Background and Question Asked
In Exercise 1, different interpolation methods were used to create a heat map of rockfish abundance based off of a large collection of point data. That blog discussed some of the challenges that arose while attempting to use a time series of point data with many points in close proximity to one another (if not overlapping). The exploring was in many ways successful: it was discovered that the Kriging method provided a more robust representation of the data than Inverse Distance Weighting. However, in the time since that post was published, my interpolation methods have been refined:
- Instead of using the entire time series as an input for the interpolation, four individual years were selected to represent the whole dataset (2003, 2007, 2011, 2015). Kriging was then used to create heat maps for each individual year.
- Additionally, the union tool was used to remove the land boundaries from the environment so that the interpolation only affected parts of the ocean
- The symbology of the abundance point data was synced across all four years being used in the analyses so that they could be easily compared to one another
- The symbology of the interpolated heat maps was also modified to be consistent throughout the analyses
For this exercise, I plan to compare my new, interpolated data to an already existing set of data, effectively comparing my two variables. Specifically, I hope to answer the question “Is there a spatial relationship between areas of significantly high and low rockfish abundances and specific seascape classes?”
An example of the most recent Kriging output using the 2007 data.
Name of Tool or Approach Used
I will be using a neighborhood analysis to seek an answer to this question. The neighborhood analysis requires taking areas of interest and examining the environmental conditions around that area from the perspective of another variable. By varying the distance from your original point of interest, a researcher is able to infer about the spatial relationship between the two variables.
Methods
Data Used
- “Points of Interest” chosen from plot below
- Buffers created around points of interest at 5km, 10km, and 25km radii
- YOYRock Kriging Abundance Interpolation for 2007
- Seascape NetCDF Raster File for May 5th, 2007
Rockfish abundance plotted against water column depth for trawls from 2007.
The first thing that was needed to complete this analysis was points of interest. I chose to use four points form the year 2007, as the data from this year provided the largest spatial footprint of all of the years of interest. Two of the points represented trawls that found significantly high rockfish abundance, and the other two represent trawls which found no rockfish. All four points vary spatially and physically (latitude, longitude, water column depth, etc). All points were selected from interpolated areas with different modeled outputs. Next, circular buffers were created around each point of interest with 5km, 10km, and 25km radii.
Map showing Points of Interest with circular buffers overlaid on seascape NetCDF file.
In order to use the overlay tool in ArcGIS, two polygon features are needed. In order to convert my NetCDF Raster files into a polygon, I used the Raster to Polygon tool. Once the seascape classes were converted to polygons, the Intersect tool was used to measure the shape area of each seascape class within each buffer. Those statistics were then converted to .xlsx files and summarized in Excel.
Results and Discussion
Example of data after Raster to Polygon and Intersect tools used
The neighborhood analysis found evidence that specific seascape classes may have impacted young of the year rockfish abundances in the locations selected to be a part of this analysis.
The low-abundance trawls were dominated by three seascape classes: Class 14 (Temperate Upwelling Blooms), Class 19 (Subpolar Shelves), and Class 21 (Warm, Blooms, high Nutrients). While there were more classes represented overall by the high abundance trawls, those areas were mostly dominated by two seascape classes: Class 7 (Unnamed) and Class 12 (Subpolar Nutrient). Additionally, there was very little overlap between the two areas – the only seascape class that appeared in both the high abundance radii and the low abundance radii was Class 14. Further analyses would be needed to determine if these trends are representative to the entire region or year, but this neighborhood analysis provides results that give us a place to start. Overall, I found this analysis to be extremely useful despite the number of steps needed to make it work. In addition to working in GIS normally, the data type of my seascapes had to be changed and much of my analysis had to be done in Excel, as ArcGIS cannot summarize key statistics. However, I feel as though streamlining this method could be done now that I am familiar with it.
