Overview:
In lab today I built upon what I have learned about working with attribute data by engaging in spatial joins, and doing some spatial analysis.
Objectives:
Deliverables:
The lab document with answered questions, along with cartographically correct maps.
Process:
Importing shapefiles into the geodatabase: Until now, I have always worked with feature classes within my database, but sometimes I will encounter data that needs to be imported. I copied over a shapefile into my file, and then imported that into my geodatabase.
I first established a folder connection within ArcGIS Pro for the Censusdata folder. Our professor, Dr. Hupy, provided us with the cenus dataset that we used in the lab.
This was a shapefile, but shapefiles cannot be used for many forms of analysis, and their size is greatly reduced when brought into a geodatabase. We then went about importing this file. To do this, we used the ‘feature to feature’ tool within the geoprocessing tab within ArcGIS Pro. The feature class to feature class tool is used to convert a feature class or feature layer to a feature class. I imported this into my geodatabase and named it stateboundaries.
Fig. 1: stateboundaries with continental states selected and highlighted in lambert.
The main issue with this feature class is that it included other states outside of the continental United States. To solve this I used the 'select with lasso tool'. With this tool I was able to grab and select only the data that I wanted, in this case the continental US. Once selected, I exported it as a new feature class and named it ‘continental_us_states’. I then projected this feature class into lambert conformal conic and named it ‘cont_us_state_lambert’.
Fig. 2: Continental United States in lambert conformal conic.
I then began to do some analysis with the airmen points. Because this analysis takes some time with such a big data set, I started working within just Indiana. So, I selected and created a feature class just for the Indiana State Boundary. From there, I did a select by location query. The types of ‘Select by location’:
Intersect: Anything that crosses the location.
Contains: Anything that may be inside of the location.
Within: Anything that is only inside the location.
Dr. Hupy requested that we engaged in the Query using:
Airmen points as the input
Intersect
Our Indiana boundary file as the selecting feature
I exported this as a new feature class called ‘Indianaairmen’.
Next, I engaged in some spatial analysis. For this I only used two layers, the Indiana boundary feature layer and the Indiana airmen feature layer. I used the 'Kernel Density' tool within the ArcGIS Pro toolbox which is was the best option to find where points are clustered.
Most of the clustering was in the middle near Indianapolis, near Lafayette, and the northwest corner near Chicago.
Fig. 3: Indiana airmen kernel density map layout.
Next, I engaged in hotspot analysis, which is kind of like kernel analysis, but using information gained in a spatial join. For this analysis, I again used Indiana, but first needed to do some prep work first. I had to extract only Indiana information from my polygon feature classes.
I performed a 'select by location' query with:
Us_zipcode as Input
Intersect relationship
Indiana Boundary as Selecting features
Fig. 4: Selection by location of US zip codes that intersect with the Indiana boundary.
Since I chose intersect, zip codes that intersect with the Indiana zip codes will also appear, therefore it will not be a perfect outline of Indiana.
In certain cases, I can’t really do a location query, and instead need to clip out what we want. For this I did a clip operation using the 'clip' tool in the toolbox. The input feature I used was the us_zipcode feature file in Lambert projection. The clip feature was called Indian state boundary feature file. I named the new feature: Indianazipcode. I now had an exact cookie cutter of only the Indiana zip codes.
Fig. 5: US zip codes that was clipped into Indiana Lambert.
Next, I did the same clip operation on the fishnet_10km feature class in Lambert projection.
Fig. 6: US fishnet that was clipped into Indiana Lambert.
Now that I have prepared the data, I can do a spatial join operation. Within the Geoprocessing tools tab, I used 'Spatial Join'. I started with the Indiana counties feature with the following settings:
Target Feature: Indiana counties feature class
Join Features: Indianaairmen feature class
Output feature class: Indianacounties_joined
Join operation should be one to one
Match operation should be intersect
After it finished I opened the attribute table. Notice how there is a join count field. I sorted the join count field in descending order. The top five counties are listed below
Marion
Hamilton
Lake
Allen
Hendricks
I repeated the same operation on:
Indiana zip code feature class
Indiana fishnet feature class
I used that join count to perform a hotspot analysis. The hotspot is somewhat like the kernel operation, but now I am looking at the number of points within a given area such as the county, zipcode, or 10km square. The example below will be for the 10km fishnet grid. I used the ‘optimized hot spot analysis’ within the geoprocessing tools window with the setting below:
Input: your joined 10km fishnet.
Output: Indiana_fishnet_10km_hotspot
Analysis Field: Join_count
The hotspots are in the middle of Indiana and the northeast corner just like the kernel feature because that is where there is a high concentration of airmen. I then did the same thing for the joined county feature and the joined zipcode features.
Fig. 7: Map layout of the four different airmen analysis that were completed.
I think that the zip code hotspot does the best form of analysis because it is the most detailed and is more specific in the area of distribution compared to the kernel and the county analysis where it is much less detailed and more generalized of the airmen distribution. The fishnet also did a decent job of analysis, but I would prefer the zip code because it also included the “cold spots” as well.
I then created a map of the Continental U.S. showing kernel density of airmen with P type.
Fig. 8: Map layout of the kernel density of type p airmen in the US.
This map has less density than the U rating but there is a pattern of high density in big cities especially the ones that have major airline hubs associated with them.
Next, a map of the Continental U.S. showing kernel density of airmen with U type.
Fig. 9: Map layout of the kernel density of type U airmen in the US.
Again, this map is more dense than the P rating because there is a higher distribution of U ratings than P ratings.
Finally, an optimized hotspot analysis of airmen in the continental U.S. using the 10km fishnet grid.
Fig. 8: Map layout of the hotspots of airmen in the US fishnet.
This map does a good job of showing the airmen hotspots in the US. Like I mentioned above, there appears to be more hotspots over cities that are home to an airline hub like Atlanta, New York, and Minneapolis just to name a few.
Comments