机器学习在微塑料识别与环境风险评估中的应用研究进展

微塑料是一种新型污染物,可以在环境中长期存在并造成生态风险.目前,微塑料污染已成为全球性的重大环境问题.借助新的技术途径,提高微塑料识别的简便性和可靠性,并系统分析各类环境介质中微塑料的污染特征,明确微塑料的环境效应,对科学准确评价微塑料污染的环境风险具有重要意义.机器学习技术通过学习和解析大量数据建立结果评估或预测模型,目前已广泛应用于微塑料领域的相关研究.机器学习的应用可以提高视觉和光谱识别微塑料的自动化程度和识别效率,为微塑料污染溯源提供方法支撑并有助于揭示微塑料的复杂环境效应机制.通过综述机器学习技术在微塑料识别与环境风险评估中的应用研究进展,概括了机器学习在上述方向的应用特点和局限性,为机器学习在相关方向的发展和应用提出建议与展望.     

Vertical distribution of atmospheric aerosol size distribution over south-central New Mexico

The vertical distribution of background atmospheric aerosols was measured over south-central New Mexico as a part of the Atmospheric Lidar Verification Experiment (ALIVE) during four research field periods in the summers and winters of 1989 and 1990. Aerosol size distribution was measured from the surface up to 4500 m above sea level (asl) over the particle size range 0.1∼32 μm, using two Particle Measuring Systems (PMS) probes mounted on the wings of the NOAA King Air research aircraft. Vertical profiles of aerosol number concentrations of both fine- (0.1–2.0 μm) and coarse- (>2.0 μm) particle modes show seasonal differences, with higher number concentrations and higher mixed layer heights during summers. The measured aerosol size distribution data of each ALIVE intensive were averaged for boundary layer and free troposphere regions. These data mostly exhibit bi-modal distributions, typical for the continental atmospheric aerosols. Exceptions were the free troposphere size distributions measured during December 1989 (ALIVE III) and June 1990 (ALIVE IV), which resemble Junge's power-law distribution. Each of the averaged aerosol size distributions was approximated by the sum of log-normal distributions. Different characteristics of aerosol size distribution were observed between the two summer measurements of 1989 and 1990. Back-trajectory analysis revealed that decreased aerosol concentrations were observed during June 1990 when the air mass was transported from the southwestern U.S.A.     

Catalytic Pyrolysis of Biomass in a Fluidized Bed Reactor: Influence of the Acidity of H-Beta Zeolite

Catalytic pyrolysis of pine biomass was carried out in a fluidized bed reactor at 450°C. The acidic catalysts used as bed material in the reactor were β-zeolites with varying silica to alumina ratio, thus having different acid strengths. Quartz sand was used as bed material in the non-catalytic experiment. The yield of the different product phases was clearly influenced by the zeolites acidity. The chemical composition of the bio-oil was dependent on the catalyst. Zeolites with stronger acidity formed less organic oil, and respectively more water and polyaro-matic hydrocarbons than less acidic zeolites.     

Life cycle assessment of a railway tracks substructures: Comparison of ballast and ballastless rail tracks

The increase of train speed and axle load is an essential goal to make the railway transport more and more competitive for passengers and freights. On this basis, the unevenness of the railway track is crucial for the safety of the railway due to the high speed of the vehicle. Although ballasted tracks represent by far the most used railway track substructure, in recent years the modernization process has led the development of the ballastless track substructures.In deciding between the use of ballasted or ballastless track substructure there are many important technical, economical and environmental factors that have to be addressed. Based on the above, the principal objective of this study was to evaluate the environmental impact of different railway track substructures including ballast, cast-in sleeper and embedded track systems on the short, medium and long term. To accomplish this task, a life cycle assessment (LCA) was carried out throughout the entire life cycle of the railway infrastructure by using the ReCiPe (H) method. Although such approach is commonly included in the environmental assessment of building products and buildings, it was rarely applied in the analysis of the environmental impacts of railway track substructure.Thus, the result of these LCA showed that ballasted tracks cause the lowest environmental impact for service lives of up to 75 years. On the other hand, the embedded track beds cause the highest environmental impacts, regardless of their service life. The highest contributor for the environmental impacts of the track beds was the steel production.The results of this study will provide relevant environmental information for engineers and decision makers to select the most adequate railway track substructures for addressing issues related to the pursuit of sustainable development.     

