Modelling uncertainty for decision making on ammonia mitigation with trees in the landscape (MUDMAT).

在农业领域，有两种相互竞争的需求，一种是最经济地利用土地进行粮食生产，另一种是利用土地减轻农业产生的氮污染。在英国，农业活动排放的氨占80%以上，其中包括牲畜饲养和粪便管理通过储存和扩散释放的氨。氮以氨的形式沉积可引起半自然生态系统富营养化和酸化效应，导致物种组成变化和生物多样性降低。《2019年清洁空气战略》重点关注氨排放和随后的大气氮负荷对生态系统的影响，以及影响人类健康结果的氨颗粒形式。超过60%的英国半自然栖息地的氮沉积超过了其环境极限。政府提出了缓解措施，以支持农民减少氨的排放。一种有效的减排策略是在畜舍和泥浆仓库的下风处种植树木防护带，以“清除”氨。（10-25% Bealey et al. 2014）。由于树木对湍流的影响，它们是特别有效的空气污染物清除者。由于其粗糙的表面和高表面积，树木是特别有效的空气污染物的清除者。最近已经创建了一个决策支持工具，以帮助土地管理者根据简单的用户输入选择快速预测植树减轻氨污染的潜力。这一基于网络的决策工具可以帮助解决利用农业用地进行粮食生产和减轻污染之间的利益冲突，方法是允许土地管理者评估以最小的土地利用最大化减少氨排放的植树战略。因此，该网络工具已被证明非常受主要食品生产商（如蛋业）和农业决策者（特别是污染监管机构、保护机构和规划人员）的欢迎。然而，没有量化的可靠性预测从网络工具。这是一个关键的遗漏，阻碍了它在决策中的使用。由于模型永远不能完美地代表景观，模型预测只有在量化这些预测的确定程度时才有效。更糟的是，在网络工具中，计算成本高昂的湍流-沉积耦合模型MODDAAS-THETIS被一个非常简单的经验模型所取代，目前还没有办法量化与这种简化相关的预测精度降低。在这里，我们建议使用统计方法来创建一个更快但可量化追踪的MODDAAS-THETIS模拟器，它将取代网络工具中的经验模型，并且还将允许MODDAAS-THETIS预测的不确定性通过模拟器进行量化。我们的主要目标是：1)提供一个决策工具，帮助土地管理者最经济有效地利用农业用地，同时最大限度地减少氨污染。2)通过量化预测的不确定性，将最先进的统计方法引入基于网络的决策分析系统，逐步改进景观氨污染缓解决策。3)通过仿真显著改进基于web的工具背后的模型，使其可统计地追溯到基于底层过程的模型（MODDAAS-THETIS），并增加决策者可用于预测工具中氨缓解的树带种植选项。4)建立一种模拟景观中污染输送模型的多变量时空输出的方法，使这些计算成本高昂的模型预测中的不确定性得以量化。

In the agricultural landscape there are competing needs of making the best economical use of the land for food production and the use of land to mitigate against Nitrogen pollution from agriculture. Agricultural practises accounts for over 80% of ammonia emissions within the UK with releases from livestock housing and manure management through storage and spreading. Deposition of nitrogen in the form of ammonia can cause eutrophication and acidification effects on semi-natural ecosystems, leading to species composition changes and reduced biodiversity. The Clean Air Strategy 2019 has given significant focus to the impact of ammonia emissions and the subsequent atmospheric nitrogen load on ecosystems together with the particulate form of ammonia affecting human health outcomes. Over 60% of the UK's semi-natural habitats exceed their environmental limit for nitrogen deposition. Mitigation measures have been proposed by Government to support farmers in providing reductions in ammonia emissions. One effective abatement strategy is to plant tree shelter-belts downwind of livestock housing and slurry stores to 'scavenge' ammonia. (10-25% Bealey et al. 2014). Trees are particularly effective scavengers of air pollutants due to their effect on turbulence. Because of their rougher surface and high surface area trees are particularly effective scavengers of air pollutants. Recently a decision support tool has been created to help land managers quickly predict the potential of tree planting to mitigate in ammonia pollution based on simple user input choices. This web-based decision tool can help resolve the competing interests of using agricultural land for food production and pollution mitigation, by allowing the land manager to assess tree planting strategies that maximise ammonia abatement for the minimum use of land. Therefore, the web tool has proved to be very popular with key food producers (e.g. egg industry), and agricultural decision makers notably pollution regulators, conservation bodies and also planners.However there is no quantification of the reliability of the predictions made from the web tool. This is a key omission and hampers its use in decision making. Since models are never perfect representations of the landscape, model predictions are only as good as the quantification of how certain those predictions are. Compounding this, in the web tool, the computationally expensive coupled turbulence-deposition model MODDAAS-THETIS is replaced with a very simple empirical model and there is at present no way of quantifying the reduced accuracy of predictions associated with this simplification.Here we propose to use statistical methods to create a faster but quantifiably traceable emulator of MODDAAS-THETIS which will replace the empirical model in the web tool and will also allow uncertainty of the predictions from MODDAAS-THETIS to be quantified through the emulator.Our key objectives are to:1) Provide a decision tool that will help land managers to make the most economically efficient use of the agricultural land whilst minimising ammonia pollution. 2) Make a step-change improvement in decision making concerning the mitigation of ammonia pollution in the landscape by bringing state-of-the-art statistical methods to a web-based decision analysis system by quantifying the uncertainty of the predictions made. 3) Significantly improve the model behind the web-based tool making it traceable statistically to the underlying process-based model (MODDAAS-THETIS) through emulation and increasing the tree belt planting options available to the decision makers for predicting the mitigation of ammonia in the tool.4) Establish a methodology for emulating the multivariate spatial and temporal output from pollution transport models in the landscape making it possible to quantify uncertainty in the predictions from these computationally-expensive models.