When simulating air flows in an urban environment, for e.g. pollutant dispersion investigations, today's main tool is advanced computational fluid dynamics simulations. These simulations take a lot of time and resources to perform, even for small geometries. In some situations, one would like to be able to run approximate simulations, possibly with large geometries, without such a significant investment. The model described in this thesis is a graph network model which have streets and intersections of an urban environment modeled as connections and nodes in a graph.
The model uses a pressured pipe model, based on the Darcy-Weisbach equation, to simulate air flow in the network. Such a model requires only rough measurements of the urban geometry and an estimated Darcy's friction factor, to be able to solve the system. Furthermore, using the same rough geometrical parameters, together with shear velocity, the model solves atmospheric exchange rates of the streets.
Intersections play a major role when investigating urban dispersion. The way this model deals with dispersion in any complex intersections, represented by single nodes, is by using wind direction variance together with a distribution parameter based on computational fluid dynamics intersection simulations made in Comsol Multiphysics - also present in this paper.
Using the simple model described above, I have simulated urban air flows in a complex urban geometry of a part of Paris. This specific geometry has already been investigated by computational fluid dynamics simulations as well as wind tunnel experiments. By comparing the computational fluid dynamics simulation with my model, I have validated its accuracy. 40% and 45% of all streets reach a relative and absolute error below 25% respectively. Directions of the street velocities have been simulated with approximately 90% accuracy - with distinct error indications. Atmospheric exchange rates of the streets are within an order of magnitude accurate, however, showing a systematic error by overestimating the vast majority of the exchange rates.
The model could become even better by covering error sources discussed in the discussion section. Excess theory for simulating each of the above-described flows is presented, which might change the results. For example, slightly altering the modeling of the atmospheric exchange rate might fix the overestimation offset we have seen.
Potential error sources could be the varying building heights and the streets angle relative the overlaying wind direction. The pressured pipe simulated flows have shown tendencies to be bad at picking up the effects of high/low buildings following low/high buildings, as well as accurately capture the behavior of streets close to perpendicular to the wind direction. Main streets with plenty of exits have been modeled with intersections at each exit, which results in strong flow variation along a street that should have a flow close to constant. Solving main streets like this separately could improve this behavior drastically.
2016. , 51 p.
Network model, Street canyons, Intersections, Urban dispersion, CFD data, CFD simulations, Atmospheric exchange
Björnham, Oscar, Ph.D.