Research

INTRODUCTION

Coral reefs are highly diverse social-ecological systems (SESs) that provide food, jobs, recreation, and tourism for people, and habitat for more than one million aquatic species around the world (Goatley et al. 2016, Sully et al. 2019). Further, they offer protection for coastal SESs, afford opportunities for novel medical discoveries, and undergird the commercially harvested fish supply at local, regional, and global scales (Santavy et al. 2021). However, due to a confluence of issues, including climate change, ocean acidification, coastal development, overfishing, and invasive species, these essential SESs have become severely degraded over the last half-century, resulting in significant global impacts. For example, coral cover in the Caribbean decreased by 80% from 1975 to 2000 (Contreras-Silva et al. 2020), and the climate-induced El Niño and Southern Oscillation (ENSO) events that occurred from 2014 to 2017 triggered the most extreme and widespread global coral bleaching event on record (Eakin et al. 2018, 2019). Moreover, rising temperatures and other stressors have led many researchers to contend that a substantial portion of coral reefs are at risk of dying within this century (e.g., Dance 2019, Santavy et al. 2021). Although it is possible for some coral species to recover from stressor events (e.g., Graham et al. 2011, Steneck et al. 2019), proactive steps must be taken to protect these invaluable resources and the seascapes (ocean spaces that are dynamic and spatially and temporally interconnected; Pittman et al. 2021) that house them. To that end, it is critical to examine the condition of these SESs and develop strategies for monitoring and managing coral reef SESs to ensure that they persist in light of accelerating environmental change.

The concept of social-ecological resilience provides a robust framework for studying environmental change because it not only considers a system’s ability to absorb perturbations and persist while maintaining functions, structures, and processes, it also accounts for nonlinear change in SES dynamics and alternate regimes (Holling 1973, Jozaei et al. 2022). In the context of coral reefs, resilience pertains to the ability of these SESs to withstand disturbances and continue to thrive in a regime that sustains ecological functions, structures, and processes (Marshall et al. 2006, Anthony et al. 2015). Further, resilience highlights the importance for coral reef SESs to build adaptive capacity (Angeler et al. 2019) to change and resist transitioning to alternate regimes (see Allen et al. 2019) characterized by coral decline, reduction in coral diversity, and loss of the ability to provide coastal protection and habitat for aquatic species (Toth et al. 2023).

Various quantitative methods have been used to study coral reefs. For example, Vo et al. (2019) used semi-quantitative methods to examine the impact of anthropogenic effects and water temperature on aspects of coral reef resilience in Vietnam. Index-based resilience assessments have provided the basis for examining coral reef health in the Caribbean Sea (Díaz-Pérez et al. 2016), assessing relative resilience of coral reefs in Puerto Rico and American Samoa (Schumacher et al. 2018, Gibbs and West 2019), and exploring potential management actions in Indonesia (McClanahan et al. 2012). Contreras-Silva et al. (2020) and Illosvay et al. (2020) used meta-analysis to study temporal change in benthic cover patterns and subsequent drivers in the Mexican Caribbean. Modelling and Bayesian approaches have been applied in a variety of studies, such as identifying coral reef regimes in the Seychelles and Hawaiian Islands (Graham et al. 2015, Donovan et al. 2018), predicting global coral bleaching patterns (Sully et al. 2019), and examining the potential impact of management activities on coral reef ecosystems (Carriger et al. 2019). Further, researchers have investigated patterns and trends in coral reef condition to assess the impact of climate change on coral reefs (e.g., Lapointe et al. 2019, Romero-Torres et al. 2020), with some employing satellite imagery to enhance the analysis (Newnham et al. 2020, Roelfsema et al. 2020). Although this overview provides a snapshot of some of the diverse approaches used to study coral reef ecosystems, it is important to note that there has been limited exploration of the complex spatial and temporal dynamics of coral reef SESs necessary for assessing the resilience of coral reef SESs at the global scale (Lawrence et al. 2021).

Resilience-based approaches capture the effects of environmental change by assessing variation in system dynamics to detect regimes and regime shifts, which are critical aspects of social-ecological resilience (Spanbauer et al. 2014, Eason et al. 2016). Researchers studying ecosystem dynamics have often relied on the theory that a system “slows down” as it approaches a bifurcation point (Dakos et al. 2008, 2012). Univariate indicators (e.g., rising variance, skewness, kurtosis, autocorrelation) have traditionally been used to examine system behavior near a threshold (Scheffer et al. 2009, 2015, Clements and Ozgul 2018). These indicators are appealing because they help to focus on general trends in the variables without requiring a priori knowledge of mechanisms or system processes (Sundstrom et al. 2017, Eason et al. 2019). However, given that multiple variables are often needed to characterize the condition of SESs, multivariate methods are particularly useful because they afford the ability to account for the inherent complexity of SESs, which is critically important when drivers are unknown (Eason et al. 2016, Sundstrom et al. 2017). Two resilience-based methods that have been used to examine multivariate data are the variance index (VI) and Fisher information (FI).

The VI detects the dominant variance component in a multivariate system and is computed as the maximum eigenvalue of the covariance matrix (Brock and Carpenter 2006). It was initially used to monitor broad-scale pollutant dispersion and subsequent impacts on ecosystem services (Brock and Carpenter 2006), but more recently, it has been included in a suite of tools for assessing ecosystem dynamics (Eason et al. 2016, Sundstrom et al. 2017). FI is an information theory-based measure of dynamic order in data (Fisher 1922) that has been adapted to detect patterns in the variables used to characterize a complex system (Mayer et al. 2006, Karunanithi et al. 2008). The core of these methods lies in their ability to evaluate multiple system variables simultaneously to assess changes in system dynamics (Eason et al. 2016, Sundstrom et al. 2017). Multivariate resilience-based methods have been employed to examine a variety of model and real human and natural systems (e.g., Karunanithi et al. 2008, Spanbauer et al. 2014, Eason et al. 2016, Sundstrom et al. 2017). The strength of these approaches lies in their: (1) ability to identify latent characteristics (e.g., regimes and regime shifts) in complex systems in which drivers (fast and slow variables) are unknown, (2) utility in handling a variety of variable types, (3) capacity to handle noisy data, and (4) ability to collapse multiple variables into an index that can be tracked over time or space (Sundstrom et al. 2017, Eason et al. 2019).

