Scholarlink

<p>Various research approaches are adopted to assess and estimate the severity of the air pollution scenario locally as well globally. Indian cities are severely affected by the particulate matters (PM) pollution. Regular PM<sub>2.5</sub> concentration monitoring is essential to investigate the associated health and climatic effects. The region selected for this study i.e., Eastern part of India has scarcity in the observational site, and measurements of PM<sub>2.5</sub> is even more thus we must look for the alternative method to derive the concentration of it. ANN technique is an upcoming technique in different research fields. In the current work, we have evaluated the performance of an ANN with its parameters in simulating the PM<sub>2.5</sub> concentrations. The study is performed over Ranchi, capital city of Jharkhand for a time of Jan22-Feb23 with daily data. Various pollutants concentrations data and meteorological parameters supplied as input to this model, to project PM<sub>2.5</sub> concentration. Results emphasized the necessity of optimal learning rate values and iteration counts to prevent overfitting and obtain precise results at a cheap computing cost. As the number of iterations increased (from 400 to 450, 500), we saw a 12–24% increase in processing time with slight change in correlations. Therefore, for precise estimations, these parameters should be carefully chosen. Simulation results showed that the ANN model captured the pattern of variation of PM<sub>2.5</sub> concentration with good accuracy with slight over/underestimation in some cases. The majority of ANN simulated concentrations were closed to observed values. About ~ 70% of the predicted PM<sub>2.5</sub> concentration was below ± 15% of the observed concentration. Role of atmospheric boundary layer (ABL) height and ventilation coefficient (VC) in ANN performance is also investigated, and a faster convergence is notice with the incorporation these along with earlier inputs. Finding showed over a region where very low observational sites exist; this type of study will be advantageous for the researcher to use this technique to derived particulate matter concentration.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Assessment of health hazardous particulate matter (PM2.5) from artificial neural network using meteorological and pollutant parameters

  • 期刊论文

  • 发表日期：

    2025年1月8日

查看全文
  • Tanvi Gaurav,
  • Nishi Srivastava

摘要

Various research approaches are adopted to assess and estimate the severity of the air pollution scenario locally as well globally. Indian cities are severely affected by the particulate matters (PM) pollution. Regular PM2.5 concentration monitoring is essential to investigate the associated health and climatic effects. The region selected for this study i.e., Eastern part of India has scarcity in the observational site, and measurements of PM2.5 is even more thus we must look for the alternative method to derive the concentration of it. ANN technique is an upcoming technique in different research fields. In the current work, we have evaluated the performance of an ANN with its parameters in simulating the PM2.5 concentrations. The study is performed over Ranchi, capital city of Jharkhand for a time of Jan22-Feb23 with daily data. Various pollutants concentrations data and meteorological parameters supplied as input to this model, to project PM2.5 concentration. Results emphasized the necessity of optimal learning rate values and iteration counts to prevent overfitting and obtain precise results at a cheap computing cost. As the number of iterations increased (from 400 to 450, 500), we saw a 12–24% increase in processing time with slight change in correlations. Therefore, for precise estimations, these parameters should be carefully chosen. Simulation results showed that the ANN model captured the pattern of variation of PM2.5 concentration with good accuracy with slight over/underestimation in some cases. The majority of ANN simulated concentrations were closed to observed values. About ~ 70% of the predicted PM2.5 concentration was below ± 15% of the observed concentration. Role of atmospheric boundary layer (ABL) height and ventilation coefficient (VC) in ANN performance is also investigated, and a faster convergence is notice with the incorporation these along with earlier inputs. Finding showed over a region where very low observational sites exist; this type of study will be advantageous for the researcher to use this technique to derived particulate matter concentration.