, it is often desirable to blend spatially extensive satellite-derived rainfall estimates with high-fidelity rain gauge observations to obtain more accurate precipitation estimates. In this research, we use two different methods, namely, ordinary kriging and κ-nearest neighbor local polynomials, to blend rain gauge observations with the Climate Hazards Group Infrared Precipitation satellite-derived precipitation estimates in data-sparse Central America and Colombia. The utility of these methods in producing blended precipitation estimates at pentadal (five-day) and monthly time scales is demonstrated. We find that these blending methods significantly improve the satellite-derived estimates and are competitive in their ability to capture extreme precipitation.

Spatial Bayesian hierarchical modeling of precipitation extremes over a large domain

We propose a Bayesian hierarchical model for spatial extremes on a large domain. In the data layer a Gaussian elliptical copula having generalized extreme value (GEV) marginals is applied. Spatial dependence in the GEV parameters is captured with a latent spatial regression with spatially varying coefficients. Using a composite likelihood approach, we are able to efficiently incorporate a large precipitation data set, which includes stations with missing data. The model is demonstrated by application to fall precipitation extremes at approximately 2600 stations covering the western United States, −125°E to −100°E longitude and 30°N–50°N latitude. The hierarchical model provides GEV parameters on a 1/8° grid and, consequently, maps of return levels and associated uncertainty. The model results indicate that return levels and their associated uncertainty have a well-defined spatial structure. Maps of return levels provide information about the spatial variations of the risk of extreme precipitation in the western US and is expected to be useful for infrastructure planning.

Parameter tuning for a multi-fidelity dynamical model of the magnetosphere

Geomagnetic storms play a critical role in space weather physics with the potential for far reaching economic impacts including power grid outages, air traffic rerouting, satellite damage and GPS disruption. The LFM-MIX is a state-of-the-art coupled magnetospheric-ionospheric model capable of simulating geomagnetic storms. Imbedded in this model are physical equations for turning the magnetohydrodynamic state parameters into energy and flux of electrons entering the ionosphere, involving a set of input parameters. The exact values of these input parameters in the model are unknown, and we seek to quantify the uncertainty about these parameters when model output is compared to observations. The model is available at different fidelities: a lower fidelity which is faster to run, and a higher fidelity but more computationally intense version. Model output and observational data are large spatiotemporal systems; the traditional design and analysis of computer experiments is unable to cope with such large data sets that involve multiple fidelities of model output. We develop an approach to this inverse problem for large spatiotemporal data sets that incorporates two different versions of the physical model. After an initial design, we propose a sequential design based on expected improvement. For the LFM-MIX, the additional run suggested by expected improvement diminishes posterior uncertainty by ruling out a posterior mode and shrinking the width of the posterior distribution. We also illustrate our approach using the Lorenz `96 system of equations for a simplified atmosphere, using known input parameters. For the Lorenz `96 system, after performing sequential runs based on expected improvement, the posterior mode converges to the true value and the posterior variability is reduced.

An anomaly in the isotopomer shift of the hyperfine spectrum of LiI

A high-precision examination of the hyperfine spectrum of 6LiI in comparison with 7LiI shows a shift in the iodine nuclear electric quadrupole moment that cannot be accounted for by a model in which the electric field gradient at the iodine site is assumed to depend only upon the internuclear distance between Li and I. The other hyperfine interactions are consistent between the two isotopomers, including the previously reported electric hexadecapole interaction of the iodine nucleus.