Have a personal or library account? Click to login
Tracking site-scale tidal flat dynamics using time-series satellite data and Google Earth Engine over the past 40 years Cover

Tracking site-scale tidal flat dynamics using time-series satellite data and Google Earth Engine over the past 40 years

By: Maham Arif,  Qianguo Xing and  Jie Li  
Open Access
|Dec 2025

Figures & Tables

Figure 1

Geographical locations of study area. (A) KPT, Karachi, Pakistan. (B) Yangma Island, Yantai China. KPT, Karachi Port Trust.
Geographical locations of study area. (A) KPT, Karachi, Pakistan. (B) Yangma Island, Yantai China. KPT, Karachi Port Trust.

Figure 2

Distribution of sample points. True color composite image of Landsat for both study areas.
Distribution of sample points. True color composite image of Landsat for both study areas.

Figure 3

Average spectral. Curves of four land cover types by using the lowest tide images of the Landsat series. The term ‘tidal flat’ refers to a fully exposed TF pixel; ‘water’ refers to both inland and seawater; ‘vegetation’ refers to both inland and intertidal vegetation; and ‘urban’ refers to all inland buildings. TF, tidal flat.
Average spectral. Curves of four land cover types by using the lowest tide images of the Landsat series. The term ‘tidal flat’ refers to a fully exposed TF pixel; ‘water’ refers to both inland and seawater; ‘vegetation’ refers to both inland and intertidal vegetation; and ‘urban’ refers to all inland buildings. TF, tidal flat.

Figure 4

Histograms of different land covers for high and low water extent. Histograms representing various land cover classes under high-tide (A and B) and low-tide (C and D) composite conditions.
Histograms of different land covers for high and low water extent. Histograms representing various land cover classes under high-tide (A and B) and low-tide (C and D) composite conditions.

Figure 5

Flow chart of TF extraction.
Flow chart of TF extraction.

Figure 6

Spatial analysis and rates of area change (ha · y−1) of TF from 1986 to 2024, for KPT, Karachi, and Yangma Island, Yantai, respectively. (A and C) Spatial analysis (B and D) Rates of area change (ha · y−1).
Spatial analysis and rates of area change (ha · y−1) of TF from 1986 to 2024, for KPT, Karachi, and Yangma Island, Yantai, respectively. (A and C) Spatial analysis (B and D) Rates of area change (ha · y−1).

Figure 7

TF maps in KPT, Pakistan, from 1986 to 2024. KPT, Karachi Port Trust; TF, tidal flat.
TF maps in KPT, Pakistan, from 1986 to 2024. KPT, Karachi Port Trust; TF, tidal flat.

Figure 8

TF maps in Yangma Island, Yantai, China, from 1986 to 2024. TF, tidal flat.
TF maps in Yangma Island, Yantai, China, from 1986 to 2024. TF, tidal flat.

Figure 9

Frequency map of TFs for 11 composite raster’s from 1986 to 2024. (A and B) Frequency map of KPT and Yangma Island, respectively. (C) Sentinel-2 images (2024) for major sites of frequency change in TFs for KPT. (D) Frequency map of TFs for the major sites.
Frequency map of TFs for 11 composite raster’s from 1986 to 2024. (A and B) Frequency map of KPT and Yangma Island, respectively. (C) Sentinel-2 images (2024) for major sites of frequency change in TFs for KPT. (D) Frequency map of TFs for the major sites.

Figure 10

Area and trend line analysis by Sentinel-2 data for 2017–2024. (A) For KPT. (B) For Yangma Island, Yantai.
Area and trend line analysis by Sentinel-2 data for 2017–2024. (A) For KPT. (B) For Yangma Island, Yantai.

Figure 11

Sentinel-2 satellite imagery of the study areas. (A) TF for 2017 in KPT. (B) TF for 2017 in Yangma Island, Yantai. KPT, Karachi Port Trust; TF, tidal flat.
Sentinel-2 satellite imagery of the study areas. (A) TF for 2017 in KPT. (B) TF for 2017 in Yangma Island, Yantai. KPT, Karachi Port Trust; TF, tidal flat.

Figure 12

Comparative analysis. (A) KPT, Pakistan, and (B) Yangma Island, China. For KPT, the comparison utilizes the UQD dataset (Murray et al., 2019) for 1989 and 2010. For Yangma Island, the comparison is between the UQD dataset for 1989 and the WF_TF dataset (Wang et al., 2020) for 2016. The red-highlighted area is the most noticeable area of change. KPT, Karachi Port Trust; TF, tidal flat; UQD, University of Queensland Dataset.
Comparative analysis. (A) KPT, Pakistan, and (B) Yangma Island, China. For KPT, the comparison utilizes the UQD dataset (Murray et al., 2019) for 1989 and 2010. For Yangma Island, the comparison is between the UQD dataset for 1989 and the WF_TF dataset (Wang et al., 2020) for 2016. The red-highlighted area is the most noticeable area of change. KPT, Karachi Port Trust; TF, tidal flat; UQD, University of Queensland Dataset.

Figure S1

Data collection. (A) Temporal distribution of Landsat images sensors (TM, OLI) and number of good quality images from 1984 to 2024; (B) Sentinel-2 images and number of good-quality from 2017 to 2024. OLI, operational land imager; TM, thematic mapper.
Data collection. (A) Temporal distribution of Landsat images sensors (TM, OLI) and number of good quality images from 1984 to 2024; (B) Sentinel-2 images and number of good-quality from 2017 to 2024. OLI, operational land imager; TM, thematic mapper.

