Relationship between lake substrate sample point density and accuracy of ArcView 3.2 Spatial Analyst interpolated maps.

1
Relationship between lake substrate sample point
density and accuracy of ArcView 3.2 Spatial
Analyst interpolated maps.
Sam M. Stukel, and Michael L. Brown South
Dakota State University
Introduction Lake habitat inventories allow an
assessment of the availability of specific
habitat types. This information is necessary to
investigate relations between environmental
parameters and fish community and population
characteristics. An understanding of these
relationships could assist in identifying
environmental benchmarks that are optimal for
fish species, and allow new insight into
management problems. Based on these inventories,
habitat maps can be created which facilitate
communication among management agencies and the
public. Substrate composition is important in
determining productivity of all trophic levels in
a lake. It provides spawning habitat and both
ambush and escape cover for fishes. ESRIs
Spatial Analyst extension ArcView 3.2 can be used
to create habitat maps based on data collected at
discrete points. Spatial Analyst utilizes
sophisticated algorithms to estimate values for
an entire surface based on a limited number of
sample points. This interpolation procedure
allows users to estimate coverage of habitat type
and predict where certain habitats are found.
Little work has been done investigating the
effects that sample point density has on the
accuracy of interpolated maps. Since determining
sample point density is one of the initial steps
of a habitat mapping process, having some
estimate of the resulting accuracy of the
interpolated maps would be helpful. This poster
investigates the relationship between sample
point density and accuracy of Spatial Analyst
interpolations of lake substrate.

• Results
• Table 1. Test levels and percent of test points
  correctly classified.
• Madison
• Test Level of Points Acres/pt Correct SE
• 1 480 5.5 70.7
  3.6
• 330 8
  69.6 2.4
• 220 12 74.8 3.9
• 4 165 16 68.6 2.8
• 5 132 20 70.3 6.1
• 6 94 28 67.4 3.6
• 7 73 36 62.0 5.6
• 53 50 62.7 3.2
• Brandt
• Test Level of Points Acres/pt Correct SE
• 1 128 5.5 75.8
  3.9
• 85 8
  68.0 6.3
• 73 12 71.7 4.6
• 4 64 16 61.7 4.9

GIS Interpolation Accuracy Assessment The ESRI
Spatial Analyst extension was utilized to
interpolate point data and create substrate maps.
The IDW method was selected with the default
Neighbor and Power settings. Six interpolations
were done at each of the 8 sample point density
levels (table 1) on each lake. For each
interpolation a different set of random points
(50 on Madison, and 20 on Brandt) was used to
assess the accuracy of the habitat map. Results
were recorded as percent of test points correctly
classified. A mean percent accuracy was
calculated for each of the 8 test levels on each
lake.
Study Sites
Lake Madison (1132 ha) and Brandt Lake (407 ha)
in Lake County, SD were chosen as study sites.
Field Methods Substrate was sampled using a Ponar
dredge at 530 randomly selected points on Lake
Madison and 150 points on Brandt Lake. The
samples collected by the dredge were washed
through a set of sieves and classified based on
dominant particle size.

• Implications
• Accuracy of Interpolated maps does increase with
  sample point density
• To create substrate maps that are at least 65
  accurate, no less than one point per 20 acres
  should be sampled.
• Doubling sample point density (and with it, field
  time) will most likely not result in a
  proportional increase in map accuracy.

Scot Hawks