Computational region

To simplify computation while reflecting the topography and ecosystem distribution specific to given regions, we divided and categorized global coastlines into computational domains. Global coastlines have been classified on the basis of coastal topographic entities, types such as deltas or fjords, etc [18]. Based on these global coastline types, adjacent line segments shorter than 1,000 km that were of the same type were merged, yielding 198 computational areas (the mean coastline length was about 2,000 km) as shown in S1 Fig. Because SWEs have the potential to expand or shift not only in the cross-shore direction but also in the coastline direction, the entire length of each segment, which was divided based on the coastline type, was applied to the computational domain to account for its potential in this study.

Geographical data

Topographic data for land and sea were derived from the Shuttle Radar Topography Mission 15 PLUS (SRTM-15PLUS) dataset (WebGIS: http://www.webgis.com/srtm3.html). The spatial resolution of this dataset is about 450 m. Topography in the landward direction was incorporated to an elevation of up to 50 m to ensure adequate coverage when the SWEs expand in the future. Bathymetry was incorporated to 100 m depth to encompass the depth range from the maximum high-tide surface to the euphotic layer (S2 Fig). The elevation data were averaged along the shoreline direction for each of the 198 computational areas ArcGIS software (ESRI, Inc.) to generate hypsometric curves in the offshore direction with a horizontal resolution of 100 m (S3 Fig). The zero elevation point, which is the boundary between sea and land, was obtained from Open Street Map Data (https://osmdata.openstreetmap.de/).

Distribution of SWEs

The five SWEs we focused on were seagrass meadows, macroalgal beds, tidal marshes, mangroves and coral habitats. As with the topographic data, the SWE area was compiled in each of the 198 areas classified in “Computational region” (S1 Fig). These data were obtained from UNEP-WCMC except for macroalgal beds (S4 Fig and S1 Table). Because there are no available data for the global distribution of macroalgal beds, we adopted the total global area reported in the literature [19]. Because macroalgae are nearly ubiquitous in coastal areas [20], we assumed that all 198 computational regions are habitable areas for macroalgae. Macroalgal beds constitute the largest area among the SWEs (S1 Table). These SWEs are very different from one another, and each ecosystem is not only affected by climate change but also affects climate change through processes such as CO₂ absorption. Mangroves, tidal marshes and seagrass meadows were the primary focus of initial research on blue carbon [2], but macroalgal beds were also included in the study of blue carbon in IPCC AR6 and other recent studies since Krause-Jensen and Duarte [19].

Regulating factors for SWE projections

When projecting changes in the distribution and extent of SWEs due to climate change, we considered the following regulating factors: water temperature for coral habitats, chlorophyll concentration and sea-level change for seagrass meadows and macroalgal beds, and sea-level change for tidal marshes and mangroves. The IPCC special report on the ocean and cryosphere in a changing climate [9] estimated a maximum rise of 110 cm in sea level by 2100. The major global climate models (GCMs) used in climate prediction are summarized in the Coupled Model Intercomparison Project phase 5 (CMIP5) [21]. For this study, we used results from the GFDL-ESM2M model [22], which has a relatively rich set of marine chemical and physical variables, as the regulating factors for calculating changes in SWE extents. We adopted RCP2.6 and RCP8.5 because these two scenarios have the lowest and highest emissions, respectively, and thus the projections can clearly show the influence of climate change. RCP4.5 and RCP6.0 are positioned between the other two, and the influences should within the range of those projected with RCP2.6 and RCP8.5, so they were omitted from the study. As described in the next section, this study performed calculations in the present state and the future. The GCM’s output data from historical experiments were used as regulating factors in calculations of the present state, and the variables in the RCP scenarios were used for the future projections. For applying the GCM results, we extracted the data adjacent to the land area of each of the 198 computational areas, and averaged these data for use in our projections of SWE change.

