Jump to main content or area navigation.

Contact Us

Atmospheric Modeling and Analysis Research

Regional Climate Downscaling

Global Climate Models (GCMs) are used to study Earth's climate system and to simulate how climate may change in the future. Regional Climate Models (RCMs) are used to simulate Earth's climate system at a higher spatial resolution over a limited area.To meet the growing need for regional climate projections to support impact assessments, EPA is developing regional climate modeling capabilities using the Weather Research and Forecasting (WRF) model to dynamically downscale GCM simulations.

One way to develop a downscaling methodology is to apply the technique to a historical coarse-scale meteorological “reanalysis,” which is a blend of observational and model data.  For a given spatial scale, the reanalysis can be regarded as the best available representation of the meteorological fields that occurred during that time period.  Though observations that are generally available can be used to create accurate fine-scale fields for current and recent-past periods, this is not the case for the future.  Accordingly the downscaling process is tested using a coarse-scale reanalysis, comparable in resolution to that used by GCMs, with the results evaluated by comparing against finer-scale data.

When using coarse-scale data from a reanalysis or GCM as lateral boundary conditions for a regional model without further constraint, the interior meteorological fields simulated by the regional model can deviate significantly from those of the driving fields. Four-dimensional data assimilation or "nudging" techniques provide one way to constrain the RCM and keep it from diverging too far from the coarse-scale fields.

If the regional model is constrained too strongly to the GCM fields, however, there is the possibility that the benefit of using the higher-resolution RCM will not be realized. A delicate balance is needed between the amount of constraint given to the RCM and the freedom of the RCM to simulate its own mesoscale features.

EPA has used the WRF model to downscale fields from the 2.5 degree x 2.5 degree NCEP-AMIP II Reanalysis (hearafter, "R-2") for the 20-year period 1988-2007.  In these experiments, WRF has been used without nudging and with either analysis nudging or spectral nudging — two types of interior nudging available in WRF. 

The results show that large biases exist when using WRF as an RCM in the absence of interior nudging, and that these biases are greatly reduced when either type of nudging strategy is employed.  Furthermore, subsequent work has shown that extreme temperatures and precipitation events are also better simulated when interior nudging is used.

Difference in 300hPa wind fields

Seasonally-averaged (April-June) wind fields (meters per second) at 300 hectopascals as simulated by:

  1. North American Regional Reanalysis;
  2. WRF without nudging;
  3. WRF with analysis nudging; and
  4. WRF with spectral nudging.
Analysis nudging improves WRF's ability to simulate the location and intensity of the jet stream (Bowden et al., 2012).

The methodology used to downscale the R-2 reanalysis is now being used to downscale GCM simulations being conducted for the Intergovernmental Panel on Climate Change Fifth Assessment Report. Downscaling of the NASA Goddard Institute for Space Studies Model E2 has been completed, and tools needed to downscale the National Oceanic and Atmospheric Administration (NOAA) Geophysical Fluid Dynamics Laboratory (GFDL) Climate Model version 3 (CM3) and National Center for Atmospheric Research (NCAR) Community Earth System Model (CESM) are under development. Preliminary simulations indicate that interior nudging is also needed to maintain large-scale consistency between driving GCM fields and those simulated within the WRF model.

Differences in July average 500-hPa geopotential height simulated by WRF and ModelE

Mean July 500 hectopascals geopotential height in meters for:

  1. GISS ModelE;
  2. base WRF run, without any interior nudging;
  3. WRF with analysis nudging; and
  4. WRF with spectral nudging.
Though both nudging techniques are only applied above the planetary boundary layer, both serve to keep the 500 hectopascals geopotential height simulated by WRF from diverging as far from ModelE.
Differences in July average 2-m temperature simulated by WRF and ModelE

Mean July 2-m temperature in kelvins for

  1. GISS ModelE;
  2. base WRF run, without any interior nudging;
  3. WRF with analysis nudging; and
  4. WRF with spectral nudging.
Without nudging, average near-surface temperatures simulated by WRF for the Pacific Northwest are more than 6 kelvins warmer than in the GCM.

Contacts: Tanya Otte, Chris Nolte

Related Publications

Jump to main content.