A team of researchers led by Mohammad Farazmand, associate professor of mathematics at NC State, has developed a new mathematical method for estimating global sea surface temperatures (SSTs) from sparse observations. The method, called the Sparse Discrete Empirical Interpolation Method (S-DEIM), combines mathematical analysis with machine learning to provide accurate estimates while substantially reducing computational time.

Sea surface temperatures are important for understanding marine ecosystems, climate, and weather patterns. However, observations from buoys and satellites are inherently limited, making it necessary to use mathematical and computational methods to estimate temperatures between available measurements.

The researchers tested S-DEIM using 30 years of data from the National Oceanic and Atmospheric Administration (NOAA), comparing its performance with the traditional Discrete Empirical Interpolation Method (DEIM) and a leading convolutional neural network (CNN) model. S-DEIM was 40% more accurate than DEIM and 2% more accurate than the CNN, while requiring only about one minute of training time, compared with approximately 90 minutes for the CNN.

The research has potential applications in both short-term weather forecasting and long-term climate prediction.

Importantly, the work grew out of the DRUMS Research Experience for Undergraduates (REU) at NC State, a program designed to introduce undergraduate students to mathematical research. REU participants Cassidy All (University of Colorado Boulder), Kevin Ho (Mississippi State University), Maya Magnuski (Bard College), and Christopher Nicolaides (Indiana University) contributed to the research. Louisa Ebby, a graduate student in mathematics at NC State, also contributed.

“This work shows that S-DEIM is capable of utilizing sparse data to provide accurate results while reducing training and computational time,” Farazmand says.

The research was published in the Journal of Geophysical Research: Machine Learning and Computation and was supported in part by the National Science Foundation.

https://news.ncsu.edu/2026/09/new-method-estimates-sea-surface-temps-quickly-and-accurately/