Use Case

Application of the Weather Research and Forecasting (WRF) Model for Climate-Relevant Simulations on the Cloud

S.C. Pryor (Cornell) with R. Levy (NASA Goddard), F. Yu (SUNY Albany), A. Hodzic (NCAR), P. Crippa (University of Newcastle), H. Matsui (Japan Agency for Marine-Earth Science and Technology), and B. Barker, P. Vaillancourt, B. Wineholt (Cornell CAC)


  • Model the variability of wind speeds and atmospheric properties

Why Aristotle?

  • Bursting to process new data
  • Sharing of a high-value processed dataset of general interest


  • Built a physics-only version of the Weather Research and Forecasting (WRF) model using Docker and compiled it with parallel NetCDF to evaluate cloud-based performance.
  • Ran high-resolution simulations to quantify wind climate and analyze the impact of large wind turbine developments on downstream climate (local to mesoscale).
  • Evaluated 10-minute wind speeds from simulations relative to in-situ measurements from the National Weather Service Automated Surface Observing System (ASOS) on Jetstream to allow Aristotle to continue to focus on the numerical simulations.
  • Analyzed high-resolution numerical simulations of the effects of wind turbines (WT) on regional climate.
  • Completed simulations to test the sensitivity of the climate impacts to the precise description of the WT aerodynamics (the extraction of momentum and introduction of turbulence behind the turbine rotor).
  • Analyzed long-term simulations with the WRF model to examine inter-annual variability of annual mean wind speeds at/near typical wind turbine hub-heights, and applied the power curve of the most commonly deployed WT to post-process the 10-minute wind speed output into estimated annual energy production (goal is to create a more robust prediction of the value of wind energy projects: analyses rendered possible by mounting a 100TB hard drive on an Aristotle instance).
  • Analyzed output from simulation of WRF model at 12Km over eastern N. America for 2001-2016 for the assessment of year-to-year variability in the wind resource (150TB WRF runs).
  • Enhanced high-resolution simulations of wind farm wakes from 2 parameterizations to advance methods to optimized wind turbine arrays and maximize system-wide power production. Analysis of output from simulation on Aristotle were conducted on Jetstream.
  • Began new simulations for a domain centered over the Southern Great Plains on a single VM; this case would make an exceptional candidate for a trial of simulations across multiple VMs.
  • Used machine learning and Red Cloud's large RAM and multi-processors to detect and quantify wind gusts at Newark, Boston, and Chicago O'Hare Airports. Found Artificial Neural Networks exhibit a higher skill than logistic and linear regression models for wind gust occurrence and magnitude.
  • Simulated wind farm wakes from the east-coastal offshore lease areas using ultra-high resolutions with WRF on NERSC-Cori. Performed the data analyses on Aristotle's Red Cloud.


  • Complete simulations of derechos, fast-moving damaging deeply convective systems, associated with tornadoes, wind gusts, very heavy precipitation and hail.
  • Under a new grant, examine the wind resources and optimize wind farm layouts for the offshore wind turbines along the U.S. east coast. This work strongly leverages our previous simulations on Aristotle and will entail additional WRF simulations with wind farm parameterizations enabled. The goal is to define the optimal density of wind farms to optimize system-wide power production and minimize the levelized cost-of-energy.


Data streams and models need to be fully integrated to better understand the Impact of aerosol particles on climate and health
Policymakers and the wind energy industry need better climate and wind speed variability models. (click image for video)

Next   Learn more