Introduction

Tidal flats have been recognized as one of the most important ecosystems. They contribute irreplaceable ecosystem services and functions, such as carbon storage and sequestration, and they provide economic benefits as well opportunities for biodiversity conservation (Healy et al. 2002; Passarelli et al. 2018; Dang et al. 2023). In terms of biodiversity, they lie between marine and terrestrial environments; thus, they support a highly diverse range of fauna and flora (UNDP/GEF 2007). Although tidal flats are not usually covered by vegetation, the productivity of algae in the mud substrate supports many invertebrate species, such as ragworms and bivalves, which are usually foraged by waterbirds (Cai et al. 2024; VanDusen et al. 2012).

However, tidal flats have often been overlooked in contrast with coral reef and mangrove ecosystems. Tidal flat loss has been a serious threat globally (Davidson 2014). Between 1984 and 2016, approximately 16% of tidal flats disappeared in different parts of the world (Murray et al. 2019). The expansion of built-up areas, sea walls, and energy infrastructure along coastlines is widely known as the main reason for the loss of tidal flats (Gopal 2013; Ma et al. 2014). For example, the coasts in East Asia have experienced some of the most serious tidal flat losses. The expansion of infrastructure and seawall construction around the Yellow Sea, in particular, has been identified as a threat to the local tidal flat ecosystem (Choi et al. 2018).

Tidal flat loss has been recognized as the main threat to migratory waterbirds (Piersma et al. 2017; Studds et al. 2017), especially along the East Asian-Australasian Flyway (EAAF) which is the migration route for millions of migratory birds (Amano et al. 2010; Hansen et al. 2022). Along the coastline of East Asia, tidal flats that serve as stopover sites and refueling habitats for migratory waterbirds have disappeared over the last 50 years (Murray et al. 2012, 2014; Ma et al. 2014; Mathot et al. 2018). Moreover, in South Korea, the Saemangeum Seawall has impounded almost 29,000 hectares of tidal flats; this led to the loss of at least 128,000 shorebirds between 2006 and 2008 (Moores et al. 2016). Moreover, in the non-breeding grounds, habitat loss has coincided with the decline in the numbers of migratory shorebirds in Australasia (Studds et al. 2017). Recent studies have shown that even simple, predictable linkages among sites caused by “to-and-fro” migration can make migratory species especially vulnerable to habitat loss, substantially affecting conservation prioritization (Runge et al. 2014).

Despite the understanding of tidal flat change on the global scale, there are still a lot of unknowns on the smaller scale, with the stopover and wintering sites along the EAAF being a prime example. Taiwan is a vital waterbird hotspot at the midpoint of the EAAF. The sandy and muddy shores on its western coast support thousands of migratory birds during the non-breeding season (Li et al. 2019; Kasahara et al. 2020; Lin et al. 2021). However, over the last 100 years, industrial and commercial development has caused the natural coastline to shrink dramatically, with only 5% of the natural coast remaining in western Taiwan (Construction and Planning Agency 2021). For example, the six main land cover types that were used to analyze land cover changes in Taiwan for the period between 1904 and 2015 did not even include tidal flats as a category (Chen et al. 2019). Although the population trajectories of the main migratory waterbirds of Taiwan have been determined for the period between 2014 and 2021, researchers have still been unable to examine the associations between waterbird trends and tidal flat changes (Lin et al. 2023). Without these associations, it is difficult to plan the conservation strategies for migratory waterbirds.

The Kunming-Montreal Global Biodiversity Framework has issued a target that centers on reducing threats to biodiversity, with a specific focus on rehabilitating degraded coastal ecosystems (CBD 2022). This involves bringing the loss of tidal flat areas to virtually zero while simultaneously implementing effective restoration efforts. Alongside this work, Nature Positive has aimed to halt and reverse nature loss by 2030. Measuring the baseline and assessing the risks to tidal flat ecosystems will be a key strategy within this framework.

