基于QA权重NDVI时间序列重建效果评价研究————以长江流域为例

doi:10.13249/j.cnki.sgs.2022.11.016

摘要/Abstract

摘要：

通过计算质量频率和噪声比分析了长江流域MODIS NDVI质量情况，然后基于常用的S-G、A-G、D-L这3种重建方法设计了3种质量权重方案，对长江流域2001—2020年的时间序列MODIS NDVI数据进行重建，最后采用视觉对比、优质区域保真性和模拟加噪的方法对重建效果进行分析评价。结果表明：长江流域全年噪声比主要集中于75%~125%，其中冬季噪声对NDVI有较大的抑制效果，春秋季其次，而夏季噪声对NDVI有增强效果；基于质量权重方案三的S-G法对原始数据连续缺失的重建效果最好；在高质区域A-G法重建保真性较高，高质像元的R²和RMSE均值为0.9489和0.0245；在模拟加噪实验中，S-G法重建后数据丢失像元最少，与原始数据的R²平均值和标准差分别为0.8616和0.1848，RMSE为0.0035~0.4411，标准差为0.0383，表明在低质区域S-G法重建保真性较高。

关键词: NDVI时间序列重建, 长江流域, 质量保证数据（QA）权重, Savitzky-Golay

Abstract:

In this article, the quality of MODIS NDVI in the Yangtze River Basin is analyzed by calculating the mass frequency and noise ratio. Based on three commonly used reconstruction methods of S-G, A-G and D-L, three quality weight schemes are designed to reconstruct the time series MODIS NDVI data of the Yangtze River Basin from 2001 to 2020. Subsequently, the reconstruction effect is analyzed and evaluated by visual contrast, high-quality regional fidelity as well as simulation denoising. The results show that the annual noise ratio in the Yangtze River Basin is mainly 75%?125%, in which the noise in winter has a great inhibitory effect on NDVI, followed by spring and autumn, and the noise in summer has an increasing effect on NDVI. The S-G method on the basis of the third quality weight scheme exerts the optimal effect on the reconstruction of continuous missing original data. Additionally, in the high-quality region, the fidelity of A-G method is high, and the mean R² and RMSE of high-quality pixels are 0.9489 and 0.0245, respectively. Moreover, in the simulation denoising experiment, the S-G method has the least data loss pixels after reconstruction, the average value and standard deviation of the R² between the reconstructed data and the original data are 0.8616 and 0.1848 respectively the RMSE is between 0.0035 and 0.4411, and the standard deviation is 0.0383, indicating that the fidelity of the S-G method is higher in the low-quality area.

Key words: NDVI time series reconstruction, the Yangtze River Basin, QA weight, Savitzky-Golay

中图分类号:

TP79

朱慧, 胡勇, 孙芬, 王强, 马雪莹. 基于QA权重NDVI时间序列重建效果评价研究————以长江流域为例[J]. 地理科学, 2022, 42(11): 2019-2027.

Zhu Hui, Hu Yong, Sun Fen, Wang Qiang, Ma Xueying. Evaluation of Reconstruction Effect of NDVI Time Series Based on QA Weight: A Case Study of MODIS NDVI in the Yangtze River Basin[J]. SCIENTIA GEOGRAPHICA SINICA, 2022, 42(11): 2019-2027.

