HR: 08:50h
AN: H51N-04 INVITED [Abstracts]
TI: Realtime Delivery of Alarms and Key Observables in a Deployed Hydrological Sensor Network
AU: * Marshall, I W
EM: i.w.marshall@lancaster.ac.uk
AF: Lancaster Environment Centre, Lancaster University, Lancaster, LA1 4YQ, United Kingdom
AU: Price, M C
EM: pricemc@comp.lancs.ac.uk
AF: Infolab21 (Computer Science), Lancaster University, Lancaster, LA1 4YQ, United Kingdom
AU: Li, H
EM: h.li@lancaster.ac.uk
AF: Infolab21 (Computer Science), Lancaster University, Lancaster, LA1 4YQ, United Kingdom
AU: Boyd, N
EM: n.boyd@salamander-group.co.uk
AF: Salamander Group, Williams House
Manchester Science Park, Manchester, M15 6SE, United Kingdom
AU: Boult, S
EM: s.boult@manchester.ac.uk
AF: School of Earth, Atmospheric and Environmental Sciences, Manchester University,
Manchester, M60 1QD, United Kingdom
AB:
It has widely [1-3] been proposed that sensor networks are a good solution for environmental monitoring.
However, this application presents a number of major challenges for current technology. In particular
environmental science involves the study of coupled non-equilibrium dynamic processes that generate time
series with non-stationary means and strongly dependent variables and which operate in the presence of large
amounts of noise/interference (thermal, chemical and biological) and multiple quasi-periodic forcing factors
(diurnal cycles, tides, etc). This typically means that any analysis must be based on large data samples obtained
at multiple scales of space and time. In addition the areas of interest are large, relatively inaccessible and
typically extremely hostile to electronic instrumentation. Our analysis of these factors has encouraged us to focus
on this list of generic requirements;
a) Node lifetime (between visits) should be 1 yr or greater
b) Communication range should be ~250m
c) Nodes should be portable, unobtrusive, low cost, etc.
d) Networks are expected to be sparse since areas of interest are large and budgets are small
However, the characteristics of each environment, the dominant processes operating in it and the measurements
that are of interest are sufficiently different that the design of an appropriate sensor network solution is normally
most determined by site specific constraints. Most importantly the opportunities for exploiting contextual
correlation to disambiguate observations and improve the maintenance and robustness of a deployed sensor
network are always site specific. We will describe the design and initial deployment of a hydrological sensor
network we are developing to assess the hydro-dynamics of surface water drainage into Great Crowden Brook in
the Peak District (UK). The complete network will observe soil moisture, temperature and rainfall on a number of
transects across the valley, and will also investigate water quality parameters (colouration, turbidity, Ph) in the
stream. GSM access for remote real time reporting of network status is only available from the hilltops so a
multihop communication strategy is being used for communication from the valley floor. To minimise radio
usage and maximise battery life we are reporting only those alarms and events that are judged to be of high
priority by embedding a simple rule based decision engine in each node. The rule conditions are derived from
spatio-temporal cross-correlation of the available sensor inputs. . We report on our initial experiments with
correlating readings for management purposes, and offer some initial hypotheses regarding aspects of this that
might be generic based on a comparison with data obtained in an earlier experiment in a marine setting [4].
1. http://cens.ucla.edu/
2. http://eyes.eu.org/
3. http://www.ee.unimelb.edu.au/ISSNIP/
4. J. Tateson, C. Roadknight, A. Gonzalez, T. Khan, S. Fitz, I. Henning, N. Boyd, C. Vincent, and I. W. Marshall. Real
World Issues in Deploying a Wireless Sensor Network. In Workshop on Real-World Wireless Sensor Networks
REALWSN'05, Stockholm, Sweden, June 2005
DE: 0555 Neural networks, fuzzy logic, machine learning
DE: 1804 Catchment
DE: 1848 Monitoring networks
DE: 1895 Instruments and techniques: monitoring
SC: Hydrology [H]
MN: 2007 Fall Meeting