Monitoring waterbird populations in Australia is challenging for reasons of counting logistics, and because population aggregation and dispersion can shift rapidly in response to large spatio-temporal variations in resource availability. The East Australian Waterbird survey has conducted annual, aerial, systematic counts of waterbirds over eastern Australia for almost 40 years. It was designed to monitor waterbird populations using design-based inference though for many species this form of inference appears inadequate in the face of these challenges. Here we develop a state-space model-based Bayesian approach that, in addition to explicitly incorporating process noise and observation uncertainty, uses random effects and rainfall-derived covariates to model the year-to-year variation in the proportion of the total (super) population that is present on surveyed wetlands, and available to be counted. We use this model-based approach to estimate the superpopulation size of 45 waterbird species annually, and model the rate of population increase as a function of antecedent rainfall. The results confirm the strong positive effect of antecedent rainfall on population growth rates for nearly all species, and illustrate that species respond to rainfall differently in terms of habitat use, which influences whether they are present on surveyed wetlands. For many species, the year-to-year variation in the estimated proportion of the population on surveyed wetlands is very high. The results have implications for making inferences on population trends from these data, with the ability to model the year-to-year sampling variation a key requirement before the rate of population increase can be estimated with any precision. This study illustrates how to progress this approach, and infers that under average rainfall conditions, the general trend is for estimated superpopulation rates of increase to be negative, though for only a few species is this occurring with strong belief.

由于计数物流的原因，监测澳大利亚的水鸟种群具有挑战性，并且由于种群聚集和分散可以响应资源可用性的巨大时空变化而迅速变化。近 40 年来，东澳大利亚水鸟调查对澳大利亚东部的水鸟进行了年度、空中、系统的计数。它旨在使用基于设计的推理来监测水鸟种群，但对于许多物种而言，这种推理形式在面临这些挑战时显得不够充分。在这里，我们开发了一种基于状态空间模型的贝叶斯方法，除了明确结合过程噪声和观察不确定性之外，使用随机效应和降雨衍生的协变量来模拟存在于调查湿地的总（超级）人口比例的逐年变化，并可用于计算。我们使用这种基于模型的方法来估计每年 45 种水鸟的超种群规模，并将人口增长率建模为前期降雨量的函数。结果证实了前期降雨对几乎所有物种的种群增长率都有很强的积极影响，并说明物种在栖息地利用方面对降雨的反应不同，这会影响它们是否存在于所调查的湿地中。对于许多物种来说，被调查湿地人口的估计比例的逐年变化非常大。结果对从这些数据中推断人口趋势具有影响，能够对逐年抽样变化进行建模，这是在能够以任何精度估计人口增长率之前的关键要求。这项研究说明了如何推进这种方法，并推断在平均降雨条件下，总体趋势是估计的超种群增长率为负数，尽管只有少数物种强烈相信这种情况发生。