A0161
Title: Spatio-temporal deepkriging for interpolation and probabilistic forecasting
Authors: Pratik Nag - King Abdullah University of Science and Technology (Saudi Arabia) [presenting]
Abstract: Gaussian processes (GP) and Kriging are widely used in traditional spatio-temporal modelling and prediction. These techniques typically presuppose that the model has a parametric covariance structure and the data are observed from a stationary GP. However, processes in real-world applications often exhibit non-Gaussianity and nonstationarity with complex dependence structures. Moreover, it is well-known that the likelihood-based inference for GPs is computationally expensive and thus prohibitive for large datasets. We propose a deep neural network (DNN) based two-stage model for spatio-temporal interpolation and forecasting. Interpolation is performed in the first step, which utilizes a dependent DNN with the embedding layer constructed by spatio-temporal basis functions. For the second stage, we propose to use Long-Short Term Memory (LSTM) and convolutional LSTM to forecast future observations at a given location. We adopt the quantile-based loss function in the DNN to provide probabilistic forecasting. Compared to Kriging, the proposed method does not require specifying covariance functions or making stationarity assumption, and is computationally efficient. Therefore, it is suitable for large-scale prediction of complex spatio-temporal processes. We apply our method to daily evapotranspiration data at more than 1 million locations from January 2019 to December 2021 for fast imputation of missing values and provide forecasts with uncertainties.
