stilt.observations.TransportError#

class stilt.observations.TransportError(variance: 'float', noise: 'float', enhancement: 'float', enhancement_perturbed: 'float', levels: 'pd.DataFrame', length_scale: 'float | None', background: 'float' = 0.0, realizations: 'int' = 1) → None[source]#

Result of transport_error().

All values are in the flux’s units times the footprint’s (ppm for a flux in µmol m⁻² s⁻¹).

variance is the signed transport-error variance of the modelled enhancement: the extra ensemble variance the wind perturbation produced. A negative value is sampling noise. noise is the standard deviation of variance expected with no perturbation at all, estimated from random halves of the unperturbed particles; variance is resolved only when it is several times noise. sd is sqrt(variance), or 0 when the variance is negative.

realizations is how many error realizations went into the estimate. Their level variances are averaged before the difference is taken, so variance has less noise than a single realization’s, and noise is scaled to match (see transport_error()).

enhancement and enhancement_perturbed are the modelled enhancement from the unperturbed and the perturbed particles (the perturbed one averaged over the realizations). levels has one row per release level: height (m, mean release height), n particles, weight (its share of the column), mean / var of the per-particle enhancement without (_orig) and with (_err) the perturbation, dvar their difference, and sd_trans the signed square root of dvar (or X-STILT’s regression-scaled value with regression=True).

With a background field the per-particle values are the modelled mole fraction, enhancement plus background at the particle’s endpoint, so enhancement and enhancement_perturbed are then modelled mole fractions and variance includes the background’s response to the wind errors. background is the weighted background from the unperturbed particles (0 when no field was given), so enhancement - background is the enhancement alone.

Attributes#

`background`
`realizations`
`sd`	Transport-error standard deviation; `0` when `variance` is negative.
`variance`
`noise`
`enhancement`
`enhancement_perturbed`
`levels`
`length_scale`