NCCOS Assessment: An Aquaculture Opportunity Atlas for the Southern California Bight (NCEI Accession 0286986)

Submission Package ID: U3LNX6

Methods:
(1) Study Areas:
Study areas (North, Central North, Central South, South) in the Southern California Bight were needed due to the geographical breaks so Aquaculture Opportunity Area (AOA) options could be identified. Study areas were developed based on depth and jurisdictional boundaries.
Step 1.1:
Federal waters off Southern California, south of Point Conception to the U.S. and Mexico border, were selected as one of the first regions for AOA evaluation because of preexisting spatial data availability, previous analyses in the region, and industry interest in developing sustainable offshore aquaculture operations.
Step 1.2:
NOAA further narrowed the AOA site selection criteria in Southern California using a combination of spatial mapping approaches, scientific review, and stakeholder input. As described above, the Southern California AOA area of interest includes federal waters between 5.6 km (3.0 nautical miles [nm]) and 46.3 km (25.0 nm) offshore within the EEZ at depths ranging between 10 m (32.8 ft) and 150 m (492.1 ft).

(2) Suitability Models:
Planning and siting for marine aquaculture operations requires thorough synthesis and spatial analyses of critical environmental data and ocean space use conflicts (Kapetsky et al. 2013). This dataset is to aid with visualization of aquaculture opportunity areas (AOAs) suitability model in the Southern California Bight. A gridded relative suitability analysis, commonly used in a Multi-Criteria Decision Analysis (MCDA), was performed to identify the grid cells with the highest suitability for aquaculture development in the study areas (Longdill et al. 2008; Radiarta et al. 2008; Gimpel et al. 2015; Bwadi et al. 2019). Spatial data layers included in the suitability analysis identify space-use conflicts and environmental constraints such as active national security areas, maritime navigation, ocean industries, and natural resource management. We utilized a submodel structure to capture ocean use and conservation concerns including national security; natural and cultural resources; industry, navigation, and transportation; and aquaculture and fishing. Data layers with no compatibility with aquaculture development (e.g., shipping fairways) were captured in the list of incompatible constraints and removed from further analysis due to known incompatibility with aquaculture. This model structure ensures that each submodel is given equal representation in the overall suitability model regardless of how many data layers are present in each submodel. The geometric mean of the four submodels (i.e., national security; industry, navigation, and transportation; natural and cultural resources; fishing and aquaculture) was used to calculate an overall suitability score. The geometric mean #(Equation 1) was chosen because it grants equal importance to each variable (Bovee 1986; Longdill et al. 2008; Silva et al. 2011; Muñoz-Mas et al. 2012). Furthermore, all data layers and submodels had equal weight within the suitability model. Equation 1.
Step 2.1:
1. A suitability polygon with a 4.05 ha (10 ac) hexagonal grid was created using the extent of the created study area polygons (North, Central North, Central South, South; see Step 1.2), which was only in the US Federal waters of the Southern California Bight and at depths between 10 and 150 m.
Step 2.2:
Each data layer was scored on a 0 to 1 scale, with scores approaching 0 representing low suitability and 1 representing high suitability relative to the other grid cells for aquaculture. Any grid that contained a data layer with a 0 score (i.e., constraints data layer) was deemed unsuitable for aquaculture, and not considered further in the analysis. Next, an overall suitability score was calculated for each submodel (i.e., national security; industry, navigation, and transportation; natural and cultural resources; fishing and aquaculture) by taking the geometric mean of all scores within each grid cell. Scoring rationale for both categorical and continuous data can be seen in Appendix C of Riley et al. 2021. The geometric mean of the four submodels was used to calculate an overall suitability score.
Step 2.3:
Suitability scores are presented as categories (“Unsuitable,” “Low,” “Moderate,” “High”) grouped by quantiles of the calculated scores, with all scores of 0 being in the “Unsuitable” category and represented by the color red. Within the suitability submodel and overall model maps, standardized colors were used to depict categories, with orange representing “Low,” yellow “Moderate,” and blue “High” suitability and coinciding with each proportion of quantile values. With all suitability maps, relative categories still represent values ranging from 0 to 1, with the “Low” category representing the lowest quantile of the data, “Moderate” the middle quantile, and “High” the upper quantile. Presenting categories rather than actual suitability scores simplified interpretation of results and provided optimal contrast among categories. Distribution of scores varies among the suitability submodels (e.g., number of data layers, score range of data distribution depicted); for example, in one submodel a score of 0.5 could be classified as “High,” while in another submodel or region a score of 0.5 could be “Low” because the scores are relative. Thus, suitability scores among the different study areas and different submodels should not be compared, as the score is unique to each study area and submodel.

