Understanding the past to predict the future of distribution change in British Lepidoptera

由于缺乏长期数据，我们对生物多样性变化的了解受到严重限制。最好的数据集跨度只有几十年，提供的气候变化事件和极端天气事件的复制太少，无法精确估计其影响。即使是在英国，这是非常好的记录，我们的知识也仅限于1970年以来的时期。这一时期太短，无法揭示是否存在一般模式，或者最近的趋势是否具有特殊性。博物馆藏品包含了大量的信息，可以填补这一空白，但这些数据直到现在才被使用，原因有二。首先，自然历史收集数据无法鲁棒地建模，因为我们通常不太了解它们是如何收集的。如果每个人都以标准化的方式收集他们的标本，统计建模是最简单的，但我们知道博物馆的收藏品是随意组装的。幸运的是，动态占用检测模型使我们能够稳健地分析这些数据。第二，博物馆标本的数据库直到最近还很不完整。然而，自然历史博物馆（NHM）刚刚完成了50万份英国蝴蝶和尺蠖蛾标本的数字化。在这个项目中，学生将使用尖端的定量方法[1 - 3]将来自50万份英国蝴蝶和尺蠖蛾博物馆标本的数据与近几十年来数百万份观察记录整合在一起。覆盖时间较长的数据集包括气候变暖和变冷的多个事件（例如19世纪70年代、19世纪90年代和1940年代的变暖与变冷交替），以及干旱等更极端的天气事件。这些综合模型使重建世纪物种分布的动态成为可能，并揭示环境变化在驱动这些变化中的作用。了解气候事件和分布变化之间的关系，使人们有可能获得长期气候变化对本地生物多样性的影响的机制预测。通过在一个新的分析框架中分析博物馆标本数据和观测记录，彻底改变长期范围动态和分布变化的研究。2.根据为杂乱的机会主义数据量身定制的动态物种分布模型，制作英国蝴蝶和蛾类物种的历史分布图，以重建15年来英国蝴蝶和尺蠖蛾类物种分布的动态。将重建的物种分布轨迹与功能性状数据和环境层相结合，以回答有关生物多样性变化的关键问题，这些问题无法使用目前存在的短时间序列来回答，例如：-土地使用或物种生物学是否限制了物种跟踪气候变化的能力？极端事件的影响是可预测的吗？我们能发现早期的剧烈范围变化的警告吗？4.根据这些关系，得出物种分布未来变化的预测。答案将使我们对空间和时间的范围动态有更深入和更精确的理解，而不是任何其他群体，并展示使用短期视角（通常是我们所能做的最好的）的研究人员是否被误导了。该项目将开发新的定量方法&通过将发生记录和博物馆标本视为同一生态状态的独立实现，使用统计工具。分析将涉及1）大型和复杂数据集的高通量数据分析，以及2）将外部数据集成到生态模型中（博物馆标本，重建的土地使用和气候层）。该项目将采用贝叶斯统计分析（动态多物种占用模型），并涉及使用超级计算机（NERC JASMIN设施）为大型数据集构建高效的工作流程：Python，R，Shell脚本。

Our knowledge about biodiversity change is severely limited by a dearth of long-term data. The best datasets span only a few decades, providing too little replication of climate change episodes and extreme weather events to estimate their effects precisely. Even in Great Britain, which is unusually well-documented, our knowledge is limited to the period since 1970. This period is too short to reveal whether there are general patterns or whether recent trends are idiosyncratic. Museum collections contain a vast amount of information that can fill this gap, but such data could not be used until now for two reasons. First, natural history collection data could not be modelled robustly because we usually do not know much about how they were collected. Statistical modelling is easiest if everyone collected their specimens in a standardized way, but we know that Museum collections were assembled haphazardly. Fortunately, dynamic occupancy-detection models make it possible to analyse data like these robustly. Second, the databasing of museum specimens has until recently been very incomplete. However, the Natural History Museum (NHM) has just finished digitizing all 500,000 of its UK specimens of butterfly and geometrid moths.In this project, the student will use cutting-edge quantitative methods [1-3] to integrate data from half a million museum specimens of British butterflies and geometrid moths with many millions of observational records from recent decades. Datasets that cover a longer period of time include multiple episodes of climatic warming and cooling (e.g. warming in 1870s, 1890s & 1940s was interspersed with cool periods), as well as more extreme weather events such as droughts.These integrated models make it possible to reconstruct the dynamics of species distributions over a century and to reveal the role of environmental change in driving these changes. Understanding the relationship between climatic events and distribution change makes it possible to derive mechanistic predictions for the impact of long-term climate change on native biodiversity.There are four broad aims:1. Revolutionise the study of long-term range dynamics and distribution change by analysing museum specimen data and observational records in a single new analytical framework.2. Produce historical distribution maps for British butterfly and moth species based on dynamic species distribution models tailored for messy opportunistic data to reconstruct the dynamics of species distributions for British butterflies and geometrid moths over 15 decades.3. Integrate reconstructed trajectories of species distributions with functional trait data and environmental layers to answer key questions about biodiversity change that are unanswerable using the short time series that currently exist, such as:- Does land-use or species' biology limit species' ability to track climate change?- Is the impact of extreme events predictable?- Can we detect early warnings of dramatic range change?4. Based on these relationships, derive projections for future changes in species distributions.The answers will give us a deeper and more precise understanding of range dynamics in space and time than is possible in any other group, and show whether researchers using a shorter-term perspective - usually the best we can do - are being misled.This project will develop new quantitative methods & statistical tools by treating occurrence records and museum specimens as independent realisations of the same ecological state. Analyses will involve 1) high throughput data analysis of large and complex datasets, and 2) integration of external data into ecological models (museum specimens, reconstructed land-use and climate layers).The project will employ Bayesian statistical analysis (dynamic multispecies occupancy models) and involve building efficient workflows for large datasets using supercomputers (NERC JASMIN facility): Python, R, Shell scripting.