Although these methods have been effective for temporal assessments, their use for spatial analysis has been limited (Sundstrom et al. 2017, Eason et al. 2019, 2022). Given the inherently spatial nature of coral reef SESs, it is important to determine whether these methods are effective for evaluating spatiotemporal dynamics at the global scale. Such an examination fits in line with research priorities established in the growing area of seascape ecology (“spatially explicit ecological science”), which includes themes related to examining seascape change, connectivity, spatial and temporal scales, ecosystem-based management, and emerging technologies and metrics (Pittman et al. 2021).

Here, our objective is to test the utility of resilience-based approaches for examining spatiotemporal change in coral reef SESs. Using publicly available coral reef data from 2003–2016, we employ two of the more commonly used univariate indicators (variance and skewness) and two multivariate indices (FI and VI) to assess environmental change in coral reef SESs. We seek to shed light on whether and how resilience-based approaches can be used to examine spatiotemporal change over the global coral reef seascape. Our effort aims to improve the understanding of effects of environmental change on coral reef SESs, to undergird effective assessment and management of the seascapes that house them, and to provide subsequent linkages to social-ecological resilience.

MATERIALS AND METHODS

Data

To find suitable data to test the methods, we searched for an existing coral reef study that provided publicly available data with adequate spatial (i.e., global) and temporal coverage (at least 10 years). In our survey, we found that many published studies did not provide publicly accessible data (e.g., Wedding et al. 2018, Houk et al. 2020), or offered limited temporal or spatial coverage (e.g., Gibbs and West 2019, Jouffray et al. 2019, Ilosvay et al. 2020), thereby reducing their utility for our effort. The data set used for our analysis was published by researchers examining global coral bleaching (Sully et al. 2019) and was selected because of its spatial and temporal coverage: 19 years of data and up to 859 coral reef stations per year. Unlike the other case studies, it included data on bleaching and a plethora of information on temperature, which is an important factor and known driver of coral reef decline in many affected coral reef SESs (Dance 2019, Santavy et al. 2021). The data were made available through GitHub and were compiled from data collected by Reef Check and the National Oceanic and Atmospheric Administration’s National Centers for Environmental Information (Sully et al. 2019).

The data set included 30 temperature variables, but many of them were aggregated metrics (e.g., mean sea surface temperature [SST], sea surface temperature anomaly [SSTA] frequency). Given the goal of our study, we opted to use the base metrics (e.g., SSTA) and selected a time period with the highest resolution (2003–2016). Our final data set (Eason 2024) consists of nine variables: average bleaching, climatological SST (ClimSST), depth, diversity, rate of SST change, SSTA, SSTA degree heating weeks (SSTA DHW), thermal stress anomaly (TSA), and windspeed (Table 1). The spatial resolution of the data set ranged from 392 to 600 coral reef sampling stations per year, and values for all variables were provided for each station surveyed. Due to data limitations, no Arabian Gulf coral reefs were available in 2005 and 2008, and data from only one station was provided in 2006 and 2007.

Because we were performing a macro-level assessment, we amalgamated station names and associated data when the locations (latitude and longitude) were essentially the same from year to year. Before commencing the study, we explored basic patterns in the data and found that aside from windspeed and bleaching, the means of the variables tended to increase over time (Fig. A1 in Appendix 1), and there were only two statistically significant correlations between the study variables (longitude and diversity: ρ = 0.65, P < 0.05; and SSTA and TSA: ρ = 0.56, P < 0.05). Although the values of most parameters varied broadly or were skewed, the temperature variables (except for ClimSST and SSTA DHW) were largely gaussian and nearly symmetrical (Fig. A1 in Appendix 1).

Employing resilience-based approaches to assess coral reef social-ecological systems

In previous studies, time and linear proximity served as natural principles for ordering the data (e.g., Spanbauer et al. 2014, Eason et al. 2016, Sundstrom et al. 2017); however, these approaches limit the types of questions that can be examined in a spatial context (Sundstrom et al. 2017). For true geospatial analysis, the data must be assessed based on sampling station location (latitude and longitude) to capture patterns of change across the seascape. Accordingly, in line with Eason et al. (2019, 2022), we used Haversine distance as the ordering parameter because it effectively captures the curvature of the Earth. After sorting the stations (and data) by distance from a reference location (minimum latitude and longitude of the station locations), the data were then divided into moving windows that iteratively captured small geographical “regions” based on proximity to the reference station (Eason et al. 2019, 2022). Variance, skewness, and VI were computed on the sorted data using standard statistical functions in Matlab (release 2021). Because FI is a more complex approach, we first provide a basic theoretical foundation before discussing the index’s computation.

FI is based on the probability of observing particular system states, p(s), and is proportional to the change in probability, dp(s), vs. the change in state, ds (Mayer et al. 2007).

(1)

Accordingly, a system with a high probability of experiencing a particular set of conditions is biased toward a specific state and exhibits more stable patterns; hence, it has higher FI. Conversely, a highly variable system tends not to exhibit consistent patterns or conditions and is characterized by a low FI (Mayer et al. 2007).

System states relate to the condition of a system as characterized by a set of measurable variables (s: x₁, x₂, .... x_n; Karunanithi et al. 2008). Given that system conditions may vary due to internal dynamics or external stressors, a system may experience small perturbations or could undergo substantial change. Further, because uncertainty is inherent in any measurement, a tolerance or size of states (sost) is used to distinguish minor fluctuations from significant variation (Karunanithi et al. 2008). A sost is defined for each variable such that if the measured value varies within a finite range (e.g., |x₁(t₁) − x₁(t₂)| ≤ sost_x1), the two points are deemed as indistinguishable and counted (binned) as observations of the same state (Karunanithi et al. 2008, Cabezas and Eason 2010).

Using this binning approach, p(s) is calculated by counting the number of points that fit within a particular state. Through a series of derivation steps, Eq. 1 was adapted to handle empirical data from real systems (Karunanithi et al. 2008).

(2)

where p(s) is replaced by its amplitude (q²(s) ≡ p(s)) to reduce calculation errors from very small p(s) (Karunanithi et al. 2008). Further details on the derivation and calculation are found in Mayer et al. (2007), Karunanithi et al. (2008), and Cabezas and Eason (2010).