Computational term

The period for the calculation of present conditions is 1986–2005 [23, 24], and the period for predicted conditions is 2031–2100. We excluded the period from 2006 to 2030, because it includes the past despite being in the prediction period and the very near future. Sea-level change is applied as regulating factor for calculating changes in the extent of seagrass meadows, macroalgal beds, tidal marshes, and mangroves, but the results calculated by the GCM are defined as “sea level height from geoid layer” (ZOS) and “thermosteric sea level change” (ZOSTOGA). Therefore, we established ZOS+ZOSTOGA averaged over the present period as the basis and adopted the difference between the present and each month of the future period as sea-level change. Although the future predictions are performed on a monthly basis, it is plausible that the calculation results would be biased when particular months or portions of the year are used for comparison. Therefore, we averaged the calculated results for the present state over 20 years and those for the future predictions over 10 years, respectively, and compared these averages to reduce the bias (S2 Table).

Projection of seagrass meadows and macroalgal beds

The effect of light intensity is maximal on primary production in benthic ecosystems, and it also has an effect on seagrass meadows and macroalgal beds [25–29], but it has been reported that photoinhibition may not affect the growth of seagrasses [30]. They do, however, require a minimum quantity of light for growth [26, 31], indicating that low light levels are a greater constraint on seagrass growth than high light levels. Although macroalgae has been reported to exhibit photoinhibition [32], in the absence of relevant data for the diverse range of macroalgal species, we assumed that macroalgae, like seagrasses, are not subject to photoinhibition. In this study, we calculated the maximum water depth at which sufficient light exists for the growth of seagrasses and macroalgae with the following equation proposed in previous studies [26, 31, 33]: (1) (2) the values and coefficients for which are defined in S3 Table. The light attenuation coefficient (K_PAR) is calculated by using the chlorophyll-a concentration (Chl) as a proxy for light attenuation, and the photosynthetic active radiation (PAR) depth (Z_c) is calculated from K_PAR. The distribution width (L) is estimated (S5 Fig) from Z_c and the hypsometric curve for each computational area (S3 Fig). The current distribution width (L_p) is estimated from current data, and the future distribution width (L_f) is estimated by recalculating from future chlorophyll-a concentrations and sea-level change. In this study, the future area was estimated by multiplying the ratio L_f/L_p by the current habitat area (E_p). The areal change in seagrass meadows was estimated in each of the 198 computational areas and summed as the future global distribution area: (3)

For macroalgal beds, lacking data specific to the computational areas, the future global distribution area was estimated by (4)

Terms in these equations are defined in S3 Table.

In this study, the PAR depth was the governing environmental parameter for the change in SWE area. PAR depth depends on changes in the chlorophyll-a concentration. Moreover, the chlorophyll-a concentration applied in this study depends on offshore environmental changes because it is the result of GFDL-ESM2M, which is one of the GCMs used. Coastal chlorophyll-a concentration can be influenced by anthropogenic effects from land, but when the global coasts are divided into 198 segments, such as in this study, the ratio of region open to offshore increases. Because the effects of nutrients from offshore inputs are greater than those from anthropogenic inputs for ecosystems open to offshore [34, 35], we applied the chlorophyll-a concentration from the GCM as the input parameter. Although other parameters are known to be influential, these were ignored for lack of firm constraints.

Ranges of water temperatures that are optimal or sufficient for the growth of seagrasses and macroalgae have been reported, but these differ among species, and thus, if we consider multiple species, seagrasses and macroalgae can inhabit a wide range of shallow water areas [30, 36]. In general, species in mid to high latitudes favor lower water temperatures, and species in low latitudes favor higher temperatures [30, 36]. If water temperatures in mid to high latitudes increase, ecosystems based on currently dominant species may disappear while dominant species from lower latitudes may expand to higher latitudes [37]. Thus, the adaptability of these SWEs to changes in water temperature may be greater than the adaptability of specific species. Nutrients are also a limiting factor for the growth of seagrass and macroalgae. However, because the intake rate of nutrients is influenced by light intensity in the first place, light intensity is more important [38]. In addition, the nutrients in pore water affect the growth of seagrasses [30], although the global distribution of nutrient concentration and future changes in it are poorly understood. Therefore, nutrients are not accounted for in this study.

