Establishing historical benthic cover levels for coral reefs of the Western Indian Ocean

Data on coral reef health prior to large-scale disturbances are unavailable in most parts of the world including the Western Indian Ocean (WIO). Robust coral reef health baselines could improve the understanding of changes occurring to reefs in the 21st century and prevent the “shifting baseline” phenomenon, enabling researchers and managers to evaluate the success of management measures, and set achievable targets for new interventions. To make this data accessible to the WIO coral reef community, a literature review was conducted to identify and compile data collected prior to 2008 for two principal measures of reef health; hard coral and fleshy algae cover. Baseline hard coral and algae cover levels were calculated using data from selected sites that were known to be in healthy condition prior to (or just after) the 1998 bleaching event. Mayotte had the highest mean hard coral cover with 80.9 % (95 % bootstrap confidence interval (95 % CI) =65.8-95.9 %), followed by Comoros with 62.1 % (95 % CI=53.2-78.8 %) and Mad-agascar with 55.6 % (95 % CI=49.8-62.5 %). Mean fleshy algae cover varied from 8.4 % in Mayotte (95 % CI=2.4-17.4 %) to 35.4 % in Mozambique (95 % CI=20.6-50.8 %). At a regional scale, mean baseline hard coral cover is estimated to be between 41 and 47 %; reefs were in a coral-dominant state, with more than double the amount of coral compared to algae.


Introduction
Data on coral reef health prior to large-scale degradation (circa 1970s/pre-industrial era) are unavailable in most parts of the world 1970, before large-scale disturbance and degradation.
These baselines can support multiple research and management applications. Baseline levels can be compared with current conditions to improve the understanding of status and trends in reef health, and evaluate the success of management measures (Bruno et al., 2014). Reliable baselines can also help set ambitious and achievable conservation targets that are within natural ranges, thereby solving issues associated with shifting-baselines ).
Recent regional and global reef status reports Gudka et al., 2018;Souter et al., 2021) and vulnerability assessments  framed their analysis around two key benthic indicators of reef health; hard coral cover and fleshy algae cover.
These variables have been recognised as Essential Ocean Variables (EOVs) because of their importance as standard measures of coral reef functioning, and their extensive historical monitoring records (Miloslavich et al., 2018;Obura et al., 2019). Hard corals construct the reef framework, but sensitivity to marine heat waves, pollution and sedimentation threatens these reef engineers. Fleshy algae (which may include turf, macro and calcareous algae) are an increasingly important taxonomic group to monitor as the main competitor to hard corals for space, particularly following a disturbance event (Nyström et al., 2008;Jouffray et al., 2015;Mora et al., 2016;Brown et al., 2017).
Long term coral reef monitoring in the WIO began in the 1980s and early 1990s. A significant amount of data from this period were either not digitised, remain unpublished, are archived on personal and institutional databases (principally Non-Government Organizations), or are scattered across the grey literature. Efforts to secure this data in coral reef databases, e.g. Coral Reef Monitoring Database (CoReMo) and the Coral Reef Information System (CRIS), unintentionally led to significant losses in access to historical (early) data in the region as these databases became non-operational (Obura, 2013). These factors have led to very few coral reef datasets from the WIO being freely available and accessible, reducing the utility of the data. Renewed efforts to compile regional data and contribute it into regional and global outputs began around 2015 Gudka et al., 2018;Souter et al., 2021;Obura et al., 2021), nevertheless, large data gaps still exist, particularly prior to 1998 (as illustrated in . To make historical data accessible to the wider WIO coral reef research and management community, a literature review was conducted to identify, compile and consolidate available (published) data from sites across the WIO for two key indicators; hard coral and fleshy algae cover. The aim of this exercise is to establish a baseline that reflects the state of reefs around 1970, before widespread degradation occurred in the WIO region.
Because of the lack of monitoring data from that time, data collected from reefs known to be healthy prior to 2000 were used to approximate this. The main objectives of this paper are to: a) compile site-level historical hard coral and algae cover data in the WIO into a single dataset; b) estimate pre-disturbance baseline levels of hard coral and fleshy algae cover for countries and ecoregions in the WIO; and c) quantify the magnitude of coral mortality in the WIO due to coral bleaching in 1998 using data from reefs monitored before, during and soon after the event.

