Refined remote sensing classification of Yancheng coastal wetland considering tide-level changes and vegetation phenological characteristics on the GEE platform
-
摘要: 滨海湿地具有重要的经济价值与生态价值，快速准确地监测其现状对滨海湿地资源的保护和管理具有重要意义。由于潮汐动态变化、植被光谱相似性以及云覆盖等因素的影响，滨海湿地的遥感监测具有较大挑战。本文提出了一个综合考虑潮位变化及植被物候特征的滨海湿地遥感分类方法，基于GEE（Google Earth Engine）平台，首先引入Fmask（Function of mask）算法进行云检测与去云处理，然后利用S-G（Savitzky-Golay）滤波算法重构NDVI时间序列数据，提取植被物候特征参数，采用随机森林算法实现互花米草（Spartina alterniflora）、芦苇（Phragmites australis）、碱蓬（Suaeda salsa）与茅草（Imperata cylindrica）4种湿地植被类型的提取；最后利用最大光谱指数合成算法 (Maximum Spectral Index Composite, MSIC) 生成最高与最低潮位合成影像，结合大津算法（Otsu）提取光滩与海水，实现滨海湿地的精细化遥感分类。研究结果表明，生长季开始时间、生长季结束时间、生长季时长、基准值、振幅、小季节积分是区分滨海湿地植被的重要植被物候特征参数。利用本方法对盐城滨海湿地进行分类，湿地总体分类精度达96.50%，Kappa系数为0.957 1，湿地植被中互花米草的使用者精度最高，为96.59%；其次是芦苇与碱蓬；茅草最低，为93.55%。与面向对象分类相比，本方法不仅能够提取完整的光滩范围，而且将总体精度提高10.25%，体现出植被物候特征在滨海湿地动态变化遥感监测中的应用潜力。Abstract: Coastal wetlands have important economic and ecological value. Rapid and accurate monitoring of the status of coastal wetlands is of great significance for the protection and management of coastal wetland resources. Due to factors such as the variability of the tide-level changes, similarity of vegetation spectra, and frequent cloud cover, remote sensing monitoring of coastal wetlands faced certain challenges. In this paper, we proposed a multi-technology coupled remote sensing classification method of coastal wetlands that considers tide-level changes and vegetation phenological characteristics. Based on the Google Earth Engine (GEE) platform, the Fmask (Function of mask) algorithm was first performed for cloud testing and cloud removal processing. Then, the S-G (Savitzky-Golay) filtering algorithm was used to reconstruct NDVI time series data and extract vegetation phenological characteristic parameters. In this phase, the random forest algorithm was applied for the classification of four vegetation types namely Phragmites australi (PA), Suaeda salsa (SS), Spartina alterniflora (SA), and Imperata cylindrical (IC). Finally, the Maximum Spectral Index Composite (MSIC) algorithm was used to generate composite images of the highest and lowest tide levels. The tidal flats and seawater were precisely extracted using the Otsu algorithm based on these two composite images. Combining these feature types, the refined remote sensing classification of coastal wetlands was ideally obtained. The results showed that start-of-season time, end-of-season time, length of season, base value, amplitude, and small seasonal integral were the six key vegetation phenological characteristic parameters for distinguishing different types of coastal wetland vegetation. Applying this method to classify coastal wetlands on the Yancheng coast, the overall classification accuracy was 96.50%, and the Kappa coefficient reached 0.957 1. Among the wetland vegetation, the highest user accuracy was 96.59% for SA, followed by PA and SS, and the lowest was 93.55% for IC. Compared with object-oriented methods, our method can extract the complete range of tidal flats, and the overall accuracy is improved by 10.25%, reflecting the potential application of vegetation phenological characteristics in remote sensing monitoring of dynamic changes in coastal wetlands.
