Abstract

The ability to make sense of data is valuable to climate action. The goal of this talk is to weaken the emerging challenges in climate science arising from the ongoing "data explosion" which will threaten the accessibility and usefulness of climate information in the coming decade through an appropriate application of statistical methods. Our proposal seeks to help science experts to accurately and efficiently forecast weather and climate, ensuring reliable information for decision makers in key sectors for the world´s development. More specifically, the proposed methods can be useful to support the planning and seek the safety of operations, such as, for example, predicting the occurrence of floods, droughts, strong winds, cyclones or cold and heat waves on different locations.

Weather and climate forecasts from mathematical models are essential for planning actions to mitigate the effects of meteorological and climate phenomena and for the development of preventive policies. According to the Research Division of the Air Space Control Institute, currently, the only tool capable of predicting the future state of the atmosphere are numerical weather forecast models. Institutions need reliable weather and climate information to make decisions and propose policies and actions consistent with the situation. Weather and climate modeling can help in a better understanding of meteorological and climate phenomena for a good use of the results by society.

A numerical weather forecasting model is a system developed to simulate the behavior of the atmosphere through mathematical equations, in order to predict the future state. There is an urgent need to improve meteorological and climate models and their predictions. Outputs from weather and climate numerical prediction models are never perfect. Given this scenario, one of the challenges for researchers is to improve the quality of weather and climate forecasts. Technically, we explore the potential of experimental design, big data methods and regression analysis in helping science experts and decision makers at drawing knowledge from a large volume of data.

This talk aims to provide a fundamental advance on how climate model ensembles are designed and analyzed, and is expected to accelerate time to science discovery by orders of magnitude. In doing so, this proposal will allow us to provide the same high quality climate information using less data and computational resources. The proposal addresses this challenge by reducing the computational effort using a new approach based on Design of Experiments (DoE).

The next frontier in climate action is not to produce more data, but to produce more information through a targeted reduction in the volume of data and an increase in its representativeness. This makes a substantial contribution to the United Nations Sustainable Development Goal 13, “Climate Action”, by directly influencing the way we produce representative and effective climate information.