To determine the changes in tidal flats along the coast of Taiwan in different periods of the past 100 years, we collated different spatial data from different sources (e.g., spatial images, aerial photos, and historical maps) for the period between the 1920s and the 2020s. In this study, we (1) mapped the spatial distribution of tidal flats over the past 100 years and determined the trajectories of tidal flats; (2) applied the International Union for Conservation of Nature (IUCN) Red List of Ecosystems criteria (Keith et al. 2013; Bland et al. 2016) as the international standard for assessing and comparing the threatened state of tidal flat ecosystems along the EAAF; and (3) reviewed the development events to classify the threats to the tidal flat ecosystem.

Methods

Taiwan’s shoreline includes sand, rock, cliffs, gravel, and reef coasts. The sand coasts are mainly located on the western coastline (Yang et al. 2012), where is our study area between the Dansuie River (淡水河) estuary (25.17°N, 121.41°E) and the Fenggang River (楓港溪) estuary (22.19°N, 120.68°E) of Taiwan proper (Fig. 1). We obtained the spatiotemporal distribution of tidal flats from historical topographic maps and satellite data and digitized these images to estimate the area of the tidal flats (Aslam et al. 2023; Murray et al. 2014, 2019, 2022).

Mapping Tidal Flat Area

Historic Topographic Map Data

We used the historical Taiwan Fortress map published in 1921 as a baseline to estimate the area of the tidal flats. This map covers 60% of the island and includes the intertidal and lowland areas of Taiwan proper. It is the earliest map with land cover reconstruction (Academia Sinica 2003; Chen et al. 2019). The other map used was from the 110 series of 1:500,000 topographic maps reproduced in 1956 and based on the AMS L892 (Formosa [Taiwan]) topographic maps by the US Army Map Service in 1944 and in 1945 (Academia Sinica 2003). We digitized the coastal areas of these two historical maps as spatially distributed polygons and estimated the tidal flat areas in 1921 and in 1956.

Satellite Data

We followed the protocols in Murray et al. (2012) to accurately delineate tidal flat areas using Landsat images. This procedure involves extracting land areas from satellite images captured during both high tide and low tide periods, with the discrepancy in land area between these two moments serving as the basis for determining the tidal flat extent.

We used the U.S. Geological Survey (USGS) EarthExplorer to access Landsat archived images, which include the data from Landsat Collection 2 Level-1 and Landsat Collection 2 Level-2. The range covered the western coastline of Taiwan proper between 1972 and 2021. The metadata of those images includes geographic location (footprints), cloud cover, and acquisition timestamps (U.S. Geological Survey 2020). We selected the images that had low cloud cover (< 30%) and were captured within one hour of low or high tide as determined by the Taiwan Central Weather Administration (https://www.cwa.gov.tw/V8/E/D/publication.html). Our final images comprised Landsat satellite images at spatial resolutions of either 60 m (LM1 and 4) or 30 m (LT5 and LC8) (Appendix Table 2 and Appendix Table 3).

To map the tidal flat range from the satellite images, we grouped the images by footprints (path-row tiles), decades, and tide times. As Landsat Level-1 data lack atmospheric correction, we used a Semi-Automatic Classification Plugin (SCP) in QGIS version 3.18.0 (https://qgis.org/en/site/) to perform atmospheric correction for all images within the Landsat Level-1 dataset. Following this correction, we calculated the Normalized Difference Water Index (NDWI) for every pixel in each of the selected images. The equation for the NDWI was as follows:

$$ {R}_{TM/ETM}=\frac{{\rho }_{green}-{\rho }_{NIR}}{{\rho }_{green}+{\rho }_{NIR}}$$
(1)

