基于半监督小样本学习的北京市臭氧空间分布模式分类研究
孙津(1993— ),女,江西高安人,博士,讲师,主要从事大气污染与地理信息科学方向的研究。E-mail: jinsun@hhu.edu.cn |
Copy editor: 蒋树芳
收稿日期: 2023-09-25
修回日期: 2023-11-06
网络出版日期: 2024-03-31
基金资助
中央高校基本科研业务费专项资金项目(423062)
Classification for Spatial Patterns of Urban Ozone Pollution in Beijing Based on Semi-Supervised Few-Shot Learning
Received date: 2023-09-25
Revised date: 2023-11-06
Online published: 2024-03-31
Supported by
Fundamental Research Funds for the Central Universities(423062)
研究城市臭氧空间分布模式有助于分析污染成因,也能够为污染防治提供科学依据。但过去多基于站点数据分析浓度平均分布的概括性特征,对全域浓度分布平面的描述有限,也很少进行分类研究,无法更全面地看待污染分布的多种模式及其时间变化。本研究利用由卫星数据建模估计的臭氧日最大8 h滑动均值分布数据,针对臭氧空间分布模式标签化的难度提出一种面向小样本的半监督学习方法,以北京市为例进行分类实验。实验发现: ① 2020年数据经预处理后以40个训练样本对249个测试样本进行分类,总体分类精度达81.12%, kappa系数达0.741 6,说明在小样本条件下半监督方法取得了较好的分类效果;② 分类得到的8种模式中,“(东)南高(西)北低或东高西低”的模式1、“(西)北高(东)南低”的模式2以及“中心低”的模式6为主要模式,分别在暖季(3—10月)、冷季(11—次年2月)和冷暖季过渡期占据主导,这一时间规律反映出区域传输和光化学反应的季节性影响; ③ 将2020年的训练样本迁移至2019年进行分类,在取得较高精度的同时也对上述规律进行了验证。以上结果表明,本研究提出的空间分布模式分类方法能够为全面确定高污染的防治区域以及分类研究不同污染事件的成因提供支持。
孙津 . 基于半监督小样本学习的北京市臭氧空间分布模式分类研究[J]. 地球信息科学学报, 2024 , 26(3) : 725 -735 . DOI: 10.12082/dqxxkx.2024.230581
Ozone concentrations tend to be heterogeneous across a city's space due to the mixed land use and diverse landscapes. Studying spatial patterns of urban ozone pollution contributes to the knowledge of the mechanism of pollution formation and also provides scientific reference for pollution prevention and control. Nevertheless, most previous research focused on the averaged value of ozone concentration from monitoring sites, which cannot describe the spatial characteristics of the entire region's concentration surface. Additionally, the classification method was seldom used to analyze pollutants' spatial patterns, and thus very few studies paid attention to the varied types of patterns and their temporal variations. In this study, based on the distributions of ozone’s daily maximum 8-h moving average estimated from satellite data, an approach of semi-supervised few-shot learning was proposed to classify ozone's spatial patterns in Beijing. The self-training method considered the difficulty of data labeling and can utilize information from a large number of unlabeled samples to augment the training set iteratively. Three kinds of normalized features were involved in classification to describe the spatial variations of concentrations. Totally, there were 40 training samples and 249 test samples for the year of 2020, and the overall classification accuracy was 81.12% with a kappa coefficient of 0.741 6. This demonstrated the effectiveness of the semi-supervised classification method despite the small size of training samples. The classification results showed that, among the eight patterns of ozone distributions in Beijing, three of them were major patterns, including Pattern 1: high concentrations in the south/east/southeast and low in the north/west/northwest, Pattern 2: high concentrations in the north/northwest and low in the south/southeast, and Pattern 6: low concentrations in the center. They dominated the warm season (from Mar. to Oct.), the cold season (from Nov. to Feb.), and the transition period, respectively. These temporal variations of ozone's spatial patterns indicated the influence from the seasonality of regional transport and photochemical reactions. When training samples were transferred to the year of 2019, the overall classification accuracy reached 80.97%, and the kappa coefficient was 0.745 6, suggesting the high potential of sample migration. And the results of 2019 further confirmed the previous findings. Thus, the proposed classification method for spatial patterns of urban ozone pollution can not only benefit the identification of regions with heavy pollution but also support the study on mechanisms of different pollution events.
