ASOS Data Analysis in Python

Monthly-averaged diurnal temperature variability for a single ASOS surface station, from a raw NOAA/NCEI .dat file — an introduction to numpy and matplotlib on data that has real gaps in it.

Getting the data

The script reads 64060KLGA201807.dat — station KLGA (LaGuardia, New York), July 2018, in NCEI's ASOS 5-minute (DSI-6406) format. The file is not in this repository; download the month you want from NOAA/NCEI and either keep the same name or edit the filename at the top of monthly_asos_avg.py.

Running it

pip3 install numpy matplotlib
python3 monthly_asos_avg.py

It writes asos_analysis_python.png at 200 dpi and opens the plot.

What the script does, and the two choices in it worth knowing

1. It keeps only rows whose third field is NP and skips everything else. That field is a record flag, and NP selects the common case; confirm its exact meaning against NCEI's DSI-6406 format documentation before reusing this parser for anything precipitation-related. Whatever it denotes, the average below is conditioned on it rather than taken over every record.
2. The timestamp is parsed by character offset, not by delimiter ([3:7] year, [7:9] month, [9:11] day, [11:13] hour, [13:15] minute). That is what makes a fixed-width record readable in three lines, and it is also what breaks silently if you feed it a different product with a different column layout. Check one line by hand before trusting a month.

Then it is plain numpy: bucket the observations by hour, take np.mean and np.std per bucket, and shade ±1 standard deviation with plt.fill_between. The scatter underneath is every 5-minute observation in the month, which is what makes the spread legible rather than asserted.

---

Part of the Maker Portal open-source scientific computing and hardware ecosystem. Explore interactive calculators and engineering tools at makerportal.ai.