Mapping large-area tidal flats without the dependence on tidal elevations: A case study of Southern China

https://doi.org/10.1016/j.isprsjprs.2019.11.022Get rights and content

Abstract

Coastal ecosystems are under increased pressure from climate change and anthropogenic impacts. Tidal flat maps are necessary for managing, protecting, and restoring coastal ecosystems. Most of the existing tidal flat mapping approaches that are dependent on modelled tidal elevations, suffer high uncertainty over offshore, especially over areas where the tidal conditions have been changed by human-made structures and anthropogenic impacts. Meanwhile, these approaches separate the tidal flats in estuaries by a coastal distance rather than actual distance of saltwater intrusion. In this study, we propose a tidal flat mapping approach, using a new method to avoid the dependence on tidal elevations. The proposed approach synthesizes large-area satellite observations over a time period, to acquire the high and low tidal datum, and then classifies land and water under each tidal datum utilizing an adaptive binary classifier. The tidal flats are identified as areas being water during high tides and being land during low tides. Furthermore, the proposed approach uses the distribution of salt vegetation (i.e., mangroves) as a salinity proxy to particularly identify the tidal flats in estuaries. The proposed approach was applied for mapping the tidal flats in southern China (covering over a half of the coastline of China) utilizing Sentinel-1 Synthetic Aperture Radar (SAR) data from July 2016 to July 2018. The results achieved an overall accuracy of 92.4% and an agreement rate of 70.5% with the most recent global tidal flat map. The proposed approach does not depend on the modelled tidal elevations or large sample sets and could be easily applied for mapping tidal flats globally.

Introduction

Tidal flat is an important component of the intertidal zone that connects marine, terrestrial, and freshwater ecosystems, provides services such as shoreline stabilization, storm and wave buffering, and breeding and nursing ground (Barbier et al., 2011, Zohary and Gasith, 2014). However, tidal flats are under increased pressure from activities such as land reclamation, coastal development, sea-level rise, and river sediment reduction (Blum and Roberts, 2009, French, 2002, Syvitski et al., 2005), which result from anthropogenic impacts and climate change (Kirwan and Megonigal, 2013, Murray et al., 2014, Nicholls and Cazenave, 2010). For example, coastline hardening to protect human residential areas, results in rapid degradation of tidal flats. Many human-made structures (such as dams and reservoirs) constructed to control floods and saltwater intrusion not only reduce the sediment delivery to tidal flats, but also change the interactions between rivers and seas (Blum and Roberts, 2009, Kirwan and Megonigal, 2013). In areas experiencing rapid coastal development, the changes can be quick and wide-ranging. The sea-level rise worsens the situation (Barros et al., 2014). It might lead to conversion of tidal flats to open water (Craft et al., 2009) and result in increased saltwater intrusion into the fresh surface water (Bhuiyan and Dutta, 2012, Hong and Shen, 2012). Thus, a high-resolution and fast-updating tidal flat map that covers large areas is necessary for managing, protecting, and restoring coastal ecosystems.

To map the tidal flats, studies have been conducted based on traditional tidal stations or remote sensing (RS) imagery. The tidal stations provide valuable information on tidal datum at points, which could be interpolated to the coastline for mapping the tidal flats. However, for large areas of coastline, which are spatially undulated and varied (Maloney and Ausness, 1974, Milton and Farvar, 1968), the spatial sparsity of tidal stations due to high expenditure, as well as the difficulty in accessing such data, limit the availability of such information.

The development of RS techniques provides a large amount of imagery potentially available for mapping the intertidal zone (or tidal flats particularly) (Green et al., 1996). Currently many RS-based approaches for mapping the intertidal zone or tidal flats have been proposed. Existing RS-based approaches for mapping tidal flats can be classified into three types: terrain-based approaches, approaches based on simulated tidal elevations and RS images, and approaches based solely on RS images.

