PHOENICS-Direct Heat-Island Simulation Scenario:
(first version. February 2012): dbs, gwm, jcl
(latest version November 2012): dbs, jrs
Click here for general
information about Simulation Scenarios.
- The current status of HEATISLE
- How to start HEATISLE
- Viewing the current (November 2012) scenario
- Running the simulation
- Inspecting the results
- Graphical display of the results
Further details of the default settings supplied in February 2012
- Preliminary assessment of the realism of the predictions
HEATISLE was originally designed for the
simulation of the 'Heat-Island' phenomenon, i.e. the formation
of elevated-temperature regions in the vicinity of cities and large
A recent account of this phenomenon can be viewed at the web-site
contacted by clicking
Features of the phenomenon which are disclosed there are:
- That it results primarily from the reduced heat loss by radiation and convection from urban areas in comparison to surrounding vegetation- or water-covered areas, as in the typical plot of near-ground temperatures shown below:
- That it is most pronounced when the velocity of the atmospheric wind is low, so that free (i.e. buoyancy-induced) convection is more important than forced convection.
- That it is to be properly understood only if:
- Long-wave radiation (e.g. from and between buildings) is considered as well as short-wave radiation (i.e. from the sun);
- Radiation absorption, reflection and transmission properties of the atmosphere are taken into account:
- Both day and night are included in the simulation period;
- The differing heat capacities of buildings, soil and vegetation are also considered.
2. The current status of HEATISLE
HEATISLE was created by CHAM UK in order to enable users more easily to exercise a city-simulation model supplied in January 2012 by CHAM Shanghai. CHAM UK took the opportunity to add a solar-radiation model, to remove certain inconsistencies in the building-description files describing the building, and to provide for restricting attention to smaller groups of buildings; but little else was done.
During 2012, many new feature were added to PHOENICS-Direct and to SimScenes. Accordingly the original Parameterised Q1 was brought into line with these in November 2012; but little else was changed.
Consequently, HEATISLE is still far from complete. It should therefore be regarded as an exploratory tool, which has here been applied to a simple example so as to illustrate its future potential.
3. How to start HEATISLE
HEATISLE is activated via the file start.bat which resides in \phoenics\d_sapps\heatisle. The screen then shows the following image,:
Inspection and or modification of the input sat is effected via menus which appear when the appropriate bar in the second row is pressed. A typical one is as follows:
Displayed there is part of the 'geometrical' menu, which is one ot the six menus activated by pressing the boxes on the left of the screen. Each allows the values of groups of simulation-influencing parameters to be inspected and, if desired, changed by typing in the right-hand white boxes.
4. Viewing the current (November 2012) scenario
The parameter-changing session may involve introducing new objects, or deleting or moving already-present ones. Therefore clicking on the second-from-left top-row button
see what he has done, as shown below:
Here the grey objects represent buildings, and brown, green and blue areas indicate road, grass and water respectively.
This image has been displayed by the PHOENICS Virtual-Reality Editor, which permits zooming, and changing the viewing angle; but the Editor's data-changing facilities are not active in this mode of operation, lest they conflict with settings made by at of the PHOENICS-Direc menus.
5. Running the simulation
When the user is satisfied with the parameter which have been set, clicking the running-man button will cause the flow-simulating calculation to be performed. He then has nothing to do but wait until, after seconds, minutes or hours (according to the magnitude of the task), during which the scree shows the screen appears as something like the following image:
The coloured curves, here representing maximum and minimum values of solved-for variables extend farther and farther to the right as time proceeds; and the nearness to horizontality of their slopes indicates the closeness of the calculation to convergence.
6. Inspecting the results
When the run terminates, as a result of the user's allowed intervention or of his menu-made settings, the image on the screen may appear as follows
Shown here is the start of the standard PHOENICS RESULT file, which the user
can then explore by way of scrolling, or by use of the 'find'
Alternatively, clicking on the button to the right of 'find'
the user is enabled to browse for, and then inspect ,any other file in the
working directory or elsewhere.