Because other environmental conditions (such as salinity, precipitation, and bottom sediment) are not known in detail in areas adjoining but not presently containing (seagrass meadows, macroalgal beds, tidal marshes, and mangroves), we did not consider the possible expansion of these SWEs into new areas, but we did consider changes of their current distribution areas in the offshore and onshore directions.

Projection of tidal marshes and mangroves

Tidal marshes and mangroves occupy the intertidal area. In this study, we assumed that tidal marshes and mangroves are distributed from MSL to mean high water spring tide (MHWS) (S6 Fig). Because sea-level change markedly influences in changes in the distribution of tidal marshes and mangroves [39, 40], we used topographic and sea-level data to calculate the changes in the distribution width of each ecosystem after sea-level change. We then estimated the future area by multiplying the ratio of future change by the current ecosystem area. We summed the 198 computational areas to determine the total future area after estimating the area of tidal marshes from (5) and the area mangroves from (6)

The terms in these equations are defined in S4 Table. We extracted tidal data for each month from 1986 to 2005 from the TOPEX/Poseidon dataset (TPXO8-ATLAS) [41], and the averaged data were applied to the projection.

Projection of coral habitats

For coral habitats, high water temperature is fatal [42]. Previous studies have reported that coral bleaching occurs when sea surface temperature (SST) on the warmest month of the year exceeds 30°C [43–45]. It has also been reported that coral habitats can expand into previously unsuitable areas when SSTs for the annual coldest and warmest months are above 18°C and below 30°C, respectively [46]. In this study, we used these constraints when calculating the changes in coral habitat distribution from the present to the future. We assumed that coral bleaching eliminated coral habitat when SST in the warmest month exceeds 30°C for more than five years in a calendar decade (2021‒2030, 2031‒2040, etc.). We also excluded areas from coral expansion in which SST on the annual coldest month is below 18°C. The area of habitat expansion was determined by multiplying the area in which coral habitats expand by the global average proportion of present coral habitat in the area of coastal shallow waters. Areas where there are no coral habitats despite having the appropriate SST range were excluded because other environmental factors may have hindered the establishment of corals.

Although high water temperature is thought to be the main cause of coral loss, other factors have been cited such as acidification and light transmittance changes in seawater. A previous study has found that the effect of acidification was negligible [15]. To evaluate the effect of changes in light transmittance in sea water, the light demands for photosynthesis of diverse zooxanthellae are necessary, but they are poorly understood. Therefore, SST was the only environmental factor we considered for calculating the change of coral habitat distribution in this study.

Sedimentation rate

The effects of particulate matter sedimentation are more pronounced in SWEs where the sediment consists of sand and mud (seagrass beds, tidal marshes, and mangroves). Specifically, bottom friction and the drag force of submerged aquatic vegetation increase the rate of sediment accumulation [47–49], which can offset the effect of sea-level rise. In this study, we assumed a sedimentation rate of 1.68×10⁻⁴ m/month for seagrass meadows, based on Duarte et al [5], and held it constant throughout the computational period. Sedimentation was not considered for tidal marshes and mangroves because the average shoreline length in each computational area is approximately 2,000 km, and the average hypsometric curve includes relatively steep terrain outside the gently sloping intertidal area; thus, we considered future changes in sedimentation in these areas to be negligible because of the intertidal zone being absorbed into the spatial resolution in the coast-offshore direction (>100 m).

