Accelerating Programmable Web Analytics, Web Services, and Open and Reproducible Deep Learning Models for Hydroinformatics Research

The convergence of Internet of Things (IoTs), Big Data analytics, and deep learning and the ability to use them, is at a tipping point where significant changes, transformation, and innovation are poised to take place. This convergence enabled developing exceptional approaches to collect and manage seasonal, daily, and event-triggered catastrophic readiness and countless actionable data loads and tools for hydroinformatics research. The goal of this presentation is to discuss how programmable Web data processing, Web services, and open and reproducible deep learning models can benefit hydroinformatics research communities. Recent efforts toward this goal have fostered around IoT Application Programming Interfaces (APIs) development for batch and real-time collection of the data generated by IoT devices. This also includes (i) data veracity to link, match, cleanse and transform data across systems, (ii) provenance and reproducibility of deep learning pipelines to define repeatable and reusable steps for modeling processes, (iii) reusable software environments for training and deploying models, and (iv) data container development by bundling the application code together with the related configuration files, libraries, and dependencies required for it to runable to run across any platform or cloud, free of issues. This study will discuss how an ideal hydroinformatics system with data veracity, integration, reusable software environment, and reproducible deep learning combined with high-powered analytics could be designed to benefit hydroinformatics research communities.