Tutorial Section

Model areas should have a few cells at the model boundary where no building cells are digitized. Terrain, soil profiles, and vegetation should still be properly digitized. Building cells near the model boundary can block or channel wind flow, which can lead to large instabilities and ultimately simulation crashes.

The amount of open space needed at the boundary depends on building density and height. As a rule of thumb, the distance between the model boundary and the first building should be half the building height. This typically results in 4-8 cells of open space at each model area boundary. In the vertical dimension, however, the distance should be much greater (more on this in the next question).

Most ENVI-met studies analyze outdoor thermal comfort at pedestrian level and therefore require a high vertical resolution. In general, the height of the model domain should be at least twice the height of the tallest building. However, if very tall buildings (e.g., 100 m tall) are included in the model domain, 100 Z-cells with a resolution of 2 m would be needed to accurately represent the model domain and still obtain high resolution results at the pedestrian level (about 1 m above the ground). We now have several solutions to reduce the number of Z cells and thus the simulation time:

Solution A:
Use the telescoping option to stretch cells that are not in scope, i.e. starting above the highest building height, we will only stretch air cells. A stretching factor, to be specified in percent, is then applied to the size of each previous cell: cell sizes thus increase rapidly. This way, the height of the model area is achieved with fewer Z-cells. Using our example above, with a stretch factor of 20% and a starting height of 60 meters, we now need only 45 cells instead of 100 cells to reach a model surface height of more than 200 meters. However, we would start stretching within the building height range to achieve this goal.

Solution B:
Use the Split option to split the grid cell closest to the ground into 5 cells. The cells of particular interest at pedestrian level are now available at high resolution, while the default vertical resolution can be changed to a more coarse value, such as 5 m. Back to our example, with splitting enabled and a resolution of 5 m, we now only need 41 cells to reach more than 200 m in model domain height, and we still have high resolution outputs at pedestrian level (i.e., at 1.5 m height).

Solution C:
Use a combination of telescoping (Solution A) and splitting (Solution B) with 5 m vertical resolution and the telescoping settings described in Solution A. We now only need 22 cells to reach more than 200 m in model area height. However, now that we have saved so many Z-cells, we could try to improve stability even further by adding a few more Z-cells to provide more open air cells above the tallest building. We could also consider increasing the vertical resolution to 3 m or 4 m, or choosing a higher start height for telescoping / using a lower telescoping factor.

We strongly recommend that you check these settings before you start digitizing your model area. Digitized 3D information, such as special facade elements, may be lost if the vertical gridding has to be changed again at the end (and converted back from 3D to 2.5D mode). Use the Model Inspector on the Tools tab in Spaces to find the perfect vertical resolution for your model area.

Yes, ground profiles are only visually covered by terrain in Spaces, but are still used correctly in the simulation.

Please review the CSV file carefully. Time steps must be in 30 minute intervals, should not appear more than once, and no time step should be missing. All columns must exist as shown in the template image, even if they are empty or contain invalid data.

Make sure you have selected the correct value separator and decimal separator for the text file import. Check the units of the imported values. Date and time must be in the correct structure (example: 08:00:00, not 8:00:00). Temperature must be in Kelvin instead of Celsius.

The default values for these two parameters generally do not need to be adjusted. If the specific humidity values become rather large due to high humidity near the ground (this can happen in both Simple and Full Forcing), you can lower the specific humidity value at 2500 m to about 8 g/kg to avoid instabilities in the simulation.

Everything is fine as long as the Windows Task Manager shows that the ENVI-core program is still using the CPU. ENVI-met does not respond to Windows messages such as “Redraw yourself” during simulation to save processing time. It will refresh the simulation window from time to time.

The simulation time depends strongly on three aspects

Hardware: how many cores are there, what is the CPU speed, how much RAM is available. Also note that only some versions allow parallel processing.

Simulation settings: e.g. pollutant simulations with active chemistry take longer than standard simulations.

Model domain: Even if the model domain is rather small in horizontal dimensions, many users choose a very high vertical resolution. This can massively increase the simulation time and is not necessary in most cases.

See the question “How do I find the best vertical gridding for my model domain without using too many z-cells?” in the Spaces section for more details.

