Sampling schemes for digital soil survey
Alex McBratney
The University of SydneyAUSTRALIA
• Random catena sampling• Latin hypercube sampling• Fuzzy k-means
kth-order random toposequences
Seed pointUphill section
Downhill section
Odgers, N. P., McBratney, A. B., Minasny, B. Generation of kth-order random toposequences. Computers & Geosciences. (in press)
Distance (m)
Elev
atio
n (m
)kth-order random toposequences
kth-order random toposequences
Latin Hypercube Sampling
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
Probability X 2
Pro
babi
lity X
1
0
0.2
0.4
0.6
0.8
1
-2 -1 0 1 2
X 2
Cum
ulat
ive
Prob
abilit
y
0
0.2
0.4
0.6
0.8
1
-2 -1 0 1 2
X 1
Cum
ulat
ive
Pro
babi
lity
Latin Hypercube SamplingA random sampling procedure that
ensures a full coverage of the range of
each variable by maximally stratifying
the marginal distribution
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
Cum. Probability Elevation
Cum. Probability Slop
e Latin Hypercube sampling
• Ancillary information
• Latin hypercube sampling
• 300 soil profiles
Soil survey
• Fuzzy k means stratification
Monitoring in the Gulf CountrQueensland
QLD DNR&W
Fuzzy k-means ofThe environmental layersk = 50
Siteses were constrained to be located within 1.5 km from roads
• HELS useful for legacy data checking and new sampling
• Work needed on validation sampling