In order to study the wind turbine wake and its eventual interactions with neighbouring wind turbines, several numerical and physical modelling approaches are used. Some model the wind turbine with the simplest model, that is the actuator disc concept, adding a drag source (i.e. pressure loss) within the surface swept by the blades (numerical , physical ). Some use the Blade Element Momentum Theory, which takes into account the blade rotation effect on the wake and the aerodynamic features of the blades . Some use Large Eddy Simulation to compute the unsteady flow around the entire rotor . In a wind resource assessment context, the latter one is not practical enough to be used since the computation times are extremely long.
Wind tunnel experiments are performed in both neutral and stable boundary layers to study the effect of thermal stability on the wake of a single turbine and on the wakes of two axially aligned turbines, thereby also showing the influence of the second turbine on the impinging wake. In the undisturbed stable boundary layers, the turbulence length scales are significantly smaller in the vertical and longitudinal directions (up to 50 % and ≈≈40 %, respectively), compared with the neutral flow, while the lateral length scale is unaffected. The reductions are larger with the imposed inversion of a second stable case, except in the near-wall region. In the neutral case, the length scales in the wake flow of the single turbine are reduced both vertically and laterally (up to 50 % and nearly 40 %, respectively). While there is significant upstream influence of a second turbine (on mean and turbulence quantities), there is virtually no upstream effect on vertical length scales. However, curiously, the presence of the second turbine aids length-scale recovery in both directions. Longitudinally, each turbine contributes to successive reduction in coherence. The effect of stability on the turbulence length scales in the wake flows is non-trivial: at the top of the boundary layer, the reduction in the wall-normal length scale is dominated by the thermal effect, while closer to the wall, the wake processes strongly modulate this reduction. Laterally, the turbines’ rotation promotes asymmetry, while stability opposes this tendency. The longitudinal coherence, significantly reduced by the wake flows, is less affected by the boundary layer's thermal stability.
Two cases of an overlying inversion imposed on a stable boundary layer are investigated, extending the work of Hancock and Hayden (Boundary-Layer Meteorol 168:29-57, 2018; 175:93-112, 2020). Vertical profiles of Reynolds stresses and heat flux show closely horizontally homogeneous behaviour over a streamwise fetch of more than eight boundary-layer heights. However, profiles of mean temperature and velocity show closely horizontally homogeneous behaviour only in the top two-thirds of the boundary layer. In the lower one-third the temperature decreases with fetch, directly as a consequence of heat transfer to the surface. A weaker effect is seen in the mean velocity profiles, curiously, such that the gradient Richardson number is invariant with fetch, while various other quantities are not. Stability leads to a 'blocking' of vertical influence. Inferred aerodynamic and thermal roughness lengths increase with fetch, while the former is constant in the neutral case, as expected. Favourable validation comparisons are made against two sets of local-scaling systems over the full depth of the boundary layer. Close concurrence is seen for all stable cases for z/L < 0.2, where z and L are the vertical height and local Obukhov length, respectively, and over most of the layer for some quantities.
It is demonstrated that the vertical profile of gradient Richardson number, Ri , can be shaped by control of the working-section inlet temperature profile. In previous work (Hancock and Hayden in Boundary-Layer Meteorol 168:20–57, 2018; 175:93–112, 2020; 180:5–26, 2021) the inlet temperature profile had been specified but without control of the profile of Ri in the developed-flow region of the working section. Control of the inlet temperature profile is provided by 15 inlet heaters (spread uniformly across the height of the working section), allowing control of the temperature gradient over the bulk of the boundary layer, and the overall temperature level above that of the surface. The bulk Richardson number for the 11 cases covers the range 0.01–0.17 (there is no overlying inversion). In the upper ≈ 2/3 of the boundary layer the Reynolds stresses and turbulent heat flux are controlled by the gradient in mean temperature, while in the lower ≈ 1/3 they are controlled both by this gradient and by the level above the surface temperature. In three examples, Ri is approximately constant at 0.07, 0.10 and 0.13 across the bulk of the layer. The previous observation of horizontally homogenous behaviour in the temperature profiles in the top ≈ 2/3 of the boundary layer but not in the lower ≈ 1/3 is repeated here, except when, tentatively, Ri does not exceed 0.05 over the bulk of the boundary layer. Favourable validation comparisons are made against two sets of local scaling systems and field data over the full depth of the boundary layer, over the range 0.006 ≤ R i ≤ 0.3, or, in terms of height and local Obukhov length, 0.005 ≤ z / L ≤ 1.
