the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Spatio-Temporal Changes in China’s Mainland Shorelines Over 30 Years Using Landsat Time Series Data (1990–2019)
Abstract. Continuous monitoring of shoreline dynamics is essential to understanding the drivers of shoreline changes and evolution. A long-term shoreline dataset can describe the dynamic changes in the spatio-temporal dimension and provide information on the influence of anthropogenic activities and natural factors on coastal areas. This study, conducted on the Google Earth Engine platform, analyzed the spatio-temporal evolution characteristics of China’s shorelines, including those of Hainan and Taiwan, from 1990 to 2019 using long time series of Landsat TM/ETM+/OLI images. First, we constructed a time series of the Modified Normalized Difference Water Index (MNDWI) with high-quality reconstruction by the harmonic analysis of time series (HANTS) algorithm. Second, the Otsu algorithm was used to separate land and water of coastal areas based on MNDWI value at high tide levels. Finally, a 30-year shoreline dataset was generated and a shoreline change analysis was conducted to characterize length change, area change, and rate of change. We concluded the following: (1) China’s shoreline has shown an increasing trend in the past 30 years, with varying growth patterns across regions; the total shoreline length increased from 24905.55 km in 1990 to 25391.34 km in 2019, with a total increase greater than 485.78 km, a rate of increase of 1.95 %, and an average annual increasing rate of 0.07 %; (3) the most visible expansion has taken place in Tianjin, Hangzhou Bay, and Zhuhai for the three economically developed regions of the Bohai Bay-Yellow River Estuary Zone (BHBYREZ), the Yangtze River Estuary-Hangzhou Bay Zone (YRE-HZBZ) and the Pearl River Estuary Zone (PREZ), respectively. The statistics of shoreline change rate for the three economically developed regions show that the average end point rates (EPR) were 43.59 m/a, 39.10 m/a, and 13.42 m/a, and the average linear regression rates (LRR) were 57.40 m/a, 43.85 m/a, and 10.11 m/a, respectively. This study presents an innovative and up-to-date dataset and comprehensive information on the status of China’s shoreline from 1990 to 2019, contributing to related research and policy implementation, especially in support of sustainable development.
- Preprint
(2084 KB) - Metadata XML
-
Supplement
(11186 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on essd-2024-123', Anonymous Referee #1, 20 Jun 2024
This research produced the first 30-year spatio-temporal change analysis of China’s mainland shoreline based on the time series data of Landsat images from 1990 to 2019 obtained from GEE platform. The datasets are widely covered and complete. The experiments and result analysis are interesting and sufficient. The discussions and conclusions are effective. However, there are some points needs to be clarified before accepted.
Major points:
1.In the shoreline extraction step, the classic threshold segmentation method Otsu algorithm is used to segment the grayscale MNDWI images into water bodies and non-water bodies. And then, to extract water body, pixels corresponding to lakes and reservoirs are removed by geographic distribution and area sizes. In Line 175 of the manuscript, this study employed an area parameter and select the largest water body for each object to effectively eliminate interference caused by terrestrial water bodies. As well known, Otsu algorithm is only a thresholding algorithm, which only can segment image to isolate label points. However, to extract water bodies, connected segmentation regions are necessary. The thresholding algorithm and the post-processing step is too simple. We doubt the accuracies and robustness of the extracted shoreline.
Minor points:
1.In Line 190, the symbols in equations (2) and (3) are not defined, such as , , .
2.In Line 205, What are EPR and LPR? We can find the full name in Abstract part but not in the method.
3.The labels of subfigures are confusion. In most of Figures, the labels of subfigures are denoted as (a), (b), (c), (d). But in Figure 7 and Figure 10, the labels of subfigures are denoted as a, b, c. In Figure 14, A, B, C are used.
4.In Line 170, “grayscale MNDWI binary images” should be “grayscale MNDWI images” we think.
Citation: https://doi.org/10.5194/essd-2024-123-RC1 -
RC2: 'Comment on essd-2024-123', Anonymous Referee #2, 25 Jul 2024
General comments
I thank the authors for their paper presenting a new method for mapping coastal change using satellite remote sensing data over large spatial extents. The approach presented here is scientifically sound, and I believe it will be of interest to readers of Earth System Science Data with an interest in large-scale coastal mapping. However, I believe there are several major and minor areas where the paper should be improved prior to publication.
