Estimation of Lidar backscatter gap, and aerosol size distribution using Artificial Neural Network Algorithms
González Chévere, David
MetadataShow full item record
This work presents the design and implementation of two Artificial Neural Network (ANN) algorithms to (1) estimate data-gap in Lidar 1064 nm, and (2) determine and plot Aerosol Size Distribution (ASD) based on four available wavelengths at the University of Puerto Rico, Mayaguez (UPRM). In (1), the network was trained using Ceilometer data at 910nm as input and available Lidar data at 1064 nm from the same time and range as target. Results of the error analysis show good matches with better than 0.82 of correlation and 0.844 of Root mean Square Error (RMSE) values. In (2), the backscatter column profile data at the three Lidar wavelengths (355, 532, 1064 nms) and single wavelength Ceilometer (910 nm) were used to derive their respective Aerosol Optical Depth (AOD) values. To determine ASD using Lidars, an ANN was trained using the Aerosol Robotic Network (AERONET) data. Subsequently, the AODs from Lidars (obtained at 355, 532, 910, and 1064nms) were used as inputs to the trained ANN to generate ASD at the output. Comparison between the estimated ASD based on Lidar’s four wavelengths and AERONET ASD of eight wavelengths showed good results with RMSE better than 0.016.