Comparative Evaluation of Satellite- Based Merra-2 against Ground-based Data for PM

This study compares satellite-based observations of PM2.5 and Total Carbon (TC) concentrations derived from NASA’s MERRA-2 reanalysis dataset for the Ghaziabad region with ground-based observations at two sites (Site A and Site B) from June 2018 to May 2019. Results reveal consistent underestimation by MERRA-2, with daily mean biases of -85.34 (Site A) and -111.31 (Site B) for PM2.5, and -54.77 (Site A) to -59.08 (Site B) for TC, alongside monthly mean biases of -81.30 to -103.74 for PM2.5 and -54.77 to -59.08 for TC. The absolute error indicates a 49.01% to 53.85% underestimation for both PM2.5 and TC. Daily FAC2 values show around 39% agreement for PM2.5, reducing to 0.27 (Site A) and 0.3 (Site B) monthly, indicating reduced agreement over time. For TC, daily FAC2 is notably low at 0.04 (Site A) and 0.06 (Site B), with no monthly agreement within a factor of 2 of ground-level data. Strong correlations (R²=0.68-0.84 for PM2.5; R²=0.94-0.96 for TC) between biases and ground-level data are observed, indicating proportional relationships. However, discrepancies increase with higher PM2.5 mass concentration, highlighting MERRA-2’s limitations during elevated pollution periods. Significant monthly variations are observed in GLC-PM2.5 concentrations at Site A (F= 61.42, p= 1.68×10-11) and Site B (F=25.15, p=2.96×10-06), and For GLC-TC concentrations at Site A (F=103.85, p=4.10×10-16) and Site B (F=55.70, p=7.54×10-11). Both GLC and MERRA-2 PM2.5 and TC concentrations follow a consistent monthly pattern, with higher levels during post-monsoon and winter seasons and lower levels during monsoon and pre-monsoon periods. This study underscores MERRA-2’s limitations in estimating PM2.5 and TC concentrations compared to ground-based observations and emphasises the necessity for further refinement and validation of the MERRA-2 model to enhance accuracy across different spatial and temporal scales.