The use of Twitter to disseminate weather information presents need for the analysis of what types of messages, and specifically warning messages, incur exposure and attention. Having this knowledge could increase exposure and attention to messages and perhaps increase retransmission through Twitter. Two models describe the cognitive processing of tweets and warnings. The extended parallel process model describes components of an effective warning message. Even in a tweet, ignoring one or both critical components of a warning—threat and efficacy—could inhibit a user from taking the correct protective action. The protective action decision model (PADM) describes risk perception and factors that enable or disable one from giving attention to a message. The PADM also helps to define impressions, retweets or likes as metrics of exposure or attention to a tweet. Tweets from three Twitter accounts within one television market during two high-impact weather events were examined. From an individual account, impressions, retweets and likes were collected to identify commonalities to tweets with much exposure and attention. Results indicate photographs and geographically specific messages were popular. Second, from two competing television weather accounts, warning tweet formats were compared to identify exposure and attention to each. Warning tweets providing threat and efficacy performed best. The purpose of this work is twofold. First is to identify local trends to compliment findings from studies with large sample sizes. Second is to apply existing theory on warning message content to Twitter. This approach should benefit communication strategies of key information nodes—local meteorologists—during high-impact weather events.

使用Twitter传播天气信息意味着需要分析哪些类型的消息（尤其是警告消息）引起曝光和关注。掌握这些知识可以增加对消息的曝光和关注，并可能增加通过Twitter进行的重新传输。两种模型描述了推文和警告的认知处理。扩展的并行过程模型描述了有效警告消息的组成部分。即使在推文中，忽略警告的一个或两个重要组成部分（威胁和功效）也可能会阻止用户采取正确的保护措施。保护性行动决策模型（PADM）描述了风险感知和使人们无法关注消息的风险感知和因素。PADM还有助于定义印象，转推或喜欢，作为曝光度或关注度的指标。在两次高影响天气事件中，检查了来自一个电视市场中三个Twitter帐户的推文。从个人账户中收集印象，转推和赞，以识别具有大量曝光和关注度的推文的共性。结果表明，照片和特定地理位置的信息很受欢迎。其次，从两个相互竞争的电视气象账户中，比较了警告鸣叫格式，以识别每种消息的曝光度和关注度。提供威胁和功效的警告推文效果最好。这项工作的目的是双重的。首先是确定本地趋势，以补充大样本研究的结果。第二是将有关警告消息内容的现有理论应用于Twitter。