Scalar dispersion from ground-level sources in arrays of buildings is investigated using wind-tunnel measurements and large-eddy simulation (LES). An array of uniform-height buildings of equal dimensions and an array with an additional single tall building (wind tunnel) or a periodically repeated tall building (LES) are considered. The buildings in the array are aligned and form long streets. The sensitivity of the dispersion pattern to small changes in wind direction is demonstrated. Vertical scalar fluxes are decomposed into the advective and turbulent parts and the influences of wind direction and of the presence of the tall building on the scalar flux components are evaluated. In the uniform-height array turbulent scalar fluxes were dominant, whereas the tall building causes an increase of the magnitude of advective scalar fluxes which become the largest component. The presence of the tall building causes either an increase or a decrease to the total vertical scalar flux depending on the position of the source with respect to the tall building. The results of the simulations can be used to develop parametrizations for street canyon dispersion models and enhance their capabilities in areas with tall buildings.
Large-eddy simulation (LES) is used to calculate the concentration fluctuations of passive plumes from an elevated source (ES) and a ground-level source (GLS) in a turbulent boundary layer over a rough wall. The mean concentration, relative fluctuations and spectra are found to be in good agreement with the wind-tunnel measurements for both ES and GLS. In particular, the calculated relative fluctuation level for GLS is quite satisfactory, suggesting that the LES is reliable and the calculated instantaneous data can be used for further post-processing. Animations are shown of the meandering of the plumes, which is one of the main features to the numerical simulations. Extreme value theory (EVT), in the form of the generalized Pareto distribution (GPD), is applied to model the upper tail of the probability density function of the concentration time series collected at many typical locations for GLS and ES from both LES and experiments. The relative maxima (defined as maximum concentration normalized by the local mean concentration) and return levels estimated from the numerical data are in good agreement with those from the experimental data. The relative maxima can be larger than 50. The success of the comparisons suggests that we can achieve significant insight into the physics of dispersion in turbulent flows by combining LES and EVT. Present address: School of Engineering Sciences (Aero), University of Southampton, Southampton SO17 1BJ, UK.
Pollutant mass fluxes are rarely measured in the laboratory, especially their turbulent component. They play a major role in the dispersion of gases in urban areas and modern mathematical models often attempt some sort of parametrisation. An experimental technique to measure mean and turbulent fluxes in an idealised urban array was developed and applied to improve our understanding of how the fluxes are distributed in a dense street canyon network. As expected, horizontal advective scalar fluxes were found to be dominant compared with the turbulent components. This is an important result because it reduces the complexity in developing parametrisations for street network models. On the other hand, vertical mean and turbulent fluxes appear to be approximately of the same order of magnitude. Building height variability does not appear to affect the exchange process significantly, while the presence of isolated taller buildings upwind of the area of interest does. One of the most interesting results, again, is the fact that even very simple and regular geometries lead to complex advective patterns at intersections: parametrisations derived from measurements in simpler geometries are unlikely to capture the full complexity of a real urban area.
The CFD model Fluidyn-Panache was configured to model atmospheric transport from an area source. Modelled flow and turbulence were evaluated by comparison with on-site meteorological measurements, whilst atmospheric dispersion was compared with wind tunnel measurements. The results showed that higher rates of vertical and lateral dispersion were modelled than were determined in the wind tunnel, though modelled and measured ground-level centreline concentration data were within a factor of two. Uncertainties in wind tunnel and numerical modelling were highest close to the source. Consideration of fine-scale features was only necessary for receptors in the immediate near-field.
