DSM in Tasmania

Report
USING DIGITAL SOIL MAPPING FOR
ENTERPRISE SUITABILITY ASSESSMENT IN
SUPPORT OF TASMANIAN IRRIGATION
DEVELOPMENT
Darren Kidd1, Brendan Malone2, Alex McBratney2,
Budiman Minasnay2, Mathew Webb1, Chris Grose1, Rob
Moreton1, Raphael Viscarra-Rossel3, William Cotching4,
Leigh Sparrow4, Rowan Smith4
1 Department of Primary Industries, Parks, Water and Environment, Prospect, TAS.
2 University of Sydney, Faculty of Agriculture and Environment, Eveleigh NSW.
3 CSIRO Land and Water, Canberra, ACT.
4 Tasmanian Institute of Agriculture, University of Tasmania, Launceston TAS.
WEALTH FROM WATER
Wealth from Water Pilot Program – 2 Year, joint initiative between The Tasmanian
Department of Primary Industries, Parks, Water & Environment (DPIPWE), the
Tasmanian Institute of Agriculture (TIA), and the Department of Economic
Development, Tourism & the Arts (DEDTA)
Develop a Decision Support Tool - Enterprise Suitability Assessment to aid
irrigation development in Tasmania for a range of different enterprises
AIMS:
1. Generate comprehensive soil, climate, crop and enterprise suitability data.
2.
Classify Land (within Tasmanian Irrigation Schemes) according to its suitability
for various agricultural enterprises (Approximately 70,000ha as a pilot, 20
Enterprises)
3.
Provide Farm Business Planning Tools, Market and Technological Information
to help farmers or investors develop, diversify or intensify into new irrigated
enterprises
PILOT AREAS
Total - 70,000 ha
• Meander Irrigation Scheme, (43,000 ha)
• Midlands Irrigation Scheme (Tunbridge District, 27,000 ha)
Areas chosen to
cover a diverse
range of soils,
existing land
uses, terrain
and climatic
conditions
EXISTING DATA INADEQUACIES……..
Only available mapping– 1:100,000 Quamby (1959) and Interlaken
(1963) – partial coverage
Soil Property (eg. pH)
would be determined from
a Modal Soil Type applied
to entire polygon
Highly complex alluvial
plains – mapped as one
Miscellaneous Soil Unit
Sustainable Land Use and Information Management Section
Department of Primary Industries Parks Water & Environment
APPROACH
•
Digital Soil Mapping Approach (Predictive Soil Mapping – raster based,
associated uncertainties of predictions)
Soil Data (point and/ or polygon)
Sp/Sc = ∫(S,C,O,R,P,A,N)
……..McBratney et al 2003

Climate (rainfall, temp)
Organisms (vegetation, land use)
Relief (DEM terrain analysis)
Parent Material (geological maps)
Age (age of material, temporal
components)
N (spatial coordinates, spatial
variability)
•
DPIPWE has formed partnerships with the University of Sydney (Faculty of
Agriculture and Environment) to apply and develop departmental capacity in
the latest Digital Soil Mapping (DSM) techniques, through and ARC linkage
Project.
•
There has also been collaboration with the Australian Collaborative Land
Evaluation Program (ACLEP).
