Classification of lidar measurements using supervised and unsupervised machine learning methods
Bibliographic record
Abstract
Abstract. While it is relatively straightforward to automate the processing of lidar signals, it is more difficult to choose periods of “good” measurements to process. Groups use various ad hoc procedures involving either very simple (e.g. signal-to-noise ratio) or more complex procedures (e.g. Wing et al., 2018) to perform a task that is easy to train humans to perform but is time-consuming. Here, we use machine learning techniques to train the machine to sort the measurements before processing. The presented method is generic and can be applied to most lidars. We test the techniques using measurements from the Purple Crow Lidar (PCL) system located in London, Canada. The PCL has over 200 000 raw profiles in Rayleigh and Raman channels available for classification. We classify raw (level-0) lidar measurements as “clear” sky profiles with strong lidar returns, “bad” profiles, and profiles which are significantly influenced by clouds or aerosol loads. We examined different supervised machine learning algorithms including the random forest, the support vector machine, and the gradient boosting trees, all of which can successfully classify profiles. The algorithms were trained using about 1500 profiles for each PCL channel, selected randomly from different nights of measurements in different years. The success rate of identification for all the channels is above 95 %. We also used the t-distributed stochastic embedding (t-SNE) method, which is an unsupervised algorithm, to cluster our lidar profiles. Because the t-SNE is a data-driven method in which no labelling of the training set is needed, it is an attractive algorithm to find anomalies in lidar profiles. The method has been tested on several nights of measurements from the PCL measurements. The t-SNE can successfully cluster the PCL data profiles into meaningful categories. To demonstrate the use of the technique, we have used the algorithm to identify stratospheric aerosol layers due to wildfires.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".