J4C.6 Estimating Near-Surface Marine Air Temperature Using High-resolution Infrared Radiation Sounder Data and In Situ Observations

Monday, 29 January 2024: 5:30 PM
338 (The Baltimore Convention Center)
Yuhan Rao, North Carolina State University, Asheville, NC; Cooperative Institute for Satellite Earth System Studies, Asheville, NC; and J. L. Matthews and L. Shi

Near-surface air temperature is one of the essential climate variables that play a key role in climate monitoring and applications. Historically, sea surface temperature (SST) has been used to monitor temperature change over the ocean surface as a proxy due to the lack of reliable and consistent near-surface marine air temperature (MAT). Since the late 1970s, polar-orbiting environmental satellites carried High-resolution Infrared Radiation Sounder (HIRS) to collect information across different pressure levels of the atmosphere globally. HIRS data has been used to estimate the temperature and humidity values at standard pressure levels including near-surface air temperature. However, the estimated temperature and humidity data have limitations because it is only available during clear-sky conditions. Despite the clear-sky-only nature, HIRS-based temperature estimates still provide useful information for monitoring the change of near-surface MAT. In this research, we present our method to create a blended near-surface MAT data set using a gradient-boosting tree model to combine HIRS temperature retrievals with ship- and buoy-based measurements of MAT from the International Comprehensive Ocean and Atmospheric Data Set (ICOADS). The gradient-boosting tree model was first trained using match-up data pairs between HIRS temperature retrievals and height-adjusted MAT observations from ICOADS and other ancillary data. The trained model was then evaluated using independent observations of near-surface MAT from diverse data sources including NOAA research ships and international field campaigns. The overall evaluation shows that the gradient-boosting tree model-based MAT estimates are accurate when compared with the independent observations (bias: 0.08K; RMSE: 1.1K). This method will allow us to estimate near-surface MAT globally from 1979 and can be extended using the most recent infrared sounding instruments which can provide more than 40 years of long-term records for climate monitoring and applications.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner