rolodex.forecast.forecast_index

Module Contents

rolodex.forecast.forecast_index.Timestamp

rolodex.forecast.forecast_index.Timedelta

rolodex.forecast.forecast_index.create_lazy_valid_time_variable(*, reference_time, period)

class rolodex.forecast.forecast_index.Model

Bases: enum.StrEnum

Enum where members are also (and must be) strings

HRRR

class rolodex.forecast.forecast_index.ModelRun(time)

The complete results for a single run is a model run dataset.

Parameters:

time: Timestamp-like

Initialization time for model run

time: pandas.Timestamp

get_indexer(model, time_index, period_index)

class rolodex.forecast.forecast_index.ConstantOffset(step)

A constant offset dataset is created from all the data that have the same offset time. Offset here refers to a variable usually named step or lead or with CF standard name forecast_period.

step: pandas.Timedelta

get_indexer(model, time_index, period_index)

class rolodex.forecast.forecast_index.ConstantForecast(time)

A constant forecast dataset is created from all the data that have the same forecast/valid time.

Parameters:

time: Timestamp-like

time: pandas.Timestamp

get_indexer(model, time_index, period_index)

class rolodex.forecast.forecast_index.BestEstimate

For each forecast time in the collection, the best estimate for that hour is used to create a best estimate dataset, which covers the entire time range of the collection.

asof: pandas.Timestamp | None = None

offset: int = 0

__post_init__()

get_indexer(model, time_index, period_index)

class rolodex.forecast.forecast_index.Indexes

reference_time: xarray.indexes.PandasIndex

period: xarray.indexes.PandasIndex

get_names()

class rolodex.forecast.forecast_index.ForecastIndex(variables, valid_time_name, model=None)

Bases: xarray.Index

An Xarray custom Index that allows indexing a forecast data-cube with forecast_reference_time (commonly init) and forecast_period (commonly step or lead) dimensions as _Forecast Model Run Collections_.

Examples

To do FMRC-style indexing, you’ll need to first add a “valid time” variable.

>>> from rolodex.forecast import (
...     BestEstimate,
...     ConstantForecast,
...     ConstantOffset,
...     ForecastIndex,
...     ModelRun,
...     Model,
...     create_lazy_valid_time_variable,
... )
>>> ds.coords["valid_time"] = create_lazy_valid_time_variable(
...     reference_time=ds.time, period=ds.step
... )

Create the new index where time is the forecast_reference_time dimension, and step is the forecast_period dimension. >>> newds = ds.drop_indexes([“time”, “step”]).set_xindex(

[“time”, “step”, “valid_time”], ForecastIndex, model=Model.HRRR )

>>> newds

Use valid_time to indicate FMRC-style indexing

>>> newds.sel(valid_time=BestEstimate())

>>> newds.sel(valid_time=ConstantForecast("2024-05-20"))

>>> newds.sel(valid_time=ConstantOffset("32h"))

>>> newds.sel(valid_time=ModelRun("2024-05-20 13:00"))

valid_time_name

model = None

names

classmethod from_variables(variables, options)

Must be created from three variables: 1. A dummy scalar forecast variable. 2. A variable with the CF attribute`standard_name: “forecast_reference_time”. 3. A variable with the CF attribute`standard_name: “forecast_period”.

sel(labels, **kwargs)

Allows three kinds of indexing 1. Along the dummy “forecast” variable: enable specialized methods using

ConstantOffset, ModelRun, ConstantForecast, BestEstimate

2. Along the forecast_reference_time dimension, identical to ModelRun

3. Along the forecast_period dimension, identical to ConstantOffset

You cannot mix (1) with (2) or (3), but (2) and (3) can be combined in a single statement.

__repr__()