Advancing the integration of spatial data to map human and natural drivers on coral reefs

A major challenge for coral reef conservation and management is understanding how a wide range of interacting human and natural drivers cumulatively impact and shape these ecosystems. Despite the importance of understanding these interactions, a methodological framework to synthesize spatially explicit data of such drivers is lacking. To fill this gap, we established a transferable data synthesis methodology to integrate spatial data on environmental and anthropogenic drivers of coral reefs, and applied this methodology to a case study location–the Main Hawaiian Islands (MHI). Environmental drivers were derived from time series (2002–2013) of climatological ranges and anomalies of remotely sensed sea surface temperature, chlorophyll-a, irradiance, and wave power. Anthropogenic drivers were characterized using empirically derived and modeled datasets of spatial fisheries catch, sedimentation, nutrient input, new development, habitat modification, and invasive species. Within our case study system, resulting driver maps showed high spatial heterogeneity across the MHI, with anthropogenic drivers generally greatest and most widespread on O‘ahu, where 70% of the state’s population resides, while sedimentation and nutrients were dominant in less populated islands. Together, the spatial integration of environmental and anthropogenic driver data described here provides a first-ever synthetic approach to visualize how the drivers of coral reef state vary in space and demonstrates a methodological framework for implementation of this approach in other regions of the world. By quantifying and synthesizing spatial drivers of change on coral reefs, we provide an avenue for further research to understand how drivers determine reef diversity and resilience, which can ultimately inform policies to protect coral reefs.

Annual catch values were joined to reporting blocks using the Area ID. Using the Polygon to 23 Raster conversion tool, average annual commercial catch data was converted from polygon to 24 raster with 100 m resolution for each gear type. Map pixels within marine protected areas that 25 are full no-take, explicitly prohibit commercial fishing, or prohibit specific gear groupings, were 26 set to zero respectively for each gear layer. Catch in Defacto MPAs and other areas with 27 restricted access were reduced according to expert input and local knowledge. Next each map  Shore-based fishing 39 To quantify boat-based fishing at a within-island spatial resolution we combined MRIP estimates 40 with two measures of shoreline accessibility (steepness and presence of roads), and gear specific 41 footprints.
Slope of the shoreline was calculated in degrees using the USGS 10 m Digital Elevation Model 43 (DEM). Then focal statistics was used to calculate the average slope within a 100 m radius of 44 each pixel. Next the string of raster cells that fall along the coastline were extracted and 45 reclassified into 3 categories (Table C). Exploration of the data showed that 0-3 degrees average 46 slope characterize coastline that would be easily accessible to anyone, 3-20 degrees includes areas that are more difficult but possible to access and fish from, and average slope greater than 48 20 degrees is very rugged coastline that is not possible to access and high cliffs that are too tall to 49 fish off of ( Fig A1).

50
Next, the coastal raster cells with associated slope information were converted from Raster to 51 Point data and a Near Analysis was used to calculate the distance within 1 km to the nearest 52 roads of various type. Topologically Integrated Geographic Encoding and Referencing (TIGER) 53 road data from the US Census Bureau were used based on completeness compared to other road 54 data available from the Hawaii Statewide GIS Program. TIGER roads are classified into 14 types 55 of roads, which we further grouped into 3 classes: 1) paved public roads, 2) 4WD roads, and 3) 56 private roads and foot trails (Private roads and foot trails were grouped into a single class 57 because both are relatively rare across the dataset and both are believed to represent a much 58 lower level of accessibility than public and 4WD roads). Coastal points were classified by 59 presence and type of roads within distances of 100 m, 200 m, 500 m, and 1 km. Type of roads 60 present were determined using a priority ranking based on ease of accessibility with paved public 61 roads ranking highest, 4WD roads next, and private roads and foot trails ranked as lowest (Table   62   D). For example, if all road types are present within 500 m of a point, that point would be 63 assigned "paved public road" because that ranks highest in ease of accessibility, regardless of 64 which road type was closest to the shore. Final map layers were created using the 500 m cutoff, as this layer provided the most precise information without compromising our ability to make 66 conclusions about this proxy for coastal access ( Fig A2).

67
Attributes for slope and road accessibility were then combined into a single accessibility criteria.

68
A weighting scheme was created that assumes easily accessible shorelines with flat slopes and 69 paved public road access have the highest catch, and therefore the highest weight, and that catch 70 and weight decreases incrementally with level of accessibility (Table E). Any combination that 71 includes no accessibility due to steep slopes received a zero weight (and therefore zero fishing). buffer. These population values were then used to assign weights to each boating facility in order to allocate a proportion of total island catch estimates to each boat harbor or ramp (more 110 described below). These weights sum to 1 for each island.

111
Next, boating facility cost allocation footprints were created by calculating distance to boating 112 facility using the Cost Allocation tool iteratively for each boat harbor/ramp, with a maximum 113 distance of 80 km. This allows for fishing influence from one harbor to overlap with nearby ones 114 as well as with neighboring islands. A cost surface was created by converting island polygons to 115 a 100 m raster with land pixels assigned a value of 1,000,000, and ocean pixels a value of 1.

116
During rasterization, priority was set in the Polygon to Raster tool for ocean areas -this ensures 117 that boating facility points do not fall on land. The cost distance surface output shows the 118 distance from the nearest ramp/harbor to a given pixel without traveling over land. The resulting 119 raster was then clipped to the footprint of inshore commercial reporting blocks.

120
In order to allocate catch proportionally to each boat harbor/ramp, estimated annual catch at the Where is the relative catch at a given map pixel, is the island-scale catch estimate, and is