图/表 10

图1

表1

图2

表2

图3

图4

表3

图5

表4

图6

参考文献 26

[1]	马新萍, 白红英, 贺映娜, 等. 基于NDVI的秦岭山地植被遥感物候及其与气温的响应关系——以陕西境内为例[J]. 地理科学, 2015, 35(12): 1616-1621 doi: 10.13249/j.cnki.sgs.2015.12.016
	Ma Xinping, Bai Hongying, He Yingna et al. The vegetation remote sensing phenology of Qinling Mountains based on NDVI and it’s response to temperature: Taking within the territory of Shaanxi as an example. Scientia Geographica Sinica, 2015, 35(12): 1616-1621 doi: 10.13249/j.cnki.sgs.2015.12.016
[2]	甘春英, 王兮之, 李保生, 等. 连江流域近18年来植被覆盖度变化分析[J]. 地理科学, 2011, 31(8): 1019-1024 doi: 10.13249/j.cnki.sgs.2011.08.013
	Gan Chunying, Wang Xizhi, Li Baosheng et al. Change of vegetation coverage during recent 18 years in Lianjiang River watershed. Scientia Geographica Sinica, 2011, 31(8): 1019-1024 doi: 10.13249/j.cnki.sgs.2011.08.013
[3]	白建军, 白江涛, 王磊. 2000—2010年陕北地区植被NDVI时空变化及其与区域气候的关系[J]. 地理科学, 2014, 34(7): 882-888 doi: 10.13249/j.cnki.sgs.2014.07.002
	Bai Jianjun, Bai Jiantao, Wang Lei. Spatio-temporal change of vegetation NDVI and its relations with regional climate in Northern Shaanxi Province in 2000-2010. Scientia Geographica Sinica, 2014, 34(7): 882-888 doi: 10.13249/j.cnki.sgs.2014.07.002
[4]	王建国, 张飞. 2000—2019年新疆植被覆盖度时空格局及重心变化分析[J]. 农业工程学报, 2020, 36(20): 188-194 doi: 10.11975/j.issn.1002-6819.2020.20.022
	Wang Jianguo, Zhang Fei. Spatial-temporal pattern and gravity center change of fractional vegetation cover in Xinjiang, China from 2000 to 2019. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2020, 36(20): 188-194 doi: 10.11975/j.issn.1002-6819.2020.20.022
[5]	陈雨思, 李丹, 黎臻, 等. 多时相MODIS影像的黑龙江省水稻种植面积提取[J]. 农业工程学报, 2020, 36(23): 201-208 doi: 10.11975/j.issn.1002-6819.2020.23.023
	Chen Yusi, Li Dan, Li Zhen et al. Rice planting area extraction based on multi-temporal MODIS images in Heilongjiang Province of China. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2020, 36(23): 201-208 doi: 10.11975/j.issn.1002-6819.2020.23.023
[6]	刘哲, 汪雪滢, 刘帝佑, 等. 基于MODIS数据的黄淮海夏玉米高温风险空间分布[J]. 农业工程学报, 2018, 34(9): 175-181 doi: 10.11975/j.issn.1002-6819.2018.09.021
	Liu Zhe, Wang Xueying, Liu Diyou et al. Spatial distribution of high temperature risk on summer maize in Huang-Huai-Hai Plain based on MODIS data. Transactions of the Chinese Society of Agricultural Engineering (Transactions of the CSAE), 2018, 34(9): 175-181 doi: 10.11975/j.issn.1002-6819.2018.09.021
[7]	耿丽英, 马明国. 长时间序列NDVI数据重建方法比较研究进展[J]. 遥感技术与应用, 2014, 29(2): 362-368
	Geng Liying, Ma Mingguo. Advance in method comparison of reconstructing remote sensing time series data sets. Remote Sensing Technology and Application, 2014, 29(2): 362-368
[8]	Yu Wentao, Li Jing, Liu Qinhuo et al. Gap filling for historical landsat NDVI time series by integrating climate data[J]. Remote Sensing, 2021, 13(3): 484-505 doi: 10.3390/rs13030484
[9]	Carlos Granero-Belinchon, Karine Adeline, Aude Lemonsu et al. Phenological dynamics characterization of Alignment Trees with Sentinel-2 Imagery: A vegetation indices time series reconstruction methodology adapted to urban areas[J]. Remote Sensing, 2020, 12(4): 1613-1635
[10]	李杭燕, 马明国, 谭俊磊. 时序NDVI数据集重建综合方法研究[J]. 遥感技术与应用, 2010, 25(6): 891-896 doi: 10.11873/j.issn.1004-0323.2010.6.891
	Li Hangyan, Ma Mingguo, Tan Junlei. Intergrated reconstruction method of time series NDVI Dataset. Remote Sensing Technology and Application, 2010, 25(6): 891-896 doi: 10.11873/j.issn.1004-0323.2010.6.891
[11]	Suman Kumar Padhee, Subashisa Dutta. Spatio-temporal reconstruction of MODIS NDVI by regional land surface phenology and harmonic analysis of time-series.[J] GIScience& Remote Sensing, 2019, 56(8):1261-1288.
[12]	朱慧, 李军. 三种时序NDVI重建方法的对比研究——以重庆为例[J]. 山地学报, 2017(6): 919-925.
	Zhu Hui, Li Jun. Three timed-series NDVI reconstruction methods: A case study of Chongqing. Mountain Research, 2017(6): 919-925.
