Weather Model Upgrades Offer Better Forecasts

by Jordanna Sheermohamed of Weather Forecast Solutions

Weather forecast models have become much more sophisticated as computing processing power has increased. It stands to reason that a better forecast means more accurate and better lead time to prepare for environmental disasters.  

But what exactly happens under the hood with these weather models and how do these upgrades help with producing a better weather forecast?  

A weather model is a computer program designed to take actual weather observational data as input and predict a potential outcome based on the laws of physics – the movement of mass, energy and forces through space and time. Somewhere in its output, a forecast is made with some degree of accuracy in place, but this will rely on several important factors.

First is the quality of the input data; ideally, frequent data with a high degree of accuracy would be best. Frequent can refer to the geographical distribution of measurements or in the frequency at which the measurement occurs – continuous, every 15 seconds, every minute, every hour or even every day.

Every day, twice a day, 900 weather balloons are simultaneously released across the world, allowing a vertical snapshot of the atmosphere’s major weather parameters at several levels: temperature, pressure, dew point, wind speed and direction. This real time data will be used as input for a weather model.

The second major factor is the computing power of the machine that will be processing the data. One processor is great, but a million is clearly better. A faster machine gets a forecast out earlier or allows for multiple potential outcomes to simultaneously run.

The third major factor is the way a specific weather model weights the microphysical processes of the atmosphere. The program algorithms will incorporate atmospheric physics, the laws of fluid dynamics, mathematical laws and a statistical analysis of the data crunched through the program.

The unique skill of a weather model rests on the ability to juggle all of the above, in addition to integrating how Earth’s spheres may influence weather conditions, especially ocean dynamics and land/water formations. In recent years, there has been concern over the lack of capturing ocean/atmospheric interaction in the models; having that closely interpreted in the model physics would likely create a better forecast.

When it comes to global models, the major players include the American Global Forecast System (GFS), the European Center for Medium Range Forecasts (ECMWF), the U.K. MetOffice’s UKMET, and the Canadian Meteorological Center (CMC).Regional models that focus on smaller regions under a National Oceanic and Atmospheric Administration jurisdiction also tend to do well up through the short range, such as the NOAA High Resolution Rapid Refresh (HRRR) and the National Blended Model (NBM).

The U.K. recently indicated an intention to spend about $1.6 billion on a new weather and climate model supercomputer to replace its current model, which will yield a sixfold performance increase. This massive upgrade in computing power will allow for better resolution in both the horizontal surface of the earth, and the vertical levels of the atmosphere. In comparison, the American GFS was given an upgrade in 2019, which was to increase performance by 50% and adding 60% more storage capacity for observations.

Because weather balloons can’t be launched over the ocean with the same frequency as they are over land, the open ocean is practically a data desert when it comes to weather models. This makes it one of the hardest types of forecast outputs to produce. Surface wind measurements over the ocean are often derived from satellite data, which brings its own difficulties in frequent global coverage. This is also why many of the maritime forecasts lean heavily on the forecaster’s knowledge of geographical influences and nuances of atmospheric pressure patterns.