This study trials the Key Biodiversity Area (KBA) approach to conservation planning on a wide-ranging marine species. We used marine turtle nesting data to test the global thresholds and criteria devised for terrestrial systems and adapted to the marine environment. Data were collated for the Melanesian region and used in the first region-wide marine application of the KBA approach. Using the standard criteria, we tested a range of population thresholds to identify and delineate KBAs for five marine turtle species. A series of decision rules were devised during this analysis specific to marine turtles. The study showed that the standard KBA thresholds could be applied to marine turtles in Melanesia, with some modifications. Using the threshold of 10 breeding females, 54 preliminary marine turtle KBAs were identified and delineated for six Melanesian nations. A threshold of 1% of a species genetic stock aggregating at a site was trialled and showed that it made no difference to the identification of KBAs, with the exception of Lepidochelys olivacea. The study demonstrates that this approach to conservation planning, which is most suited to site-associated species, can also be applied to wide-ranging species displaying seasonal congregatory behaviour, such as marine turtles.
The Key Biodiversity Area (KBA) approach identifies sites of global significance for biodiversity conservation (Eken et al., 2004; Langhammer et al., 2007) through the application of global standards and thresholds. This approach uses quantitative criteria that can be consistently applied using available biological and point occurrence data; it is appropriate for species that will benefit from conservation and management actions at the site scale (Eken et al., 2004). KBA sites are identified and prioritized within a regional or a national context where global priorities may otherwise be overlooked because of political boundaries and national or local economics. The use of the KBA process to identify sites of global significance ensures that action at a local level is in line with international conservation efforts aimed at reducing biodiversity loss.
The KBA approach is based on quantitative criteria established three decades ago by BirdLife International for the designation of Important Bird Areas worldwide, and adapted to include other species groups by a number of organizations, including Plantlife International, IUCN and Conservation International. Over the past few years, KBA criteria originally developed for terrestrial systems have been modified to make them applicable to the marine environment, where greater connectivity, faster turnover rates and three-dimensional habitats provide additional challenges for site conservation (Day & Roff, 2000). Provisional criteria and thresholds have been adapted for marine KBAs (Table 1), but these require testing before being standardized for all marine species (Langhammer et al., 2007). The process for marine KBA identification is described in detail by Langhammer et al. (2007) and further by Edgar et al. (2008b). The first marine KBA analysis was undertaken in the Galapagos Marine Reserve on a range of threatened and restricted range taxa. However, wide-ranging species, including marine turtles, were excluded from this case study as vagrants (Edgar et al., 2008a). Provisional thresholds and criteria were proposed in the Galapagos study, but the authors noted that field testing is required to ensure that the number of KBAs identified is appropriate.
Table 1. Criteria and thresholds provisionally considered for the identification of marine KBAs (from Langhammer et al., 2007)
Provisional thresholds for triggering KBA status
KBA, Key Biodiversity Area.
Vulnerability Regular occurrence of a globally threatened species (IUCN Red List) at the site
Regular presence of a single individual for Critically Endangered (CR) and Endangered (EN) species; regular presence of 30 individuals or 10 pairs for vulnerable (VU) species
Irreplaceability Site holds x% of a species' global population at any stage of the species' lifecycle
Restricted-range species Species with highly restricted global ranges.
Species with a global range less than 100 000 km2; 5% of global population at site
Species with large but clumped distributions Species with highly clustered distributions.
5% of global population at site
Globally significant congregations Species that temporarily aggregate in particular sites.
1% of global population seasonally present at site
Globally significant source populations Species with small sub-populations that are responsible for generating a significant proportion of recruitment (‘source’ species).
Site is responsible for maintaining 1% of global population
KBAs are triggered by the presence of species that meet the irreplaceability or vulnerability criteria, the premise being that the loss of any of these species is a loss of global significance. Irreplaceability and vulnerability are principles used in systematic conservation planning, a data-driven process with set criteria and clear targets for achieving conservation goals (Margules & Pressey, 2000). The sites of highest conservation urgency are those with high irreplaceability (a measure of uniqueness) and high vulnerability (a measure of threat). For KBAs, a site meets the vulnerability criterion if it holds a species with a high likelihood of becoming extinct in the short to medium term. These are species listed as Critically Endangered (CR), Endangered (EN) or Vulnerable (VU) on the IUCN Red List (IUCN, 2001, 2009). A site meets the irreplaceability criteria for KBA selection if it holds a significant proportion of the global population of a species at some point in the species life cycle (Eken et al., 2004; Langhammer et al., 2007). Identifying and protecting a KBA network based on the presence of species meeting the irreplaceability or vulnerability criteria ensures that sites critical to maintaining the global population of species are not lost.