[13]	马明国, 宋怡, 王旭峰, 等. AVHRR, VEGETATION和MODIS时间系列遥感数据产品现状与应用研究进展[J]. 遥感技术与应用, 2012, 27(5): 663-670 doi: 10.11873/j.issn.1004-0323.2012.5.663
	Ma Mingguo, Song Yi, Wang Xufeng et al. Development status and application research of the Time Series Remote Sensing Data Products based on AVHRR, VEGETATION and MODIS. Remote Sensing Technology and Application, 2012, 27(5): 663-670 doi: 10.11873/j.issn.1004-0323.2012.5.663
[14]	周增光, 唐娉. 基于质量权重的Savitzky-Golay时间序列滤波方法[J]. 遥感技术与应用, 2013, 28(2): 232-239.
	Zhou Zengguang, Tang Ping. VI-Quality-Based Savitzky-Golay Method for filtering time series data. Remote Sensing Technology and Application, 2013, 28(2): 232-239.
[15]	程琳琳, 李玉虎, 孙海元, 等. 京津冀MODIS长时序增强型植被指数拟合重建方法适用性研究[J]. 农业工程学报, 2019, 35(11): 148-158 doi: 10.11975/j.issn.1002-6819.2019.11.017
	Cheng Linlin, Li Yuhu, Sun Haiyuan et al. Applicability of fitting and reconstruction method of MODIS long-time enhanced vegetation index in Beijing-Tianjin-Hebei. Transactions of the Chinese Society of Agricultural Engineering, 2019, 35(11): 148-158 doi: 10.11975/j.issn.1002-6819.2019.11.017
[16]	王正兴, 曹云锋. 东南亚地区2010年MODIS-NDVI再分析数据集[J]. 全球变化数据学报(中英文), 2017, 1(3): 317-323
	Wang Zhengxing, Cao Yunfeng. Reanalysis dataset on MODIS-NDVI in Southeast Asia, 2010. Journal of Global Change Data & Discovery, 2017, 1(3): 317-323
[17]	Tao Jianbin, Wu Wenbin, Zhou Yong et al. Mapping winter wheat using phenological feature of peak before winter on the North China Plain based on time-series MODIS data[J]. Journal of Integrative Agriculture, 2017, 16(2): 348-359 doi: 10.1016/S2095-3119(15)61304-1
[18]	卫炜, 吴文斌, 李正国, 等. 时间序列植被指数重构方法比对研究[J]. 中国农业资源与规划, 2014, 35(1): 34-43
	Wei Wei, Wu Wenbin, Li Zhengguo et al. Comparison of three methods for reconstructing time series vegetation index. Chinese Journal of Agricultural Resources and Regional Planning, 2014, 35(1): 34-43
[19]	张晗, 任志远. 多种时序NDVI重建方法比较与应用分析[J]. 中国农业科学, 2014, 47(15): 2998-3008 doi: 10.3864/j.issn.0578-1752.2014.15.010
	Zhang Han, Ren Zhiyuan. Comparison and application analysis of several NDVI time-series reconstruction methods. Scientia Agricultura Sinica, 2014, 47(15): 2998-3008 doi: 10.3864/j.issn.0578-1752.2014.15.010
[20]	熊文, 李志军, 黄羽, 等.中华人民共和国长江保护法要点解读[M].武汉: 长江出版社,2021.
	Xiong Wen, Li Zhijun, Huang Yu et al. Interpretation of key points of the Yangtze River protection law of the People's Republic of China. Wuhan: Changjiang Publishing House,2021.
[21]	彭俊, 凌敏, 俞珊妮, 等. 长江流域径流量变化过程及其对ENSO和PDO的响应[J]. 地理科学,2022,42(3):515-526.
	Peng Jun, Ling Min, Yu Shanni et al. Variation process of water discharge and its response to ENSO and PDO in the Yangtze River Basin. Scientia Geographica Sinica,2022,42(3):515-526.
[22]	李军, 朱慧. 重庆地区MODIS/NDVI时间序列数据重建研究[J]. 地理科学, 2017, 37(3): 437-444 doi: 10.13249/j.cnki.sgs.2017.03.014
	Li Jun, Zhu Hui. The reconstruction of MODIS/NDVI time series data in Chongqing. Scientia Geographica Sinica, 2017, 37(3): 437-444 doi: 10.13249/j.cnki.sgs.2017.03.014
[23]	Lovell J L, Graetz R D. Filtering Pathfinder AVH RR land NDVI data for Australia[J]. International Journal of Remote Sensing, 2001, 22(13): 2649-2654 doi: 10.1080/01431160116874
[24]	朱慧. 重庆地区NDVI时间序列重建研究[D]. 重庆: 重庆师范大学, 2017.
	Zhu Hui. Study on the reconstruction of NDVI time series data in Chongqing. Chongqing: Chongqing Nomal University, 2017.
[25]	Chen J, Jönsson P, Tamur M et al. A simple method for reconstructing a high-quality NDVI time-series data set based on the Savitzky–Golayﬁlter[J]. Remote Sensing of Environment, 2004, 91: 332-344 doi: 10.1016/j.rse.2004.03.014
[26]	Jönsson P, Eklundh L. Seasonality extraction by function fitting to time-series of satellite sensor data[J]. IEEE Transaction on Geoscience and Remote Sensing, 2002, 40(8): 1824-1832 doi: 10.1109/TGRS.2002.802519