•
•
•
Radiometric Mapping
MiR analyses
DSM training
Sustainable Land Use and Information Management Section
Department of Primary Industries Parks Water & Environment
ESM REQUIREMENTS
Map Suitability for 20
different enterprises
Based on Enterprise-Specific Soil; Terrain; and
Climate Parameters
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
Poppies
Carrots
Carrot Seed
Barley
Blueberries
Hazelnuts
Industrial Hemp
Pyrethrum
Rye Grass
Lucerne
Cherries
Wheat
Onions
Strawberry
Raspberry
Potatoes
Wine grapes
Linseed
Olives
•
•
•
•
•
pH (water) 0 to 15cm
ECse 0 to 15cm
Clay% 0 to 15cm
Soil Drainage Class (Yellow Book)
Stone% Class (2 to 200mm, > 60mm, >200mm)
0 to 15cm
Soil Depth/ Depth to Impeding Layer
Depth to Sodic Layer (ESP > 6%)
Duplex Clay Presence @ 0 to 40cm (carrots)
Exch Ca 0 to 15cm (onions)
Exch Mg 0 to 15cm (onions)
•
Slope %
•
•
•
•
•
Frost Risk (Enterprise time specific)
Heat Risk (Enterprise time specific)
Rainfall (Enterprise time specific)
Chill Hours (Enterprise time specific)
Growing Degree Days (Enterprise time specific)
SAMPLE SUITABILITY RULE
Onions (Allium cepa)
Soil
Depth
(Depth to
heavy
clay)
pH H2O
(top
15cm)
ECse*
(top 15
cm)
Drainage
Well Suited
>25cm
> 6.0
<2.0
Suitable
>25cm
> 6.0
Marginally
Suitable
20-25cm
Unsuitable
<20cm
Suitability
class
Heat at harvest
> 3 days where
Tmax > 31oC at
harvest
(January or
February)
Rainfall at
harvest (> 3 days
in any 7 day
period with ≥ 5
mm rain/day
during January –
March)
Spring frost
At least 1 day
where
Stoniness
(stones> 60
mm in the
top 15 cm)
Slope
Excessive;
Well
< 2 % (1)
<5%
> 2000 ppm
> 120 ppm
<1/10 years
<1/5 years
≤ 2/5 years
2.0 - 4.0
Mod Well
< 2 % (1)
5 - 10 %
> 2000 ppm
> 120 ppm
1/10 - 2/10
1/5-3/10
≤ 2/5 years
5.8 6.0
2.0 - 4.0
Imperfect
2 - 10 %
(2)
10 - 20
%
> 2000 ppm
> 120 ppm
2/10 – 3/10
3/10-2/5
>2/5 years
< 5.8
> 4.0
Poor; very
poor
> 10%
(>=3)
>20%
< 2000 ppm
< 120 ppm
> 3/10
>2/5 years
>2/5 years
*Uses Most-limiting Factor approach
Exch Ca
(top 15 cm)
Exch Mg
(top 15 cm)
Tmin <0oC in
November
METHODS
Logistics
• 30m Resolution Mapping
• 930 soil cores
(650 training/ 280 validation)
• 271 temperature sensors (tiny-tag loggers)
• 6 climate stations – temperature, humidity, rainfall
Covariates (Explanatory Environmental
Variables)
•
•
•
•
•
•
Legacy Soil and Land Capability Mapping
Land Use Mapping
Geology
SRTM-DEM and terrain derivatives (Slope, TWI,
MrVBF, MrRTF etc)
NDVI/ FVC (SPOT, RapideEye, LandSat)
Gamma Radiometrics (Dose, U, Th, K)
Prediction Methods
•
•
•
•
•
Regression Trees (Cubist/ R) (Continuous Data)
Logistic Modeling (See5) (Discrete/ Ordinal Data)
Random Forests (R)
Universal Kriging/ Regression Kriging (SAGA/ R)
