A Wavelet-Based Automated Aerosol Layer Identification Algorithm on Vaisala CL61 and CL51 Ceilometer Backscatter Measurements

Over the last decade, increased wildfire frequency and severity in Canada has created a greater need for the tracking of associated smoke and forecasting its movement. The risk to public health due to elevated PM2.5 concentrations necessitate continuous improvements to air quality forecasting. Ceilometers can perform vertical atmospheric profiling measurements at high temporal and spatial resolution, filling the gaps left by localized surface measurements and satellites’ long orbital paths by providing tropospheric aerosol distribution, dynamics, and cloud coverage. The new Vaisala CL61 ceilometer and the older-generation Vaisala CL51 ceilometer are operating, co-located in Edmonton, AB, in a campaign by Environment and Climate Change Canada's (ECCC) Meteorological Service of Canada (MSC). They have continuously operated since September 2022 through two major wildfire seasons. The introduction of this instrumentation follows similar pushes in Europe with the E-PROFILE network, EARLINET, NASA's MPLNET, and Canada's MPLCAN to monitor vertical aerosol distributions continuously. Automating the process of layer identification and aerosol typing increases the accessibility of the measurements for forecasting and facilitates automatic model verification.

This work presents an algorithm designed to identify aerosol layers using signal-to-noise (SNR) filtering and the commonly used wavelet analysis. The algorithm uses SNR to filter the data, and then applies a continuous wavelet transform to identify and classify layers using backscatter levels. The CL61 depolarization signal can additionally differentiate the cloud phase and increase the accuracy of aerosol classification. One primary obstacle is the differentiation between boundary layer aerosol and precipitation, which we address by observing the spatial and temporal continuity of the backscatter signal. Additionally, applying our algorithm across generations of ceilometers and beyond requires a normalization to an operating standard for cross-instrument application. Our results demonstrate that our algorithm effectively identifies aerosol layers. Future work will refine the algorithm to support a broader range of instruments, and add more predictor variables to improve its accuracy, allowing for real-time smoke detection, analysis, and air quality forecasting.