Journal of Geo-information Science >
Review on Monitoring Open Surface Water Body Using Remote Sensing
Received date: 2019-09-12
Request revised date: 2019-10-17
Online published: 2019-12-11
Supported by
Strategic Priority Research Program(No.XDA19040301)
Key Research Program of Frontier Sciences of Chinese Academy of Sciences(No.QYZDB-SSW-DQC005)
Copyright
Open surface water bodies play important roles in industrial and agricultural production, climate regulation, and ecosystem maintenance. The spatial distributions of surface water bodies are always changing due to climate change and anthropogenic activities. Therefore, rapid and accurate monitoring of the spatiotemporal dynamics of surface water bodies is of great significance for water resources management and protection, as well as prediction of climate change. Remote sensing technology with the advantages of broader perspective, stronger timely effectiveness, larger information, and the ability of not affected by geographical environment provides a new way to monitor the dynamics of open surface water bodies over large extents, especially in remote and inaccessible mountain regions. The approach of the era of big earth data leads to the continuous improvements of water body mapping algorithms and increasing amounts of remote sensing data. However, there still lacks systematic review and evaluation about the evolution of relevant algorithms and data sources. In this context, based on the relevant literature ranging from the 1980s to 2018, this paper reviewed and assessed the existing algorithms and remote sensing data sources used in open surface water body mapping, and concluded the evolution processes of the common algorithms, such as single-band threshold approach, multi-band spectral relationship approach, spectral- and index-based approach, Support Vector Machine (SVM), Random Forest (RF), and Deep Learning (DL). Besides, we summarized the evolution of remote sensing data from coarse spatial resolutions (e.g. MODIS) to medium (e.g. Landsat) and high (e.g. GF-1/2) spatial resolutions. Furthermore, the different performances between these algorithms and data used in the studies of water body changes were analyzed. Also, we demonstrated the development of computing platforms from local computer to high performance cloud computing platforms such as Google Earth Engine (GEE) and Amazon Web Service (AWS), and highlighted typical cases that conduct retrospective and continuous monitoring of land cover changes over the global or regional scales. Then, we discussed the progress of studies focusing on the monitoring of open surface water body changes, from epoch-based analyses to interannual change analyses. Finally, we discussed the significance of the combined use of multi-source remote sensing data fusion and cloud computing platforms in the continuous monitoring of surface water body changes, and the uncertainties in detecting the different types of surface water bodies.
