The site suitability analysis consisted of 1) ranking the coastal wetlands based on specific sub-models or categories consisting of hydrology, resilience to sea-level rise, soils, vegetation, bird habitat value, and community support and 2) carrying out the actual suitability analysis in ArcGISPro based on the sub-models in step 1). Rankings (1 (lowest)–5 (highest)) were based on the best available data from the rapid assessment, the literature, and database sources. Site data were very limited, so we relied on both professional judgement as well as the available data to develop criteria that best encapsulated the functions and values of the sites. Following are descriptions of the ranking criteria for each of the sub-models and the suitability analysis procedure that was used.

Hydrologic parameters were ranked in three ways and then averaged for an overall score. Depth to groundwater from the land surface was used to assess the availability of water in the rooting zone of common wetland macrophytes (approximately 0–30 cm in depth; Moffett and Gorelick, 2016). The ranking was as follows: <30 cm from land surface to groundwater = 5; 31–50 cm = 3; >50 cm = 1. The second measure was the range of estimated groundwater (gw) discharge at the coastline (Oki et al., 2019) and/or the number of springs mauka of the site (Stearns and Macdonald, 1947). The gw discharge/spring ranking was as follows: >500,000 gallons gw discharge/day and the presence of springs/seeps mauka of site = 5; 100,000–500,000 gallons of gw discharge/day and springs on site = 3 or high gw discharge at the coast but no springs on site = 3; low gw discharge (<500,000 gallons/day) and no springs on site = 1. The third hydrologic category was salinity stress, which was ranked as follows: <35 part per thousand (ppt) in gw = 5, 35–50 ppt in gw = 3, and >50 ppt in gw or extensive salt pan = 1. If groundwater was not accessible at the site, then the condition of the vegetation during the rapid field assessment was used as an indication of the hydrologic status because good condition during drought is indicative of good hydrological status.

The resilience of each wetland to future SLR was ranked in the following way. First, we determined the surface elevation of each site using the Molokaʻi digital elevation model provided by Sweet et al. (2022). For 2050, the median elevation of each wetland was then estimated by subtracting 0.29 m (the projected SLR, Table 2) and adding a conservative estimate for vertical accretion over the ensuing years (3 mm/yr * 28 years). The median elevation of each wetland in 2100 was estimated by subtracting 1.1 m (the projected SLR, Table 2) and adding vertical accretion at the same rate but over 78 years. Data were not available for the elevation range of dominant wetland vegetation, which is required to determine the elevation of marsh drowning (Morris et al., 2022). For this reason, we used the elevation capital of each wetland (sensu Cahoon et al., 2020) to rank its relative resilience to inundation. Rankings for both 2050 and 2100 were <0 m elevation relative to the mean sea level (MSL) = 1; 0–0.3 m = 2, 0.31–0.6 m = 3, 0.61–0.9 m = 4, and >0.9 m = 5. Wetland migration potential was then determined by the availability of flat land (which was available for all wetlands, so no ranking was needed) and the distance to the nearest mauka road. The rankings for the distance to the nearest mauka road were <100 m = 1; 100–250 m = 2, 251–400 m = 3, 401–550 m = 4, and >551 m = 5. Because the coastal lowland zone of Molokaʻi is very flat and migration space is limited, we weighted resilience to inundation to be twice as important as migration space in the overall ranking of resilience to SLR. Finally, because the overall resilience to SLR is of critical importance to wetland sustainability, we weighted this sub-model to be three times more important than all the other sub-models except community support, which also received the same weighting.

The soil ranking was based on two parameters, soil texture and % of organic carbon (OC) content. Soil texture is a key measure of water holding capacity and water availability, which are both essential for successful plant colonization (Sprecher, 2001). The percentage of OC (or organic matter content, which is ∼2*OC) is typically at least 5% in wetland soils and is essential for biogeochemical cycling of nutrients and adequate friability for plant growth (Mitsch and Gosselink, 2007, Ch. 13; Munkholm, 2011). For soil texture, the ranking criteria were as follows: silty clay loam and silt loam = 5; silty clay and loam = 3; and sandy loam and loamy sand = 1. For % OC, the ranking was simply >2.5% OC (∼5% OM) = 5 and <2.5% OC = 1. Both of these ranking criteria were averaged to obtain the final soil rankings for each of the sites.