Many marine species are not suitable for KBA selection because of their pelagic or wide-ranging nature. The species most likely to benefit from the identification of KBAs are those that are habitat forming, site associated or that occur regularly at a site for part of their life cycle, for example for breeding, feeding or migration. Marine turtles present an interesting case study, because they travel widely, but also aggregate during their breeding season and nest on shore.
Marine turtles generally nest every 2–3 years (depending on species), with females returning to the region of birth to lay eggs (Miller, 1997). They tend to nest along the same beach during subsequent nesting attempts within a season (Miller, 1997). This predictable nesting behaviour means that conservation efforts can be focused on the nesting beach to reduce mortality of nesting females, eggs and hatchlings. These nesting beaches are critical to maintaining the global population of these globally threatened species; hence, they meet the vulnerability criterion for KBA selection (Table 1).
This study presents the identification and delineation of KBAs for the five species of marine turtle that are known to nest in the Western Pacific region (Melanesia): leatherback Dermochelys coriacea, loggerhead Caretta caretta, green Chelonia mydas, hawksbill Eretmochelys imbricata and Lepidochelys olivacea. All of these species are globally threatened (IUCN, 2009); D. coriacea and E. imbricata are Critically Endangered (CR), Ca. caretta and Ch. mydas are Endangered (EN) and L. olivacea is Vulnerable (VU). The confirmed regular presence of Ca. caretta and Ch. mydas or 10 breeding pairs of L. olivacea fulfills the vulnerability criterion for KBA identification. The aim of this analysis is to use data on marine turtles in Melanesia to test the proposed marine KBA thresholds. By testing these thresholds, we generate a preliminary set of marine turtle KBAs for Melanesia, which would require validation and refinement within each country.
The general methods for identifying and delineating marine KBAs used for this analysis follow those detailed in Edgar et al. (2008b) and Langhammer et al. (2007). We tested species-specific thresholds for KBA identification in this study. Rules for KBA delineation were also developed, based on marine turtle ecology and the management context in Melanesia. While the decision rules developed are presented below, they constitute an important result of this analysis and are also presented in Appendix S1.
The first stage in the KBA process is identifying trigger species that meet the irreplaceability or vulnerability criteria. Seven species of globally threatened marine turtles are listed on the IUCN Red List of Threatened Species. Five of these are known to occur and nest in Melanesian countries (Papua New Guinea, the Papuan provinces of Indonesia, the Solomon Islands, Vanuatu, New Caledonia, Fiji) and therefore meet the vulnerability criterion.
The second stage of KBA identification is collation of data layers. KBA identification requires a comprehensive synthesis of distributional data for trigger species. Biological and occurrence data were sourced from available published and unpublished literature, reports, surveys and the UNEP-WCMC Marine Turtle nesting sites for Melanesia database (extracted from the database 24 January 2007). The dataset is limited by the availability of data, and while every effort has been made to make the dataset as comprehensive as possible, we cannot guarantee that all nesting survey data for this region has been included in this study.
Compiling point locality data is the basis of KBA identification, but care must be taken, given that data points are often only loosely geo-referenced. The points do not always represent precisely where the species was observed or collected. Points may have coordinates recorded to the nearest degree making them inaccurate to several kilometers. Therefore, we refined point localities if further information about the point could be obtained from satellite imagery, first-hand knowledge or from the location description in the literature. However, points were discarded if no coordinates were provided and the location was not known.
The KBA guidelines (Langhammer et al., 2007) caution against the use of historical records, as habitat may no longer exist or location names may have changed. In the case of marine turtles, the decline in populations in recent decades is more of a concern, because nesting populations of turtles may be locally extinct or so reduced as to no longer be significant. Some turtle species have undergone a decline of 70% or greater in the global population in one to three generations (Limpus, 1994, 1995; Limpus et al., 2001; IUCN, 2009). The generation time of marine turtle species can range from 13 to 50 years (Limpus, 1992; Zug & Parham, 1996; Chaloupka & Musick, 1997; Limpus & Miller, 2000); therefore, a threshold of 40 years will ensure that at least one generation is included. Also, most of the intensive nesting surveys in Melanesia only began in the early 1970s. Thus, any records before 1970 were considered historical and omitted from this analysis.
