Determining the influence of meteorological, environmental, and anthropogenic activity variables on the atmospheric CO2 concentration in the arid and semi-arid regions: A case study in the Middle East

Abstract

Climate change and the resulting warming caused by human activities, such as releasing greenhouse gases (GHG), notably carbon dioxide (CO₂), are paramount concerns in the contemporary era. Therefore, it is essential to understand spatial and temporal variations of atmospheric carbon dioxide concentration (XCO₂) on national and international scales and the role of controlling variables. This research examines the role of controlling variables on XCO₂ by utilizing spatiotemporal modeling, correlation analysis, and machine learning methods to determine the sinks and sources of CO₂ and their changes in the Middle East (ME). First, XCO₂ data from the OCO-2 satellite spanning from 2015 to 2021 were employed for spatiotemporal analysis. The temporal analysis of XCO₂ showed peak levels during spring (May) and the lowest levels during summer (September). Additionally, the spatial dispersion of XCO₂ revealed significant spatial heterogeneity influenced by land cover. Areas with high vegetation abundance and suitable weather conditions exhibited minimum XCO₂ levels, while vice versa occurred. Next, the correlation analysis between the principal controlling variables and XCO₂ revealed that, except for vegetation abundance and anthropogenic CO₂ emissions, they exhibited predominantly negative and positive correlations throughout the year, respectively. However, the correlation patterns for temperature, precipitation, humidity, and wind speed varied across different months, showing both negative and positive relationships depending on the month. Recognizing that simple correlation alone cannot determine which variables played the most significant role in XCO₂ each month, this study employed machine learning and Permutation Feature Importance (PFI) techniques. The results showed that, except for March and September, when precipitation and wind speed respectively had the most significant influence on determining XCO₂, air temperature played the dominant role in other months. Additionally, the average monthly results revealed that air temperature, wind speed, precipitation, and vegetation abundance, respectively, played the most significant roles in XCO₂. In contrast, CO₂ emissions had the most negligible impact. These results highlight the considerable influence of meteorological and environmental variables on regulating XCO₂ levels and distributions throughout the ME. By elucidating the seasonality of the carbon cycle and identifying key XCO₂ drivers over the ME, this study provides valuable insights to guide regional climate change mitigation policies and further analysis of vulnerable semi-arid ecosystems.