Ordinarily, and not as illustrated above, the first-offered file will not be RESULT; for that contains too much general information for HEATISLE specialists. Smaller files, containing only what the specialists need, are commonly preferred. They are chosen via an 'output-related-items' menu.
7. Graphical display of the results
Clicking on the graphical-display button
activates whichever of the PHOENICS display packages the user has chosen, If it is the VR Viewer, what the screen then shows is something like this:
This is the starting screen of the Viewer, with, in the top-right-hand
corner, a 'macro' button .
Clicking on this elicits whatever sequence of results-display operations
have been prepared and placed in the input file. The user has only to press
RETURN to advance the sequence of pictures.
Those who are familiar with the VR-Viewer's capabilities are free thereafter to generate further images
8. Further details of the default setting supplied in February 2012
8.1 General description
The scenario represents a small section of a city, as shown below:
The scene contains:
- Some buildings, dark grey,
- Some roads brown
- A grass area, green,
- A section of river, blue and
- The surrounding ground, light grey
The blue arrow points North, and the red arrow indicates the wind direction, in
this example from the East. The scene is illuminated by the sun:
The big orange ball shows the sun location at the time of the simulation. The
small orange balls show the position of the sun at each hour during daytime, and
the small grey balls show the position of the sun during the night.
8.12 Parameters which users may alter via menus
- The domain size in X, Y and Z directions. The default sizes are 1250m * 625m *
- The origin of the scene within the domain.
- The name of the buildings file.
- Whether to include the road or not.
- Whether to include the river or not.
- Whether to include the grass or not.
- Whether to include clipping planes (which can help to visualise the scene later)
- Whether to include buoyancy or not.
- The thermal conductivities of the buildings, road, river and ground. The grass
area if included has the same conductivity as the ground.
- Emissivities of the buildings, road, grass, river and ground.
The properties of the elements are not well known. In steady-state, the only
important properties are the thermal conductivity and emissivity. The
conductivity of the ground will depend on the composition and water content. The
buildings are represented very simplistically as a single body. In reality they
comprise a mixture of materials and internal air space. The conductivity of the
buildings should reflect this mixture of materials. The emissivity will control
how well the surface of the element can radiate away the heat from the sun.
The best values for these properties can be estimated by comparing solutions to
known results and adjusting the properties until acceptable agreement, of trends
at least, is achieved.
- The direction the wind is blowing from. By default from the East.
- The orientation of the domain relative to North. By default the Y axis points
- The wind speed at the reference height of 10m above the ground, by default
- The ambient air temperature, by default 32 deg C.
- The latitude at which the scene is located, by default 51deg.
- The date on which the simulation is taking place, by default 1st June.
- The time of day of the simulation, by default midday, 12.00hrs.
- The direct solar radiation, by default 500 W/m2. This represents the
radiation which has passed directly through the atmosphere from the sun at the
particular latitude, date and time of day.
- The diffuse solar radiation, by default 100 W/m2. This represents the
radiation scattered by the atmosphere back to the earth at the particular
latitude, date and time of day.
- The temperature of the sky. This controls how much heat radiates back to the
sky. it is usually 10-20 deg C lower than the ambient temperature.
- Whether the simulation is to be run steady state or in transient mode.
- For transient operation only the duration of the simulation, and
- The time-step size.
- The number of outer iterations (sweeps), by default 1000.
- Whether to restart from an earlier solution or not.
- The number of computational cells in the X direction within the buildings, by
default 100. This gives a cell size of around 9.5m.
- The number of computational cells in the X direction outside the region containing buildings. They increase in size from the buildings to the domain edge.
- The number of computational cells in the Y direction within the building region, by
default 90. This gives a cell size of around 5m.
- The number of computational cells in the Y direction outside the region containing buildings.