All data points and relevant biological information and management layers were entered into a Microsoft Access database. Contextual layers, including bathymetry, reef and seagrass and validated locality data points were entered in the gis software, ESRI ArcGIS 9.1. With this information, KBAs were identified for marine turtles. The standard method to define KBAs was followed and clarified by adding decision steps applicable to marine turtles (Appendix S1).
Points were verified and nesting beaches defined with arcs, based on the description and satellite imagery. An arc was drawn around the entire island where islands were under 2 km in diameter and where the nesting beach was unknown. This decision rule was based on series of trials on variously sized islands. This process forms the basis of the results.
The KBAs identified in this study are listed in Appendix S2; however, the coordinates for the sites are not given for conservation reasons. These data are available to legitimate researchers upon request.
Turtle nesting data
There is a high level of uncertainty and inconsistency in adult turtle nesting records due to the diverse methods of data collection. Data can take the form of actual counts of tagged or sighted adults, estimates of nests, crawls or nesting adults. Estimates may be made from anecdotal evidence, seasonal nesting populations, surveys done over 1 day, a number of days or weeks, one-off sightings during an aerial survey or island visit, or from intensive annual nesting surveys. Some of the major nesting grounds have had annual surveys undertaken, while other sites have only been surveyed once or several times over a period of decades. It is important to note that where numbers of nests per season are given, the population could range by a factor of five, based on the assumption that a female, on average, nests five times per season (Spotila et al., 1996). Dutton et al. (2007) illustrated the problem of drawing conclusions on population status from estimates of numbers of nesters derived from nest counts. Therefore, where possible, the number of nesting females recorded was entered for this analysis. Total nesting populations were not derived from the raw data.
Thus, while our analysis does not try to provide population size estimates for turtle species, it does provide a relative indication of the significance of nesting female turtle populations at each site. Turtle abundance shown on the maps was grouped into categories, some of which were quite large, to accommodate inconsistent data recording.
Clarification of thresholds
Strict application of the vulnerability criteria results in the identification of KBAs where we have the confirmed presence of one or more individuals of EN or CR species, and provisionally 10 breeding pairs or 30 individuals for VU species. Preliminary analyses showed that the inclusion of all records for CR and EN species resulted in an excessively high number of KBAs for Melanesia, given that the five species that occur in this region have much larger rookeries in other parts of the world. Thus, the use of more strict thresholds is critical to identifying and conserving nesting sites that are globally significant.
We then looked again at the thresholds under the vulnerability criterion. Through mapping different selections of nesting data to identify preliminary KBAs, we refined the standard thresholds to include a definition of vagrancy. After a number of trials, we formulated the decision rule that sites with fewer than 10 nesting females annually represented vagrant individuals that do not contribute significantly towards maintaining the global populations of marine turtle species. We excluded such records from the KBA analysis, a decision based largely on data quality. Because of the inconsistent nature of turtle survey variables, many datasets present a range of one to 10 to reflect the difference in turtle track or nesting data, as compared with adult counts (e.g. one female may nest three to eight times in a season [Dodd, 1988; Addison, 1996; Broderick et al., 2002)]. However, we recognize that by excluding these data, it is also possible that significant nesting beaches are missed due to poor or one-off surveys.
One of the challenges of the KBA process pointed out by Edgar et al. (2008b) is in determining a threshold to ensure that an excessively high number of KBAs is not recognized for widely distributed threatened species by using every confirmed record. They recommended that KBAs should not be recognized for wide-ranging EN and VU species that are well represented in existing KBAs, unless at least 1% of the global population is present at a site. We test this 1% threshold for EN and VU species to see how it compares to the number of KBAs defined using the vulnerability criterion.
Once identified, boundaries for KBAs were drawn in ArcGIS. As with terrestrial KBA delineation, as far as possible, each site should:
•be different in character, habitat or conservation importance from the surrounding area;
•exist as an actual or potential protected area, or be an area that can be managed in some way for nature conservation;
•be alone or in combination with other sites, provide all the requirements of trigger species during the time they are present (from Langhammer et al., 2007).