The terrain-based approaches mainly depend on digital elevation model (DEM). Such DEMs have been proposed to be built from waterlines that are extracted from images and tidal elevations at the precise time of captured images (Liu et al., 2012, Mason et al., 1995, Ryu et al., 2008), or to be built by Light Detection and Ranging (LIDAR) and Synthetic Aperture Radar (SAR) at low tides (Bell et al., 2016, Cracknell, 1999, Gade et al., 2008). Due to the scarcity of coastal DEMs with high resolution in spatial and temporal domain, the spatial variation of water levels, rapidly changing coastal topography, and high cost of airborne acquisitions (Liu et al., 2012, Murray et al., 2012), the terrain-based approaches are not applicable to large-area tidal flat mapping.

For mapping large areas of intertidal zones or tidal flats, approaches based on simulated tidal elevations and RS images were proposed. These approaches assign a simulated tidal height to each of the images covering large areas and maps of the intertidal zones or tidal flats were made by composing those images within a certain simulated tidal interval to classify them (Murray et al., 2014, Murray et al., 2012, Sagar et al., 2017). Certainly, the performance of these approaches is highly dependent on the simulation of tidal elevations over offshore regions. However, the tidal elevation simulation has uncertainties in accuracy over offshore regions (Turner et al., 2013), especially in large areas where anthropogenic impacts (such as aquacultures, harbor constructions, dams, and roads) have changed the tidal conditions. Meanwhile, in these approaches the tidal information of each image is represented by only a single value, which ignores the complexity in both tidal patterns and geographic heterogeneity. Therefore, these approaches suffer high uncertainty caused by the difference in simulated tidal elevations in large areas, especially in rapidly changing regions.

To avoid dependence on the modelled tidal elevations at the time of image acquisition, approaches based solely on RS images were proposed. Chen et al. (2016) manually selected the images with the highest or lowest shoreline to map the tidal flats of the Yangtze Estuary. The method of manually selecting images with the highest or lowest tides is laborious to be applied to large areas. Wang et al. (2018) classified each pixel from the satellite imagery during a period as water and non-water pixel, by setting a global threshold on its derived water indices and then differentiated the intertidal zone by using another global threshold on inundation frequency. The tidal flats were identified by applying a third global threshold of vegetation frequency to each pixel in intertidal zone over the period. This approach relies highly on the global thresholds, which ignores the complexity in geographic heterogeneity. In a state-of-the-art study based solely on RS images, Murray et al. (2019) mapped the tidal flats around the world using a machine learning classifier mainly based on features of statistics of the Landsat-derived water indices on each individual pixel location. The features were calculated from percentiles and interval means of Landsat observation values within a two-year observation period. Murray et al. (2019)'s approach depends on large training samples and numerous statistics on water indices. The large training samples mean a laborious work; meanwhile these numerous statistics are comparatively redundant in describing the water information. Therefore, existing ways of avoiding the dependence on the modelled tidal elevations in current approaches based solely on RS images might not work effectively for large areas.

Meanwhile, mapping tidal flats in estuaries is still comparatively ignored in the existing approaches for mapping tidal flats, although globally, tidal flats are present in estuaries and are key areas where intertidal zone connects freshwater ecosystems with the sea. Few station observation in these river-sea interaction areas is available because these areas are hardly monitored for accurate vertical tidal datum, due to the complexity of the tidal influence interfered by fresh water runoff (Maloney and Ausness, 1974) and thus are normally excluded from building tidal stations. In the existing RS-based tidal flats mapping approaches, estuaries are considered using a global threshold of coast distance. This method is over-simplified and would cause large errors in estimating the tidal flats in estuaries. In fact, the saltwater induced by tides can penetrate upstream to a few to tens of kilometers from the estuarine mouth (Garvine, 1975, Meade, 1966). Meanwhile, tidal flats in estuaries could vary because the river-sea interactions are highly dynamic due to the influences from sea-level rise, dams, river channel morphology change, etc. (Hoitink and Jay, 2016).