Vegetation at each of the sites was ranked based on two main criteria: 1) the proportion of native vs. non-native wetland plant species and 2) the species richness of native wetland species. In Hawaiʻi, the high prevalence of invasive species has largely transformed lowland areas (Staples and Cowie, 2001; Erickson and Puttock, 2006), so the presence of native wetland species is a strong indicator of vegetation status and is key to the successful restoration of wetland function. The ranking for native vs. non-native species was as follows: the site contains intact native-dominated wetland plant communities = 5, a small area of the site contains native-dominated wetland plant communities = 4, the site is mostly dominated by non-native alien wetland plant species but with some natives = 3, the site is dominated by non-native wetland plants but no invasive mangroves (Rhizophora mangle L.) = 2, and the site has no native wetland plants and high cover of mangrove, pickleweed (Batis maritima L.), and other non-native wetland species = 1. For species richness, ranking was as follows: ≥ five native wetland species = 5, four native wetlands species = 4, three native wetland species = 3, one or two native wetland species = 2, and no native wetland species = 1. The two criteria were averaged to obtain the overall vegetation ranking for each of the sites.

The current and potential (restored) habitat value of each wetland for threatened and endangered (T&E) wetland birds and migratory birds was ranked using data from the scientific literature, wetland management plans, the rapid assessments, the National Wetland Inventory (NWI), the 10-year database (2010–2020) of bird observations on Molokaʻi by the Hawaiian Islands Conservation Collective (Supplementary Table S2), GIS databases, and expert opinions from managers and landowners on Molokaʻi (Pacific Coast Joint Venture, 2006; US Fish and Wildlife Service (USFWS), 2011a, 2011b; Chutz, 2014; Malachowski and Dugger, 2018; Malachowski et al., 2018; Sustainable Resource Group International, 2018; Henry and Fredrickson, 2019; Henry and Fredrickson, 2022; USFWS, 2022). Sites were scored using the 12 criteria in Table 3. A multiplier of two was applied to the ‘suitability for T&E and migratory birds after restoration’ expert scores, as this criterion was evaluated at all sites and provides a more important indication of potential future habitat value than past bird observations, which were not available at all the sites. To ensure that sites with severe hazards for avian species were distinctly noted, a multiplier of three was applied to this category. The rankings used for each criterion and the details on how ranks were determined are provided in Table 3.

The sixth sub-model was community support. This component is of key importance as without broad-scale support from the largely Native Hawaiian community on Molokaʻi, wetland restoration is unlikely to come to fruition or be successful in the long term. Community support was carried out in several steps. First, the project community liaison, Pūlama Lima of Ka Ipu Makani, held one-on-one meetings with key Molokaʻi community members. This was followed by a larger community meeting during the rapid assessment work in which the project was explained and there was an open discussion about the 12 sites chosen for prioritization. Next, a score sheet for community support regarding the sites was sent out by the community liaison to each member of the Molokaʻi Wetland Partnership (MWP), a coalition of local NGOs and state and federal agencies with extensive knowledge of the natural resources and socio-cultural landscape of Molokaʻi. A follow-up meeting was then held to discuss any scores, which were in dispute or not well understood. In this way, each site was ranked as follows: likely community opposition (1), possible community opposition (2), community neutral (3), community supportive (4), and community highly supportive (5). Due to the overarching importance of this sub-model, it received a weight three times that of all the other sub-models except for sea-level rise, which also received the same weight.

All six sub-models were used to carry out the site suitability analysis in ArcGISPro version 3.0.2. Site suitability analyses have been used in a broad range of applications including wind farm placement, refugee camp establishment, and wetland restoration (White and Fennessy, 2005; Darwiche-Criado et al., 2017; Nasery et al., 2021; Younes et al., 2022) to identify locations with desired characteristics by using weighted criteria defined by the analyst to rank and score potential sites based on how closely they match an ideal set of criteria. In order to prepare the data for the analysis, several steps first needed to be followed. Study area boundaries were manually delineated using high-resolution digital imagery (EagleView Technologies, Inc., 2008) and converted to polygons in the GIS using the NAD 1983 UTM Zone 4N projected coordinate system. Feature classes were created in the GIS for each of the datasets collected in the rapid field assessment. For datasets that contained null values, a proxy was assigned to prevent ArcGIS Pro from excluding the site from the analysis. Each dataset was then spatially joined to the wetland site polygons. Each data attribute was then converted to a raster (cell size of 3 m) with the Polygon to Raster tool. Due to the soil data being collected at two pits within each study area boundary, single rankings for each site had to be calculated for use in the final model scenarios to match the study area boundaries of the other sub-models. After the soil sub-model was run at the pit level, the ranks for each site were determined by averaging the soil pit ranks. The new ranks were then assigned to their respective study area boundary polygons and then converted to a raster (cell size of 3 m) with the Polygon to Raster tool. All the raster data were then uploaded to the ArcGIS Pro (version 3.0.2) Site Suitability Modeler tool.