ZHOU Yan , DONG Jinwei . Review on Monitoring Open Surface Water Body Using Remote Sensing[J]. Journal of Geo-information Science, 2019 , 21(11) : 1768 -1778 . DOI: 10.12082/dqxxkx.2019.190518
表1 基于水体指数与阈值的水体提取算法总结Tab. 1 A summary of open surface water mapping algorithms based on water indices and thresholds |
算法 | 水体指数 | ||
---|---|---|---|
TCWCrist[19,20] | TCWCrist = 0.1509 × Bblue + 0.1973 × Bgreen + 0.3279 × Bred + 0.3406 × BNir - 0.7112 × BSWIR-1 - 0.4572 × BSWIR-2 | ||
NDWI[21] | NDWI=(Bgreen - BNir) / (Bgreen + BNir) | ||
mNDWI[22,23] | mNDWI=(Bgreen - BSWIR-1) / (Bgreen + BSWIR-1) | ||
SNN[25,26] | Sum457 = BNir + BSWIR-1 + BSWIR-2 ND5723 = [(BSWIR-1 + BSWIR-2) - (Bgreen + Bred)] / [(BSWIR-1 + BSWIR-2) + (Bgreen + Bred)] ND571 = [(BSWIR-1 + BSWIR-2) - Bblue] / [(BSWIR-1 + BSWIR-2) + Bblue] | ||
AWEIsh[27] | AWEIsh = Bblue + 2.5 × Bgreen - 1.5 × (BNir + BSWIR-1) - 0.25 × BSWIR-2 | ||
AWEInsh[27] | AWEInsh = 4 × (Bgreen - BSWIR-1) - (0.25 × BNir + 2.75 × BSWIR-1) | ||
NDWI+ VI[28] | EVI = 2.5 × (BNir - Bred) / (BNir + 6.0 × Bred - 7.5 × Bblue +1) NDVI = (BNir - Bred) / (BNir + Bred) LSWI = (BNir - BSWIR-1) / (BNir + BSWIR-1) mNDWI=(Bgreen - BSWIR-1) / (Bgreen + BSWIR-1) NDWI=(Bgreen - BNir) / (Bgreen + BNir) | ||
mNDWI+VI[28] | |||
LSWI + VI[28] | |||
算法 | 阈值选取 | 研究区 | 总体精度/% |
TCWCrist[19,20] | TCWCrist> 0 | 北卡罗来纳州 | |
NDWI[21] | NDWI > 0 | 内布拉斯加州 | |
mNDWI[22,23] | mNDWI > 0 | 厦门市 | 99.85 |
SNN[25,26] | (Sum457 < 0.188)或(ND5723 < -0.457)或(ND571 < 0.04)或(Sum457 < 0.269且ND5723 < -0.234且ND571 < 0.40) | 密苏里州 | 96.00 |
AWEIsh[27] AWEInsh[27] | AWEIsh > 0 AWEInsh > 0 | 丹麦、瑞士、新西兰、 埃塞俄比亚、南非 | 93.00~98.00 |
NDWI+ VI[28] mNDWI+VI[28] LSWI + VI[28] | EVI<0.1且(NDWI > NDVI或NDWI > EVI) EVI<0.1且(mNDWI > NDVI或mNDWI > EVI) EVI<0.1且(LSWI> NDVI或LSWI > EVI) | 佛罗里达州、 拉斯维加斯 | 92.23~99.1294.34~99.5893.60~99.52 |
表2 常用遥感卫星用于水体提取的若干典型波段波长(μm)与空间分辨率(m)Tab. 2 Wavelengths and spatial resolutions of several typical bands for open surface water mapping from commonly used remote sensing satellites |
蓝光 | 绿光 | 红光 | 近红外 | 短波红外-1 | 短波红外-2 | |||
---|---|---|---|---|---|---|---|---|
低空间分辨率传感器 | ||||||||
MODIS | 0.46~0.48/500 | 0.55~0.57/500 | 0.62~0.67/250 | 0.84~0.88/250 | 1.63~1.65/500 | 2.11~2.14/500 | ||
NOAA/AVHRR | 0.55~0.68/1100 | 0.73~1.10/1100 | ||||||
Suomi NPP-VIIRS | 0.49/750 | 0.56/750 | 0.64/370 | 0.87/370 | 1.61/750 | 2.25/750 | ||
MERIS | 0.49/300 | 0.56/300 | 0.67/300 | 0.87/300 | ||||
中等空间分辨率传感器 | ||||||||
Landsat 1-5 MSS | 0.50~0.60/80 | 0.60~0.70/80 | 0.70~0.80/80 | 0.80~1.10/80 | ||||
