Return periods

glofas_ewds.return_period

Return levels of a river from its own GloFAS history: the 2-, 5-, 20-year flow.

Pure: nothing here does I/O.

The EWDS does not serve GloFAS's flood thresholds, and the 90th percentile of two years (:func:~glofas_ewds.reference.references) is "an ordinary high flow", not "a flood that comes once in five years". This module fits a Gumbel distribution to the annual maxima of a node's reanalysis, by L-moments, and reads the levels off it.

Two cautions. These are not the official GloFAS thresholds: same family of method, but your years, your cells, your fit. And a fit needs decades: the reanalysis starts in 1979, and fewer than min_years complete years give None.

annual_maxima

annual_maxima(by_day: Mapping[date, float], *, min_days: int = MIN_DAYS_PER_YEAR) -> dict[int, float]

The highest value of each calendar year with at least min_days values.

by_day is one node of :func:~glofas_ewds.glofas.discharge_by_day.

gumbel_fit

gumbel_fit(maxima: ArrayLike) -> tuple[float, float]

Location and scale of a Gumbel, by L-moments (Hosking 1990).

Needs at least two values.

return_level

return_level(location: float, scale: float, years: float) -> float

The flow exceeded on average once every years years.

return_levels

return_levels(maxima: Iterable[float], *, periods: Sequence[float] = RETURN_PERIODS, min_years: int = MIN_YEARS) -> dict[float, float] | None

{period: level} from a node's annual maxima; None below min_years or when the maxima do not vary (a dry cell).