Thus, we examine both biological and management data in delineating KBAs for marine turtles. During this analysis, we determined a series of decision rules to guide site delineation. Sites should be biologically sensible, in that they include sufficient area for nesting and inter-nesting habitat for marine turtles. However, they also need to be practical in terms of management for conservation. The size of a KBA can vary depending on the habitat and needs of trigger species and the scale of local management areas. If there are existing marine-protected areas (MPAs), then KBA boundaries will generally be equivalent to the MPA boundary. The size of the KBA may be dependent on a proposed management area or extent of the habitat required to safeguard the trigger species. Where possible, boundaries should follow depth contours, habitat edges or existing MPAs.
Where no biogeographical features are evident, an offshore boundary of 3 miles (5 km) is drawn from the beach, which is generally promoted in Papua New Guinea and the Solomon Islands as an appropriate dimension for local marine tenure (J. Kinch, pers. comm., 2008). However, anecdotal information suggests that 4 miles (6 km) may be more appropriate in Indonesia. If the extent of nesting on a beach is not known, a 5 km circular buffer is drawn from a central point on the beach.
We devised rules regarding the size of the nesting beach or island and the proximity to other KBAs based on case studies. Where an island was <2 km across its longest diameter, we included the entire island as a nesting site. The determination of a 2-km-diameter cut-off point was based on an average measurement of small islands with potential nesting beaches around the entire perimeter. The accuracy of data records was a factor in deciding on this distance, as coordinates were often not accurate, and so presumably nesting could occur anywhere on a small island with suitable beaches. In such cases, the KBA boundary was extended to the reef edge. Where there was no reef or habitat data, or where the island was one of an island chain in a large reef complex, a 3 mile (5 km) buffer was drawn around the island based on marine tenure (to account for manageability).
In the case where one or more KBAs were in close proximity to each other (e.g. a series of beaches along a coast or islands in a reef complex), decision rules were developed regarding when these sites should be joined. Where KBAs were within the distance of the length of one KBA, they would be joined to form one larger KBA. Where the boundaries of two or more buffers overlapped, in the case of islands in a reef complex, a simple polygon was drawn around both buffers to form the KBA boundary. Where an existing MPA overlaps with a potential KBA, the KBA boundary is aligned with that of the MPA; however, in cases where the MPA is extremely small, it is included in its entirety.
Preliminary marine turtle KBA selection was sent to seven experts and researchers following the presentation of these results at the 29th Symposium on Sea Turtle Biology and Conservation (February 2009). From this review process, some data gaps were filled and KBA delineations were refined. The best available data at the time was used for this analysis, but the authors recognize that it may not include all turtle survey data recorded for Melanesia.
Management units (MU) for marine turtles
Based on feedback from the expert review process, we tested the KBA results against the concept of marine turtle MUs or genetically discrete populations (Moritz et al., 2002). The term ‘management unit’ is used to denote genetically distinct populations that represent a logical focus for conservation planning and action. Breeding populations have a different genetic make-up and are not likely to exchange individuals, and so they will respond differently to threats and management actions. Research on green turtles in the Indo-Pacific shows that groups of adjacent rookeries isolated from other rookeries by a few hundred kilometres can be expected to support a genetically distinct MU, and where a chain of adjacent rookeries extend over a large geographical area, the entire area can be assumed to represent a single MU (Dethmers et al., 2006). The study by Moritz et al. (2002) recognizes 17 different MUs for marine turtles in South-East Asia and the Western Pacific; therefore, these genetic boundaries need to be considered in the delineation and prioritization phase of the KBA process.
Identification of KBAs
Strict application of the vulnerability criterion suggests that the regular presence of a single individual of a CR or EN species or 10 breeding pairs of a VU species is enough to trigger a KBA. To test this threshold to see whether it resulted in a manageable number of KBAs for marine turtles in Melanesia, we trialled a range of thresholds (Table 2). These represent total numbers of single-species KBAs. When summed, they represent total numbers of species–site relationships rather than true KBAs, given that, in many cases, there will be overlaps where multiple species nest within the same KBA and where adjacent KBAs are grouped. Given this potential for overlap, the final number of KBAs depends in part on site delineation (discussed below).
