2.1 Fiji geographic overview

Fiji is an archipelagic nation in the southwest Pacific with over 330 islands and 550 smaller islets, covering a land area of 18,270 km². Larger watersheds are located on the major high islands of Viti Levu, Vanua Levu, Taveuni, Kadavu, and Ovalau. Mean annual rainfall ranges between <2,000 mm on the northwestern sides of the larger islands in the shadow of prevailing southeasterly trade winds and >3,200 mm on the southeastern sides [11]. As with most other Pacific Islands, the original Indigenous settlers significantly changed the natural vegetation structure, with forests replaced by herbaceous communities [42]. Following arrival of European colonizers in the 1800s, further large-scale landscape changes within watersheds resulted from commercial logging and agriculture (e.g., sugarcane), livestock, and urban and coastal development. As of the 2017 census, Fiji had a population of 884,887, of which 44.1% reside in rural areas [43]. In 2007, the most recent records of population breakdown by ethnicity, 56.8% of the population identified as Indigenous (iTaukei), while 37.4% identified as Indo-Fijian (of Indian descent) and 5.8% as other [44]. iTaukei Fijians have tenure, and thus decision-making rights, over 88% of Fiji’s land, held at the mataqali (similar to clan) level [45]. The largest administrative units in geographical size are divisions (Central, Western, Northern, and Eastern), followed by provinces (14 in total), tikina (86 in total), and enumeration areas (the smallest unit for population census that typically include 80 to 120 households).

2.2 Watershed management in Fiji

There is no formal plan or policy that provides an overarching framework for watershed management, though the Integrated Coastal Management (ICM) Framework 2011 lays out a process that Fiji could follow to develop a national coastal plan, inclusive of coordinating and regulating activities in upstream watersheds [46]. At present, despite the Department of Waterways’ strategic objective for “sustainable management of waterways and watersheds” [47], policies regulating upstream activities are piecemeal and poorly coordinated across agencies that sometimes have overlapping jurisdictions, which confounds responsibilities for enforcement [48]. Individual communities or collectives of communities have drafted ecosystem-based management (EBM) plans that include rules governing use and access of ecosystems and resources to which they commit themselves to follow on a voluntary basis [40], and some ICM plans have been developed at the provincial level [49]. In October 2022, Fiji’s Cabinet endorsed a new National Drinking Water Quality Committee, with a mandate to provide evidence of safe drinking water through sanitary surveys, water safety plans, and drinking water quality monitoring and surveillance programs, which may help facilitate improved coordination for water management and governance.

2.3 The WISH Fiji project

The WISH Fiji project involves a research consortium between two Australian universities, a Fijian university, the Fiji Ministry of Health and Medical Services (MoHMS), the World Health Organization (WHO), the United Nations Children’s Fund (UNICEF), the Pacific Community (SPC) and the Wildlife Conservation Society (WCS) [6, 37]. WISH Fiji has five goals, to: reduce the incidence of water-related diseases in people and downstream ecosystems; empower communities to access and maintain their fundamental right to clean water; strengthen connections to place to enhance environmental stewardship and maintain cultural practice; develop a coordinated mechanism for systems health governance; and facilitate approaches to sustainable finance and scale interventions. The project was designed to use knowledge co-production approaches in order to encourage uptake and ownership of watershed management and governance by landowners and government. WISH Fiji has been undertaken through a series of steps to identify, communicate and reduce risk through an adaptive management approach (Fig 1). Each of the steps are described in brief below.

Download:

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

Fig 1. Project risk reduction methodology steps within an adaptive management cycle.

https://doi.org/10.1371/journal.pwat.0000102.g001

2.4 Ethics

WISH Fiji received ethics approval from the Fiji National Health Research and Ethics Review Committee (FNHRERC No: 2018.231.CEN), Fiji National University’s College Health Research Ethics Committee (CHRED ID: 009.19), the University of Sydney’s Human Research Ethics Committee (2019/588) and Edith Cowan University’s Human Research Ethics Committee (#2019–00618).