Status and trends of terrestrial arthropod abundance and diversity in the North Atlantic region of the Arctic

Ambio - The Circumpolar Biodiversity Monitoring Programme (CBMP) provides an opportunity to improve our knowledge of Arctic arthropod diversity, but initial baseline studies are required to...     

A Comparison of Data Sets Varying in Spatial Accuracy Used to Predict the Occurrence of Wildlife-Vehicle Collisions

Wildlife-vehicle collisions (WVCs) pose a significant safety and conservation concern in areas where high-traffic roads are situated adjacent to wildlife habitat. Improving transportation safety, accurately planning highway mitigation, and identifying key habitat linkage areas may all depend on the quality of WVC data collection. Two common approaches to describe the location of WVCs are spatially accurate data derived from global positioning systems (GPS) or vehicle odometer measurements and less accurate road-marker data derived from reference points (e.g., mile-markers or landmarks) along the roadside. In addition, there are two common variable types used to predict WVC locations: (1) field-derived, site-specific measurements and (2) geographic information system (GIS)-derived information. It is unclear whether these different approaches produce similar results when attempting to identify and explain the location of WVCs. Our first objective was to determine and compare the spatial error found in road-marker data (in our case the closest mile-marker) and landmark-referenced data. Our second objective was to evaluate the performance of models explaining high- and low-probability WVC locations, using congruent, spatially accurate (<3-m) and road-marker (<800-m) response variables in combination with field- and GIS-derived explanatory variables. Our WVC data sets were comprised of ungulate collisions and were located along five major roads in the central Canadian Rocky Mountains. We found that spatial error (mean ± SD) was higher for WVC data referenced to nearby landmarks (516 ± 808 m) than for data referenced to the closest mile-marker data (401 ± 219 m). The top-performing model using the spatially accurate WVC locations contained all explanatory variable types, whereas GIS-derived variables were only influential in the best road-marker model and the spatially accurate reduced model. Our study showed that spatial error and sample size, using road-marker data for ungulate species, are important to consider for model output interpretation, which will impact the appropriate scale on which to apply modeling results. Using road-marker references <1.6 km or GPS-derived data locations may represent an optimal compromise between data acquisition costs and analytical performance. Electronic supplementary material  The online version of this article (doi:) contains supplementary material, which is available to authorized users.     

What humanlike errors do autonomous vehicles need to avoid to maximize safety?

Introduction: The final failure in the causal chain of events in 94% of crashes is driver error. It is assumed most crashes will be prevented by autonomous vehicles (AVs), but AVs will still crash if they make the same mistakes as humans. By identifying the distribution of crashes among various contributing factors, this study provides guidance on the roles AVs must perform and errors they must avoid to realize their safety potential. Method: Using the NMVCCS database, five categories of driver-related contributing factors were assigned to crashes: (1) sensing/perceiving (i.e., not recognizing hazards); (2) predicting (i.e., misjudging behavior of other vehicles); (3) planning/deciding (i.e., poor decision-making behind traffic law adherence and defensive driving); (4) execution/performance (i.e., inappropriate vehicle control); and (5) incapacitation (i.e., alcohol-impaired or otherwise incapacitated driver). Assuming AVs would have superior perception and be incapable of incapacitation, we determined how many crashes would persist beyond those with incapacitation or exclusively sensing/perceiving factors. Results: Thirty-three percent of crashes involved only sensing/perceiving factors (23%) or incapacitation (10%). If they could be prevented by AVs, 67% could remain, many with planning/deciding (41%), execution/performance (23%), and predicting (17%) factors. Crashes with planning/deciding factors often involved speeding (23%) or illegal maneuvers (15%). Conclusions: Errors in choosing evasive maneuvers, predicting actions of other road users, and traveling at speeds suitable for conditions will persist if designers program AVs to make errors similar to those of today’s human drivers. Planning/deciding factors, such as speeding and disobeying traffic laws, reflect driver preferences, and AV design philosophies will need to be consistent with safety rather than occupant preferences when they conflict. Practical applications: This study illustrates the complex roles AVs will have to perform and the risks arising from occupant preferences that AV designers and regulators must address if AVs will realize their potential to eliminate most crashes.