Basic steps to compute FI are as follows: (1) gather data for each station location sorted by Haversine distance; (2) compute the sost for each variable; (3) for each window, group points into states of the system based on the sost; (4) divide the number of points grouped into each state by the total number of points in the window to produce p(s); and (5) use Eq. 2 to compute FI, where q(s) = √p(s). These steps are repeated to compute FI for each window. Computations are completed for each year, producing a set of values at the corresponding survey station locations over the global coral reef seascape. The data for this study were managed, analyzed, and visualized using Microsoft Excel 365, Matlab (release 2021), Python, and JMP 14.

Using the indicators and indices to assess change in social-ecological systems dynamics

As previously noted, variance, skewness, and VI are expected to increase approaching a bifurcation (Scheffer et al. 2009, 2015, Clements and Ozgul 2018). FI typically declines prior to a regime shift (e.g., Mayer et al. 2007, Eason and Cabezas 2012, Eason et al. 2016); however, researchers have found that the behavior of FI in the neighborhood of a transition heavily depends on trends in the underlying variables as the system approaches a regime shift (Eason et al. 2014, González-Mejía et al. 2015). While a high FI value is typically associated with a greater degree of dynamic order (stable patterns), the level of dynamic order is not as important as the ability of the system to remain within a desirable regime; hence, a system regime is denoted as a period that is relatively stable and has a high mean FI (↑μFI), low standard deviation in FI (↓σFI), low coefficient of variation in FI (↓cvFI), and low mean VI (↓μVI). Conversely, transitions are periods in which the structures and processes in a particular regime begin to degrade; accordingly, they are characterized by disorder (instability) and are identified as periods with relatively low mean FI, high standard deviation in FI, high coefficient of variation in FI, and high mean VI (↓μFI, ↑σFI, ↑cvFI, and ↑μVI, respectively; González-Mejía 2011, Eason and Garmestani 2012, Eason et al. 2022). Here, high↑ (or low↓) values are those that are above (or below) the mean value for the period or region being assessed (e.g., high μFI: μFI_year > μFI_{study period}). These summary statistics facilitate the comparison of different systems, time periods, and geographical regions by classifying the variation in conditions and distinguishing dynamic behavior indicative of regimes and transitions (González-Mejía et al. 2014, Eason et al. 2016, 2022, Sundstrom et al. 2017).

Once the indicators and indices were computed for each year, we evaluated the results based on the criteria established for each metric to determine whether they were able to capture patterns of change across the global coral reef seascape. Spearman rank order correlations were used to assess whether there were linkages between the spatiotemporal trends in the global coral reef seascape and the regional patterns observed within the coral reef SES for each waterbody.

RESULTS

To help provide a foundational understanding of the results, we begin by detailing our findings from the first year of the study, 2003 (Fig. A2 in Appendix 1). While bleaching was rare (except at a few Indian and Pacific Ocean stations) and ClimSST remained relatively high, other parameters varied across the global seascape, with a few peaking near the center of the plots (Fig. A2a–b in Appendix 1). There were similar patterns in temperature-related parameters, but temperature appeared to be largely decoupled from coral diversity and bleaching events.

Univariate indicators computed for each variable (Fig. 1a–b) showed that although both variance and skewness increased near the center of the plots for depth, average bleaching, and SSTA, there was little correspondence between the indicators for other variables (e.g., ClimSST, diversity). Linear plots of the multivariate indices captured changes in overall system dynamics by simultaneously tracking the behavior of all the variables over the seascape (Fig. 2). While VI peaked and dropped in clusters at varying intensities over the global coral reef seascape, FI displayed two stable regions (↑μFI, ↓σFI, ↓cvFI) separated by periods of decline, indicating transitional dynamics (shaded in red: ↓μFI, ↑σFI, ↑cvFI; Fig. 2). By combining the plots, we found that declining FI corresponded with the highest VI peaks approaching the center of the plots, but there were other regions where this complementary pattern could not be seen or was less pronounced. Using GIS plots, we visualized the results in a spatial context to show how the indices captured system dynamics from region to region (Fig. 3). Much like the linear plots, VI spiked highest at certain locations (primarily the Pacific), and FI captured more nuanced patterns over the global coral reef seascape (Fig. 3a–b). Given the criteria for interpreting the indices, summary statistics were used to classify more stable (↑μFI, ↓σFI, ↓cvFI, ↓μVI) and less stable (↓μFI, ↑σFI, ↑cvFI, ↑μVI) regions. These results indicate that the Pacific Ocean and Arabian Gulf coral reefs experienced the most and least variation, respectively, in 2003 (Fig. 3c–d).

Because the results from univariate indicators conflicted and did not provide clear insight into overall system behavior, we focused the remainder of the assessment on the multivariate approaches. In lieu of providing a detailed description of the results for each year, we placed the linear and GIS plots of the VI and FI results in Appendix 1 (Figs. A3–A12) and used summary statistics to capture the patterns visualized in the figures and to synthesize the assessment of spatiotemporal variation over the global coral reef seascape. By examining these statistics, we could classify the years in which the coral reefs around the globe experienced more or less stability (Fig. 4). The shaded areas (Fig. 4a–b) highlight the values that met the “stable” regimes criteria for the individual measures (↑μFI, ↓σFI, ↓cvFI, and↓μVI). Correspondence in the results indicated that 2003 and 2006–2009 were more stable years, whereas 2016 was the least stable (↓μFI and ↑σFI). This result largely converges with the years that met the criteria for cvFI (Fig. 4b–c). Given this congruence, we assessed the association between the summary statistics over time (data from Fig. 4b) by computing Spearman rank order correlations for each pair (e.g., μFI vs. cvFI) and found strong, statistically significant correlations (P < 0.05) between cvFI and the other parameters (cvFI vs.: μFI [ρ = −0.70], σFI [ρ = 0.81], μVI [ρ = 0.72]). This high level of correlation suggests that cvFI alone can be employed as an effective measure of spatiotemporal variation over the coral reef seascape. Viewing cvFI on a temporal scale provides insight on the stability of the global patterns and how they varied from year to year (Fig. 4c). Results show that cvFI rose initially before declining to its lowest point in 2007, after which it increased through the remainder of the study period, indicating declining dynamic order and increasing spatiotemporal change over the global coral reef seascape.