Interaction between environmental variables and feedbacks and tipping point

In seagrass meadows, transparency changes with topographic changes that are induced by sedimentation, and seagrass meadows shifted landward also change the region of sedimentation. For the other SWEs, because only one environmental factor is applied as described above, interaction and feedback cannot be considered. Furthermore, multiple SWEs can be included in one region, but each SWE is highly independent because they do not necessarily overlap at the same point in a region along a coastline of few thousand kilometers and are often scattered. Therefore, interaction and feedback between the SWEs are not taken into account in this study.

Subsidence

This study estimates the effects of subsidence on the distribution of tidal marshes and mangroves. Subsidence in several deltas [50] was assigned to 7 tidal marsh regions and 12 mangrove regions (S5 and S6 Tables). Furthermore, relative subsidence was accounted for by adding it to sea-level rise following the method of Schuerch et al [14]. Future global subsidence is unknown, so we applied the present subsidence.

Seagrass meadows and macroalgal beds

The area of seagrass meadows was projected to be stable in RCP2.6, but increase by 11% in RCP8.5, from the present to 2100 (Fig 1b). By region, seagrass meadows were predicted to expand the most in East-South Africa and West Africa and to shrink the most in the Mediterranean (Fig 2a–2e, S7 and S8 Tables). The projected extent of macroalgal beds before 2100 varied within a narrow range from −2% to 0.9% (Fig 1c).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Predicted global distribution of seagrass meadows.

a Present, b 2050s in RCP2.6, c 2090s in RCP2.6, d 2050s in RCP8.5, e 2090s in RCP8.5. The color bar shows changes in area; the present area is 100%. The coastline data was obtained from Open Street Map (openstreetmap.org/copyright).

https://doi.org/10.1371/journal.pclm.0000298.g002

The difference in the area change of seagrass meadow between the two RCP scenarios is due to the photosynthetic active radiation depth (Z_c), which was close to the current value in RCP2.6 (around 22 m) but slightly deeper in RCP8.5 (23 m) by 2100 because of a continuing decrease in the chlorophyll concentration (S7 Fig). The future Z_c varied between 21 and 23 m (for seagrass meadows) and 62 and 66 m (for macroalgal beds) and became slightly deeper starting around 2050 in RCP8.5. The maximal differences in Z_c between the two RCP scenarios were about 1.6 m in seagrass meadows and about 3.7 m in macroalgal beds. The decrease in chlorophyll concentrations can be accelerated by reduced nutrient supplies as upwelling weakens under global warming [51]. As a result, the transparency of sea water increases, and the expansion of favorable habitat for seagrass meadows and macroalgal beds may be particularly noticeable in RCP8.5.

Our results show a 11% expansion of seagrass meadows by 2100, in contrast to a previous study showing a decrease of −8.6% in RCP8.5 [16]. This discrepancy may reflect a difference in the assumed changes in the euphotic layer. The previous study assumed a constant euphotic layer depth, whereas this study assumed that changes in chlorophyll concentration would deepen the euphotic layer. If our model assumes no change in chlorophyll concentration (and thus no change in euphotic layer depth), the extent of seagrass meadows would decrease by 2.5% from the present (S8 Fig), similar to the previous study. However, the assumption of no sea-level changes permits the extent of seagrass meadows to expand 8.9% by 2100. Therefore, if the water quality is good (i.e., the degree of transparency is high), for example, seagrass meadows can tolerate greater increases in water temperature. In practice, actions to reduce runoff in the form of sediments and nutrients under the European Water Directive (https://environment.ec.europa.eu/topics/water/water-framework-directive_en) should continue to improve water transparency. Production in seagrass ecosystems has been observed to recover with increased water transparency [52], and the global distribution of seagrass has expanded [8]. Therefore, if management actions taken to reduce the runoff of sediments and nutrients are carried out in the future, seagrass meadows further expanded.

In addition, seagrass meadow has been expanding along the coast of Greenland since 1900, primarily driven by climate change (however, Greenland was not included in the computational area of this study due to its non-habitat area in the present distribution; S4b Fig). This expansion illustrates how climate change can influence the spread of seagrass ecosystems depending on regional environmental conditions. Therefore, it is important to note that climate change does not always have negative impacts everywhere.

