Wireless Sensor and SensorNet Applications

1
Wireless Sensor and SensorNet Applications for
Forest Ecological Research
Todd Dawson and collaborators, U.C. Berkeley and
Intel Berkeley
2
Wireless networks for the analysis of forest
function a program of design, measurement and
modeling in forest biogeosciences

• PREMISE
• Forested ecosystems play key roles in the
  dynamics of Earth systems
• Modulating global, regional and local climate
• Modulating the carbon, nitrogen, trace-gas,
  water, and energy cycles

The biogeochemical cycles of forests are in turn
sensitive to physical and chemical forcing and
the growing influence of human activities
3
Characterizing the modulating influences that
forests have on climate and biogeochemical cycles
requires that we better understand the impacts
that the trees themselves are having on forested
ecosystem characteristics

• Our focus has been on the role of trees in the
  hydro-
• logical cycle and how best to view their impacts

• 60-70 of H2O cycles through trees, not off of
  the ground (evaporation)
• We can usually only observe the top or bottom
  surfaces of the forest
• We need to understand what happens within the
  forest and what feedbacks exist between trees and
  atmosphere

4
From Van Pelt and Sillett (2007) drawings by R.
Van Pelt
Yes, people to scale
BUT . . . . . The vertical structure of forest
trees is usually very complex
How do we adequately characterize the
microclimatic complexity and tree response to it?
5
From Koch et al. 2004
Variation in needle characteristics in ONE tree
crown
6
From Burgess Dawson (2007)
Based on wired static sensor deployments
Redwood
Eucalyptus
Light extinction is NOT uniform across tree
species or forest types (e.g., young vs. old,
natural vs. managed)
7
From Sillett and Van Pelt (2008)
Redwood
Based on event sampling
Light extinction is NOT uniform across a suite of
different methods either
8
SOLUTIONDense Self-Organized, WirelessMultihop
Network
9
Wireless Micro-weather Mote

• Incident Light Sensors
• TAOS total solar
• Hamamatsu PAR
• Mica2 dot mote
• Power board
• Power supply
• SAFT LiS02 battery, 1 Ah _at_ 2.8V
• Weatherproof Packaging
• HDPE tube with coated sensor boards on both ends
  of the tube
• O-ring seal for two water flows
• Additional PVC skirt to provide extra shade and
  protection against the rain
• Radiant Light Sensors
• PAR and Total Solar
• Environmental Sensors
• Sensirion humidity temp
• Intersema Pressure temp

mote
10
Custom Sensor Subsystem
Top Sensing Surface(incident PAR and TSR)
11
Custom Mechanical Design Power Subsystem
Weatherproofthe package while carefully exposing
the sensors
Top endcap
Battery
Shielding skirt
O-rings
Choose the power source according to the
deployment envelope and needs of the researchers
Cylindrical enclosure
Bottom endcap
12
(No Transcript)
13
From Tolle et al. 2005
14
Redwood sensor network

• All units were placed on a rack, calibrated in
  growth chambers and out-of-doors and then
  deployed in various parts of tree canopies

15
Redwood sensornetwork
From Dawson et al., in review
16
Install array between trees too
17
Red line a month
From Tolle et al. 2005
18
After 45 days with a 55 node deployment we had
49 usable data (gt1,700,000 points)
From Tolle et al. 2005
19
Light used by leaves Photosynthetically Active
Radiation
Evaporative demand from leaves Vapor Pressure
Difference
Temperature around leaves
From Dawson et al., in review
20
Temperature
Rel. Humidity
Height Within Tree (m)
Incident Light
From Tolle et al. 2005
21
5.2C range
Temperature
Temperature
53 range
Relative Humidity
Rel. Humidity
98 range
Incident Light
Incident Light
Off to the movies
From Tolle et al. 2005
22
NEXT Biological monitoring - water useCosts and
logistics in monitoring tree water balance, water
use and ecosystem function

• Sap flow sensors to measure plant transpiration
  are relatively cheap(120 each)

• Sensor wires (10 m) add about 50-70 each

23
Biological monitoring water use

• Connecting multiple sensors together
  (multiplexing) costs 150 per sensor

• More wires are needed to connect multiplexers to
  data loggers e.g. 2.6 km of wirefor a 12 tree
  set-up

24
Biological monitoring water use

• Data loggers cost 2,200 or up to 150 per
  sensor

• A portable computer (2,000) must be physically
  transported to the site (travel ) for routine
  data collection

25
Biological monitoring water use

• These large systems are also power hungry
  (requiring a 500 power system!)

