Downloads the metoffice rainfall radar images, summarises them daily, weekly, monthly, and yearly, and creates time laps videos for selected areas.
If pdflatex is installed, it will also create pdf with the summarised images for the selected areas to provide a nice overview. In the pdfs the centre of the images is marked with a small white dot.

Installation/Requirements

git clone https://github.com/ronnyerrmann/rainfall
cd rainfall
python -m venv rain_venv
. rain_venv/bin/activate
pip install Image requests opencv-python psutil

Add crontab to run automatically every 8 hours:

1 */8 * * * /home/<user>/rainfall/rain_venv/bin/python /home/<user>/rainfall/rainfall_images.py

Run

rain_venv/bin/python rainfall_images.py

All data will be stored in the folder of the script.

As data only a few days back is available, the script should be run at least every 24 hours. For courtesy, do not run it more than once per hour.

Description

For the daily images the same scale as metoffice uses in their rainfall radar is used, but instead of indicating mm/h it's mm/day. Additionally purple for 64 to 128 and white for 128 to 999 mm/day were added. If more than that value is reached, black will be shown, and you should build am arch.

As the publicly available data only provides ranges from x to 2x mm/h (with x being the lower rainfall limit for a colour), the resulting summaries for the day can be off by a similar amount. The program uses the value at 33% of the range to reflect the assumed Poisson distribution of rainfall and that the rainfall measurement is done at discrete times. The final values, given in the grey scaled image, have therefore a relative uncertainty of [+50% -25%].

The metoffice files are timestamped as UTC and this time is used to define a day, which means for BST (British Summer Time) the daily images cover 1:00 to 1:00.

Add your own areas

This can be done either by creating a file subareas.csv in the script folder, with comma-separated values for

1. (int) x-Pixel position in the original file in image_data, starting from 0. In the future, that should be derived from lat/lon but for now I used Irfanview to see the pixel position during a click (which is already 0 based). To know which pixel, I looked for rain reaching the position on the metoffice website.
2. (int) y-Pixel position in the original file in image_data.
3. longitude
4. latitude
5. semi-size of the box around the selected pixel, e.g. 1 leads to a 3x3 array, 10 to 21x21, a pixel being about 0.65 km in size.
6. scale of a pixel in the final image, e.g. a scale of 5 on a 3x3 array leads to a 15x15 png
7. name of the place

Alternatively, the script can be edited to add the lines containing SUBAREAS.append with the same entries (name needs to be in quotes)

Example output

The full frames of the summed daily images are stored as image_daily/YYYY-MM-DD_sum.png (using the same colour coding as metoffice) and image_daily/greyscale-mm_YYYY-MM-DD_sum.png (using 8 bit grey-scale):

The full frames of the summed weekly, monthly, and yearly images are stored as image_weekly_monthly_yearly/[wmy]YYYY-MM-DD_sum.png with w, m, y indicating weekly, monthly, and yearly, respectively. The colour coding uses the same colours for the same numeric value ranges as metoffice, but values indicate cm/week, cm/month, and dm/year with 1 cm = 10 mm and 1 dm = 100 mm (1 m = 10 dm). The grey scale images are stored in image_weekly_monthly_yearly/greyscale-<unit>_[wmy]YYYY-MM-DD_sum.png (8 bit grey-scale), using the same units as above: cm, cm, and dm.

The subframes are stored in subframes_daily/Place_Name_YYYY-MM-DD.png or subframes_weekly_monthly_yearly/Place_Name_[wmy]YYYY-MM-DD.png:

Finally, the subframes are added to a pdf document in tex/Place_Name_rainfall.pdf, starting with the yearly subframes, followed by monthly, weekly, and daily data. In it, bracketed numbers behind the date indicate the number of frames, if not all the 288 daily or 2016 weekly rainfall images could be downloaded from metoffice. For monthly and yearly summaries, the number of frames are always given. A 3-week example of the daily and a 3-month example of weekly summaries for my home (half box size is 10 pixel so the box spans about 13 km) is shown below as image (showing that it often rains less at home than the in surrounding area) and here is the whole pdf for of Place.