# Ryan Connaghan Lab 6

Skip to end of metadata
Go to start of metadata

Ryan Connaghan Lab 6

The distribution of qocha is not normal, but very highly clustered.  In addition to that, there are several outliers in both distance and size.  Below are the outliers in distance.  Those highlighted in blue are the outliers by distance. I am using ArcMap 10, which allowed me to find the nearest neighbor without converting the qochas to points.  However, the density tool required points rather than polygons, so I was still required to convert the polygons to points to complete the later parts of the lab.

By calculating the nearest neighbors using the polygons, I found that some polygons overlapped.  These are either errors, or are qochas that overlapped or has a figure 8 shape to them and were mapped as 2 units.  These would have been missed if the near value was computed from points.  This meant that my calculated mean was much lower than when calculated from the center points.  Below is an image with 2 overlapping polygons, which gave near values of 0. The point density map is below.  The cell size is 30, and I left the default radius in place.  This map clearly shows the voids where no qochas are present, but missing is the large qocha. The kernel density uses area in the population field, and uses a weighted value to calculate the density.  This is seen below, and in this case, the lake is a large circle. You can also see several others that are large outliers. The point density tells us the density of the qochas numerically, and does not take into account the size of the qocha.  The kernel density takes the size into account, and gives the qochas that are larger carry more weight in the calculation.

Below are the images zoomed in on the large qocha, showing the way it was taken into account in both point and kernel.  The cell size here is 30m.

Point density Kernel Density I was interested in the way the radius affected the kernel density, so below are maps with radii of 160(average + 2 STDEV), 482(ArcMap given value), and 1000(chosen by me for contrast)

160 482 1000 You can see that a larger radii gives you a courser set of data.  A radii that has meaning would need to be selected for this to be useful.

• No labels