Artificial Neural Networks (R/ JMP)
METHODS
Sampling
• Condition Latin Hyper-cube (stratified covariates)
• Fuzzy k-means Clustering (stratified-random)
Data Collection
•
•
•
•
“Yellow Book” field descriptions
Soil Cores sampled by horizon to 1.5m
MiR analyses (with 15-20% wet-chem calibrations)
Fitted Depth-Splines for standardised depths
Validation
• Independent Sites, sampled at time of training sampling
Soil Property Uncertainty
• Upper/ Lower Limits (based on distance to fuzzy k-means centroids of covariates used for prediction)
Suitability Model
• Most-limiting factor
• Queries based on Enterprise Suitability rules, trials, industry experts, agronomists (TIA)
• Compiled with ESRI Model-builder
Suitability Uncertainties/ Probabilities
• Monte-Carlo simulations, based on suitability classification for each parameter, for each pixel, between the upper and
lower limits (normally distributed)
SUITABILITY MODELS
Pyrethrum
Land Suitability Model –
applies rules for each
enterprise to soil and
climate surfaces – results
in a Land Suitability
Rating plus the limitations
for each 30m pixel
SOIL PROPERTY SURFACES
pH (0 to 15cm) Tunbridge – Cubist
Training (Lin’s) Concordance* = 0.82
Validation Concordance = 0.45
Residual Standard Error = 0.23
*Concordance – correlation coefficient around the 1:1 line
SOIL PROPERTY SURFACES
Soil Drainage Index
Cubist training
Cubist validation
SOIL PROPERTY SURFACES
Coarse Fragments 2 to 200mm
Soil Depth (cm)
R-2 Validation = 0.56
Concordance = 0.69
CLIMATE MAPPING - METHODOLOGY
•
Locate temperature loggers in the study areas at a
density of 1 logger per 250ha, using stratified random
sampling of terrain covariates
•
Record temperature for 1 year at 10 minute intervals
•
Correlate to surrounding Bureau of Meteorology (BOM) stations to
obtain a linear relationship, use equation to derive 20 years worth of
temperature data at daily and hourly intervals for each logger
•
Model climatic parameters using spatial interpolation (via adopted
DSM techniques) to model the quantified risk values (i.e. frost risk)
of each logger.
EXAMPLE OF GRID GENERATED FROM BOM STATION RECORDINGS (1/10/2011)
Maximum
temperature for
1/10/2011
Logger data correlated with
BoM historical data at for
corresponding grid
Using Elevation as
explanatory variable
Minimum
temperature for
1/10/2011
A total of 7305 temperature grids
produced!
Sustainable Land Use and Information Management Section
Department of Primary Industries Parks Water & Environment
19
0.92
0.96
22
logger
0.93 Temp
number (also
0.95
59
0.92
0.96
0.94
0.95
0.95
0.96
0.93
0.93
0.92
0.92
0.93
0.89
0.93
0.98
0.96
0.96
0.96
0.96
0.96
0.91
0.93
0.86
0.93
0.94
0.94
0.96
0.96
0.96
0.96
