RLS-based data assimilation with unknown uncertainty

Implementation of the Recursive Least Squares (RLS)-based uncertainty quantification and least-squares data assimilation algorithms.

The algorithms are described in the following papers:

• Lightweight Assimilation of Open Urban Ambient Air Quality Monitoring Data and Numerical Simulations with Unknown Uncertainty, DOI: 10.1007/s10666-023-09909-x (experiments from example1.py).
• Lightweight Open Data Assimilation of Pan-European Urban Air Quality (experiments from example2.py).

Python: 3.*

Testing

To be able to reproduce the results presented in the paper, install requirements from requirements.txt. It will install the packages needed to read and plot the data.

Project package dependencies can be installed using pip and a virtual environment manager:

Create and activate a virtual environment:

python -m venv rls-venv
source rls-venv/bin/activate

In the virtual environment, install requirements from the requirements.txt file:

pip install -r requirements.txt

Lightweight Assimilation of Open Urban Ambient Air Quality Monitoring Data and Numerical Simulations with Unknown Uncertainty

After installing the requirements, run the script example1.py:

python example1.py

The script will generate plots in plots/Liivalaia/ and print metrics for the autumn Tallinn dataset data/liivalaia_aq_meas_with_forecast.csv. Modify plots' location to plots/Liivalaia2/ for the winter Tallinn dataset data/liivalaia_aq_meas_with_forecast2.csv.

The IoT PM10 sensor data corresponds to the data/liivalaia_pm10_iot.csv dataset. The plots for the sensor data are located in the plots/Liivalaia/IoT directory.

Sequential Assimilation of Open European Urban Ambient Air Quality Data of Different Scales with Unknown Uncertainty

After installing the requirements, run the script example2.py:

python example2.py

The script run the experiments and prints the validation results described in the paper. The plots are generated for the data/eu-eq.csv dataset file. The statistics are collected for the datasets from the data/Europe_AQ/ directory.

Repository content

rls_assimilatiion is the lightweight package for uncertainty quantification and least-squares data assimilation.

The used data is stored in the data/ directory, plots are generated to plots/ directory.

Directory download/ contains script to download data from the SILAM cloud storage.