Spaghetti… what? Forecast models broken down

By now I’m sure you have seen the “spaghetti plot”- ensemble of forecast models-regarding Tropical Depression Nine. If you’re not a meteorologist or everyday weather guru, then you have every right to be confused by this graphic of overwhelming lines. With a little background on the models, you’ll be sure to impress your coworkers when they begin to bring up TD Nine or soon to be Tropical Storm Nine.

Numerical Weather Prediction (NWP) is the process of numerical weather data (initial weather conditions) that is ingested into complex supercomputers, then undergo numerous mathematical equations to determine a future state of atmospheric parameters like temperature, humidity, pressure, rainfall, and wind. These models differ on initial conditions, physics, and data assimilation resulting in varying forecast evolutions. A meteorologist’s toughest decision on any day can be interpreting the model data and what model will perform the best.

The multi-layer global dynamical models used for day-to-day forecasting can be some of the most accurate hurricane forecast models.  These models solve mathematical equations that govern the atmosphere at every point on the globe. Below are a few of the global models:

ECMWF (European Center for Medium-range Weather Forecasting) Model- produced twice daily at 00Z and 12Z. This 4-D data assimilation, high spatial resolution model has also shown skill in forecasting tropical systems. Most argue this model is the premier global dynamical model. Access to this model’s data usually requires an expensive subscription.

GFS (Global Forecast System) Model- recently upgraded in May of 2016 to a 4-D ensemble hybrid data assimilation. With the new upgrade, the GFS Model now accounts for time as the fourth dimension. This model is produced four times a day, however the recent upgrade allows for hourly forecast guidance of up to 5 days.

UKMET (United Kingdom Met Office) Model- 4-D data assimilation model produced twice daily at 00Z and 12Z

All of these models are continuously upgraded in regards to resolution and initialization.  

A typical model used in hurricane forecasting is the BAM model:

BAM (Beta and Advection) Model – Follows a single-layer trajectory of vertically averaged upper level winds of GFS. Therefore, the BAM model is highly dependent on the GFS model. There are three different BAM models using different layers of the atmosphere to calculate the trajectory. The Shallow model (BAMS) calculates 850-700 hPa, the Medium (BAMM) model calculates 850-400 hPa, and the Deep (BAMD) model calculates the 850-200 hPa. These different levels help account for variation in the existing steering flow with height. The model’s trajectory is independent of the surrounding atmosphere and focuses merely on the existing flow pattern.

In addition to the well-known global models, there are three models specific to hurricane forecasting:

GFDL (Geophysical Fluid Dynamics Laboratory) Model – coupled with Princeton Ocean Model to help capture the interaction of the tropical system with the ocean. This model provides a specific intensity forecast of the hurricane.

GFDN– the Navy’s version of GFDL.

HWRF (Hurricane Weather Research and Forecast) Regional Model – also coupled with the Princeton Ocean Model in the Atlantic and NE Pacific. Due to the detailed nature of this model similar to that of the GFDL, the HWRF has the skill to forecast the specific intensity and track of hurricanes.

WWT Spaghetti Plot for Tropical Depression Nine
“Spaghetti plot” of forecast tracks for Tropical Depression Nine