2.5 Watershed selection process (Step 1)

To facilitate the project site selection, the WISH Fiji team: devised a list of selection criteria for project watersheds; held a national-level workshop with key stakeholders from government and civil society to apply the criteria; and then presented the proposed watersheds to the interim National Drinking Water Quality Committee chaired by MoHMS for consideration and final decision. Invitations to the national workshop were sent to key sectors engaged in public health, WASH and environment active in project geographies under consideration, with particular focus on organizations engaged in the national WASH cluster. Key stakeholders participating included: staff from MoHMS (including sub-divisional medical officers); Ministry of Agriculture; Water Authority of Fiji (WAF); Ministry of iTaukei Affairs (MiTA); project partners from WHO, SPC and UNICEF; and members of NGOs (e.g., World Wide Fund for Nature; Live and Learn Fiji) and WASH consultancies. To be suitable for selection, a watershed needed to have all the following primary characteristics: sufficient records to demonstrate recent outbreaks of at least two of the three LTDs in the prior two years; at least six identifiable communities within its boundaries; and known concerns about drinking water quality, health-related climate vulnerability, impacts of recent natural disasters and/or poor water and sanitation infrastructure. To ensure consideration of the whole linked watershed-to-reef system, we also required at least two of the watersheds to be coastal and to discharge to the ocean. Upon satisfying these primary criteria, short-listed watersheds were evaluated according to the following secondary criteria: accessibility; characterization as primarily rural; not concurrently receiving other significant assistance/funded support in WASH, environmental management, or other areas that would compromise the ability of the project to detect changes in risk factors; and potential for leveraging resources from other agencies to support implementation of prioritized interventions. These processes resulted in the selection of five project watersheds (Fig 2), for which the major defining features are described in Table 1.

Download:

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

Fig 2.

Locations of WISH Fiji project watersheds in Fiji: (a) Dawasamu and Waibula; (b) Upper Navua; (c) Bureta; and (d) Dama. Black circles indicate project villages. Base layer map was obtained from GADM at https://gadm.org/download_country.html (license information available at https://gadm.org/license.html).

https://doi.org/10.1371/journal.pwat.0000102.g002

Download:

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

Table 1. Major demographic, geographic, development and management characteristics of five project watersheds.

EBM: ecosystem-based management; WASH: water, sanitation and hygiene.

https://doi.org/10.1371/journal.pwat.0000102.t001

Across all five watersheds there is a total population of 13,206, ranging from the lowest population in the smallest watershed of Bureta (1,089 people) to the greatest population in the largest watershed of Waibula (6,119 people). The headwaters of all watersheds are well-forested. Waibula and Dawasamu are low gradient, coastal watersheds with alluvial and depositional hydrology in the lower reaches. The Upper Navua River forms the headwater section of the larger Navua River watershed and is steep and mountainous, with erosional, colluvial and depositional features. In Dama and Bureta watersheds, rivers flow through moderately steep, coastal watersheds with erosional and colluvial features.

2.6 Community selection and free, prior and informed consent (Step 2)

Project communities were selected through consultations with provincial government staff who had knowledge of presence of prior outbreaks of LTDs and local knowledge of where there was likely to be disease risk that could be addressed through project interventions. Across the 29 communities selected, most of the population is of iTaukei origin, though two communities have a majority Indo-Fijian population. Our free, prior and informed consent (FPIC) process began with a series of consultations with MiTA, responsible for developing, implementing, and monitoring government programs focused on the governance and well-being of iTaukei people. In the absence of a formal government process for community-level FPIC, we co-designed a process with MiTA tailored to the Fijian context based on international best practice guidelines [50]. Prior to approaching the 29 communities, detailed discussions on WISH Fiji planned activities were held with key government ministries influential in the watersheds, including MoHMS, Agriculture, Forestry, Lands and Mineral Resources, and MiTA. Over a five-month period, we conducted a three-phased FPIC process in all 29 communities which focused on: initial visits to local and provincial government partners to describe project objectives; comprehensive community awareness sessions with participation of broad segments of each community, including men, women, elders and youth; and, following adequate time for internal community discussion, a final visit to each community with representatives from MiTA to obtain granted signed consent. When all phases of FPIC were completed for all communities, we then undertook household-level consent for the 311 households enrolled in the project (see Step 4 below).

