The Essence of True Weather Forecasting: Understanding the Inner Workings of a Weather Model (credit: Colin Zarzycki)
DISCUSSION: When it comes to weather forecasting, the first-order explanation involves the following process: Determine the region of interest, select an appropriate dataset, run a numerical model with the dataset to generate outputs, and create a forecast based on the model output. That may be the most straightforward answer but in truth, there is a plethora of information that lies within each individual step that would make this article become excessively long if driven into with such detail. However, as models become more complex and with the advent of supercomputing powerhouses such as Cheyenne (National Center for Atmospheric Research), broaching this topic has many useful insights and applications for a greater public understanding.
The Weather Research and Forecasting (WRF) model is currently the flagship numerical weather prediction engine that is capable of ingesting large quantities of data from a currently-operational model and in turn produces a forecast based on many billions of calculations of the physical and atmospheric governing equations. These model datasets range from a fine-scale, regional model such as the North American Model (NAM, which can use a grid scale of as small as 3km) to a coarser-scale yet fully global-coverage model such as the Global Forecast System (GFS, usually 0.25 degrees). On the time scale, one could utilize a more transient, rapidly-updating model such as the High Resolution Rapid Refresh (HRRR; updates each hour) to the less frequent but more comprehensive data from the European Center for Medium-range Weather Forecasting model (ECMWF; updates every 12 hours). The options are abundant, and through the help of powerful tools that can retrieve and process data from satellites and ground-based data, WRF also has the ability to perform calculations using these extra pieces of information. All in all, the central goal of such a powerful forecasting model is to provide as clear a depiction of the state of the atmosphere in the present and future through the help of weather model data and observations.
In many cases, forecasters often derive their outlook from a series of different outcomes known as an ensemble forecast. In general, ensembles are a group of forecasts that use a slightly different set of conditions in order to get a better understanding of the. These changes can range from the way that WRF utilizes a different atmospheric physics scheme to the way that WRF. WRF is re-run multiple times with subtle changes to the initial conditions, and that can yield vastly different forecasts which is where the element of uncertainty in forecasting comes into play. Bear in mind that the atmosphere is a dynamic fluid and changes are consistently occurring, so it is essential to understand and judge the forecasts appropriately. However, power still lies within the forecaster’s skill to interpret the most reasonable forecast given the expected changes in the short-term, and that holds valid when using multiple model products. A great example of this was the most recent “snow squall” that impacted portions of the Mid-Atlantic states of Pennsylvania and New York. The consensus between the 12km NAM and the 14km Community Atmosphere (CAM) model both showed snow in their forecast, but due to inherent differences in the techniques between the two models, the CAM was able to resolve, or show the development of, a squall-like feature in the forecast several hours in advance. It once again highlights the importance of analyzing multiple products to develop a precise forecast, but the availability of such vast options means more potential for forecasters to make sound decisions for short-term weather forecasts.
So what’s the future of weather forecasting and forecast models? It’s still a fresh research topic for researchers and forecasters alike and is applicable to many facets of daily and sub-daily forecasting. Model configurations at different space and time scales have potentially differing outcomes and greater computing power and increasingly efficient techniques will give forecasters the knowledge they need to issue more accurate forecasts in the coming years. Will we see a new system supersede WRF in the near-future? Possibly, as change is essential to the improvement of weather forecasting. But the recent improvements are a welcome sign for the near-future.
To learn more about other topics in weather research, be sure to click here!
© 2019 Meteorologist Brian Matilla