基于Argo剖面浮标观测的海洋溶解氧空间-质量迭代插值方法
岳林峰(1998—),男,河南安阳人,硕士生,研究方向为海洋时空数据挖掘分析。E-mail: yuelinfeng20@mails.ucas.ac.cn |
收稿日期: 2022-09-23
修回日期: 2023-02-02
网络出版日期: 2023-09-22
基金资助
中国科学院A类战略先导专项(XDA19060103)
可持续发展大数据国际研究中心创新研究计划项目(CBAS2022IRP05)
An Iterative Space-quality Interpolation Method for Marine Dissolved Oxygen Data Observed by Argo Floats
Received date: 2022-09-23
Revised date: 2023-02-02
Online published: 2023-09-22
Supported by
Strategic Priority Research Program of the Chinese Academy of Sciences(XDA19060103)
Innovative Research Program of International Research Center of Big Data for Sustainable Development Goals(CBAS2022IRP05)
溶解氧是探究海洋内部各种生物地球化学循环过程以及评估海洋健康的重要参数。目前,基于Argo剖面浮标的观测方式已经成为获取海洋溶解氧信息的主要来源之一,但是这种方式获取到的海洋溶解氧资料在空间上是离散分布的。如何利用空间插值方法将离散的Argo溶解氧剖面观测数据构建成连续的空间网格数据是当前研究关注的热点,本文提出了一种面向Argo溶解氧剖面观测数据的空间-质量特征的迭代插值方法(ASOSQIM)。针对观测样点空间上分布离散、不规则的特点,ASOSQIM首先对溶解氧观测数据进行空间初始化来构建初始背景场,然后对Argo溶解氧剖面观测样点进行空间异质性分区,采用空间-质量定权的方式赋予不同质量标示数据相应的权重系数来解决参与插值的数据量及其精度问题,最后利用迭代插值的思想并集成溶解氧空间分区特征和质量标识,实现了基于Argo剖面浮标观测的溶解氧浓度空间插值。利用交叉验证法和WOA18资料对ASOSQIM进行方法验证和对比分析,结果表明:① ASOSQIM的插值精度相比于其他空间插值方法具有大幅度的提升,与2022年同期Argo浮标溶解氧观测结果进行对比,验证了该方法的优越性和可行性; ② 利用ASOSQIM得到的溶解氧浓度插值结果与WOA18溶解氧资料具有高度一致性,溶解氧浓度绝对误差在20 μmol/kg范围内的区域占比达到85%以上,表明利用该方法得到的插值结果具有可靠性。
岳林峰 , 薛存金 , 王振国 , 牛超然 . 基于Argo剖面浮标观测的海洋溶解氧空间-质量迭代插值方法[J]. 地球信息科学学报, 2023 , 25(10) : 2000 -2011 . DOI: 10.12082/dqxxkx.2023.220720
Dissolved oxygen is an important parameter for exploring various biogeochemical cycles in the ocean interior and assessing ocean health. At present, the observation method based on Argo floats has become one of the main sources to obtain the information of Marine dissolved oxygen, but the data obtained by this method are distributed discretely in space. How to use spatial interpolation method to construct discrete dissolved oxygen profiles of Argo floats into continuous spatial grid data is the focus of current research. Therefore, an iterative interpolation method (ASOSQIM) for spatial-quality characteristics of dissolved oxygen observation data using Argo floats is proposed in this paper. It is of great significance to integrate discrete Marine dissolved oxygen data to obtain spatially distributed continuous grid data for exploring and revealing the variation law of Marine dissolved oxygen concentration and providing reliable data support and information services for related fields. In view of the discrete and irregular spatial distribution of the observed sample points, ASOSQIM carries out a series of processing on the dissolved oxygen observation data using Argo floats, such as monthly statistics, quality recontrol and standardized interpolation. Based on the processed dissolved oxygen observation data, the spatial initialization was carried out, and the dissolved oxygen concentration on the regular grid after initialization was taken as the initial value of spatial interpolation calculation. Then, the dissolved oxygen observation data using Argo floats were divided into different spatial heterogeneity zones, and the corresponding weight coefficients were assigned to data of different quality flags by using the space-quality weighting method to solve the problem of the amount of data involved in interpolation and its accuracy. Finally, the idea of iterative interpolation was used to integrate the spatial partitioning characteristics and quality flags of dissolved oxygen. The spatial interpolation of dissolved oxygen concentration based on observation data of Argo floats is realized. The cross-validation method and WOA18 data were used to verify and compare ASOSQIM. The results show that: (1) the interpolation accuracy of ASOSQIM is significantly improved compared with other spatial interpolation methods. The advantages and feasibility of this method are verified by comparing with the dissolved oxygen observation results of Argo floats in 2022. (2) The dissolved oxygen concentration interpolation results obtained by ASOSQIM are highly consistent with the dissolved oxygen data of WOA18, and the absolute error of dissolved oxygen concentration accounts for more than 85% in the range of 20μmol/kg, indicating that the spatial interpolation results of dissolved oxygen concentration obtained by this method are reliable.
