We will have a meeting on Friday (February 27) at 3pm in SAS 5270 with speaker Louisa Ebby.
Title: Estimating global sea surface temperatures from sparse in situ observations
Abstract: Accurate estimates of global sea surface temperatures (SST) are essential for weather forecasting and climate projections. We infer high-resolution SST estimates from sparse in situ measurements using a recent data-driven and equation-free state estimation technique called Sparse Discrete Empirical Interpolation Method (S-DEIM). S-DEIM has two components: a deterministic component which is evaluated from instantaneous in situ measurements, and a kernel component which is inferred from time series of past in situ measurements. We use historical SST reanalysis data to train a recurrent neural network that infers the kernel component. S-DEIM exhibits significantly reduced reconstruction errors compared to existing methods. We demonstrate the utility of S-DEIM for rapid state estimation from sparse streaming observational data in real-time.