To capture spatiotemporal change at the regional scale, we computed cvFI values for each year from the sampling stations within each waterbody. These values were then visualized as a heatmap using a uniform scale to compare the cvFI values for each year and classify which waterbodies most frequently had the highest and lowest values (Fig. 5a). Results indicate that although high cvFI values persisted in the Pacific and Atlantic, the highest cvFI values were typically found in the Pacific Ocean coral reef SES (57% frequency), and the lowest cvFI was commonly in the Red Sea (43% frequency). Using linear plots of the cvFI values to illustrate how the trends changed over time (Fig. 5b), we found a general increasing trend in cvFI globally that was reflected regionally in the Pacific Ocean, Indian Ocean, and Arabian Gulf coral reef SESs, but not in the Atlantic and Red Sea (declined). Although cvFI values were typically lower (and in a declining trend) from 2003–2009 for all waterbodies, corresponding peaks were found in 2010 at the global scale and in the Pacific, Atlantic, and Indian Ocean coral reef SESs, after which the trends either increased (Global, Pacific, Indian, and Arabian Gulf) or decreased (Atlantic and Red Sea). The 2010 peaks in the larger water bodies preceded peaks found in the Arabian Gulf (2012) and Red Sea (2013; red circles in Fig. 5b). Spearman rank order correlations showed that cvFI for the global coral reef seascape was highly correlated with the cvFI trends in the Pacific Ocean (ρ = 0.89, P < 0.05) and Indian Ocean (ρ = 0.71, P < 0.05). There were no statistically significant correlations between cvFI of the global coral reef seascape and the other coral reef SES or between the coral reef SESs in distinct waterbodies.

DISCUSSION

Coral reefs span the globe and provide critical ecosystem services to human and natural systems, yet are under perpetual threat of degradation and, in some cases, demise (Dance 2019, Santavy et al. 2021). Given their essential roles in fisheries and for safeguarding vulnerable coastal communities around the globe, protecting coral reefs should be facilitated by initiatives that not only account for the ecosystem goods and services they provide, but also prioritize development of comprehensive monitoring and management strategies (e.g., multiscale adaptive management; Garmestani et al. 2023). Researchers have used a variety of approaches, from meta-analysis and semi-quantitative assessments (e.g., Vo et al. 2019, Ilosvay et al. 2020) to statistical modeling and Bayesian approaches (e.g., Carriger et al. 2019, Sully et al. 2019), to study coral reef ecosystems. However, the complex spatial and temporal dynamics inherent in coral reef SESs is a critical gap that needed to be addressed for a comprehensive assessment of their resilience at the global scale (Lawrence et al. 2021). Researchers have underscored the power of resilience-based approaches for analyzing coupled human and natural systems, emphasizing their efficacy in capturing latent patterns and trends in SESs (including regimes and regime shifts; Spanbauer et al. 2014, Eason et al. 2016, Sundstrom et al. 2017). Consequently, resilience-based approaches have significant potential for enhancing the understanding of spatiotemporal change in coral reef SESs.

Using data gathered from a global coral bleaching study (Sully et al. 2019), we tested the utility of resilience-based approaches for assessing spatiotemporal dynamics over the global coral reef seascape. Although univariate indicators tended to be less “noisy” versions of the raw data (Fig. 1; Fig. A1 in Appendix 1), helping to focus on general trends in the variables, they provided conflicting views (e.g., correspondence between variance and skewness for some variables but not others; Fig. 1). Moreover, the need to assess each variable separately posed challenges in determining how trends in separate variables related to the condition of the overall SES. Our results reinforce previous findings indicating that univariate indicators have limited utility when assessing complex systems, particularly when drivers are unknown, and highlight the utility of multivariate approaches (Eason et al. 2016, Sundstrom et al. 2017). VI and FI provided simultaneous evaluation of the variables by condensing their behavior into indices that tracked SES dynamics (Figs. 2 and 3; Figs. A3–A12 in Appendix 1). Our findings revealed complementary patterns in some cases (e.g., Fig. 2: largest FI declines coincide with highest VI values), but VI tended to spike at similar intensities and then abruptly drop, making VI signals less consistent and more challenging to interpret. Notably, the largest VI peaks consistently occurred in the same region (Pacific) throughout the study period and, when mapped, provided little insight into how the coral reef seascape evolved over time (Figs. A11–A12 in Appendix 1). Given that VI measures the dominant variance component (Brock and Carpenter 2006), this result is intuitive; however, it underscores a limitation of VI in capturing more refined patterns. This result aligns with previous research suggesting that although VI may be valuable within a suite of tools, it is difficult to decipher when used alone (Eason et al. 2016, Sundstrom et al. 2017).

In contrast, FI identified distinctive trends indicative of changing SES dynamics, providing a more robust analysis of spatiotemporal change. The approach helped to distinguish periods of relative stability (↑μFI, ↓σFI, ↓cvFI) from transitional dynamics (e.g., declining FI; ↓μFI, ↑σFI, ↑cvFI), which may warn of an impending shift (e.g., Fig. 2). Given the strong correspondence between cvFI and the other metrics, one of our key findings was that cvFI alone can serve as a useful proxy for assessing spatiotemporal variation at multiple scales. Figs. 4 and 5 synthesize the results mapped in Figs. A7–A10 (Appendix 1) and demonstrate the utility of the method for capturing spatiotemporal change at the global and regional scale. We found that while the global coral reef seascape became increasingly unstable over time, denoting degraded resilience, there was a period of relative stability from 2006–2009 (Fig. 4). This global trend was reflected regionally, with increasing instability also found in the Pacific Ocean, Indian Ocean, and Arabian Gulf (Fig. 5b). Additionally, we detected cvFI peaks in 2010 that partitioned the temporal trends and reflected distinctive patterns of variation. From 2003–2009, cvFI for all of the coral reef SESs were in declining trends, indicating increasing stability, which coincided with a period of relatively minor ENSO events (National Oceanic and Atmospheric Administration 2009, 2024). Afterwards, the cvFI for the global coral reef seascape (and most of the regional coral reef SESs) peaked in 2010 and continued to increase, indicating declining stabilty and increased variation in coral reef SES dynamics. This declining stability may be linked to more extreme ENSO events, including the 2014–2017 El Niño, the most damaging and widespread on record, which caused a coral bleaching event that affected > 70% of coral reefs globally (Eakin et al. 2018, 2019).

