Laser-Induced Incandescent Signals for Laboratory Samples of Absorbing Aerosols Detected by the Single Particle Soot Photometer

This repository contains a labeled, machine-learning ready data set of Laser-Induced Incandescent (L-II) time series acquired by the NOAA Single Particle Soot Photometer (SP2). The data set includes examples of 7 different types of aerosols that demonstrate measurable incandescence in the SP2: Fullerene soot (FS), Fullerene soot+glycerol (FS+glyc), Arizona Test Dust (ATD), volcanic ash (VA), coal fly ash (CFA), Fe2O3, Fe3O4). The data set is described in detail in the paper "Classification of iron oxide aerosols with a single particle soot photometer using supervised machine learning" by K.D. Lamb, Atmospheric Measurement Techniques, 12, 3885-3906 (2019). The data is split into training (50%), validation (25%), and test (25%) data sets and stored in numpy arrays. Lab_X_{}.npy contains the time series, with the array dimensions corresponding to 0 - aerosol sample number, 1 - the SP2 detection channel, and 2 - the 400 point time series associated with that channel. The SP2 detection channels correspond to 0 - scattering channel, 1 - blue incandescent channel, 2 - red incandescent channel, and 3 - position-sensitive scattering channel. Lab_Y_{}.npy contains the class labels. The total number of aerosols of each type and its corresponding class label is given in the table below: Label Class Total 0 FS 20004 1 FS+glyc 20018 2 CFA 20009 3 ATD 20001 4 VA 20005 5 Fe2O3 20008 6 Fe3O4 20037