Publications

Albero: A Visual Analytics Approach for Probabilistic Weather Forecasting

A. Diehl, L. Pelorosso, K. Matkovic, J. Ruiz, M. E. Gröller, and S. Bruckner

Abstract

Probabilistic weather forecasts are amongst the most popularways to quantify numerical forecast uncertainties. The analogregression method can quantify uncertainties and express them asprobabilities. The method comprises the analysis of errorsfrom a large database of past forecasts generated with a specificnumerical model and observational data. Current visualizationtools based on this method are essentially automated and provide limitedanalysis capabilities. In this paper, we propose a novelapproach that breaks down the automatic process using the experience andknowledge of the users and creates a new interactivevisual workflow. Our approach allows forecasters to study probabilisticforecasts, their inner analogs and observations, theirassociated spatial errors, and additional statistical information bymeans of coordinated and linked views. We designed thepresented solution following a participatory methodology together withdomain experts. Several meteorologists with differentbackgrounds validated the approach. Two case studies illustrate thecapabilities of our solution. It successfully facilitates theanalysis of uncertainty and systematic model biases for improveddecision-making and process-quality measurements.

A. Diehl, L. Pelorosso, K. Matkovic, J. Ruiz, M. E. Gröller, and S. Bruckner, "Albero: A Visual Analytics Approach for Probabilistic Weather Forecasting," Computer Graphics Forum, vol. 36, iss. 7, p. 135–144, 2017. doi:10.1111/cgf.13279
[BibTeX]

Probabilistic weather forecasts are amongst the most popularways to quantify numerical forecast uncertainties. The analogregression method can quantify uncertainties and express them asprobabilities. The method comprises the analysis of errorsfrom a large database of past forecasts generated with a specificnumerical model and observational data. Current visualizationtools based on this method are essentially automated and provide limitedanalysis capabilities. In this paper, we propose a novelapproach that breaks down the automatic process using the experience andknowledge of the users and creates a new interactivevisual workflow. Our approach allows forecasters to study probabilisticforecasts, their inner analogs and observations, theirassociated spatial errors, and additional statistical information bymeans of coordinated and linked views. We designed thepresented solution following a participatory methodology together withdomain experts. Several meteorologists with differentbackgrounds validated the approach. Two case studies illustrate thecapabilities of our solution. It successfully facilitates theanalysis of uncertainty and systematic model biases for improveddecision-making and process-quality measurements.
@ARTICLE {Diehl-2017-AVA,
author = "Alexandra Diehl and Leandro Pelorosso and Kresimir Matkovic and Juan Ruiz and Meister Eduard Gr{\"o}ller and Stefan Bruckner",
title = "Albero: A Visual Analytics Approach for Probabilistic Weather Forecasting",
journal = "Computer Graphics Forum",
year = "2017",
volume = "36",
number = "7",
pages = "135--144",
month = "oct",
abstract = "Probabilistic weather forecasts are amongst the most popularways to quantify numerical forecast uncertainties. The analogregression method can quantify uncertainties and express them asprobabilities. The method comprises the analysis of errorsfrom a large database of past forecasts generated with a specificnumerical model and observational data. Current visualizationtools based on this method are essentially automated and provide limitedanalysis capabilities. In this paper, we propose a novelapproach that breaks down the automatic process using the experience andknowledge of the users and creates a new interactivevisual workflow. Our approach allows forecasters to study probabilisticforecasts, their inner analogs and observations, theirassociated spatial errors, and additional statistical information bymeans of coordinated and linked views. We designed thepresented solution following a participatory methodology together withdomain experts. Several meteorologists with differentbackgrounds validated the approach. Two case studies illustrate thecapabilities of our solution. It successfully facilitates theanalysis of uncertainty and systematic model biases for improveddecision-making and process-quality measurements.",
pdf = "pdfs/Diehl-2017-AVA.pdf",
images = "images/Diehl-2017-AVA.jpg",
thumbnails = "images/Diehl-2017-AVA.png",
youtube = "https://www.youtube.com/watch?v=-yqoeEgkz28",
doi = "10.1111/cgf.13279",
keywords = "visual analytics, weather forecasting, uncertainty"
}
projectidprojectid

Media

Downloads

Full paper [PDF]