Our results also reveal that while the Red Sea experienced a longer period of increasing stability prior to a dramatic shift to more variable SES dynamics, the leading trends in the larger coral reef SESs (e.g., 2010 peaks in Pacific) may have served as a warning of an impending shift in the Red Sea and Arabian Gulf coral reef SESs. Further, we found that the dynamics of the global coral reef seascape were most strongly linked to the condition of the Pacific Ocean and Indian Ocean coral reef SESs. The lack of association between the dynamics of the global coral reef seascape and the condition of the other water bodies may be attributed to regional differences in coral reef conditions, stressors, and responses. The identification of changing spatiotemporal patterns detected at global and regional scales suggests shifting spatial regimes and degraded resilience in the coral reef seascape, which corresponds with the predicted increases in coral reef decline linked to accelerating environmental change (Dance 2019, Santavy et al. 2021).

Although nearly 40% of the coral reefs in our study experienced some level of bleaching, at an aggregated scale, the temperature variables were uncorrelated with diversity and bleaching. This result helps to highlight the fact that coral reef degradation is not simply related to high temperatures, as temperature variation and intensification through time are also important factors in coral reef decline. In line with this result, Sully et al. (2019) found that coral bleaching was more commonly associated not just with high sea surface temperature but also greater intensity and frequency of thermal stress anomalies. In addition, researchers have found common drivers of coral reef decline in systems that are not geographically linked (Gardner et al. 2003). A study of barrier coral reefs in Belize noted that although overfishing has typically been identified as a key cause of coral reef decline along the Central American coast, nutrient pollution driving the development of harmful algal blooms is also a driver of coral reef decline in the region (Lapointe et al. 2021). Similar to the Florida Keys, USA, researchers found that nitrogen-to-phosphorous (N:P) ratios in the Belize Barrier Reef lagoon have drastically increased in the last ≥ 30 years. The N:P ratio has doubled in Belize, resulting in only 17% of hard coral cover remaining, while the N:P ratio tripled in parts of the Florida Keys, where only 2% of hard coral cover remains (Lapointe et al. 2021). Accordingly, there is often a combination of spatiotemporal factors contributing to coral reef decline, further highlighting the need for multivariate approaches to assess coral reef resilience.

LIMITATIONS

Given that the primary goal of our study was to test resilience-based approaches for spatiotemporal analysis, we did not gather data on every variable that might impact coral reefs. Instead, we searched for a pre-existing coral reef study that used publicly available data with appropriate temporal and spatial coverage for our research. Although the data set selected (Sully et al. 2019) did not include all parameters researchers have identified, it contained many important variables (e.g., multiple temperature measures, bleaching, diversity). Previous coral reef studies have successfully used limited variables to study coral reef ecosystems. For example, Pisapia et al. (2019) assessed changes in temperature, coral cover, and coral composition to develop models to explore variation in bleaching susceptibility and mortality related to a 2016 mass bleaching event in the Maldives. Skirving et al. (2019) used daily coral heat stress and DHW data to examine the impact of heat stress on coral reefs from 1985–2017. Hence, although the objectives and methods employed in our study are different, the scale, scope, and variable coverage are in line with other, high-end coral reef studies. Further, because resilience-based approaches are not limited by the amount or type of data, the methods we used will remain effective as more spatiotemporal data sets become available.

CONCLUSION

We assessed resilience-based approaches for analyzing spatiotemporal change over the global coral reef seascape. We demonstrated the effectiveness of the approaches and found that FI yielded more distinctive results that not only facilitated the identification of SES patterns, but also allowed us to differentiate between relatively stable periods and transitional dynamics indicative of possible shifts to alternate regimes. Moreover, we found links between the global coral reef seascape and the dynamics of regional coral reef SESs, providing evidence of degraded resilience that coincided with climatological patterns. Our research contributes to the development of techniques and guidance for managing the resilience of SESs in the face of global change. Future work could involve focusing on specific geographical regions (e.g., Florida Keys, Great Barrier Reef) to elucidate further understanding of the causes and drivers of coral reef decline and to facilitate better connection between quantitative approaches, monitoring, and management strategies (Thrush et al. 2016).

LITERATURE CITED

Allen, C. R., D. G. Angeler, B. C. Chaffin, D. Twidwell, and A. Garmestani. 2019. Resilience reconciled. Nature Sustainability 2(10):898-900. https://doi.org/10.1038/s41893-019-0401-4

Angeler, D. G., H. B. Fried-Petersen, C. R. Allen, A. Garmestani, D. Twidwell, W.-C. Chuang, V. M. Donovan, T. Eason, C. P. Roberts, S. M. Sundstrom, and C. L. Wonkka. 2019. Adaptive capacity in ecosystems. Pages 1-24 in D. A. Bohan and A. J. Dumbrell, editors. Advances in ecological research. Volume 60. Academic Press, London, UK. https://doi.org/10.1016/bs.aecr.2019.02.001

Anthony, K. R. N., P. A. Marshall, A. Abdulla, R. Beeden, C. Bergh, R. Black, C. M. Eakin, E. T. Game, M. Gooch, N. A. J. Graham, A. Green, S. F. Heron, R. van Hooidonk, C. Knowland, S. Mangubhai, N. Marshall, J. A. Maynard, P. McGinnity, E. McLeod, P. J. Mumby, M. Nyström, D. Obura, J. Oliver, H. P. Possingham, R. L. Pressey, G. P. Rowlands, J. Tamelander, D. Wachenfeld, and S. Wear. 2015. Operationalizing resilience for adaptive coral reef management under global environmental change. Global Change Biology 21(1):48-61. https://doi.org/10.1111/gcb.12700

Brock, W. A., and S. R. Carpenter. 2006. Variance as a leading indicator of regime shift in ecosystem services. Ecology and Society 11(2):9. https://doi.org/10.5751/ES-01777-110209

Cabezas, H., and T. Eason. 2010. Fisher information and order. Pages 163-222 in M. T. Heberling and M. E. Hopton, editors. San Luis basin sustainability metrics project: a methodology for assessing regional sustainability. EPA/600/R-10/182. U.S. Environmental Protection Agency, Washington, D.C., USA. http://nepis.epa.gov/Exe/ZyPURL.cgi?Dockey=P100BSKA.txt

Carriger, J. F., S. H. Yee, and W. S. Fisher. 2019. An introduction to Bayesian networks as assessment and decision support tools for managing coral reef ecosystem services. Ocean and Coastal Management 177:188-199. https://doi.org/10.1016/j.ocecoaman.2019.05.008

Clements, C. F., and A. Ozgul. 2018. Indicators of transitions in biological systems. Ecology Letters 21(6):905-919. https://doi.org/10.1111/ele.12948