Our results showing almost no change in the area of macroalgal beds (-2% to 0.9%) by 2100 also differ markedly from the previous study, which predicted a decrease of kelp (seaweeds belonging to Laminariaceae) by 20.6% in the RCP8.5 scenario [16]. First, we considered all groups of macroalgae, whereas the previous study considered only the kelp group. Second, while the previous study set an upper water temperature limit of 26°C for predicting sustainable kelp habitats [16], our study ignored water temperature because each macroalgal group has different optimal water temperatures [36]; thus, even if changing water temperature induces a shift in macroalgae species, the distribution changes for the whole macroalgal group are unpredictable given current knowledge. For this reason, comparisons with the previous study are problematic.

The effects of water temperature were not considered in the projection of seagrass meadows and macroalgal beds because these ecosystems can adopt to climate change by a change in species composition (or dominant species) as described in the methods section. However, if there is a change from temperate species to tropical ones as water temperature increases, seagrasses can alter sediment retention and accretion because of changes in shoot density, canopy height, and other features. Additionally, the ecosystem service can change; for example, food values can decrease if macroalgal species that favor lower water temperatures change to species that favor higher water temperatures. Further study is necessary because differences in methodology can cause differences in reported values [53], and each ecosystem habitat has a different function [54].

Tidal marshes and mangroves

The areas of tidal marshes and mangroves were projected to be almost stable in both RCP scenarios by 2100 (Fig 1d and 1e). Our results showed that the areas of these SWEs did not change under sea-level change.

The area of tidal marshes was projected to continuously expand in the Black Sea (in the Mediterranean region) in both RCP scenarios, while the projected distribution was almost stable between the RCP scenarios in other regions (Fig 3a–3e and S7 and S8 Tables). These spatial differences can be attributed to regional differences in sea-level change and the relationship between sea-level change and the topographic gradient in the cross-shore direction. These details are discussed below.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Predicted global distribution of tidal marshes and mangroves.

a Tidal marshes in the present, b 2050s in RCP2.6, c 2090s in RCP2.6, d 2050s in RCP8.5, e 2090s in RCP8.5. f Mangroves in the present, g 2050s in RCP2.6, h 2090s in RCP2.6, i 2050s in RCP8.5, j 2090s in RCP8.5. The color bar shows changes in area; the present area is 100%. The coastline data was obtained from Open Street Map (openstreetmap.org/copyright).

https://doi.org/10.1371/journal.pclm.0000298.g003

In both RCP scenarios, the projected distribution area of mangroves also showed no marked differences from the present (Fig 3f–3j, S7 and S8 Tables).

Future sea-level change in the global coastal regions is presented in S9 Fig. In both RCP scenarios, average sea-level changes in each computational region by 2100 were 0.2 m in RCP2.6 and 0.4 m in RCP8.5. The range of sea-level changes in the 2090s was between 0.3 m and -0.2 m in RCP2.6 (difference: 0.5 m) and between 0.6 m and -0.3 m in RCP8.5 (difference: 0.9 m). Although sea level is rising as a global average due to climate change, sea level can locally fall because of changes in ocean currents [55]. Thus, changes in the area of SWEs are dependent on sea-level change in each region.

Consequently, the relationship between geomorphic gradients and sea level determines the future change in the area of tidal marshes and mangroves (Fig 4 and S10 Fig). If sea-level rise occurs in areas where slopes grow gentler away from the shoreline, tidal marshes and mangroves in the intertidal zone will expand into more area than is lost (Fig 4a and 4b). In turn, when the sea level falls, the area lost is greater than the area expanded (S10a and S10b Fig). If the landward slope is steeper than the seaward slope, sea-level rise reduces more area than it expands (Fig 4d and 4e); the opposite is true when sea level falls (S10c and S10d Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Hypsometric relations between sea-level rise and area of SWEs.

