Going from the virtual to the real world - using alternative data sources to tackle real-world challenges in hydrology

应用水文科学在很大程度上依赖于水文气象信息。水文模型（随机和确定性）的性能在很大程度上取决于输入数据的质量。气象站观测（通常测量温度和降水数据）为水文和其他模拟模型提供关键信息。然而，它们受到诸如时间范围短、数据缺失、误差、仪器偏差以及由于设备更换或台站移位而引入的偏差等问题的困扰。加拿大北方的一个更严重的缺点是台站的空间密度低，几乎没有具有长期观测记录的台站。所有这些问题都严重限制了在流域尺度上充分表现水文过程时空变异性的能力。也许更重要的是，这些问题仍然存在，因为北美的气象站数量正在减少，因为更多的气象站正在退役而不是增加到现有网络中。考虑到环境监测在不断变化的气候中的重要性，这些网络的状况变得更加重要。面对这一令人担忧的现实，本研究计划旨在评估使用半虚拟和虚拟数据集作为传统台站数据（降水，温度，流量）的代理。该计划将专门研究观测（半虚拟数据集）和高分辨率，有限区域动态生成的数据（来自区域气候模型的虚拟数据集）的再分析。再分析是通过同化来自不同来源的几个观测变量，从海洋-大气耦合模式产生的网格数据。动态生成的数据来自各种复杂的动态模型（气候模型），这些模型类似于用于再分析的模型，但没有数据同化过程。当区域气候模式在其边界上受到再分析的驱动时，区域气候模式将遵循驱动模式（再分析）中存在的大尺度环流，并有能力再现历史数据序列，尽管比纯粹的数据再分析有更多的自由度。 这些数据集在水文学中的使用既令人兴奋又有争议：令人兴奋是因为它们在现实世界中的应用构成了水文研究的新范式，但有争议是因为它们与观测数据集相比显示出空间和时间偏差。具体而言，该研究计划将利用几个再分析和高分辨率区域气候模型数据集，为以下两个突出的水文问题提供替代解决方案：轴1。降水和气温历史时间序列缺失数据的填补。2轴无资料流域的径流预报。在这两种情况下，几个水文模型的输出将被用作主要的性能基准时，比较传统的方法与本项目开发的。我们有很高的希望，这些半虚拟和虚拟数据集的使用将提供有效的替代解决方案，这些问题，而这个研究计划将发挥先锋作用，在建立这些数据集的现实世界的应用。

Applied hydrological science strongly relies on hydrometeorological information. The performance of hydrology models (stochastic and deterministic) is strongly dependent on the quality of input data. Weather station observations (typically measuring temperature and precipitation data) provide critical information to hydrology and other simulation models. They are however plagued with problems such as short temporal horizons, missing data, errors, instrument’s biases and biases introduced through equipment change or station displacement. A more significant drawback in Northern Canada is the low spatial density of stations and the quasi absence of stations with long observational records. All of these problems severely limit the ability to adequately represent the spatial and temporal variability of the hydrological processes at the basin scale. Perhaps more importantly, these problems are here to stay since the number of weather stations is decreasing in North America, as more stations are being decommissioned than added to the existing networks. The state of these networks is made even more critical considering the importance of environmental monitoring in a changing climate. In face of this alarming reality, this research programs aims at evaluating the use of semi-virtual and virtual datasets as proxies for traditional station data (precipitation, temperature, streamflow). This program will specifically look at reanalysis of observations (semi-virtual datasets) and high-resolution, limited-area dynamically generated data (virtual datasets from regional climate models). Reanalysis are gridded data generated from coupled ocean-atmosphere models through the assimilation of several observed variables from different sources. Dynamically generated data come from dynamical models of various complexities (climate models) which are similar to the ones used for reanalyses, however without the data assimilation process. When driven at their boundaries by reanalysis, regional climate models will follow the large-scale circulation present in the driving model (the reanalysis) and have the ability to reproduce the historical sequences of data, albeit with more degrees of freedom than a pure reanalysis of data. The use of these datasets in hydrology is both exciting and controversial: Exciting because their use in real-world applications constitutes a new paradigm in hydrological research, but controversial because they display spatial and temporal biases when compared against observation-derived datasets. Specifically, this research program will make use of several reanalysis and high-resolution regional climate model datasets to provide alternatives solutions to the following two outstanding hydrologic problems: Axis 1. Infilling of missing data in historical time series’ of precipitation and temperature. Axis 2. Streamflow forecasting in ungauged watersheds. In both cases, the outputs of several hydrology models will be used as the main performance benchmarks when comparing traditional approaches against the ones developed in this project. We have high hopes that the use of these semi-virtual and virtual datasets will offer efficient alternative solutions to these problems, and that this research program will play a pioneer role in the establishment of these datasets for real-world applications.