Deployment and Maintenance of Groundwater Sensor Networks

1
Deployment and Maintenance of Groundwater Sensor
Networks
Rick Johnson OGI School of Science
Engineering Oregon Health Science
University Portland, OR
30 November 2004
2
Essential elements for deployment and
maintenance of a sensor network
Detect parameter of interest with sufficient
specificity and sensitivity Stable over months to
years Provide time series data at minute to hour
intervals Economical Easily calibrated (or
requiring no field calibration) Easily networked
(e.g., electrical signals that can be
transmitted)
3
Essential elements for deployment and
maintenance of a sensor network
Detect parameter of interest with sufficient
specificity and sensitivity Stable over months to
years Provide time series data at minute to hour
intervals Economical Easily calibrated (or
requiring no field calibration) Easily networked
(e.g., electrical signals that can be
transmitted) At present we can easily do this
for pressure, temperature, conductance Not quite
so easily with geochemical parameters (pH,
ORP) In some cases with specific ion
electrodes Perhaps with a few other sensors Not
for most of the array of contaminant molecules of
interest
4
Here is a schematic drawing of the infiltration
gallery example I mentioned at the beginning of
the talk. I might argue that this case could
justify a sensor network because we need to
understand changes in water quality over years,
but on the time scale of minutes to hours
Infiltration Gallery
River
Public Water Supply
5
In this setting we might be interested in a wide
range of potential contaminants which may have
complex time/concentration histories These would
probably include pesticides, hormones,
pharmaceuticals, pathogens
Infiltration Gallery
River
Public Water Supply
Sensors
6
We need to know what the contaminant
concentrations are in the river and in the
influent to the infiltration system on a real
time basis in order to make decisions about when
to re-infiltrate water. We also need to know if
the contaminants propagate through the subsurface
to the water supply well.
Infiltration Gallery
River
Public Water Supply
Sensors
7
To make this happen we would probably
need Sub-nanogram-per-liter detection Real-time,
continuous detection Molecular
specificity Detection of a wide range of
compounds A networked system (probably coupled to
an automatic decision-making network)
Infiltration Gallery Control
Public Water Supply
8
Let me focus for a minute on one class of sensors
that has the POTENTIAL to meet all of those
criteria INTERDIGITATED MICROSENSOR ELECTRODES
9
These sensors can use immobilized co-factors,
enzymes, antibodies, enzyme-antibody conjugates,
stabilized receptors, and DNA fragments to gain
molecular specificity. They can be probed
electronically to detect the presence of target
molecules associated with the immobilized
structure of the electrode They can potentially
be configured as micro-arrays to provide
detection of a number of different molecules on a
single chip
10
But, practical examples of these devices is still
a ways off (i.e., years).
For the foreseeable future, we will be need
hybrid networks that can couple available sensors
with ancillary data to provide real-time
measurements of parameters of interest.
11
In the earlier example we wanted to determine
pesticides, hormones, pharmaceuticals and
pathogens in real time. We dont have real-time
sensors for any of them, so we need to use the
information we do have (river flow, turbidity,
conductance, precipitation, time of year,
groundwater geochemistry), together with discrete
collected samples analyzed for the target
compounds. From these we can develop
correlation databases. Over time, and as new
sensors become available, the system can become
smarter
Infiltration Gallery
River
Public Water Supply
12
A real-world example of this approach comes from
the Equus infiltration gallery in south central
Kansas. The graph shows the correlation
between measured and predicted atrazine
concentrations (along with the correlation
equation). In the absence of a atrazine
-specific real time detector, they used month of
the year, flow (pressure) and specific
conductance to estimate atrazine. I think we can
currently be a lot more sophisticated than this
example.
U.S. Geological SurveyWater-Resources
Investigations Report 00-4126
13

• Summary
• Deployment and maintenance of groundwater sensor
  networks for some parameters is straightforward.
• For many parameters of interest we do not have
  real-time sensors that are stable over the
  timeframes we are interested in for groundwater
  systems.
• New sensors will clearly come on line over the
  next decade.
• However, we need a strategy that allows us to
  start making decisions in the near term, while
  allowing the new sensors to be integrated into
  our networks
• Data fusion, correlation analysis, time-series
  modeling, and other approaches will be important
  in making those networks deployable and
  maintainable.

14
(No Transcript)
15
When should we deploy groundwater sensor networks?
From the Center for Embedded Networked Sensing

• At the risk of being controversial, I would argue
  that this may not be the best application
• (admittedly, this is a cartoon, but allow me to
  use it as an example)
• The network is perhaps more dense than necessary
• The timeframe may be too long (calibration and
  stability are issues)
• Not sufficient impact to justify expense