When this problem occurs, it is most likely due to numerical instability. Billions of operations are performed in ENVI-met every minute, and yet every user (including us) wishes for even greater complexity.

Almost all variables calculated within the simulation depend on other spatial and temporal factors, which can lead to a rogue variable containing unrealistic values. The software corrects this in most cases, but it can also happen that the variable unexpectedly goes to a value such as zero before it is used for a division in the next operation, triggering an error.

This is not because of a programming error, but because the data set was corrupted during the calculations. It is not possible to check the validity of the data before each operation, as this would dramatically increase the calculation time. An immense number of “intelligent” routines have been introduced into ENVI-met to automatically correct the most common problems. After all, it is a sophisticated numerical tool, and these routine errors are inherent in numerical modeling.
You may also want to check our support center where many individual cases are answered.

There is no general reason why a model may not run correctly. In most cases, you will need to try different things to get a stable simulation when errors occur. However, if your configuration is not working, here are a few things to check:

Was Windows running properly when the simulation crashed? ENVI-met allocates large amounts of memory for data storage. If a program crashes or Windows has serious problems before or during the model run, stored data may be lost. Do not run ENVI-met if you are low on memory, and make sure that the simulation is running in your physical memory, NOT in virtual memory.
Does ENVI-met crash on startup? Check the simulation log output on the screen. Use the Check Model option to generate a review of the output. Check that the input files and database files are OK and contain realistic values. Problems with model domain design often cause the simulation to crash.
Check the meteorological boundary conditions (especially important for full forcing!): Do the radiation values seem unreasonable? Are the wind speeds too low (<0.8 m/s) or too high (> 5 m/s)? Does the wind direction change rapidly and strongly between time steps (e.g. from 0° to 180° within one hour)? Are the relative humidity values too high for your high temperature values, resulting in a very high specific humidity?

Some additional tips to promote a successful simulation are

Move complex buildings away from the model boundary
Increase the vertical extent of the model
Decrease the time step if the model becomes unstable in the normal calculation loop.
Simplify your model, complex geometries of certain buildings can be adjusted to a simpler configuration.

This feature is not currently implemented. If a simulation is aborted, it must be restarted from the beginning.

This is not yet possible.

If the license is from a version prior to ENVI-met 4.4.5, the BIO-met license will only work for the old BIO-met version 1.5, which is available separately from our homepage (https://envi-met.info/doku.php?id=files:downloadv4).

The new BIO-met 2.0, which is included in the latest ENVI-met versions starting with 4.4.5, works in parallel mode (faster). A new license is required to use this version.

UTCI cannot be calculated for wind speeds below 0.5 m/s. For more information, visit https://envi-met.info/doku.php?id=apps:biomet_utci.

Pollutant simulations in ENVI-met are complex. There are many circumstances that may cause some or all pollutants to be simulated.

Please read this document about air pollution simulations (http://www.envi-met.info/doku.php?id=kb:sources) to learn more.

If Terrain is used together with Splitting and the Leonardo map is extracted with the option “Follow Terrain”, the elevation in meters is only valid for cells with an elevation of 0 m. As an example, let us take a model with a vertical resolution of 2 m, which we want to extract in Leonardo with Follow Terrain for the elevation level 3 (since the index counting in Leonardo starts at 0, we extract the 4th cell above the ground):

Cell A: This cell has a terrain height of 0 m. The height information is extracted at a height of 1.4 m.
Cell B: This cell has a terrain height of 2 m. Height extraction takes place at a height of 9 m.

Leonardo always shows the height that would be extracted without terrain, and therefore (in this example) it shows 1.4 m above ground. This means that if a 5 m high building is placed in the model, it would only be displayed in cell A and not in cell B.

It is recommended to disable splitting when using terrain in simulations to avoid these complications.

The Air Temperature variable is now called Potential Air Temperature. This is a more accurate name for the variable, since ENVI-met always simulates at standard air pressure. However, it is planned to add air pressure as a variable in the future and thus there will be a difference between potential and absolute air temperature.

More information about all output variables of the atmosphere folder can be found here: https://envi-met.info/doku.php?id=filereference:output:atmosphere. Output variable descriptions of the other output folders are also found on the https://envi-met.info website.