Figure S2

High tide image composite and low tide image composite images, respectively. (A) Composite by mNDWI using MVCAmax. (B) Composite by NDWI_p10 using MQCAmin. mNDWI, modified normalized difference water index; MQCAmin, minimum quantile composite algorithm; MVCAmax, maximum value composite algorithm; NDWI, normalized difference water index.
High tide image composite and low tide image composite images, respectively. (A) Composite by mNDWI using MVCAmax. (B) Composite by NDWI_p10 using MQCAmin. mNDWI, modified normalized difference water index; MQCAmin, minimum quantile composite algorithm; MVCAmax, maximum value composite algorithm; NDWI, normalized difference water index.

Figure S3

Otsu threshold application on spectral index histograms for composite (2022–2024). (A) mNDWI composite histogram for high-tide conditions (MVCAmax). (B) NDWI 10th percentile composite histogram for low-tide conditions (MQCAmin).
Otsu threshold application on spectral index histograms for composite (2022–2024). (A) mNDWI composite histogram for high-tide conditions (MVCAmax). (B) NDWI 10th percentile composite histogram for low-tide conditions (MQCAmin).

Figure S4

Methodology steps. The top row shows the initial binary water masks derived from the maximum mNDWI (left) and the 10th percentile NDWI (right). The middle row displays the results after post processing. Last shows extents before and after refinement. mNDWI, modified normalized difference water index; NDWI, normalized difference water index.
Methodology steps. The top row shows the initial binary water masks derived from the maximum mNDWI (left) and the 10th percentile NDWI (right). The middle row displays the results after post processing. Last shows extents before and after refinement. mNDWI, modified normalized difference water index; NDWI, normalized difference water index.

Figure S5

Landsat TM images for Yangma Island. TM, thematic mapper.
Landsat TM images for Yangma Island. TM, thematic mapper.

Figure S6

Frequency distribution of TFs. (A) Chinna Creek (1986–2024) shows widespread persistence; (B) Karachi Harbour (2008–2024) highlights dominant low-frequency zones linked to human disturbance. TFs, tidal flats.
Frequency distribution of TFs. (A) Chinna Creek (1986–2024) shows widespread persistence; (B) Karachi Harbour (2008–2024) highlights dominant low-frequency zones linked to human disturbance. TFs, tidal flats.

Temporal variation of TFs area with CI error (ha)

Year_RangeKPT_AreaKPT_CI_ErrorYear_RangeYangma_AreaYangma_CI_Error
1984–19861984–1986972.0445.39
1986–19881678.8178.41887–1989517.0624.15
1989–19911651.0777.11990–1992450.8321.05
1992–19941226.5057.281993–1995415.4419.4
1995–19971146.2553.531996–1998383.2317.9
1998–20001136.9753.11999–2001308.7414.42
2001–2003947.2644.242002–2004308.9014.43
2004–20072005–2007372.5317.4
2008–20101338.5162.512008–2010351.6416.42
2011–20122011–2012
2013–20151344.8462.82013–2015298.0413.92
2016–20181396.8065.232016–2018299.0113.96
2019–20211394.7665.142019–2021258.2512.06
2022–20241408.1765.762022–2024290.1213.55

Formulas of the spectral indices used in this study

1mNDWI= Green  SWIR1  Green + SWIR1 {\rm{mNDWI}} = {{{\rm{ Green }} - {\rm{ SWIR1 }}} \over {{\rm{ Green }} + {\rm{ SWIR1 }}}}
2NDWI= Green NIR Green +NIR{\rm{NDWI}} = {{{\rm{ Green }} - {\rm{NIR}}} \over {{\rm{ Green }} + {\rm{NIR}}}}
3NDVI=NIRRedNIR+Red{\rm{NDVI}} = {{{\rm{NIR}} - {\rm{Red}}} \over {{\rm{NIR}} + {\rm{Red}}}}

Confusion matrix of TF validation

ClassTFNon-TFTotalUA
Map pixelsTF2341625093.6%
Non-TF1019920995.2%
Total 244215459OA = 94.3%
PA 96.0%92.6% Kappa = 0.88

Annual TF area with 95% CIs (2017–2024)

YearKPT areaKPT CI errorYangma areaYangma CI error
20171492.2369.69291.9313.63
20181624.9275.88299.4813.99
20191605.0674.96401.0118.73
20201227.0157.3438.3420.47
20211489.7569.57232.2210.84
20221498.9470.0351.3216.41
20231236.2457.73278.813.02
20241492.2869.69290.9113.59
DOI: https://doi.org/10.26881/oahs-2025.1.28 | Journal eISSN: 1897-3191 | Journal ISSN: 1730-413X
Language: English
Page range: 328 - 352
Submitted on: Oct 28, 2025
|
Accepted on: Nov 28, 2025
|
Published on: Dec 31, 2025
Published by: University of Gdańsk
In partnership with: Paradigm Publishing Services
Publication frequency: 4 issues per year

© 2025 Maham Arif, Qianguo Xing, Jie Li, published by University of Gdańsk
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.