Statistical identification of nitrous oxide hot moments and their significance across global ecosystems

Nitrous oxide (N2O) emissions from agricultural soils contribute 4% of total anthropogenic greenhouse gas (GHG) emissions globally. Events known as ‘hot moments’ can occur following environmental changes that favor N2O production, which contribute disproportionately to annual cumulative emissions. Despite their significance, hot moments and their impact have not been statistically well defined, particularly on a global scale. We collected 13,787 soil N2O flux measurements from 42 publications and evaluated 14 methods of statistical anomaly detection for their ability to identify hot moments within datasets. Two methods achieved highest overall performance by Matthews correlation coefficient (MCC): median absolute deviation (MCC: 0.80) and minimum covariance determinant (MCC: 0.80), the latter which also performed evenly across highly dissimilar datasets and identified more contextually important minor hot moments (39%) that other methodologies may misidentify. Interquartile range, which..., , , # Hot Moment Identification

This work uses several methods of statistical outlier detection for the detection of hot moments of nitrous oxide emissions using a dataset of daily average emissions collected from publications across the globe. Three files are included: first is a CSV file containing all data collected from publications (HotMomentTreatments.csv). Second, “Supplemental_Material.pdf” contains further description of statistical concepts and the final optimized model parameters used. The third file “Hot_Moment_Identification_Code-_Actuals.ipynb” is a Jupyter notebook containing all code used to perform data analysis and figures.

## Sharing/Access information

The source of each data point is cited within HotMomentTreatments.csv.

## Code/Software

All code for data analysis is contained in the file “Hot_Moment_Identification_Code-_Actuals.ipynb”, which is a Jupyter notebook file.

Analysis was performed using Python 3.8, Pyod 1.0.9, Fitter 1.5.2, Pandas 1.4.3, Numpy 1.21.2, S...,

全球人为温室气体（Greenhouse Gas, GHG）排放中，农田土壤一氧化二氮（Nitrous oxide, N₂O）排放占比达4%。当出现利于一氧化二氮产生的环境变化时，会出现被称为“热时刻”的事件，其对年度累积排放量的贡献远超平均水平。尽管这类热时刻具有重要研究意义，但目前尚未对其及影响进行完善的统计学定义，在全球尺度上尤为如此。

本研究从42篇学术文献中收集了13787组土壤一氧化二氮通量观测数据，评估了14种统计异常检测方法在数据集内识别热时刻的性能。基于马修斯相关系数（Matthews Correlation Coefficient, MCC），两种方法取得了最高综合性能：中位数绝对偏差（MCC=0.80）与最小协方差行列式（MCC=0.80）。其中最小协方差行列式在差异极大的数据集上同样表现稳定，且能识别出39%的、被其他方法误判的具有重要生态学意义的小型热时刻。四分位数间距（Interquartile Range, IQR）……

# 热时刻识别

本研究采用多种统计离群点检测方法，基于全球范围内公开文献收录的日均一氧化二氮排放数据集，识别农田一氧化二氮排放的热时刻。本数据集包含3个文件：其一为收录所有文献来源数据的CSV文件HotMomentTreatments.csv；其二为Supplemental_Material.pdf，其中详细阐述了相关统计学概念与最终优化的模型参数；其三为Hot_Moment_Identification_Code-_Actuals.ipynb，这是一份包含所有数据分析与绘图代码的Jupyter Notebook文件。

## 共享与获取说明
每个数据点的来源均在HotMomentTreatments.csv中予以标注。

## 代码与软件
所有数据分析代码均收录于Hot_Moment_Identification_Code-_Actuals.ipynb这一Jupyter Notebook文件中。本分析基于Python 3.8、Pyod 1.0.9、Fitter 1.5.2、Pandas 1.4.3、NumPy 1.21.2等工具完成。