Weather conditions are an important factor that influences everyday life and decision making in specialized fields like agriculture, industry, transportation, water management etc. Often decisions are taken under uncertain information due to the probabilistic nature of weather forecasting which is inherently uncertain. This makes the communication task very challenging for forecasters attempting to communicate the uncertainties surrounding the predicted event. This is particularly true for precipitation forecasts, since precipitation is perceived by general public as the most important weather component influencing decisions. In many countries precipitation forecasts are conveyed through probabilistic information while in others, like Italy, through linguistic expressions. In both cases lay people should interpret uncertainty to make decisions.
 
This work examine two years of current day weather bulletins for Tuscany Region (n: 772 issues, from July 2009 to August 2011), provided by the forecasters of the official meteorological agency (LaMMA), in order to identify the textual and visual elements expressing uncertainty within the bulletin. The analysis is focused on the specific subset (n: 484) of forecast of precipitation events. To quantify uncertainty two ranking score indexes were proposed for the visual and textual component respectively. The visual uncertainty was measured by evaluating only the icons pertaining precipitation events and assigning to each one a definite descending score, starting from the more complex and ambiguous. Likewise, textual uncertainty was calculated by assigning a ranking score to terms (using a selected Italian subset of words) expressing probability (like “probable” or “possible”) and ambiguous terms referred to type, amount, time and spatial prediction of precipitation events. Preliminary results show that it is suitable to build a statistically consistent framework to evaluate uncertainty in weather bulletins. Indeed textual uncertainty have shown association with variables as: the season, less uncertainty in winter and higher in spring; the forecaster’s experience, less experienced expressing more uncertainty; the visual complexity and icons heterogeneity of the weather map; the visual icons score, high occurrence of composite icons calling for more uncertain terms. Finally a strong connection emerges between textual metrics parameters of bulletin statement and the value of textual uncertainty measured by the defined indexes.
">
 [PCST]
PCST Network

Public Communication of Science and Technology

 

Communicating uncertainty
The case of precipitation forecasts in Tuscany, Italy

Valentina Grasso   CNR Ibimet - Consorzio LaMMA

Federica Zabini   CNR Ibimet - Consorzio LaMMA

Alfonso Crisci   CNR Ibimet

Valerio Capecchi   CNR Ibimet - Consorzio LaMMA

Claudio Tei   CNR Ibimet - Consorzio LaMMA

Weather conditions are an important factor that influences everyday life and decision making in specialized fields like agriculture, industry, transportation, water management etc. Often decisions are taken under uncertain information due to the probabilistic nature of weather forecasting which is inherently uncertain. This makes the communication task very challenging for forecasters attempting to communicate the uncertainties surrounding the predicted event. This is particularly true for precipitation forecasts, since precipitation is perceived by general public as the most important weather component influencing decisions. In many countries precipitation forecasts are conveyed through probabilistic information while in others, like Italy, through linguistic expressions. In both cases lay people should interpret uncertainty to make decisions.
 
This work examine two years of current day weather bulletins for Tuscany Region (n: 772 issues, from July 2009 to August 2011), provided by the forecasters of the official meteorological agency (LaMMA), in order to identify the textual and visual elements expressing uncertainty within the bulletin. The analysis is focused on the specific subset (n: 484) of forecast of precipitation events. To quantify uncertainty two ranking score indexes were proposed for the visual and textual component respectively. The visual uncertainty was measured by evaluating only the icons pertaining precipitation events and assigning to each one a definite descending score, starting from the more complex and ambiguous. Likewise, textual uncertainty was calculated by assigning a ranking score to terms (using a selected Italian subset of words) expressing probability (like “probable” or “possible”) and ambiguous terms referred to type, amount, time and spatial prediction of precipitation events. Preliminary results show that it is suitable to build a statistically consistent framework to evaluate uncertainty in weather bulletins. Indeed textual uncertainty have shown association with variables as: the season, less uncertainty in winter and higher in spring; the forecaster’s experience, less experienced expressing more uncertainty; the visual complexity and icons heterogeneity of the weather map; the visual icons score, high occurrence of composite icons calling for more uncertain terms. Finally a strong connection emerges between textual metrics parameters of bulletin statement and the value of textual uncertainty measured by the defined indexes.

A copy of the full paper has not yet been submitted.

BACK TO TOP