geocomputation 2007 ph - mcgill...
TRANSCRIPT
1
Viacheslav Adamchuk Biological Systems EngineeringDavid Marx and April Kerby StatisticsAshok Samal and Leen-Kiat Soh Computer Science and EngineeringRichard Ferguson and Charles Wortmann Agronomy and Horticulture
University of Nebraska-Lincoln(Lincoln, Nebraska, USA)
Guided Soil Sampling for Guided Soil Sampling for Enhanced Analysis of Enhanced Analysis of
GeoreferencedGeoreferenced Sensor Based DataSensor Based Data
GeocomputationGeocomputation 20072007MaynoothMaynooth, Kildare, Ireland, Kildare, Ireland
September 3, 2007September 3, 2007BackgroundBackground
• Assessment of soil variability is essential in site-specific management
• Variable rate application requires accurate information about soil spatial structure
• Obtaining adequate spatial information for a field is expensive using conventional soil sampling and analysis methods
• Accurate mapping of soil attributes requires high density on-the-go sampling
• Recent on-the-go sensors can reveal spatial variation in soils, but improved prescription algorithm are needed
OnOn--thethe--go Soil Sensorsgo Soil Sensors
Electrical and Electromagnetic
Acoustic
Mechanical Electrochemical
H+
H+ H+H+
H+
Pneumatic
Optical and Radiometric
Applicability of OnApplicability of On--thethe--Go Soil SensorsGo Soil Sensors
OKSomeSomeResidual nitrate (total nitrogen)
OKOKCEC (other buffer indicators)
OKSomeOther nutrients (potassium)
GoodSomeSoil pH
SomeOKSomeDepth variability (hard pan)
SomeGoodSoil compaction (bulk density)
SomeOKSoil salinity (sodium)
GoodGoodSoil water (moisture)
GoodSomeSoil organic matter or total carbon
SomeOKGoodSoil texture (clay, silt and sand)
Soil property H+
H+ H+H+
H+
Integrated Soil Physical Properties Integrated Soil Physical Properties Mapping SystemMapping System
Dual wavelength soil reflectance sensor
Soil mechanical resistance profiler with an array of
strain gage bridgesCapacitor-based
sensor
UNL (Lincoln, Nebraska)
Mobil Sensor Platform (MSP)Mobil Sensor Platform (MSP)
Veris Technologies, Inc.(Salina, Kansas)
http://www.veristech.com
EC Surveyor 3150
Soil pH Manager
2
Direct Soil MeasurementDirect Soil MeasurementPurdue University
(West Lafayette, Indiana)Veris Technologies, Inc.
(Salina, Kansas)UNL (Lincoln, Nebraska)
Water Nozzle
Soil Sampler
Ion-selective Electrodes
Soil pH MapsSoil pH Maps
Soil pH Maps of a 23-ha
Kansas Field
Directed Soil Sampling
ObjectiveObjective
• The ultimate goal of our research is to delineate sensor-based field areas that require differentiated soil treatments
• One of intermediate objectives is to develop an algorithm for prescribing guided (targeted) soil sampling
• In this presentation we discuss criteria for evaluating different combinations of guided samples
General RulesGeneral Rules
• Guided samples should be collected from relatively homogeneous field areas away from the field boundary and other transitional areas
• They should cover the entire range of sensor-based measurements, especially toward low and high ends
• Guided samples should be physically spread across the entire field
• It should be possible to process multiple sensor-based data layers
Our CriteriaOur CriteriaHomogeneity
Neighborhood variability
Even data spread
D-optimality
Even field coverage
Spatial predictability
MethodologyMethodology• Property: Soil pH • Instrument: Veris® Mobile Sensor Platform• Field area: 23 ha• Number of valid measurements: 598• Number of guided samples: 10• Different sets of samples considered: 63
– Random selection: 20– Grid cell spread: 19– Even soil pH spread: 20– Maximum homogeneity: 1– Grid cell centers: 1– Subjunctive selection: 2
3
Soil pH HistogramSoil pH Histogram
0
10
20
30
40
50
60
70
80
90
100
5.0
5.1
5.2
5.3
5.4
5.5
5.6
5.7
5.8
5.9
6.0
6.1
6.2
6.3
6.4
6.5
6.6
6.7
6.8
6.9
7.0
7.1
7.2
7.3
7.4
7.5
7.6
7.7
7.8
7.9
8.0
Soil pH
Num
ber o
f poi
nts
Complete Complete RandomizationRandomization
Grid Cell Grid Cell SpreadSpread
Even Soil pH Even Soil pH SpreadSpread
Maximum Maximum HomogeneityHomogeneity
Grid Cell Grid Cell CentersCenters
4
Subjunctive Subjunctive SelectionSelection
ZZ--Score ComparisonScore Comparison
-4
-2
0
2
4
6
8
Homogeneity criterion D-optimality criterion Field spread criterion Total score
z-sc
ore
Random selectionGrid cell spreadEven soil pH spreadMaximum homogeneityGrid cell centersSubjunctive selection
Summary and Future WorkSummary and Future Work• Satisfactory optimization of the three diverse
criteria is feasible• Certain objective field parameters should be
used to set the weight of each criteria• The number of guided points should be a part
of optimization process• It will be challenging to generate an algorithm
dealing with multiple data layers obtained with inconsistent densities
• Any comments and recommendations are welcome
http://bse.unl.edu/adamchukE:mail: [email protected]