Development and evaluation of correction models for a low-cost fine particulate matter monitor
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Abstract. Four correction models with differing forms were developed on a training dataset of 32 PurpleAir–Federal Equivalent Method (FEM) hourly fine particulate matter (PM2.5) observation colocation sites across North America (NA). These were evaluated in comparison with four existing models from external sources using the data from 15 additional NA colocation sites. Colocation sites were determined automatically based on proximity and a novel quality control process. The Canadian Air Quality Health Index Plus (AQHI+) system was used to make comparisons across the range of concentrations common to NA, as well as to provide operational and health-related context to the evaluations. The model found to perform the best was our Model 2, PM2.5-corrected=PM2.5-cf-1/(1+0.24/(100/RH%-1)), where RH is limited to the range [30 %,70 %], which is based on the RH growth model developed by Crilley et al. (2018). Corrected concentrations from this model in the moderate to high range, the range most impactful to human health, outperformed all other models in most comparisons. Model 7 (Barkjohn et al., 2021) was a close runner-up and excelled in the low-concentration range (most common to NA). The correction models do not perform the same at different locations, and thus we recommend testing several models at nearby colocation sites and utilizing that which performs best if possible. If no nearby colocation site is available, we recommend using our Model 2. This study provides a robust framework for the evaluation of low-cost PM2.5 sensor correction models and presents an optimized correction model for North American PurpleAir (PA) sensors.
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.
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 | 0.003 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| 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) | 0.000 | 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 it