-
图 3 不同植被类型NDVI时间序列拟合曲线
Fig. 3 Time series fitting curves of different vegetation types
图 4 盐城滨海湿地植被物候特征参数对比
Fig. 4 Comparison of vegetation phenological characteristics parameters of Yancheng coastal wetland vegetation
图 5 盐城滨海湿地分类结果
a. 盐城滨海湿地分类结果图；b. 植被混生带；c. 植被与内陆水体相接地带；d. 植被与光滩相接地带
Fig. 5 Classification results of Yancheng coastal wetland
a. Classification results map of Yancheng coastal wetland; b. the mixed zone of vegetation; c. the adjacent zone of vegetation and inland water; d. the adjacent zone of vegetation and tidal flats
图 6 面向对象方法分类结果
a. 盐城滨海湿地分类结果图；b. 植被混生带；c. 植被与内陆水体相接地带；d. 植被与光滩相接地带
Fig. 6 Classification results of the object-oriented methods
a. Classification results map of Yancheng coastal wetland; b. the mixed zone of vegetation; c. the adjacent zone of vegetation and inland water; d. the adjacent zone of vegetation and tidal flats
图 7 不同时间序列数据重构方法分类精度对比
Fig. 7 Comparison of classification accuracy of different time series data reconstruction methods
图 8 判别碱蓬的3个重要物候特征参数对比
Fig. 8 Comparison of three important phenological characteristics parameters of the discriminate Suaeda salsa
图 9 盐城滨海湿地植被分类结果图 ((a) QA60 + HANTS; (b) QA60 + S-G; (c) Fmask + HANTS)
Fig. 9 Classification results map of coastal wetland vegetation in Yancheng
表 1 Fmask算法的公式与阈值
Tab. 1 Formula and threshold of Fmask algorithm
名称 公式与阈值 公式编号 基本检测 ${ \begin{gathered} {\rho _{SWIR2}} > 0.03{\text{ }}and{\text{ }}NDSI < 0.8{\text{ }}and{\text{ }}NDVI < 0.8,{\text{ }} \\ \left\{ \begin{gathered} NDSI = ({\rho _{Green}} - {\rho _{SWIR1}})/({\rho _{Green}} + {\rho _{SWIR1}}) \\ NDVI = ({\rho _{NIR}} - {\rho _{Red}})/({\rho _{NIR}} + {\rho _{Red}}) \\ \end{gathered} \right. \\ \end{gathered}} $ （1） 白度检测(Whiteness) ${ \begin{gathered} Whiteness = \left| {\left( {{\rho _{Red}} - MeanVis} \right)/MeanVis} \right|+\left| {\left( {{\rho _{Green}} - MeanVis} \right)/MeanVis} \right| + \\ \left| {\left( {{\rho _{Blue}} - MeanVis} \right)/MeanVis} \right| < 0.7,{\text{ }}MeanVis = \left( {{\rho _{Red}} + {\rho _{Green}} + {\rho _{Blue}}} \right)/3 \end{gathered} }$ （2） 霾优化转换检测(HOT) $ {HOT = {\rho _{Blue}} - 0.5 \times {\rho _{Red}} - 0.08 > 0 }$ （3） 比值检测 $ {{\rho _{NIR}}/{\rho _{SWIR1}} > 0.75 }$ （4） 卷云检测(Cir) ${ Cir = {\rho _{Cir}}/0.04 > 0.01 }$ （5） 水检测(Water) ${ \begin{gathered} Water = (NDVI < 0.01{\text{ }}and{\text{ }}{\rho _{NIR}} < 0.11){\text{ or }}(NDVI < 0.1{\text{ }}and{\text{ }}{\rho _{NIR}} < 0.05) \\ {\text{ or }}(GSWO > {O_{water}}{\text{ }}and{\text{ }}snow/ice = false) \\ \end{gathered} }$ （6） 陆地云概率(lCloudp) ${ \begin{gathered} lCloudp = lVar \times lHOT + 0.5 \times Cir > lCloudt, \\ \left\{ \begin{gathered} lHOT = \frac{{HOT - \left( {HO{T_{low}} - 0.04} \right)}}{{\left( {HO{T_{high}} + 