Landsat 4-5 TM | 0.45~0.52/30 | 0.52~0.60/30 | 0.63~0.69/30 | 0.76~0.90/30 | 1.55~1.75/30 | 2.08~2.35/30 | ||
Landsat 7 ETM+ | 0.45~0.52/30 | 0.52~0.60/30 | 0.63~0.69/30 | 0.76~0.90/30 | 1.55~1.75/30 | 2.08~2.35/30 | ||
Landsat 8 OLI | 0.45~0.51/30 | 0.53~0.59/30 | 0.64~0.67/30 | 0.85~0.88/30 | 1.57~1.65/30 | 2.11~2.29/30 | ||
Sentinel-2 MSI | 0.46~0.52/10 | 0.55~0.58/10 | 0.64~0.67/10 | 0.78~0.90/10 | 1.57~1.65/20 | 2.10~2.28/20 | ||
SPOT 1-3 | 0.50~0.59/20 | 0.61~0.68/20 | 0.79~0.89/20 | |||||
SPOT 4 | 0.50~0.59/20 | 0.61~0.68/20 | 0.78~0.89/20 | 1.58~1.78/20 | ||||
SPOT 5 | 0.49~0.61/10 | 0.61~0.68/10 | 0.78~0.89/10 | 1.58~1.78/20 | ||||
SPOT 6-7 | 0.46~0.53/6 | 0.53~0.59/6 | 0.63~0.70/6 | 0.76~0.89/6 | ||||
ASTER | 0.52~0.60/15 | 0.63~0.69/15 | 0.76~0.86/15 | 1.60~1.70/30 | 2.15~2.43/30 | |||
HJ-1A/B | 0.43~0.52/30 | 0.52~0.60/30 | 0.63~0.69/30 | 0.76~0.90/30 | 1.55~1.75/150 | |||
高空间分辨率传感器 | ||||||||
RapidEye | 0.44~0.51/5 | 0.52~0.59/5 | 0.63~0.69/5 | 0.76~0.85/5 | ||||
IKONOS | 0.45~0.53/4 | 0.52~0.61/4 | 0.64~0.72/4 | 0.77~0.88/4 | ||||
Quickbird | 0.45~0.52/2.44~2.88 | 0.52~0.60/2.44~2.88 | 0.63~0.69/2.44~2.88 | 0.76~0.90/2.44~2.88 | ||||
Worldview 2 | 0.45~0.51/0.46 | 0.51~0.58/0.46 | 0.63~0.69/0.46 | 0.77~0.90/0.46 | ||||
Worldview 3 | 0.45~0.51/0.31 | 0.51~0.58/0.31 | 0.63~0.69/0.31 | 0.77~0.90/0.31 | 1.64~1.68/0.31 | 2.15~2.29/0.31 | ||
Worldview 4 | 0.45~0.51/0.31 | 0.51~0.58/0.31 | 0.66~0.69/0.31 | 0.78~0.92/0.31 | 1.64~1.68/0.31 | 2.15~2.37/0.31 | ||
高分1号 | 0.45~0.52/8 | 0.52~0.59/8 | 0.63~0.69/8 | 0.77~0.89/8 | ||||
高分2号 | 0.45~0.52/4 | 0.52~0.59/4 | 0.63~0.69/4 | 0.77~0.89/4 | ||||
资源3号 | 0.45~0.52/6 | 0.52~0.59/6 | 0.63~0.69/6 | 0.77~0.89/6 |
注:“/”前面的数值为波长;“/”后的数值为空间分辨率。 |
表3 全球地表水体覆盖数据集Tab. 3 Datasets of global surface water body |
数据产品 | 空间范围 | 空间分辨率 | 时间跨度 |
---|---|---|---|
Global Inundation Extent from Multi-Satellites (GIEMS)[51] | 全球 | 0.25 ° | 1993-2007 |
Global Raster Water Mask at 250 meter Spatial Resolution[52] | 全球 | 250 m | 2000-2002 |
GLOWABO[53] | 全球 | 30 m | 2000 |
Global Land 30-water[54] | 全球 | 30 m | 2000, 2010 |
Global 3 arc-second Water Body Map (G3WBM)[55] | 全球 | 90 m | 2010 |
Global water cover map in 2013[56] | 全球 | 500 m | 2013 |
GLCF GIW[57] | 全球 | 30 m | 2000 |
Joint Research Centre (JRC)[11] | 全球 | 30 m | 1984-2015 |
500 m 8-day Water Classification Maps[58] | 全球 | 500 m | 2000-2015 |
500 m Resolution Daily Global Surface Water Change Database[49] | 全球 | 500 m | 2001-2016 |
[1] |
饶萍, 王建力 . 最优分区与最优指数联合的水体信息提取[J]. 地球信息科学学报, 2017,19(5):702-712.
[
|
[2] |
梁立, 刘庆生, 刘高焕 等. 基于遥感影像的海岸线提取方法综述[J]. 地球信息科学学报, 2018,20(12):1745-1755.
[
|
[3] |
李艳华, 丁建丽, 闫人华 . 基于国产GF-1遥感影像的山区细小水体提取方法研究[J]. 资源科学, 2015,37(2):408-416.
[
|
[4] |
张超, 彭道黎 . 基于遥感的水体信息提取技术研究进展[J]. 河南农业科学, 2013,42(6):16-20.
