Sentinel-2 mapping of a turbid intertidal seagrass meadow in Southern Vietnam

Figures & data

Figure 1. The study area, with captured UAV image on the background of a true colour image from high resolution satellite, with ground truth points and their elevation in meters.

Table 1. List of images used for the analysis.

Index	Date of acquisition	Sensing orbit number	Sensing orbit direction	Spacecraft name	Cloud free area	Estimated tide level (m above mean sea level)
1	2019-11-12	118	DESCENDING	Sentinel-2B	100%	0.164
2	2019-11-15	18	DESCENDING	Sentinel-2B	100%	0.208
3	2019-11-17	118	DESCENDING	Sentinel-2A	87%	0.294
4	2019-11-22	118	DESCENDING	Sentinel-2B	100%	0.211
5	2019-11-27	118	DESCENDING	Sentinel-2A	100%	0.187
6	2019-11-30	18	DESCENDING	Sentinel-2A	100%	0.306
7	2019-12-02	118	DESCENDING	Sentinel-2B	100%	0.289
8	2019-12-07	118	DESCENDING	Sentinel-2A	100%	0.174
9	2019-12-10	18	DESCENDING	Sentinel-2A	100%	0.153
10	2019-12-15	18	DESCENDING	Sentinel-2B	100%	0.258
11	2019-12-20	18	DESCENDING	Sentinel-2A	97%	0.175
12	2019-12-22	118	DESCENDING	Sentinel-2B	100%	0.088
13	2019-12-25	18	DESCENDING	Sentinel-2B	98%	0.139
14	2020-01-01	118	DESCENDING	Sentinel-2B	100%	0.179
15	2020-01-04	18	DESCENDING	Sentinel-2B	100%	0.103
16	2020-01-06	118	DESCENDING	Sentinel-2A	100%	0.085
17	2020-01-09	18	DESCENDING	Sentinel-2A	100%	0.107
18	2020-01-19	18	DESCENDING	Sentinel-2A	100%	0.009
19	2020-01-21	118	DESCENDING	Sentinel-2B	100%	0.011
20	2020-01-24	18	DESCENDING	Sentinel-2B	100%	0.143
21	2020-01-26	118	DESCENDING	Sentinel-2A	99%	0.187
22	2020-02-05	118	DESCENDING	Sentinel-2A	100%	−0.015
23	2020-02-13	18	DESCENDING	Sentinel-2B	100%	0.072
24	2020-02-15	118	DESCENDING	Sentinel-2A	100%	−0.045
25	2020-02-18	18	DESCENDING	Sentinel-2A	85%	−0.099
26	2020-02-20	118	DESCENDING	Sentinel-2B	100%	−0.018
27	2020-02-23	18	DESCENDING	Sentinel-2B	100%	0.095
28	2020-02-25	118	DESCENDING	Sentinel-2A	100%	0.081
29	2020-03-11	118	DESCENDING	Sentinel-2B	100%	0.055
30	2020-03-14	18	DESCENDING	Sentinel-2B	100%	−0.124
31	2020-03-19	18	DESCENDING	Sentinel-2A	100%	−0.136
32	2020-03-21	118	DESCENDING	Sentinel-2B	100%	−0.036
33	2020-03-31	118	DESCENDING	Sentinel-2B	100%	−0.255

Table 5. Correlation between classification accuracy and tide levels.

Pearson correlation between overall accuracy of images without water column correction and tide level	Pearson correlation between overall accuracy of images with water column correction and tide level
−0.351 (p = 0.109)	−0.378 (p = 0.083)

Figure 2. The overall flow chart of image analysis.

Figure 3. The orthomosaic of the UAV images.

Figure 4. (a) Close ups on the UAV image, (b) object-based segmentation result, and (c) classification result.

Table 2. List of indices used in this paper.

