Data assimilation experiments inform monitoring needs for near-term ecological forecasts in a eutrophic reservoir: data, forecasts, and scores

This data publication contains zipped parquet from the Beaverdam Reservoir forecasting data assimilation experiments using the FLARE (Forecasting Lake And Reservoir Ecosystems) system: drivers.zip contains NOAA driver forecast files, targets.zip contains in-situ water temperature observations and meteorological data, forecasts.zip contains forecast parquet files generated from the BVR FLARE DA experiment workflow, and scores.zip contains forecast skill metrics required for analysis. Within the forecasts and scores folders, there are four runs that were conducted with different parameter tuning and uncertainty quantification. The "all_UC" folder includes forecasts run with process, driver, parameter, and initial condition uncertainty quantification. The "IC_off" folder includes forecasts run without initial conditions uncertainty included (i.e., only process, driver, and parameter uncertainty). The "constant_bad_pars" folder includes forecasts run with constant parameters (but daily updating of initial conditions) that were not tuned for Beaverdam Reservoir before forecasts were generated. Finally, the "tuned_bad_pars" folder includes forecasts that were run with daily updating of initial conditions and parameters, but the parameters started out at random values that were not tuned for Beaverdam Reservoir.

本数据集公开内容包含基于FLARE（Forecasting Lake And Reservoir Ecosystems，湖泊与水库生态系统预报系统）开展的比弗丹水库预报数据同化实验的压缩Parquet格式文件：其中drivers.zip包含美国国家海洋和大气管理局（NOAA）驱动预报文件，targets.zip包含原位水温观测数据与气象数据，forecasts.zip包含由比弗丹水库FLARE数据同化实验流程生成的预报Parquet文件，scores.zip包含分析所需的预报技巧评分指标。在预报与评分文件夹下，共开展了4组采用不同参数调优与不确定性量化方案的试验运行。"all_UC"文件夹包含同时考虑过程、驱动、参数与初始条件不确定性量化的预报结果；"IC_off"文件夹包含未纳入初始条件不确定性的预报结果（即仅考虑过程、驱动与参数不确定性）；"constant_bad_pars"文件夹包含使用固定参数（但每日更新初始条件）生成的预报结果，此类参数未针对比弗丹水库进行调优；"tuned_bad_pars"文件夹包含采用每日更新初始条件与参数的预报结果，但参数初始值为未针对比弗丹水库调优的随机取值。