Esben Almkvist

Politecnico di Torino, Torino, Piedmont, Italy

Are you Esben Almkvist?

Claim your profile

Publications (8)1.32 Total impact

  • Knowledge-Based Intelligent Information and Engineering Systems, 11th International Conference, KES 2007, XVII Italian Workshop on Neural Networks, Vietri sul Mare, Italy, September 12-14, 2007, Proceedings, Part III; 01/2007
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The study described in this paper, analyzed the urban and suburban air pollution principal causes and identified the best subset of features (meteorological data and air pollutants concentrations) for each air pollutant in order to predict its medium-term concentration (in particular for the PM 10). An information theoretic approach to feature selection has been applied in order to determine the best subset of features by means of a proper backward selection algorithm. The final aim of the research is the implementation of a prognostic tool able to reduce the risk for the air pollutants concentrations to be above the alarm thresholds fixed by the law. The implementation of this tool will be carried out using machine learning methods based on some of the most wide-spread statistical data driven techniques (Artificial Neural Networks, ANN, and Support Vector Machines, SVM).
    01/2007;
  • [Show abstract] [Hide abstract]
    ABSTRACT: The research activity described in this paper concerns the study of the phenomena responsible for the urban and suburban air pollution. The analysis carries on the work already developed by the NeMeFo (neural meteo forecasting) research project for meteorological data short-term forecasting. The study analyzed the air pollution principal causes and identified the best subset of features (meteorological data and air pollutants concentrations) for each air pollutant in order to predict its medium-term concentration (in particular for the particulate matter with an aerodynamic diameter of up to 10 mum called PM10). The selection of the best subset of features was implemented by means of a backward selection algorithm which is based on the information theory notion of relative entropy. The final aim of the research is the implementation of a prognostic tool able to reduce the risk for the air pollutants concentrations to be above the alarm thresholds fixed by the law. The implementation of this tool will be carried out using data-driven models based on some of the most wide-spread statistical data-learning techniques (artificial neural networks and support vector machines).
    Proceedings of the International Joint Conference on Neural Networks, IJCNN 2007, Celebrating 20 years of neural networks, Orlando, Florida, USA, August 12-17, 2007; 01/2007
  • Source
    Esben Almkvist, Moniaci Walter
    [Show abstract] [Hide abstract]
    ABSTRACT: The road weather stations (RWIS) are constructed to measure the conditions of the road. The sensor equipment normally consists of sensors for surface temperature, a ir temperature, relative humidity, wind speed, precipitation and type of precipitation. This study tries to answer which variables should be measured at an RWIS-station. This equipment has remained si milar since 1979 when the RWIS-stations were first introduced. At a test site outside Göt eborg some 100 climate variables, apart from the normal variables of an RWIS-station are measur ed. A neural network model is used to select the variables that give the best prediction of the surface temperature. Thereby recommendations of how to equip an RWIS-station can be made. Some climatic variables would be difficult to include in the RWIS-system because of high maintenanc e level, it may be practically impossible or simply too expensive. Results show that more temperatur e sensors in the ground help the neural network model predict the surface temperature. Ground heat f lux and net radiation also improved the output of the model. The temperature predictions by the model were good when common variables were used as input and were improved when the additional variables were included. A forecast model from the Swedish meteorological office ( SMHI) was also given as input for the neural network model. While the model from SMHI alone performed rather poorly, when combining it with the measured variables and the neural network mode l a very large improvement was achieved. The neural network had adapted and improved the output from the SMHI model to the site specific conditions. The analyzed time series was only two months long, so it was too short for the neural network model to learn how to predict occasions of spec ial interest for road climate. A next step is to use a longer time series and more stations to improve the forecasts and so the model can learn to predict frost events. In the future the neural net work model can be used as nowcasting system to improve the output from forecasting models, such as the one from SMHI.
    01/2006;
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Since the start of field station manufacturing approximately 25 years ago, the configuration of Road Weather Information System (RWIS) stations and the type of sensors used has changed very little. Little is known about the variation of climatic variables from the 2 m level, where instruments are normally placed, down to the road surface. The `Road Climate Room' can be defined as the volume of air, road and surroundings influencing the conditions on the road. This study attempts to describe the Road Climate Room both theoretically and experimentally. Mobile measurements as well as a permanent station were used. At the permanent station detailed temperature measurements above and beside the road were made. These were related to other climatic variables such as wind, cloudiness, humidity and ground heat flux. The results show that the most important processes occur below the 10 cm level. The air column above this level is well represented by the 2 m level temperature sensor. For most situations the cooling of the air closest to the ground is much more intense in the vegetation than over the road. The temperature difference can be as large as 8 °C and can be represented statistically with high determination by wind and cloudiness. The heat storage in the road is a key factor for keeping the temperature of the road high throughout a full diurnal cycle. The cold air in the vegetation can be considered as a potential source for cold air drainage onto the road surface in terrain where that is a problem. No advection affecting the road surface was observed at the permanent station. This suggests that the Road Climate Room can be defined as being well within the internal boundary layer of the road. The mobile measurements, however, show that advection from the surroundings is likely to have taken place, which makes the definition of the Road Climate Room more difficult. Further studies are needed to fully understand the complicated processes.
    Meteorlogical Applications 01/2005; 12(04):357-370. · 1.32 Impact Factor
  • Source
    Esben Almkvist
    01/1999;
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The research activity described in this paper concerns the study of the phenomena responsible for the urban and suburban air pollution. The analysis carries on the work already developed by the NeMeFo (Neural Meteo Forecasting) research project for meteorological data short-term forecasting, Pasero (2004). The study analyzed the air-pollution principal causes and identified the best subset of features (meteorological data and air pollutants concentrations) for each air pollutant in order to predict its medium-term concentration (in particular for the PM10). The selection of the best subset of features was implemented by means of a backward selection algorithm which is based on the information theory notion of relative entropy. The final aim of the research is the implementation of a prognostic tool able to reduce the risk for the air pollutants concentrations to be above the alarm thresholds fixed by the law. The implementation of this tool will be carried out using the most wide-spread statistical data-learning techniques (Artificial Neural Networks, ANN, and Support Vector Machines, SVM).
  • Esben Almkvist, Crister Jansson
    [Show abstract] [Hide abstract]
    ABSTRACT: In Road Climatology great emphasis is put on predicting the surface temperature of a road. Good predictions will give information of the road conditions and warn for risk of slipperiness. Thereby accidents can be avoided by maintaining the road. In this study the emphasis is put on validating the models, not only with surface temperature, but also by studying the various heat uxes to and from the road surface. Radiation and ground heat ux can be measured with well known techniques, but the turbulent latent and sensible heat ux is very dicult to measure for a road surface. In order to measure the turbulent uxes it is necessary to get very close to the road surface, so that inuence from the surroundings is minimized. Various techniques from micrometeorology have been tested at a test site outside Göteborg in south western Sweden. The Eddy covariance technique seems to underestimate the ows, but the Bowen-ratio technique seemed to work under good conditions.