表1 全部观测样点的交叉验证结果Tab. 1 Cross-validation results of all observed samples |
MAE/(μmol/kg) | RMSE/(μmol/kg) | MARE/% | |
---|---|---|---|
ASOSQIM_1 | 2.90 | 5.94 | 1.06 |
ASOIM_1 | 3.26 | 6.18 | 1.06 |
IDW_1 | 4.21 | 7.06 | 1.56 |
KRG_1 | 4.12 | 6.85 | 1.53 |
GPI_1 | 8.20 | 12.64 | 3.18 |
Cressman_1 | 4.05 | 8.23 | 1.66 |
ASOSQIM_7 | 3.47 | 8.40 | 1.44 |
ASOIM_7 | 4.08 | 8.76 | 1.52 |
IDW_7 | 4.94 | 9.43 | 2.01 |
KRG_7 | 4.83 | 9.24 | 1.97 |
GPI_7 | 8.69 | 15.74 | 3.68 |
Cressman_7 | 5.20 | 10.17 | 1.88 |
表2 部分观测样点的交叉验证结果Tab. 2 Cross-validation results of some observed samples |
MAE/(μmol/kg) | RMSE/(μmol/kg) | MAPE/% | |
---|---|---|---|
ASOSQIM_1 | 4.38 | 7.67 | 1.46 |
ASOIM_1 | 5.83 | 10.64 | 2.20 |
IDW_1 | 7.27 | 11.42 | 2.72 |
KRG_1 | 7.12 | 11.06 | 2.67 |
GPI_1 | 11.87 | 15.15 | 4.07 |
Cressman_1 | 8.22 | 13.45 | 3.62 |
ASOSQIM_7 | 4.39 | 12.23 | 2.65 |
ASOIM_7 | 6.04 | 13.84 | 3.48 |
IDW_7 | 8.04 | 16.09 | 3.97 |
KRG_7 | 7.98 | 16.00 | 3.91 |
GPI_7 | 13.11 | 24.62 | 6.90 |
Cressman_7 | 9.28 | 15.06 | 4.39 |
表3 ASOSQIM插值结果与实测值比较Tab. 3 Comparison between ASOSQIM interpolation results and measured values |
编号 | 浮标ID | 周期 | 观测值/(μmol/kg) | 插值结果/(μmol/kg) | 绝对误差/(μmol/kg) | 相对误差/% |
---|---|---|---|---|---|---|
1 | 1902381 | 1 | 210.831 | 211.825 | 0.994 | 0.47 |
2 | 5906035 | 99 | 240.255 | 241.454 | 1.199 | 0.50 |
3 | 5904844 | 183 | 215.156 | 216.732 | 1.576 | 0.73 |
4 | 5905103 | 144 | 325.733 | 327.549 | 1.816 | 0.56 |
5 | 5905138 | 187 | 298.229 | 293.810 | 4.419 | 1.48 |
6 | 5905986 | 174 | 195.506 | 197.730 | 2.224 | 1.14 |
7 | 5905109 | 159 | 223.558 | 221.725 | 1.833 | 0.82 |
8 | 5906208 | 72 | 251.583 | 256.715 | 5.132 | 2.04 |
9 | 5906218 | 67 | 252.084 | 254.358 | 2.274 | 0.90 |
10 | 5906221 | 70 | 313.106 | 318.269 | 5.163 | 1.65 |
11 | 5905974 | 181 | 200.025 | 195.742 | 4.283 | 2.14 |
12 | 5906035 | 98 | 246.273 | 249.353 | 3.080 | 1.25 |
13 | 5906303 | 51 | 194.703 | 195.370 | 0.667 | 0.34 |
14 | 5906310 | 39 | 229.277 | 234.260 | 4.983 | 2.17 |
15 | 5906343 | 30 | 201.162 | 209.158 | 7.996 | 3.97 |
16 | 6902981 | 52 | 294.120 | 295.728 | 1.608 | 0.55 |
17 | 5906475 | 5 | 203.489 | 201.456 | 2.033 | 1.00 |
18 | 5906503 | 3 | 199.267 | 202.761 | 3.494 | 1.75 |
19 | 5906624 | 113 | 320.351 | 326.284 | 5.933 | 1.85 |
20 | 7900566 | 75 | 288.801 | 295.325 | 6.524 | 2.26 |
[1] |
|
[2] |
|
[3] |
|
[4] |
|
[5] |
|
[6] |
|
[7] |
|
[8] |
|
[9] |
|
[10] |
|
[11] |
刘烨, 邢小罡. Argo浮标观测溶解氧数据的原理与质量控制[J]. 海洋科学, 2019, 43(1):28-37.
