Sea surface temperature mapping using data from satellite-based MTVZA-GYa microwave radiometer
A.O. Maslyashova, A.B. Uspenskiy
Scientific Research Center of Space Hydrometeorology «Planeta», Moscow, Russia
Keywords: sea surface temperature, microwave radiometer MTVZA-GYa, ERA5 reanalysis, ICOADS database, artificial neural network
Abstract
A method for remote mapping of the sea surface temperature field (SST) in a cloudless and cloudy atmosphere has been proposed and tested based on SST measurements of MTVZA-GYa microwave radiometer from Meteor M satellite Nos. 2-2 and 2-4. The method includes the preliminary SST estimation using an artificial neural network (ANN) algorithm of multilayer perceptron type and statistical filtering of the preliminary estimates using climatic SST values calculated from the ERA5 reanalysis. The neural network algorithm uses antenna temperatures measured in five scanner channels of the MTVZA-GYa radiometer as predictors. Reference SST values from the open access ICOADS database are used to train the ANN. The statistical filtering procedure makes it possible to reduce the influence of clouds and precipitation in the satellite radiometer field of view and provides a root-mean-square error of the SST estimates on the order of 1.2-1.7 °C and a coefficient of determination of about 0.8-0.9 when compared with in situ observations. The proposed approach is applicable to operational global “all-weather" monitoring of sea surface temperature and can be adapted to analyze SST measurements of MTVZA-GYa type radiometers with improved technical and information characteristics.
|
