Modelling the dispersion of vehicular carbon monoxide pollution along Deerfoot Trail in Calgary

Abstract (MGIS Thesis)

The main objective of this study is to model the dispersion pattern of vehicular carbon monoxide along Deerfoot Trail in Calgary by using CALINE4 software combined with GIS techniques. A road network (14.28 km long), extending from south-east to north-east Deerfoot Trail, is considered as the road segment under study. The study is broadly divided into three parts viz. CALINE4 analysis, GIS analysis and S-PLUS analysis. The CALINE4 analysis involves two different run type models to calculate 1-hour average CO concentrations at the receptors under ‘Standard’ run type model and then 8-hour average CO concentration under ‘Multi-Run/WorstCase hybrid’ model. A typical day (June 1, 2006) is chosen to calculate 1-hour average CO concentration at the receptors from 7:00 am to 6:00pm. The predicted CO is used to evaluate the model by comparing to the observed values of CO at south-east monitoring station. The output from standard run type model shows that the predicted CO concentration is mainly dependent on wind speed and wind direction. The R-square value of 0.47 suggests that the model explained 47% of the original variability in the data. The predicted 8-hour average CO concentration from 7:00 am to 2:00 pm is used for interpolating the surface to predict the values of CO at unknown locations within the study area. The predicted CO concentrations at the receptors are interpolated using Inverse Distance Weighting (IDW), Global Polynomial (GP), Ordinary Kriging (J-Bessel model) and Universal Kriging (Spherical model) interpolation methods. The Oridnary Kriging model is statistically more significant than the Universal Kriging model with lower Root Mean Square (RMS) value of 0.32 and higher RMS standardized value of 0.92. However, the Trend Surface Analysis shows the presence of first order trend in the data and the Geometric Anisotropy shows the presence of anisotropy in the data. Therefore, Ordinary Kriging model uses a biased estimator and is not reliable. The Universal Kriging model accounts for both trend and anisotropy in the data. As such, it is the best model that can be used reliably for decision-making purposes. This model has RMS and RMS standardized values of 0.33 and 0.87 respectively. The S-PLUS analysis is used to validate the statistical significance of the GIS analysis and perform Universal kriging. The Universal kriging model shows CO concentration ‘hot-spots’ especially around the areas close to the highway.

Read entire thesis

Poster Presentation of MGIS Thesis (awarded as the best MGIS Poster from The Department of Geography at University of Calgary in 2010)

Read more

Modeling the dispersion of vehicular carbon-monoxide pollution in Kathmandu valley

Modeling the dispersion of vehicular carbon-monoxide pollution in Kathmandu valley, Nepal: A CALINE4 approach combined with GIS Techniques

Abstract

Kathmandu valley is more vulnerable to air pollution than other rapidly growing Asian cities because of the bowl like structure of the valley and poor wind speed inside the valley. The main objective of this study is to model the dispersion pattern of vehicular carbon monoxide in Kathmandu valley by using CALINE4 software combined with GIS techniques. CALINE4 uses vehicular count, pollution and meteorological data to predict the carbon monoxide (CO) concentration. A typical day (15th of February 2007) is chosen to calculate 1-hour average CO concentration at the receptor points during peak hour (8:00 – 9:00 am). A road network extending from Maitighar to Koteshwor, which is approximately 4 km in length, is considered as the main road network. Ninety receptor points are created within the 500 meters buffer area of the main road network and CALINE4 is used to predict CO concentration at these points. The predicted CO concentration at the receptor points are then interpolated using K-Bessel universal kriging. The resulting map is reclassified to create ‘hot-spots’ where the areas are classified based on the predicted CO concentration. Root mean square error (RMSE) method is carried out to evaluate the model performance by comparing the predicted and observed CO concentration within 10 meters buffer from the study site. The RMSE value is found to be 0.77 and the accuracy of the model performance as 74%. Read entire article

Read more