2.7 Survey instrument design (Step 3)

Our next step was to understand to what extent: individuals in a community were at risk of being exposed to an LTD infection or a diarrheal disease; and downstream ecosystems were at risk from upstream land-based activity. A search of the literature from reviews, case-control studies and models relating environmental, climate, and socioeconomic variables to disease incidence or seroprevalence revealed 117 potential risk factors which could be assigned within nested spatial scales, from watershed (consisting of largely environmental and landscape factors), to community, to the household and individual-level. Watershed risk factors were defined by environmental and landscape parameters, evaluated at the ‘sub-catchment’ level, which we define here as inclusive of all upstream areas that drain to primary water sources identified during community mapping (see S1 Table). Community-level risk factors were inclusive of infrastructure and services, as well as proximity of swamp and proximity of livestock to water sources, which affected each community. Various demographic, socio-economic, and behavioural factors, as well as some household-level infrastructure, were relevant at the household and individual-levels. We selected risk factors that were applicable to one or more individual diseases or downstream ecosystem impacts. Our literature review was also used to identify survey instruments that could be used to gauge the degree to which communities were vulnerable to these risk factors. Accordingly, we identified and adapted existing instruments and developed new instruments that could be applied at an appropriate level. Relevant instruments are shown in Table 2, which represents a subset of the full set of instruments applied over the duration of WISH Fiji (see [51, 52] for additional instruments).

Download:

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

Table 2. Types of data collection instruments designed by the WISH Fiji team to measure risk factors at watershed/sub-catchment, community and household/individual scales.

References are indicated when the instruments were adapted from prior sources.

https://doi.org/10.1371/journal.pwat.0000102.t002

2.8 General methodology for data collection (Step 4)

Instruments were applied during a baseline data collection phase between August and December 2019. Within the 29 communities, 311 (21%) out of 1,502 households were selected for survey and observation. To select households, each was assigned a unique number, all numbers were placed in a bowl and were selected one by one until at least 15% of households per community were selected (or a minimum of 6 households for communities with less than 40 households; [51]). Data collection was supervised by a nominated coordinator for each watershed (“Catchment Coordinator”) with a team of trained project staff and volunteers during an intensive phase of interviews, surveys and observations, over a period of about one week per community. A water quality monitoring program was designed and simultaneously implemented to assess risk at three scales: (1) watershed-level: river and creek water (from ridge to reef); (2) community-level drinking water sources and piped distribution networks; and (3) household-level drinking water (piped and stored). This longitudinal approach produced a dataset that could be used to observe changes in the watersheds and communities. Water was sampled in sterile 500 mL bottles and all analysis was conducted using portable field kits (Wagtech, Palin, UK). Faecal indicator bacteria were measured in the field using membrane filtration with m-ColiBlue24 reagent (method 10029 Hach, USA) which gives counts of Escherichia coli (E. coli cfu/100 mL). All survey data were used to evaluate pre-intervention systems health state (see Step 5), as well as to highlight risk factors requiring attention (see Step 8).

2.9 Identifying risks across nested spatial scales (Step 5)

Implementation of the instruments described for Steps 3 and 4 generated a significant amount of data from each community. The first phase of data analysis focused on 22 known risk factors by: removing variables that were not reliably measured; removing variables that were not able to discriminate between communities or between households; removing highly correlated variables that repeatedly showed the same response; and combining variables into a composite indices to represent a more comprehensive risk factor (e.g., for “livestock near water”, see S1 Table). In addition, certain factors (i.e., socio-economic and demographic variables), which have been documented to have associations with disease risk, are not considered here because we could not intervene to change them. S1 Table provides a detailed explanation of the 22 risk factors, with key studies from the literature providing the evidence base for their selection, arranged according to: watershed factors determined from geospatial and water quality data; community factors measured through observations or water quality data collected by field teams using instruments (instruments B-F and J-L; Table 2); and household/individual factors measured from survey instruments (instruments G-I; Table 2) and household water quality sampling (instrument J; Table 2). The source data and measurement methods are described for each risk factor, along with threshold values used to categorize low, medium and high risk, where the thresholds between low, medium and high risk represent testable assumptions drawn from the literature, field observations, or discussions in each community arising from the water and sanitation safety planning process (see Step 7).