The study investigates thermal stratification effects of approach flows on dispersion in urban environments. This is in some ways analogous to a well-developed non-neutral flow (e.g. through a large urban area) approaching a neighbourhood-scale urban region, where the effect of the local heat transfer was assumed less important. A generic urban-type geometry, i.e. a group of staggered cubes, was taken as the first test case. The DAPPLE site, which was about a one-km2 region near the intersection of Marylebone Road and Gloucester Place in central London, was taken as the second test case. Only weakly unstable conditions (i.e. bulk Richardson number Rb ≥ - 0.2) of approach flows were considered, with adiabatic boundary conditions at the ground and building surfaces. A number of numerical experiments were performed. The modelled mean concentration for Rb = -0.1 gave the best agreement with the field data at all DAPPLE stations. This suggests that stratification effects on dispersion in weakly unstable conditions (e.g. in London) are not negligible
A family of wall models is proposed that exhibits more satisfactory performance than previous models for the large-eddy simulation (LES) of the turbulent boundary layer over a rough surface. The time and horizontally averaged statistics such as mean vertical profiles of wind velocity, Reynolds stress, turbulent intensities, turbulent kinetic energy and also spectra are compared with wind-tunnel experimental data. The purpose of the present study is to obtain simulated turbulent flows that are comparable with wind-tunnel measurements for use as the wind environment for the numerical prediction by LES of source dispersion in the neutral atmospheric boundary layer.
The statistics of the fluctuating concentration field within a plume is important in the analysis of atmospheric dispersion of toxic, inflammable and odorous gases. Previous work has tended to focus on concentration fluctuations in single plumes released in the surface layer or at ground level and there is a general lack of information about the mixing of two adjacent plumes and how the statistical properties of the concentration fluctuations are modified in these circumstances. In this work, data from wind tunnel experiments are used to analyse the variance, skewness, kurtosis, intermittency, probability density function and power spectrum of the concentration field during the mixing of two identical plumes and results are compared with those obtained for an equivalent single plume. The normalised variance, skewness and kurtosis on the centre-lines of the combined plume increase with distance downwind of the stack and, in the two-source configuration, takes lower values than those found in the single plumes. The results reflect the merging process at short range, which is least protracted for cases in which the sources are in-line or up to 30° and more, the plumes are effectively side-by-side during the merging process and the interaction between the vortex pairs in each plume is strong. Vertical asymmetry is observed between the upper and the lower parts of the plumes, with the upper part having greater intermittency (i.e. the probability that no plume material is present) and a more pronounced tail to the concentration probability distribution. This asymmetry tends to diminish at greater distances from the source but occurs in both buoyant and neutral plumes and is believed to be associated with the 'bending-over' of the emission in the cross-flow and the vortex pair that this generates. The results allowed us to identify three phases in plume development. The first, very near the stack, is dominated by turbulence generated within the plume and characterised by concentration spectra with distinct peaks corresponding to scales comparable with those of the counter-rotating vortex pair. A second phase follows at somewhat greater distances downwind, in which there are significant contributions to the concentration fluctuations from both the turbulence internal to the plume and the external turbulence. The third phase is one in which the concentration fluctuations appear to be controlled by the external turbulence present in the ambient flow. © 2013 Springer Science+Business Media Dordrecht.
A wind-tunnel simulation of an atmospheric boundary layer, artificially thickened as is often used in neutral flow wind-loading studies, has been investigated for weakly unstable stratification, including the effect of an overlying inversion. Rather than using a uniform inlet temperature profile, the inlet profile was adjusted iteratively by using measured downstream profiles. It was found that three cycles are sufficient for there to be no significant further change in profiles of temperature and other quantities. Development to nearly horizontally-homogeneous flow took a longer distance than in the neutral case because the simulated layer was deeper and therefore the length scales larger. Comparisons show first-order and second-order moments quantities are substantially larger than given by 'standard forms' in the mixed layer but are close in the surface layer. Modified functions, obtained by matching one to the other, are suggested that amount to an interpolation in the mixed layer between the strongly unstable and the weakly unstable cases.
Experimental results on the wake properties of a non-rotating simplified wind turbine model, based on the actuator disc concept, and a rotating model, a three-blade wind turbine, are presented. Tests were performed in two different test sections, one providing a nominally decaying isotropic turbulent inflow (turbulence intensity of 4% at rotor disc location) and one providing a neutral atmospheric boundary layer above a moderately rough terrain at a geometric scale of 1:300 (determined from the combination of several indicators), with 13% of turbulence intensity at hub height. The objective is to determine the limits of the simplified wind turbine model to reproduce a realistic wind turbine wake. Pressure and high-order velocity statistics are therefore compared in the wake of both rotor discs for two different inflow conditions in order to quantify the influence of the ambient turbulence. It has been shown that wakes of rotating model and porous disc developing in the modeled atmospheric boundary layer are indistinguishable after 3 rotor diameters downstream of the rotor discs, whereas few discrepancies are still visible at the same distance with the decaying isotropic turbulent inflow.
