Global ocean dimethyl sulfide climatology estimated from observations and an artificial neural network

Weilei Wang, University of California, Irvine, Earth System Science, Irvine, CA, United States, Francois Primeau, University of California Irvine, Earth System Science, Irvine, CA, United States, Eric S Saltzman, University of California Irvine, Irvine, CA, United States and Jefferson Keith Moore, University of California Irvine, Earth System Science, Irvine, United States
Abstract:
Marine dimethyl sulfide is of importance to the Earth’s climate due to its ability to alter the Earth’s radiation budget. However, the global-scale seasonal and annual DMS distributions and the factors controlling it have been elusive. Here we apply artificial neural network (ANN) techniques to a global database of 57,810 ship-based DMS measurements in surface waters to explore if a suite of environmental parameters can be used to predict DMS concentrations. We use the trained network to extrapolate the available DMS measurements into a global climatology with a monthly temporal resolution. We find that on global scales mixed layer depth and solar radiation are the strongest predictors, but they capture only 15% and 13% of the DMS variance, respectively. An ANN using lat-lon coordinate, time-of-year, solar radiation, sea surface temperature, salinity, nitrate, phosphate, silicate, and oxygen captures ~52% of the DMS concentration variance in our database. Like prior climatologies our results show a strong seasonal cycle in DMS with the highest concentrations occurring at high latitudes during the summer. The new DMS climatology will be compared to previous parameterizations and climatologies.