Contreras-Silva, A. I., A. Tilstra, V. Migani, A. Thiel, E. Pérez-Cervantes, N. Estrada-Saldívar, X. Elias-Ilosvay, C. Mott, L. Alvarez-Filip, and C. Wild. 2020. A meta-analysis to assess long-term spatiotemporal changes of benthic coral and macroalgae cover in the Mexican Caribbean. Scientific Reports 10(1):8897. https://doi.org/10.1038/s41598-020-65801-8

Dakos, V., M. Scheffer, E. H. van Nes, V. Brovkin, V. Petoukhov, and H. Held. 2008. Slowing down as an early warning signal for abrupt climate change. Proceedings of the National Academy of Sciences 105(38):14308-14312. https://doi.org/10.1073/pnas.0802430105

Dakos, V., E. H. van Nes, P. D’Odorico, and M. Scheffer. 2012. Robustness of variance and autocorrelation as indicators of critical slowing down. Ecology 93(2):264-271. https://doi.org/10.1890/11-0889.1

Dance, A. 2019. These corals could survive climate change – and help save the world’s reefs. Nature 575(7784):580-582. https://doi.org/10.1038/d41586-019-03629-7

Díaz-Pérez, L., F. A. Rodríguez-Zaragoza, M. Ortiz, A. L. Cupul-Magaña, J. D. Carriquiry, E. Ríos-Jara, A. P. Rodríguez-Troncoso, and M. C. García-Rivas. 2016. Coral reef health indices versus the biological, ecological and functional diversity of fish and coral assemblages in the Caribbean Sea. PLoS ONE 11(8):e0161812. https://doi.org/10.1371/journal.pone.0161812

Donovan, M. K., A. M. Friedlander, J. Lecky, J.-B. Jouffray, G. J. Williams, L. M. Wedding, L. B. Crowder, A. L. Erickson, N. A. Graham, J. M. Gove, C. V. Kappel, K. Karr, J. N. Kittinger, A. V. Norström, M. Nyström, K. L. L. Oleson, K. A. Stamoulis, C. White, I. D. Williams, and K. A. Selkoe. 2018. Combining fish and benthic communities into multiple regimes reveals complex reef dynamics. Scientific Reports 8:16943. https://doi.org/10.1038/s41598-018-35057-4

Eakin, C. M., G. Liu, A. M. Gomez, J. L. De la Couri, S. F. Heron, W. J. Skirving, E. F. Geiger, B. L. Marsh, K. V. Tirak, and A. E. Strong. 2018. Sidebar 3.1: unprecedented three years of global coral bleaching 2014–17 (in State of the climate in 2017). Bulletin of the American Meteorological Society 99(8):S74-S75. https://journals.ametsoc.org/supplemental//journals/bams/99/8/
2018bamsstateoftheclimate.1.xml/10.1175_2018BAMSStateoftheClimate.3.pdf

Eakin, C. M., H. P. A. Sweatman, and R. E. Brainard. 2019. The 2014–2017 global-scale coral bleaching event: insights and impacts. Coral Reefs 38(4):539-545. https://doi.org/10.1007/s00338-019-01844-2

Eason, T. 2024. Data used for “Assessing spatiotemporal change in coral reef social-ecological systems”. [Data set]. Zenodo. https://zenodo.org/doi/10.5281/zenodo.6394620

Eason, T., and H. Cabezas. 2012. Evaluating the sustainability of a regional system using Fisher information in the San Luis Basin, Colorado. Journal of Environmental Management 94(1):41-49. https://doi.org/10.1016/j.jenvman.2011.08.003

Eason, T., W.-C. Chuang, S. Sundstrom, and H. Cabezas. 2019. An information theory-based approach to assessing spatial patterns in complex systems. Entropy 21(2):182. https://doi.org/10.3390/e21020182

Eason, T., and A. S. Garmestani. 2012. Cross-scale dynamics of a regional urban system through time. Région et Développement 36. https://regionetdeveloppement.univ-tln.fr/wp-content/uploads/4_EasonGarmestani.pdf

Eason, T., A. Garmestani, and D. G. Angeler. 2022. Spatiotemporal variability in Swedish lake ecosystems. PLoS ONE 17(3):e0265571. https://doi.org/10.1371/journal.pone.0265571

Eason, T., A. S. Garmestani, and H. Cabezas. 2014. Managing for resilience: early detection of regime shifts in complex systems. Clean Technologies and Environmental Policy 16(4):773-783. https://doi.org/10.1007/s10098-013-0687-2

Eason, T., A. S. Garmestani, C. A. Stow, C. Rojo, M. Alvarez-Cobelas, and H. Cabezas. 2016. Managing for resilience: an information theory-based approach to assessing ecosystems. Journal of Applied Ecology 53(3):656-665. https://doi.org/10.1111/1365-2664.12597

Fisher, R. A. 1922. On the mathematical foundations of theoretical statistics. Philosophical Transactions of the Royal Society A 222(594-604):309-368. https://doi.org/10.1098/rsta.1922.0009

Gardner, T. A., I. M. Côté, J. A. Gill, A. Grant, and A. R. Watkinson. 2003. Long-term region-wide declines in Caribbean corals. Science 301(5635):958-960. https://doi.org/10.1126/science.1086050

Garmestani, A., C. R. Allen, D. G. Angeler, L. Gunderson, and J. B. Ruhl. 2023. Multiscale adaptive management of social-ecological systems. Bioscience 73(11):800-807. https://doi.org/10.1093/biosci/biad096

Gibbs, D. A., and J. M. West. 2019. Resilience assessment of Puerto Rico’s coral reefs to inform reef management. PLoS ONE 14(11):e0224360. https://doi.org/10.1371/journal.pone.0224360

Goatley, C. H. R., R. M. Bonaldo, R. J. Fox, and D. R. Bellwood. 2016. Sediments and herbivory as sensitive indicators of coral reef degradation. Ecology and Society 21(1):29. https://doi.org/10.5751/ES-08334-210129

González-Mejía, A. M. 2011. Fisher information - sustainability analysis of several US metropolitan statistical areas. Dissertation. University of Cincinnati, Cincinnati, Ohio, USA. http://rave.ohiolink.edu/etdc/view?acc_num=ucin1318610140

González-Mejía, A. M., T. N. Eason, H. Cabezas, and M. T. Suidan. 2014. Social and economic sustainability of urban systems: comparative analysis of metropolitan statistical areas in Ohio, USA. Sustainability Science 9(2):217-228. https://doi.org/10.1007/s11625-013-0227-3

