Large-sample comparative hydrologic modelling computational laboratory

拟议的研究计划将建立一个新的大样本水文建模计算实验室，世界各地的环境建模者将利用和受益。研究计划的目标是：1）产生开放和比较性的大样本分布式水文建模相互比较研究和模型挑战，以激发，基准，然后随着时间的推移利用知识生成2）寻求理解深度学习相对预测能力背后的原因（例如，水文建模的人工智能方法，以设计下一代混合数据驱动和水文模型3）发现新的水文模型校准方法，提高我们模拟和预测未来条件（如洪水和干旱）的能力，4）设计和进行独特的大样本比较经验测试，帮助建模人员清楚地证明模型开发选择最有效。与我一起工作的学生将接受现代水文建模软件（乌鸦水文建模框架）的培训，这些软件今天在加拿大各地由行业和政府流量预测机构以及世界各地的国际研究人员使用。学生还将接受培训，并为下一代水文建模方法做好准备，这些方法利用传统水文建模方法的大数据革命。改进的水文模型和模型开发过程通过生成更准确的洪水和干旱预测来改善公共安全。现场推进成果将包括：1）改进传统水文模型构建方法，2）下一代模型和3）新的基准模型相互比较研究。从根本上说，我的领域的进展只能通过研究社区的基准标准来跟踪（Nearing等人，2021年），这种创建易于访问的大数据存储库的基准测试工作被视为我所在领域的最佳投资之一（Nearing et al.，2021年）。我的团队将与世界各地的许多研究人员合作建立这样的基准，因此他们将实际使用它们，然后重要的是随着时间的推移传播它们的进一步使用。总的来说，5年后，我的大样本比较水文建模计算实验室将成为我所在领域的从业者和研究人员找到他们信任的世界级证据的地方，这些证据指导他们如何建立水文模型。专家们还将进行虚拟访问，以便与我的团队和来自世界各地的其他人合作，以客观的方式比较他们的水文模型，从而推动我们领域的科学发展。

The proposed research program will build a new large-sample hydrological modelling computational laboratory that environmental modellers around the world will utilize and benefit from. Research program goals are to: 1) produce open and comparative, large-sample distributed hydrologic modelling intercomparison studies and model challenges to inspire, benchmark and then leverage knowledge generation over time 2) seek to understand the reasons behind the relative predictive power of deep learning (e.g., Artificial Intelligence) methods for hydrological modelling in order to design next generation hybrid data-driven and hydrological models 3) discover new hydrologic model calibration methods that enhance our ability to simulate and predict future conditions such as floods and drought and 4) design and conduct unique, large-sample comparative empirical tests that help clearly demonstrate to modellers which model development choices work best. Students working with me will be trained on modern hydrological modelling software (The Raven Hydrological Modelling Framework) used today across Canada by industry and governmental flow forecasting agencies, as well as international researchers around the world. Students will also be trained and ready for next generation hydrological modelling approaches that leverage the big data revolution with traditional hydrologic modelling approaches. Improved hydrological models and model development processes improve public safety by generating more accurate predictions of flooding and drought. Field advancement outcomes will include: 1) improved traditional hydrological model building methods 2) next-generation models and 3) new benchmark model intercomparison studies. Fundamentally, the progress in my field can only be tracked via a research community standard for benchmarking (Nearing et al., 2021) and such a benchmarking effort to create an easily accessible big data repository is viewed as one of the best investments in my field (Nearing et al., 2021). My team will have built such benchmarks in collaboration with numerous researchers around the world who will therefore actually use them and then importantly propagate their further use over time. Overall, after 5 years, my large-sample comparative hydrological modelling computational laboratory will be a place where both practitioners and researchers in my field come to find world-class evidence they trust that guides how they build their hydrological models. Experts will also virtually visit in order to collaborate with my team and others from around the world to compare their hydrological models in an objective way that will advance the science in our field.