IoT and Satellite Sensor Data Integration for Assessment of Environmental Variables: A Case Study on NO2

Do Machine Learning Methods Improve Prediction of Ambient Air Pollutants with High Spatial Contrast? A Systematic Review

BACKGROUND & OBJECTIVE

The use of machine learning for air pollution modelling is rapidly increasing. We conducted a systematic review of studies comparing statistical and machine learning models predicting the spatiotemporal variation of ambient nitrogen dioxide (NO2), ultrafine particles (UFPs) and black carbon (BC) to determine whether and in which scenarios machine learning generates more accurate predictions.

METHODS

Web of Science and Scopus were searched up to June 13, 2024. All records were screened by two independent reviewers. Differences in the coefficient of determination (R2) and Root Mean Square Error (RMSE) between best statistical and machine learning methods were compared across categories of methodological elements.

RESULTS

A total of 38 studies with 46 model comparisons (30 for NO2, 8 for UFPs and 8 for BC) were included. Linear non-regularized methods and Random Forest were most frequently used. Machine learning outperformed statistical models in 34 comparisons. Mean differences (95% confidence intervals) in R2 and RMSE between best machine learning and statistical models were 0.12 (0.08, 0.17) and 20% (11%, 29%) respectively. Tree-based methods performed best in 12 of 17 multi-model comparisons. Nonlinear or regularization regression methods were used in only 12 comparisons and provided similar performance to machine learning methods.

CONCLUSION

This systematic review suggests that machine learning methods, especially tree-based methods, may be superior to linear non-regularized methods for predicting ambient concentrations of NO2, UFPs and BC. Additional comparison studies using nonlinear, regularized and a wider array of machine learning methods are needed to confirm their relative performance. Future air pollution studies would also benefit from more explicit and standardized reporting of methodologies and results.

Five Years of Accurate PM2.5 Measurements Demonstrate the Value of Low-Cost PurpleAir Monitors in Areas Affected by Woodsmoke

Low-cost optical sensors are used in many countries to monitor fine particulate (PM2.5) air pollution, especially in cities and towns with large spatial and temporal variation due to woodsmoke pollution. Previous peer-reviewed research derived calibration equations for PurpleAir (PA) sensors by co-locating PA units at a government regulatory air pollution monitoring site in Armidale, NSW, Australia, a town where woodsmoke is the main source of PM2.5 pollution. The calibrations enabled the PA sensors to provide accurate estimates of PM2.5 that were almost identical to those from the NSW Government reference equipment and allowed the high levels of wintertime PM2.5 pollution and the substantial spatial and temporal variation from wood heaters to be quantified, as well as the estimated costs of premature mortality exceeding $10,000 per wood heater per year. This follow-up study evaluates eight PA sensors co-located at the same government site to check their accuracy over the following four years, using either the original calibrations, the default woodsmoke equation on the PA website for uncalibrated sensors, or the ALT-34 conversion equation (see text). Minimal calibration drift was observed, with year-round correlations, r = 0.98 ± 0.01, and root mean square error (RMSE) = 2.0 μg/m3 for daily average PA PM2.5 vs. reference equipment. The utitilty of the PA sensors without prior calibration at locations affected by woodsmoke was also demonstrated by the year-round correlations of 0.94 and low RMSE between PA (woodsmoke and ALT-34 conversions) and reference PM2.5 at the NSW Government monitoring sites in Orange and Gunnedah. To ensure the reliability of the PA data, basic quality checks are recommended, including the agreement of the two laser sensors in each PA unit and removing any transient spikes affecting only one sensor. In Armidale, from 2019 to 2022, the continuing high spatial variation in the PM2.5 levels observed during the colder months was many times higher than any discrepancies between the PA and reference measurements. Particularly unhealthy PM2.5 levels were noted in southern and eastern central Armidale. The measurements inside two older weatherboard houses in Armidale showed that high outdoor pollution resulted in high pollution inside the houses within 1-2 h. Daily average PM2.5 concentrations available on the PA website allow air pollution at different sites across regions (and countries) to be compared. Such comparisons revealed major elevations in PA PM2.5 at Gunnedah, Orange, Monash (Australian Capital Territory), and Christchurch (New Zealand) during the wood heating season. The data for Gunnedah and Muswellbrook suggest a slight underestimation of PM2.5 at other times of the year when there are proportionately more dust and other larger particles. A network of appropriately calibrated PA sensors can provide valuable information on the spatial and temporal variation in the air pollution that can be used to identify pollution hotspots, improve estimates of population exposure and health costs, and inform public policy.

Reflection Symmetry Detection in Earth Observation Data

The paper presents a new algorithm for reflection symmetry detection, which is specialized to detect maximal symmetric patterns in an Earth observation (EO) dataset. First, we stress the particularities that make symmetry detection in EO data different from detection in other geometric sets. The EO data acquisition cannot provide exact pairs of symmetric elements and, therefore, the approximate symmetry must be addressed, which is accomplished by voxelization. Besides this, the EO data symmetric patterns in the top view usually contain the most useful information for further processing and, thus, it suffices to detect symmetries with vertical symmetry planes. The algorithm first extracts the so-called interesting voxels and then finds symmetric pairs of line segments, separately for each horizontal voxel slice. The results with the same symmetry plane are then merged, first in individual slices and then through all the slices. The detected maximal symmetric patterns represent the so-called partial symmetries, which can be further processed to identify global and local symmetries. LiDAR datasets of six urban and natural attractions in Slovenia of different scales and in different voxel resolutions were analyzed in this paper, demonstrating high detection speed and quality of solutions.

Resource Allocation for Cognitive LEO Satellite Systems: Facilitating IoT Communications

Due to the characteristics of global coverage, on-demand access, and large capacity, the low earth orbit (LEO) satellite communication (SatCom) has become one promising technology to support the Internet-of-Things (IoT). However, due to the scarcity of satellite spectrum and the high cost of designing satellites, it is difficult to launch a dedicated satellite for IoT communications. To facilitate IoT communications over LEO SatCom, in this paper, we propose the cognitive LEO satellite system, where the IoT users act as the secondary user to access the legacy LEO satellites and cognitively use the spectrum of the legacy LEO users. Due to the flexibility of code division multiple access (CDMA) in multiple access and the wide use of CDMA in LEO SatCom, we apply CDMA to support cognitive satellite IoT communications. For the cognitive LEO satellite system, we are interested in the achievable rate analysis and resource allocation. Specifically, considering the randomness of spreading codes, we use the random matrix theory to analyze the asymptotic signal-to-interference-plus-noise ratios (SINRs) and accordingly obtain the achievable rates for both legacy and IoT systems. The power of the legacy and IoT transmissions at the receiver are jointly allocated to maximize the sum rate of the IoT transmission subject to the legacy satellite system performance requirement and the maximum received power constraints. We prove that the sum rate of the IoT users is quasi-concave over the satellite terminal receive power, based on which the optimal receive powers for these two systems are derived. Finally, the resource allocation scheme proposed in this paper has been verified by extensive simulations.