We present results from laboratory and computational experiments on the turbulent flow over an array of rectangular blocks modelling a typical, asymmetric urban canopy at various orientations to the approach flow. The work forms part of a larger study on dispersion within such arrays (project DIPLOS) and concentrates on the nature of the mean flow and turbulence fields within the canopy region, recognis- ing that unless the flow field is adequately represented in computational models there is no reason to expect realistic simulations of the nature of the dispersion of pollutants emitted within the canopy. Comparisons between the experimental data and those ob- tained from both large-eddy simulation (LES) and direct numerical simulation (DNS) are shown and it is concluded that careful use of LES can produce generally excellent agreement with laboratory and DNS results, lending further confidence in the use of LES for such situations. Various crucial issues are discussed and advice offered to both experimentalists and those seeking to compute canopy flows with turbulence resolving models
Methods used to convert wind tunnel and ADMS concentration feld data for a complex building array into effective radiation dose were developed based on simulations of a site in central London. Pollutant source terms were from positron emitting gases released from a cyclotron and clinical PET radiotracer facility. Five years of meteorological data were analysed to determine the probability distribution of wind direction and speed. A hemispherical plume cloud model (both static and moving) was developed which enabled an expression of gamma-ray dose, taking into account build-up factors in air, in terms of analytic functions in this geometry. The standard building wake model is presented, but this is extended and developed in a new model to cover the concentration feld in the vicinity of a roof top structure recirculation zone, which is then related to the concentration in the main building wake zone. For all models presented the effective dose was determined from inhalation, positron cloud immersion and gamma ray plume contributions. Results of applying these models for determination of radiation dose for a particular site are presented elsewhere.
The simulation of horizontally homogeneous boundary layers that have characteristics of weakly and moderately stable atmospheric flow is investigated, where the well-established wind engineering practice of using ‘flow generators’ to provide a deep boundary layer is employed. Primary attention is given to the flow above the surface layer, in the absence of an overlying inversion, as assessed from first- and second-order moments of velocity and temperature. A uniform inlet temperature profile ahead of a deep layer, allowing initially neutral flow, results in the upper part of the boundary layer remaining neutral. A non-uniform inlet temperature profile is required but needs careful specification if odd characteristics are to be avoided, attributed to long-lasting effects inherent of stability, and to a reduced level of turbulent mixing. The first part of the wind-tunnel floor must not be cooled if turbulence quantities are to vary smoothly with height. Closely horizontally homogeneous flow is demonstrated, where profiles are comparable or closely comparable with atmospheric data in terms of local similarity and functions of normalized height. The ratio of boundary-layer height to surface Obukhov length, and the surface heat flux, are functions of the bulk Richardson number, independent of horizontal homogeneity. Surface heat flux rises to a maximum and then decreases.
Stable and convective boundary layers over a very rough surface have been studied in a thermally-stratified wind tunnel. Artificial thickening by means of spires was used to accelerate the formation of a sufficiently deep boundary layer, suitable for urban-like boundary layer flow and dispersion studies. For the stable boundary layer, the methodology presented in Hancock and Hayden (2018) for low-roughness offshore surface conditions has been successfully applied to cases with higher-roughness. Different levels of stratification and roughness produced modifications in the turbulence profiles of the lower half of the boundary layer, but little or no change in the region above. Data for a stronger stability case suggested that the employed spires may not be suitable to simulate such extreme condition, though further studies are needed. The results were in reasonably good agreement with field measurements. For the convective boundary layer, great attention was given to the flow uniformity inside the test section. The selection of a non-uniform inlet temperature profile was in this case found not as determinant as for the stable boundary layer to improve the longitudinal uniformity, while the application of a calibrated capping inversion considerably improved the lateral uniformity. The non-dimensional vertical profiles of turbulent quantities and heat fluxes, did not seem to be influenced by roughness.
