Initial steps:
Figure 9: A raster of the mean annual wind speeds at 80m elevation.
First of all, the wind speed shapefile was converted into a raster which had a cell size of 6500 (default) (See Figure 9).
Secondly, the bathymetric data surrounding the UK was in the ASC format and was also retrieved as many multiple separate rasters (since it was gridded bathymetry). Using the conversion tools each ASC grid was converted into a raster, after they were joined using the Mosaic to New Raster tool (See Figure 1).
Figure 10: Distance from the UK shoreline raster
It is more economically feasible to build wind farms closer to the coast, and this means creating a distance- from-the-shoreline raster. This was created by buffering the UK shapefile by 12 nautical miles (the extent of the UK territorial waters) and thereafter increasing (and decreasing) the distance from the shoreline by 6nm (arbitrarily chosen). The final distances were: 0, 6, 12,18, and 24nm. The maximum distance was chosen because it is close to the average distance for offshore wind projects in Europe in 2015 (43.7km ~ 23.4nm) (Ho et al. 2016). The distances were then ranked 1-6 (higher rank = more preferable). All the distance polygons were merged and then converted into a raster (Figure 10).
Normalisation:
The bathymetric, wind speed, and distance rasters were then normalised so that they could be compared with one another. This was done by using the Fuzzy Membership tool; which reclassifies each raster and assigns vales ranging from 0-1. For all normalisations, the linear transformation for the fuzzy membership determination was used. This involves adding a minimum and maximum value (specified range). The raster values closer (greater) than the maximum are assigned values closer to 1 (values of 1).
Figure 11: The Fuzzy Membership Raster for UK bathymetry
The maximum value for the bathymetric raster was set to 0 (sea level) and the minimum value was set to -50m. This value was chosen as the minimum because it is currently the maximum depth for commercial wind turbine structures and thus is the economic limit (EWEA, 2013).
Figure 12: The Fuzzy Membership Raster for Wind Speeds in the UK
For the wind speed raster, the maximum value was the maximum value found in the dataset, the minimum was set at 6 m/s, which is within the ranges suggested by various sources (The City of Calgary, 2016; WEDPEIS, 2016; Government of Alberta Agriculture and Forestry, 2016; Wind Energy Foudation, 2016).
Figure 13: The Fuzzy membership for distance from shoreline.
The distance raster was easily normalised since it had previously been categorised by rank, the minimum was set at rank 1, the maximum at rank 6.
Relative weight determination:
The AHP method was used in order to determine weights for the respective factors using the website My Choice My Decision. The following weights were determined:
Figure 14: Criteria Importance of the chosen factors. Critical Ratio: 0.0706
The values were rounded to: 0.6, 0.3, and 0.1 for depth to seabed, wind speed, and distance from the coast respectively. Depth was subjectively determined to be slightly more important than wind speed since it is a more serious economic constraint to wind power generation. Depth was also chosen to be significantly more important than distance, whereas windspeed was slightly more important than distance. The critical ratio (0.0706) was <0.1 therefore we can assume that the judgements used are justified and not close to randomness (International Hellenic University, Unknown).
Multi-Criteria Evaluation Model and Sensitivity Analysis
The three normalised rasters were then overlaid using the Weighted Sum tool, and were assigned the weights according to the AHP results. The resulting raster was then overlaid with the constraint rasters (Oil and gas fields/wells and wind energy activity) using the Raster Calculator. The resulting raster was then classified so as to show the top 30% of cells which showed the highest suitability (values >0.7) (see Results Section).
Sensitivity Analysis was carried out to see the changes in the results if the three normalised rasters had equal weights (0.33 each). The output was then overlaid on the constraint rasters using the Raster Calculator as done previously and the top 30% of cells selected. Now for the equally-weighted MCE raster and the weighted MCE raster were reclassified so that only the top 30% of cells were shown, and then overlaid using the Weighted Sum tool in order to see how much the different methods affected the number of cells selected as suitable (See Results).