SUMMA Simulations using CAMELS Datasets for HPC use with CyberGIS-Jupyter for Water

This resource, configured for execution in connected JupyterHub compute platforms using the CyberGIS-Jupyter for Water (CJW) environment's supported High-Performance Computing (HPC) resources (Expanse or Virtual ROGER) through CyberGIS-Compute Service, helps the modelers to reproduce and build on the results from the VB study (Van Beusekom et al., 2022) as explained by Maghami et el. (2023).

For this purpose, four different Jupyter notebooks are developed and included in this resource which explore the paper goal for four example CAMELS site and a pre-selected period of 60-month simulation to demonstrate the capabilities of the notebooks. The first notebook processes the raw input data from CAMELS dataset to be used as input for SUMMA model. The second notebook utilizes the CJW environment's supported HPC resource (Expanse or Virtual ROGER) through CyberGIS-Compute Service to executes SUMMA model. This notebook uses the input data from first notebook using original and altered forcing, as per further described in the notebook. The third notebook utilizes the outputs from notebook 2 and visualizes the sensitivity of SUMMA model outputs using Kling-Gupta Efficiency (KGE). The fourth notebook, only developed for the HPC environment (and only currently working with Expanse HPC), enables transferring large data from HPC to the scientific cloud service (i.e., CJW) using Globus service integrated by CyberGIS-Compute in a reliable, high-performance and fast way. More information about each Jupyter notebook and a step-by-step instructions on how to run the notebooks can be found in the Readme.md fie included in this resource. Using these four notebooks, modelers can apply the methodology mentioned above to any (one to all) of the 671 CAMELS basins and simulation periods of their choice. As this resource uses HPC, it enables a high-speed running of simulations which makes it suitable for larger simulations (even as large as the entire 671 CAMELS sites and the whole 60-month simulation period used in the paper) practical and much faster than when no HPC is used.

本资源旨在JupyterHub计算平台上通过CyberGIS-Jupyter for Water（CJW）环境支持的（Expanse或虚拟ROGER）高性能计算（HPC）资源，利用CyberGIS-Compute服务执行，为模型构建者复现并基于VB研究（Van Beusekom等，2022年）的结果提供支持，正如Maghami等（2023年）所阐释。为此，本资源包含四个不同的Jupyter笔记本，旨在探讨四个示例CAMELS站点和预选的60个月模拟周期的论文目标。第一个笔记本处理CAMELS数据集的原始输入数据，作为SUMMA模型的输入。第二个笔记本利用CJW环境支持的HPC资源（Expanse或虚拟ROGER）通过CyberGIS-Compute服务执行SUMMA模型。该笔记本使用第一个笔记本的输入数据，采用原始和修改后的强迫项，具体细节在笔记本中进一步描述。第三个笔记本利用第二个笔记本的输出，使用Kling-Gupta效率（KGE）可视化SUMMA模型输出的敏感性。第四个笔记本专门为HPC环境开发（目前仅适用于Expanse HPC），通过CyberGIS-Compute集成的Globus服务，以可靠、高性能和快速的方式将大量数据从HPC传输到科学云服务（即CJW）。关于每个Jupyter笔记本的更多信息以及如何运行笔记本的逐步说明，可在包含在本资源中的Readme.md文件中找到。使用这四个笔记本，模型构建者可以将上述方法应用于任何（一个或全部）671个CAMELS流域和所选择的模拟周期。由于本资源使用HPC，它能够实现模拟的高速运行，使得大规模模拟（甚至包括整个671个CAMELS站点和论文中使用的整个60个月模拟周期）变得切实可行，并且比不使用HPC时快得多。