0.92
0.93
0.96
0.96
0.92
0.92
0.93
0.94
0.97
0.95
0.95
0.96
COEFFICIENT OF DETERMINATION (R2 ) OBTAINED BETWEEN TEMPERATURE LOGGER READINGS
20
0.89 FROM BOM
0.94 STATION0.92
0.94
0.96 OF
0.96
AND GRIDS
PRODUCED
RECORDINGS60USING 3 MONTHS
WORTH
TEMPERATURE
LOGGING
DATA 0.96
(1/08/2012 TO0.94
1/11/2012). 61
21
0.91
0.87
0.92
0.92
23
24
25
0.89
0.85
0.93
refer to figure 6)
1
2
0.97
0.86
3
0.89
33
6
0.86
0.96
0.94
0.95 0.92
46
7
0.9
0.97
0.95
47
0.9
0.96
0.92
0.96
0.93 0.93
48
0.93
0.96
49
0.91
0.95
9
0.96
0.91
0.96
0.96 0.95
50
0.95
12
0.93
0.93
0.95
52
51
13
0.97
0.91
0.95
0.95 0.93
14
0.9
0.96
0.95
54
15
0.9
0.97
0.94
0.96 0.93
55
16
0.93
0.94
0.96
56
0.94
17
0.87
0.95
0.91
0.95
0.95 0.92
57
0.93
0.92
58
0.94
19
0.92
0.95
0.93
0.92
0.92
18
20
0.94
0.89
0.94
0.97 0.92
59
21
0.91
0.96
0.94
61
60
22
0.94
0.93
0.96
0.95 0.95
23
0.89
0.97
0.93
63
24
0.85
0.94
0.93
0.95 0.92
64
25
0.93
0.96
0.95
65
0.89
0.96
0.96
0.96
0.96 0.93
66
0.94
0.96
67
0.91
0.95
28
0.92
53
0.96
0.92
40
44
11
35
39
43
45
26
38
42
0.92
0.92
37
0.92
0.92 0.91
34
36
0.93 0.92
0.85
0.94
10
32
0.93
0.94
0.94
31
0.96
Temp logger
number
(continued)
41
0.89
27
30
2
value
0.93
0.86
8
29
Hourly
0.95
temperature R
5
0.89
0.91
Daily maximum
temperature R2
value
0.97
4
26
28
0.96
Daily minimum
temperature R2
value
0.93
27
62
0.96
0.93
0.97
0.97
0.96 0.96
68
29
30
0.94
0.95
0.95
70
69
31
0.95
0.94
0.94
0.94 0.97
32
0.92
0.94
0.95
72
33
0.92
0.96
0.94
0.95 0.95
73
34
0.92
0.96
0.96
74
0.93
35
0.92
0.95
0.96
0.95
0.95 0.96
75
0.92
0.94
76
0.9
37
0.93
0.95
0.93
0.91
36
71
0.95
0.95 0.95
77
38
0.96
0.93
0.96
39
0.9
0.96
0.95
79
40
0.96
0.91
0.94 0.94
0.96
Statistic
Statistic
Mean
78
80
Daily min
0.91
Daily min
62
63
Daily minimum
temperature R2
value
0.87
64
0.94
66
0.94
0.9
0.9
67
0.92
0.95
68
0.89
0.94
69
0.92
0.86
70
0.86
71
0.92
0.87
0.94
0.87
72
0.93
0.9
73
0.89
74
0.93
75
0.92
0.93
76
0.91
0.96
0.95
0.96
0.93
0.96
0.94
0.93
0.93
0.94
0.96
0.96
0.97
0.97
0.92
0.93
0.97
0.94
0.92
0.93
0.88
0.9
0.94
0.92
0.91
0.93
0.91
0.96
0.94
0.92
0.96
0.92
0.92
0.95
0.93
0.91
0.92
0.89
0.93
0.96
0.92
0.95
0.93
0.92
0.93
0.93
0.98
0.96
0.96
0.93
0.9
0.89
0.92
0.92
0.94
0.93
0.86
0.94
0.96
0.89
0.96
0.92
0.92
0.96
0.93
0.94
0.96
0.96
0.93
0.96
0.96
0.92
0.92
0.92
0.93
0.97
0.94
0.95
0.94
0.95
0.96
Hourly
0.94
Daily
max0.94
0.02
0.02
0.85
0.85
0.90
0.98
0.97
0.94
0.95
0.91
0.93
80
0.93
0.93
0.96
0.88
0.93
Daily max
0.96
0.96
0.96
0.91
79
0.89
0.89
0.96
0.9
0.96
0.92
78
0.96
Hourly
temperature R2
value
0.91
0.94
0.92
77
0.91
0.95
0.96
0.91
Minimum
0.910.95
0.96
0.88
0.02
Maximum
0.9
0.97
0.92
0.92
Standard dev.
Mean
0.94
0.93
65
0.93
Daily maximum
temperature R2
value
0.92
Hourly
0.94
Standard dev.
