I have some points that represent animal locations over the course of a few months at the scale of ~20 sqkm. I am emulating the work of previous researchers in this field, but I'm having trouble understanding why they chose the parameters they did. Specifically, I want to know how they chose the bandwidth and the scaling factor. Furthermore, what is the difference between the scaling factor and the bandwidth? It is my understanding that the scaling factor helps to transform bandwidths between different types of kernels, but I don't understand how that would apply with just one kernel in this case.

Since they didn't provide an explanation of how they arrived at the parameters, I was hoping someone would be able to estimate, back of the envelope, if these parameters look appropriate for the locations and kde provided. If not, how would I determine an appropriate bandwidth and scaling factor myself? The irregular rectangle depicted is roughly 15km x 15km. The citation is Management of pinyon-juniper woodlands at Kirtland Air Force Base: Pinyon Jay summer and winter home ranges and habitat use 2009 final report and below.

Referenced literature:

For the kernel analysis we used the Hawth’s Tool, Fixed Kernel Density Estimator. Kernel analysis is a nonparametric statistical method for estimating probability densities from a set of points. When used to analyze home range data, kernel density methods describe the probability of finding an animal in any one place. The Fixed Kernel Density Estimator calculates a fixed kernel density estimate and produces contour lines representing the boundary of the area that contains a specified percent of the volume of a probability density distribution. A 95% volume contour, for example, typically contains 95% of the points used to generate the kernel density estimate. Kernel parameters were set as follows: scaling factor- 1000000, kernel- bivariate normal, single parameter smoothing factor (h)-1000, raster cell size-100, and percent volume contours- 50, 90, and 95.

  • Perhaps this may help. In "GIS Fundamentals" by Paul Bolstad he writes that for a Gaussian kernel, h(optimal) = sigma(2/3n)^.25 where sigma is the estimated standard deviation parameter (5th edition, Ch12, pg 554). In practical terms, what does sigma represent? Is it the standard deviation of the distances from points on a map? Would the units be in meters? – PIJA Nov 7 '18 at 21:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.