Table 2. Number of KBAs in Melanesia for each species using minimum thresholds of 10, 20 and 50 nesting females
KBA, Key Biodiversity Area.
Dermochelys coriacea KBAs
Eretmochelys imbricata KBAs
Chelonia mydas KBAs
Caretta caretta KBAs
Lepidochelys olivacea KBAs
Total number of species–site relationships
By excluding sites with fewer than 10 nesting females as representing instances of vagrancy, the number of species–site relationships is narrowed from 257 to 88. If we were to apply the threshold of 50, a smaller number of species–site relationships would result (57). This simple comparison shows that for D. coriacea, Ch. mydas and E. imbricata, the majority of nesting sites in Melanesia appear to have fewer than 10 nesting females; thus by excluding vagrants, the number of KBAs is reduced significantly. The number of KBAs for Ca. caretta and L. olivacea is not affected considerably by the range of thresholds tested. A threshold of >50 only makes a noticeable difference for D. coriacea, reducing the number of KBAs by almost half. This does not, however, account for sampling errors or data paucity, such that there is some risk of omitting potentially significant sites.
The extent of all turtle nesting records in Melanesia compared with those identified as KBAs is shown in Fig. 1. The nesting abundance is indicated by the size of the dots. It appears that while the number of nesting sites may be large for some species, they are not necessarily highly significant sites, and in many cases likely reflect the presence of vagrant individuals.
Delineation of KBAs
While methods outlined in Langhammer et al. (2007) for delineating KBA boundaries were followed, a series of decision rules specific to marine turtles were formulated (discussed in ‘Methods’ and outlined in Appendix S1). Examples of the application of these decision rules include merging neighbouring sites (Fig. 2) and merging sites with overlapping buffers (Fig. 3). KBA boundaries developed during this analysis are provisional and will be refined through in-country processes and through the inclusion of new data.
A total of 54 marine turtle KBAs were identified for Melanesia once vagrants were excluded, overlapping spatial units were accounted for and adjacent sites were merged (Fig. 4 and Appendix S2). Of these 54 KBAs, 15 were identified as being KBAs for multiple turtle species.
We calculated the 1% threshold of nesting females required to identify a KBA for EN and VU species (Ch. mydas, Ca. caretta and L. olivacea) that do not occur at a multi-species KBA (Table 3), based on a previous study by Edgar et al. (2008b). They recommended from trials in the Galapagos that if wide-ranging EN and VU species are well represented in existing KBAs, a 1% threshold of the global population should be present at a site for it to be identified as a KBA.
Table 3. Calculation of 1% population threshold for known genetic breeding stock populations in Melanesia
Marine turtle species
Management units located in Melanesia
Estimated population of a management unit
1% threshold of MU
Most recent records of adult females nesting at a single site
Based on current global estimates for L. olivacea (Limpus, 2008), 1% of the population would require a site to hold up to 10 000 nesting female turtles; therefore, no significant nesting sites would be identified for L. olivacea in Melanesia using this threshold. A population of c. 100–1500 at a site would be required for Ch. mydas (National Marine Fisheries Service, 2007a), 340–940 for D. coriacea (National Marine Fisheries Service, 2007b), 1200 for Ca. caretta (National Marine Fisheries Service, 2007c) and 100 for E. imbricata (Meylan & Donnelly, 1999). This would result in potentially 11 individual sites in Melanesia being identified: five for Ch. mydas, four for D. coriacea and two for E. imbricata. However, this global approach is not suitable for marine turtle species because they may not all be of the same genetic stock, and current global population estimates may be inaccurate to the order of thousands.
Calculating the 1% threshold for EN and VU turtle species in a MU provides some interesting results (Table 3). Chelonia mydas has three MUs in Melanesia: Aru Island, Long Island and the east coast of New Caledonia. KBAs were identified at all three sites, and so applying the 1% threshold does not alter the identification of these KBAs as each of them holds the total genetic stock for that breeding population. This is also the case for Ca. caretta, as two KBAs were identified in New Caledonia.
Lepidochelys olivacea is an exception; it has two distinct MUs, with Melanesia forming part of the Indo-West Pacific MU. Two KBAs were identified in this area, both in Indonesia. Approximately 500 nesting females were recorded along the coast of Papua, representing 0.25% of the Indo-West Pacific MU. This implies that the Papuan L. olivacea population may not form a significant part of the Indo-West genetic stock or global population. This suggests that a 1% threshold may be more appropriate for this VU species.