表1 北京市臭氧分布数据经站点监测值验证结果Tab. 1 Validation of near-surface ozone concentrations in Beijing using monitoring data |
年份 | 总体 | 分站点 | |||||
---|---|---|---|---|---|---|---|
R2 | RMSE/ (μg/m3) | R2最 小值 | R2最 大值 | RMSE最小值/(μg/m3) | RMSE最大值/(μg/m3) | ||
2019 | 0.93 | 12.17 | 0.86 | 0.99 | 0.33 | 28.82 | |
2020 | 0.94 | 13.33 | 0.81 | 1.00 | 5.72 | 28.44 |
表2 北京市臭氧空间分布的8类模式Tab. 2 Eight types of ozone's spatial patterns in Beijing |
模式类型 | 模式描述 |
---|---|
1 | (东)南高(西)北低或东高西低 |
2 | (西)北高(东)南低 |
3 | 东北高西南低 |
4 | 西(南)高东(北)低 |
5 | 中心高 |
6 | 中心低 |
7 | 空间差异较小的均匀型 |
8 | 难以分入上述类型的混杂型 |
表3 半监督分类的混淆矩阵Tab. 3 Confusion matrix of semi-supervised classification |
模式类型 | 1 | 2 | 3 | 4 | 5 | 6 | 8 | 制图精度/% | 用户精度/% |
---|---|---|---|---|---|---|---|---|---|
1 | 99 | 0 | 0 | 0 | 2 | 1 | 1 | 96.12 | 89.19 |
2 | 0 | 55 | 4 | 0 | 1 | 0 | 3 | 87.30 | 90.16 |
3 | 0 | 0 | 14 | 0 | 0 | 0 | 0 | 100.00 | 58.33 |
4 | 2 | 0 | 0 | 8 | 1 | 0 | 0 | 72.73 | 72.73 |
5 | 7 | 0 | 0 | 0 | 4 | 0 | 0 | 36.36 | 44.44 |
6 | 0 | 5 | 1 | 2 | 0 | 20 | 5 | 60.61 | 90.91 |
8 | 3 | 1 | 5 | 1 | 1 | 1 | 2 | 14.29 | 18.18 |
总体分类精度/% | 81.12 | ||||||||
Kappa系数 | 0.741 6 |
表4 2020年北京市臭氧日最大8 h滑动均值分布模式分类结果统计Tab. 4 Statistics of spatial patterns of ozone's daily maximum 8-h moving averages in Beijing during 2020 |
模式类型 | 分类结果 | 实际 | |||
---|---|---|---|---|---|
天数 | 占比/% | 天数 | 占比/% | ||
1 | 121 | 33.06 | 113 | 30.87 | |
2 | 71 | 19.40 | 73 | 19.95 | |
3 | 29 | 7.92 | 19 | 5.19 | |
4 | 16 | 4.37 | 16 | 4.37 | |
5 | 15 | 4.10 | 17 | 4.64 | |
6 | 71 | 19.40 | 82 | 22.40 | |
7 | 32 | 8.74 | 32 | 8.74 | |
8 | 11 | 3.01 | 14 | 3.83 |
[1] |
|
[2] |
|
[3] |
陈浪, 赵川, 关茗洋, 等. 我国大气臭氧污染现状及人群健康影响[J]. 环境与职业医学, 2017, 34(11):1025-1030.
[
|
[4] |
李红, 彭良, 毕方, 等. 我国PM2.5与臭氧污染协同控制策略研究[J]. 环境科学研究, 2019, 32(10):1763-1778.
[
|
[5] |
|
[6] |
|
[7] |
|
[8] |
严心田. 基于WRF-Chem的南京市春夏季空气质量研究:时空特征和减排效果分析[D]. 南京: 南京信息工程大学, 2019.
[
|
[9] |
赵丽敏. 京津冀及周边地区近地表O3浓度的遥感估算研究[D]. 南京: 南京大学, 2019.
[
|
[10] |
|
[11] |
陈菁, 彭金龙, 徐彦森. 北京市2014—2020年PM2.5和O3时空分布与健康效应评估[J]. 环境科学, 2021, 42(9):4071-4082.
[
|
[12] |
苏志华, 韩会庆, 李莉, 等. 贵阳市臭氧的时空分布、气象作用及其与前体物的关系[J]. 中山大学学报(自然科学版), 2020, 59(5):102-112.
[
|
[13] |
赵洁, 丁俊傑, 刘芮伶, 等. 重庆市臭氧污染特征分析及天气分型研究[J]. 环境科学与技术, 2022, 45(11):62-69.
[
|
[14] |
|
[15] |
|
[16] |
|
[17] |
|
[18] |
徐竟泽, 吴作宏, 徐岩, 等. 融合PCA、LDA和SVM算法的人脸识别[J]. 计算机工程与应用, 2019, 55(18):34-37.
[
|
[19] |
|
[20] |
何云, 吴怀宇, 钟锐. 基于多种LBP特征集成学习的人脸识别[J]. 计算机应用研究, 2018, 35(1):292-295.
[
|
[21] |
李倩玉, 蒋建国, 齐美彬. 基于改进深层网络的人脸识别算法[J]. 电子学报, 2017, 45(3):619-625.
[
|
[22] |
|
[23] |
张浩, 陈昌红. 基于深度学习的极光序列自动分类方法[J]. 激光与光电子学进展, 2018, 55(11):346-354.
[
|
[24] |
韩冰, 贾中华, 高新波. 改进的主成分分析网络极光图像分类方法[J]. 西安电子科技大学学报, 2017, 44(1):83-88.
[
|
[25] |
韦晶, 李占清. 中国高分辨率高质量地面臭氧数据集(2013-2020)[DB/OL]. 2023.https://doi.org/10.5281/zenodo.4400042.
[
|
[26] |
|
[27] |
王占山, 李云婷, 陈添, 等. 北京市臭氧的时空分布特征[J]. 环境科学, 2014, 35(12):4446-4453.
[
|
[28] |
窦晶晶. 北京城区近地面气象要素精细化时空分布特征[D]. 北京: 中国气象科学研究院, 2014.
[
|
[29] |
曹杨, 何文英, 施红蓉, 等. 2018年北京城区和远郊区低层大气风场特征分析[J]. 气候与环境研究, 2021, 26(4):403-412.
[
|
[30] |
符传博, 周航. 中国城市臭氧的形成机理及污染影响因素研究进展[J]. 中国环境监测, 2021, 37(2):33-43.
[
|
/
〈 |
|
〉 |