James Denholm-Price

A Climatological Study of Internal Gravity Waves in the Atmospheric Boundary Layer Overlying the Brunt Ice Shelf, Antarctica

Internal gravity waves are frequently observed in stably stratified regions of the atmospheric boundary layer. In order to determine the statistical influence of such waves on the dynamics of the boundary layer it is necessary to compile information concerning properties of the waves such as frequency of occurrence, propagation, and spectral characteristics. Gravity wave climatologies have been compiled from relatively few locations. In this paper a climatological study of gravity waves, in the period range 1‐20 min, propagating in the stably stratified atmospheric boundary layer overlying an Antarctic ice shelf is presented. An extensive set of boundary layer measurements were compiled throughout 1991. Surface pressure fluctuations were recorded from a spatial array of six sensitive microbarographs. Wind and temperature records from an instrumented mast were also available. A beam-steering technique has been used to determine wave parameters from the surface pressure data. The microbarographs detected the presence of internal gravity waves throughout the observational campaign. Rootmean-square pressure values were typically in the region 16‐40 mb, but a significant number of isolated events with amplitudes of up to 180 mb were also found. Wave properties have been studied in conjunction with the mean wind and temperature profiles in the boundary layer. It was found that most of the wave activity did not originate locally, but from shear layers aloft, or, more commonly, from the katabatic flow regime where the ice shelf joins the Antarctic continent.

Cattle, weather and water: mapping Escherichia coli O157:H7 infections in humans in England and Scotland

Entero-haemorrhagic Escherichia coli O157:H7 is a zoonotic pathogen, responsible for a relatively small number of food poisoning and illness outbreaks each year, when compared with other food-borne bacteria capable of causing infections in the population. Nevertheless, E. coli O157:H7 is a bacterial pathogen associated with severe human illnesses including bloody diarrhoea and haemolytic uremic syndrome occurring in both outbreak and sporadic settings. In England and Wales approximately 1% of all laboratory-confirmed cases of food poisoning are the result of E. coli O157:H7; however, in Scotland this figure increases to 3%. When the size of the population is taken into account and the rate of E. coli O157:H7 confirmed cases per 100,000 population is examined, the rate of E. coli 0157:H7 infections in Scotland is much greater than England and Wales. The routes of transmission have changed over time, with new routes of transmission such as farm visits emerging. The prevalence of E. coli O157:H7 has a seasonal dependency, with greater faecal shedding of the organism in the warmer months; this is directly mirrored in the increased reporting of E. coli O157:H7 infection among hospitalized patients. This review attempts to suggest why this phenomenon occurs, paying particular attention to weather, animal movement and private water supplies.

A Practical Example of Low-Frequency Trend Removal

Three different ‘classical’ methods of removing low-frequency trends are used to detrend some instrumental data, and their effect is evaluated. The examples given here highlight problems that may occur whenever detrending is necessary. The trend present in the data arises from the passage of synoptic weather systems over the Antarctic during the operation of the instrument. The resultant signal varies over a period of hours. This contrasts with the phenomenon to be studied (atmospheric gravity waves), which have periods of a few tens of minutes at most and amplitudes typically one or two orders of magnitude smaller than for synoptic systems. The trend removal procedure is relevant to many more situations than the analysis of meteorological pressure data.

Detecting Waves Using an Array of Sensors

Abstract The method of frequency-wavenumber “beamsteering” is a familiar tool in seismic applications and radar work where it is commonly used to detect coherent, propagating disturbances in “noisy” data. However, its use in meteorological situations is not well documented. In this paper the beamsteering method is summarized and its application in a meteorological context is discussed. Three different algorithms are used to detect single waves in artificial data with varying signal-to-noise ratios. Two of these are different beamsteering algorithms:one is a low-resolution scheme where the frequency-wavenumber spectrum is estimated using a fixed wavenumber window and the other a high-resolution method where the wavenumber window is effectively optimized for each wavenumber. These are compared with a method based on the cross-correlation function, which is more routinely used in meteorological studies. The results from this analysis provide a quantifiable measure of the performance and reliability of the th...

Can an ensemble give anything more than Gaussian probabilities

Can a relatively small numerical weather predic- tion ensemble produce any more forecast information than can be reproduced by a Gaussian probability density func- tion (PDF)? This question is examined using site-specific probability forecasts from the UK Met Office. These fore- casts are based on the 51-member Ensemble Prediction Sys- tem of the European Centre for Medium-range Weather Fore- casts. Verification using Brier skill scores suggests that there can be statistically-significant skill in the ensemble forecast PDF compared with a Gaussian fit to the ensemble. The most significant increases in skill were achieved from bias- corrected, calibrated forecasts and for probability forecasts of thresholds that are located well inside the climatological limits at the examined sites. Forecast probabilities for more climatologically-extreme thresholds, where the verification more often lies within the tails or outside of the PDF, showed little difference in skill between the forecast PDF and the Gaussian forecast.

TalkMaths Better ! Evaluating and Improving an Intelligent Interface for Creating and Editing Mathematical Text

We outline a system, called TalkMaths, which has been under development for some time, to allow users to create, access and edit Mathematical text documents using speech. This interface system could prove to be of particular advantage to people with a range of disabilities, amongst others. We describe initial results of a study to evaluate how easy this system is to use, and recent developments aimed at improving it and making it more “user friendly” and intelligent.

SWIMS (Speech-Based Web Interface for Mathematics Using Statistical Language Models): An Intelligent Editing Assistant for Mathematical Text

This paper describes the development and evaluation of an intelligent web-based interface for editing mathematical text that assists the user with the aid of the predictive and corrective power of statistical language models. It offers options for predicting what will appear next (analogous to predictive text for SMS messages) and identifying likely errors due to simple mistakes on the users part in order to assist in correcting the errors. Using text-stream input, we investigate the utility of the error identification by studying the proportion of times the correct version of the complete mathematical expression appears within the M most likely alternatives suggested by our system. We aim to integrate these facilities into our existing Talk Maths system.

Interactive error correction using statistical language models in a client-server interface for editing mathematical text

Learning and using mathematical notation poses particular difficulties for people with various disabilities, partly due to its wide range of symbols and rather complicated layout. These pose great challenges, often affecting the educational and career opportunities of people who are visually impaired or have limited (or no) use of their hands or arms. Assistive systems to alleviate these difficulties would be of considerable benefit to such groups of people

Random forest explorations for URL classification

Phishing is a major concern on the Internet today and many users are falling victim because of criminals deceitful tactics. Blacklisting is still the most common defence users have against such phishing websites, but is failing to cope with the increasing number. In recent years, researchers have devised modern ways of detecting such websites using machine learning. One such method is to create machine learnt models of URL features to classify whether URLs are phishing. However, there are varying opinions on what the best approach is for features and algorithms. In this paper, the objective is to evaluate the performance of the Random Forest algorithm using a lexical only dataset. The performance is benchmarked against other machine learning algorithms and additionally against those reported in the literature. Initial results from experiments indicate that the Random Forest algorithm performs the best yielding an 86.9% accuracy.