For efficient wind farm management and optimized power generation under adverse weather conditions, understanding the causal meteorological drivers is essential. In this paper, we investigate the temporal causal influences of wind speed-related meteorological processes within a wind farm using the Heterogeneous Graphical Granger model (HMML). HMML is applied to synthetically generated wind power production data from Eastern Austria. To assess the plausibility of the identified causal processes, we compare the results with those obtained using the state-of-the-art LiNGAM method. Both methods are applied and evaluated across six different scenarios, each defined by distinct hydrological periods. The scenarios are defined by a set of time intervals characterized by either low/high extreme wind speeds or moderate wind speeds. We applied both methods across these scenarios and conducted causal reasoning to identify potential causes of extreme wind speeds within the wind farm. The sets of causal parameters obtained using HMML were found to be more realistic than those derived from LiNGAM. Combining the knowledge of causal variables affecting wind speed at the turbine hub, identified by HMML in each scenario, with weather forecasts can offer practical guidance for wind farm operators. Specifically, this knowledge can support more informed planning regarding when wind turbines should or should not be generating energy. For instance, the strong Granger-causal linkage identified between wind speed and temperature can inform curtailment strategies. In scenarios where rising temperatures are predictive of declining wind speeds, operators may preemptively adjust turbine output or schedule maintenance to optimize efficiency and reduce wear. Moreover, such predictive insights can feed into energy market models, where anticipated curtailment due to meteorological dependencies affects both generation forecasts and pricing strategies. By integrating these causal relationships into operational planning, the proposed tool offers a pathway toward more adaptive and economically efficient wind energy management.

This study presents a framework that combines Bayesian inference with reinforcement learning to guide drone-based sampling for methane source estimation. Synthetic gas concentration and wind observations are generated using a calibrated model derived from real-world drone measurements, providing a more representative testbed that captures atmospheric boundary layer variability. We compare three path planning strategies—preplanned, myopic (short-sighted), and non-myopic (long-term)—and find that non-myopic policies trained via deep reinforcement learning consistently yield more precise and accurate estimates of both source location and emission rate. We further investigate centralized multi-agent collaboration and observe comparable performance to independent agents in the tested single-source scenario. Our results suggest that effective source term estimation depends on correctly identifying the plume and obtaining low-noise concentration measurements within it. Precise localization further requires sampling in close proximity to the source, including slightly upwind. In more complex environments with multiple emission sources, multi-agent systems may offer advantages by enabling individual drones to specialize in tracking distinct plumes. These findings support the development of intelligent, data-driven sampling strategies for drone-based environmental monitoring, with potential applications in climate monitoring, emission inventories, and regulatory compliance.

As the volume of meteorological observations continues to grow, automating the quality control (QC) process is essential for timely data delivery. This study evaluates the performance of three machine learning algorithms—autoencoder, variational autoencoder, and long short-term memory (LSTM) autoencoder—for detecting anomalies in air temperature data. Using expert-quality-controlled data as ground truth, all models demonstrated anomaly detection capability, with the LSTM outperforming others due to its ability to capture temporal patterns and minimize false positives. When applied to raw data, the LSTM achieved 99.6% accuracy in identifying valid observations and replicated 79% of manual flags, with only five false negatives and six false positives over a full year. Its sensitivity to subtle meteorological changes, such as those caused by rainfall or cloud cover, highlights its robustness. The LSTM’s performance using a three-day timestep, combined with basic QC checks in SaQC (System for Automated Quality Control), suggests a scalable and effective solution for automated QC at Met Éireann, with potential for expansion to include additional variables and multi-station generalization.