Deep-Pathfinder: A boundary layer height detection algorithm based on image segmentation

Abstract. The mixing layer height (MLH) indicates the change between vertical mixing of air near the surface and less turbulent air above. MLH is important for the dispersion of air pollutants and greenhouse gases, and for assessing the performance of numerical weather prediction systems. Existing lidar-based MLH detection algorithms typically do not use the full resolution of the ceilometer, require manual feature engineering, and often do not enforce temporal consistency of the MLH. To address these limitations, a novel MLH detection approach has been developed based on deep learning techniques for image segmentation. The concept of our Deep-Pathfinder algorithm is to represent the 24-hour MLH profile as a mask and directly predict it from an image with lidar observations. Therefore, range-corrected signal data was obtained from Lufft CHM 15k ceilometers at five locations in the Netherlands that were part of the operational ceilometer network. Input samples of 224 × 224 pixels were extracted, each covering a 45-minute observation period. A customised U-Net architecture was developed with a nighttime indicator and MobileNetV2 encoder for fast inference times. The model was pre-trained on 19.4 million samples of unlabelled data and fine-tuned using 50 days of high-resolution annotations. Qualitative and quantitative results showed competitive performance compared to two benchmark models: the Lufft and STRATfinder algorithms. Existing path optimisation algorithms have good temporal consistency, but can only be evaluated after a full day of ceilometer data has been recorded. Deep-Pathfinder retains the advantages of temporal consistency but can also provide real-time estimates. This makes our approach valuable for operational settings, as real-time MLH detection better meets the requirements of users such as in aviation, weather forecasting and air quality monitoring.