hourly_kernel_qdm() creates a complete multi-year future-weather method
from matching three-hourly historical and future model data plus an hourly
observed reference. CMIP6 model roles require point-sampled 3hrPt tas,
ps, huss, uas, and vas, together with interval-mean 3hr rsds
and rsdsdiff. Daily tasmin and tasmax are optional interpolation
anchors. The observed role requires tas, ps, hurs, sfcWind, rsds,
and rsdsdiff.
Continuous model state variables and wind-vector components are interpolated
to an hourly lattice. Relative humidity is derived from huss, tas, and
ps; scalar speed and meteorological direction are derived from uas and
vas. Shortwave variables are allocated with solar geometry before the
centered three-month kernel-density Quantile Delta Mapping calculation.
Every corrected future-model year is then mapped to the 365-day EPW calendar
and passed through the common absolute_model_fields physical policy, with
the future model's wind direction retained alongside the corrected speed.
The method returns one addressable output member per complete future-model
year. Numerical kernel, bandwidth, grid, tail, and zero-denominator settings
not reported by the source publication remain explicit experimental defaults.
The high-level workflow expands historical and future extraction windows by
one source timestep so bounded interpolation can retain every requested year.
Pre-extracted ShiftClimate inputs must retain equivalent edge support.
hourly_kernel_qdm(
reference = NULL,
observed_reference = NULL,
signal_overrides = list()
)A required historical_reference(),
shift_reference_plan(), or extracted ShiftClimate stage containing
matching variable-specific 3hrPt, 3hr, and optional daily historical
model output.
A required shift_reference_plan() or extracted
ShiftClimate stage containing hourly observed weather.
Optional named list of variable-specific kernel-QDM settings passed to the signal component.
A complete ShiftMorphMethod for shift_future_epw().
Wang, Z. et al. (2023). Climate data for building simulations in EnergyPlus. doi:10.1038/s41467-023-41458-5