0.04} \right) - \left( {HO{T_{low}} - 0.04} \right)}} \\ lVar = 1 - \max \left( {\left| {NDVI} \right|,\left| {NDSI} \right|,\left| {Whiteness} \right|} \right) \\ \end{gathered} \right. \\ \end{gathered}} $ （7） 水体云概率(wCloudp) ${ wCloudp = wBright + 0.5 \times Cir > wCloudt,{\text{ }}wBright = \min ({\rho _{SWIR1}},0.11)/0.11} $ （8） 云阴影检测 ${ \left\{ \begin{gathered} {\rho _{NIR}} < 0.25 \\ {\rho _{SWIR1}} < 0.11 \\ \end{gathered} \right. }$ （9） 注：ρBlue、ρGreen、ρRed、ρNIR、ρSWIR1、ρSWIR2与ρcir分别为蓝、绿、红、近红外、短波与卷云波段的反射率。MeanVis为平均反射率。GSWO（Global Surface Water Occurrence）为全球地表水发生率数据，为每个像元提供水发生率，其中0%表示永久的陆地，100%表示永久的水体；snow/ice为雪、冰；Owater为水发生率阈值，取光谱识别的水体像元集的低值（17.5百分位值）再减去5%的GSWO数据误差。lHOT为陆地HOT云概率，HOTlow等于陆地绝对无云像元的HOT低值（17.5百分位值），HOThigh等于陆地绝对无云像元的HOT高值（82.5百分位值）；lCloudt为陆地云概率阈值，等于陆地绝对无云像元的lCloudp高值（82.5百分位值）再加上常数0.2。lVar为陆地光谱差异概率。wCloudt为水体云概率阈值，等于水体绝对无云像元的wCloudp高值（82.5百分位值）再加上常数0.2。wBright为水体亮度概率。 
下载: 导出CSV
表 2 不同植被类型的植被物候特征判别参数与阈值
Tab. 2 Discriminative parameters and thresholds for vegetation phenological characteristics of different vegetation types
植被类型 植被物候特征判别参数与阈值 芦苇 120 < SOS < 150 碱蓬 SI < 50 and BV > 0.22 and AV < 0.4 互花米草 340 < EOS < 380 茅草 LOS < 130 and AV > 0.55
下载: 导出CSV
表 3 盐城滨海湿地分类精度评价结果
Tab. 3 Accuracy evaluation of coastal wetland classification
芦苇 碱蓬 互花米草 茅草 内陆水体 光滩 海水 使用者精度/% 芦苇 113 1 2 1 0 0 0 96.58 碱蓬 1 52 1 1 0 0 0 94.55 互花米草 1 2 85 0 0 0 0 96.59 茅草 0 0 2 29 0 0 0 93.55 内陆水体 0 0 1 0 36 0 0 97.30 光滩 0 0 0 0 0 35 0 100.00 海水 0 0 0 0 0 1 36 97.30 生产者精度/% 98.26 94.55 93.41 93.55 100.00 97.22 100.00 总体精度/% 96.50 Kappa系数 0.957 1
下载: 导出CSV
表 4 不同分类方法分类精度对比
Tab. 4 Comparison of classification accuracy of different techniques
指标 面向对象方法 与本文方法相比 精度/% 芦苇 82.76 −13.82 碱蓬 88.89 −5.66 互花米草 82.02 −14.57 茅草 93.33 −0.22 内陆水体 78.57 −18.73 光滩 96.88 −3.12 海水 97.30 0.00 总体精度 86.25 −10.25 Kappa系数 0.831 6 −0.125 5
下载: 导出CSV
-
[1] 解雪峰, 孙晓敏, 吴涛, 等. 互花米草入侵对滨海湿地生态系统的影响研究进展[J]. 应用生态学报, 2020, 31(6): 2119−2128.Xie Xuefeng, Sun Xiaomin, Wu Tao, et al. Impacts of Spartina alterniflora invasion on coastal wetland ecosystem: Advances and prospects[J]. Chinese Journal of Applied Ecology, 2020, 31(6): 2119−2128. [2] Kirwan M L, Megonigal J P. Tidal wetland stability in the face of human impacts and sea-level rise[J]. Nature, 2013, 504(7478): 53−60. doi: 10.1038/nature12856 [3] Gabler C A, Osland M J, Grace J B, et al. Macroclimatic change expected to transform coastal wetland ecosystems this century[J]. Nature Climate Change, 2017, 7(2): 142−147. doi: 10.1038/nclimate3203 [4] 刘瑞清, 李加林, 孙超, 等. 基于Sentinel-2遥感时间序列植被物候特征的盐城滨海湿地植被分类[J]. 