Optimal DEM resolution and neighborhood size for soil resource inventory using the SoLIM approach

1
Optimal DEM resolution and neighborhood size for
soil resource inventory using the SoLIM approach

• Michael Smith
• Department of Geography

2
Research Objective

• The objective of this research is to investigate
  the effect of DEM resolution and neighborhood
  size on soil resource inventory.
• Two different hypotheses will be tested in order
  to achieve this goal
• 1). For a given resolution DEM, there is an
  optimal neighborhood size for conducting soil
  survey,
• 2). The optimal neighborhood size is the same
  across DEMs of different resolutions.

• Optimal creating the most accurate soil resource
  inventories.

3
DEMs

• Digital Elevation Models (DEMs) are continuous
  raster layers in which data values represent
  elevation.

4
Soil resource inventories

• Traditional soil resource inventories examine,
  describe, classify and map soils

5
Using DEMs to characterize soil-formative
environments

• DEMs can be used to create terrain
  characteristics (slope, aspect, curvature, etc.)
  which characterize a soil formative environment.

Slope
Profile Curvature
Planform Curvature
6
DEM resolution and neighborhood size

• DEM resolution cell size
• Neighborhood size spatial extent over which
  terrain characteristics are computed.
• Increasing the neighborhood size has the effect
  of smoothing the data (ie, removing short-range
  variations).

10ft. Resolution, 10ft. NS
10ft. Resolution, 180ft. NS
10ft. Resolution, 90ft. NS
90ft. Resolution, 180ft. NS
90ft. Resolution, 90ft. NS
7
Hypothesis testing

• 1). For a given resolution DEM, there is an
  optimal neighborhood size for conducting soil
  survey,
• 2). The optimal neighborhood size is the same
  across DEMs of different resolutions.
• The SoLIM (Soil-Landscape Inference Model)
  approach is used to derive soil series maps

• GIS databases created from 10ft. (3m), 30ft.
  (10m), and 90ft. (30m) resolution DEMs and many
  different neighborhood sizes are used to produce
  soil series maps.
• These maps are then validated for accuracy by
  field investigations.

8
The SoLIM approach

• SoLIM uses (Zhu, 2001)
• 1) Raster-based GIS and remote sensing techniques
  to characterize the soil-formative environment,
• 2) A set of knowledge acquisition techniques to
  extract soil-landscape knowledge, and
• 3) Fuzzy logic to link the extracted knowledge
  and GIS databases and map the distribution of
  soils across the landscape.

9
SoLIM technical implementation
Knowledge Acquisition
(Knowledgebase)
Zhu, 2001
10
Study site location
11
Soils
12
Sampling scheme

• Three different sample sets

St. Peter Transect (43), Galena Transect (32),
Selected Points(33)
13
Methods the GIS databases

• Elevation data source a 10ft. DEM
• Created from mass points and breaklines
  photogrammetrically collected.

• Coarsening of the 10ft. DEM to 30 and 90ft.
  resolutions occurred using a bilinear
  interpolation technique in Arc/INFO.

14
Environmental data layers (the GIS input into
the SoLIM model)

• The terrain characteristic values at each cell
  will be different for each GIS database!

15
The soils knowledgebase (the knowledge input
into the SoLIM model)
16
The soils knowledgebase (the knowledge input
into the SoLIM model)
17
Experiment one

• The goal of this experiment was to investigate
  the effect of using both coarser resolution DEMs
  and larger neighborhood sizes.

18
Results Maps (DEM resolution, neighborhood size)
10, 120
10, 90
10, 10
30, 90
30, 120
30, 30
90, 90
90, 180
19
Interpreting the Map results

• More continuous distribution of soils as the
  neighborhood size is increased (e.g, removal of
  linear units).
• Pixilated terrain

10, 120
10, 90
10, 10
90, 90
90, 180
20
Experiment 1 results

• The effect of using both coarser resolution DEMs
  and larger neighborhood sizes.

Experiment 1 90ft DEM Resolution
90NS 180NS
21
Experiment 1 analysis

• Trend The highest accuracy value for all sets is
  not at the finest resolution with the smallest
  neighborhood size.
• Highest overall accuracy value is 65.12, St.
  Peter transect, 30 ft. DEM resolution and 90ft.
  neighborhood size.
• Trend The highest accuracy value for a given
  resolution DEM is not at the smallest
  neighborhood size (90ft. resolution exception).
• For the 10ft. resolution DEM (62.79), 100ft.
  neighborhood size, for the 30ft. resolution DEM
  (65.12), 90ft. neighborhood size, both on the
  St. Peter transects.

22
Conclusions

• The results of the experiments in this research
  do not reveal a single optimal neighborhood size
  to use for a given resolution DEM .
• Trend The highest accuracy value for all sets is
  not at the finest resolution with the smallest
  neighborhood size.
• Trend The highest accuracy value for a given
  resolution DEM is not at the smallest
  neighborhood size (90ft. resolution exception).
• Suggestion The highest accuracy values for a
  given experiment and validation set seem to
  center around the 90ft. neighborhood size for
  each DEM resolution.

23
Significance

• The need for very fine scale DEMs.
• Creating the most accurate soil resource
  inventories.
• Implementation of SoLIM in new areas.

24
Thank You!