Discrepancy between buffer and point distance function

[KeriO_Leary]

Hi! Hoping I can get some guidance on this. I am getting a discrepancy between using a buffer and a point distance approach. What I am trying to do is to see if there are industries located within one mile of patients' homes. I have coordinates for the homes and coordinates for the industries and have mapped them in ArcMap. I then created a one mile buffer around the homes. Because I also need to calculate the distance between the homes and the industries (for all industries located within one mile of the home), I also used the Point Distance function using the Python script example described below. I used a search radius of 1 mile and the output looked great. My problem is that I was checking the output to make sure it matched up with the buffers in my map and found that it did not. For example, one home had 4 industries that fell within the 1 mile buffer (by sight) but the Point Distance analysis indicated that only 3 industries were located within a one mile radius. I am assuming the Point Distance approach is more reliable but it is bothering me that there is a clear discrepancy between the two approaches and that I may be doing something wrong. Any help or advice would be appreciated! TIA!

# Description: Finds distance of each near point from each input point and outputs to a table. # import system modules import arcpy

# set workspace environment arcpy.env.workspace = "C:/data/pointdistance.gdb"

# set variables in_features = "police_stations"

near_features = "crime_location"

out_table = "crime_distance4"

search_radius = "22000 Feet"

# find crime locations within the search radius arcpy.PointDistance_analysis(in_features, near_features, out_table, search_radius)

[JimCousins]

In order of likelihood:

Buffer will look to do an intersect, and rasterizes the features. So, dependent on the cell size used, you may miss points, or erroneously include points. 

Another consideration is the projection you are using - does it preserve distance?

The last item I can think of is the buffer parameter for METHOD - Euclidean vs Geodesic. This will only effect you if your coordinate system is not planar. 

More information is available in the ArcGIS help under Buffer Analysis.

Best Regards,

Jim

[KeriO_Leary]

Jim, 

Thank you so much for your response--I really appreciate it!  First, I used the Planar v. Geodesic methods for the buffer but did not find a difference. So, next I measured the distances (using the Measure tool) between the industries and the home (where I found the discrepancy) and found that the distances that I measured were different from the distances that were generated using the PointDistance approach/Python code. My home layer and my industry layers use the GCS_WGS_1984 geographic coordinate system and the D_WGS_1984 Datum. I am looking at the industries in all of the counties in one state and so I also have the USA_States_Generalized and USA_Counties_Generalized layers that I downloaded from ArcGIS online. Those layers are in the WGS_1984_Web_Mercator_Auxiliary_Sphere projected coordinate system. The data frame's coordinate system is GCS_WGS_1984. From doing some google searches, I am finding that the WGS_1984_Web_Mercator_Auxiliary_Sphere projection can distort distances. Therefore, I am guessing this is my problem. Does that mean I need to do a transformation? I have been examining a table of ArcGIS supported map projections and looking at the ones with the equidistant property but there are at least 6 to choose from. Any thoughts on this? Thanks again!!!

K

[JimCousins]

Apologies, I neglected to say that I would trust the results of the point distance analysis over the buffer analysis.

Jim