2.10 Communication tool design (Step 6)

We contracted cChange, a nonprofit organization specializing in designing communications products for nature-positive behavioural change, to develop flip charts with graphics that illustrate how activities at the watershed, community and household/individual level create risks for LTDs and ill-health in downstream ecosystems. Each graphic illustrating risk was paired with a solution-space graphic that indicated recommended interventions to reduce risk: versions were produced with accompanying text in both English and iTaukei. Our project team developed a script to guide facilitators in explaining the graphics during meetings with each community. Flip chart discussions were paired with Powerpoint presentations of the baseline data results in each community, highlighting where medium and high-risk factors were observed. These presentations were made in concert with meetings to undertake water and sanitation safety planning (see Step 7).

2.11 Water and sanitation safety planning (Step 7)

Our water safety and sanitation plan (WSSP) process engaged communities to identify and address risks related to drinking water, solid waste, sanitation and hydrological systems using a combination of UNICEF’s Drinking Water Safety and Security Planning (DWSSP) implementation cycle [54] and WHO’s Sanitation Safety Planning process [55] tailored to the Fiji context and with added attentiveness to activities occurring in the sub-catchment area around drinking water sources. The six iterative components of this community-level, adaptive management process were: preparation, by collating community details, assembling the team and assessing any existing water or sanitation plans; documentation, by describing in detail community and household-level drinking water supply and sanitation systems; hazard mapping, by identifying and assessing hazards and exposure risks (within 0.5 km of drinking water sources), hazardous events, and existing control measures; planning, led by community members to identify priority actions to minimize risks; implementation, through community stewardship of each WSSP through coordination of intervention implementation and infrastructure maintenance; and cyclic review, to improve and document all aspects of WSSP implementation. The initial WSSP process started in mid-August 2020 and was completed by mid-October 2020. Because several communities were severely damaged by tropical cyclones Yasa and Ana in December 2020 and January 2021, respectively, these communities’ WSSPs were updated in early 2021 to reassess their post-cyclone WASH needs.

2.12 Intervention prioritization process (Step 8)

Decisions about resourcing watershed interventions by the project team were influenced by: cost, including balancing investment as equitably as possible across project watersheds; urgency, given impacts to water infrastructure by tropical cyclones Yasa and Ana, and COVID-19 transmission mitigation; feasibility, given travel restrictions and supply chain issues associated with COVID-19; ability to obtain landowner permissions; and knowledge of partner resources that could be leveraged to support other interventions. These decisions were made considering the complexity and financial viability of each proposed intervention in the context of community capacity. We categorized proposed interventions into five types of work: (A) watershed or sub-catchment scale (e.g., long-term reforestation activities across large scales), requiring major resources (> US$500), complex procurement to outsource external skills, and coordination of multiple stakeholders; (B) community-level (e.g., infrastructure construction by tradesmen), requiring major resources (> US$500) and complex procurement to outsource external skills; (C) community-level (e.g., simple infrastructure construction), requiring major resources (>US$500) but where there was local capacity to complete the work; (D) community-level (e.g., small repairs, simple construction), requiring minor resources (<US$500) and where there was local capacity to complete the work; and (E) community-level (e.g., policy enforcement, community decisions or basic repairs), where no physical resources were needed and the community has the capacity to do the work. Costs were estimated through quotes obtained from vendors and service providers. Determination of the complexity of work and local capacity available was done with local WISH Fiji project managers and Catchment Coordinators.

3.1 Community-level risk

By far, the most ubiquitously high risks were associated with the poor coverage of safely managed sanitation in the community and high numbers of enrolled households had damaged or overflowing sanitation infrastructure (Table 3 and S2 Table). A suite of factors had either medium or high-risk for more than 70% of communities. These included: the average E. coli calculated for environmental water samples, from each community; the presence of swamps proximal to the community; issues associated with livestock near water; perceived adequacy of drinking water supply; and frequent reports of householders working in wet environments (Table 3).