地理学报, 2021, 76(7): 1680−1692. doi: 10.11821/dlxb202107008Liu Ruiqing, Li Jialin, Sun Chao, et al. Classification of Yancheng coastal wetland vegetation based on vegetation phenological characteristics derived from Sentinel-2 time-series[J]. Acta Geographica Sinica, 2021, 76(7): 1680−1692. doi: 10.11821/dlxb202107008 [5] Chung C H. Forty years of ecological engineering with Spartina plantations in China[J]. Ecological Engineering, 2006, 27(1): 49−57. doi: 10.1016/j.ecoleng.2005.09.012 [6] Tang Long, Gao Yang, Wang Chenghuan, et al. A plant invader declines through its modification to habitats: A case study of a 16-year chronosequence of Spartina alterniflora invasion in a salt marsh[J]. Ecological Engineering, 2012, 49: 181−185. doi: 10.1016/j.ecoleng.2012.08.024 [7] Zeng Jing, Sun Yonghua, Cao Peirun, et al. A phenology-based vegetation index classification (PVC) algorithm for coastal salt marshes using Landsat 8 images[J]. International Journal of Applied Earth Observation and Geoinformation, 2022, 110: 102776. doi: 10.1016/j.jag.2022.102776 [8] Murray N J, Phinn S R, Clemens R S, et al. Continental scale mapping of tidal flats across East Asia using the Landsat archive[J]. Remote Sensing, 2012, 4(11): 3417−3426. doi: 10.3390/rs4113417 [9] 张晨宇, 陈沈良, 李鹏, 等. 现行黄河口保护区典型湿地植被时空动态遥感监测[J]. 海洋学报, 2022, 44(1): 125−136.Zhang Chenyu, Chen Shenliang, Li Peng, et al. Spatiotemporal dynamic remote sensing monitoring of typical wetland vegetation in the Current Huanghe River Estuary Reserve[J]. Haiyang Xuebao, 2022, 44(1): 125−136. [10] Zhang Zhen, Xu Nan, Li Yangfan, et al. Sub-continental-scale mapping of tidal wetland composition for East Asia: A novel algorithm integrating satellite tide-level and phenological features[J]. Remote Sensing of Environment, 2022, 269: 112799. doi: 10.1016/j.rse.2021.112799 [11] Dong Jinwei, Xiao Xiangming, Menarguez M A, et al. Mapping paddy rice planting area in northeastern Asia with Landsat 8 images, phenology-based algorithm and Google Earth Engine[J]. Remote Sensing of Environment, 2016, 185: 142−154. doi: 10.1016/j.rse.2016.02.016 [12] 程丽娜, 钟才荣, 李晓燕, 等. Sentinel-2密集时间序列数据和Google Earth Engine的潮间带湿地快速自动分类[J]. 遥感学报, 2022, 26(2): 348−357. doi: 10.11834/jrs.20211311Cheng Li’na, Zhong Cairong, Li Xiaoyan, et al. Rapid and automatic classification of intertidal wetlands based on intensive time series Sentinel-2 images and Google Earth Engine[J]. National Remote Sensing Bulletin, 2022, 26(2): 348−357. doi: 10.11834/jrs.20211311 [13] Sun Chao, Li Jialin, Liu Yongchao, et al. Tracking annual changes in the distribution and composition of saltmarsh vegetation on the Jiangsu coast of China using Landsat time series-based phenological parameters[J]. Remote Sensing of Environment, 2023, 284: 113370. doi: 10.1016/j.rse.2022.113370 [14] Jia Mingming, Wang Zongming, Mao Dehua, et al. Rapid, robust, and automated mapping of tidal flats in China using time series