Schematic views of the relationship between geomorphic slope where a the landward slope is gentler than the seaward slope and the resulting SWE changes under sea-level rise in a coastal setting b without hard infrastructure and c with hard infrastructure, and where d the landward slope is steeper than the seaward slope and the resulting SWE changes under sea-level rise e without hard infrastructure and f with hard infrastructure. MSL_t1, present mean sea level; MHWS_t1, present mean high water spring tide level; MSL_t2, future mean sea level; MHWS_t2, future mean high water spring tide.

https://doi.org/10.1371/journal.pclm.0000298.g004

Our results from tidal marshes and mangroves differ greatly from those of the IPCC special report on the ocean and cryosphere in a changing climate (SROCC) [9], which projected that coastal wetlands (tidal marshes, mangroves, and seagrass meadows) will shrink by 20%‒90% from present conditions [10, 11, 56]. Although direct comparisons of the two studies are difficult, the reason for this discrepancy may be that SROCC assumed 1 m of sea-level rise by 2100 [10], which is an overestimate compared to the average value of RCP8.5, but the studies also differ in other data used. However, if sea-level rise is set at 1 m in our model, then the area of tidal marshes is stable (S11a Fig) and that of mangroves expands by 3.5% (S11b Fig). Additionally, we carried out a numerical prediction in order to assess the effects of the landward shift of SWEs relative to sea-level rise. This projection assumed that the landward shift of SWEs relative to sea-level rise was limited due to the presence of global coastal hard infrastructures and land use (Fig 4c and 4f). Our projection showed that 91.9% of tidal marshes and 74.3% of mangroves are lost by 2100 in RCP8.5 (Fig 5). Similarly, Kirwan et al [12] and Lovelock et al [13] showed that SWE areas shrank considerably due to sea-level rise (similar to Fig 5) because coastal structures were assumed to exist. In addition, they predicted that sedimentation prevented SWE shrinkage despite the fact the SWEs could not shift landward because the seafloor rise by sedimentation offset the effects of the sea-level rise. In this study, sedimentation was not considered in tidal marshes and mangroves, although the results showed that the area will at least be sustained if a landward shift of SWEs is allowed. Sedimentation does not necessarily follow sea-level rise, and it has also been shown that the self-weight can accelerate subsidence as discussed below [57]. Therefore, it terms of sea-level rise, it is important to consider whether open spaces have been maintained (i.e., to consider the effect of hard infrastructure) such that SWEs can migrate in the future. These points support the view that the projections from our model are an improvement over those of previous studies. However, there are still large gaps in these future projections. Hard coastal structures (e.g., seawalls) and related land use (e.g., culture ponds) are distributed along the coasts of many countries (e.g., hard infrastructure covers 14% of the U.S. coastline [58]), but global data on these subjects are inadequate for the present and nonexistent for future projections. It may be possible to project the distribution of SWEs with higher accuracy in the future given adequate global data on hard infrastructure.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. SWE areas gained and lost with sea-level rise.

Shown are percentages relative to the present of area lost (dashed line) and gained (solid line) in habitat for a tidal marshes and b mangroves due to sea-level rise. Blue and red colors represent changes in RCP2.6 and RCP8.5, respectively.

https://doi.org/10.1371/journal.pclm.0000298.g005

There are also regions where sea level falls, but the effects of the disappearance of tidal marshes and mangroves are considerably less than those caused by sea-level rise (S12 Fig). Furthermore, our model considers globally calculated sea-level change as an external forcing, but discrepancies can be created if local sea-level change based on observed data is applied [56].