The KBA process provides a mechanism for marine sites of global significance to be systematically identified and given priority for conservation action (Edgar et al., 2008b). This global perspective to conservation planning is most appropriate for site-associated species. However, this study demonstrates that by the inclusion of additional definition and preliminary delineation rules, this methodology can also be applied to wide-ranging species displaying seasonal congregatory behaviour, such as marine turtles. A comparative study using the same methods in a relatively data-rich region with generally larger rookeries would be very useful; the Caribbean might be a good option for such an analysis.
A main goal of this analysis was to test the applicability of proposed marine KBA thresholds under the vulnerability criterion using data for marine turtles in Melanesia. The strict application of the proposed thresholds resulted in the number of sites being too high for conservation purposes in the region. However, a slight modification to include a definition of vagrancy (<10 nesting females) resulted in a total of 54 KBAs for marine turtles in Melanesia. We also tested higher thresholds of 20 and 50 nesting females, but this only made a noticeable difference in the number of KBAs for D. coriacea. What this meant, in essence, was that we were raising the threshold for EN and CR marine turtles to equal the proposed threshold for VU species of 10 breeding pairs (as represented by nesting females in this case). Therefore, we concluded that the thresholds defined for vulnerability could be applied to marine turtles to produce a set of KBAs that represent the major nesting sites for Melanesia.
Given the number of countries in Melanesia and the extensive coastal area, a total of 54 preliminary KBAs for marine turtles was considered to be a manageable number for conservation planning purposes. These results of this analysis need to be prioritized using additional data (including threats to sites) to determine which KBAs need to be conserved most urgently. The preliminary marine KBAs and their boundaries need to be validated and refined by experts within each country; refining site boundaries based on finer-scale management data would be particularly useful, but needs to be done as a bottom-up process. These data can be integrated into ongoing conservation planning processes (such as protected area expansion to meet Convention on Biodiversity targets). Ideally, they would become part of wider marine KBA analyses for multiple taxonomic groups (including, for instance, coral, fish, seagrass and mangroves). In the Melanesian context, most conservation action occurs at the community level; thus, community engagement and further refinement of site boundaries is recommended.
The further analysis of percentage thresholds as recommended by Edgar et al. (2008b) for EN and VU species suggests some interesting areas for further research. Application of a threshold of 1% of the global population would lead to an identification of 11 individual KBAs in Melanesia for three marine turtle species (and very few sites globally). However, this is not a relevant threshold for marine turtles that have numerous different breeding stocks. Looking at a 1% threshold within MUs defined by genetic stock analysis did provide some interesting results that correlated with the KBA analysis. The genetic stock populations need to be taken into account when prioritizing KBAs for conservation action to ensure each breeding stock is represented in the conservation of sites.
Conservation practitioners cannot wait for perfect information, given the urgency of threats to KBA trigger species. Marine KBAs are identified based on the best available data and represent global priorities for conservation. As such, it is important to integrate these sites into conservation planning processes within each Melanesian country. At the same time, KBA identification is an iterative process. New data can be incorporated to further refine the sites themselves as well as the methods used to identify and delineate KBAs. This will help fill existing gaps and biases due to survey effort and the availability of unpublished sources. Additional KBAs will be identified, and site boundaries will be modified over time not only based on new data on marine turtles but also based on the incorporation of other taxonomic groups and finer-scale management data (including formal protected areas, community-based reserves and other land-tenure information).
Finally, it is important to note that site conservation is only a small part of what is needed to conserve marine turtles. The migratory nature of marine turtles means that conservation action is required at a much broader scale to address the threats of bycatch and pollution. Conservation of turtles in the open ocean requires moving towards multinational agreements and enforcement (clearly an exceedingly difficult undertaking). The management of land-based nesting beaches within a given country is much more achievable, and KBAs represent a useful approach in identifying priority sites for such management and conservation.
The authors would like to thank Col Limpus for providing data and advice. Bryan Wallace provided detailed comments on the draft paper and information on MUs. Graham Edgar provided advice on marine KBAs and the definition of vagrancy. We thank the many researchers, NGO and Government staff who provided and assisted us in obtaining data and reports. We also appreciate the suggestions provided by two anonymous reviewers.