Modeling Larval Connectivity of Coral Reef Organisms in the Kenya-Tanzania Region

Most coral reef organisms have a bipartite life-cycle; they are site attached to reefs as adults but have pelagic larval stages that allow them to disperse to other reefs. Connectivity among coral reef patches is critical to the survival of local populations of reef organisms, and requires movement across gaps that are not suitable habitat for recruitment. Knowledge of population connectivity among individual reef habitats within a broader geographic region of coral reefs has been identified as key to developing efficient spatial management strategies to protect marine ecosystems. The study of larval connectivity of marine organisms is a complex multidisciplinary challenge that is difficult to address by direct observation alone. An approach that couples ocean circulation models with individual based models (IBMs) of larvae with different degrees of life-history complexity has been previously used to assess connectivity patterns in several coral reef regions (e.g., the Great Barrier Reef (GBR) and the Caribbean). We applied the IBM particle tracking approach to the Kenya-Tanzania region, which exhibits strong seasonality in the alongshore currents due to the influence of the monsoon. A 3-dimensional (3D) ocean circulation model with 2 km horizontal resolution was coupled to IBMs that track virtual larvae released from each of 661 reef habitats, associated with 15 distinct regions. Given that reefs provide homes to numerous species, each with distinctive, and in aggregate very diverse life-histories, several life-history scenarios were modeled to examine the variety of dispersal and connectivity patterns possible. We characterize virtual larvae of Acropora corals and Acanthurus surgeonfish, two coral reef inhabitants with greatly differing pelagic life-histories, to examine the effects of short (50 days) pelagic larval durations (PLD), differences in swimming abilities (implemented as reef perception distances), and active depth keeping in reef connectivity. Acropora virtual larvae were modeled as 3D passive particles with a precompetency period of 4 days, a total PLD of 12 days and a perception distance of 10 m. Acanthurus virtual larvae were characterized by 50 days precompetency period, a total PLD of 72 days and a perception distance of 4 km. Acanthurus virtual larvae were modeled in


1
Introduction 50 Tropical coral reef ecosystems are very important from both the ecological and economical points of 51 view (Spalding et al., 2001). However, they are also particularly fragile, and have been declining in 52 recent years in most regions of the world (Melbourne-Thomas et al., 2011; Hughes et al., 2003;53 Pandolfi et al., 2003), since they are highly susceptible to anthropogenic stressors operating at global 54 scales (e.g., global warming and ocean acidification) and local scales (e.g., pollution/eutrophication, 55 fishing, over-commercialization for recreation). Coral reef ecosystems are complex communities 56 with very high species diversity. Most reef species have bipartite life histories with a planktonic 57 larval stage and a benthos associated adult life. As adults, coral reef organisms exhibit various 58 degrees of site attachment ranging from completely sessile, like corals and sponges, to highly mobile, 59 like fish and crustaceans. Generally, even fish capable of swimming several kilometers in a few hours 60 P r o v i s i o n a l have restricted home ranges, since they are relatively territorial and are associated with specific reef 61 habitats that are patchily distributed (Sale, 2006). Most adult reef organisms are distributed in 62 metapopulations connected by pelagic larvae that disperse subject to the ocean currents ( Hamilton and Brakel, 1984). Western 72 Indian Ocean coral reef communities are characterized by high levels of species diversity and may be 73 centers of biodiversity (Spalding et al., 2001). Coastal communities of Kenya and Tanzania depend 74 on the reef for food. Since there is little regulation on the use of these resources through formal 75 resource management strategies, reef areas in Kenya and Tanzania have been degraded due to  76 overfishing, destructive fishing techniques, coastal pollution and other activities affecting the coastal 77 environment (Hamilton and Brakel, 1984;Spalding et al., 2001). Increasing interest in coral reef 78 tourism is simultaneously leading to increased pressure on some coral reefs while providing a 79 powerful local incentive for conservation (Spalding et al., 2001). There are 26 Marine Protected 80 Areas (MPAs) in Kenya  Population connectivity plays a fundamental role in local and metapopulation dynamics, community 95 dynamics and structure, genetic diversity, ecosystem responses to environmental changes, and the 96 resiliency of populations to human exploitation . Connectivity among marine 97 metapopulations is controlled by physical transport and dispersion, temperature, and biological 98 processes such as the timing of spawning, pelagic larval duration (PLD), larval behavior, and 99 mortality. The net combined effect of these processes determines the spatial scales over which a 100 population is connected (Gawarkiewicz et al., 2007). Connectivity is therefore a function of several 101 interacting variables including species, geographical area, and ocean conditions, and is highly 102 variable in both time and space (e.g. Cowen  (self-replenishing) (e.g. Schultz and Cowen, 1994). For decades the spatial connectivity of larval fish and invertebrates was thought to be a passive 148 process governed primarily by the ocean physics and the duration of the larval period (e.g. Shanks, 149 2009). The pelagic larval duration (PLD) of coral reef organisms varies greatly; from a few hours for 150 some coral species to a few months for some fish and crustaceans (Shanks, 2009  The relative lack of physiological and behavioral data for larvae of coral reef species in the  Tanzania region led us to examine connectivity among coral reefs using idealized particle tracking 197 experiments that simulate larvae with characteristics of two ubiquitous and ecologically important 198 species groups: the Acropora branching corals with short PLD (ca. 12 days, (Babcock and Heyward, showing the origin locations on one axis and destination locations on the other axis is used to 301 visualize the geographic connections among habitat patches for simple alongshore linear systems. 302 However, the two dimensional nature of the reef systems bordering East Africa, with multiple reefs at 303 the same latitude (e.g., mainland fringing reefs, atolls or patch reefs in the channels between the 304 islands and mainland, fringing reefs on the west and east coast of the islands), make the reef to reef 305 connectivity matrices organized by the latitude of the centroid of the reef polygons insufficiently 306 informative regarding inshore-offshore connections. Due to the spatial complexity of the reef habitat, 307 P r o v i s i o n a l we simplified the connectivity matrices by assigning individual reefs to one of fifteen geographic 308 subregions ( Figure 1). Geographic regions considered mainland continuity of reefs, but also national 309 borders and offshore island masses, many of which have both shoreward facing and offshore facing 310 fringing reefs (Figure 1). This allowed a more meaningful visualization of the results. Based on the 311 number of particles released within a region, the percentage of particles that successfully connect 312 from region to region was calculated. Summing the percentages in the horizontal direction (all 313 destination regions) on the connectivity matrices shows the percent of successful recruits from each 314 region of origin. 315

