### Kernel Density Analysis

```Density Estimation
• Converts points to a raster
• The density of points in the
neighborhood of a pixel
• No “Z” value is used
• ArcMap has a simple “Point Density” tool
– Each pixel=number of points within radius
• Kernel Density is related to Kernel
Smoothing but different
Density Estimation
• Simple point density: Golf courses
Rockware
Point Density in ArcMap
Distance=0.3
Distance=3
Point Density in ArcMap
Distance=10
Kernel Smoothing
• Kernel Smoothing is interpolation
Density Estimation Using Kernels
• Creates a raster from points
– Weight (attribute) optional
– Not really interpolation
• “Kernel function” applied to points near
target pixel
• Different functions are available
• High parameters make a “wide” pile,
small values make a “narrow” pile
Width of Kernel
• Determines
smoothness of
surface
– narrow kernels
produce bumpy
surfaces
– wide kernels
produce smooth
surfaces
Kernel Density in ArcGIS 10
• Under Spatial Analyst -> Kernel
Density
• The kernel function is based on the
Silverman (1986, p. 76, equation 4.5).
Overview
• This analysis show where point features are
concentrated.
• Estimations are based on probability “kernels”
– regions around each point location containing some
likelihood of point presence.
• The width of the kernel is based on the
smoothing parameter (h)
• The output is often called a Utilization
Distribution (UD) Grid.
• Methods include: minimum convex polygons,
bivariate ellipses, adaptive and fixed kernels
Kernel Density in ArcGIS
Kernel Density
• Cell Size = 0.05
Kernel Density
• Cell Size = 0.05
How to select parameters?
• What should the cell size be?
• What should the search radius be?
Origins of Computer Viruses
Origins of Email Spam
Kernel Density Analysis
Amelia O’Connor
Kernel Density Output
Other tool extensions for kernel density:
•
•
•
•
•
Home Range Tools
Animal Movement
Biotas
Home Ranger 1.5
KernelHR
Spatial Stats Toolbox
• New in ArcGIS 10
ArcGIS 10.2
• By Lauren
Rosenshein
Hot-Spot Analysis
• Layer may show “hot-spot” but is it
really?
• Z-score and P-value are required
– Z-score = high or low values together?
– P-value = random?
Hot-Spot Analysis
• High z-values indicate a
significantly high or low
value
– 2.5=cluster of high or low
values
• P-value is the chance a
pattern is random
– 0.01=probably not random
Hot-Spot Analysis Tool
Citations
• Bugoni, L., D'Alba, L., and Furness, R. W. (2009) Marine
habitat use of wintering spectacled petrels Procellaria
conspicillata, and overlap with longline fishery. Marine
Ecology Progress Series 374:273-285.
• Mitchell, Brian R. (2007) Comparison of Programs for
Fixed Kernel Home Range Analysis