Sentinel-2 images and Google Earth Engine[J]. Remote Sensing of Environment, 2021, 255: 112285. doi: 10.1016/j.rse.2021.112285 [15] Sun Chao, Liu Yongxue, Zhao Saishuai, et al. Classification mapping and species identification of salt marshes based on a short-time interval NDVI time-series from HJ-1 optical imagery[J]. International Journal of Applied Earth Observation and Geoinformation, 2016, 45: 27−41. doi: 10.1016/j.jag.2015.10.008 [16] 王敏钰, 罗毅, 张正阳, 等. 植被物候参数遥感提取与验证方法研究进展[J]. 遥感学报, 2022, 26(3): 431−455. doi: 10.11834/j.issn.1007-4619.2022.3.ygxb202203002Wang Minyu, Luo Yi, Zhang Zhengyang, et al. Recent advances in remote sensing of vegetation phenology: Retrieval algorithm and validation strategy[J]. National Remote Sensing Bulletin, 2022, 26(3): 431−455. doi: 10.11834/j.issn.1007-4619.2022.3.ygxb202203002 [17] Descals A, Verger A, Yin Gaofei, et al. Improved estimates of arctic land surface phenology using sentinel-2 time series[J]. Remote Sensing, 2020, 12(22): 3738. doi: 10.3390/rs12223738 [18] Wu Nan, Shi Runhe, Zhuo Wei, et al. A classification of tidal flat wetland vegetation combining phenological features with Google earth engine[J]. Remote Sensing, 2021, 13(3): 443. doi: 10.3390/rs13030443 [19] 薛朝辉, 钱思羽. 融合Landsat 8与Sentinel-2数据的红树林物候信息提取与分类[J]. 遥感学报, 2022, 26(6): 1121−1142. doi: 10.11834/jrs.20221448Xue Zhaohui, Qian Siyu. Fusion of Landsat 8 and Sentinel-2 data for mangrove phenology information extraction and classification[J]. National Remote Sensing Bulletin, 2022, 26(6): 1121−1142. doi: 10.11834/jrs.20221448 [20] Savitzky A, Golay M J E. Smoothing and differentiation of data by simplified least squares procedures[J]. Analytical Chemistry, 1964, 36(8): 1627−1639. doi: 10.1021/ac60214a047 [21] 王明, 刘正佳, 陈元琰. 基于Sentinel-2波段/产品的图像云检测效果对比研究[J]. 遥感技术与应用, 2020, 35(5): 1167−1177.Wang Ming, Liu Zhengjia, Chen Yuanyan. Comparsions of image cloud detection effect based on Sentinel-2 bands/products[J]. Remote Sensing Technology and Application, 2020, 35(5): 1167−1177. [22] Coluzzi R, Imbrenda V, Lanfredi M, et al. A first assessment of the Sentinel-2 Level 1-C cloud mask product to support informed surface analyses[J]. Remote Sensing of Environment, 2018, 217: 426−443. doi: 10.1016/j.rse.2018.08.009 [23] 陈君, 王义刚, 张忍顺, 等. 江苏岸外辐射沙脊群东沙稳定性研究[J]. 海洋工程, 2007, 25(1): 105−113. doi: 10.3969/j.issn.1005-9865.2007.01.017Chen Jun, Wang Yigang, Zhang Renshun, et al. Stability study on the Dongsha Sandbanks in submarine radial sand ridges field off Jiangsu Coast[J]. The Ocean Engineering, 2007, 25(1): 105−113. doi: 10.3969/j.issn.1005-9865.2007.01.017 [24] 陈玮彤, 张东, 施顺杰, 等. 江苏中部淤泥质海岸岸线变化遥感监测研究[J]. 