Methods
Background
The SMORES application was built from the spatial planning processes developed for A Wind Energy Area Siting Analysis for the Oregon Call Areas and the Aquaculture Opportunity Atlas’s. These projects provided the marine spatial planning framework that is used by this application.
- Additional Information about A Wind Energy Area Siting Analysis for the Oregon Call Areas can be found in this documentation. This groups work can be cited as: Carlton J, Jossart JA, Pendleton F, Sumait N, Miller J, Thurston-Keller J, Reeb D, Gilbane L, Pereksta D, Schroeder D, Morris Jr JA. 2024. A wind energy area siting analysis for the Oregon Call Areas. Camarillo (CA): U.S. Department of the Interior, Bureau of Ocean Energy Management. 237 p. Report No.: OCS Study BOEM 2024-015.
Components
This application is broken up into different components. These components are carried through the application and provide clarity fir when different functions are applied. These components are as follows:
- Individual Layers
Individual Layers are any layer or data set that is being used in a model run before it is combined with any additional layers.
- Submodel Components
Submodel components are the subgroups of each submodel. These are the result of combining individual layers that all fall within the same general category. An example of a submodel component for the application would be the combined individual layers that fall within the habitat category. This combination of indivudal layers would result in the habitat submodel component of the natural resources submodel.
- Submodels
Submodels are the result from combining submodel components. An example of a submodel for the application would be the natural resources submodel, which is made up of two submodel components: habitat and species.
- Full Model
The full model is the result of combining submodels. In the application you can choose which submodels you would like to include in the full model.
Pre-processing
The app reads in pre-processed data layers in order to generate outputs. These data are processed at 2 different scales 2km and 5km grid cells. Larger spatial areas utilize the larger grid size in order to create the most efficient memory usage for the app. There are two data types that this app uses, presence data and continuous data.
Presence Data
Presence data is pre-processed using a threshold method. The raw data is overlaid with the 2km and 5km grid cells, and the area of overlap is calculated. If a grid cell has 50% or greater coverage by a presence data set, that grid cell is retained. These retained grid cells are then processed so that a user can select their score of choice.
Continuous Data
Continuous data is pre-processed using an area-weighted average method to transition from a 2x2 km to a 5x5 km resolution. The 2km data is overlaid with the 5x5 km grid cells, and the area of overlap is calculated. For each overlapping fragment, its original score is multiplied by the percentage of the 5km cell it occupies. These proportionally weighted scores are then mathematically summed together to calculate the final continuous score for each 5x5 km grid cell.
All data are transformed to be in the WGS 84 coordinate system for compatibility with leaflet() mapping. Leaflet is the mapping engine utilized by this shiny app. Data are saved as a geoparquet file format. This format was chosen for its ability to efficiently store large spatial data files.
Score Options
A motivation of the SMORES application is to allow for NMFS to model the impacts of different scoring scenarios before recommending scoring scenarios to BOEM. In order to achieve this, different individual layers within the app have score drop down menus to allow for different score selection options. Based on how data were created and the best practices for that data type, individual layers have three options for the scores presented: static, z-membership, and ranked importance.
Static Score
Static scored data is selected when data are represented with a single value (i.e. 0, 0.9, etc.). These data represent the presence of a feature for each individual layer for a given 2x2-km cell. Cells are assigned values between 0 to 1 with a step of 0.1, or values to represent constraints such as 0.01 or 0.001. Values are determined by the individual who is running model scenarios. Cells that do not have a feature present for that individual layer are assigned a score of 1, based off of the methods used in the Aquaculture Opportunity Area Atlas for the Southern California Bight. To determine which value to use as a score, there are scoring recommendations and guidelines from the Oregon Call Areas Report, presented on the application for the component that is being viewed (i.e. recommended species scores are shown on the species sub-tab of the natural resource submodel).
Z-Membership
Z-membership data is selected when data has been re-scaled using a fuzzy logic Z-shaped membership function from 0 to 1. The Z membership function is applied in the data_production.R file. Cells that do not have a feature modeled by the z-membership function are assigned a score of NA.
Ranked Importance
Ranked Importance data is used to represent fisheries data per the fisheries data and methods provided by NMFS in Appendix E of A Wind Energy Area Siting Analysis for the Oregon Call Areas Report. For more information please visit the Data page. Cells that do not have a ranked importance score for that individual layer are assigned a score of 1.
*Note that fisheries layers were developed specifically for the Oregon call area siting analysis and therefore only cover the geographic extent of these call areas at this time. Fisheries data should only be included in the model if you are interested in the Coos Bay and Brookings WEA’s, or the Washington/Oregon OCS Planning area. Configuring fisheries layers for other areas will result in a high suitability score (1) being applied to each of these cells.
Individual layers to submodel components
Geometric Mean
The geometric mean equation used was:
\[\text{g} = \sqrt[n]{x_1 \times x_2 \times \ldots \times x_i}\]
n = number of variables
x₁ = variable 1
x₂ = variable 2
xᵢ = additional variables
Lowest Value
The lowest equation used was:
\[\text{l} = \min(x_1, x_2, \ldots, x_i)\]
x₁ = variable 1
x₂ = variable 2
xᵢ = additional variables
Product
The product equation used was:
\[\text{p} = x_1 \times x_2 \times \ldots \times x_i\]
x₁ = variable 1
x₂ = variable 2
xᵢ = additional variables
Components to Combined Submodels
Geometric Mean
The geometric mean equation used was:
\[\text{g} = \sqrt[n]{x_1 \times x_2 \times \ldots \times x_i}\]
n = number of variables
x₁ = variable 1
x₂ = variable 2
xᵢ = additional variables
Combined submodels to Full Model
Weighted Geometric Mean
To generate the full model map a weighted geometric mean calculation was used:
\[\text{wg} = ({x_1^{w_1} \times x_2^{w_2} \times \ldots \times x_i^{w_i}})^{\frac{1}{\sum_{i=1}^{n} w_i}}\]
x₁ = variable 1
x₂ = variable 2
xᵢ = additional variables
w₁ = weight for variable 1
w₂ = weight for variable 2
wᵢ = weights for additional variables