Bias-correction is the process of mathematically scaling climate model
outputs to account for their systematic errors, in order to improve
their fitting to observations. Bias correction methods can be
mathematically simple, e.g. subtraction, or more complex, e.g. fitting
statistical curves. No climate model or data processing technique can
entirely eliminate uncertainty, but bias-correction can help reduce
mismatches between modeled gridded data and measured historical data
for point locations. Bias correction is applied to most modeled
climate data.
Downscaling is a collection of methods which can transform
lower-resolution (coarser, more pixelated) datasets into
higher-resolution products. This is useful for providing increased
detail on local conditions. Downscaling techniques can be divided into
two broad categories: dynamical and statistical.
Dynamical downscaling involves setting up boundary conditions for
a region using a relatively coarse Global Climate Model (GCM),
then creating a higher resolution regional model based on physical
principles. This method is computationally intensive.
Statistical downscaling involves defining a mathematical
relationship between historic observed climate data and the output
of GCMs or other climate models for the same time period, then
using this relationship to create future projections.
Emission scenarios reflect plausible future human greenhouse gas and
other aerosol emissions. In the past, the Intergovernmental Panel on
Climate Change (IPCC) referenced emission scenarios when describing
possible pathways for climate change. Later, Representative
Concentration Pathways (RCPs) were designed to help climate modelers
connect emissions trajectories with Radiative Forcing values. The
RCP4.5 scenario represented a future with fewer increases to emssions,
and the RCP8.5 scenario represented a future with greater global
emissions. RCP6.0 was a middle-of-the-road scenario. Today, the
scenarios used by the IPCC are known as Shared Socio-economic
Pathways, or SSPs. SSPs include information about demographics and
policy, which underlie each scenario.
Exceedance probability is the statistically determined probability
that a certain value will be exceeded in a specified future time
period. In hydrology, the exceedance probability is the inverse of the
annual recurrence interval, return interval, or return period.
A global climate model (GCM) is a mathematical representation of the
interactions and energy balance among Earth’s atmosphere, land
surface, ocean, and sea ice. Climate models divide the globe into a
three-dimensional grid of cells that interact with one another as a
coupled system. Outputs from GCMs provide long-term climate
projections.
A gridded dataset is a continuous grid-based representation of a
variable (e.g. temperature) in two dimensions (e.g latitude and
longitude). In contrast, point datasets or line (vector) datasets
offer values for only a subset of possible locations (e.g. cities,
rivers, or roads). Gridded datasets may be low-resolution (large grid
cells) or high-resolution, but do not have data gaps or
discontinuities. If gridded datasets are created from point or vector
datasets, mathematical interpolation and modeling is required.
Climate reanalyses combine past weather observations with models to
generate time series of climate variables. ERA5 is the latest climate
reanalysis produced by the European Centre for Medium-Range Weather
Forecasts (ECMWF). It provides hourly data on multiple parameters,
with estimates of uncertainty.