- The number of computational cells in the Z direction under ground, by default 3.
The ground is 2m deep.
- The number of computational cells in the Z direction in the first layer of
buildings, by default 5. This is an 11m high zone from the ground to the top of
the lowest buildings.
- The number of computational cells in the Z direction in the second layer of
buildings, by default 15. This is an 82m high zone extending from the top of the
lowest buildings to the top of the highest building.
- The number of computational cells in the Z direction above the buildings, by
default 10. This zone extends from the top of the highest building to the top of
the domain. the cells expand away from the buildings towards the top of the
This image shows the default grid in the X-Y plane:
This image shows the default grid in the X-Z plane:
9. Preliminary assessment of the satisfactoriness of the predictions
9.1 The main features of the flow
The following animated image displays calculated shapes of stream lines, which start at an elevation of 7m at the upwind boundary. Also displayed are the pressure contours on the surfaces of the buildings. The wind is evidently blowing diagonally across the city, from left to right in the diagram.
The flow pattern there displayed can be regarded as qualitatively plausible. The tall buildings evidently experience elevated pressures on their upwind faces; they cause the local wind direction to be deflected ; and there is some evidence of recirculating flow in their wakes.
It is temperature distributions which HEATISLE is desiged to simulate; therefore these also should be examined. The following image provides a first oppportunity.
The surface temperature distributions thus revealed do not appear to be qualitatively plausible; for the temperatures on the surface of the taller buildings, which are exposed to the greaterer wind velocities, would be expected to be lower than those of the smaller ones; yet they are predicted as being higher. Moreover, there are some implausible cell-sized vertical striations in the surfaces of tall buildings.
The calculations were performed with the PARSOL variable set to FALSE. When it was set to TRUE, the fluid-flow pattern remained plausible:
However the temperature solution exhibited non-physical high and low values. Evidently there are errors somewhere which must be discovered and eliminated.
9.2 The fineness of the grid
(a) for hydrodynamics
The following image shows a plan view of a small portion of the grid close to the edge of the region containing the buildings, which appear as grey-coloured blocks. PARSOL
Evidently the horizontal dimensions of the buildings, and of the spaces between them, are in places only from one to five times the dimensions of the cells; of which the consequences are:
- No faith whatever can be placed in the quantitative accuracy of the predictions; for where the gap between buildings is no more than one cell thickness, even the direction of the predicted flow may be erroneous.
- This would be true even if the flow were being regarded as laminar, with a uniform-viscosity fluid; but it is even more so when, as here, a turbulence model is being employed. The reason is that the model requires gradients of velocity to be computed in order to that the generation rate of turbulence energy can be balanced against its diffusion, convection and dissipation rates in each cell. None of these quantities can be meaningfully calculated for one-cell-thick gaps. Ten cell thickness would be needed even for mere qualitative correctness.
- The confidence which may have been inspired by the animated streamlines displayed above is thus revealed as illusory.
(b) for temperature
The dubious striations in the surface-temperatures, remarked upon in section 9.1 can be seen also in the closer-up view below.
This view reveals that the striations are aligned with the cell edges; and, as better seen in the previous image, the buildings themselves appear to have protruberances on their walls of (coincidentally?) the very same size.
Evidently it will be wise for would-be users of HEATISLE to place no practical reliance on the predictions until these oddities have been explained.
(c) for diagnostic purposes
Explanation must be based upon diagnosis; and this on more detailed study. For this the grid is simultaneously too coarse and too fine; the former because it is necessary to reduce the cell dimensionsi to one tenth if the true distributions around them of temperature, turbulence energy, etc are to be computed; the latter because to apply that increase to all the cells in the current grid would far out-strip the available computer resources.
What is the solution? To refine the grid only around very few of the buildiings; and, by way of compensation, to coarsen it elsewhere. There are several ways of doing this namely:
- Restrict the domain of study to the volume surrounding a few buildings as shown below: