• Fiorella 2004 Drivers Side Mirror
• Fiorella 2004 Drivers Ed
• Fiorella 2004 Drivers Wanted
• Fiorella 2004 Drivers License

Fiorella 2.0 - New! Fiorella; Ribbons. Clear; Mylos; Plain; Films. Films - Fiorella Printer; Films - Fiorella Printer 2.0; Everyday Ribbons. Aspy; Metallic Curling; Regular Curling; Raffia; Satin; Wired; Support; Samples; About; Contact. Limited Time Offer - Expires December 31: Buy a Printer and Get 4 Rolls of Ribbons for Free ($56 value. Fiorella FIX Just like all other members of the family, the Fiorella Fix is a spot that is embedded in the ceiling. In contrast to the long and short version, this variant cannot move.

Research

• Introduction
• Methods

ABSTRACT

Freshwater social-ecological systems are shaped by site-specific characteristics, environmental change and annual fluctuations, and the actions of resource users and managers. Our primary aim is to analyze the relative effects of these multifacted forces on fishery productivity and biodiversity using the case of community-managed protected areas within Cambodia’s rice field fisheries, known as community fish refuges. Our secondary aim is to take advantage of fixed-effects regression models, a rigorous approach that is common in other disciplines, yet novel to analysis of social-ecological system productivity, and allows comparison of each site to itself across time using observational data. We analyze the relative associations of site characteristics, annual fluctuations, biophysical modifications (e.g., deepening, creation of inlet/outlets, etc.), and governance capacity (i.e., community committee meeting intensity, fundraising to support illegal fishing patrols, etc.) with fish biomass and biodiversity within 40 refuges over three years. We find that seasonal and site-specific effects within this system relate strongly to biomass and species richness patterns. We also find an association between biomass and biodiversity and some elements of governance capacity building. Our findings suggest that in this setting solutions tailored and responsive to the local context may be most appropriate given the strong interannual and site-specific drivers. Methodologically, the dynamism and site specificity of such systems befits within-site comparisons over long time horizons to appreciate the factors that drive biomass and biodiversity.

INTRODUCTION

Where present, inland fisheries in poor countries provide a disproportionately large share of livelihoods and fish for local consumption relative to coastal fisheries (Cooke et al. 2016). Despite the importance of inland fisheries, the extent of inland catch has been underappreciated for decades (Fluet-Chouinard et al. 2018) and the inland fisheries sector is widely afflicted by management and restoration challenges, and competing uses for freshwater.

Freshwater protected areas, often called fish sanctuaries, represent an important management strategy to bolster fish production. However, the role and effectiveness of freshwater protected areas are underexamined. Remarkably few well-documented examples of fish sanctuaries exist and developing country examples are even rarer, though freshwater protected areas have been established in Cambodia (Chheng 2010), India (Sarkar et al. 2013, Gupta et al. 2015), Laos (Baird 2000, Baird and Flaherty 2005), and the nations around Lake Victoria (Kiwango et al. 2013). Such community-managed inland protected areas often draw elements of their approach from a range of fishery management strategies, including adaptive comanagement (Kura et al. 2004, Olsson et al. 2004, Pomeroy and Andrew 2011, Ratner et al. 2012), agricultural landscape management (Dela Cruz et al. 1992, Gregory and Guttman 2002, Gurung and Wagle 2005), aquatic restoration (Zedler 2000, Society for Ecological Restoration International Science & Policy Working Group 2004, Young et al. 2005), and marine protected areas (Gill et al. 2017).

Inland fisheries and community protected areas in particular represent an important yet inherently complex nexus of integration between natural systems, social systems, and their management. Ostrom’s social-ecological systems framework is a tool to identify interactions among the governance system, a specific resource, and a resource system, or the ecosystem (Ostrom 2007, 2009) that facilities cross-system comparisons to identify key relationships and system components (McGinnis and Ostrom 2014). Such complex systems demand integrated analyses of social-ecological processes (Hughes et al. 2017) and new modeling approaches (Schlüter et al. 2012), but limited studies quantitatively integrate ecological components into social-ecological systems analyses (Vogt et al. 2015). Research efforts that quantitatively analyze social-ecological system dynamics have generally analyzed components within the context of limited comparative sites (e.g, Laborde et al. 2016, Ziegler et al. 2017, Filbee-Dexter et al. 2018) or at a single time point (e.g., Leslie et al. 2015).

We have two research aims. First, we aim to analyze the interacting relationships between site characteristics, water quality, annual fluctuations in flood dynamics, and management actions taken by resource users and their association with fishery productivity and biodiversity in community-managed freshwater fish refuges. Our second aim is to apply a new approach to the quantitative analysis of social-ecological systems that extends the application of fixed effect regression models, commonly used in other disciplines, to study the productivity dynamics within a social-ecological system.

We analyze three years (2012–2015) of observational data on 40 community fish refuges (CFRs) in Cambodia’s Tonle Sap floodplain using panel fixed effect regression models. Our analyses examine the impact of site-specific characteristics, seasonal and annual differences, and aggregated management actions designed to address governance capacity or provide for biophysical improvements. CFRs were adaptively managed, therefore we aggregate management intensity across governance capacity and biophysical improvement activities. Rather than examining individual actions, e.g., deepening ponds or committee meetings, we examine the suite of real-world activities as a whole, including deciding to act and selection of management strategies that plays out as resource users respond to changing environmental conditions. Our analyses rely on observational data, though repeated observations of the system enable us to account statistically for site characteristics and annual weather and flood fluctuations, e.g., exogenous annual effects of rainfall, flooding, temperature. Unlike much prior work on social-ecological systems, we examine a large number of sites, track relationships over time, and compare each site to itself over time, allowing us to statistically account for the characteristics at each site that could otherwise confound the social-ecological relationships of interest.

Cambodian fisheries

The approximately 4.5 million Cambodians that work in the fisheries sector contribute 6–11% to national gross domestic product and depend on the annual flood for food and income (Asian Development Bank 2012, Hap et al. 2016). The rice field fisheries within the Tonle Sap Lake floodplain contributed 20% of annual fish production (550,000 metric tons) in 2013 (Hap et al. 2016). Rice field fisheries are embedded within a multifunctional agro-ecosystem that includes channels, lakes, and rivers and supports fish and other aquatic species (Balzer et al. 2006, Hap et al. 2016). A common pool resource, rice fields have been identified as a particularly important habitat for fisheries productivity and fish for consumption (Balzer et al. 2006). During the annual wet season, May through November, fish migrate from the lake and local tributaries to seasonally flooded rice fields to feed and breed (Campbell et al. 2006). From November to January, fish migrate back from rice fields and waterways to perennial water bodies, where they await the return of the rainy season in May (Campbell et al. 2006).

Concern for rice field fisheries’ ability to meet the demands of a growing rural population led to the establishment of CFRs by the Government of Cambodia. CFRs are managed protected areas that are governed by community committees in conjunction with the Fisheries Administration, and located within rice field fisheries habitats (including rice fields, streams, channels, and lakes). As of 2013, there were 779 CFRs in the country (Fisheries Administration 2009). Fishing within CFRs is prohibited throughout the year so these perennial bodies of water can provide fish and other aquatic species with refuge during the dry season so that they can move out of the refuge during the wet season and increase fish catch in the surrounding area (Joffre et al. 2012). Our study focuses on 40 CFRs that received support from WorldFish and its partners from 2012 to 2016 to implement site-specific adaptive management actions to enhance fish production and diversity.

METHODS

Our analyses focus on isolating and describing relationships between CFR biomass and biodiversity and site characteristics, water quality, governance, and annual flood fluctuations. Figure 1 depicts our hypothesized determinants of CFR biomass and biodiversity and the potential relationships between them. We discuss each relationship in turn below.

Study site and WorldFish intervention

Our study centers around 40 CFRs supported by WorldFish as part of the Rice Field Fisheries Enhancement Project (RFFEP), which was funded by USAID’s Feed the Future program. The CFR sites were situated in the four provinces surrounding Tonle Sap Lake: Battambang, Kampong Thom, Pursat, and Siem Reap (Fi. 2). Through extensive consultation with community groups and fisheries experts at the start of the RFFEP project, four hydrologically distinct types of CFRs were identified: irrigation reservoirs, community ponds without flooding, community ponds with flooding, and segments of a large water body (Brooks et al. 2015). Site selection was done by WorldFish staff to capture diverse CFR types and informed by their assessments of likely success in project implementation and approximate representation across four CFR types and four provinces.

Management was responsive to user needs and perceptions (e.g., household preferences, user perceptions of the state of the fishery) and site characteristics (e.g., observed patterns of fish migration, hydrology) as identified through biomonitoring data on water quality, fish biomass surveys, and community perception. This adaptive management led to heterogeneous, tailored management approaches (Appendix 1; Brooks and Sieu 2016). Management actions included biophysical modifications aimed to improve water retention during the dry season, quality of fish habitat, and connectivity with rice fields (e.g., deepening of ponds, removal of vegetation, installation of brush parks, and creation of CFR inlet/outlets and channels to rice field habitats) and were largely carried out during the dry season (Hortle et al. 2004). Governance support included assistance to CFR committees in developing intervention plans, electing representatives to governing committees, coaching on fundraising strategies, and implementing fishing prohibitions, demarcations, signage, and water use agreements; governance support also sought to improve coordination with local authorities (Joffre et al. 2012, Miratori and Brooks 2015). In addition, CFR committees and communities were provided with training regarding the rice field fisheries environment and its management, as well as capacity building in the areas of bookkeeping, financial planning, fundraising, and public speaking (Brooks and Sieu 2016). The management activities were based on WorldFish’s longstanding research within the rice field fishery setting and project documentation provides detailed information about the conditions under which different activities were indicated (Hortle et al. 2004, Joffre et al. 2012, Miratori and Brooks 2015).

Data sources and variable construction

Fiorella 2004 Drivers Side Mirror

Three types of data were used for this analysis: (1) biomonitoring data; (2) intervention activities, biophysical and governance information; and (3) site characteristics. All data were collected by project staff at different time points and frequencies from 2012 to 2015 and made available to the authors by WorldFish.

Biomonitoring data were collected in February, May, August, and November from November 2012 through November 2015 and include water quality measurements and fish samples. Water quality data (nitrogen, phosphate, pH, temperature, and turbidity) were obtained from 500 ml water samples, taken mid-day between 11h and 13h, 10–20 m from shore and 30 cm below the surface in six locations around the CFR. Temperature was taken at each of the six water sample sites and then averaged. All six water samples were mixed in a bucket and a single sample of the mixture was used for pH, phosphate, and nitrogen measurements. Limited variation in phosphate and nitrogen variables within the CFRs precluded their use in our analyses. Water quality variables used in the statistical models below were set to reflect optimal levels in tropical fish ponds and the values CFR managers were trained to recognize, with the variable being one if the measure falls in the defined range, zero otherwise: (a) pH, normal range 6.5–8.5; (b) temperature, normal range 25–32 ºC, high range > 32 ºC; and (c) Secchi depth, a measure of turbidity and phytoplankton productivity, normal range 30–45 cm; a larger value indicating clearer water (Rouse 1979, USAID 2012, Dodds and Whiles 2010). These adjustments were made to account for the fact that increases in these variables are not uniformly problematic, but constitute a problem for fish health when they exceed a certain range. A binary variable was used because few measurements exceeded the acceptable ranges, meaning multiple bins would have few observations.

Fish samples were collected using gill nets. Pond size varied seasonally, as ponds contracted in the dry season, and sampling was responsive to these seasonal changes to avoid excessive sampling in small water bodies. In ponds larger than 0.35 hectares, long gill nets (18 m in length) were used, and in ponds less than 0.35 hectares, short gill nets (9 m in length) were used. Gill nettings were replicated eight times per time point, unless the pond was smaller than 0.25 hectares, in which case, five replicates were used. Both long and short gill nets were 2.5 m in height with alternating mesh sizes (e.g., 1.5 m length per mesh size in short gill net, 3 m length per mesh size in long gill net) of 1.5, 2, 4, 6, 8, and 10 cm (Jensen 1990). Gill nets were set between 16h and 17h and remained in the water for 12–14 hours (Minns and Hurley 1988). The depth of the gill net was recorded and we used this to calculate the submerged net area. Fish caught were grouped by species and the number and total weight of fish of each species was recorded. At a given time point, the total number and weight of fish were summed across replicates to calculate biomass and biodiversity values; because methods were standardized across CFRs, effort was equivalent across sites and time points.

Using the biomonitoring data, we constructed two measures to understand the fish community: biomass and species richness (Magurran 2004). Biomass provides a measure of the quantity of fish and species richness describes the diversity of fish species, or biodiversity. Biomass was defined as the average weight of fish caught across gill net survey replicates per square meter of submerged gill net. Species richness refers to the number of species. To account for differences in sampling effort, in our analysis we defined species richness as the total number of species caught across gill net survey replicates. Because the species accumulation function was unknown in this setting and net area varied considerably, we did not adjust species richness by gill net area and instead explicitly controlled for submerged net area in models of species richness.

Intervention activities data were collated by calendar year from 2012 to 2015. Because CFRs were adaptively managed by CFR committees, a variety of management actions were undertaken and we thus included biophysical modification and governance activities, which were coded as count variables within categories defined by WorldFish classifications on the nature of the activity, e.g., to address water quality and governance capacity building (Appendix 1).

Site characteristics were collected once in 2012 and considered fixed over time. Site characteristics included CFR type, surface area of the CFR in the dry season at baseline (e.g., minimum habitat size), location, and number of households associated with the CFR.

Statistical methodology and variable construction

To analyze the factors that are associated with CFR biomass and species richness, we used a combination of cross-sectional and panel linear regression models, including fixed effects models (Wooldridge 2001), where possible. CFR fixed effects regressions identified the associations between biomass or species richness and explanatory variables of interest using the within-CFR intertemporal variation, controlling for all time invariant differences among CFRs, both observed and unobserved. Our sample comprised 40 CFRs monitored over three years. Our data contain time-invariant (CFR characteristics), annual (CFR interventions, year), and seasonal (13 time points; water quality) variables. To compile and analyze data, we used R vegan package (Oksanen et al. 2019) and Stata version 14 (Stata 2017).

Species richness and biomass in CFR i, at time t, represented by B_it, are conceptualized as a function of annual covariate flood, temperature, and rainfall fluctuations, T_t, that affect all CFRs; time-invariant CFR characteristics X_i; time-varying CFR governance and biophysical interventions, V_it, and time-varying CFR characteristics such as water quality, W_it:

Bit = f(Tt,Xi,Vit,Wit)

(1)

We consider relationships 1, 2, and 3 from Figure 1 each in a separate regression model, and then examine relationship 4 in a model that simultaneously nests relationships 1, 2, and 3. Relationship 5, also shown in Figure 1, is accounted for in all models as a vector of year indicator variables, T_t, the coefficient on which tell us the role of annual weather, flood, or other events experienced uniformly across the sample, while better isolating the other relationships. Specific regression models for each relationship are detailed in turn below. We aim to describe the interacting dynamics associated with biomass and biodiversity within the CFRs rather than identify causal relationships. Although our analysis provides the opportunity to assess factors potentially associated with CFR biomass and biodiversity, limitations include sample size, problems of multiple comparisons, the inability to specify multiple components of the biological community (biomass and species richness) within the same model, and the challenges of postintervention analysis using observational data.

Relationship 1: Site characteristics related to biomass and species richness

To understand the association between CFR characteristics (CFR category, province, agro-ecological zone, number of households, and CFR area) and biomass and species richness, we estimate a pooled ordinary least squares regression as follows:

Bit = α + β1Ci + β2Pi + β3Ei + β4Hi + β5Zi + β6Nit + β7Tt + eit

(2)

C_i represents a vector of binary variables, each indicating one of the four CFR categories (community ponds without flooding, community ponds with flooding, and segments of a large water body, with irrigation reservoirs as the omitted reference category). P_i is a vector of indicators for the province in which the CFR is located; and E_i is a vector of binary variables representing agro-ecological zones (upland irrigated rice, upland crops/rainfed rice, rainfed lowland rice, with lowland deepwater rice as the omitted reference category). Agro-ecological zones are distinct from CFR type and describe the area surrounding the CFR. Additionally, population pressure on the CFR is captured by, H_i, which is the number of households living in the CFR’s zone of influence at baseline, and baseline surface area of the CFR during the dry season is captured in Z_i. Finally, we included a vector of year indicators (T_t) to capture annual drivers of biomass and species richness that are common to all CFRs, such as those caused by annual weather and flood variation. We expected that each of these factors would affect CFR biomass and biodiversity, with larger CFR size and smaller population having a positive impact but no a priori predictions for CFR type, province, or agro-ecological zone.

Because of population changes over time and intra-annual fluctuations of CFR size (m²) depending on season and timing of rainfall, H_i and Z_i are likely measured with some degree of error, but we include them to control for vast background differences between CFRs’ size and population base. When species richness is the outcome, we also include a control for the average submerged net area (m²) during biomonitoring, N_it, in order to account for net size adjustments in the biomonitoring sampling protocol.

We interpret coefficients β₁ - β₆ as the conditional relationships between the respective variables and the outcome across all years and CFRs, on average. In this equation, α represents the y-intercept, or the constant if all other variables are set to 0, and e_it is the error term. Pooled ordinary least squares (OLS) regressions are particularly vulnerable to bias because of unobserved heterogeneity among the CFRs, therefore we favor the CFR fixed effects regressions below, which control for this heterogeneity. We present the results from Equation (2) as descriptive associations that highlight the importance of site-specific, time invariant characteristics that are statistically accounted for in the fixed effects regressions described below and further motivate the use of fixed effects analysis in the models that follow.

Relationship 2: Water quality variables related to biomass and species richness

To understand the impact of water quality on CFR biomass and species richness, we used a 13-period panel of biomonitoring data. Biomonitoring data was collected every three months and we model relationships of CFR i at period p, where subscript p represents three-month periods and differs from the annual t subscript in model (2) above. We estimate the fixed effects regression

Bip = γ1mi,p + γ2mi,p-1 + γ3Pi,p + γ4Pi,p-1 + γ5si,p + γ6si,p-1 + γ7s²i,p + γ8s²i,p-1 + γ9ni,p + γ10Op + μi + εi,p

(3)

in which the outcome is a function of whether the water temperature was above 32 ºC in the current period, m_i,p, and, because temperature effects may not always be immediate, in the previous period, m_i,p-1. Similarly, we included a binary indicator equal to one if the pH was in the ideal range in the current period, P_i,p, or in the previous period, P_i,p-1. Finally, we included continuous values of Secchi depth (cm), s_i,p, and its square, s²_i,p, in the current and previous period (s_i,p-1 and s²_i,p-1, respectively). We included a square term for Secchi depth to account for diminishing returns to water clarity as it moves out of the ideal range (see Appendix 2 for summary statistics). We used the natural log of all Secchi depth values. We also included, when species richness was the outcome, a control for the average submerged net area during biomonitoring, n_i,p. We included a vector of month and year indicators, O_p, to control for seasonality, including the seasonal life cycles and growth patterns of fish. The composite error term consisted of μ_i, the unobserved CFR effect, and ε_i,p, the idiosyncratic error. The fixed effects model allowed us to control for any time-invariant, unobservable CFR characteristics that may influence the outcomes of interest.

Relationship 3: Management actions related to water quality

To assess how management interventions impacted water quality, we modeled the relationship between water quality, W_ip, in CFR i during period p and biophysical interventions carried out by the CFR management committees in the preceding year. We estimated the following fixed effects regression:

Wip = αi + δ1Vi,t-1 + δ2Qi,t-1 + δ3Mp+ δ3Tp + μi + εip

(4)

Outcome variables, W_ip, are a binary indicator of whether the water temperature was ≥ 32 ºC, the natural log of Secchi depth (cm), and a binary indicator of whether the pH levels fell into the ideal range (6.5–8.5 moles/liter). Binary outcomes are estimated as a linear probability model, using an OLS estimator with fixed effects. The model includes a count of CFR environment-related activities, V_i,t-1, and a count of water quality activities, Q_i,t-1. Note that all outcomes vary intra-annually, while some explanatory variables (V_i,t-1 and Q_i,t-1) vary at the annual level only. We also included a vector of month and year indicators, (M_p and T_p), to control for seasonal and annual fluctuations common to all CFRs, including the seasonal life cycles and growth patterns of fish and covariate annual rainfall, temperature, and flood patterns. The composite error term consisted of μ_i, the unobserved CFR effect and ε_ip, the idiosyncratic error. Again, we were able to estimate a CFR fixed effects model, allowing us to control for unobservable CFR characteristics. Outcomes in a given year were regressed on the activities of the prior year to ensure that interventions took place prior in time to outcomes.

Environment-related activities (V_i,t-1) included building fences and ecoshelters, adding fertilizer, fingerlings, broodstock, adding or removing vegetation, expanding the CFR, planting flooded forests or purchasing a boat for monitoring and patrols (Appendix 1). Water quality related (Q_i,t-1) activities included adding channels, inlets, outlets, and pumps, building dams, and deepening the CFR. Because CFR committees implemented activities in response to conditions within their CFR, our measure provided a real-world metric that captured the impact when CFR committees reacted to changes in their system, elected a site-specific management action, and carried it out. The wide range of activities and frequencies CFR committees elected to implement were too low to disaggregate by individual activity. Consequently, we were not able to address the productivity effects of specific actions taken by CFRs.

Relationship 4: Management actions related to biomass and biodiversity

Finally, we examined the relationship between the governance and biophysical activities of the CFR committees and species richness and biomass, in effect nesting all of the preceding models. We estimated model (2), while controlling for CFR fixed characteristics (model 1) with a CFR fixed effect and included water quality controls as in model (3). We can then attribute any remaining association to relationship 4. In addition to controlling for unobserved physical characteristics, the CFR fixed effect controls for unobserved differences in CFR governance capacity. As governance actions were adaptive in response to local conditions, we characterized governance using a count variable, i.e., number of actions taken by the CFR committee, that allowed us to examine the composite effect of committees’ decision to act, the selection of management action, and the outcome of those efforts. Although our characterization accounted for the real-world action taken by committees, it does restrict the inference we can make and individual management actions cannot be assessed.

Bit = σ1Vit + σ2Qit + σ3Mit + σ4Ait + σ5Nit + σ6WFit + σ7Tt + ζi + eit

(5)

Outcome variables, B_it, were aggregated by calendar year and regressed on count variables indicating the number of CFR activities in a given category in a given year. Activities were categorized into CFR environment-related activities (V_it) and water quality activities (Q_it) exactly as in model (4), but using the current period rather than the previous, as we did not have enough annual time periods to support analysis with annual lags. Additionally, M_it represents a count of various governance meetings, including awareness-raising events on multiple topics, coaching sessions on management skills, and reflection worskhops; A_it, represents governance activities such as building a new CFR, demarcating the boundaries of the CFR, adding signs, paths, guard infrastructure, holding elections of CFR committees, establishing bylaws, holding events, organizing savings groups, and integrating CFR actions into broader community investment plans (Appendix 1). These sets of variables reflect the categorizations made by WorldFish staff in communications with CFR managers. As in model (2), N_it, is included for all species richness regressions to explicitly control for the area of net submerged during biomonitoring data collection (m²). We included W^F_it, a vector of water quality variables from February of each year, the month in which we would expect the poorest water quality because of lack of rainfall during the dry season, so that we can interpret σ₁-σ₄ as the relationship between each activity, independently of water quality.

RESULTS

Descriptive statistics for the CFRs (n = 40) across all years (2012–2015) are reported in Appendix 2 and a summary of results is provided in Appendix 3.

Relationship 1: Site characteristics related to biomass and species richness

To understand the site characteristics that correlate with biomass and biodiversity across CFRs, we estimated via pooled OLS regression the conditional associations with CFR category, agro-ecological zone (AEZ), number of households, CFR size, and year (Table 1).

In community ponds without flooding, biomass is significantly higher relative to irrigation reservoirs. CFRs in Battambang province had greater biomass compared to CFRs in Pursat and Siem Reap, and greater species richness than CFRs in Kampong Thom and Pursat. CFRs in rainfed lowland rice agro-ecological zones had significantly higher biomass and species richness compared to those in lowland deepwater rice agro-ecological zones. Larger CFR area during the dry season was associated with lower biomass and higher species diversity. Finally, species richness was systematically higher in 2012 (reference year) compared to the three years that followed. An alternate model that estimates these relationships separately by year is available in Appendix 4.

The effect sizes of our coefficient estimates represented an increase or decrease in the log of kilograms fish caught per meter squared of gillnet (biomass) and the number of species (species richness). Thus, the effect sizes we found of approximately 0.2–0.7 in biomass and 1.7–7 in biodiversity represent a biologically meaningful difference in these metrics. For reference, the annual mean log kg catch/m² was 3.52, with a standard deviation of 0.77, while mean species richness was 14.6, with a standard deviation of 5.2.

Relationship 2: Water quality variables related to biomass and species richness

Using fixed effects for CFR and time period to account for observable and unobservable time-invariant site-specific features of CFRs, some of which emerge above, we found significant relationships between biomass with seasonal (month) and annual effects when controlling for differences in water quality in the current and previous periods (Table 2). Within CFRs, species richness was significantly lower in August and November, compared to February. Within CFRs, biomass was significantly lower in May and August and significantly higher in November, compared to February. Species richness was significantly higher in 2015, compared to 2013.

We also find associations between biomass and water quality in the previous period (Secchi depth, high water temperature, pH). In the previous period, pH in the ideal range and a high water temperature were negatively associated with species richness. Although the pH result is counterintuitive, the large majority of pH readings were in the ideal range (Appendix 1). Previous period Secchi depth was positively associated with biomass, with diminishing marginal effects.

Relationship 3: Management actions related to water quality

Using fixed effects for CFR, we found very limited evidence of relationships between activities designed to benefit CFR water quality or environment and dimensions of water quality (Table 3). We use a broad measure of management activities, a count of biophysical interventions that influence CFR water quality or CFR environment in the previous period. We found that the year 2015 was negatively associated with high water temperature and positively associated with ideal pH, and that the number of activities intended to improve water quality appears to be negatively associated with ideal pH. However, we cannot conclude definitively that management activities did or did not impact water quality. Further, management activities might be important in forestalling waning water quality as opposed to improving water quality, and within this observational study we are unable to disentangle this possible effect.

Relationship 4: Management actions related to biomass and species richness

Using fixed effects for CFR, we examined the relationship between CFR activities, governance, and fundraising while controlling for water quality and annual effects (Table 4). Meetings held to coach or enhance knowledge were positively associated with species richness, though the effect size was small. CFR fundraising is positively associated with biomass. See Appendix 5 for logit and probit estimates of the same model.

Relationship 5: Annual weather and flood fluctuations

We found a broad influence of exogenous seasonal and annual weather and flood variations across the biophysical elements of CFRs. The relationship of seasonal and annual weather and flood variations is accounted for in all models. Seasonal patterns within the floodplain were responsible for changing biomass and species diversity, independent of water quality, throughout the course of the year from the dry season beginning around December and ending in April to the connectivity from CFRs to the rice fields in May to November. Further 2015 emerged as a particularly unique year (Tables 2, 3, 4). However, the site and water quality characteristics that influence biomass and biodiversity differed across years, suggesting, as predicted, that annual covariate effects may strongly shape this system (Appendix 4).

DISCUSSION

The rice field fisheries environment is home to a tremendous diversity of fish and aquatic species and supports access to food and livelihoods (Baran et al. 2014). Our results demonstrate that within CFRs, biomass and biodiversity of these inland protected areas are associated with season, annual weather, and flood fluctuations, and characteristics of the CFR site, i.e., province, agro-ecological zone, and surface area. Independently of these important characteristics, elements of CFR governance, e.g., fundraising or governance meetings, are associated with species richness and biomass. Our findings shed light on the function of inland protected areas in dynamic environments and suggest a strong role of seasonal, annual, and site-specific factors.

Previous quantitative analyses of social-ecological systems have used linear regressions (Leslie et al. 2015), random forest (Gill et al. 2017), structural equation modeling (Filbee-Dexter et al. 2018), and modeling projected results (Ziegler et al. 2017). Our analytic approach of fixed effects panel regression models is widely used in economics and other disciplines (Wooldridge 2001) but represents a novel approach to analyzing social-ecological systems. It extends the quantitative analysis to investigate the productivity of the CFR social-ecological systems through comparison of each site to itself across time. Because social-ecological systems rarely present the opportunity for controlled experiments, our approach allows for comparisons of each site to itself over time to identify generalizable trends with observational data.

Our analysis highlights that the drivers of diversity and productivity within inland protected areas may vary annually. For example, comparisons of the site characteristics of CFRs across multiple years indicate that particular provinces, agro-ecological zones, and surface area are associated with fish productivity and biodiversity (Table 1). However, comparing site characteristics in only a single year results in different characteristics emerging as significant in different years, e.g., rainfed lowland rice agro-ecological zone in 2012, CFR surface area in 2015 (Appendix 4). Within these dynamic social-ecological systems, accounting for strong site and annual effects over the longest time horizon possible is thus necessary to assess how shifting elements of their management, including habitat quality, biophysical changes, and governance, shape outcomes over time.

In our results, evidence as to whether the intensity of CFR management activities is associated with biomass or species richness within CFRs, however, is limited. Purposive site selection based on likelihood of success potentially introduces bias in observational data, potentially biasing our results toward a positive effect of management actions, and the absence of this finding thus strengthens the likelihood of there being no effect. However, limitations in our ability to detect a relationship due to imprecision and measurement error in intervention variables or too short a time period of study to observe associations are also possible explanations.

The absence of an effect between CFR biophysical management activities and biomass or species richness could also be due in part to the challenges in tailoring management actions to a particular setting. The extent to which management intensity should be linked to improved outcomes is difficult to predict because those sites with more management actions might also be facing more challenges or working harder to forestall declining biomass, obscuring a potential relationship. There may also be a mismatch between the intent of the resource managers, e.g, to improve water quality, and the outcome of their efforts. Although managers responded to what they observe and understand about their system and were provided technical support and tools to monitor water quality, identifying and carrying out appropriate management is complex and may have idiosyncratic outcomes. Actions that have a positive effect in one protected area may have a neutral or negative impact on productivity in another. Further, the productivity of social-ecological systems over time is entwined in the bidirectional feedbacks between natural system conditions and resource manager actions that respond to them, and thereby cyclically change natural conditions.

The positive associations we do observe between governance capacity building suggest that strengthening management capacity and motivating management investment in inland protected areas is of value. These mechanisms further provide for community-led adaptive management, whereby community members respond to changing environmental conditions and alter their management strategies. Although strong site-specific and annual effects may mean that detecting effects of management efforts within CFRs may be challenging or that their effects may be minimal relative to seasonal changes, the implementation of the CFR itself may still increase the productivity of the surrounding rice field fisheries, which we do not measure here. Analysis of the benefits of CFRs for fishery productivity and particularly to local populations are important next steps.

Our three-year time period, while longer than most previous studies of inland protected areas, may not be long enough to fully assess the impact of committee actions on the productivity of this dynamic inland ecosystem where CFRs may foster fish recruitment over longer time horizons. Other studies have found that it can take from 3 to 10 years before self-sufficient community organizing is in place to support sustainable fisheries management (Pomeroy and Berkes 1997, Pomeroy and Carlos 1997), and that observing effects of protection on marine species may take more than five years (Babcock et al. 2010). Further, the planning of a no-take reserve has also increased fishing pressure prior to implementation (McDermott et al. 2019), a dynamic that could alter the baseline biomass. These areas are ripe for future explorations of the relationships between interventions in inland protected areas and key outcomes, using precisely tracked interventions over an even longer time period than we offer here.

Our findings demonstrate the annual and seasonal effect of the flood pulse system on sites within the Tonle Sap floodplain. The established role of the flood pulse system in driving system productivity (Kummu and Sarkkula 2008) and biodiversity (Thomas et al. 2017) is particularly highlighted by the patterns seen in year 2015. In 2015, the flood arrived late and it was the start of an El Niño year, which has been shown to affect the annual flood dynamics (Ward et al. 2016). For several CFRs, the annual flooding failed to connect some CFRs and rice fields to the Tonle Sap Lake and other large water bodies, with some communities anecdotally reporting that fish from their CFRs were instrumental in providing for fish catch within rice fields. CFRs and fishery management in this dynamic system could play an important role in forestalling declines in water quality or resolving water user disputes in variable or bad years. However, CFRs function within the flood pulse system and are unlikely to be a viable management strategy if the system is broadly disrupted by irrigation or dam development along the Mekong River because disruption of the annual flood-pulse will prevent flooding waters from reaching CFRs and disrupt connectivity within the system (Lamberts and Koponen 2008, Arias et al. 2014).

GROUNDED SPECULATION AND CONCLUSIONS

Our findings highlight the important role of site-specific and seasonal characteristics in shaping social-ecological systems. Although successful elements of biophysical changes and governance may not always be identifiable, equipping local communities to monitor their social-ecological systems and with opportunities to strengthen adaptive management holds considerable promise for sustainable social-ecological systems. In the context of complex,_blank'>https://doi.org/10.1577/1548-8675(1988)008<0216:EONLAS>2.3.CO;2

Miratori, K., and A. Brooks. 2015. Good governance of rice field fishery management. WorldFish, Penang, Malaysia.

Oksanen, J., F. G. Blanchet, M. Friendly, R. Kindt, P. Legendre, D. McGlinn, P. R. Minchin, R. B. O'Hara, G. L. Simpson, P. Solymos, et al. 2019. vegan: Community ecology package. Version 2.5-5. [online] URL: https://cran.r-project.org/package=vegan

Olsson, P., C. Folke, and F. Berkes. 2004. Adaptive comanagement for building resilience in social-ecological systems. Environmental Management 34:75-90. https://doi.org/10.1007/s00267-003-0101-7

Ostrom, E. 2007. A diagnostic approach for going beyond panaceas. Proceedings of the National Academy of Sciences of the United States of America 104:15181-15187. https://doi.org/10.1073/pnas.0702288104

Ostrom, E. 2009. A general framework for analyzing sustainability of social-ecological systems. Science 325:419-422. https://doi.org/10.1126/science.1172133

Pomeroy R. S. and N. L. Andrew. 2011. Small-scale fisheries management: frameworks and approaches for the developing world. CABI, London, UK. https://doi.org/10.1079/9781845936075.0000

Pomeroy, R. S., and F. Berkes. 1997. Two to tango: the role of government in fisheries co-management. Marine Policy 21:465-480. https://doi.org/10.1016/S0308-597X(97)00017-1

Pomeroy, R. S., and M. B. Carlos. 1997. Community-based coastal resource management in the Philippines: a review and evaluation of programs and projects, 1984-1994. Marine Policy 21:445-464. https://doi.org/10.1016/S0308-597X(97)00016-X

Ratner, B. D., E. J. V. Oh, and R. S. Pomeroy. 2012. Navigating change: second-generation challenges of small-scale fisheries co-management in the Philippines and Vietnam. Journal of Environmental Management 107:131-139. https://doi.org/10.1016/j.jenvman.2012.04.014

Rouse, D. 1979. Water quality management in pond fish culture. Research and Development Series No. 22. International Center for Aquaculture, Agricultural Experiment Station, Auburn University, Auburn, Alabama, USA.

Sarkar, U. K., A. K. Pathak, L. K. Tyagi, S. M. Srivastava, S. P. Singh, and V. K. Dubey. 2013. Biodiversity of freshwater fish of a protected river in India: comparison with unprotected habitat. Revista de Biología Tropical 61:161-172. https://doi.org/10.15517/rbt.v61i1.10942

Schlüter, M., R. R. J. McAllister, R. Arlinghaus, N. Bunnefeld, K. Eisenack, F. Hölker, E. J. Milner-Gulland, B. Müller, E. Nicholson, M. Quaas, and M. Stöven. 2012. New horizons for managing the environment: a review of coupled social-ecological systems modeling. Natural Resource Modeling 25:219-272. https://doi.org/10.1111/j.1939-7445.2011.00108.x

Society for Ecological Restoration International Science & Policy Working Group. 2004. The SER primer on ecological restoration. Society for Ecological Restoration, Washington, D.C., USA.

Stata. 2017. Stata statistics/data analysis. StataCorp LP, College Station, Texas, USA.

Thilsted, S. H., A. Thorne-Lyman, P. Webb, J. R. Bogard, R. Subasinghe, M. J. Phillips, and E. H. Allison. 2016. Sustaining healthy diets: the role of capture fisheries and aquaculture for improving nutrition in the post-2015 era. Food Policy 61:126-131. https://doi.org/10.1016/j.foodpol.2016.02.005

Thomas, P., G. Holtgrieve, V. Elliott, K. McCann, B. McMeans, N. Rooney, A. Smits, T. Phanara, M. Cooperman, S. Clark, C. Phen, and S. Chhuoy. 2017. Seasonal increases in fish trophic niche plasticity within a flood-pulse river ecosystem (Tonle Sap Lake, Cambodia). Ecosphere 8:e01881. https://doi.org/10.1002/ecs2.1881

USAID. 2012. Cambodia HARVEST: Helping Address Rural Vulnerabilities and Ecosystem Stability. Technical Bulletin #33: How to Fertilize a Fish Pond. USAID, Phnom Penh, Cambodia.

Vogt, J. M., G. B. Epstein, S. K. Mincey, B. C. Fischer, and P. McCord. 2015. Putting the “E” in SES: unpacking the ecology in the Ostrom social-ecological system framework. Ecology and Society 20(1):55. https://doi.org/10.5751/ES-07239-200155

Fiorella 2004 Drivers Ed

Ward, P. J., M. Kummu, and U. Lall. 2016. Flood frequencies and durations and their response to El Niño Southern Oscillation: global analysis. Journal of Hydrology 539:358-378. https://doi.org/10.1016/j.jhydrol.2016.05.045

Wooldridge, J. 2001. Econometric analysis of cross section and panel data. MIT Press, Cambridge, Massachusetts, USA.

Fiorella 2004 Drivers Wanted

Young, T. P., D. A. Petersen, and J. J. Clary. 2005. The ecology of restoration: historical links, emerging issues and unexplored realms. Ecology Letters 8:662-673. https://doi.org/10.1111/j.1461-0248.2005.00764.x

Zedler, J. B. 2000. Progress in wetland restoration ecology. Trends in Ecology & Evolution 15:402-407. https://doi.org/10.1016/S0169-5347(00)01959-5

Fiorella 2004 Drivers License

Ziegler, J. P., E. J. Golebie, S. E. Jones, B. C. Weidel and C. T. Solomon. 2017. Social-ecological outcomes in recreational fisheries: the interaction of lakeshore development and stocking. Ecological Applications 27:56-65. https://doi.org/10.1002/eap.1433