(3) High-High Clusters:
Step 3.1:
A Local Index of Spatial Association (LISA) analysis, which identifies statistically significant clusters and outliers, was performed on the final relative suitability modeling results (Anselin 1995). All grid cells with a score of 0 were not included in the cluster analysis, as these areas are unsuitable for aquaculture and are not considered further. The ArcGIS Pro Cluster and Outlier Analysis tool was used to implement the LISA analysis (Esri 2021a). The fixed-distance spatial conceptualization was utilized within this analysis as it allows the identification of localized clusters. The function inputs were a 250-m search distance and 9,999 iterations with row standardization and a false discovery rate correction applied to allow for more conservative results by estimating the number of false positives for a given confidence level, adjusting the critical p-value accordingly (Esri 2021b). Statistically significant clusters (p < 0.05) of the highest suitable scores (i.e., high-high clusters) were identified, with any clusters smaller than 202 ha (500 ac) excluded, as this was the minimum AOA target size. Anselin L. 1995. Local Indicators of Spatial Association—LISA. Geog Anal. 27(2):93–115. Esri. 2021a. ArcGIS Pro: Release 2.8.0. Redlands, CA: Environmental Systems Research Institute. Esri. 2021b. What is a z-score? What is a p-value? Esri ArcGIS Pro online. Available from: https://pro.arcgis.com/en/pro-app/latest/tool-reference/spatial-statistics/what-is-a-z-score-what-is-a-p-value.htm

(4) Refined High-High Clusters:
Step 4.1:
A Local Index of Spatial Association (LISA) analysis, which identifies statistically significant clusters and outliers, was performed on the final relative suitability modeling results (Anselin 1995). All grid cells with a score of 0 were not included in the cluster analysis, as these areas are unsuitable for aquaculture and are not considered further. The ArcGIS Pro Cluster and Outlier Analysis tool was used to implement the LISA analysis (Esri 2021a). The fixed-distance spatial conceptualization was utilized within this analysis as it allows the identification of localized clusters. The function inputs were a 250-m search distance and 9,999 iterations with row standardization and a false discovery rate correction applied to allow for more conservative results by estimating the number of false positives for a given confidence level, adjusting the critical p-value accordingly (Esri 2021b). Statistically significant clusters (p < 0.05) of the highest suitable scores (i.e., high-high clusters) were identified, with any clusters smaller than 202 ha (500 ac) excluded, as this was the minimum AOA target size. Next multiple iterations of a 500 acre square site were fit within this shape and dissolved to create the final shape used for the precision siting model. Anselin L. 1995. Local Indicators of Spatial Association—LISA. Geog Anal. 27(2):93–115. Esri. 2021a. ArcGIS Pro: Release 2.8.0. Redlands, CA: Environmental Systems Research Institute. Esri. 2021b. What is a z-score? What is a p-value? Esri ArcGIS Pro online. Available from: https://pro.arcgis.com/en/pro-app/latest/tool-reference/spatial-statistics/what-is-a-z-score-what-is-a-p-value.htm

(5) AOA Options:
Step 5. 1:
The first step in the precision siting model evaluated the high-high cluster output from the LISA cluster analysis and refined each cluster to accommodate at minimum a square option that is 500 ac (i.e., the minimum AOA size requirement). For each of those clusters, an iterative process was developed, where the first iteration was to identify every possible location accommodating a square that is 2,000 ac. Next, all remaining areas within that cluster were examined to determine if additional square options less than 2,000 ac could be placed. Using 500-ac increments, three further iterations were run using 1,500 ac, 1,000 ac, and 500 ac to identify any additional areas within each cluster. Larger size options were prioritized over smaller options, as increased size would support more farms, space to optimally configure farming locations, and maximum flexibility in mooring configurations. However, it is important to note that size was not considered when ranking the options in the next parts of the precision siting model. All potential options identified within a single high-high cluster were ranked using the within-cluster model, which is structured to identify the highest suitable option according to closest proximity to an inlet, lowest relative fishing effort, and lowest relative vessel traffic. The data within these three submodels of the within-cluster model were rescaled using a 0 to 1 range, with 0 being less suitable for aquaculture and 1 being more suitable for aquaculture. This is the same method used in the suitability model; however, it is important to note that the rescaling is performed for the data in each individual cluster in the within-cluster model.

File Information:
10.8 MB total, 79 files in 9 folders (unzipped)
Data File Format(s): ShapeFile .SHP and ancillary files
Data File Compression: none
Data File Resolution: none
GIS Projection: Projected Coordinate System: NAD 1983 (2011) California (Teale) Albers (Meters) WKID 6414 / Geographic Coordinate System: NAD 1983 (2011) WKID 6318 and WGS 1984 WKID 4326