Download:

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

Table 3. Number of communities categorized in low, medium and high risk categories for each of the risk factors, from 2019 baseline data.

HEA(%): The amount of highly erodible soil area in the sub-catchment; HFRA(%): amount of high flood risk area in the sub-catchment; CC/km: number of creek crossings per km of road; FF/km: forest fragments in the riparian buffer zone per km of river. Data are derived from S2 Table.

https://doi.org/10.1371/journal.pwat.0000102.t003

Some risk factors revealed remarkably similar patterns across the communities. For example, there was the same distribution of communities spread across risk categories for factors related to standing water around the house and cutting of bushes in the yard. In other cases, patterns varied. While most communities were low risk for hygiene factors related to the frequency of washing fruit and vegetables and frequency that the food preparer washes hands before cooking, 41% of communities had medium to high risk associated with hand washing (Table 3).

3.2 Watershed-level risk

At the watershed level, patterns also emerged (Table 4). In addition to the risk factors that were elevated across a majority of communities described above, Bureta communities showed elevated risk due to large quantities of high flood risk area within sub-catchment boundaries. Dawasamu, Waibula and Upper Navua communities had elevated risk from large areas of highly erodible soil within sub-catchment boundaries. Dawasamu and Waibula communities had elevated risk from the presence of various types of mosquito breeding habitat. Further, compared to the other watersheds, Upper Navua and Bureta communities had elevated risk due to higher levels of E. coli detected in river water, primary drinking water sources, and piped and/or stored water. Individual communities within each watershed also showed elevated risk for specific factors, such as standing water around households, low frequency of cutting bushes near households and working in wet environments (S2 Table). These community-level risks provided specific guidance for WSSP processes and required interventions.

Download:

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

Table 4. Patterns of elevated risk across the five project watersheds.

Here ‘x’ represents instances where risk factors or groups of risk factors were high and/or medium across all communities in a project watershed at baseline.

https://doi.org/10.1371/journal.pwat.0000102.t004

3.3 Prioritizing interventions to manage risk

Based on the presentation of baseline risks and the WSSP process undertaken in each project community, 339 watershed interventions were prioritized for implementation across nine broad categories related to: animal management; drainage; health systems surveillance; hygiene; integrated planning; land use management; sanitation systems; waste management; and water systems (Table 5). Interventions related to water systems were most frequently prioritized (29.2%), followed by land use management (21.5%). Priorities for water systems interventions related to needs for maintenance, repair and new infrastructure, as well as general awareness on the factors causing unsafe water and best practice related to water systems governance and management. Priorities for land use management were inclusive of: nature-based solutions, such as riverbank stabilization (i.e., with vetiver grass), reforestation, and forest protected areas; relocation of farms away from water sources; improved policy regulation with respect to development permitting and monitoring; and general awareness raising on agricultural best practices, forest ecosystem services, and fishpond management.

Download:

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

Table 5. Total number of watershed interventions prioritized by category as a result of baseline risk factor assessments and water and sanitation safety plan (WSSP) processes, compared with number of interventions implemented as of October 2022.

https://doi.org/10.1371/journal.pwat.0000102.t005

Implementation of interventions began in mid-August 2020, starting with the participatory WSSP processes in each community, which were counted under the integrated planning category. As of October 2022, 154 completed interventions were documented, with the majority falling under water systems (31.8%), followed by integrated planning (19.5%), land use management (14.3%), waste management (11.7%), and hygiene (10.4%; Table 5). All 29 communities reported implementing watershed interventions on their own accord, while nearly 17% (26 of 154) implemented interventions were done so completely with human and financial resources from project partners, including government agencies, Water Authority of Fiji, and other NGOs (e.g., Rotary Pacific). Despite the high risks presented by large numbers of inadequately managed sanitation systems across all communities (Table 3 and S2 Table), needed sanitation interventions are yet to be undertaken due to procurement challenges partly due to a limited pool of experienced sanitation contractors. During follow-up monitoring, while the project team observed that some drainage issues identified were addressed by communities, these specific interventions have not yet been quantified through the iterative review of WSSPs. Follow-up monitoring carried out between May and August 2022 indicated reduced risks in some communities against five specific risk factors that may at least partially be attributed to project interventions: environmental water quality (E. coli); primary drinking water source quality (E. coli); drinking water supply; piped drinking water quality (E. coli); and washing hands (Fig 3 and S3 Table).