Literature review
A systematic literature review was conducted to locate and extract benthic cover data (percent cover) collected through quantitative coral reef surveys (e.g., Line or Point-Intercept-Transects, photo or visual quadrats) and visual estimates. Data were invariably reported as summarised mean cover values. Sources included grey literature (technical reports, books, and book chapters), scientific journal papers (articles and reviews) and project reports (Fig. S1).
Particular attention was given to extracting live hard coral cover and macro and turf algae cover data due to their importance as principal measures of coral reef health, with hard corals recognised as keystone reef builders and fleshy algae as their main competitors for space. These Essential Ocean Variables (EOVs) are used in national, regional and global biodiversity reporting on coral reefs (Bruno et al., 2014;Jouffray et al., 2015;Mora et al., 2016;Gudka et al., 2018;Miloslavich et al., 2018;Bang et al., 2021). Benthic cover data for the following benthic groups was also extracted: dead coral/recently dead coral (publications recorded these differently); calcareous algae (mainly Halimeda) and bleached coral; as well as combinations of these categories where reported and relevant (e.g., rock + algae and dead coral + algae).
Most effort was focused on locating data from as early as possible, ideally prior to 1998. However, since significant data gaps still exist for the early 2000s, data was also compiled from surveys conducted up until 2008. The primary focus was to extract site level data, but data aggregated at broader geographic scales was also included (e.g., "northern Kenya"), or other classes (e.g., "unspecified 9 sites (protected)"). In such cases, the number of sites included was noted. Where available, site coordinates were also included in the compiled dataset. Publications were reviewed, and data were obtained from tables or in-line text in the main sections as well as supplementary materials and appendices. During the search process, if the data provided by the document was from secondary sources, the source reference with the primary data was located and used instead. A total of over 70 documents were reviewed, from which 58 documents were found to provide relevant data for this study (Table S3, Supplementary Information), with the other publications either having no useful data or data was provided in a way which could not be extracted efficiently, such as in graphs or figures (e.g. van Katwijk et al., 1993;Ballesteros and Afonso-Carrillo, 1995;Johnstone et al., 1998;Muthiga et al., 1998;Wilkinson et al., 1999;McClanahan, 1999

Establishing baseline cover values
Baseline (pre-disturbance) hard coral and algae cover levels were calculated for each country/territory as well as for 10 ecoregions conceptualised by Obura et al., • only pre-1998 data were used, unless sites sampled shortly after the 1998 bleaching event (1998)(1999)(2000) had recent dead coral (and/or bleached) cover data collected and there was confidence that the monitoring was conducted using verified quantitative survey methods, and by experienced surveyors able to accurately distinguish and classify these categories at the necessary resolution.
Adding the recent dead coral cover to the living cover values enabled estimation of the pre-1998 hard coral cover level.
• care was taken to ensure that where possible there was equal distribution of sites within a country or ecoregion (herein referred to as a geo-unit) and each site was only represented once in the calculation of the baseline.
Using the filtered dataset, the mean hard coral and algae baseline cover was calculated by averaging across all the selected site mean values within an ecoregion or country/territory (geo-unit). Baseline and had low sample size, therefore data variability is reported using standard deviation, coefficient of variation (CV, ratio of standard deviation to the mean) and inter-quartile range (Rowland et al., 2021). As an inferential statistic, bootstrap resampling of the indicator mean was performed to provide a 95 % confidence interval using 10,000 iterations (Rowland et al., 2021). The regional mean range (95 % CI) for the WIO was calculated using the same bootstrap resampling method across all the data.
There was less algae data available than hard coral cover, with 162 sites for corals and 101 for algae. All countries/territories and ecoregions (except Madagascar South) had data for hard coral cover calculations, with some of the ecoregions only having a single data point (Madagascar East, and Madagascar North).
There was no algae data for South Africa. Where site data was presented as a range (Hardman, 1999), the median value was used for calculations.
For sites with both hard coral and algae data, the algae-coral ratio (ACR) was calculated as ACR = Algae cover/ (Algae + Coral cover) (Bajjouk et al., 2019), and selected site values were averaged to calculate baseline ACR values per geo-unit. This metric provides a useful index for describing the relationship between the competing corals and algae. Using this formula, the values are bound between zero and one, simplifying its interpretation. For this analysis, the percent cover of fleshy algae was a combination of any or all three erect algae types reported in a study (turf, macro and calcareous algae (not including crustose coralline algae)). This follows the regional practice in Gudka et al., 2018;Obura et al., 2021)

Results
For baseline hard coral cover levels (Fig. 2 (Fig. 3). For the WIO, averaging across eco-regional means, mean fleshy algae cover was 18.9 %.  When considering the ratio of fleshy algae to hard coral cover, reefs across all geo-units were dominated by hard coral, though algae cover was above 50 % at some sites ( Fig. 4

