Evaluation of geostatistical and multiple regression models for assessment of spatial characteristics of carbon monoxide concentration in a data-limited environment
(2022) In Applied Geography 149.- Abstract
Efforts to manage air pollution problems caused by motorized transport in developing countries have frequently failed due to a lack of air quality data, both quantitatively and qualitatively. Air quality data from sparse monitoring stations is typically insufficient for spatial modelling, which is required for understanding and managing air quality conditions. This study aimed at assessing the empirical Bayesian kriging regression prediction (EBKRP) model's ability to predict the spatial concentration of CO in the study area using data from moderately sparse monitoring stations. Multiple Linear Regression (MLR) and Analysis of Variance (ANOVA) statistics were used to model the relationship between meteorological and traffic-related... (More)
Efforts to manage air pollution problems caused by motorized transport in developing countries have frequently failed due to a lack of air quality data, both quantitatively and qualitatively. Air quality data from sparse monitoring stations is typically insufficient for spatial modelling, which is required for understanding and managing air quality conditions. This study aimed at assessing the empirical Bayesian kriging regression prediction (EBKRP) model's ability to predict the spatial concentration of CO in the study area using data from moderately sparse monitoring stations. Multiple Linear Regression (MLR) and Analysis of Variance (ANOVA) statistics were used to model the relationship between meteorological and traffic-related variables and CO concentration, as well as the difference in CO concentrations between the different traffic scenarios (morning, afternoon and evening traffic periods). Data on CO concentration and CO predictor variables were collected simultaneously, every day, at 2 m and 5 m above ground level (agl), during each traffic scenario, at 21 monitoring stations from January to March 2022. The data was aggregated based on height agl and traffic scenarios. ArcGIS Pro 3.0 was used to implement the EBKRP model and MLR statistics, and SPSS was used to run the ANOVA. To assess the accuracy of the EBKRP model, various error metrics (mean error, mean standard error, root mean square error, and Pearson correlation R) were calculated. The results show that EBKRP performed optimally, with the root mean square error (RMSE) of 0.51 ppm and 0.23 ppm recorded for the predicted CO surfaces at the 2m and 5m agl, respectively. CO concentration were found to have a significant relationship with traffic volume, relative humidity, temperature, and wind speed, with statistical coefficients of 0.4, 0.15, −0.16, and −2.45; and 0.02, 0.21, −0.39, and −3.23 at 2m and 5m agl, respectively. The CO concentration in the study area differed significantly (p < 0.05) between the traffic periods, and the CO-predictor variables explained approximately 93% and 91% of the CO concentration at 2 m and 5 m, respectively. CO concentrations were higher at 2 m than at 5 m and increased from west to east. Overall, concentrations of CO in the study area were found to be within regulatory permissible limits. The current study lays the groundwork for using the EBKRP model in a data-limited context as well as understanding the meteorological and traffic-related factors that influence CO concentration in Uyo urban.
(Less)
- author
- Njoku, Elijah Akwarandu ; Akpan, Patrick Etim ; Effiong, Augustine Edet ; Babatunde, Isaac Oluwatosin ; Owoseni, Olujimi Afolabi and Olanrewaju, Joel Omoniyi
- publishing date
- 2022-12
- type
- Contribution to journal
- publication status
- published
- subject
- keywords
- Air quality, Carbon-monoxide, EBKRP, Geostatistics, Regression analysis, Uyo
- in
- Applied Geography
- volume
- 149
- article number
- 102816
- publisher
- Elsevier
- external identifiers
-
- scopus:85141798611
- ISSN
- 0143-6228
- DOI
- 10.1016/j.apgeog.2022.102816
- language
- English
- LU publication?
- no
- id
- 79427270-466c-41eb-909a-48ff2caa8dde
- date added to LUP
- 2022-11-30 12:32:31
- date last changed
- 2022-11-30 12:32:31
@article{79427270-466c-41eb-909a-48ff2caa8dde,
abstract = {{<p>Efforts to manage air pollution problems caused by motorized transport in developing countries have frequently failed due to a lack of air quality data, both quantitatively and qualitatively. Air quality data from sparse monitoring stations is typically insufficient for spatial modelling, which is required for understanding and managing air quality conditions. This study aimed at assessing the empirical Bayesian kriging regression prediction (EBKRP) model's ability to predict the spatial concentration of CO in the study area using data from moderately sparse monitoring stations. Multiple Linear Regression (MLR) and Analysis of Variance (ANOVA) statistics were used to model the relationship between meteorological and traffic-related variables and CO concentration, as well as the difference in CO concentrations between the different traffic scenarios (morning, afternoon and evening traffic periods). Data on CO concentration and CO predictor variables were collected simultaneously, every day, at 2 m and 5 m above ground level (agl), during each traffic scenario, at 21 monitoring stations from January to March 2022. The data was aggregated based on height agl and traffic scenarios. ArcGIS Pro 3.0 was used to implement the EBKRP model and MLR statistics, and SPSS was used to run the ANOVA. To assess the accuracy of the EBKRP model, various error metrics (mean error, mean standard error, root mean square error, and Pearson correlation R) were calculated. The results show that EBKRP performed optimally, with the root mean square error (RMSE) of 0.51 ppm and 0.23 ppm recorded for the predicted CO surfaces at the 2m and 5m agl, respectively. CO concentration were found to have a significant relationship with traffic volume, relative humidity, temperature, and wind speed, with statistical coefficients of 0.4, 0.15, −0.16, and −2.45; and 0.02, 0.21, −0.39, and −3.23 at 2m and 5m agl, respectively. The CO concentration in the study area differed significantly (p < 0.05) between the traffic periods, and the CO-predictor variables explained approximately 93% and 91% of the CO concentration at 2 m and 5 m, respectively. CO concentrations were higher at 2 m than at 5 m and increased from west to east. Overall, concentrations of CO in the study area were found to be within regulatory permissible limits. The current study lays the groundwork for using the EBKRP model in a data-limited context as well as understanding the meteorological and traffic-related factors that influence CO concentration in Uyo urban.</p>}},
author = {{Njoku, Elijah Akwarandu and Akpan, Patrick Etim and Effiong, Augustine Edet and Babatunde, Isaac Oluwatosin and Owoseni, Olujimi Afolabi and Olanrewaju, Joel Omoniyi}},
issn = {{0143-6228}},
keywords = {{Air quality; Carbon-monoxide; EBKRP; Geostatistics; Regression analysis; Uyo}},
language = {{eng}},
publisher = {{Elsevier}},
series = {{Applied Geography}},
title = {{Evaluation of geostatistical and multiple regression models for assessment of spatial characteristics of carbon monoxide concentration in a data-limited environment}},
url = {{http://dx.doi.org/10.1016/j.apgeog.2022.102816}},
doi = {{10.1016/j.apgeog.2022.102816}},
volume = {{149}},
year = {{2022}},
}