To overcome the above-mentioned problems in existing RS-based approaches for mapping tidal flats on a large spatial scale, we proposed an approach for mapping tidal flats over large area. For removing the dependence on tidal elevations efficiently, the proposed approach adopts a new way, which is different with those of relying on either global thresholds (Wang et al., 2018), or large training samples and numerous statistic values (Murray et al., 2019). The issue of separating tidal flats along estuaries is also specifically considered in the proposed approach. Section 2 presents the basic idea of the proposed approach, including composition of the high and low tidal datum based solely on RS data, the water and land classification by an adaptive binary classifier without needing large sample set, and the separation of tidal flats along estuaries by considering the intrusion of saltwater into fresh surface water in detail. We applied the proposed approach to mapping the tidal flats of southern China for the year 2017, based on Sentinel-1 SAR data (Section 3). Section 4 shows the results of the proposed approach and compares them with the state-of-the-art tidal flat maps produced by Murray et al., 2019, Wang et al., 2018. We also analyze the results of the proposed approach, based on the distance of saltwater intrusion into fresh surface water along estuaries. Section 5 discusses the effectiveness of the proposed approach in estimating tidal datum based solely on satellite imagery, as well as the limitations and strengths of the proposed approach compared with the two state-of-the-art approaches. Conclusions are presented in Section 6.

Section snippets

Basic idea

Tides are affected by the phases of the moon and meteorological conditions, thus they are time-variant and perform inconsistently even at any fixed time of each day. The instantaneous tidal state could be observed by satellites with a certain revisit time and multiple instantaneous observations could be used in estimation of tides and tide-related tidal flats. Li and Gong (2016) regarded each observation from satellite imagery during multiple revisits for the same location as a random sample

Study area

The study area is located in south China, covering about half of China’s coastline (Fig. 2). It includes Zhejiang, Fujian, Guangdong, Guangxi Zhuang Autonomous, Hong Kong, Macao, Taiwan, and Hainan. The region has complex tidal patterns (including diurnal tides, semidiurnal tides and mixed tides), which are calculated by the data from Oregon State University (OSU) TOPEX/POSEIDON Crossover database (TPXO9) (Egbert and Erofeeva, 2010) according to Reeve et al. (2012). The abundance in

Tidal flat map and its evaluation

The tidal flat map integrating tidal salinity influence, of southern China for 2017 is shown in Fig. 7.

According to the randomly generated 1377 validation samples, the tidal flat map has an overall accuracy value of 92.4% (Table 1) and the 95% quantile for confidence interval of overall accuracy is 90.8–93.7%. The user accuracies for the non-tidal and the tidal flats are 88.3% and 100.0%, respectively. The lower user accuracy for the non-tidal areas on the resulted map might have resulted from

Area statistics from different tidal flat maps of the study area

The areas of tidal flats reported in existing studies covering the southern China vary largely. In a country-scale study, Wang et al. (2018) reported a tidal flat area of 292,282 ha in southern China in 2016, while Murray et al. (2019)’s global tidal flat map records an area double the size of the former one, i.e., a tidal flat area of 581,082 ha for the same area in 2014–2016. Both the studies used the same satellite data source. Using a different data source, the proposed approach in this

Conclusions

In this paper, we proposed a large-area tidal flat mapping approach with a new way to avoid the dependence on tidal elevations. The approach synthesizes Sentinel-1 SAR data over a two-year time period (to acquire high and low tidal datums) and classifies land and water under each tidal datum using an adaptive binary classifier. Thus, tidal flats could be identified as the areas being water during high tides and being land during low tides. Furthermore, the proposed approach uses time series of

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This research was funded by the Science and Technology Basic Resources Investigation Program of China (No. 2017FY100706). The authors thank Dr. Mingming Jia of Northeast Institute of Geography and Agroecology, Chinese Academy of Sciences, for her kindly providing the mangrove map of China for 2010, which was produced by her workgroup.

