Optimising UK Landscapes for Agroecosystem Resilience

英国退出欧盟共同农业政策(CAP)为增强农业景观的弹性提供了机会。农业生态系统复原力与提供非市场生态系统服务，如土壤健康、病虫害防治和授粉，以及支撑这些服务的功能生物多样性相互关联。然而，农业实践和景观复杂性对具有重要功能的生物群落的交互影响还不是很清楚。该项目将为农业生态系统多功能的几个生物指标(蚯蚓、弹尾目、飞蝇)开发一个机械景观尺度模型。这些生物指标代表了不同的生活史、行为和扩散能力，因此对环境条件、农业管理做法和景观特征表现出不同的敏感性。该模型允许我们检验核心假设，即虽然不同的生物指标对特定景观(土地利用类型的配置)和管理做法表现出不同的敏感性，但农业景观可以被设计为同时使每个生物指标受益。我们希望这种对种群、农业实践和景观之间相互作用的新理解将为优化管理计划提供依据，例如即将推出的环境土地管理(ELM)计划将取代CAP生态重点区域(EFA)。该项目的具体目标包括：1)确定每个生物指标最具代表性的物种，考虑它们在农业生态系统中的功能重要性，每个物种的现有模型，以及模型开发和验证的数据可用性。每个物种的个体生活史和行为特征将从文献中综合出来，其中大部分可以在现有的模型中找到。2)发展标准化的机械子模型，告知人口模型。这些方法包括既定的能源收支建模方法(许多物种的参数可从个人一级的数据中很容易获得)和行为子模型，它们将使用状态-空间建模方法来确定运动模式在相关空间尺度上的关键驱动因素。个人层面的运动数据在文献中是有限的，因此我们在这里设想有必要从Sonning Farm.3)正在进行的实验中收集新的飞蝇数据。3)设计基于主体的建模框架，该框架代表空间上明确的景观、不同的农业实践暴露、个体与环境的相互作用以及物种种群动态产生的个体特征。首先，模型环境将复制验证数据集的条件(例如天气、管理)的时间序列以及CEH(LCMS，开放源码)和Cranfield(Landis)持有的土地覆盖和土壤地图。4)使用雷丁、克兰菲尔德和先正达持有的大量数据集，以及英国环境变化网络和国家生物多样性网络地图集等已公布的数据集，验证模型。近似贝叶斯计算是开发复杂生态模型的有力工具，将用于简化模型分析。有人口数据的不同农业生态系统的模型验证将能够量化景观构成和农业管理对生物指标种群的交互影响(例如，对抗或协同)。5)利用Cranfield高性能计算机模拟新的景观组成和管理情景，以量化每个情景对每个生物指标种群的影响。方案将比较英国目前的分级ELM计划和欧盟CAP下的全民教育计划。

The UK's withdrawal from the EU's Common Agricultural Policy (CAP) provides an opportunity to enhance the resilience of agricultural landscapes. Agroecosystem resilience is interlinked with the provision of non-market ecosystem services, such as soil health, pest regulation and pollination, and the functional biodiversity that underpins them. However, the interactive effects of agricultural practices and landscape complexity on functionally important biological communities are not well known. This project will develop a mechanistic landscape-scale model for several bioindicators (earthworms, collembola, hoverflies) of agroecosystem multifunctionality. These bioindicators represent diverse life histories, behaviours, and dispersal capabilities, and so exhibit varying sensitivities to environmental conditions, agricultural management practices and landscape features. The model allows us to test the central hypothesis that while different bioindicators show differing sensitivities to specific landscape (configuration of land use types) and management practices, agricultural landscapes can be designed to benefit each bioindicator simultaneously. We expect this new understanding of the interplay between populations, agricultural practices, and landscapes to inform optimisation management plans, such as the upcoming Environmental Land Management (ELM) scheme that replaces the CAP Ecological Focus Area's (EFA's).The specific objectives of the project include: 1) Identify the most representative species for each bioindicator, giving weight to their functional importance in agroecosystems, existing models for each species, and data availability for model development and validation. Individual life history and behavioural traits for each species will be synthesised from the literature, the majority of which can be found in existing models. 2) Develop the standardised mechanistic submodels that inform the population models. These include established energy budget modelling approaches (the parameters for which are available for many species, and for others can be easily obtained from individual-level data) and behavioural submodels which will use a state-space modelling approach to identify key drivers of movement patterns at their relevant spatial scales. Individual-level movement data is limited in the literature, and so here we envisage a need to collect new data for hoverflies from ongoing experiments at Sonning Farm.3) Design the agent-based modelling framework, which represents spatially explicit landscapes, heterogeneous exposure to agricultural practices, individual-environment interactions, and the individual traits from which species population dynamics emerge. In the first instance, the model environment will replicate the time-series of conditions (e.g. weather, management) of the validation datasets together with land cover and soil maps held by CEH (LCMs, open source) and Cranfield (LandIS). 4) Validate the model, using extensive datasets held by Reading, Cranfield, and Syngenta together with published datasets such as the UK Environmental Change Network and the National Biodiversity Network atlas. Approximate Bayesian Computation, a powerful tool for developing complex ecological models, will be used to streamline model analysis. Model validation across diverse agroecosystems, for which population data is available, will enable quantification of the interactive effects (e.g. antagonistically or synergistically) of landscape composition and agricultural management on the bioindicator populations. 5) Simulate novel landscape composition and management scenarios, using the Cranfield high-performance computer, to quantify the effect of each scenario on each bioindicator population. Scenarios will compare the current tiered ELM scheme in the UK and EFA's under the EU's CAP.