Geographic Information Technology Applied to Soil Mapping

Sabine GrunwaldAssistant Professor GIS & Land ResourcesDistance Education Coordinator

Conceptual Models of Real World Phenomena

Conceptual Models of Space

1. Entity (object) model = vector model

2. Continuous fields model = raster model

1. Geographical Data Model: Object View

basic geographical data primitives (geographical entities):

Point

Line

Area Pixel

Polygon

(Goodchild, 1992)

Voxel

Polyhedrons

Volumes

Vector Data Model

x1, y1

x2, y2

x3, y3

x4, y4

x5, y5

x6, y6

x7, y7

x8, y8x1, y1

x2, y2

x3, y3

x4, y4

x5, y5

Topology

Land cover: corn

Conceptual Models of Real World Phenomena

Tree is represented bypolygon structure in a visual scene with animage stamped on it

Tree is represented aspoint feature with specificattributes (e.g. height, species, age, etc.)

Conceptual Models of Real World Phenomena

LakeKendrickBonneauMillhopperNorfolkCandlerAredondoApopkaCadillacJonesvilleUrbanWater

Soil map units

N0 3,500 7,000 m

Entity-Based Soil MapAlachuaCounty, FL

2. Geographical Data Model: Continuous Fields

Attribute valuese.g.

1 23

or:slssl

Raster

Raster-Based DEM North Florida

111.0

-2.72

Elevations (m)

N30-m pixel resolution

01 km

1 km

Topographic Models

320.0 - 321.5321.5 - 323.0323.0 - 324.5324.5 - 326.0326.0 - 327.5327.5 - 329.0

> 329.0

Elevation (m)

N(S. Grunwald, 2001)

Raster-Based Land Cover Model

South Florida, Everglades(Sofia, 2002)

30 m

Geographic Soil Model

SSURGO Suwannee County, FL

Entity model:Crisp soil classes

Continuous field model(raster)

N

Geo-Data Modeling

distance

Statistics vs. Geostatistics

Observations are independentor:no correlation between observations

Autocorrelationmodeled as a functionof distance

distance

corr

elat

ion

Z x m x x( ) ( ) '( ) ' '? ? ?? ?

Z: value of a random variablex: locationm(x): deterministic function describing the structural

component of Z at x (“deterministic trend”)

?’(x): stochastic, locally varying but spatiallydependent residual from m(x) -the regionalized variable (“autocorrelated errors”)

?’’: residual; a spatially independent noise termhaving zero mean and variance

Geostatistics

Visualization of Regionalized Variable Theory

(Burrough and McDonnell. 1998. Principles of GIS, Oxford University Press; page 134)

Experimental Semivariogram

?( ) { ( ) ( )}? hn

z x z x hi ii

n

? ? ???1

22

1

zi: value of attribute zx: locationn: number of pairs of sample points of observations

distance (h)sem

ivar

ianc

e.

. .. .

.?

nugget

sill

range

d1 d2 d3

Multi-Dimensional Variogram

0 - 900

90 - 1800

180 - 2700

270 - 3600

z

(S. Grunwald, 2001)

WCA1 (n: 251)

WCA2(n: 280)

WCA3(n: 350)

Holeyland(n: 204)

Water Conservation Areas

0-10 cm depth

range

sill

nugget

10-20 cm depth

range

sill

nugget

20-30 cm depth

range

sillnugget

30-40 cm depth

range

sillnugget

Total Phosphorus

Predictions (Ordinary Kriging)WCA2 / Total Phosphorus (TP) / Depth 0–10 cm

217.0 – 349.1349.1 – 424.2424.2 – 466.9466.9 – 491.2491.2 – 533.9533.9 – 609.0609.0 – 741.1741.1 - 973.5973.5 – 1,381.9

1,381.9 – 2,100.3

TP (mg/kg)

N0 10,000 20,000 m

Prediction Error (Ordinary Kriging) WCA2 / Total Phosphorus (TP) / Depth 0–10 cm

0.27 – 0.330.33 – 0.370.37 – 0.390.39 – 0.400.40 – 0.420.42 – 0.460.46 – 0.520.52 – 0.620.62 – 0.770.77 – 1.0

Prediction standard error

N0 10,000 20,000 m

Predictions (Ordinary Kriging)WCA2 / Total Phosphorus (TP) / Depth 30–40 cm

0.0 – 68.368.3 – 105.9

105.9 – 126.7126.7 – 138.2138.2 – 159.0159.0 – 196.7196.7 – 264.9 264.9 – 388.6388.6 – 612.6612.6 – 1018.5

TP (mg/kg)

N0 10,000 20,000 m

Prediction Error (Ordinary Kriging)WCA2 / Total Phosphorus (TP) / Depth 30–40 cm

16.4 – 44.044.0 – 64.064.0 – 78.478.4 – 88.888.8 – 96.396.3 – 101.7

101.7 – 109.2109.2 – 119.6119.6 – 133.9133.9 – 153.9

Prediction standard error

N0 10,000 20,000 m

Link to 3D models

Conclusion

Geostatistical methods facilitate to describe:

• Changes of physical, chemical, and biologicalsoil properties continuously across landscapes

• Output GIS format: raster or voxel• Variograms describe the spatial variability• Variograms can be used to guide future sampling (targeted, efficient sampling)

x

y

z

3D Soil-Landscapes

(S. Grunwald, 2001)

3D Soil-Landscape Model Types

Characteristics: • Continuous • 3-D

Stratigraphic models

Block models

Link to 3D models

Alachua County, Florida

0 500 1000 mN

TavaresMillhopperMonteochaPomonaLochloosaPlummerWater

Soil Series Elevation (m)707580859095100105

Soil and Topographic Data

(Soil Survey GeographicDatabase, NRCS)

(5-foot contour lines1:24,000, USGS & WMD, 1997)

Soil ProfilesEntisol

Tavares

A

C1

C2

C3

C4

0

25

50

75

100

125

150

175

200

225

250

cm

A

E1

E2

E3

E4

Bt

Btg

BCg

Millhopper

AE1

E2

Bt1Bt2

Btg

BCg

Cg

Lochloosa

A

Eg1

Eg2

Eg3

PlummerUltisol

AE1

E2

Bh1Bh2BE

E’

Btg1

Btg2

Pomona

A

E

Bh

BC

E’

Btg1

Btg2

Cg

MonteochaSpodosol

CCgBCgBtgE’BEBCBhBtEgEA

Stratigraphic Model

Elevations range: 70 – 105 mVertical exaggeration factor: 5 (S. Grunwald, 2001)

3D Scientific VisualizationWCA2; 0-10cm depthTP is represented as heightParameters are represented by color

Ca

Ca

200,000

150,000

100,000

50,000

18,850

Mg

Mg8,000

6,000

4,000

2,000

Fe

Fe1,750

1,500

1,250

1,000

750

500

225Al

Al (mg kg-1)1,500

1,250

1,000

750

500

250

(S. Grunwald, 2001)

3D Scientific VisualizationWCA2; 0-10cm depthTP is represented as heightParameters are represented by color

Sodium bicarbonate extractable P(“bioavailable P”)

NAHCO3 25

20

15

10

5

0

Hydrochloricacid extractable P(“stable P”)

HCLP1 (mg kg-1)1,250

1,000

750

500

250

0 (S. Grunwald, 2001)

Conclusions

3D GIT, geostatistics, and SciVisimproves our understanding of the spatial distribution and variabilityof soil-landscape properties

Home page: http://grunwald.ifas.ufl.edu