The field of environmental forecasting is moving towards more accurate and reliable predictions using deep learning-based approaches. Researchers are focusing on developing benchmark datasets and frameworks to support long-term temporal modeling and large-scale spatial coverage. This includes the integration of multiple covariates such as fuel conditions, meteorology, topography, and human activity to improve wildfire risk prediction. Additionally, there is a growing interest in data-driven ocean forecasting, with a emphasis on standardized benchmarks and evaluation methods to facilitate model development and comparison. Other areas of research include the detection of autoregressive-generated images and the early detection of forest fires using deep learning neural networks. Noteworthy papers include: BCWildfire, which presents a 25-year daily-resolution wildfire dataset and evaluates various time-series forecasting models. OceanForecastBench, which proposes a comprehensive benchmarking framework for data-driven ocean forecasting. PRADA, which introduces a simple and interpretable approach for detecting autoregressive-generated images. Probabilistic Wildfire Spread Prediction, which uses an autoregressive conditional generative adversarial network for probabilistic wildfire spread prediction.