Reconstructing Stream Flow and Water Quality Time Series based on Sparse Measurements: Applications of Sparse Sensing

High-dimensional states can often leverage a latent low-dimensional representation. This inherent compressibility enables those high-dimensional states to be reconstructed or predicted from sparse measurements through sparse sensing. As a promising technique in data compression, reconstruction, and prediction, sparse sensing has not been widely used in environmental engineering and geosciences. In this talk, some recent efforts in reconstructing and/or predicting streamflow and water quality (e.g., nitrate and phosphorus concentrations) time series across watersheds using sparse sensing will be introduced. These works focused on exploring the applicability of sparse sensing on environmental signals and pursuing effective strategies to reduce the required sampling efforts. Other potential applications of sparse sensing include sensor location optimization, gap filling, and making predictions, especially through integration with data fusion.