[
|
[12] |
陈大可, 许建平, 马继瑞, 等. 全球实时海洋观测网(Argo)与上层海洋结构、变异及预测研究[J]. 地球科学进展, 2008, 23(1):1-7.
[
|
[13] |
|
[14] |
|
[15] |
杨明远. 基于Argo数据的海洋要素场三维网格化技术研究[D]. 郑州: 战略支援部队信息工程大学, 2018.
[
|
[16] |
李直龙, 左军成, 纪棋严, 等. 基于Argo剖面和SST以及SLA数据重构三维网格温度场[J]. 海洋预报, 2020, 37(4):66-75.
[
|
[17] |
|
[18] |
张春玲, 王振峰, 李宏. 基于最优插值法的Argo数据网格化实验[J]. 海洋测绘, 2012, 32(3):29-31.
[
|
[19] |
|
[20] |
李宏, 许建平, 刘增宏, 等. 利用逐步订正法构建Argo网格资料集的研究[J]. 海洋通报, 2012, 31(5):502-514.
[
|
[21] |
李佳霖, 樊子德, 邓敏. 基于空间异质分区的残差IDW插值方法[J]. 地理与地理信息科学, 2015, 31(5):25-29.
[
|
[22] |
杨胜龙, 马军杰, 伍玉梅, 等. 基于Kriging方法Argo数据重构太平洋温度场研究[J]. 海洋渔业, 2008, 30(1):13-18.
[
|
[23] |
张韧, 黄志松, 王辉赞, 等. 基于Argo资料的三维盐度场网格化产品重构[J]. 解放军理工大学学报(自然科学版), 2012, 13(3):342-348.
[
|
[24] |
王文娟, 俞胜宾, 杨凡, 等. 基于距平分析的Argo海表温度场重构[J]. 海洋预报, 2014, 31(6):68-73.
[
|
[25] |
樊子德, 龚健雅, 刘博, 等. 顾及时空异质性的缺失数据时空插值方法[J]. 测绘学报, 2016, 45(4):458-465.
[
|
[26] |
|
[27] |
高原, 朱娅男, 陈传法, 等. 高精度DEM建模的加权径向基函数插值方法[J]. 武汉大学学报·信息科学版. https://kns.cnki.net/kcms2/article/abstract?v=3uoqIhG8C45S0n9fL2suRadTyEVl2pW9UrhTDCdPD67Dqpioc424yYx-DW27qisUMliCeNj_am5vsDvhqzxFIpY_OIEUWI2f&uniplatform=NZKPT.
[
|
[28] |
吕海洋, 盛业华, 段平, 等. 局部最优形态参数的RBF分块地形插值方法与实验[J]. 地球信息科学学报, 2015, 17(3):260-267.
[
|
[29] |
张海平, 周星星, 代文. 空间插值方法的适用性分析初探[J]. 地理与地理信息科学, 2017, 33(6):14-18,105.
[
|
[30] |
朱会义, 刘述林, 贾绍凤. 自然地理要素空间插值的几个问题[J]. 地理研究, 2004, 23(4):425-432.
[
|
[31] |
李新运, 郑新奇, 闫弘文. 坐标与属性一体化的空间聚类方法研究[J]. 地理与地理信息科学, 2004, 20(2):38-40.
[
|
[32] |
李亚, 刘丽平, 李柏青, 等. 基于改进K-Means聚类和BP神经网络的台区线损率计算方法[J]. 中国电机工程学报, 2016, 36(17):4543-4552.
[
|
[33] |
|
[34] |
段平, 盛业华, 李佳, 等. 自适应的IDW插值方法及其在气温场中的应用[J]. 地理研究, 2014, 33(8):1417-1426.
[
|
[35] |
|
[36] |
|
[37] |
Albert Noël Declercq F. Interpolation methods for scattered sample data: Accuracy, spatial patterns, processing time[J]. Cartography and Geographic Information Systems, 1996, 23(3):128-144. DOI:10.1559/152304096782438882
|
[38] |
|
[39] |
|
/
〈 |
|
〉 |