Bowen ratio-constrained global dataset of bulk air–sea turbulent heat fluxes from 1993 to 2017

Abstract

Air-sea turbulent heat fluxes, including the sensible heat flux (SHF) and latent heat flux (LHF), along with the Bowen ratio (β, ratio of SHF to LHF), are crucial for understanding air–sea interaction and global energy and water budgets. However, the existing products, primarily developed using the semi-empirical bulk aerodynamic methods and data-driven machine learning approaches, are often weak in accuracy and physical rationality, due to the uncertainties in the environmental forcings and inappropriate parameterizations. In this study, we generated a global daily 0.25° product of bulk air–sea turbulent heat fluxes using the Bowen ratio-constrained Neural Network (NN) model (referred to as the BrTHF model) that could coordinately estimate the SHF and LHF, along with the observations from 197 globally distributed buoys and multi-source remote sensing and reanalysis inputs. The spatial ten-fold cross-validation results showed that the BrTHF model achieved root mean square errors of 6.05, 23.67 W m−2 and 0.22 and correlation coefficients of 0.93, 0.91 and 0.25 for the SHF, LHF and β, respectively. Compared with the physics-agnostic NN model and seven widely used air–sea turbulent heat flux products (including JOFURO3, IFREMER, SeaFlux, ERA5, MERRA2, OAFlux, and OHF), the BrTHF model shows better agreement with observations, primarily attributable to its improved representation of β. Furthermore, the inter-comparison of the spatial distribution of multi-year means, as well as intra-annual and inter-annual change patterns showed that the BrTHF product reliably simulated global SHF, LHF and β, in contrast to the machine learning-based OHF product that failed to replicate these patterns. The main advantage of the BrTHF model lies in its improved rationality of β estimates, successfully eliminating the outliers observed in the physics-agnostic NN model and the seven typical products. The improved SHF, LHF, and β estimates can allow for more accurate quantification of the global air–sea energy and water budgets, enhance our understanding of air-sea interaction, and improve projections of climate change under global warming. The 0.25° daily global product from 1993 to 2017 can be freely accessed from the National Tibetan Plateau Data Center (TPDC) (https://doi.org/10.11888/Atmos.tpdc.302578, Tang and Wang, 2025).