Software
The software used to produce FP2026.1, by processing step. Versions are given where the notebooks or the project’s lock file record them.
| Step | Software | Version | Use |
|---|---|---|---|
| Raw data | bico | to be checked | Conversion of the raw binary eddy covariance files to ASCII. The wrong-calibration-gas factor (0.974 on the CO2 concentration, 14 December 2017 to 15 March 2019) is applied in this step. |
| Level 0 and Level 1 | EddyPro (LI-COR), run with fluxrun | to be checked | Flux calculation, one run per setup period. Level 0 is a preliminary run; Level 1 is the final run whose FLUXNET output enters the dataset. |
| Merging, post-processing, meteo products, finalize | diive (documentation) | v0.93.0 | Merging of the Level-1 runs, quality flags, storage term, outlier removal, USTAR filtering, gap-filling (XGBoost and MDS), uncertainty, the meteorological products and the final dataset. |
| USTAR threshold detection | REddyProc (R) | to be checked | Detection of the USTAR thresholds with sEstimateUstarScenarios, on the NEE of the LI-7200 (2016-2024) and of the LI-7500 (2005-2019); see Level 3.3 for how the three constant thresholds were chosen from both. The input was prepared with diive v0.90.0. |
The version of diive is the one printed by every notebook of the IRGA72 flux chain (20_L1_FLUXES/IRGA72, 30_FLUX_POST-PROCESSING_CHAIN/IRGA72), of the meteo products 01 to 07 and 99, and of 80_FINALIZE. The meteo products 08 to 13 were last run with diive v0.91.0 and v0.91.1.
Python environment
The notebooks run in Python 3.12 in an environment managed with uv. The project declares two direct dependencies in pyproject.toml, diive (with its database extra) and SciPy; everything else arrives through diive. Versions from uv.lock:
| Package | Version | Use in this repository |
|---|---|---|
| diive | 0.93.0 | all processing steps above |
| pandas | 3.0.3 | data handling in every notebook |
| NumPy | 2.3.5 | numerical work |
| SciPy | 1.17.1 | trend statistics (Theil-Sen slopes, Kendall’s tau) |
| XGBoost | 3.2.0 | gap-filling models, called through diive |
| scikit-learn | 1.8.0 | used in the meteo stage |
| Matplotlib | 3.10.9 | figures |
| PyArrow | 19.0.1 | parquet files |
| influxdb-client | 1.50.0 | access to the site database in the meteo screening |
The documentation website is built with Quarto (quarto-cli 1.9.38).
Repository
All notebooks are in the repository under workflow/, and rendered on this site under Notebooks.