arXiv:2502.01550v2 Announce Type: replace 
Abstract: With climate change intensifying fire weather conditions globally, accurate seasonal wildfire forecasting has become critical for disaster preparedness and ecosystem management. We introduce FireCastNet, a novel deep learning architecture that combines 3D convolutional encoding with GraphCast-based Graph Neural Networks (GNNs) to model complex spatio-temporal dependencies for global wildfire prediction. Our approach leverages the SeasFire dataset, a comprehensive multivariate Earth system datacube containing climate, vegetation, and human-related variables, to forecast burned area patterns up to six months in advance. FireCastNet treats the Earth as an interconnected graph, enabling it to capture both local fire dynamics and long-range teleconnections that influence wildfire behavior across different spatial and temporal scales. Through comprehensive benchmarking against state-of-the-art models including GRU, Conv-GRU, Conv-LSTM, U-TAE, and TeleViT, we demonstrate that FireCastNet achieves superior performance in global burned area forecasting, with particularly strong results in fire-prone regions such as Africa, South America, and Southeast Asia. Our analysis reveals that longer input time-series significantly improve prediction robustness, while spatial context integration enhances model performance across extended forecasting horizons. Additionally, we implement local area modeling techniques that provide enhanced spatial resolution and accuracy for region-specific predictions. These findings highlight the importance of modeling Earth system interactions for long-term wildfire prediction.

FireCastNet هو هيكل جديد للتعلم العميق مصمم للتنبؤ بموسمية حرائق الغابات، مما يعالج الحاجة المتزايدة للتنبؤات الدقيقة بسبب تغير المناخ. يستخدم الترميز التلافيفي ثلاثي الأبعاد والشبكات العصبية البيانية لنمذجة الاعتماديات الزمانية والمكانية المعقدة. من خلال الاستفادة من مجموعة بيانات SeasFire، التي تتضمن متغيرات مناخية ونباتية متنوعة، يمكن لـ FireCastNet التنبؤ بأنماط المناطق المحترقة حتى ستة أشهر مقدمًا، مع اعتبار الأرض كرسوم بيانية مترابطة لالتقاط ديناميات الحرائق عبر مقاييس زمنية ومكانية مختلفة.

FireCastNet es una nueva arquitectura de aprendizaje profundo diseñada para la predicción estacional de incendios forestales, abordando la creciente necesidad de pronósticos precisos debido al cambio climático. Utiliza codificación convolucional 3D y redes neuronales gráficas para modelar dependencias espaciotemporales complejas. Al aprovechar el conjunto de datos SeasFire, que incluye diversas variables climáticas y de vegetación, FireCastNet puede predecir patrones de áreas quemadas hasta seis meses por adelantado, tratando a la Tierra como un grafo interconectado para capturar la dinámica d…

FireCastNet est une nouvelle architecture d'apprentissage profond conçue pour la prévision saisonnière des incendies de forêt, répondant à la nécessité croissante d'une prévision précise en raison du changement climatique. Elle utilise un encodage convolutionnel 3D et des réseaux de neurones graphiques pour modéliser des dépendances spatio-temporelles complexes. En s'appuyant sur le jeu de données SeasFire, qui comprend diverses variables climatiques et de végétation, FireCastNet peut prédire les zones brûlées jusqu'à six mois à l'avance, considérant la Terre comme un graphe interconnecté pour…

FireCastNet is a new deep learning architecture designed for seasonal wildfire prediction, addressing the growing need for accurate forecasting due to climate change. It utilizes 3D convolutional encoding and Graph Neural Networks to model complex spatio-temporal dependencies. By leveraging the SeasFire dataset, which includes various climate and vegetation variables, FireCastNet can predict burned area patterns up to six months in advance, treating the Earth as an interconnected graph to capture fire dynamics across different scales.

FireCastNet: Earth-as-a-Graph for Seasonal Fire Prediction

arXiv:2512.02260v1 Announce Type: cross 
Abstract: Increasing climate change and habitat loss are driving unprecedented shifts in species distributions. Conservation professionals urgently need timely, high-resolution predictions of biodiversity risks, especially in ecologically diverse regions like Africa. We propose EcoCast, a spatio-temporal model designed for continual biodiversity and climate risk forecasting. Utilizing multisource satellite imagery, climate data, and citizen science occurrence records, EcoCast predicts near-term (monthly to seasonal) shifts in species distributions through sequence-based transformers that model spatio-temporal environmental dependencies. The architecture is designed with support for continual learning to enable future operational deployment with new data streams. Our pilot study in Africa shows promising improvements in forecasting distributions of selected bird species compared to a Random Forest baseline, highlighting EcoCast's potential to inform targeted conservation policies. By demonstrating an end-to-end pipeline from multi-modal data ingestion to operational forecasting, EcoCast bridges the gap between cutting-edge machine learning and biodiversity management, ultimately guiding data-driven strategies for climate resilience and ecosystem conservation throughout Africa.

EcoCast هو نموذج زمني مكاني جديد مقترح يهدف إلى التنبؤ المستمر بمخاطر التنوع البيولوجي والمناخ، خاصة في المناطق الغنية بيئيًا مثل إفريقيا. يستخدم هذا النموذج صور الأقمار الصناعية متعددة المصادر، وبيانات المناخ، وسجلات علم المواطن للتنبؤ بالتغيرات قصيرة المدى في توزيعات الأنواع من خلال تقنيات التعلم الآلي المتقدمة.

EcoCast es un nuevo modelo espaciotemporal propuesto para la previsión continua de riesgos de biodiversidad y climáticos, especialmente en regiones ecológicamente diversas como África. Este modelo utiliza imágenes satelitales de múltiples fuentes, datos climáticos y registros de ciencia ciudadana para predecir cambios a corto plazo en las distribuciones de especies mediante técnicas avanzadas de aprendizaje automático.

EcoCast est un nouveau modèle spatio-temporel proposé pour la prévision continue des risques de biodiversité et de climat, en particulier dans des régions écologiquement diversifiées comme l'Afrique. Ce modèle utilise des images satellites multisources, des données climatiques et des enregistrements de science citoyenne pour prédire les changements à court terme dans les distributions d'espèces grâce à des techniques avancées d'apprentissage automatique.

EcoCast is a newly proposed spatio-temporal model aimed at continual biodiversity and climate risk forecasting, particularly in ecologically diverse regions like Africa. This model leverages multisource satellite imagery, climate data, and citizen science records to predict near-term shifts in species distributions using advanced machine learning techniques.

FireCastNet: Earth-as-a-Graph for Seasonal Fire Prediction

Was this article worth reading? Share it

Attentive AI

Botcast AI

FETCH HIVE