Because some criteria had higher importance than others, the multiplier weighting method was used (White and Fennessy, 2005) in the site suitability analysis to weight the criteria relative to one another within each sub-model and in the overall suitability model. Any weighting within the individual sub-models was already described above. Weighting among the sub-models for the overall site suitability analysis was carried out in three ways: 1) equal weighting for each sub-model, 2) SLR and community support weighted 2x as high as all the other sub-models, and 3) SLR and community support weighted 3x as high as all the other sub-models. All other parts of the suitability analysis remained the same for the three different runs.

During rapid assessments, we determined that all the wetland sites receive at least some groundwater inflow. Because many of the springs are located outside of the wetland boundaries of our sites, we relied on the spring survey by Stearns and MacDonald (1947), which is the most recent to date, to identify springs mauka from our sites. Additional springs/seeps were noted by us at the upland edge of MOKU (Table 1 for site names and their abbreviations). For mauka of KOK, there must be springs or seeps as the groundwater discharge was the strongest of all the sites. However, none were found by Stearns and MacDonald (1947), suggesting either an oversight or a new flow path. Streamflow is intermittent at all sites except PUNA, MOKU, KOK, IPUKA, and KAIR, which have no streams (US Geological Survey, 2019). The measurements of depth to groundwater and salinity illustrate the impact of the ongoing drought on the lowland coastal zone of Molokaʻi (Figure 3). Although the water table is usually near the surface in coastal wetlands, the mean depth to groundwater at the sites was 36.36 cm ± 25.00 (sd). At three soil pits, PAIA1, POHO1, and POHO2, the bottom was dry at 80 cm (Supplementary Table S3). The mean groundwater salinity at all the sites was 30.35 psu ±34.25 (sd) and at soil pits KAMA1, KAUN, OHI2, and PAHUA2 salinities were in the hypersaline range (>40 psu) (Figure 3B). In contrast, the low salinities and shallow depths to groundwater at KOK1, KOK2, KNWR1, OHI1, and IPUKA demonstrate a high discharge of fresh groundwater at several places along the southern coastline (Supplementary Table S3, Figure 3C). The wetland sites that ranked the highest overall (5) for hydrology were IPUKA, KNWR, KOK, KAIR, and MOKU (Figure 3C, Supplementary Table S3).

In the analysis of wetland resilience to future sea-level rise, there were major differences in the migration space and elevation capital among the sites. The mean distance to the nearest upland road (167 m ± 173 (sd)) ranged from 9 m at KOK to almost 600 m at PAHU (Figure 4A). The resilience of the wetland sites to sea-level rise was strong in 2050, as almost all the sites maintained their elevation above MSL except for KAIR (Figure 4B). In 2100, resilience waned with 10 out of 12 sites having projected surface elevations below MSL, which is likely near the lower end of the vegetation elevation range (Figure 4B). Overall, the highest overall sea-level rise rankings (3 or higher) were found for KNWR, KOK, and PAHU (Supplementary Table S4; Figure 4C).

Basic soil characteristics varied broadly across the sites. Soil texture ranged from loamy sands to silt and silty clays (Supplementary Table S5). The mean % OC across the wetland sites was 10.93% ± 12.24 (sd) (Figure 5A). IPUKA, which had no soil per se, had a floating mat of vegetation estimated to contain ∼44% OC (Byrd et al., 2018). Most sites contained mineral soil except KOK and PUNA, which have organic soils due to their high % OC content (Figure 5A) (USDA-NRCS, 2014). All of the sites except for KOK and IPUKA are situated under reddish brown steep slopes described by gSSURGO as “very stony land-eroded” (Soil Survey Staff, 2021). Reddish brown material was seen in the top layers of many soil pits and the coastal waters, demonstrating chronic erosion and transport of materials from the uplands to the sea. Six sites had overall soil rankings above 4 (Figure 5B).