My primary critism of the manuscript is its reliance on validating modelled remote sensing-based coastal change results against other modelled remote sensing-based coastline datasets. This particularly applies to the use of GSV and Coastline_ECS, which are both also Landsat-based shoreline mapping datasets. This comparison does not effectively verify the accuracy of the data being presented in this study: it serves more as a test of "consistency" with previous approaches (with consistency not necessarily being a good thing if these previous datasets were inaccurate themselves) rather than a "validation". I feel the paper would strongly benefit from additional validation comparing the results here to real-world validation data (e.g. beach surveys etc) at least a number of coastal sites, providing additional confidence that this study is indeed producing accurate results and not simply re-producing (potentially inaccurate) existing datasets. In addition, caveats and limitations of comparing modelled results against other modelled datasets should be discussed in detail in the paper.
The paper also uses shoreline length as a key metric for comparing coastal change over time. Shoreline length is a notoriously problematic metric, being essentially unmeasurable and scale-dependent due to the "Coastline paradox", and highly influenced by noise which can be variable over time or between different satellite sensors (e.g. Landsat 5 vs Landsat 8). While I would strongly advice the authors choose another metric for comparing coastal change over time, if they wish to continue using shoreline length the limitations of this metric should be discussed and documented clearly in the paper.
Finally, the current Discussion section feels very brief and poorly referenced. I believe a significant amount of material currently contained in the Results section could be moved to discussion, and the existing Discussion material could cite and discuss existing literature in more detail. I have also suggested a number of areas below where limitations and caveats of the proposed method could be discussed to allow readers to gain a more informed understanding of the advantages and limitations of the approach.
Specific comments
- Lines 130: As the authors recognise, accounting for tide in large-scale coastal remote sensing analyses is critical. However, the current manuscript does not provide sufficient detail about how Landsat imagery was filtered by tide. In particular, "based on high tide times" should be replaced with specifics about how these high tide images were selected (e.g. tide height threshold? top X percent of tides etc?).
- Lines 130: In additional, more detail should be provided about how point tide gauge locations were mapped to continuous coverage satellite imagery. Were tide heights interpolated to each image, or assigned based on the closest tide gauge? How did the authors ensure that tides observed at this small number of locations (17) were representative and applicable to satellite imagery away from these gauges, particularly in areas of complex tide dynamics or in areas located far from the nearest tide gauge? This ideally would include some discussion around alternative approaches used for accounting for tide in complex coastal environments (e.g. the use of global ocean tide modelling; Vos et al. 2019, Bishop-Taylor et al. 2021).
- Line 160: The HANTS method presented here sounds very promising for a tool for handling noisy/sparse remote sensing time series. So that readers can appreciate how this approach works, please provide an additional figure demonstrating the HANTS approach being applied to several example pixels from this study (e.g. showing the effect of smoothing and gap filling).
- Line 172: Was OTSU thresholding applied to each annual timestep individually, producing different thresholds for each year? Or was a consistent threshold derived and applied across the entire time series?
- Line 190: Did this offset calculation include directionality? (e.g. bias on the inland or seaward directions)
- Section 4.3.2: This section currently goes into a little too much locally-specific detail for a journal with global readership - would suggest simplifying it and removing some of the current content. In addition, some of this material feels like it would more appropriately belong in Discussion instead of Results.
- Line 398: What is a "reconstruction" based threshold? Please clarify or use this term more consistently throughout the manuscript.
- Line 450: Based on the current validation, I don't think the statement "more accurate shoreline data compared to previous global shoreline datasets" can be justified, given that those global datasets were themselves used as a point of truth in the validation. Perhaps this could be justified if results of this study and those global datasets could all be compared to real-world, independent validation data.
Technical corrections
- Line 405: Should this read "20 yearly observations"?
Data comments- The transect point and line features are currently split into 7 individual features ("*_1.shp", "*_2.shp" etc). These would be much easier to use if these individual files were combined, so that users could analyse them at once without having to combine them manually first.
- Similiarly, the shoreline datasets would be easier to use if all years were combined into a single shapefile with a "year" attribute column.
Citation: https://doi.org/10.5194/essd-2024-123-RC2
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
323 | 66 | 19 | 408 | 32 | 18 | 16 |
- HTML: 323
- PDF: 66
- XML: 19
- Total: 408
- Supplement: 32
- BibTeX: 18
- EndNote: 16
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1