Bit	转换后DN值	含义
注：Bit代表转换前的数据质量。
00	0	VI产品具有良好的质量
01	1	VI产品需参考描述信息使用
10	2	VI产品受云污染
11	3	VI产品质量差

$ {NDVI}_{noise} $	1月	2月	3月	4月	5月	6月	7月	8月	9月	10月	11月	12月	平均值
注：加粗数值为该月分布最大占比。
<−200	0.63	0.63	1.36	2.28	0.93	0.35	0.23	0.16	0.22	0.39	1.33	0.65	0.76
−200~0	1.05	0.98	2.83	6.04	3.09	0.86	0.54	0.19	0.36	1.03	4.56	1.40	1.91
0~25	18.74	15.81	6.01	9.25	4.84	2.30	2.25	3.61	14.56	13.55	8.89	20.73	10.05
25~50	11.97	9.78	6.45	9.84	6.48	5.40	7.51	6.66	12.50	10.91	7.78	8.30	8.63
50~75	17.08	12.16	9.64	12.15	9.45	9.88	17.04	11.91	17.69	16.32	10.66	10.76	12.90
75~100	17.82	17.76	21.26	23.43	22.49	21.64	27.61	17.14	23.50	23.21	16.53	16.28	20.72
100~125	14.37	17.31	23.00	15.67	26.45	25.72	25.86	17.25	15.32	17.97	19.13	17.52	19.63
125~150	7.73	9.85	11.69	8.43	14.07	21.44	12.66	22.55	10.21	10.02	12.97	11.85	12.79
150~175	3.75	5.95	5.82	3.94	4.98	6.82	3.12	10.40	2.64	3.11	6.69	5.45	5.22
175~200	2.21	3.13	3.05	1.99	2.61	2.76	1.26	3.59	0.88	1.13	3.24	2.44	2.36
>200	4.66	6.64	8.88	6.97	4.62	2.83	1.92	6.54	2.13	2.34	8.21	4.61	5.03

重建方法	权重	R²					RMSE
重建方法	权重	最大值	最小值	平均值	中位数	标准差	最大值	最小值	平均值	中位数	标准差
注：加粗数值为最佳值。
A-G	方案一	0.9971	0.9150	0.9499	0.9670	0.0375	0.1162	0.0295	0.0243	0.0229	0.0090
A-G	方案二、三	0.9969	0.9143	0.9489	0.9664	0.1040	0.1177	0.0296	0.0245	0.0231	0.0112
D-L	方案一	0.9966	0.9133	0.9481	0.9658	0.0391	0.1285	0.0298	0.0247	0.0235	0.0090
D-L	方案二、三	0.9966	0.913	0.9471	0.9652	0.0402	0.1299	0.0298	0.0249	0.0238	0.0091
S-G	方案一	0.9850	0.8843	0.9059	0.9352	0.0567	0.1380	0.0344	0.0346	0.033	0.0123
S-G	方案二、三	0.9847	0.8817	0.9042	0.9336	0.0738	0.1391	0.0348	0.0349	0.0334	0.0131

重建方法	R²					RMSE
重建方法	最大值	最小值	平均值	中位数	标准差	最大值	最小值	平均值	中位数	标准差
注：加粗字体为最佳值。
A-G	0.9999	−0.8916	0.8520	0.9416	0.2389	0.9832	0.0016	0.0692	0.0547	0.0512
D-L	0.9999	−0.9741	0.8333	0.9278	0.2449	0.9168	0.0016	0.0760	0.0610	0.0534
S-G	0.9983	−0.8987	0.8616	0.9238	0.1848	0.4411	0.0035	0.0714	0.0628	0.0383