eht-met-forecast creates radio-astronomy-relevant weather forecast graphs for the Event Horizon Telescope.

This code downloads raw weather forecast data from the NOAA GFS weather model. It then runs this forecast through Scott Paine's AM radiative transfer code to compute values of interest to radio astronomy, such as the opacity at millimeter frequencies. These derived values are then archived. These values are also used to create graphs of near-future weather, which are embedded in a webpage. These webpages are then used by the EHT Array Operations Center for daily GO/NO-GO decisions during our observations.

% eht-met-forecast -h
% eht-met-forecast --backfill 168 --dir . --vex Sw --wait --log LOG

This code depends on two dependencies, pygrib and am. The azure pipelines configuration file for this repo shows a working solution for both in the azure pipelines Ubuntu and MacOS-based environment. Here are some rough notes:

The python pygrib package requires a few OS packages:

# Ubuntu 18.04 or later -- libeccodes isn't available earlier -- tested in the CI
apt-get install libeccodes-dev proj-bin libproj-dev libcairo2-dev
# RedHat flavored distros -- not tested
yum install eccodes-devel proj proj-devel cairo-devel
# Homebrew -- tested in the CI on MacOS
brew install eccodes proj cairo
# conda-forge: guesses, not tested
conda install -c conda-forge eccodes proj cairo

Once these OS packages are installed, the following sequence is needed in order:

• pip install cython
• pip install pygrib

The setup.py file does this for you.

Finally, on MacOS XCode 12 and later, you need:

export CFLAGS="-Wno-implicit-function-declaration"

because cython generates C code that triggers this warning (which becomes a fatal error).

The am code is straightforward C and does not require any unusual libraries.

The included Makefile has instructions to build it:

make am12.2

Before running eht-met-forecast, you need to set an environment variable:

export AM=./am-12.2/src/am
$AM -v

Once the OS pacakges for pygrib and am are installed,

pip install .
pip install .[test]  # if you want to run tests
pytest

• almost all configuration is in *.sh
• see [ANNUAL.md] for clues about the annual cycle
• this cronjob line is optimal for GFS's cycles (assumes UTC)

# UTC
3 1,7,13,19 * * * "bash ~/github/eht-met-forecast/weatherwrapper.sh"

The file tau255.txt is downloaded from the vlbimon website at Radbaud.

#7 0,12 * * * bash ~/github/eht-met-forecast/eurodownload.sh # UTC
#17 7,19 * * * bash ~/github/eht-met-forecast/eurodownload.sh # UTC
7 4,16 * * * bash ~/github/eht-met-forecast/eurodownload.sh # PST
17 11,23 * * * bash ~/github/eht-met-forecast/eurodownload.sh # PST
#7 5,17 * * * bash ~/github/eht-met-forecast/eurodownload.sh # PDT
#17 0,12 * * * bash ~/github/eht-met-forecast/eurodownload.sh # PDT

This code mainly a refactoring of Scott Paine's sma-met-forecast repo. The graph code is descended from code written by Scott Paine and Lindy Blackburn. Many members of the EHT Collaboration have made helpful comments.

CK Chan and Phani Datta Velicheti have independently refactored Scott Paine's code.

Alex Raymond has used MERRA-2 historical weather data to evaluate potential next-generation EHT sites.

Our European colleagues have similar code to process a European weather forecast.