316
The term local retention refers to the ratio of virtual larvae settling at their released location and the 317 total number of virtual larvae released at that location, while self-recruitment is the ratio of virtual 318 larvae settling at their released location and the total number of larvae settling at that location. In the 319 results section the comparatives "weaker" and "stronger" are used to refer to the magnitude of 320 connections between two specific sites, indicating the proportion of particles connecting from one 321 reef or region to another. Strong connections appear as large color-coded circles in the connectivity 322 matrices, while weak connections are small black circles. Conversely "few" and "more/lots" are 323 used to refer to the number of sites that are connecting to a reef or region. The number of connections 324 for a region will be represented by the number of circles on each row or column for origin and 325 destination regions, respectively. 326

327
We use the terms "source" and "origin" interchangeably to refer to reefs or regions from which 328 virtual larvae are released. Similarly, we use the terms "sinks" and "destinations" interchangeably to 329 refer to reefs or regions into which virtual larvae successfully settle. We are not referring to 330 population source/sinks according to the classical population ecology definition, since we do not 331 consider spatially variable reproductive input nor variable mortality during the settlement phase. In 332 this case we are referring only to source/sinks of the planktonic pool of successful virtual larvae, and 333 therefore the terms only refer to the diversity of origins/destinations of the virtual larvae that are 334 assumed to successfully settle.

Region to Region Connectivity Matrices 386
The region to region connectivity matrices allow an easier visualization of the main connectivity 387 patterns, synthesizing the information of reef to reef connectivity matrices (available upon request). Three-dimensional passive larvae tend to stay near the surface and are therefore less likely to be 486 carried away from suitable habitat by the strong northward flowing EACC core transitioning from 487 deep to shallow waters during the second half of the spawning season. 488 489