SWEs such as tidal marshes and mangroves distributed in coastline areas can be influenced not only by sea-level rise but also by subsidence. Because subsidence in several deltas in the world has been reported [50], this study also estimated the effects of subsidence to SWEs in the corresponding model regions (S5 and S6 Tables). In most of the regions, the distribution was not changed even when subsidence was taken into account, but the distribution was predicted to significantly expand in Thailand, Vietnam, and China (Label 82). Mangroves additionally expand in India and Bangladesh (Label 14) and Pakistan (Label 32). In particular, our model showed that the distributions of tidal marshes and mangroves expand by almost 3.2 times and 3.7 times current levels, respectively, in Thailand, Vietnam, and China. It is assumed that tidal marshes and mangroves inhabit the intertidal zone on land side at the present; therefore, the expansion area offsets the disappearing area by moving landward with sea-level rise. However, as the SWEs shift toward the gentler landward slope with sea-level rise, which was relatively increased by subsidence, the areas expanded markedly (Fig 4b). Conversely, despite the large subsidence, due to sea-level rise by 2100 (S9 Fig), the area of SWEs did not markedly expand because the topographic gradient becomes steeper in a landward direction (Fig 4e). Thus, differences in SWE distribution between the regions can be great if subsidence significantly differs in each region. The subsidence data, however, contain much uncertainty, are generally lacking, and have a coarse spatial resolution [59].

If SWEs expand landward relative to sea-level rise, sea-side SWEs necessarily shrink. If subsidence also occurs in a region where sea-level rise occurs, such shrinkage can be accelerated. Thus, it is very important to include such shrinkage when making predictions of various hazards and considering countermeasures.

Coral habitats

Our projections showed that 30% of global coral habitats disappeared in RCP2.6 and RCP8.5 by the 2030s (Fig 1f). They expanded slightly in RCP2.6 after the 2030s, but 26% disappeared by 2100. In RCP8.5, coral habitats continued to decrease after the 2030s, and 74% was gone by 2100. A key threshold for coral habitat is SST during the warmest month of the year, as coral bleaching begins at 30°C. In RCP2.6, SST was largely unchanged, and more than half of the global coral habitat lay below 30°C (S13 Fig). In RCP8.5, SST continued to increase in all areas until 2100, and it was projected to exceed the 30°C threshold in more than half of the area by the 2050s.

By region, our results showed that coral habitats were significantly affected in lower latitudes, particularly in southeast Asia (Fig 6 and S7 and S8 Tables). Coral habitat was projected to rise only in the Mediterranean by 2100 in RCP8.5.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Projected global distribution of coral habitat.

a Present, b 2050s in RCP2.6, c 2090s in RCP2.6, d 2050s in RCP8.5, e 2090s in RCP8.5. Green, blue and red represent areas of no change, loss due to climate change and expansion due to climate change, respectively. The coastline data was obtained from Open Street Map (openstreetmap.org/copyright).

https://doi.org/10.1371/journal.pclm.0000298.g006

Although Frieler et al [15] predicted greater habitat loss than we did, our results are similar to those of Jorda et al [16]. Jorda et al [16] adopted a water temperature with an upper limit of 30°C, similar to the procedure used in our model, whereas Frieler et al [15] applied the degree heating month, which is the integral of the difference between the monthly mean SST and the reference warmest month SST. Thus, the gap between these projections may be caused by the difference in methods used, but the studies all project that coral habitats will shrink significantly in the future, so countermeasures that focus on ecosystem services and other factors may be necessary. Additionally, the shrinkage of ecological populations may limit species’ capacity to colonize new areas and offset losses in other areas, which may contribute to even greater shrinkage of coral habitats in the future.