Index	Equation	Usage
Normalized Difference Vegetation Index (Tucker 1979)	$\mathit{NDVI}= \frac{\mathit{NIR}-\mathit{Red}}{\mathit{NIR}+\mathit{Red}}$	Identifying emerged vegetation
Normalized Difference Water Index (McFeeters 1996)	$\mathit{NDWI}= \frac{\mathit{Green}-\mathit{NIR}}{\mathit{Green}+\mathit{NIR}}$	Identifying water bodies
Modified Normalized Difference Water Index (Xu 2006)	$\mathit{mNDWI}= \frac{\mathit{Green}-\mathit{SWIR}1}{\mathit{Green}+\mathit{SWIR}2}$	Identifying water bodies
Depth Invariant Index (Lyzenga 1981)	${\mathit{DII}}_{ij}=\ln \hspace{0.17em}\left(L_i-L_{si}\right) - \left[\left(\frac{k_i}{k_j}\right) \times \ln \hspace{0.17em}\left(L_j-L_{sj}\right)\right]$	Adjustment for bottom reflectance, explained in detail below.
Percentile (p[i])	The value of the ith percentile in the list	Describing the variation of values within the list
Interval mean (intMn[ij])	The mean of values between the ith percentile and j percentile.	Describing the variation of values within the list

Table 3. Theil Sen slope and coefficient of determination (in brackets) between the natural log of blue, green, red, nir, and swir1 against tide levels for Dense seagrass, Sparse seagrass, and Sand.

	Dense seagrass	Sparse seagrass	Sand
lnblue	−0.954 (0.152)	−1.063 (0.193)	-1.674 (0.476)
lngreen	−0.857 (0.121)	−0.866 (0.151)	-1.625 (0.504)
lnred	−0.998 (0.085)	−0.966 (0.091)	-1.785 (0.422)
lnnir	−1.152 (0.099)	−1.150 (0.105)	−1.830 (0.354)
lnswir1	−1.787 (0.146)	−1.825 (0.152)	−1.948 (0.208)
lnswir2	−1.727 (0.145)	−1.712 (0.146)	−2.011 (0.212)

Figure 5. (a) The classification of the orthomosaic, and (b) the downsampled classification result of the UAV images.

Table 4. Table of inputs for classification.

Image	Inputs for classification
Single images without water column correction	Six reflectance bands (blue, green, red, near infrared band, and shortwave infrared 1 and shortwave infrared 2)
Single images with water column correction	The above six reflectance bands and three Depth Invariant Indices: DII_BG, DII_BR, DII_GR
The multitemporal composite without water column correction	The 17 covariates for the above six reflectance bands and three multi-spectral indices (NDVI, NDWI, mNDWI)
The multitemporal composite with water column correction	The 17 covariates for each of the above six reflectance bands, three multi-spectral indices (NDVI, NDWI, mNDWI), and three Depth Invariant Indices.

Figure 6. Tide levels at image acquisition time (10:20 am local time) for the Sentinel-2 images analysed.

Figure 7. True colour composite and classified images from Sentinel-2 satellites at high and low tides. (a), (b), and (c) show the true colour composite. (d), (e), (f) show the classification result without water column correction and (g), (h), (i) show the classification result with water column correction. (a), (d), (g) are captured at low tide, (b), (e), (h) at medium tide and (c), (f), (i) at high tide.

Figure 8. Logarithms of reflectance of the blue, green, red, near infrared, short-wave infrared 1 and short-wave infrared 2 of Dense seagrass, Sparse seagrass, and Sand at different tide levels. Red lines show the regression slope of a Theil Sen Robust regression.

Figure 9. Overall accuracy of classification results of seagrass at different tide levels for images with and without water column correction.

Figure 11. Comparison between the classification accuracy of images with and without water column corrections, and between classifying single images and the multitemporal composite.

Tucker CJ. 1979. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sens Environ. 8(2):127–150.

 Web of Science ®Google Scholar

McFeeters SK. 1996. The use of the Normalized Difference Water Index (NDWI) in the delineation of open water features. Int J Remote Sens. 17(7):1425–1432.

 Web of Science ®Google Scholar

Xu H. 2006. Modification of normalised difference water index (NDWI) to enhance open water features in remotely sensed imagery. Int J Remote Sens. 27(14):3025–3033.

 Web of Science ®Google Scholar

Lyzenga DR. 1981. Remote sensing of bottom reflectance and water attenuation parameters in shallow water using aircraft and landsat data. Int J Remote Sens. 2(1):71–82.

 Google Scholar

Data availability statement

Sentinel-2 datasets were obtained from Google Earth Engine data catalogue at https://developers.google.com/earth-engine/datasets/catalog/sentinel

TPXO Tide model data were obtained from the API at https://www.worldtides.info/

The UAV images, JavaScript and Python code for processing are available on request from the author.