Cases of anthrax in livestock are infrequently and irregularly reported in the state of Victoria, Australia; however, their impact on individual livestock, farming communities and the government agencies tasked with containing these outbreaks is high. This infrequency has been anecdotally associated with differences in annual and local weather patterns. In this study, we used historical anthrax cases and meteorological data from weather stations throughout Victoria to train a generalized linear mixed effects model to predict the daily odds of a case of anthrax occurring in each shire in the coming 30 days. Meteorological variables were transformed to deviations from the mean values for temperature or cumulative values for rainfall in the shire across all years. Shire was incorporated as a random effect to account for meteorological variation between shires. The model incorporated a post hoc weighting for the frequency of historic cases within each shire and the spatial contribution of each shire to the recently redefined Australian Anthrax Belt. Our model reveals that anthrax cases were associated with drier summer conditions (OR 0.96 (95% CI 0.95–0.97) and OR 0.98 (95% CI 0.97–0.99) for every mm increase in rainfall during September and December, respectively) and cooler than average spring (OR 0.20 (95% CI 0.11–0.52) for every °C increase in minimum daily temperature during November and warmer than average summer temperatures (OR 1.45 (95% CI 1.29–1.61) for every °C increase in maximum daily temperature during January. Cases were also preceded by a 40-day period of cooler, drier temperatures (OR 0.5 (95% CI 0.27–0.74) for every °C increase in maximum daily temperature and OR 0.96 (95% CI 0.95–0.97) for every mm increase in rainfall followed by a warmer than average minimum (or nightly) temperature 10 days immediately before the case (OR 1.46 (95% CI 1.35–1.58) for every °C increase in maximum daily temperature). These coefficients of this training model were then applied daily to meteorological data for each shire, and output of these models was presented as a choropleth and timeline plot in a Shiny web application. The application builds on previous spatial modelling and provides Victorian agencies with a tool to engage at-risk farmers and guide discussions towards anthrax control. This application can contribute to the wider rejuvenation of anthrax knowledge and control in Victoria and corroborates the anecdote that increased odds of disease can be linked to meteorological events.

中文翻译：

---

使用气象数据预测澳大利亚维多利亚州牲畜炭疽的周期性风险

澳大利亚维多利亚州很少和不定期地报告牲畜炭疽病例；然而，它们对个体牲畜、农业社区和负责控制这些疫情的政府机构的影响很大。这种不频繁的情况与年度和当地天气模式的差异有关。在这项研究中，我们使用来自维多利亚州各地气象站的历史炭疽病例和气象数据来训练广义线性混合效应模型，以预测未来 30 天内每个郡发生炭疽病例的每日几率。气象变量被转换为与该郡所有年份的温度平均值或降雨累积值的偏差。Shire 被纳入作为随机效应以解释各郡之间的气象变化。该模型结合了每个郡历史案例频率的事后加权，以及每个郡对最近重新定义的澳大利亚炭疽带的空间贡献。我们的模型显示，炭疽病例与较干燥的夏季条件有关（9 月和 12 月降雨量每增加一毫米，OR 0.96（95% CI 0.95–0.97）和 OR 0.98（95% CI 0.97–0.99）和比平均春季 (OR 0.20 (95% CI 0.11–0.52) 11 月的每日最低温度每升高 1 °C 且高于夏季平均温度 (OR 1.45 (95% CI 1.29–1.61) 每日最高温度每升高 1 °C在 1 月期间。病例之前也有 40 天的凉爽、干燥的温度（OR 0. 5 (95% CI 0.27–0.74) 每日最高温度每升高 1°C 或 0.96 (95% CI 0.95–0.97) 降雨量每增加一毫米，随后最低（或夜间）温度高于平均最低（或夜间）温度 10 天立即在病例之前（每日最高温度每升高 1°C，OR 1.46 (95% CI 1.35–1.58)）。该训练模型的这些系数然后每天应用于每个郡的气象数据，这些模型的输出在 Shiny Web 应用程序中显示为等值线图和时间线图。该应用程序建立在先前的空间建模之上，并为维多利亚州的机构提供了一种工具，可以让处于危险中的农民参与进来并指导有关炭疽控制的讨论。