Global validation of columnar water vapor derived from EOS MODIS-MAIAC algorithm against the ground-based AERONET observations

• Main Authors: Martins, Vitor S., Lyapustin, Alexei, Wang, Yujie, Giles, David M., Smirnov, Alexander, Slutsker, Ilya, Korkin, Sergey
• Format: Text
• Language: English
• Published: Iowa State University Digital Repository 2019
• Subjects: MCD19A2; MODIS Collection 6; MAIAC water vapor; Time series analysis; Atmospheric Sciences; Bioresource and Agricultural Engineering; Climate; Environmental Monitoring; Aerosol Robotic Network
• Online Access: https://lib.dr.iastate.edu/abe_eng_pubs/1028; https://lib.dr.iastate.edu/cgi/viewcontent.cgi?article=2311&context=abe_eng_pubs

The water vapor is a relevant greenhouse gas in the Earth's climate system, and satellite products become one of the most effective way to characterize and monitor the columnar water vapor (CWV) content at global scale. Recently, a new product (MCD19) was released as part of MODIS (Moderate Resolution Imaging Spectroradiometer) Collection 6 (C6). This operational product from Multi-Angle Implementation for Atmospheric Correction (MAIAC) algorithm includes a high 1 km resolution CWV retrievals. This study presents the first global validation of MAIAC C6 CWV obtained from MODIS MCD19A2 product. This evaluation was performed using Aerosol Robotic Network (AERONET) observations at 265 sites (2000–2017). Overall, the results show a good agreement between MAIAC/AERONET CWV retrievals, with correlation coefficient higher than 0.95 and RMS error lower than 0.250 cm. The binned error analysis revealed an underestimation (~10%) of Aqua CWV retrievals with negative bias for CWV higher than 3.0 cm. In contrast, Terra CWV retrievals show a slope of regression close to unity and a low mean bias of 0.075 cm. While the accuracy is relatively similar between 1.0 and 5.0 cm for both sensor products, Terra dataset is more reliable for applications in humid tropical areas (>5.0 cm). The expected error was defined as ±15%, with >68% of retrievals falling within this envelope. However, the accuracy is regionally dependent, and lower error should be expected in some regions, such as South America and Oceania. Since MODIS instruments have exceeded their design lifetime, time series analysis was also presented for both sensor products. The temporal analysis revealed a systematic offset of global average between Terra and Aqua CWV records. We also found an upward trend (~0.2 cm/decade) in Terra CWV retrievals, while Aqua CWV retrievals remain stable over time. The sensor degradation influences the ability to detect climate signals, and this study indicates the need for revisiting calibration of the MODIS bands 17–19, mainly for Terra instrument, to assure the quality of the MODIS water vapor product. Finally, this study presents a comprehensive validation analysis of MAIAC CWV over land, raising the understanding of its overall quality.