[
|
[5] |
叶庆华, 程维明, 赵永利 等. 青藏高原冰川变化遥感监测研究综述[J]. 地球信息科学学报, 2016,18(7):920-930.
[
|
[6] |
|
[7] |
|
[8] |
李丹, 吴保生, 陈博伟 等. 基于卫星遥感的水体信息提取研究进展与展望[J]. 清华大学学报, 2019,59(22):1-25.
[
|
[9] |
陈前, 郑利娟, 李小娟 等. 基于深度学习的高分遥感影像水体提取模型研究[J]. 地理与地理信息科学, 2019,35(4):43-49.
[
|
[10] |
林楚彬, 李少青 . Landsat数据水体自动提取研究进展[J]. 自动化技术与应用, 2015,34(2):1-4.
[
|
[11] |
|
[12] |
|
[13] |
|
[14] |
|
[15] |
陈文倩, 丁建丽, 李艳华 , 等. 基于国产GF-1遥感影像的水体提取方法[J]. 资源科学, 2015,37(6):1166-1172.
[
|
[16] |
王永丽, 宋希磊 . 基于ETM+数据的水体信息提取[J].绿色科技, 2017(10):208-210.
[
|
[17] |
周成虎, 骆剑承, 杨晓梅 , 等. 遥感影像地学理解与分析[M]. 北京: 科学出版社, 1999.
[
|
[18] |
张明华 . 用改进的谱间关系模型提取极高山地区水体信息[J]. 地理与地理信息科学, 2008,24(2):14-16.
[
|
[19] |
|
[20] |
|
[21] |
|
[22] |
|
[23] |
徐涵秋 . 利用改进的归一化差异水体指数(MNDWI)提取水体信息的研究[J]. 遥感学报, 2005,9(5):589-595.
[
|
[24] |
车向红, 冯敏, 姜浩 等. 2000-2013年青藏高原湖泊面积MODIS遥感监测分析[J]. 地球信息科学学报, 2015,17(1):99-107.
[
|
[25] |
|
[26] |
|
[27] |
|
[28] |
|
[29] |
杨浩, 孟娜, 王婧 , 等. 基于支持向量机的京津冀城市群热环境时空形态模拟[J]. 地球信息科学学报, 2019,21(2):190-200.
[
|
[30] |
李楠, 朱秀芳, 潘耀忠 , 等. 人工蜂群算法优化的SVM遥感影像分类[J]. 遥感学报, 2018,22(4):29-39.
[
|
[31] |
|
[32] |
|
[33] |
|
[34] |
闫东阳 . 基于对象的随机森林遥感分类方法优化[D]. 北京:中国地质大学(北京), 2018.
[
|
[35] |
饶品增, 蒋卫国, 王晓雅 等. 基于MODIS数据的洪涝灾害分析——以2017年洞庭湖区洪水为例[J]. 灾害学, 2019,34(1):203-207.
[
|
[36] |
赵书荟, 段会川, 高帅 , 等. 基于随机森林的MODIS遥感影像水体分类研究[J]. 山东师范大学学报, 2018,33(1):19-25.
[
|
[37] |
|
[38] |
|
[39] |
赵德刚, 占玉林, 刘成林 , 等. 面向土地覆盖分类的MODIS影像融合研究[J]. 遥感技术与应用, 2009,24(6):827-831.
[
|
[40] |
龚燃 . 哨兵-2A光学成像卫星发射升空[J].国际太空, 2015(8):36-40.
[
|
[41] |
|
[42] |
|
[43] |
沈占锋, 李均力, 于新菊 . 基于协同计算的白洋淀湿地时序水体信息提取[J]. 地球信息科学学报, 2016,18(5):690-698.
[
|
[44] |
|
[45] |
|
[46] |
|
[47] |
|
[48] |
|
[49] |
|
[50] |
|
[51] |
|
[52] |
|
[53] |
|
[54] |
|
[55] |
|
[56] |
|
[57] |
|
[58] |
|
[59] |
|
[60] |
|
[61] |
|
/
〈 |
|
〉 |