where \( {\rho }_{green}\) and \( {\rho }_{NIR}\) are the radiances of the green and near-infrared wavelengths of the Landsat images (Murray et al. 2012; Xu 2006). While NDWI generally uses zero as the threshold to classify water and land (NDWI ≥ 0 indicates water, NDWI < 0 indicates land), the implementation revealed that the optimal threshold might not be exactly zero (Xu 2006). Therefore, to achieve the utmost accuracy in land area determination for each image, we initially used zero as the threshold to classify pixels as either water or land. Subsequently, we assessed classification accuracy by randomly selecting several locations and comparing the classification results against the actual Landsat images. In cases of significant misclassification, we fine-tuned the threshold in increments of 0.01, repeating the previous steps until the optimal threshold value was attained. The threshold that was used falls within the range of -0.01 to 0.2. We then overlaid the land area polygons with satellite images and examined along the coastline to identify any discrepancies in land classification (Aslam et al. 2023). Corrections were made for water-land classification errors. Lastly, the low-tide land area was subtracted from the high-tide land area for each group to obtain the tidal flat area. All computations were conducted using QGIS version 3.18.0.

Assessing the Risk to the Ecosystem

Regarding ecosystem risk, we applied criterion A. This involves assessing the declining distribution of an ecosystem, as outlined in the IUCN Red List of Ecosystems criteria (Bland et al. 2016; Keith et al. 2013), to evaluate the risk status of the tidal flats of western Taiwan. The assessment involves three distinct time frames: (1) the Present, spanning the past 50 years; (2) the Future, which includes any 50-year period covering both the present and future; and (3) the Historic, extending back to 1895, the year in which Taiwan began its systematic development. By digitizing historical map data and utilizing satellite images, we have acquired data on tidal flat areas for seven periods over the past 100 years (the 1920s, 1950s, 1980s, 1990s, 2000s, 2010s, and 2020s). However, to fulfill the requirements of criterion A, we needed data for both the most recent 50-year period and for future periods. To bridge this data gap, we employed Loess Regression to predict missing information based on existing data.

To extrapolate the tidal flat areas into the future. We employed the loess() function in R to build a model that estimates the changes in tidal flat areas and used the predict() function to interpolate the tidal flat areas in years where data was missing (1920–2020 s) (R version 4.2.2; R Core Team 2022). The default “span” parameter value of 0.75 was used in the model. Two distinct models were built: one was for interpolating tidal flat areas and another was for extrapolating them. For interpolation, we utilized the model with the control parameter set to “loess.control (surface = ‘interpolate’)”; while for predicting future values, we used the model with the control parameter set to “loess.control (surface = ‘direct’)”.

Subsequently, we harnessed all available data, which encompassed information derived from historical maps, satellite imagery, and predicted data, to quantify changes in tidal flat areas across these three time frames and to implement the risk assessment. Specifically, the periods 1960–2010 s and 1970–2020 s represented the “Present”, the 1980–2030 s represented the “Future,” and the 1920–2020 s represented the “Historic.” An area was categorized as Vulnerable (VU) when its distribution displayed a decline of 30–50% in the “Present” or the “Future,” or a decline of 50–70% in the “Historic.” The Endangered (EN) classification was assigned if the decline ranged from 50 to 80% in the “Present” and “Future,” or 70–90% in the “Historic.” Finally, the category of Critically Endangered (CR) was applied when the decline surpassed 80% in the “Present” and “Future,” or exceeded 90% in the “Historic.” Ultimately, the most severe risk category identified via any of these criteria was utilized to determine the overall status of the ecosystem, following the approach advocated by Bland et al. (2016) and Keith et al. (2013).