0.02
0.02
0.02
Minimum
0.85
0.85
0.90
Maximum
0.95
0.98
0.97
Sustainable Land Use and Information Management Section
Department of Primary Industries Parks Water & Environment
SAMPLE SURFACES
The risk of having at least one day
where minimum temperature is less
than -2 degrees for the period
between 15 September to 15 October:
Validation RMSE = 14%
Validation Concordance = 0.81
R2 value= 0.72
Ratio of performance to deviation = 1.79
(i.e. a value above 1.4 indicates a
reasonable model)
Sustainable Land Use and Information Management Section
Department of Primary Industries Parks Water & Environment
SAMPLE SURFACES
Growing degree days:
Validation RMSE = 54.44
Validation Concordance = 0.82
R2 value= 0.66
Ratio of performance to deviation = 1.70
(i.e. a value above 1.4 indicates a
reasonable model)
Sustainable Land Use and Information Management Section
Department of Primary Industries Parks Water & Environment
SAMPLE SURFACES
Chill hours:
Validation RMSE = 36.11
Validation Concordance = 0.92
R2 value= 0.85
Ratio of performance to deviation = 2.44
(i.e. a value above 2 indicates a good
model)
Sustainable Land Use and Information Management Section
Department of Primary Industries Parks Water & Environment
20 ENTERPRISE SUITABILITY SURFACES
CLIMATE
•
•
DSM
•
•
Cubist/ See5 approach working well
with both measured and described
data
Random Forests/ ANN tend to over-fit
(better training fit/ poorer validation
fit)
Regression kriging
produces the most
consistent results
(due to spatial
correlation between
loggers).
Random Forests and
Cubist are also
generating good
results and in some
instances have
produced improved
results compared to
RK.
SUITABILITY SURFACES
SUITABILITY SURFACES
Pyrethrum – Main Limitations
• Frost
• Drainage
• Stone%
Raspberries – Main Limitations
• Frost
• Drainage
• pH
• EC
PARAMETER UNCERTAINTIES
Exchangeable Ca (meq/ 100g) – 0 to15cm
Fuzzy k-means clusters
•
•
Derive upper and lower limits by
determining Mahalnaobis Distance of
each pixel prediction to the Fuzzy kmean cluster centroids of the
covariates used for predictions
The better the prediction validations,
the lower the margin between the limits
Eg. Exch. Ca = 7 +/ 2.1 (Upper limit =
9.1, Lower Limit = 5.9) (meq/100mg)
SUITABILITY PROBABILITY RATINGS
• Assume values between the error bounds
for a soil property are normally distributed
• Use Monte-Carlo simulations to randomly
sample (10,000 times) between the upper
and lower limits (the majority will be around the predicted value due to a
‘normality’ constraint), based on suitability rules
• Tally the number of suitability ratings obtained for each parameter for
each enterprise for each pixel
•
eg. for pH – Obtain 7880 times ‘suitable’, 2100 times ‘marginally suitable’, 10 times
‘unsuitable’………..
• Gives a 79% probability of being suitable for an enterprise based on pH
at that pixel, and 21% probability of being marginally suitable
FINAL OUTPUTS - (UNDER REVIEW)
www.theLIST.tas.gov.au
WEALTH FROM WATER - TASMANIA
FUTURE:
“Enterprise Diversity Index”
(Combined All Suitability Surfaces)
Sustainable Land Use and Information Management Section
Department of Primary Industries Parks Water & Environment
ACKNOWLEDGEMENTS:
Chris Grose1, Rob Moreton1, Mathew Webb1, Zhuo Wang1,
Regan Parkinson1, Rhys Stickler1, Peter Voller1, Ashley
Bastock5,Robin Allchin1, Brendan Malone2, Alex
McBratney2, Budiman Minasny2, Raphael Viscarra Rossel3,
Seija Tuomi3, Peter Wilson3, Bill Cotching4, Leigh Sparrow4,
Rowan Smith4, Fiona Kerslake4, Land Owners of the Study
Areas, CSBP
1 Department
of Primary Industries, Parks, Water & Environment, Tasmania
2 University of Sydney, Faculty of Agriculture and Environment
3 ACLEP/ CSIRO Land & Water
4 Tasmanian Institute of Agriculture (TIA)
5 Irrigation Tasmania
Sustainable Land Use and Information Management Section
Department of Primary Industries Parks Water & Environment

similar documents