Numerical Weather Prediction (NWP) data is the form of weather modeling data that most professionals and consumers are most familiar with. NWP takes current observations of weather measurements and utilizes this data to create weather forecasts. Many systems assisting researchers in ingesting, analyzing, and storing this data do so with the help of weather modeling applications like Weather Research and Forecasting (WRF).
Weather Modeling and Forecasting Process
Weather forecasters utilize mathematical equations that factor in the physics behind the variables that influence weather – solar radiation, orbital distance from the sun, pressure, wind, temperature, and moisture – among others. These observations are obtained from sensors or satellites and then fed into the equations in a process that’s referred to as data assimilation. This data is then fed into a few different slots that assist in the process of specifying weather for future points.
There are three primary used synoptics forecast models: The North American Mesoscale Model (NAM), the Global Forecast System (GFS), and the Nested Grid Model (NGM).
North American Mesoscale Model (NAM)
The NAM model refers to a numerical weather prediction model run by National Centers for Environmental Prediction for short-term weather forecasting. Currently, the Weather Research and Forecasting Non-hydrostatic Mesoscale Model (WRF-NMM) model is run as the NAM, thus, three names (NAM, WRF, or NMM) typically refer to the same model output.
Weather Research and Forecasting (WRF)
The WRF model is a mesoscale numerical weather prediction system for both operational forecasting and atmospheric research objectives. WRF was developed through the partnerships of the National Center for Atmospheric Research (NCAR), the National Oceanic and Atmospheric Administration, the Forecast Systems Laboratory (FSL), the Air Force Weather Agency (AFWA), the Naval Research Laboratory, Oklahoma University and the Federal Aviation Administration (FAA).
WRF offers two dynamical solvers for its computation of the atmospheric governing equations: WRF-ARW (Advanced Research WRF) and WRF-NMM (nonhydrostatic mesoscale model). ARW is supported to the community by the NCAR Mesoscale and Microscale Meteorology Laboratory. NMM is supported to the community by the Developmental Testbed Center (DTC).
Global Forecast System (GFS)
The GFS is a weather forecast model that is produced by the National Centers for Environmental Prediction (NCEP). This model produces a dataset that allows for dozens of atmospheric and land-soil variables to be accessed and considered in the forecasting of weather, like temperature, wind, precipitation, soil moisture, and atmospheric ozone concentration.
The entire globe is covered by the GFS with a base horizontal resolution of 18 miles between grid points, which is used by forecasters to predict weather for out to 16 days in the future.
Hardware Recommendations
- RAM: Determines maximum model size (DOF, degrees of freedom) that can be solved. Typically large amount of memory per processor core (4 GB+ per processor core) are the standard.
- CPU: number of cores and clock speed determines how quickly a model can be solved. (Good metric to compare between CPU options is: Clock Speed x Number of Cores / Cost. Higher clock speeds and large core counts enable larger weather models to be run at higher resolutions.
- Storage: determines how much data can be held on the system, and how quickly it can be input/read.
- GPU: Speed up complex solutions. NVIDA GPUs are often utilized.
- Interconnects: Enables high speed cluster communication and lower latencies. 100 Gb / sec network fabrics from Intel (Omnipath) and Mellanox (Infiniband) are typically a standard in weather modeling HPC environments.
High resolution models like those mentioned above required massive amounts of computing power, along with expertise and experience. That capability comes from our HPC cluster, the PowerWulf ZXR1+. At PSSC Labs, we provide our clients with the partner they need in systems design, manufacturing, and installation of custom-built HPC hardware, built to ensure that your weather model performs exactly as you’ve designed it to. We focus on providing an ultra-reliable, extreme-scale platforms for your needs.
Government agencies, public utilities, and research organizations rely on our expertise to realize their goals for mitigating and managing risk associated with severe weather and the effects of climate change.