The need to balance computational speed and simulation accuracy is a key challenge in designing atmospheric dispersion models that can be used in scenarios where near real-time hazard predictions are needed. This challenge is aggravated in cities, where models need to have some degree of building-awareness, alongside the ability to capture effects of dominant urban flow processes. We use a combination of high-resolution large-eddy simulation (LES) and wind-tunnel data of flow and dispersion in an idealised, equal-height urban canopy to highlight important dispersion processes and evaluate how these are reproduced by representatives of the most prevalent modelling approaches: (i) a Gaussian plume model, (ii) a Lagrangian stochastic model and (iii) street-network dispersion models. Concentration data from the LES, validated against the wind-tunnel data, were averaged over the volumes of streets in order to provide a high-fidelity reference suitable for evaluating the different models on the same footing. For the particular combination of forcing wind direction and source location studied here, the strongest deviations from the LES reference were associated with mean over-predictions of concentrations by approximately a factor of 2 and with a relative scatter larger than a factor of 4 of the mean, corresponding to cases where the mean plume centreline also deviated significantly from the LES. This was linked to low accuracy of the underlying flow models/parameters that resulted in a misrepresentation of pollutant channelling along streets and of the uneven plume branching observed in intersections. The agreement of model predictions with the LES (which explicitly resolves the turbulent flow and dispersion processes) greatly improved by increasing the accuracy of building-induced modifications of the driving flow field. When provided with a limited set of representative velocity parameters, the comparatively simple street-network models performed equally well or better compared to the Lagrangian model run on full 3D wind fields. The study showed that street-network models capture the dominant building-induced dispersion processes in the canopy layer through parametrisations of horizontal advection and vertical exchange processes at scales of practical interest. At the same time, computational costs and computing times associated with the network approach are ideally suited for emergency-response applications.
Four cases of an overlying inversion imposed on a stable boundary layer are investigated, extending the earlier work of Hancock and Hayden (Boundary-Layer Meteorol 168:29–57, 2018), where no inversion was imposed. The inversion is imposed to one or other of two depths within the layer: midway or deep. Four cases of changed surface condition are also investigated, and it is seen that the surface and imposed conditions behave independently. A change of imposed inversion condition leaves the bottom 1/3 of the layer almost completely unaffected; a change of the surface condition leaves the top 2/3 unaffected. Comparisons are made against two sets of local-scaling systems over the full height of the boundary layer. Both show some influence of the inversion condition. The surface heat flux and the reduction in surface shear stress, and hence the ratio of the boundary-layer height to surface Obukhov length, are determined by the temperature difference across the surface layer (not the whole layer), bringing all cases together in single correlations as functions of a surface-layer bulk Richardson number.
A series of wind tunnel experiments were conducted in the University of Surrey's Environmental Flow wind tunnel with a 1:50 scale of a typical London street canyon to assess the exposure of cyclists riding in a group to the emissions of polluting vehicles. A propane source emitted from an Ahmed body was used to model a car exhaust and a fast flame ionisation detector was used to measure pollutant concentration around four cyclists for multiple configurations of the source, cyclists, and wind directions. Two cases were investigated with a vehicle driving in front of a line of cyclists and adjacent to them (as if it were overtaking them). In the first case, for small wind incidence, findings confirm that the cyclists exposure decreases exponentially with their distance from the source with a small dependence on wind direction but largely independently of the riders position within the group. For large wind incidences, typical of urban canyons, the rider position within the group becomes more important. For the second set of experiments, with the vehicle positioned adjacent to the riders, it was found to be preferable for a rider to be in front of the group regardless of the distance from the source, as this results in lower exposure to pollutants. This is likely linked with the complex aerodynamic field generated by the group of riders that can trap the vehicle exhaust fumes amongst the cyclists, hence increasing the exposure. This research suggests that group riding should be considered when designing mitigation strategies to minimise cyclists exposure to road traffic pollution within urban environments, where busy and narrow cycle lanes often results in cyclists riding in line.
