Science Leads: Sara C. Pryor (Cornell), Sue Ellen Haupt and Jared Lee (NCAR)
This use case will focus on two critical sources of uncertainty facing decarbonization of our electricity supply: the intermittency of renewable energy supplies to the grid across a range of scales and the degree to which these “weather-dependent” renewable energy sources will themselves be changed by an evolving climate . WRF is an ideal tool for performing such research because (1) there is a wealth of information regarding optimal configurations for use in renewable energy resource estimation, and (2) WRF, including customized applications of WRF like WRF-Solar [2, 3] and WRF-Solar-Wind, has been demonstrated to exhibit high fidelity and hence credibility within the renewable energy community [4, 5, 6, 7, 8, 9].
A specific, but by no means exclusive, focus of our activities will be to quantify variability and change in the frequency of occurrence and/or intensity of “production droughts”—large-scale (regional) extended periods (30–90 days) of low productivity—and “production excesses”—large-scale (regional) extended periods (30–90 days) of anomalously high productivity. Production droughts have implications for the stability of the electricity supply and in the contemporary climate are caused by the action of large-scale climate modes (e.g., the Pacific Decadal Oscillation) [10, 11]. We will apply WRF-Solar-Wind nested within ICs/LBCs from Earth System Model (ESM) output to examine if these periods of anomalously low and high-power production change over the current century. LBCs will be drawn from a Coupled Model Intercomparison Project 6th phase (CMIP6)-generation ESM that best represents characteristics of internal climate modes that influence North America in the contemporary climate .
The computational demands of such simulations are substantial but will be tractable within the multi-node containers to be constructed within the project. We envisage a domain covering all of CONUS using a horizontal grid spacing of 4 km, with approximately 41 vertical levels (for a total grid cell count of ~80 million). Assuming a 12-s time step, and a multi-node instance that has 256 cores, we envisage being able to achieve 1:40 compute time:real-time ratios. Thus, our target is to achieve a simulation of a 40-year duration spanning approximately 2015–2054. The output from the first 9 years of these simulations will first be subjected to evaluation relative to observations from the contemporary climate, thus demonstrating another configuration of METplus included within the containerized system. The simulation output for the entire period will then be subject to the advanced visualization tools embedded within the analysis/visualization container to demonstrate that component of the containerization.
In a test case using two winter wind storms that caused substantial economic losses and human mortality and morbidity, we used a simulation domain of 500×400 horizontal grid cells, grid resolution of 2 km, and 57 vertical levels, for a total of approximately 11 million grid cells. Reasonable run times of approximately 9 hours for a 5-day event were achieved using 64 cores. Thus, the use of substantially extended domains to cover a region that includes more upstream influences of the Great Lakes will be entirely achievable. The output from simulations of each storyline will first be subjected to evaluation relative to observations from the contemporary climate, thus demonstrating the model output fidelity toolbox included within the containers.