González-Mejía, A., L. Vance, T. Eason, and H. Cabezas. 2015. Recent developments in the application of Fisher information to sustainable environmental management. Pages 25-72 in J. J. Klemeš, editor. Assessing and measuring environmental impact and sustainability. Butterworth-Heinemann, Oxford, UK. https://doi.org/10.1016/B978-0-12-799968-5.00002-6

Graham, N. A. J., S. Jennings, M. A. MacNeil, D. Mouillot, and S. K. Wilson. 2015. Predicting climate-driven regime shifts versus rebound potential in coral reefs. Nature 518(7537):94-97. https://doi.org/10.1038/nature14140

Graham, N. A. J., K. L. Nash, and J. T. Kool. 2011. Coral reef recovery dynamics in a changing world. Coral Reefs 30(2):283-294. https://doi.org/10.1007/s00338-010-0717-z

Holling, C. S. 1973. Resilience and stability of ecological systems. Annual Review of Ecology and Systematics 4:1-23. https://doi.org/10.1146/annurev.es.04.110173.000245

Houk, P., A. Yalon, S. Maxin, C. Starsinic, A. McInnis, M. Gouezo, Y. Golbuu, and R. van Woesik. 2020. Predicting coral-reef futures from El Niño and Pacific Decadal Oscillation events. Scientific Reports 10:7735. https://doi.org/10.1038/s41598-020-64411-8

Ilosvay, X. E. E., A. I. Contreras-Silva, L. Alvarez-Filip, and C. Wild. 2020. Coral reef recovery in the Mexican Caribbean after 2005 mass coral mortality—potential drivers. Diversity 12(9):338. https://doi.org/10.3390/d12090338

Jouffray, J.-B., L. M. Wedding, A. V. Norström, M. K. Donovan, G. J. Williams, L. B. Crowder, A. L. Erickson, A. M. Friedlander, N. A. J. Graham, J. M. Gove, C. V. Kappel, J. N. Kittinger, J. Lecky, K. L. L. Oleson, K. A. Selkoe, C. White, I. D. Williams, and M. Nyström. 2019. Parsing human and biophysical drivers of coral reef regimes. Proceedings of the Royal Society B 286(1896):20182544. https://doi.org/10.1098/rspb.2018.2544

Jozaei, J., W.-C. Chuang, C. R. Allen, and A. Garmestani. 2022. Social vulnerability, social-ecological resilience and coastal governance. Global Sustainability 5:e12. https://doi.org/10.1017/sus.2022.10

Karunanithi, A. T., H. Cabezas, B. R. Frieden, and C. W. Pawlowski. 2008. Detection and assessment of ecosystem regime shifts from Fisher information. Ecology and Society 13(1):22. https://doi.org/10.5751/ES-02318-130122

Lapointe, B. E., R. A. Brewton, L. W. Herren, J. W. Porter, and C. Hu. 2019. Nitrogen enrichment, altered stoichiometry, and coral reef decline at Looe Key, Florida Keys, USA: a 3-decade study. Marine Biology 166:108. https://doi.org/10.1007/s00227-019-3538-9

Lapointe, B. E., A. Tewfik, and M. Phillips. 2021. Macroalgae reveal nitrogen enrichment and elevated N:P ratios on the Belize Barrier Reef. Marine Pollution Bulletin 171:112686. https://doi.org/10.1016/j.marpolbul.2021.112686

Lawrence, A., A. Heenan, A. Levine, N. R. Haddaway, F. Powell, L. Wedding, R. Roche, P. Lawrence, C. Szostek, H. Ford, L. Southworth, S. S. Pilly, L. E. Richardson, and G. J. Williams. 2021. Spatial and temporal scales of coral reef fish ecological research and management: a systematic map protocol. Environmental Evidence 10:3. https://doi.org/10.1186/s13750-021-00217-z

Marshall, P., H. Schuttenberg, and J. West. 2006. A reef manager’s guide to coral bleaching. Great Barrier Reef Marine Park Authority, Townsville, Australia. https://repository.library.noaa.gov/view/noaa/11358/noaa_11358_DS1.pdf

Mayer, A. L., C. W. Pawlowski, and H. Cabezas. 2006. Fisher information and dynamic regime changes in ecological systems. Ecological Modelling 195(1-2):72-82. https://doi.org/10.1016/j.ecolmodel.2005.11.011

Mayer, A. L., C. Pawlowski, B. D. Fath, and H. Cabezas. 2007. Applications of Fisher information to the management of sustainable environmental systems. Pages 217-244 in B. R. Frieden and R. A. Gatenby, editors. Exploratory data analysis using Fisher information. Springer, London, UK. https://doi.org/10.1007/978-1-84628-777-0_7

McClanahan, T. R., S. D. Donner, J. A. Maynard, M. A. MacNeil, N. A. J. Graham, J. Maina, A. C. Baker, J. B. Alemu I, M. Beger, S. J. Campbell, E. S. Darling, C. M. Eakin, S. F. Heron, S. D. Jupiter, C. J. Lundquist, E. McLeod, P. J. Mumby, M. J. Paddack, E. R. Selig, and R. van Woesik. 2012. Prioritizing key resilience indicators to support coral reef management in a changing climate. PLoS ONE 7(8):e42884. https://doi.org/10.1371/journal.pone.0042884

National Oceanic and Atmospheric Administration. 2009. Climate variability: oceanic Niño index. National Oceanic and Atmospheric Administration, Washington, D.C., USA. https://www.climate.gov/news-features/understanding-climate/climate-variability-oceanic-nino-index

National Oceanic and Atmospheric Administration. 2024. Multivariate ENSO index version 2 (MEI.v2). National Oceanic and Atmospheric Administration, Washington, D.C., USA. https://psl.noaa.gov/enso/mei/

Newnham, T. J., N. K. Browne, J. Bumbak, L. Loudon, H. Wellington, G. Shedrawi, J. Hacker, and M. O’Leary. 2020. Long-term (70-year) monitoring of reef structure through high-resolution multidecadal aerial imagery. Coral Reefs 39(6):1859-1870. https://doi.org/10.1007/s00338-020-02012-7

Pisapia, C., D. Burn, and M. S. Pratchett. 2019. Changes in the population and community structure of corals during recent disturbances (February 2016–October 2017) on Maldivian coral reefs. Scientific Reports 9:8402. https://doi.org/10.1038/s41598-019-44809-9

