天气预报中不确定性问题的调查分析
A Survey on Forecasters’ View about Uncertainty in Weather Forecasts
中文摘要：
围绕天气预报中的不确定性问题，对中国18个省（市、区）229名代表了国家、省和地市3级以及不同岗龄的预报员进行了问卷调查。具体调查内容包括了以下3个方面：（1）应用数值预报的情况；（2）对天气预报准确率的看法；（3）对天气预报（数值预报）中不确定性、集合预报产品以及预报发布形式等问题的看法。结果表明，数值预报产品目前已经成为预报员们制作日常天气预报的主要工具。因此，数值预报所固有的可预报性（不确定性）问题已是预报员不得不面对的一个科学问题。最后，介绍了一项新近关于美国预报员在日常天气预报中应用集合预报信息所遇到的一些挑战，这包括“资料不全面、分析工具不足、培训教育不够、缺乏检验、工作负荷重、模式分辨率低以及预报规范格式的限制”7方面。这些可供气象学科带头人和业务管理决策者思考和计划。
中文关键词：
集合数值预报和产品，天气预报准确率和不确定性，完全预报和不完全预报，预报员和问卷调查
KeyWords：
ensemble numerical weather prediction and product, weather forecast accuracy and uncertainty, complete and incomplete forecasts, forecaster and questionnaire
Abstract：
A nation-wide survey was conducted to 229 forecasters who came from 18 provinces representing both senior and junior forecasters at all three levels (national, provincial and regional) in China. Three issues are surveyed: (1) usage of numerical weather prediction (NWP) in daily operational weather forecasting; (2) opinion about weather forecast accuracy; and (3) view about uncertainty in forecasts including ensemble products and methods to convey forecast confidence. The result shows that NWP has been applied by all forecasters to their daily preparations of weather forecasting. Therefore, forecast uncertainty or predictability issue associated with highly nonlinear NWP models is a reality which forecasters have to face day by day. Long-term education is needed to train forecasters in effectively using and correctly interpreting forecast uncertainty information. Regarding weather forecast accuracy, all forecasters admit that any forecast has errors and believe that it will never be perfect no matter how advanced meteorological sciences are, given chaotic nature of the atmospheric system. With imperfect forecasts, most forecasters believe that specialized forecast products particularly tailored to a certain type of customer as well as frequent updates of forecast information are important to maximize the economical and societal values of weather forecasts. In spite of the variation in forecast accuracy, most forecasters have, however, an idea about their confidence or difficulty level of a forecast he or she is issuing beforehand. Being asked if such forecast confidence information should be released to the public and users, most forecasters responded with a positive “yes” answer although a few of them worried that the pubic and users might be confused and do not know how to use these information correctly. To fully accomplish this, forecasters acknowledge that forecast-issuing format/context must also need to be reformed accordingly. Furthermore, the majority of forecasters welcome the National NWP Center to provide ensemble-based quantitative forecast uncertainty products for them to use although the “single-solution” type of ensemble products is currently more preferred to “multi-solution” type. Therefore, the “complete forecast” concept of including quantitative uncertainty information as part of a forecast instead of the current single-valued deterministic “incomplete forecast” is generally an acceptable one to forecasters. Various formats are also voted to better convey forecast uncertainty information to the public. To let the public and users be aware of the fact that a forecast always has a certain degree of uncertainty and it is necessary and valuable to take such uncertainty into account in their decision-making process, forecasters agreed that a combination of popular scientific education, training and collaboration is a way forward to achieve the goal. Finally, a survey recently done by Stony Brook University about the challenges faced by U.S. forecasters in utilizing ensemble information in their daily forecasting activity is introduced. The seven main obstacles identified by the survey are summarized, which could be a useful reference for Chinese forecasters, scientists and managers in making their long-term plans. The seven challenges are “unavailability of full ensemble data, lack of ensemble data mining/interpretation/display tool, lack of relevant training and education, lack of ensemble product verification, forecasters’ overload of their daily routine work, coarse spatial resolution of ensemble model, and limitation/restriction of forecast-issuing format/context”.