Total shallow water ecosystems in each segment

We separated the future SWE distribution into six categories (S9 Table) and mapped the categories to visualize the regions of greatest vulnerability, least concern, and significant trade-offs (Fig 7). SWEs in Southeast Asia, Melanesia, the Mediterranean, and the Caribbean significantly decreased and were classified as categories I and II (Fig 7; S8 and S10 Tables). In Southeast Asia, Melanesia, and the Caribbean, the considerable shrinkage of coral habitat caused the decrease of total SWEs (S8 and S10 Tables). Although seagrass meadows simultaneously expanded in Melanesia and the Caribbean, the shrinkage of coral habitats was larger and more than offset the increase. In the Mediterranean, shrinking of seagrass meadows caused the decrease of total SWEs. Coral habitat was projected to rise in the Mediterranean in RCP8.5 by the 2050s, but the shrinkage of seagrass meadows was much larger. Thus, in the Mediterranean and three other regions (Southeast Asia, Melanesia, and the Caribbean), the ecosystem services of seagrass meadows and coral reefs could decline, making them the most vulnerable regions of global SWE habitat. On the other hand, West Africa was classified into category VI (Fig 7; S8 and S10 Tables) and Central America into categories VI (in RCP8.5 by 2050s) and V (in RCP8.5 by 2090s) (Fig 7c and 7d; S8 and S10 Tables), showing an expansion of total SWEs. The increase was caused by expanding seagrass meadows (S8 and S10 Tables), so the ecosystem services of seagrass meadows will increase in these two regions, indicating that they could be of less concern in terms of global SWEs.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Global category map of predicted total SWE distribution.

a 2050s in RCP2.6, b 2090s in RCP2.6, c 2050s in RCP8.5, and d 2090s in RCP8.5. Category I: total SWE shrinks as each ecosystem shrinks; Category II: overall total SWE shrinks, but shrinkage and expansion of ecosystems are intermixed; III: total SWE is sustained because each ecosystem is sustainable; IV: total SWE is sustained because the shrinkage and expansion offset each other; V: total SWE expands but shrinkage and expansion are intermixed; VI: total SWE expands as each ecosystem expands. One ecosystem (only one species was present) was not categorized, and macroalgal beds are not included because there were no available data. The coastline data was obtained from Open Street Map (openstreetmap.org/copyright).

https://doi.org/10.1371/journal.pclm.0000298.g007

Segments were most frequently placed in category I (i.e., a decreasing trend), followed by category III (a sustaining trend) (S14 Fig). The relationship between change in content of an ecosystem in total SWEs and the change in total SWE distribution was positively correlated for coral habitats, indicating that most of the total SWE shrinkage was caused by the shrinkage of the habitats (S15d Fig). Conversely, changes in the content of the other ecosystems (seagrass meadows, tidal marsh and mangroves) were negatively correlated to the change in total SWE distribution (S15a–S15c Fig). The relationship between changes in the relative content of each ecosystem in the total SWE distribution for respective pairs of ecosystems is shown in S16 Fig; clear negative correlations can be seen for seagrass meadows to coral habitat (S16c Fig) and mangroves to coral habitat (S16f Fig). This finding indicates that seagrass meadows and mangroves can expand or be sustained in segments where coral habitat will shrink. Coral reefs, seagrass meadow, and mangroves have many common ecosystem services, such as habitat for other marine organisms, coastal protection, and water quality [60–62]. Because seagrass meadows and mangroves have much higher CO₂ absorption than coral reefs [63], however, the values of their ecosystem services in term of climate change mitigation will increase in the future. This means that climate change will not necessarily have negative impacts on ecosystem services in each segment, but the projection assumes that there is no influence from hard infrastructure (this is particularly important for mangroves).

Macroalgal beds were excluded from this analysis because of inadequate global data. In future studies, the ecosystem services in each segment could be altered by considering macroalgal beds. If the impacts of hard infrastructure are also considered, macroalgae could create new habitat on the hard infrastructure while tidal marshes and mangroves could shrink as sea level rises, all of which would have an impact on ecosystem services.

Moreover, although various physical-chemical variables can alter the habitat of SWEs, many of these remain unknown at the present, so we projected the SWE distribution using only important parameters (seagrass and macroalgae: SST and sea level rise, tidal marsh and mangroves: sea level rise, and coral habitats: SST) in this study. In the future, it will be necessary to conduct projections considering more specific processes as research on the influences of other physical-chemical variables and other relevant data become available.