Pittman, S. J., K. L. Yates, P. J. Bouchet, D. Alvarez-Berastegui, S. Andréfouët, S. S. Bell, C. Berkström, C. Boström, C. J. Brown, R. M. Connolly, R. Devillers, D. Eggleston, B. L. Gilby, M. Gullström, B. S. Halpern, M. Hidalgo, D. Holstein, K. Hovel, F. Huettmann, E. L. Jackson, W. R. James, J. B. Kellner, C. Y. Kot, V. Lecours, C. Lepczyk, I. Nagelkerken, J. Nelson, A. D. Olds, R. O. Santos, K. L. Scales, D. C. Schneider, H. T. Schilling, C. Simenstad, I. M. Suthers, E. A. Treml, L. M. Wedding, P. Yates, and M. Young. 2021. Seascape ecology: identifying research priorities for an emerging ocean sustainability science. Marine Ecology Progress Series 663:1-29. https://doi.org/10.3354/meps13661

Roelfsema, C. M., E. M. Kovacs, J. C. Ortiz, D. P. Callaghan, K. Hock, M. Mongin, K. Johansen, P. J. Mumby, M. Wettle, M. Ronan, P. Lundgren, E. V. Kennedy, and S. R. Phinn. 2020. Habitat maps to enhance monitoring and management of the Great Barrier Reef. Coral Reefs 39(4):1039-1054. https://doi.org/10.1007/s00338-020-01929-3

Romero-Torres, M., A. Acosta, A. M. Palacio-Castro, E. A. Treml, F. A. Zapata, D. A. Paz-García, and J. W. Porter. 2020. Coral reef resilience to thermal stress in the Eastern Tropical Pacific. Global Change Biology 26(7):3880-3890. https://doi.org/10.1111/gcb.15126

Santavy, D. L., C. L. Horstmann, L. M. Sharpe, S. H. Yee, and P. Ringold. 2021. What is it about coral reefs? Translation of ecosystem goods and services relevant to people and their well-being. Ecosphere 12(8):e03639. https://doi.org/10.1002/ecs2.3639

Scheffer, M., J. Bascompte, W. A. Brock, V. Brovkin, S. R. Carpenter, V. Dakos, H. Held, E. H. van Nes, M. Rietkerk, and G. Sugihara. 2009. Early-warning signals for critical transitions. Nature 461(7260):53-59. https://doi.org/10.1038/nature08227

Scheffer, M., S. R. Carpenter, V. Dakos, and E. H. van Nes. 2015. Generic indicators of ecological resilience: inferring the chance of a critical transition. Annual Review of Ecology, Evolution, and Systematics 46:145-167. https://doi.org/10.1146/annurev-ecolsys-112414-054242

Schumacher, B. D., B. Vargas-Ángel, and S. F. Heron. 2018. Identifying coral reef resilience potential in Tutuila, American Samoa based on NOAA coral reef monitoring data. NMFS-PIFSC-SP-18-03. National Oceanic and Atmospheric Administration, Washington, D.C., USA. https://doi.org/10.7289/V5/SP-PIFSC-18-003

Skirving, W. J., S. F. Heron, B. L. Marsh, G. Liu, J. L. De La Cour, E. F. Geiger, and C. M. Eakin. 2019. The relentless march of mass coral bleaching: a global perspective of changing heat stress. Coral Reefs 38(4):547-557. https://doi.org/10.1007/s00338-019-01799-4

Spanbauer, T. L., C. R. Allen, D. G. Angeler, T. Eason, S. C. Fritz, A. S. Garmestani, K. L. Nash, and J. R. Stone. 2014. Prolonged instability prior to a regime shift. PLoS ONE 9(10):e108936. https://doi.org/10.1371/journal.pone.0108936

Steneck, R. S., S. N. Arnold, R. Boenish, R. de León, P. J. Mumby, D. B. Rasher, and M. W. Wilson. 2019. Managing recovery resilience in coral reefs against climate-induced bleaching and hurricanes: a 15 year case study from Bonaire, Dutch Caribbean. Frontiers in Marine Science 6:265. https://doi.org/10.3389/fmars.2019.00265

Sully, S., D. E. Burkepile, M. K. Donovan, G. Hodgson, and R. van Woesik. 2019. A global analysis of coral bleaching over the past two decades. Nature Communications 10:1264. https://doi.org/10.1038/s41467-019-09238-2

Sundstrom, S. M., T. Eason, R. J. Nelson, D. G. Angeler, C. Barichievy, A. S. Garmestani, N. A. J. Graham, D. Granholm, L. Gunderson, M. Knutson, K. L. Nash, T. Spanbauer, C. A. Stow, and C. R. Allen. 2017. Detecting spatial regimes in ecosystems. Ecology Letters 20(1):19-32. https://doi.org/10.1111/ele.12709

Thrush, S. F., N. Lewis, R. Le Heron, K. T. Fisher, C. J. Lundquist, and J. Hewitt. 2016. Addressing surprise and uncertain futures in marine science, marine governance, and society. Ecology and Society 21(2):44. https://doi.org/10.5751/ES-08574-210244

Toth, L. T., C. D. Storlazzi, I. B. Kuffner, E. Quataert, J. Reyns, R. McCall, A. Stathakopoulos, Z. Hillis-Starr, N. H. Holloway, K. A. Ewen, C. G. Pollock, T. Code, and R. B. Aronson. 2023. The potential for coral reef restoration to mitigate coastal flooding as sea levels rise. Nature Communications 14:2313. https://doi.org/10.1038/s41467-023-37858-2

Vo, S. T., T. T. Hua, and K. H. Phan. 2019. A study of coral reef resilience and implications of adaptive management and rehabilitation in Khanh Hoa Province, Vietnam. Acta Oceanologica Sinica 38:112-117. https://doi.org/10.1007/s13131-019-1377-7

Wedding, L. M., J. Lecky, J. M. Gove, H. R. Walecka, M. K. Donovan, G. J. Williams, J.-B. Jouffray, L. B. Crowder, A. Erickson, K. Falinski, A. M. Friedlander, C. V. Kappel, J. N. Kittinger, K. McCoy, A. Norström, M. Nyström, K. L. L. Oleson, K. A. Stamoulis, C. White, and K. A. Selkoe. 2018. Advancing the integration of spatial data to map human and natural drivers on coral reefs. PLoS ONE 13(3):e0189792. https://doi.org/10.1371/journal.pone.0189792