"""Consumer groups data."""
import logging
from copy import deepcopy
from typing import TYPE_CHECKING, Dict
import pandas as pd
import xarray as xr
from genno import Quantity
if TYPE_CHECKING:
from genno import Computer
log = logging.getLogger(__name__)
[docs]def prepare_computer(c: "Computer") -> None:
"""Prepare `rep` for calculating transport consumer groups."""
from .key import cg, pop_at
c.add("indexers:n-cd", "indexers_n_cd", "config")
# Population shares by area_type
c.add(pop_at, urban_rural_shares, "population:n-y", "config")
# Exogenous data for consumer group sizes
keys = c.infer_keys(
[
"population suburb share:*:exo",
"ma3t attitude:*:exo",
"ma3t driver:*:exo",
"ma3t population:*:exo",
]
)
c.add(cg, cg_shares, pop_at, *keys, "indexers:n-cd", "indexers:cg")
[docs]def cg_shares(
ursu_ru: Quantity,
su_share: Quantity,
ma3t_attitude: Quantity,
ma3t_driver: Quantity,
ma3t_pop: Quantity,
n_cd_indexers: Dict[str, xr.DataArray],
cg_indexers: Dict[str, xr.DataArray],
) -> Quantity:
"""Return shares of transport consumer groups.
Parameters
----------
ursu_ru : Quantity
Population shares with "UR+SU" and "RU" on the ``area_type`` dimension.
su_share : Quantity
Share of suburban ("SU") population within "UR+SU".
ma3t_attitude: Quantity
ma3t_driver : Quantity
ma3t_pop : Quantity
Population shares between urban, suburban, and rural.
DLM: “Values from MA3T are based on 2001 NHTS survey and some more recent
calculations done in 2008 timeframe. Therefore, I assume that the numbers here
are applicable to the US in 2005.”
NB in the spreadsheet, the data are also filled forward to 2110.
Currently not used.
n_cd_indexers : dict
cg_indexers : dict
Returns
-------
.Quantity
Dimensions: n, y, cg. Units.dimensionless.
"""
from genno.operator import concat, mul
cg_indexers = deepcopy(cg_indexers)
consumer_group = cg_indexers.pop("consumer_group")
# Assumption: split of population between area_type 'UR' and 'SU'
# - Fill forward along years, for nodes where only a year 2010 value is assumed.
# - Fill backward 2010 to 2005, in order to compute.
su_share = su_share.ffill("y").bfill("y")
# Split the 'UR+SU' population share to "UR", "SU" using `su_share`
assert {"UR+SU", "RU"} == set(ursu_ru.coords["area_type"].data)
pop_share = (
concat(
ursu_ru.sel(area_type="UR+SU", drop=True) * (1 - su_share),
ursu_ru.sel(area_type="UR+SU", drop=True) * su_share,
ursu_ru.sel(area_type="RU", drop=True),
dim=pd.Index(["UR", "SU", "RU"], name="area_type"),
)
.ffill("y")
.bfill("y")
)
# - Compute the group shares.
# - Select using matched sequences, i.e. select a sequence of (node,
# census_division) coordinates.
# - Drop the census_division.
# - Collapse area_type, attitude, driver_type dimensions into consumer_group.
# - Convert to short dimension names.
groups = (
mul(pop_share, ma3t_attitude, ma3t_driver)
.sel(n_cd_indexers)
.sel(cg_indexers)
.assign_coords(consumer_group=consumer_group.values)
.rename(dict(node="n", year="y", consumer_group="cg"))
)
# Assert that the sum across groups is 1
assert isinstance(groups, Quantity)
assert (groups.sum("cg") - 1 < 1e-5).all(), "Groups not balanced"
return groups
[docs]def urban_rural_shares(pop: Quantity, config: dict) -> Quantity:
"""Return shares of urban and rural population.
Parameters
----------
pop : .Quantity
Must have n and y dimensions.
config : dict
The ``regions`` and ``data source/population`` keys are used.
Returns
-------
.Quantity
Dimensions: at least area_type, possibly also n, y. Units: dimensionless.
"""
from genno.operator import div
from message_ix_models.util import broadcast
if "area_type" in pop.dims:
result = div(
pop.sel(area_type=["UR+SU", "RU"]), pop.sel(area_type="total", drop=True)
)
else:
log.warning("Population data lack 'area_type' dimension")
df = (
pd.DataFrame([["UR+SU", 0.6], ["RU", 0.4]], columns=["area_type", "value"])
.assign(n=None, y=None)
.pipe(broadcast, n=pop.coords["n"].data, y=pop.coords["y"].data)
.set_index(["area_type", "n", "y"])
)
result = Quantity(df)
assert {"area_type", "n", "y"} == set(result.dims), result.dims
return result