Knowledge-Guided Machine Learning: Illustrating the use of Explainable Boosting Machines to Identify Overshooting Tops in Satellite Imagery
Home / Papers / Knowledge-Guided Machine Learning: Illustrating the use of Explainable Boosting Machines to Identify Overshooting Tops in Satellite Imagery
Knowledge-Guided Machine Learning: Illustrating the use of Explainable Boosting Machines to Identify Overshooting Tops in Satellite Imagery
Authors: Nathan Mitchell, Lander Ver Hoef, Imme Ebert‐Uphoff, Kristina Moen, Kyle Hilburn, Yoonjin Lee, Emily J. King Date: 2026-06-19 Paper ID: openalex:2507.03183
Summary
This paper introduces the use of Explainable Boosting Machines (EBMs) in meteorology to address the lack of transparency in complex forecasting models. The authors propose a knowledge-guided pipeline that transforms satellite imagery into scalar features, allowing for the construction of a fully interpretable model for identifying overshooting tops. Through human-machine collaboration, the model is refined to align with domain expert heuristics, providing a balance between predictive performance and explainable decision-making.
Key Contributions
- Demonstrates the application of Explainable Boosting Machines (EBMs) for meteorological feature detection to improve interpretability in high-stakes weather forecasting.
- Proposes a knowledge-guided feature extraction pipeline to map complex meteorological imagery to scalar inputs suitable for EBMs using techniques like Gray-Level Co-occurrence Matrices.
- Provides a framework for human-in-the-loop refinement of ML models by analyzing and adjusting EBM decision paths to align with domain expert strategies.
Links
Metadata & Links
- url
- https://arxiv.org/abs/2507.03183
- paper_id
- 2507.03183
- paper_source
- openalex
- domain
- computer-vision
- tags
- explainable-aimeteorologyhuman-in-the-loop
- architectures
-
- datasets
-
- concept_slugs
-
- dataset_slugs
-
- skill
- TimeSeriesSkill
- created_at
- 2026-06-22T06:49:19Z