The Micropulse Lidar Cloud Mask Machine Learning value-added product (MPLCMASKML VAP) uses a machine learning (ML) model to produce a cloud mask from the lidar attenuated backscatter and linear depolarization ratio.
This VAP uses as input the attenuated backscatter and linear depolarization ratio from the original Micropulse Lidar Cloud Mask (MPLCMASK) VAP.
MPLCMASKML produces a daily file at a 30-second temporal resolution that contains the ML-predicted cloud mask, the ML model’s confidence in its prediction of each time and height bin, the number of cloud layers, and the cloud layer boundaries.
MPLCMASKML reduces cloud misclassification by almost half compared with the MPLCMASK cloud mask when evaluated against hand-labeled cloud masks created by a lidar expert. MPLCMASKML can capture clouds below 500 meters and prevent cloud layer merging. In addition, the VAP captures finer details of the cloud layers.
Information on the ML model and how it was trained and evaluated can be found in the article “Lidar Cloud Detection With Fully Convolutional Networks” (Cromwell and Flynn 2019).
Reference: Cromwell E and D Flynn. 2019. “Lidar Cloud Detection With Fully Convolutional Networks.” In 2019 IEEE Winter Conference on Applications of Computer Vision (WACV), 619-627, doi:10.1109/WACV.2019.00071.