Discussion
Reefs in the WIO have unequivocally been altered over the past 30 to 40 years (Ateweberhan et al., 2011;McClanahan et al., 2014). Increasing pollution, fishing, and temperatures, coupled with inadequate protection have set coral reefs on a path of diminishing returns to society (Cinner et al., 2009;Halpern et al., 2015;Samoilys et al., 2017). The data presented in this paper provides insights into what reef state was likely to have been for different parts of the WIO before various threats began causing widespread degradation.
Reef condition prior to the 1998 El Nino was non-uniform within and across eco-regions, with coral cover varying from below 10 % to above 95 %. Coral dominance was common, with fleshy algae cover less than half that of coral cover in all but 3 eco-regions, with particularly low algae levels in Mayotte, Seychelles, and Madagascar. Reefs in Mayotte had the highest  coral cover and the lowest algae cover, though this was from just four sites. When combined with four more sites from Comoros (Eco-region=Comoros), hard coral cover levels for all eight sites were above 50 %.
Hard coral cover was highest in the Northern Mozambique channel, reflecting the vibrancy and diversity of this region as the 2 nd most biodiverse coral region in the world (Obura, 2012). Baseline estimates use a selection of sites which are assumed to be representative of general reef condition around 1970. Sites were rigorously selected to mainly use data prior to 1998 when reef degradation was less widespread in the WIO, and only sites known or assumed to have had minimal degradation since 1970 were included. This selection of the 'best' sites is an example of space-for-time substitution similar to using pristine or uninhabited reefs as references, and has been used when long-term data is lacking (Sandin et al., 2008;Blois et al., 2013).
There are large differences in the amount of data Some caveats exist with the process described in this paper. Although a thorough search for literature containing suitable data was conducted, invariably these efforts were not exhaustive. There are other datasets which exist but remain unpublished, hidden or have only been presented as aggregations in the literature, minimising their suitability for these and other purposes. Baseline averages are not disaggregated by reef zone, as zonal information was not available for some sites, resulting in too few data points for some geo-units. This is consistent with the protocols of other recent regional analyses Gudka et al., 2018;Obura et al., 2021). An element of subjectivity with the method comes from determining the sites  (Khamala, 1971;McClanahan and Muthiga, 1988), which is clear from the low hard coral cover of between 1 and 7 %.
Data from very specific habitats which are not representative of the regional reef system were also excluded e.g., data collected in 1987 for offshore reefs in Lamu, Kenya by Samoilys (1988). This selection protocol was consistently followed, resulting in some low hard coral cover sites being included as these conditions were deemed to be normal for the site (or area) and not due to prior degradation or disturbance (e.g., Mike's Cupboard in Inhambane, Mozambique). Data were included as much as possible to increase the sample size.
Recent benthic cover values published in regional reports (  (Nicet et al., 2017)), but algae may be on an upward trajectory. Mauritius had a coral cover of ~50 % with corresponding algae levels of ~20 % in 2015 , though the data in this study is from an extensive rapid assessment of the entire Mauritius  (Wilkinson, 2000;Ateweberhan et al., 2011;, though the coral loss values vary because of differences in methodologies. Obura et al.  , the data can be incorporated into frameworks or other analyses that provide policy or management relevant results.
As reef condition continuously changes, there is heightened need for more, and better-quality data to be made available for management and research.
Scaled national investment in long-term, highresolution monitoring is recommended through regional entities like the Global Coral Reef Monitoring

Network (GCRMN) nodes and Nairobi Convention
Coral Reef Task Force. Project donors are also encouraged to enforce strict data publishing measures that follow the FAIR principles (Findable, Accessible, Interoperable, and Reusable) (Wilkinson et al., 2016), but researchers are also requested to voluntarily avail data particularly for data-deficient regions as well as other key taxonomic groups such as fish and urchins. Coupled, this will enable historical baseline conditions to be calculated at a higher precision across more ecoregions in the WIO, and using other variables, as well as to trace trends in reef health over time, particularly after acute disturbance events.
It is the belief of these authors that the prospects of conservation are greater with open data practices than without them.

Supplementary material
Provided in separate PDF file Table S1. Baseline (pre-disturbance) hard coral cover levels (%) for 10 countries/territories (top) and 11 eco-regions (bottom) in the Western Indian Ocean (WIO). Table S2. Baseline (pre-disturbance) fleshy algae cover levels (%) with comparative (same sites) hard coral cover levels (%) and algae-coral-ratios for 9 countries/territories (top) and 11 eco-regions (bottom) in the Western Indian Ocean. Table S3. List of all publications where data was extracted from. Table S4. Change in hard coral cover (%) during 1998 bleaching. Figure S1. Types of publication from which the data was extracted (n=58).