Wind tunnel experiments on regular arrays of buildings were conducted in the environmental wind tunnel in the EnFlo laboratory at the University of Surrey. The model canopy comprised a square array of 14×21 rectangular blocks (1h × 2h) with height h = 70 mm. Preliminary measurements of velocity, turbulence and tracer concentrations were made for 3 wind directions: 0, 45 and 90◦. The results from this first experimental campaign along with numerical simulations have shown that the canopy has obstacles sufficiently long compared with their heights to yield extensive flow channelling along streets. Across the whole of the downwind half of the long street the flow for the present canopy is closely aligned with the obstacle faces, despite the 45◦ flow orientation aloft. This supports the suggestion that the streets are long enough to be representative for street network modelling approaches; shorter streets would probably not be sufficient and it will be interesting to see how well network models can predict concentrations in the present canopy. The extensive array and the small scale of the model posed challenging problems for reaching the desired high accuracy needed to validate the numerical simulations. The improvements in the methodology will be presented and discussed at the conference. The wind tunnel data, along with LES and DNS simulations, are being used to understand the behaviour of flow and dispersion within regular array with a more realistic geometry than the usual cuboids. This integrated methodology will help developing parametrisations for improved street network dispersion models
It is demonstrated that the vertical profile of gradient Richardson number, Ri, can be shaped by control of the working-section inlet temperature profile. In previous work (Hancock and Hayden in Boundary-LayerMeteorol 168:20–57, 2018; 175:93–112, 2020; 180:5–26, 2021) the inlet temperature profile had been specified but without control of the profile of Ri in the developed-flow region of the working section. Control of the inlet temperature profile is provided by 15 inlet heaters (spread uniformly across the height of the working section), allowing control of the temperature gradient over the bulk of the boundary layer, and the overall temperature level above that of the surface. The bulk Richardson number for the 11 cases covers the range 0.01–0.17 (there is no overlying inversion). In the upper ≈ 2/3 of the boundary layer, the Reynolds stresses and turbulent heat flux are controlled by the gradient in mean temperature, while in the lower ≈ 1/3 they are controlled both by this gradient and by the level above the surface temperature. In three examples, Ri is approximately constant at 0.07, 0.10 and 0.13 across the bulk of the layer. The previous observation of horizontally homogenous behaviour in the temperature profiles in the top ≈ 2/3 of the boundary layer but not in the lower ≈ 1/3 is repeated here, except when, tentatively, Ri does not exceed 0.05 over the bulk of the boundary layer. Favourable validation comparisons are made against two sets of local scaling systems and field data over the full depth of the boundary layer, over the range 0.006 ≤ Ri ≤ 0.3, or, in terms of height and local Obukhov length, 0.005 ≤ z/L ≤ 1.
The EMPAS collaborative project covered development and demonstration of an electric motor powered aero-engine simulator in the QinetiQ 5-metre wind-tunnel. As part of the project, the University of Surrey was responsible for background research and development of various flow measurement and system monitoring technologies, for use in the EMPAS windtunnel tests. Development of an integrated digital pressure rake system is presented, which has enabled assessment of duct flow characteristics and derivation of fan pressure ratio for the EMPAS system. A novel approach based on use of miniature, low cost microphones was developed, which successfully provided in-situ fan RPM measurement and highlighted several promising areas for future monitoring of fan/system health. A 7-hole pressure probe was used to great effect for extensive flow-field surveys, to investigate and validate the aerodynamic flow field effects associated with the aero-engine simulator. Wind-tunnel results from use of the microphone and 7-hole probe systems are analyzed and presented in detail.
Following a malicious or accidental release in an outdoor environment (industrial or urban), first responders will ensure public safety by cordoning off and/or evacuating areas where human life may be in danger. Information on the source (strength and location) and the type of chemical agent released is needed for this to happen reasonably promptly and accurately. A simple inverse modelling technique has been developed to estimate the source strength and location of such a release using measurements of concentration from chemical sensors. The technique relies on either a fixed installation or rapid deployment of chemical sensors to gather and return data to a base station. These measurements are there used, together with meteorological information, as the input data to an inverse algorithm that attempts to make a “best” estimate of the source strength and location. The algorithm works to minimise a penalty function that measures the difference between the concentration observations and predictions based on the current estimate of the source parameters. This is an iterative procedure that should converge to a best estimate of those parameters and, in doing so, provide a measure of the uncertainty in that estimate. There is, in this, a trade-off between the desire for an early prediction and the error implicit in that prediction. Wind tunnel experiments have been used to investigate the propagation of error through the inverse modelling procedure. Firstly, very detailed dispersion measurements were made in a deep boundary layer so that an accurate dispersion model could be established. Four fast flame ionisation detectors were then used to provide long, simultaneous concentration records in the plume from a ground level point source. The output was used to study the sensitivity to sensor placement and then sample duration. Simultaneous sub-samples were taken from the main records and used with the inversion algorithm to quantify the degradation of its performance with decreasing sample duration; i.e. with increasing uncertainty in the concentration observations. Guidelines for application of the inversion technique could then be proposed. The final stage was to move from a simple Gaussian plume to an urban dispersion model, in this case a street network model.