Abstract. Identifying, detecting, and localizing extreme weather events is a crucial first step in understanding how they may vary under different climate change scenarios. Pattern recognition tasks such as classification, object detection, and segmentation (i.e., pixel-level classification) have remained challenging problems in the weather and climate sciences. While there exist many empirical heuristics for detecting extreme events, the disparities between the output of these different methods even for a single event are large and often difficult to reconcile. Given the success of deep learning (DL) in tackling similar problems in computer vision, we advocate a DL-based approach. DL, however, works best in the context of supervised learning – when labeled datasets are readily available. Reliable labeled training data for extreme weather and climate events is scarce. We create “ClimateNet” – an open, community-sourced human-expert-labeled curated dataset that captures tropical cyclones (TCs) and atmospheric rivers (ARs) in high-resolution climate model output from a simulation of a recent historical period. We use the curated ClimateNet dataset to train a state-of-the-art DL model for pixel-level identification – i.e., segmentation – of TCs and ARs. We then apply the trained DL model to historical and climate change scenarios simulated by the Community Atmospheric Model (CAM5.1) and show that the DL model accurately segments the data into TCs, ARs, or “the background” at a pixel level. Further, we show how the segmentation results can be used to conduct spatially and temporally precise analytics by quantifying distributions of extreme precipitation conditioned on event types (TC or AR) at regional scales. The key contribution of this work is that it paves the way for DL-based automated, high-fidelity, and highly precise analytics of climate data using a curated expert-labeled dataset – ClimateNet. ClimateNet and the DL-based segmentation method provide several unique capabilities: (i) they can be used to calculate a variety of TC and AR statistics at a fine-grained level; (ii) they can be applied to different climate scenarios and different datasets without tuning as they do not rely on threshold conditions; and (iii) the proposed DL method is suitable for rapidly analyzing large amounts of climate model output. While our study has been conducted for two important extreme weather patterns (TCs and ARs) in simulation datasets, we believe that this methodology can be applied to a much broader class of patterns and applied to observational and reanalysis data products via transfer learning.

The number and intensity of individual hot days affecting Finland in the current and future climate is investigated together with the circulation patterns associated with the hot days. In addition, the number, length and intensity of heat waves lasting at least 3 days is also considered. ERA‐Interim reanalysis data and both direct model output and bias‐corrected data for historical and future climate [representative concentration pathway 4.5 (RCP4.5) scenario] simulations from 17 global climate models are analysed. Three intensities of heat waves and hot days are defined based on daily mean temperature thresholds of 20, 24 and 28 °C. The percentage of summertime days which exceed these temperature thresholds is shown to increase in the future. In ERA‐Interim, 24% of summertime days in southern Finland exceed the lowest temperature threshold while none exceed the highest temperature threshold. Under the RCP4.5 scenario these values increase to 47 and 1%, respectively. Larger relative changes occur in northern Finland. Heat waves are also longer in the RCP4.5 simulations than in the historical simulations. In southern Finland, the mean length of a heat wave where the 20 °C daily mean temperature is exceeded is 6.1 days in the historical simulations but increases to 9.4 days in the RCP4.5 simulations. The hot days in both northern and southern Finland are associated with a statistically significant positive pressure anomaly over Finland and to the east to Finland and a statistically significant negative pressure anomaly over Russia between 90 and 120°E. These pressure anomalies were evident for all intensities of hot days in the current climate and the future climate. The magnitude of the pressure anomalies increases as the daily mean temperature threshold increases. However, for hot days which exceed the same daily mean temperature threshold, the pressure anomalies are weaker in the RCP4.5 simulations than in the historical or ERA‐Interim data.

The range of synoptic patterns that North Pacific landfalling atmospheric rivers form under are objectively identified using genesis day 500 hPa geopotential height anomalies in a self-organizing map (SOM). The SOM arranges the synoptic patterns to differentiate between two groups of climate modes -the first group with ENSO (El Niño Southern Oscillation), PDO (Pacific Decadal Oscillation), PNA (Pacific North American) and NP (North Pacific index) and the sec-