The vegetation survey demonstrated that the coastal flora on Molokaʻi contains a high proportion of non-native species dominated by mangroves, pickleweed, and Indian fleabane (Supplementary Tables S6–S10). There were 39 plant species found overall at the wetland sites. Of this total, 22% or 13 were not wetland species. Of the 26 wetland plant species, 11 were native, one was introduced by Polynesians, and 14 were non-native (Supplementary Table S7). The most diverse plant life form was trees. However, seven of the nine tree species were exotic (non-native) in origin (Supplementary Table S7). Sedges were the next most numerous plant form (seven species), with three native species. Overall, the non-native wetland plant species richness (mean = 3.33 ± 1.92 (sd)) at the wetland sites was slightly greater than the native wetland plant species richness (mean = 2.08 ± 2.02 (sd)) (Figure 6A, Supplementary Table S8), but the difference was not statistically significant (Student’s t-test, p > 0.05). The number of native species and/or the proportion of native to non-native wetland species resulted in KAIR, KOK, and OHI ranking the highest overall for vegetation (all >4; Figure 6B, Supplementary Table S9).

The assessment of bird habitat value, which consisted of 12 separate criteria (Table 3), ranged from a total score of 55 at POHO to 105 at OHI (Table 4, Supplementary Table S11). It is important to note that PAIA was not included in the rankings due to strong community opposition to restoration of the site (see below). The criteria with the highest consistent rankings among the sites were potential suitability for T&E birds (mean = 3.5) and managed human disturbance and hazards (means of both = 3.4). The criteria with the lowest scores across the sites were restoration status (mean = 1.9) and adequate predator control (mean = 2.1) (Table 4, Supplementary Table S11). Four sites (KNWR, KAMA, PAHU, and PUNA) ranked above “4” in “potential suitability after restoration for all threatened and endangered waterbirds, migratory shorebirds, and migratory waterfowl” (Table 4). The top ranking sites based on suitability for birds were OHI (5), PUNA (5), KAMA (4), PAHU (4), and KNWR (4) (Table 4).

The local community expressed support for coastal wetland restoration at eight of the 12 wetland sites (Figure 7). The site rankings were particularly high (rank of 5) for KNWR, KAMA, and KOK. This high level of support was indicative of a broad, shared community vision of restoration at these sites. In contrast, for PAIA, there was no shared vision for restoration and, thus, no discussion of support due to past trauma from island development efforts, issues around maintaining access for hunting and fishing, and a lack of trust regarding state and federal agencies (Figure 7; Supplementary Table S12). The main concerns raised about wetland restoration included maintaining access for hunting and fishing, protecting sites with iwi kupuna (ancestral bones), the need to include the community in the planning and restoration process, and a desire to see indigenous agro-ecology (loʻi kalo and loko iʻa) included in wetland restoration planning (Supplementary Table S12).

All of the sites receive at least some groundwater inflow. However, erosion from the uplands has resulted in some springs and seeps being at least partially buried under deposited materials (e.g., MOKU and KNWR) and fine silts being carried across the wetlands to nearshore environments (Stock et al., 2011). At some of the sites, the restoration process will need to include the excavation of springs and regrading of the wetland surface so that the site can sit closer to the water table. The degradation of springs from sedimentation and drought has caused salinity levels to climb to the hypersaline range in four of the sites (KAMA, KAUN, OHI, and PAHU; Figure 3B). This situation will need to be remedied in order to create conditions conducive for native plant recruitment (Zedler, 1982; Alexander and Dunton, 2002). Half the sites had % cover of bare ground greater than 10% (Supplementary Table S7) and in most of these sites, there was precipitated salt on the soil surface (Supplementary Figure S1). Despite these impacts, however, half the sites had soil texture and % OC highly suitable for restoration (a ranking of “4” or above, Figure 5B).

At several sites, the water quality was fresh due to high groundwater discharge along the coastline (Figure 3). Previous work on Molokaʻi has shown that a great volume of groundwater travels along flow paths from the ridge to the reef (Oki et al., 2019). As a case in point, groundwater salinity at the KOK soil pits, practically adjacent to the sea, were both 0.5 psu, which is indicative of a freshwater ecosystem (Cowardin et al., 1979). The ubiquitous presence of groundwater resulted in half the sites scoring a “4” or higher in their overall hydrologic ranking (Figure 3C).

The results of the sea-level rise analysis showed that the elevation capital of the microtidal coastal wetlands on Molokaʻi appears to be modest. Most sites are likely resilient to SLR until 2050, but after that, resilience will likely wane dramatically (Figure 4, Supplementary Table S4). It is important to note that these projections for sustainability under sea-level rise have large uncertainties. Due to a major deficit in wetland data such as the vertical accretion rates of sediments and peat, elevation ranges in the tidal frame suitable for dominant vegetation, and primary productivity of coastal wetlands, no marsh sustainability modeling could be performed. Furthermore, the sea-level rise estimates used were for passive flooding only and, thus, did not include wave-induced flooding, yearly variability of sea-level from natural processes including the El Niño–Southern Oscillation, or any additional impacts due to coastal erosion, which can all increase the extent and depth of flooding (Drexler and Ewel, 2001; Anderson et al., 2018; Goodman et al., 2018).

