Ensemble projections of hydro-biogeochemical fluxes under climate change

水文地球化学模拟中的不确定性估计是一个正在进行的研究领域，主要集中在随机模型不确定性的调查。结构模型不确定性的评估仍然是不寻常的，但有各种技术可用于量化结构的不确定性。Enhancement建模是一种通常用于气候学和气象学的技术，这些学科长期以来一直建立预测模型的结构不确定性。然而，它在水文模拟中的应用却很少。在这里，我们建议评估结构的不确定性，通过 *P集成建模，使用一组四个模型来预测水文和氮通量：SWAT，LASCAM，HBV-N和CMF-N。选择的模型，以代表在流域尺度建模的复杂性范围，从概念模型，以物理为基础的方法，并从集总到完全分布式的描述。GLUE的概念被应用到量化参数的不确定性。这种方法导致制定单一模式合奏。然后将这些单模型集合组合以产生不同的概率性和确定性多模型集合。这些多模型集成用于量化结构误差对整体预测不确定性的贡献。开发有条件的多模式组合是工作计划的一个重要组成部分。在这种情况下，多模型集成成员的选择是基于不同的模型结构和参数化的能力，以捕捉某些条件的调查集水区，如高低流量，冻融循环，或再湿润后，延长干旱。该集合模型适用于德国、瑞典和澳大利亚集水区，涵盖了广泛的不同气候边界条件、土地利用和人为干扰水平。

Uncertainty estimation in hydro-biogeochemical modeling is an ongoing area of research that focuses primarily on the investigation of stochastic model uncertainty. The evaluation of structural model uncertainty remains unusual, however there are various techniques available to quantify structural uncertainty. Ensemble modeling is one such technique that is commonly used in climatology and meteorology; disciplines where the structural uncertainty of predictive models has long been established. Its application in hydrological modeling is, however, much less common. Here we propose to evaluate structural uncertainty through *P ensemble modeling, using a set of four models to predict hydrological and nitrogen fluxes: SWAT, LASCAM, HBV-N and CMF-N. The models were selected to represent the range of complexity found in catchment scale modeling, from conceptual models to physically-based approaches, and from lumped to fully distributed descriptions. The GLUE concept is applied to quantify parameter uncertainty. This approach leads to the formulation of single-model ensembles. These single-model ensembles are then combined to produce different sets of probabilistic and deterministic multi-model ensembles. These multi-model ensembles are used to quantify the contribution of structural errors to overall predictive uncertainty. The development of conditional multi-model ensembles represents a large component of the work plan. In this case, the selection of the multi-model ensemble members is based on the capability of different model structures and parameterizations to capture certain conditions of the investigated catchments such as high-low flow, freeze-thaw cycles, or rewetting after extended droughts. The ensemble model is applied to German, Swedish and Australian catchments, and covers a broad range of different climatic boundary conditions, land uses and levels of anthropogenic disturbances.