Impact of the Interdecadal Pacific Oscillation on Tropical Cyclone Activity in the North Atlantic and Eastern North Pacific.

Tropical cyclones (TCs) are among the most devastating weather systems affecting the United States and Central America (USCA). Here we show that the Interdecadal Pacific Oscillation (IPO) strongly modulates TC activity over the North Atlantic (NA) and eastern North Pacific (eNP). During positive IPO phases, less (more) TCs were observed over NA (eNP), likely due to the presence of stronger (weaker) vertical wind shear and the resulting changes in genesis potential. Furthermore, TCs over NA tend to keep their tracks more eastward and recurve at lower latitudes during positive IPO phases. Such variations are largely determined by changes in steering flow instead of changes in genesis locations. Over the eNP, smaller track variations are observed at different IPO phases with stable, westward movements of TCs prevailing. These findings have substantial implications for understanding decadal to inter-decadal fluctuations in the risk of TC landfalls along USCA coasts.

Adaptive Mixture Modelling Metropolis Methods for Bayesian Analysis of Non-linear State-Space Models.

We describe a strategy for Markov chain Monte Carlo analysis of non-linear, non-Gaussian state-space models involving batch analysis for inference on dynamic, latent state variables and fixed model parameters. The key innovation is a Metropolis-Hastings method for the time series of state variables based on sequential approximation of filtering and smoothing densities using normal mixtures. These mixtures are propagated through the non-linearities using an accurate, local mixture approximation method, and we use a regenerating procedure to deal with potential degeneracy of mixture components. This provides accurate, direct approximations to sequential filtering and retrospective smoothing distributions, and hence a useful construction of global Metropolis proposal distributions for simulation of posteriors for the set of states. This analysis is embedded within a Gibbs sampler to include uncertain fixed parameters. We give an example motivated by an application in systems biology. Supplemental materials provide an example based on a stochastic volatility model as well as MATLAB code.

THE STEAM-JET AEROSOL COLLECTOR

A new principle of sampling aerosol particles by means of steam injection with the consequent collection of grown droplets has been established. An air stream free of water-soluble gases is rapidly mixed with steam. The resulting supersaturation causes aerosol particles to grow into droplets. The droplets containing dissolved aerosol species are then collected by two cyclones in series. The solution collected in the cyclones is constantly pumped out and can be on- or off-line analysed by means of ion chromatography or flow injection analysis. On the basis of the new sampling principle a prototype of an aerosol sampler was designed which is capable of sampling particles quantitatively down to several nanometres in diameter. The mass sampling efficiency of the instrument was found to be 99\%. The detection limit of the sampler for ammonium, sulphate, nitrate and chloride ions is below 0.7 mu g m(-3). By reduction of an already identified source of contamination, much lower detection limits can be achieved. During measurements the sampler proved to be stable, working without any assistance for extended periods of time. Comparison of the sampler with filter packs during measurements of ambient air aerosols showed that the sampler gives good results.

Nested dictionary learning for hierarchical organization of imagery and text

A tree-based dictionary learning model is developed for joint analysis of imagery and associated text. The dictionary learning may be applied directly to the imagery from patches, or to general feature vectors extracted from patches or superpixels (using any existing method for image feature extraction). Each image is associated with a path through the tree (from root to a leaf), and each of the multiple patches in a given image is associated with one node in that path. Nodes near the tree root are shared between multiple paths, representing image characteristics that are common among different types of images. Moving toward the leaves, nodes become specialized, representing details in image classes. If available, words (text) are also jointly modeled, with a path-dependent probability over words. The tree structure is inferred via a nested Dirichlet process, and a retrospective stick-breaking sampler is used to infer the tree depth and width.