Towards Greener Nights: Exploring AI-Driven Solutions for Light Pollution Management
Paras Varshney, Niral Desai, Uzair Ahmed
TL;DR
This work tackles light pollution by building predictive pipelines that estimate night-sky brightness (limiting magnitude) across space and time using a fusion of Globe at Night and GaN-MN datasets. The approach integrates spatio-temporal context via KNN-derived features, text representations with DeBERTa v3 embeddings, and TF-IDF with PCA, evaluated through multi-fold cross-validation and ensemble learning (XGBoost, LightGBM, RandomForest). The study reports a Mean F1-Score of up to $0.81883$ on the leaderboard, with LightGBM selected as the primary engine for its balance of speed and accuracy, supported by a rigorous comparison against CatBoost and DeBERTa-based variants. The resulting framework offers actionable insights for targeting lighting interventions, reducing sky glow, and informing policy and urban planning to mitigate ecological, energy, and human health impacts, while providing a data-rich basis for ongoing monitoring of nocturnal luminosity. $F1 = 2 \cdot \frac{\text{Precision} \cdot \text{Recall}}{\text{Precision} + \text{Recall}}$ is employed as the core evaluation metric to ensure balanced performance.
Abstract
This research endeavors to address the pervasive issue of light pollution through an interdisciplinary approach, leveraging data science and machine learning techniques. By analyzing extensive datasets and research findings, we aim to develop predictive models capable of estimating the degree of sky glow observed in various locations and times. Our research seeks to inform evidence-based interventions and promote responsible outdoor lighting practices to mitigate the adverse impacts of light pollution on ecosystems, energy consumption, and human well-being.