Moreover, to understand the threats altering the expanse of tidal flat areas, we employed land use data spanning the last century to compile significant coastal development events as well as development occurrences along rivers. These events encompass notable occurrences like the establishment of ports as well as developmental undertakings along rivers, such as the construction of reservoirs and dams. The compilation includes the event, the establishment year, the type of development, and the type of threat. All of these events have been classified into categories according to the IUCN direct threats framework, which is based on land use and land cover types (IUCN 2022). This framework delineates four categories, each representing a distinct type of threat: (1) Residential and Commercial Development, which encompasses areas designated for residential and commercial purposes. This includes the establishment of structures such as harbors, thermal power plants, industrial factories, and terrestrial regions. (2) Agriculture and Aquaculture, which revolves around activities related to agriculture and aquaculture. These activities are further subdivided into agriculture and aquaculture areas. (3) Energy Production and Mining, which encompasses energy production and mining endeavors and activities such as mining and quarrying, the reclamation of deserted salt ponds, and the deployment of renewable energy sources like solar plate farms. (4) Natural System Modifications, which focuses on alterations made to the natural system. This includes transformations like the construction of dams as well as various water management or utilization practices, such as the establishment of reservoirs and river weirs. By categorizing these events, we aim to evaluate and address the potential impacts of various developmental activities on the coastal ecosystem.

Results

Our data cover 438 km of Taiwan’s coastline, and our results reveal a rapid decline of tidal flats along the western coastline between 1921 and 2021 (Fig. 1). Historical topographic maps indicate that tidal flats occupied 368.29 km2 in the 1920s and 459.72 km2 in the 1950s, an increase of 24.83% for the first period (1920–1950 s). As per satellite image estimations, the area of tidal flats decreased rapidly to 339.78 km2 in the 1980s, resulting in a second period (1950–1980 s) decrease of 26.09%. In the 1990s, the area decreased further to 270.51 km2. After the 2000s, the decline rate slowed down (11.53%); the tidal flat area was 239.31 km2 in the 2000s and 194.05 km2 in the 2010s. Overall, between the 1950s and the 2010s, the area decreased rapidly from 459.72 km2 to 194.05 km2. However, in the recent decade, the area of tidal flats expanded slightly to 236.08 km2 in the 2020s (see Table 1; Fig. 2, and see Appendix Fig. 3 for more details).

Fig. 1
figure 1

The detection of tidal flat change for the past 100 years. The area of the tidal flats in western Taiwan was 236.08 km2 (red patches) in the 2020s. The grey patches represent the area lost over the last 100 years. The blue line is the coastline, and the white borders delineate the study area. A-E are the zoom-in images for the part of western coast respectively

Full size image
Table 1 The net loss of tidal flats of the western coast of Taiwan between the 1920s and the 2030s
Full size table
Fig. 2
figure 2

The total area of tidal flats in western Taiwan for the period 1920–2030. The tidal flat area decreased between 1950 and 2010, but somewhat increased in the following decade. The solid points represent the data used to generate the estimated rates of change. The solid line is the trend (both past and future) of the tidal flat area, and the dash lines represent the 95% confidence interval

Full size image

During the last 100 years, the lowest tidal flat area (194.05 km2) occurred in 2010. This was the culmination of a 60–year decline (1950–2010 s) in tidal flat area in this region, representing a loss of 57.79%. According to the extrapolated data, 56.46% of the tidal flat coverage was lost in the 50–year period of 1960–2010 s, and 42.27% was lost between 1970 and 2020 s, indicating that the ecosystem has reached VU-EN status based on the ecosystem distribution decline rate. Based on the extrapolated data, the tidal flat area will reach 313.09 km2 in the 2030s. This results in a 7.86% decline for the 1980–2030 s and corresponds to a future decline status of Least Concern (LC). The total decline of the tidal flat area since the 1920s has been 35.90–47.31%, and the ecosystem had an LC status due to historic changes (Table 1; Fig. 2). Taiwan’s tidal flat ecosystem is now estimated to be Endangered.

We discovered 48 reservoirs and river dams in 17 major western river systems, nine commercial and industrial harbors, six thermal power plants and industrial factories, and seven agriculture and aquaculture developments in the western coastal region, all of which were built in the 1900s. The lost tidal flat areas were converted into agriculture and aquaculture areas (53%; 9% agriculture, and 44% aquaculture), residential and commercial development (46%), and energy production and mining (2%) (Fig. 3).

