Sensor network data integrity: multilevel calibration

1
Sensor network data integrity multilevel
calibration
Pilot-scale in situ
Lab-scale
beaker

• Bench-top calibration
• Pilot deployment in situ calibration
• calibrate models early in the deployment
• recalibrate sensors using model later (as sensors
  become less sharp)
• Incorporate uncertainty into the analyses

2
In situ calibration of individual sensors

• How does the sensor response change in the
  embedded position?
• Absolute position, response relative to potential
  obstructions, flow diversions is difficult to
  control

3
In situ calibration of individual sensors (contd)

• Concept
• Investigator delivered solution pulse
• Vary solution concentration
• Quantitative (peak ht., area, variance)
• Compare response over time for insight into
  sensor degradation
• Challenges
• Delivery system must be reproducible in the field
• And must stay that way over extended time periods
• Longer-term challenge
• Autonomous delivery, analysis

4
Deterministic sensor network calibration

• The problem is really multi-dimensional
• Controlled perturbations (e.g., controlled
  irrigation event)
• Iterative calibration of distributed parameter
  soil flow and transport models

Accounting for multiple contributions (as from
various flow vectors)
5
Coupling simulations to sensor network calibration

• At the field scale
• rigorous conventional characterization sampling
  plans still required
• for soils, geostatistical parameterization
  techniques may be needed
• Report network information and associated
  uncertainty
• Iteration between models and sensor perpetually
  to decrease uncertainty

indicator kriging (probability Ks
exceeds...)
kriging (interpolating Ks)
6
Data integrity in sensor networks multilevel
calibration

• Bench-top calibration
• Pilot deployment
• develop in situ calibration protocol
• characterize longevity, degradation
• Early in the deployment
• Take advantage of the sensors integrity
• Calibrate model (distributed parameters)
• Integrate DAQ with simulator to accelerate
  process
• Later (as sensors become suspect)
• Reverse the process
• Let the network identity bad sensors
• Incorporate uncertainty into the process