New neural-network-based method to infer total ozone column amounts and cloud effects from multi-channel, moderate...

Area of Science:

• Atmospheric Science
• Data Analysis
• Remote Sensing

Background:

• Ultraviolet (UV) irradiance data analysis is crucial for atmospheric studies.
• Traditional methods like Look-up Table (LUT) may have limitations in accuracy and data retrieval.
• Accurate Total Ozone Column (TOC) measurements are vital for climate and atmospheric research.

Purpose of the Study:

• To introduce and evaluate a new method using a radial basis function neural network (RBF-NN) for analyzing UV irradiance data.
• To compare the performance of the RBF-NN method against the traditional LUT method.
• To assess the accuracy and validity of TOC retrievals using the RBF-NN method.

Main Methods:

• Development and application of a radial basis function neural network (RBF-NN) model.
• Analysis of approximately three years of data from a NILU-UV multi-channel instrument.
• Comparison of RBF-NN results with a traditional Look-up Table (LUT) method and Ozone Monitoring Instrument (OMI) TOC data.

Main Results:

• The RBF-NN method demonstrated improved agreement with OMI TOC values, showing a 1% decrease in relative difference and a 0.03 increase in correlation.
• The RBF-NN method retrieved a higher number of valid daily average TOC values (200-500 DU) compared to the LUT method.
• Under cloudy conditions, OMI TOC values were found to be underestimated, consistent with prior research.

Conclusions:

• The RBF-NN method offers a more accurate and reliable approach for analyzing UV irradiance data and retrieving TOC.
• The RBF-NN method outperforms the traditional LUT method in terms of accuracy and the range of valid data retrieved.
• The study confirms the underestimation of TOC by OMI under cloudy conditions, highlighting the need for robust data analysis techniques.