Stochastic methods for temporal augmentation and quality improvement of time series datasets

This document provides a detailed description of the methodologies developed within EIFFEL project aiming to address aspects related with the temporal augmentation and quality improvement of time series datasets (see WP4 and T4.1). More specifically, the proposed approaches consist of a toolkit designed to cope with three, common in time series modelling, challenges/problems, that is: 1) the generation of statistically consistent stochastic realizations, 2) infilling of time series missing values, and 3) Lower-scale extrapolation (i.e., downscaling) of timeseries statistics. Further to the above, the report demonstrates the application of the methods via representative datasets of a variety of variables, which are also of high interest for the project’s pilots (e.g., precipitation, temperature, streamflow, air quality related quantities, etc.).