The development of tidal flat ecosystems in this region can be classified into four stages. In the first stage (1920–1950 s), the tidal flat area increased because of the continual supply of sediments from the upstream rivers. From the 1950s to the 1970s, tidal flat abstraction and reclamation for aquaculture and agriculture caused a great deal of shrinkage in the southwestern Taiwan. The third stage (1980–2010 s) was the industrial development period, during which the tidal flat area continuously decreased. Many harbors, thermal power plants, and industrial factories have been constructed on the coastline since the 1980s, and the last huge industrial factory area was built in 1998. In the fourth stage, representing the most recent decade (2010–2020 s), industrial construction slowed down and the tidal flat area slightly increased.

Fig. 3
figure 3

Land cover type over the past 100 years of predominantly tidal flat areas on the western shore of Taiwan. The loss tidal flat areas were converted into agriculture and aquaculture areas (53%; 9% agriculture, and 44% aquaculture), residential and commercial development (46%), and energy production and mining (2%). Black solid lines indicate the ranges of study areas

Full size image

Discussion

The extent of tidal flat loss is alarming and irreparable. Tidal flats are major wetlands with high productivity and biodiversity in both marine and terrestrial areas. The disappearance of tidal wetlands disrupts the entire ecosystem, impacting various species that depend on these habitats for breeding, foraging, or shelter. The loss of tidal flats negatively affects migratory shorebirds by reducing vital rest and foraging areas, leading to habitat scarcity and food shortages. This can result in decreased bird health and lower chances of survival and successful migration (Buehler and Piersma 2008; Runge et al. 2014, Mathot et al. 2018).

Our study underscores the urgency of addressing wetland loss on a global scale and highlights the importance of conserving tidal flat ecosystems; it also provides information on the speed and progress of mudflat accumulation before systematic human development. The number of migratory shorebirds that use the EAAF has declined dramatically over the last few decades. Evidence suggests that this decline is due to habitat loss at important stopover sites in the Yellow Sea (Moores et al. 2016; Murray et al. 2014; Studds et al. 2017). As Taiwan is a crucial stopover site for migrating birds in the EAAF, the tidal flats in Taiwan serve as essential refueling sites for these migrations. The degradation of Taiwan’s tidal flat ecosystem is not only a local threat but also a threat to migratory shorebirds in the EAAF (Lin et al. 2023).

We identified distinct periods of change in tidal flat ecosystems in Taiwan: expansion in the 1920s to 1950s, rapid decline from the 1980s to the 2010s, and recent stabilization and regrowth from the 2010s to the 2020s. Notably, a reduction in human disturbance in recent years has contributed to the slowing of tidal flat disappearance. Coastal areas have rapidly developed and are now occupied by large populations, resulting in the global loss of tidal flats (Kirwan and Megonigal 2013; Murray et al. 2019). The situation in Taiwan mirrors broader Asian and global trends. In Taiwan in 2017, 45% of the coastlines were artificial; in the western part, only 10% of the coastline was natural (Construction and Planning Agency 2021). From the 1950s to the 2010s, 57.79% of the tidal flat area was lost. The results also showed that more than half of the tidal flats in Taiwan have disappeared in the last 60 years, a pattern similar to that observed in the Yellow Sea (Moores et al. 2016; Murray et al. 2014, 2015, 2019). The changes in socioeconomic patterns in East Asia were similar to those in Taiwan, but Taiwan experienced these changes a decade earlier.

Our results reconstructed the spatial and temporal patterns of the coastal area over the last 100 years, and we conducted local and regional coastline long-term assessments. While satellite imagery is accessible from the 1960s, historical topographic maps were crucial for reconstructing past land cover data. Despite the potential influence of cartographic generalization, these maps are invaluable records offering insights into historical environments as well as evidence of environmental alterations (Stein et al. 2010). The foundational topographic maps spanning from 1921 to 1956 featured precise survey techniques (Academia Sinica 2003; Chen et al. 2019). These materials make us able to estimate the distribution shift of tidal flats over the past century, though we approach comparisons with historical maps with due caution (Murray et al. 2014). Although the quality and coverage of Landsat images used in this study were insufficient to estimate tidal flat extent in Taiwan during the 1960s and 1970s, but Shin (1980) through fieldwork and map comparisons, found that the area of tidal flats in western Taiwan increased continually until the 1970s. Furthermore, in the process of our research, we enhanced the spatial resolution of the regional mapping, avoiding situations of image misinterpretation such as misidentifying fishponds as intertidal mudflats, as has been seen in the other global-scale studies (Murray et al. 2022).

