A Systematic Literature Review on Numerical Weather Prediction Models and Provenance Data

Abstract

The weather has been an important issue for mankind since the earliest times. The need to predict the weather accurately increases every day when considering the effects of natural disasters such as floods, hails, extreme winds, landslides, etc. on many sectors from transportation to agriculture, which all depend on weather conditions. Numerical weather prediction (NWP) models, today's the de-facto tools used for weather forecasting, are scientific software that models atmospheric dynamics in accordance with the laws of physics. These models perform complex mathematical calculations on very large data (gridded) and require high computational power. For this reason, NWP models are scientific software that is usually run on distributed infrastructures and often takes hours to finish. On the other hand, provenance is another key concept as important as weather prediction. Provenance can be briefly defined as metadata that provides information about any kind of data, process, or workflow. In this SLR study, a comprehensive screening of literature was performed to discover primary studies that directly suggest systematic provenance structures for NWP models, or primary studies in which at least a case study was implemented on an NWP model even if considered in a broader perspective. Afterward, these primary studies were thoroughly examined in line with specific research questions, and the findings were presented in a systematic manner. An SLR study on primary studies which combines the two domains of NWP models and provenance research has never been done before. So we think that this work will fill an important gap in literature regarding studies combining the two domains and increase the interest in the subject.