Performance analysis and synergistic use of cloud flagging algorithms of ground-based remote sensing instrumentations

Sun photometers retrieve Aerosol Optical Depth (AOD) from direct solar irradiance observations when sun disk is not obscured by clouds. Therefore, identifying cloud-contaminated measurements is crucial for achieving high-quality AOD retrieval. Cloud flagging algorithms are instrumental in this process, as failure to accurately flag clouds can have significant impact on AOD retrieval and related applications e.g., air quality monitoring, numerical weather prediction, atmospheric transport models and data synergism.

This work aims at synergistically leverage ground-based instruments to analyze the performance of existing stand-alone algorithms in different scenarios, including periods of high-variability due to clouds, and extreme weather events like wildfires, dust storms, etc. To this direction, we used co-located Pandonia Global Network Pandora spectroradiometer and Global Atmosphere Watch Precision filter radiometer (PFR) network instrument and analyzed the performance of existing algorithms over the course of 2023 at Izana station.

PFR and Pandora measurements were synchronized with a time window of  1 min. For this analysis, PFR flags comprise two scenarios for clear-sun and flagged data, while Pandora has nine double-digit quality flags (QF) representing a combination of uncertainty (high: 0, medium: 1, low: 2) and data quality assurance (assured: 0, not assured:1, unusable data: 2). The analysis revealed good agreement between Pandora QF10 and PFR clear-sun flag (96.45%) followed by QF11 and QF12 (2.29% and 1.26%, respectively). Conversely, for PFR flagged data, corresponding percentages for Pandora QF10, QF11 and QF12 were 57.83%, 12.58% and 29.6%, respectively. Daily variation of Pandora QFs for 1-year indicated 34 days with Saharan dust episodes (concluded from HYSPLIT 24-h backward trajectories at levels from 500 to 6000 m above ground level originating from the Saharan region) on which for PFR clear sun flag, QF10 flag was more than 90% of daily number of comparison points indicating that dust events did not deteriorate much the flagging agreement between the two instruments. On the contrary, there were 20 dust free days with QF10 below 90% (most days had number of daily comparison points below 40, indicating high variability due to clouds). PFR clear-sun flag based and Pandora QF10 based AOD distribution showed respective geometric mean ranging from 0.02 to 0.04 and respective standard deviation ranging from 3.24 to 2.35 at 862 nm to 367 nm, respectively. This attempt of synergistic use of different measuring instruments can be useful for enhancing cloud flagging algorithms, thereby contributing to higher quality retrieval product.

 

Acknowledgement

This research has been supported by European Space Agency (ESA) in the frame of Instrument Data quality Evaluation and Assessment Service – Quality Assurance for Earth Observation (IDEAS-QA4EO) project, ACTRIS Switzerland project funded by Swiss State Secretariat for Education Research and Innovation and COST Action HARMONIA CA21119, supported by COST (European Cooperation in Science and Technology).