Lightning Probability - Warning Decision Training Division (WDTD)

Navigation Links

Products Guide

Lightning Probability

Short Description

Probability of cloud-to-ground (CG) lightning at a given location over the next 30 minutes and next 60 minutes.

Subproducts

Lightning Probability Next 30 minutes

Lightning Probability Next 60 minutes

Primary Users

NWS: WFO, CWSU, AWC, SPC

FAA: Tower, TRACON, ARTCC, ATCSCC, AFSS

Other: EM

Input Sources

Earth Network’s In-Cloud Lightning

Vaisala’s Cloud-to-Ground Lightning

MRMS products including Maximum Estimated Size of Hail (MESH) and Vertically Integrated Liquid (VIL)

Near-Storm Environmental Data

Resolution

Spatial Resolution: 0.01^o Latitude (~1.11 km) x 0.01^o Longitude (~1.01 km at 25^oN and 0.73 km at 49^oN)

Temporal Resolution: 2 minutes

Product Creation

The cloud-to-ground (CG) lightning probability product was created using "random forest" machine learning methodology. With storm-based inputs of Earth Network’s in-cloud lightning, Vaisala’s cloud-to-ground lightning, MRMS products including the Maximum Estimated Size of Hail (MESH) and Vertically Integrated Liquid (VIL), and near storm environmental data including lapse rate and CAPE, a random forest algorithm was trained to produce probabilities of CG lightning.

Technical Details

Latest Update: MRMS Version 12

References

Strengths

This product has proven more successful than the steady-state approach of simply advecting the lightning density field.

The major strength of using “Random forests” is that they can easily take large datasets and create decision trees. The decision tree is how the algorithm arrives at the “answer.”

Another strength of using random forest methodology in the creation of the lightning probability products is that it can show which inputs are most important to the algorithm by testing for statistical significance. The data inputs that were not statistically significant were removed from the algorithm, which makes the algorithm computationally more efficient by reducing the amount of data processing that needs to occur.

Limitations

Does not incorporate satellite-based data (GLM).

Quality Control

Applications

The lightning probability product is especially helpful in both tactical and short-term air traffic routing decisions, as well as with Impact-Based Decision Support Services, since it provides a lightning forecast, rather than just an observation

Example Images

Fig. 1: Cloud-to-Ground Lightning Probability over the next 30 minutes on 26 September 2018 at 1458Z.