Sensitivity to perception distance 525
The analysis with increased (reduced) perception distance for Acropora (Acanthurus) is presented to 526 provide insight on one of the processes responsible for the large difference in settlement success 527 between the modeled species groups, and to illustrate the variability that might be expected among 528 coral reef species with different life-history strategies. The percentage of settlement success increased 529 with greater perception distance for both the Acanthurus surgeonfish and the Acropora coral 530 simulations ( Figure 6). However, the increase in settlement success for the short PLD coral was 531 always much higher than that of the surgeonfish for the same increase in perception distance, with the 532 coral reaching 95% settlement success with a 4 km perception distance. The difference in settlement 533 success between the two genera was significant for all perception distance scenarios (Figure 6), 534 indicating that with the same perception distance short PLD virtual larvae will always be more The dominant pattern of connectivity for both Acanthurus and Acropora in the KT region is southern 568 reefs providing virtual larvae to northern reefs. The spatial scale of connectivity is much smaller for 569 the short PLD coral group; successful connections are restricted to a 1° radius (~100 km) around 570 source reefs. 8.2% of Acropora larvae that successfully settle, recruited to their source reef (self-), an 571 important proportion compared to only 1-2 % for Acanthurus. Some Acropora were capable of long 572 P r o v i s i o n a l distance dispersal, particularly larvae spawned at the reef offshore of Dar es Salaam peninsula. This 573 indicates that they can take advantage of the strong offshore EACC to reach distant northern reefs, 574 and that even for short PLD, latitudinal isolation may be minimal, especially at longer (i.e. 575 evolutionary) timescales. 576 577 In contrast to the generally short dispersal distances of Acropora, long distance connections from the 578 southern to the northern most reefs (~950 km) are common for virtual Acanthurus. Their longer 579 pelagic durations lead to greater transport distances and reduced local retention. Overall settlement 580 success was significantly greater in Acanthurus (24%) than in Acropora (<0.5%). This is due to 581 several factors that enhance Acanthurus successful settlement probabilities: longer competency 582 period, greater reef perception distance and swimming ability. Kenyan and Somali reefs are however the most common sink reefs, receiving larvae from many reefs 650 to their south ( Figure 11). These northern reefs may be more resilient to local threats. The source and 651 sink patterns reflect the strong, mostly unidirectional south to north flow along the coast. For 652 Acropora larvae, the source and sink maps are much more patchy (Figures 10 and 11, respectively), 653 reflecting the effects of the smaller dispersal scale of Acropora larvae. 654 655 A recent genetic study of Acropora tenuis connectivity in the KT region, reports high but variable 656 connectivity between sample sites, which cluster in 3 different groups: 1) Kenya and northern-657 Tanzania Pemba, and south Mafia. In the Acropora simulations these regions receive virtual larvae from less 685 than 10 different source reefs (Figure 11). The regional Acropora connectivity matrices (Figure 3b, 686 d) show that local retention is important for these regions. In our regional connectivity results, 687 however, only west Pemba, east Mafia and south Tanzania show relative isolation, receiving 688 Acropora larvae from only three and two other regions respectively; south Kenya in contrast receives 689 settlers from many regions further south. This discrepancy between our model results and Souter et 690 al.'s (2009) genetic study may reflect the different spatial scales considered, since our regional 691 grouping aggregates connections for several reefs that might have different degrees of isolation. Larvae in the ocean are subject to mixing at scales smaller than those represented in the ocean 791 circulation model. In particle tracking models these unresolved motions are often implemented as a 792 random walk scaled by the model diffusivity. Simulations that implement a random walk to mimic 793 diffusion are considered more realistic but computationally expensive. We conducted a few 794 sensitivity experiments that included 3-dimensionally variable vertical diffusion. Simulations that 795 included vertical diffusion (not shown) reproduced the main connectivity patterns produced by the 796 3D advective only experiments, but with smaller connectivities-mostly due to greater vertical 797 dispersion that subjected larvae to greater horizontal flow variation. These results are probably more 798 realistic for early or weakly swimming larvae (e.g., coral species) that are unable to maintain their 799 vertical position in the water column in the presence of vigorous vertical mixing. 800 801 While reef-to-reef connectivity is important in metapopulation ecology, regional connectivity is 802 expected to be more robust to the uncertainty introduced by the oceanographic and biological 803 assumptions made in these models. Region to region connectivity matrices synthesize the 804 information of reef to reef connectivity matrices, making it more manageable and easier to interpret. 805 The regional summary could assist managers, policy makers and the general public to understand the 806 interconnections among coral reef regions due to pelagic larval dispersion of their local populations. 807 Previous bio-physical connectivity studies highlight the importance of considering larval connectivity 808 at regional levels when trying to prioritize the implementation of management strategies for both 809 conservation and fisheries enhancement goals. One of the insights of examining connectivity at a 810 regional scale is that the importance of international connections becomes obvious, as has been 811 shown by Kough  This modeling study is a first approach to understanding the connectivity among coral reef 818 populations in a data poor region. The information provided, even though preliminary, presents a 819 general pattern of the potential regional connectivity and identifies particularly resilient and 820 vulnerable areas as well as the hydrodynamic features driving the connections. Spatial scales of 821 connectivity and settlement success rates are within the ranges reported by other bio-physical 822 modeling studies for similar genera in other coral reef regions Dorman et al., 823 2015). However, the robustness of the connectivity patterns presented needs to be further evaluated 824 by performing experiments for more years and longer spawning seasons, and carrying out more 825 extensive sensitivity analysis to the model assumptions. After gaining more confidence in the 826 modeled connectivity patterns, the information provided by this modeling study could be carefully 827 and critically evaluated, in order to be applied to optimize the effectiveness of marine protected area 828 management and other marine protection efforts. Further modeling experiments similar to those 829 presented here, but better informed by empirical data, and including the capability of larvae to 830 respond to the ocean conditions will provide greater detail on the complex biophysical interactions 831 that occur in the sea, and will provide a more realistic, and less uncertain, representation of 832 connectivity patterns. These results will aid in understanding how a range of species specific 833 individual responses influence the distribution and connectivity patterns and should enable more 834 specific guidelines for spatial management that provide better resource resiliency and protection 835 throughout the Kenya- Tanzania