海洋学报, 2017, 39(5): 138−148.Chen Weitong, Zhang Dong, Shi Shunjie, et al. Research on monitoring coastline changes by remote sensing in muddy coast, central Jiangsu coast[J]. Haiyang Xuebao, 2017, 39(5): 138−148. [25] Zhu Zhe, Woodcock C E. Object-based cloud and cloud shadow detection in Landsat imagery[J]. Remote Sensing of Environment, 2012, 118: 83−94. doi: 10.1016/j.rse.2011.10.028 [26] Zhu Zhe, Wang Shixiong, Woodcock C E. Improvement and expansion of the Fmask algorithm: cloud, cloud shadow, and snow detection for Landsats 4-7, 8, and Sentinel 2 images[J]. Remote Sensing of Environment, 2015, 159: 269−277. doi: 10.1016/j.rse.2014.12.014 [27] Shi Qiu, Zhu Zhe, He Binbin. Fmask 4.0: Improved cloud and cloud shadow detection in Landsats 4-8 and Sentinel-2 imagery[J]. Remote Sensing of Environment, 2019, 231: 111205. doi: 10.1016/j.rse.2019.05.024 [28] Zhang Yanli, Song Yuyu, Ye Changqing, et al. An integrated approach to reconstructing snow cover under clouds and cloud shadows on Sentinel-2 Time-Series images in a mountainous area[J]. Journal of Hydrology, 2023, 619: 129264. doi: 10.1016/j.jhydrol.2023.129264 [29] Sun Lin, Liu Xinyan, Yang Yikun, et al. A cloud shadow detection method combined with cloud height iteration and spectral analysis for Landsat 8 OLI data[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2018, 138: 193−207. doi: 10.1016/j.isprsjprs.2018.02.016 [30] Steinbach S, Hentschel E, Hentze K, et al. Automatization and evaluation of a remote sensing-based indicator for wetland health assessment in East Africa on national and local scales[J]. Ecological Informatics, 2023, 75: 102032. doi: 10.1016/j.ecoinf.2023.102032 [31] 于信芳, 庄大方. 基于MODIS NDVI数据的东北森林物候期监测[J]. 资源科学, 2006, 28(4): 111−117. doi: 10.3321/j.issn:1007-7588.2006.04.023Yu Xinfang, Zhuang Dafang. Monitoring forest phenophases of northeast China based on MODIS NDVI data[J]. Resources Science, 2006, 28(4): 111−117. doi: 10.3321/j.issn:1007-7588.2006.04.023 [32] Jönsson P, Eklundh L. TIMESAT-a program for analyzing time-series of satellite sensor data[J]. Computers & Geosciences, 2004, 30(8): 833−845. [33] White M A, Thornton P E, Running S W. A continental phenology model for monitoring vegetation responses to interannual climatic variability[J]. Global Biogeochemical Cycles, 1997, 11(2): 217−234. doi: 10.1029/97GB00330 [34] Otsu N. A threshold selection method from gray-level histograms[J]. IEEE Transactions on Systems, Man, and Cybernetics, 1979, 9(1): 62−66. doi: 10.1109/TSMC.1979.4310076 [35] Xing Huaqiao, Niu Jingge, Feng Yongyu, et al. A coastal wetlands mapping approach of Yellow River Delta with a hierarchical classification and optimal feature selection framework[J]. CATENA, 2023, 223: 106897. doi: 10.1016/j.catena.2022.106897 -
点击查看大图
图(9) / 表(4)
计量
- 文章访问数: 55
- HTML全文浏览量: 31
- PDF下载量: 13
- 被引次数: 0
