# Outputs

This page describes generated outputs (CSV tables, LUTs, corrected fluxes, and diagnostic figures).

## Output folder structure

Outputs are written to the folder configured by `output_path`. In the repository example run, outputs are
available under:

`examples/output/`

Filenames include a site name and run metadata (gas, workflow, time window, timestamp).

## CSV outputs

The following CSV files are typically produced.

### Cut-off frequency LUT

- `4_LUT_cof__<site>__<gas>__<mode>__... .csv`

This table stores the cut-off frequency look-up table used by the workflow. In the example run it contains
one row per wind-direction / wind-speed class (depending on configuration), with columns such as:

- `wd_mean`, `wd_max`, `unc_wdclass`
- `ws_mean`, `ws_max`
- `cof_L`, `cof_G`: cut-off frequencies for the Lorentzian / Gaussian transfer-function fits
- `fn_L`, `fn_G`: normalized frequencies
- `unc_L_tf`, `unc_G_tf`: transfer-function-related uncertainty on the fitted cut-off frequencies (per class)

### Mean (co)spectra

- `4_mean_cosp__<site>__<gas>__<mode>__... .csv`

This table stores averaged (co)spectra per class and frequency bin. It is primarily used for diagnostics
and plotting.

### Correction factor LUT

- `5_LUT_CF__<site>__<gas>__<mode>__... .csv`

This table stores correction factors (stable/unstable) aggregated by class. In the example output, the file
contains two blocks (`unstable` then `stable`), each providing (among others):

- `wd_mean`, `ws_mean`
- `CF_L`, `CF_G`: correction factors for Lorentzian / Gaussian approaches
- `unc_L_tf`, `unc_G_tf`: transfer-function-related uncertainty component (computed as the maximum deviation from the central CF value "M")
- `unc_L_sd`, `unc_G_sd`: within-class spread of CF values (standard deviation computed after outlier removal)

Note: total uncertainties are not written to the output anymore; users can recompute them as needed from the component uncertainties.

### Corrected flux output

- `6_flux_out__<site>__<gas>__<mode>__... .csv`

This is the main half-hourly output time series. In the example output it includes:

- `Timestamp`
- meteorological descriptors used for classification (e.g. `wind_speed (m s-1)`, `wind_dir (deg)`, `(z-d)/L (-)`)
- original flux columns (e.g. `Fc (µmol m-2 s-1)`, `FcEP (µmol m-2 s-1)`)
- corrected fluxes (e.g. `FCcorL`, `FCcorG`)
- correction factors applied (`CF_L`, `CF_G`)

## Diagnostic figures

When plotting is enabled, FREQCOR generates diagnostic figures to help evaluate selection steps, transfer
function fitting, LUT construction, and correction-factor behavior.

The following figures are shown from the example BE-Lon CO2 cospectral run.

### Individual (co)spectra

This figure shows individual (co)spectra contributing to the computation (after initial selection), used to
evaluate variability and potential outliers.

![Individual cospectra (example output)](../{{ ex_2_all_individual_co2 }})

### Filtering diagnostics

These figures illustrate how data were filtered when estimating cut-off frequencies and correction factors.

![Cut-off frequency filtering (example output)](../{{ ex_3_filtering_cof_gas }})

![Correction factor filtering (unstable) (example output)](../{{ ex_3_filtering_cf_h_unst }})

### Mean transfer functions and mean cospectra

These figures summarize transfer functions and (co)spectra averaged over classes.

![Mean transfer functions by class (example output)](../{{ ex_4_mean_tf_all_classes }})

![Mean cospectra by class (example output)](../{{ ex_4_mean_cosp_all_classes }})

### Correction factors

These figures show the correction factors as a function of wind speed for stable and unstable regimes.

![Correction factors vs wind speed (stable) (example output)](../{{ ex_5_cf_vs_ws_st }})

![Correction factors vs wind speed (unstable) (example output)](../{{ ex_5_cf_vs_ws_unst }})