According to the IUCN Red List of Ecosystems criteria, the tidal flat ecosystem in Taiwan has reached EN status. This assessment is useful for evaluating the impact of human activities on ecosystems and the major direct threats to the tidal flat ecosystem in Taiwan: (1) residential and commercial development, (2) agriculture and aquaculture, (3) natural system modifications, and the potential threat of (4) energy production and mining. The extensive degradation of tidal flats in Taiwan, emblematic of a global issue, stems from a systematic reclamation process initiated in 1947, marked by a governmental development plan that repurposed 10% of tidal flat area for agriculture, fishery, and salt production over three decades (Shin 1980). Half of the lost tidal flat areas have been converted into aquaculture farms, aquaculture farms, and abandoned salt ponds have become alternative habitats for migratory shorebirds in the coastal region. Subsequently, heavy, petrochemical, and shipbuilding industries and large public construction projects, such as transportation and electric power, have been developed in the coastline areas since the 1970s. Nine commercial and industrial harbors, six thermal power plants and industrial factories, and seven major agricultural and aquaculture developments have been built in the western coastal region since the 1900s.

Tidal flats situated near estuarial delta areas, which rely on river-delivered sediments, have faced significant degradation in Taiwan due to factors such as sediment supply disruption, sediment deposition, and the construction of dams and reservoirs in upstream river systems (Chang and Chen 2001; Hood 2004; Hsu et al. 2007; Walling and Fang 2003; Walling 2006; Beninger et al. 2018). While the 20th century initially witnessed an expansion of these areas, a persistent decline has been recorded since the 1970s, with recent years showing only a slight resurgence. In Taiwan, the transport of sufficient sediment from mountainous areas to the western coastline has led to the creation of extensive tidal flats (Hsu et al. 2007). However, a substantial reduction in sediment output has resulted from at least 48 reservoirs and river dams in 17 major western river systems. The construction of the Jiji Weir in 2002, which locates on the Taiwan’s largest river, had a critical impact, disrupting sediment transport from upstream and endangering the last surviving tidal flat located on the north side of the Choshui River estuary (Huang 2010), which is the largest and the last remaining patch of complete tidal flats in Taiwan (Fig. 1C). A huge industrial project called the Kuokuang Petrochemical Project was planned for this area in the 2000s, but the project stopped in 2011. This project included refineries and petrochemical midstream/downstream factories, and the expected development area was approximately 1,900 ha. However, there is another threat to the horizon in the form of solar panels, a new type of energy-producing building which needs a huge area to set the construction. These will cover on tidal flats that are also the cause of wetlands being endangered in recent years.

Our research significantly contributes to the Kunming-Montreal Global Biodiversity Framework by providing crucial insights into the status and changes in tidal flat ecosystems, which are integral to global biodiversity (CBD 2022). It offers valuable data for biodiversity assessment, particularly in the context of the EAAF, where migratory species heavily rely on these ecosystems. Designating tidal wetlands as “endangered” under the IUCN Red List of Ecosystems criteria is of immense conservation significance. Beyond the wetlands themselves, this status underscores their critical role in supporting diverse species. This approach extends beyond past species-centric conservation efforts, embracing a more comprehensive view of conservation. Furthermore, it underscores the interconnectedness of ecosystems across borders, encouraging international collaboration, raising public awareness, and promoting adaptive management, all vital components in achieving a Nature Positive world.