26
Alternative Biological MonitoringWireless
Tree-Sensor Networks

• Use the same sensor technology

• Connect to miniaturised wireless data loggers
  (developed by University of California and Intel
  Research, Berkeley)

27
Wireless sapflow
mote
probe-set
micromet
team
radio
28
Heat Ratio Method
Because of its symmetric configuration, the HRM
can resolve zero flows as well as reverse flows
Zero sap flow
Diagram by Al Grigg
29
Heat Ratio Method
Flow velocity (V) is logarithmically related to
the ratio of temperature increases up- and
down-stream from a heater
V thermal diffusivity x Ln T1
probe distance T2
Active sap flow
Active sap flow
Diagram by Al Grigg
30
Wireless sapflow
From Burgess, Dawson and Kranz, in press
31
Wirelesssapflow
From Burgess Dawson (2008)
Sequoia sempervirens August 2006
Drawing R. Van Pelt
Capacitance or aspect?
32
Other Wireless MOTE sensor packaging
Soil Moisture Wood Moisture Leaf
Wetness Growth Bands Soil CO2 -
Vaisala Vegetation Indicies NDVI, PRI, WBI,
Hyperspectral Soil Nitrate
Prototypes exists as of 12/07
33
Beyond the footprint emerging technologies and
FluxNetworks
FluxNetworks have been successfully
established to both monitor and extend our
understanding of what controls ecosystem CO2 and
H2O exchange across diverse systems,
Sites in these networks are nonetheless
challenged to extrapolate beyond the
footprint, Yet . . . new technologies do exist
that possess the power to extend the footprint
while also helping to validate existing
measurements.
34
Merging Wireless with Tower Measurements
35
The Good News
Merging Wireless with Tower Measurements
Spatial variability in surface fluxes may scale
with the spatial variability in surface
properties that can be assessed with the
wireless motes . . . and then via remote sensing
36
Merging Wireless with Tower Measurements
Tonzi Ranch Site
OAK SAVANNA - Seasonally dry Mediterranean
ecosystem winter active grass-dominated
understory summer active tree (Quercus)
canopy overstory Seven years of eddy
flux, ecophysiological data and stable isotope
research since 2004
37
Spatial Heterogeneity at Tonzi Oak Savanna,Ione,
CA
Prevailing wind
38
Merging Wireless, Tower and Satellite information
Use motes to scale and validate by relating
spatial variability within tower footprint to
satellite pixel
Scaling Oak Savanna Information to MODIS grid
Analyst Qinhua Guo/J Kim
Coefficient of Variation
Distance (m)
39
HydroWatch Understanding the LifeCycle of
Water - A geo-hydro-eco-atmospheric
program - Using Wireless Watersheds
Angelo Reserve Watershed Eel River, Northern
California
40
HydroWatch Angelo Reserve Wireless
Infrastructure
Cahto Peak Wired Access Point
BOB Relay
FORB Relay
COB Relay
Wilderness Lodge VOIP, WiFi WAP
Uber Relay
Science Center Data server AC power WiFi WAP
REB Relay
Gothic Pool Robotic Camera
41
Rivendell test watershed 1 -smaller scale
starting points -geo / hydro / eco / atmo /
engineering expertise all represented -model
and GIS framework in place
-Measuring an entire watershed is
impossible -Well positioned sensor arrays
composed of the relevant sensors provide a
solution -When constituents cannot be sensed
directly use relevant surrogates -Integration of
sensed information with GIS and modeling is
critical to scaling-up -Teams of those who know
and those who can is essential to success
42
Rivendell test watershed 1
Tree node (sapflow and vertical microclimate)
Deep well
Soil moisture nodes
Wireless backbone node
43
Rivendell test watershed 1
Tree node (sapflow and vertical microclimate)
Deep well
Soil moisture nodes
Wireless backbone node
44
Opportunities, needs and challenges
? We now have the power and opportunity to
characterize environments and the biological
response to them in ecologically meaningful
ways ? We are on the cusp of being able to ask
and answer questions we never could before ?
But, we need the scientists who know the relevant
questions to be asking to be in leader- ship
roles and guide the sensor choices and deployment
strategies - teamwork among the engineering,
biological, computing, modeling and statistical
experts is essential ? Simple is better, but
must also be robust and relevant to the questions
posed ? New sensors or surrogates are needed ?
Statistical algorithms connected to the network
design is essential ? Error analysis, sensor
yield, network integrity, gap filling procedures
must be in place ? Data fusion strategies with
remotely sensed, tower-based and/or modeling will
extend our knowledge beyond what we do not or
cannot measure wirelessly
45
Concluding remarks

• ? Wireless technologies are permitting the
  characterization of micro-environments and the
• biological responses to them in VERY NEW ways . .
  . ways that are important to the organisms,
• environments and/or questions we need answered
• ? Wireless technology is allowing us to learn
  things we never knew or could before
• ------------ DO NOT UNDERESTIMATE THE POWER AND
  UTILITY OF WIRELESS ------------
• ? Teamwork among the natural / life scientists
  and the engineering, computing (IT), modeling
• and statistical communities is essential
• ? New sensors or surrogates to what cannot be
  sensed as well as new seneornet designs
• and deployments must be an ongoing development
  activity
• ? Data acquisition and statistical algorithm
  development and implementation should be
• an ongoing activity matched to evolving research
  needs and questions
• ? Analysis, modeling and visualization activities
  will help extend what sensors measure

46
Thanks to David Culler Steve Burgess and
funding from
W. M. KECK FOUNDATION