climpred是什么?
有许多与计算初始化地球科学预测的指标有关的软件包。但是,我们没有找到任何一个包可以统一我们的所有需求。
地球系统预测回报(也叫重新预报)试验的输出是很难处理的。一个典型的输出文件可能包含维度初始化、超前时间、集合成员、经/纬度、深度,climpred利用xarray的标注维度为你处理令人头疼的记账问题。我们提供HindcastEnsemble
和PerfectModelEnsemble
的对象,这些对象携带产品与您的数十年预测输出结果一起进行验证(例如,控制试验、重建、未初始化的集合成员)。
当计算与超前相关的技巧评分时,climpred 会为您处理所有的滞后相关,正确地对齐后方预测和验证数据集之间的多个时间维度。我们提供了一套可应用于时间序列和网格的向量确定性和概率性指标。这就像将您的数十年预测输出结果添加到一个对象并运行verify()
命令一样简单。HindcastEnsemble.verify(metric='rmse', comparison='e2o', dim='init', alignment='maximum')
.
安装
pip install climpred
或者
conda install -c conda-forge climpred
实例
Dask
Pre-Processing
- Setting up your own output
- intake-esm for cmorized output
Subseasonal
- Calculate the skill of a MJO Index as a function of lead time -Calculate the skill of a MJO Index as a function of lead time for Weekly Data
Monthly and Seasonal
- Calculate ENSO Skill as a Function of Initial Month vs. Lead Time
- Calculate Seasonal ENSO Skill
Decadal
- Demo of Perfect Model Predictability Functions
- Hindcast Predictions of Equatorial Pacific SSTs
- Diagnosing Potential Predictability
- Significance Testing
相关的工具包
在climpred,我们是开源软件
的忠实粉丝,并希望支持和配合任何其他预报相关的软件包。以下是在地球系统预报领域有一定地位的软件包列表。如果您知道这个领域有任何未在列表中的开源软件包,请与我们联系。
- Microsoft forecasting: A collection of best practices for time series forecasting.
- MurCSS: A tool for standardized evaluation of decadal hindcast systems. See also this and this page for additional resources on scoring from MurCSS.
- properscoring: Probabilistic forecast metrics in python. (We’ve since wrapped these functions in xskillscore)
- pySTEPS: Framework for short-term ensemble forecasting of precipitation.
- S2D Verification: An R package for a common set of tools for forecast verification.
- SwirlsPy: Analysis and prediction of nowcasts for precipitation and weather phenomena.
- xskillscore: Metrics for verifying forecasts (a key dependency to climpred).
项目地址:
https://github.com/pangeo-data/climpred/tree/stable