The field of environmental time series forecasting is rapidly advancing with the development of innovative models and techniques. Recent studies have demonstrated the effectiveness of large time series models, such as foundation models, in predicting water levels and other environmental variables. Additionally, multi-scale graph learning methods have shown promise in downscaling data to fine spatial resolutions, enabling more accurate predictions of stream water temperatures and other variables. The integration of remote sensing data and artificial intelligence models has also led to significant improvements in the detection and classification of environmental phenomena, such as algal blooms. Furthermore, advances in probabilistic forecasting methods, including denoising diffusion models, have enhanced the ability to predict uncertain environmental events, such as coastal inundation. Noteworthy papers in this area include the introduction of the Non-stationary Diffusion method, which relaxes the fixed uncertainty assumption of traditional diffusion models, and the proposal of the STRGCN model, which captures asynchronous spatio-temporal dependencies in irregular multivariate time series forecasting. The AI-driven multi-source data fusion approach for algal bloom severity classification also presents a promising direction for future research.