Despite these limitations, the SLR analysis raises some important points about the state of knowledge concerning coastal wetlands on Molokaʻi. First, there are some key opportunities (e.g., KAMA and PAHU) for marsh migration (Figure 4) on the island. Second, there is a great need to improve the process-level understanding of coastal wetlands of Hawaiʻi. In particular, greater knowledge of the two components of vertical accretion, organic matter accumulation and inorganic sedimentation rates, is needed as the first step toward improving projections of coastal wetland resilience.

Non-native, invasive plant species were ubiquitous at the sites, particularly mangroves and pickleweed (Table 1). Mangrove cover was ≥36% at KAMA, KAUN, PAHU, PAIA, and PUNA. Monotypic cover of pickleweed was ≥26% at KAMA, KAUN, OHI, PAHU, POHO, and PUNA (Supplementary Table S6). Invasive vegetation will need to be controlled and native species will need to be planted in order to regain some of original plant diversity. Only three sites (OHI, KAIR, and KOK) scored “4” or higher in the vegetation rankings, demonstrating that even having a handful of native wetland species is a high bar (Supplementary Table S7). Overall, plant species richness across all the sites was low (39 species), with only 11 native wetland plants (Figure 6; Supplementary Tables S7–S9). This may be due to competition from invasive plants, or similar to native Hawaiian forests, low plant diversity may have always been the case due to the extreme isolation of the Hawaiian archipelago (Barton et al., 2021).

Although proper hydrology, vegetation, soils, and resilience to SLR are critical for restoring tidal wetlands, these basic wetland components are insufficient to adequately support sensitive wildlife. Endangered species, such as the Hawaiian stilt, Hawaiian coot, and Hawaiian duck, are conservation reliant (Harmon et al., 2022), requiring special protections against predators, avian botulism, the encroachment of invasive plant species, and human disturbance (Greer, 2005; US Fish and Wildlife Service, 2011a; Underwood et al., 2014; Malachowski et al., 2022). Predator control, in particular, is of critical importance to protect endangered waterbirds from a suite of non-native species including cats (Felis catus), dogs (Canis familiaris), pigs (Sus scrofa), mongooses (Herpestus javanicus), Black, Norwegian, and Polynesian rats (Rattus rattus, Rattus norvegicus, and Rattus exulans), bullfrogs (Rana catesbeiana), introduced barn owls (Tyto alba), cattle egrets (Bubulcus ibis), and the common myna (Acridotheres tristis) (Greer, 2005; US Fish and Wildlife Service, 2011a; Underwood et al., 2014). Our analysis showed that breeding and bird counts were low for endangered waterbirds across the sites (Table 4). Furthermore, several of the top rated sites for bird habitat (OHI, PUNA, KNWR, KAMA, and PAHU) had low ratings for predator and botulism control due to a lack of management (Table 4). Although even simple ponds can provide habitat for endangered waterbirds (including wastewater treatment ponds containing raw sewage (KAUN)), unmanaged sites can act as ‘sinks’ with high mortality, especially for chicks, due to the abovementioned threats (Christensen et al., 2021; Harmon et al., 2021). The recovery of endangered waterbirds in Hawaiian wetlands depends on restoring and expanding high-quality habitat combined with strong protections from predators, hazards, and diseases.

Public support and local stakeholder involvement have been shown to be critical for conservation efforts, restoration projects, and sustainable resource management (Lee, 2011; Doyle-Capitman et al., 2018; Wilkins et al., 2018). We engaged with the Molokaʻi community at multiple times in our study in order to achieve coproduction of knowledge (Meadow et al., 2015). This process relied strongly on our community liaison and the Molokaʻi Wetland Partnership to effectively engage with residents. This effort allowed us to gain valuable insights about community support regarding our sites. For example, in the beginning of our work, we considered PAIA to be a strong contender for restoration, but as mentioned above, the failed development efforts at this site caused deep trauma for local residents, causing them to omit PAIA from consideration. Ultimately, PAIA received very low scores overall (Table 5). Knowledge of site history and perspectives on “landscape frame,” the environmental vision for the landscape of a person or community, are essential for truly understanding community preferences for restoration (Aggestrom, 2014). Furthermore, there must be strong consideration of local community needs (Doyle-Capitman et al., 2018), which for Molokaʻi include hunting, fishing, and socio-cultural resources.