Download:

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

Fig 3. Examples of risks, interventions implemented to address each risk, and changes in the number of communities in each risk category between 2019 baseline and 2022 follow-up monitoring (S2 and S3 Tables).

Red = high risk; yellow = medium risk; green = low risk. Communities were assigned risk categories based on thresholds for each risk factor indicated in S1 Table.

https://doi.org/10.1371/journal.pwat.0000102.g003

Recognizing that gender roles shape the collection and use of water, and in response to recommendations from Nelson et al. [52] who suggested that water resource governance could be strengthened in WISH Fiji project communities by increasing representation of women and community health workers on water committees, the WISH Fiji project team made a concerted effort through our adaptive management cycle (Fig 1) to facilitate more inclusive participatory planning in reviews of WSSPs and implementation of interventions. As a result, by October 2022, 69% of communities (20 of 29) increased representation of women on community water committees, and 83% of community water committees (24 of 29) included community health workers. Community health workers are community representatives trained by district health nurses to assist their communities to maintain proper child and maternal health and promote overall health and well-being. They work alongside the district health nurses to deliver community outreach and provide nurses with information regarding notable WASH issues requiring attention.

4.1 Outcomes from a portfolio approach to integrated watershed management

Under WISH Fiji, our expectation, based on best available evidence from the literature [57, 58], is that systems health risks will occur within all nested subsystems, and that a combination of nature-based solutions (e.g., forest protection, restoration around water sources, riverbank stabilization, coastal wetland management), WASH and behavior change interventions, implemented across nested scales within watersheds, will reduce the incidence of microbial disease in humans and aquatic organisms. We also hypothesize that upstream nature-based solutions that provide flood risk mitigation benefits (recognizing that these are variable, [59]), combined with other specific community and household-level interventions that reduce mosquito-breeding habitat (e.g., cutting bushes, eliminating standing pools, covering containers, improved solid waste management), will reduce incidence of dengue and other mosquito-borne illnesses [18]. As described below, some of our expectations were met. We found communities in high-risk categories for risk factors at the watershed, community and household/individual scale (S2 Table). Some risks could be reduced by quick-fix, low-cost improvements measures. For example, we observed: fewer communities at high risk for drinking water supply and quality, which were likely related to improvements to water infrastructure; lower risk E. coli concentrations from river and source water samples in several communities, which may have resulted from fencing water sources and tethering livestock; and more food preparers practicing regular handwashing, likely due to awareness about best hygiene practices and distribution of soap, which was heightened due to community COVID-19 transmission (Fig 3 and S3 Table). Changes across other risk factors were less immediately apparent, both because of anticipated, lagged temporal responses (e.g., from ecosystem restoration [59]) and due to challenges in the way health systems surveillance data are collected.

Health systems surveillance data in Fiji, like in many countries, are generally not geolocated to the residences of individuals presenting at health centers but are enumerated by health facilities [18]. As a result, we are challenged with an inability to link specific watershed socio-ecological variables and management actions to specific disease incidence because health facilities in Fiji generally do not record the home address of those visiting the facility. We attempted to overcome this under WISH Fiji by asking about suspected case incidence and reviewing hospital and rural health clinic records for confirmed case incidence. These investigations yielded very low case numbers, likely due to the limitations of our project geography and many confirmed cases not being geolocated. New opportunities are emerging with the use of digital platforms to link disease clusters to place-based factors [60–62]. However, ethics considerations of digital surveillance need more scrutiny, and the technological innovations may not necessarily be suitable for remote locations where people are not connected online or resolved at fine enough geographic scales for smaller watersheds. In the absence of confirmed and reliable case data collected within watershed boundaries to enable identification of key local drivers of disease risk, and as described above, the WISH Fiji approach has been to measure a suite of potential risk factors and then co-design portfolios of interventions with communities and partners based on these measured potential risks to improve systems health (Fig 1).

