Journal article
Machine learning-driven prediction of Visual Range under changing climate conditions over complex terrain using AOD and CMIP6 climate simulations
Remote sensing applications, Vol.39, 101712
08/01/2025
DOI: 10.1016/j.rsase.2025.101712
Abstract
Visibility through the atmosphere, or Visual Range (VR), is a key indicator of ambient air quality, especially in areas with complex topography and vulnerability to climate change. The specific aims of this study were to (1) evaluate the ability of the Coupled Model Intercomparison Project Phase 6 (CMIP6) climate model outputs and satellite Aerosol Optical Depth (AOD) to predict VR across diverse topography; (2) select important meteorological parameters for VR; and (3) design an ensemble Machine Learning (ML) model with high accuracy using Bagged Extreme Gradient Boosting (BG-XG) for long-term VR trends under future climate scenarios. This study contributes to the significant gap in regional visibility prediction by combining climate model projections, remotely sensed AOD, and ML to project future VR through 2100 across Pakistan. The BG-XG model was trained using in situ meteorological data, AOD, and six CMIP6 models (Euro-Mediterranean Centre on Climate Change Climate Model 2 High Resolution – version SR5 (CMCCCM2-SR5 (Italy))was the most consistently accurate model across all the topography). For the results computed at Lahore (LHR), the BG-XG model achieved the highest correlation coefficient of R = 0.98 and Root Mean Square Error (RMSE) = 0.24 km for the validation dataset. It is expected that the region will observe an average VR of 5.88 km with a standard deviation of 1.66 km by the end of 2100. The predictive strength of climate model parameters for VR was high (>90 %), with significant dependencies on sea-level pressure (SLP), relative humidity (RH), eastward wind (EW), and AOD. The region is expected to witness a significant decrease in average VR at a rate of −281.3 m/year due to an increase in AOD at a rate of 0.14/year from 2003 to 2100. Among the regions, Karachi (KHI) is anticipated to experience the most substantial reduction in VR by 2100, followed by Sindh and the northwestern areas. This study provides the first long-term, region-specific VR forecasts for Pakistan by integrating ML with CMIP6 climate projections. These findings can guide climate adaptation strategies, particularly for regions at considerable risk of declining air quality due to reduced visibility.
Details
- Title: Subtitle
- Machine learning-driven prediction of Visual Range under changing climate conditions over complex terrain using AOD and CMIP6 climate simulations
- Creators
- Sadaf Javed - COMSATS University IslamabadMuhammad Imran Shahzad - COMSATS University IslamabadMuhammad Zeeshaan Shahid - University of the PunjabJun Wang - University of IowaImran Shahid - Qatar University
- Resource Type
- Journal article
- Publication Details
- Remote sensing applications, Vol.39, 101712
- DOI
- 10.1016/j.rsase.2025.101712
- ISSN
- 2352-9385
- eISSN
- 2352-9385
- Publisher
- ELSEVIER
- Language
- English
- Date published
- 08/01/2025
- Academic Unit
- Electrical and Computer Engineering; Civil and Environmental Engineering; Physics and Astronomy; Chemical and Biochemical Engineering
- Record Identifier
- 9984962545902771
Metrics
1 Record Views