Troubleshooting

Here we will collect information of possible problems and how they have been solved

Problems installing the Python modules

If you get errors from the Python setuptools when installing the Python modules, such as:

"Unfortunately the Python module installation did not work, updating setuptools could help."

try upgrading pip with this command:

python3 -m pip install --upgrade pip

and then run the setup.py again.

Problems installing NetCDF4

Unfortunately installing NetCDF4 is not always smooth, and the situation could escalate like in the figure above.

Installation order seems to matter; first gfortran, then NetCDF4, then NetCDF4 for Fortran, then NetCDF4 for Python.

We will add more information when we get more success stories. Meanwhile, try to stick to the installation procedures described in the videosHow to Install ARCA!

Incompatible hdf5 libraries (problem with Windows CygWin)

For some reason CygWin is installing the latest hdf5 library instead the one defined in "packagelist", leading to incompatible installation.

Solution: After installing cygwin, manually install libhdf5_103 version 1.10.6-1. Please let us know if that didin't help!

Surface plotting not working (missing plot or displaced axis) in GUI

In terminal, type python3 -m pip list (or python -m pip list if that calls your Python 3x), and check the version of pyqtgraph. If it is 0.12.1 or larger, downgrade to 0.12.0:

python3 -m pip uninstall pyqtgraphpython3 -m pip install pyqtgraph==0.12.0

Please let us know if that didn't help!

ARCA on Windows: HD5 library version mismatch error

This is a problem that comes from incompatible hdf libraries, and seems to be related to cygwin. To fix this, start your cygwin setup tool (setup-x86_64.exe, download it from cygwin.com if you can not find it), and after you have proceeded to the main window ("Select packages"), choose "Full" for the View, and search for "libhdf5_103". There the "Current" will probably be version 1.10.7, but you should choose 1.10.6 (you do this by selecting the version in the next column).

Then click "Next" and accept whatever comes next. Then try again running ARCA (you should not have to recompile).

Garbled GUI, buttons or text not properly visible, scaling problems

On systems with screen scaling (e.g. Windows) or high resolution, the scaling sometimes does not work properly. This problem has so far been only reported from Windows users. The problem seems to occur because of scaling done by QT and OS for high dpi displays doesn't always work the way it is intended, and the behavior in different QT versions seems to be different. The solution is to modify the QT scaling factors, and finding the correct factor has to be made iteratively.

Solution 1

The QT variables affecting the scaling are QT_SCALE_FACTOR and QT_SCREEN_SCALE_FACTORS. These are set as environmental variables before the GUI object is created. Setting values to these variables can be done in the gui code, or before launching the gui. A solution is to edit the calling script run_arca.sh and add the next line before "python3 ModelLib...":

export QT_SCALE_FACTOR=n.n

where n.n is the scaling factor. You will have to try which one works best, values between 0.5 and 1.5 could do the trick. So edit, save and launch.

Doesn't make any difference? Try setting QT_SCREEN_SCALE_FACTORS:

export QT_SCREEN_SCALE_FACTORS=n.n;m.m

where n.n and m.m are factors for primary and secondary display (if present). The values are separated with semicolon. trial and error is your friend.

Solution 2 (this only affects QT_SCALE_FACTOR)

Edit the "run_arca.sh" ("run_arca.command" in Mac), and add --scaling_1.0 after "ModelLib/gui/ARCA_gui.py". After modification, the last line should look like this:

python3 ModelLib/gui/ARCA_gui.py --scaling_1.0

or

python ModelLib/gui/ARCA_gui.py --scaling_1.0

depending which "python" is used to call Python 3. You might have to experiment with a proper scaling factor (the number after "--scaling_") to get everything working.

PyQT5 Plugin not found when starting the GUI (Windows)

This problem could be solved by reinstalling Python, but usually it doesn't help. Instead, add a Windows environment variable "QT_PLUGIN_PATH" in the User variables. The "value" should be the path to Qt5 Plugin directory, and varies from one system and Python instaalation to another. It is generally in the form of:

c:/Users/<User>/AppData/Roaming/Python/<Python version>/site-packages/PyQt5/Qt/plugins

IMPORTANT! Before adding the path, replace the <User> and <Python version> with the current username and Python version (major and minor, e.g.  for Python 3.8 it would be "Python38" and make sure this path exists. Then start new cygwin session and run "sh run_arca.sh"

Why is the model not producing particles/gases?

Some possible causes:

• Check the units in tab Time dependent variables - If you have #/cm3 and your measurements are in ppb,it is not good.
• In tab Time dependent variables, maybe you have a Multi set to 0?
• Are you are missing some important compound in the input? For example, You won't produce much OH if you are missing relative humidity (in fact, the only way to get OH in dry air is with VOCs).
• Do you have a chemistry scheme that has the necessary reactions?
• Check the Date: if the day is January 1st, you will not get much sunlight if the latitude is the default 65 N. Then no matter what SW_RADIATION you send, the model thinks the sun is below horizon for most of the day (as it is in the winter here...). Switching to Index or using the proper date helps.
• If ACDC is not producing much even if the monomer concentrations are ok, the most probable reason is high background particle concentration and/or high temperature, both which affect cluster formation.

I get strange values and errors or the model crashes at some point due to SIGFPE

There can be many reasons for this, one probable is that somewhere a division with zero or square root of negative number is being tried. This should not happen, but never say never.

• Check your input. Are the values physically sensible? Are the units correct?
• Are you using the custom command NO_NEGATIVE_CONCENTRATIONS=.false. and using negative values? Note that you can give temperature in C° but they are converted to Kelvins upon startup and should then never be negative inside the model.
• You might have some extremely large concentration time derivatives, which can choke the condensation routine. Try decreasing the timestep or use optimized timestep and try different tolerances.
• Try different PSD scheme, sometimes Moving average is more stable than Fixed grid but opposite could be true sometimes.
• Try increasing the number of aerosol bins
• Try decreasing the number of aerosol bins

Let us know if you cannot find a reason from any of the above. Use the Export current case tool to save the model parameters.

Mac installation

Here is a link to some help: thiagodossantos.com/post/1-mac-science-software/

Garbled plot area in GUI, such as duplicate x and y axis (on Windows dual monitor)

This can be a scaling issue. Try if the problem is only appearing an one of the screens. If it happening on both screens, try if the following helps:

1) Open run_arca.sh on notepad or similar

2) on last line, after "ModelLib/gui/ARCA_gui.py", add space and -ns. It should now read:

<whatever path or call to python> ModelLib/gui/ARCA_gui.py -ns

This will override the QT scaling of the gui on Windows. You might have to set the font size from menu Help→ Set Fonts