Through WISH Fiji, we confirmed that watershed-level characteristics are important in most communities in all sub-catchments. Soil erosion associated with rainfall and/or poor land use practices changes hydrological and water quality characteristics downstream [63, 64]. These disturbances influence more proximal determinants of human health, like access to clean water, habitat for mosquito vectors that carry disease like dengue, and direct exposures to contaminants and infection [24]. Similarly, a high amount of high flood risk area in the sub-catchment poses a risk for exposure to: zoonotic disease like leptospirosis, which can be transmitted through mammalian urine and excreta that are mobilized by flood waters [31]; and bacterial disease like typhoid which is transmitted through faecal-oral pathways [11]. Flood risks and rainfall-associated erosion are likely to accelerate under predicted future climate scenarios for Fiji, with high probability of greater intensity and frequency of days of extreme rainfall [65]. Perceptions of frequent flooding were recognized by householders in our surveys, suggesting that interventions around placement of future houses constructed in communities may not be difficult to implement. In both cases, more careful planning at the village and district level about the placement of communities and houses on tenured land is a warranted intervention.

We found an almost universal need for improvement of sanitation back-end infrastructure so that faecal sludge is safely contained or treated [66]. This is not uncommon in rural communities in low to medium income countries, where there are documented links to unsafely managed sanitation and poor human health [14]. In Fiji, inadequate placement and upkeep of sanitation facilities will increase the likelihood of exposure in downslope communities and contamination of waterways downstream. Identifying the highest priority (most damaged, most poorly placed) latrines continues to be an important part of the WSSP process and intervention activities for WISH Fiji. We were, however, challenged with the ability to locate a contractor who could complete critical sanitation upgrades during the project timeline, even after testing the market with more than six open tendering calls for septic tanks. We recommend that future projects either develop in-house expertise to oversee sanitation construction to code or ensure adequate time and resources to facilitate training for local contractors if there is not an adequate marketplace to provide services.

At the community-level, we found that primary drinking water sources were nearly always from spring-fed dams, from which piped water was drawn to reservoir tanks and then delivered to households. The land surrounding springs is rarely protected from human activity and livestock incursion, and there is a distance between the spring and the dam where faecal contamination can easily occur. Increased risks of exposure to faecal pathogens from drinking water supplies have been documented to occur in other tropical, rural settings due to unimproved drinking water infrastructure and the use of surface water (rivers or creeks) as an alternate drinking water source [67]. All of these matters provided us with opportunities for cost-effective interventions for systems health outcomes.

Finally, we found patterns of risk factors related to behaviours of residents, including reporting infrequent hand washing (and/or without soap) and high frequency of working in wet environment (including without appropriate protective equipment). Under these circumstances, awareness raising, education and health promotion activities are worthy interventions in rural communities in Fiji where mosquito habitat remains in proximity to households [18], where exposure to contaminated water in the environment is likely to occur [68], and where increased attention to hygienic practices are warranted [69]. The timing of the COVID-19 pandemic and its associated travel restrictions created some barriers for our project team to raise awareness in person with communities about certain behavioural risks, though at the same time, opportunities for broadcast messaging about hygiene practices were enhanced (e.g., through public services announcements on television and radio).

Interventions will change risk at different spatial and temporal scales and will have variable impacts across geographic and socio-economic contexts [59]. For example, evidence indicates that changes in water quality, including bacteria levels, post-wastewater management and water infrastructure improvements can occur in as little as one to two years [70], whereas ecosystem-level changes in downstream communities in response to upstream interventions, particularly from restoration, are more likely to require decadal timescales for recovery [70, 71]. The time lag from intervention planning to response is also influenced by the complexities of land tenure in Fiji. For instance, engaging in forest restoration is complex, requiring mapping erosion-prone areas near water sources, identifying and verifying rightful landowners, and only then sourcing or growing of seedlings for outplanting once landowner consent is granted–a process which may on its own take well over a year to achieve.

