[AAS15-P03] Statistical reconstruction of the heterogeneity of a dense urban wind field using a sparse network of weather stations
Keywords:micrometeorology, spatial wind distribution, sparse meteorological network, Mahalanobis distance
The vectorial wind field is a major factor in the transport and dispersion of air pollution in urban regions. This field is frequently characterized by an inherent spatial heterogeneity. This heterogeneity may be manifested by noticeable differences between rooftop level measurements in adjacent locations. Quite often the degree of heterogeneity changes through the day.
A possible way to obtain real-time information on the current state of the urban wind field is to use a dense weather stations’ network. Such endeavor is expansive and technically demanding. This leads to networks that are too sparse to accurately describe the changing degree of the urban wind vectors’ heterogeneity.
In situations where there is not sufficient information regarding a specific urban region, it is possible to conduct a two phase scheme. First, deploy a dense weather stations network for a limited time. Then leave a sparse network to continuously monitor the region. Such scheme requires the use of statistical methods in order to estimate essential features of the dense wind field from the sparse network’s measurements.
A useful method can be a tolerance region that contains a given proportion of the wind vectors’ spatial realizations. For bivariate normal distributed vectors, an ellipse shaped tolerance region can be constructed using an analytical Mahalanobis distance function. However, because actual measurements of urban wind vectors does not always follow this distribution, another approach is required.
This study used wind data collected in the metropolitan area of Tel-Aviv using a network of more than 20 weather stations that were separated by less than a kilometer from each other. The results show that the spatial wind distribution can be very well represented by a small sample of merely four stations. Based on this sample stations, empirical Mahalanobis distance functions were calculated for each season. These functions were found to fit well the logistic distribution. Using these functions, tolerance regions were applied on a different data set in order to validate the statistical projection.
A possible way to obtain real-time information on the current state of the urban wind field is to use a dense weather stations’ network. Such endeavor is expansive and technically demanding. This leads to networks that are too sparse to accurately describe the changing degree of the urban wind vectors’ heterogeneity.
In situations where there is not sufficient information regarding a specific urban region, it is possible to conduct a two phase scheme. First, deploy a dense weather stations network for a limited time. Then leave a sparse network to continuously monitor the region. Such scheme requires the use of statistical methods in order to estimate essential features of the dense wind field from the sparse network’s measurements.
A useful method can be a tolerance region that contains a given proportion of the wind vectors’ spatial realizations. For bivariate normal distributed vectors, an ellipse shaped tolerance region can be constructed using an analytical Mahalanobis distance function. However, because actual measurements of urban wind vectors does not always follow this distribution, another approach is required.
This study used wind data collected in the metropolitan area of Tel-Aviv using a network of more than 20 weather stations that were separated by less than a kilometer from each other. The results show that the spatial wind distribution can be very well represented by a small sample of merely four stations. Based on this sample stations, empirical Mahalanobis distance functions were calculated for each season. These functions were found to fit well the logistic distribution. Using these functions, tolerance regions were applied on a different data set in order to validate the statistical projection.