References (66)

  • J.-H. Ryu et al.

    Detecting the intertidal morphologic change using satellite data

    Estuar. Coast. Shelf Sci.

    (2008)
  • S. Sagar et al.

    Extracting the intertidal extent and topography of the Australian coastline from a 28 year time series of Landsat observations

    Remote Sens. Environ.

    (2017)
  • S. Schlaffer et al.

    Flood detection from multi-temporal SAR data using harmonic analysis and change detection

    Int. J. Appl. Earth Obs. Geoinf.

    (2015)
  • J. Siddorn et al.

    Detecting the Zambezi River plume using observed optical properties

    Mar. Pollut. Bull.

    (2001)
  • D. Sirjacobs et al.

    Cloud filling of ocean colour and sea surface temperature remote sensing products over the Southern North Sea by the Data Interpolating Empirical Orthogonal Functions methodology

    J. Sea Res.

    (2011)
  • E.T. Slonecker et al.

    The new Landsat 8 potential for remote sensing of colored dissolved organic matter (CDOM)

    Mar. Pollut. Bull.

    (2016)
  • T. Wu et al.

    Spatio-temporal characteristics of the mainland coastline utilization degree over the last 70 years in China

    Ocean Coast. Manag.

    (2014)
  • J. Xu et al.

    Optical models for remote sensing of chromophoric dissolved organic matter (CDOM) absorption in Poyang Lake

    ISPRS J. Photogramm. Remote Sens.

    (2018)
  • W. Zhu et al.

    An assessment of remote sensing algorithms for colored dissolved organic matter in complex freshwater environments

    Remote Sens. Environ.

    (2014)
  • S. Abdikan et al.

    Land cover mapping using sentinel-1 SAR data

    Int. Arch. Photogram., Rem. Sens. Spatial Informat. Sci.

    (2016)
  • A. Alvera Azcarate et al.

    Enhancing temporal correlations in EOF expansions for the reconstruction of missing data using DINEOF

    Ocean Sci.

    (2009)
  • Y. Bai et al.

    Remote sensing of salinity from satellite-derived CDOM in the Changjiang River dominated East China Sea

    J. Geophys. Res. Oceans

    (2013)
  • E.B. Barbier et al.

    The value of estuarine and coastal ecosystem services

    Ecol. Monogr.

    (2011)
  • Barros, V., Field, C., Dokke, D., Mastrandrea, M., Mach, K., Bilir, T., Chatterjee, M., Ebi, K., Estrada, Y., Genova,...
  • J.-M. Beckers et al.

    DINEOF reconstruction of clouded images including error maps? Application to the Sea-Surface Temperature around Corsican Island

    Ocean Sci.

    (2006)
  • M.D. Blum et al.

    Drowning of the Mississippi Delta due to insufficient sediment supply and global sea-level rise

    Nat. Geosci.

    (2009)
  • L. Breiman

    Random forests

    Mach. Learn.

    (2001)
  • J. Chen et al.

    Estimation of colored dissolved organic matter from Landsat-8 imagery for complex inland water: case study of Lake Huron

    IEEE Trans. Geosci. Remote Sens.

    (2017)
  • Y. Chen et al.

    Land claim and loss of tidal flats in the Yangtze Estuary

    Sci. Rep.

    (2016)
  • L.D. Clarke et al.

    The mangrove swamp and salt marsh communities of the Sydney district: III. Plant growth in relation to salinity and waterlogging

    J. Ecol.

    (1970)
  • M. Clement et al.

    Multi-temporal synthetic aperture radar flood mapping using change detection

    J. Flood Risk Manage.

    (2018)
  • A. Cracknell

    Remote sensing techniques in estuaries and coastal zones an update

    Int. J. Remote Sens.

    (1999)
  • C. Craft et al.

    Forecasting the effects of accelerated sea-level rise on tidal marsh ecosystem services

    Front. Ecol. Environ.

    (2009)
  • Cited by (0)

    View full text