Insights into the development of automated near real-time high-resolution rainfall maps in Hawai

A new Hawaii climate data collection system is being developed to automate the near real-time production of spatial continuous precipitation using station-based observations statewide. This effort builds on prior and ongoing work to compile a comprehensive high-resolution, spatially and temporally complete precipitation observation dataset for the Hawaiian Islands. A crucial component of this work is determining how to best leverage and integrate data from over 250 daily or sub daily reporting weather stations across steep precipitation gradients and often during extreme events. We share our findings regarding the effectiveness of various automated and data science methodologies tested and used across Hawaii's climatically heterogenous space. While approaches have been developed, assessed, and optimized to minimizing error for the intended time-step, comparison of alternative approaches at each stage in the process yields important insights into how different methods perform in particular climates and at differing temporal scales. Particular emphasis is given to the possible difficulties, utilities and performance of all components of the automated rainfall mapping process including, digital rainfall data acquisition from multiple types of online sources, automatic error detection using machine learning algorithms, gap filling data using statistical relationships, and several spatial interpolation methods at various temporal scales. We also describe how we have engaged rainfall stakeholders and climate data users through an interactive web-map based interface.