Augmented mapping of the Extreme Heat and Cold Events (EHE/ECE) at continental scale with cloud-based computing

Harvard Medical School, Boston MA

Investigators

Linked publications & trials

Abstract

Project Summary/Narrative Extreme heat and cold events (EHE/ECE) have been linked to a range of adverse health outcomes from exacerbated pre-existing conditions to mental illness and respiratory and cardiovascular disease Previous research has often determined the areas and population impacted by EHE/ECE through simplistic methods that assign temperature data from the closest weather station to the population being studied (e.g., a census tract or postal zipcode). Our preliminary analysis has demonstrated that dynamic spatial-temporal methodologies significantly alleviate misclassifications that tend to occur in conventional approaches. Implementing more sophisticated models with higher spatial and temporal resolution can pose computational complexity which hinders application and scalability of the dynamic models. Here we propose a hybrid method to leverage cloud computing resources to streamline and scale up EHE/ECE identification workflows with improved specification to help configuration of on-premises computing. Aim 1: Improving scalability and computational efficiency of detecting extreme climate events in-cloud versus on-premises computing We will develop and implement computational methodologies to (1) scale up the spatial interpolation methods at continental scale using parallel and distributed computing algorithms, (2) monitor and assess the performance of these algorithms in terms of computational time, memory allocation and storage resources compared to the dedicated server utilization and conventional High-Performance Computing (HPC) approach.We hypothesize that cloud computing will improve efficiency of current methods which have been implemented on on-premises computing infrastructure using all-in memory solutions and serialized data pipeline. We will leverage efficiencies of spatially enabled databases along with DevOps tools and services such as containerization, and automated deployment to streamline our research workflows. Aim 2: Improving accuracy and robustness of extreme climate events identification in-cloud versus on-premises computing We will assess the robustness of dynamic EHE/ECE delineation methods when applied to heterogeneous climatological data at continental geographies and beyond. Specifically, we will evaluate the extent to which cloud computing improves the accuracy of the spatial-temporal methods in identifying populations and areas impacted by EHE/ECE using different model parameterization scenarios. We hypothesize that cloud computing will improve the accuracy and robustness of EHE/ECE identification methods by (1) facilitating development of more complex models that take into account additional environmental variables, (2) by streamlining reproducibility practices that enables the wider scientific community to test and validate the models at multiple scales that results in more reliable models