While we observed fewer communities in higher risk categories post-interventions related to adequacy of drinking water supply, source, piped and environmental water E. coli, and frequency of handwashing, we acknowledge that there was not a lot of time between implementing interventions and follow-up monitoring to affect change. We also did not have enough time to complete interventions targeting all high-risk factors, especially for sanitation systems. Differences in risk factor measurements post-interventions could also be a result of: natural stochasticity, sampling variability and seasonal/climate differences (e.g., for environmental water quality); different individuals responding as heads of households; influence of COVID-19 hygiene messaging or respondents telling us what they think we want to hear (e.g., for handwashing); or other activities happening within the communities of which we are not aware.

Given that interventions within the project are delivered as a portfolio of actions across multiple nested scales, tools are needed to quantify the risk reduction potential of the combined effect of these interventions. Bayesian Belief Networks (BBNs) are one increasingly popular analytical platform that can incorporate knowledge of different uncertainties, from different scales and sources, and easily handle missing data [72–74]. Using BBNs can help identify co-benefits across and within nested scales and where simultaneous implementation of multiple interventions across different scales could have a larger effect than the complete reduction of risk factors at any one level [75]. Using decision-support tools such as BBNs does not, however, eliminate the need to balance trade-offs in different aspects of systems health: for example, proximity of swamps may increase risk of vector-borne disease such as dengue, while at the same time the wetlands provide important ecosystem services for flood mitigation and nutrient cycling that may reduce risk from pathogenic bacteria and other contaminants to people and ecosystems downstream [76, 77]. Intervention planning ultimately needs to take into consideration these trade-offs and balance risks, particularly with attention to what interventions can produce the most net improvements for overall systems health.

4.2 Lessons for knowledge co-production within research-action arenas

The collaborative and cross-sectoral nature of project implementation allowed the WISH Fiji team to leverage unanticipated outcomes that support long-term durability of the approach. The project results and outcomes are featured prominently in the Fiji Government’s 2023 Voluntary National Report on progress against the SDGs [78], particularly for good health and well-being (SDG3) and clean water and sanitation (SDG6): this demonstrates broad-scale government support of the WISH Fiji approach. The co-produced WSSPs for each community were supportive in several ways that enabled external partners from NGOs and government to directly contribute to intervention implementation in project communities. First, priorities from the WSSPs were integrated into broader Integrated Village Development Plans, which form the basis for annual resource allocation at the provincial level. Secondly, through co-development of WSSPs with communities, government and staff from WAF, community leaders became better aware of the process for notifying WAF of priorities for water infrastructure improvements, which involves raising the issues at provincial meetings so that they can be reported to WAF for inclusion in annual budget allocation.

These actions to embed ongoing implementation of watershed management within regular government and community systems became increasingly important as WISH Fiji project staff were unable to access project sites for lengthy periods due to COVID-19 related restrictions on movement in Fiji. Reviews of community WSSPs in 2023 showed that almost 40% of priority interventions were funded partly or wholly by communities, indicating their level of commitment to risk reduction. When international border closures prevented overseas-based research partners from traveling to Fiji between April 2020 and December 2021, increased responsibility for project oversight and coordination was transferred to in-country staff at Fiji National University (FNU), which helped build national capacity for planetary health research. As required, FNU and other in-country project staff coordinated village-based activity implementation and data collection (e.g., key informant surveys) via telephone when sites could not be reached.

WISH Fiji also played an important convening and brokering role to bring important sectoral actors together (i.e., from ministries of Health, Rural Development [Department of Water and Sewerage], Forestry, Agriculture, Environment, iTaukei Affairs, and WAF) for joint stakeholder planning at the district level. The WISH Fiji Catchment Coordinators were responsible for leading facilitation, documentation and educating participants about potential impacts and synergies of each sector’s planned local activities. This type of cross-sector coordination and collaboration can improve efficiency of resource allocation and minimize implementation of actions that too narrowly focus on single-sector strategic objectives at the expense of overall systems health [38, 79]. Such brokering, intermediary and boundary spanning roles are increasingly recognized as essential components of successful interdisciplinary and cross-sectoral collaboration [80].