Researchers have developed a new method for estimating sea surface temperatures (SST) from limited data that is significantly more accurate than traditional mathematical models and only slightly less accurate than the best artificial intelligence (AI) models, while requiring far less time to train. This advancement could improve both short-term weather forecasts and long-term climate predictions. SST is crucial for understanding marine ecosystems, climate change, and weather patterns, but collecting comprehensive data is challenging. Current methods include buoys, which provide accurate but limited measurements, and satellites, which cover more area but can be affected by atmospheric conditions. To overcome these limitations, oceanographers have traditionally used complex mathematical models to estimate SST from sparse data. Federal agencies like the National Oceanic and Atmospheric Administration (NOAA) have historically relied on these models, which use complicated differential equations. More recently, AI models, such as convolutional neural networks (CNNs), have been developed to improve accuracy, but they are computationally expensive and require extensive training time. A previous method called the Discrete Empirical Interpolation Method (DEIM) attempted to address missing data by using a set of patterns to estimate SST. However, DEIM struggled with sparse data. To improve on this, Farazmand and their team created a new approach called the Sparse Discrete Empirical Interpolation Method (S-DEIM). This method uses historical data to estimate a "kernel vector," which helps fill in gaps in the data. Unlike DEIM, S-DEIM is better suited for sparse data. In testing, the S-DEIM method outperformed both DEIM and the best-performing CNN model by 40% and 2%, respectively. It also required only one minute of training time compared to the CNN's 1.5 hours. The researchers hope to further refine the method's accuracy. This work, published in the Journal of Geophysical Research: Machine Learning and Computation, was conducted as part of a research experience for undergraduates. Contributors included students from various universities, including the University of Colorado Boulder, Mississippi State University, Bard College, Indiana University, and a graduate student from North Carolina State University.