The National Weather Sensor Grid
Hock Beng Lim, Keck Voon Ling, Wenqiang Wang, Yuxia Yao, Mudasser Iqbal,
Boyang Li, Xiaonan Yin, Tarun Sharma
Intelligent Systems Centre
Nanyang Technological University
Categories and Subject Descriptors
C.2.4 [Distributed Systems]: Distributed applications
General Terms
Design, Experimentation, Measurement
Sensor Networks, Grid Computing, Sensor Grid, Sensor
Data Management, Data Visualization, Web Integration
1 Introduction
With the rapid advances in technologies such as MEMS
sensors, low-power embedded processing and wireless net-
working, sensor networks are becoming more powerful in
terms of data acquisition and processing capabilities. Sensor
networks can now be deployed in the physical world for var-
ious important applications such as environmental monitor-
ing, weather monitoring and modeling, military surveillance,
healthcare monitoring, tracking of goods and manufacturing
processes, smart homes and offices, etc.
The field of sensor networks has grown dramatically in
recent years. However, it remains a daunting challenge to
deploy large-scale sensor networks. It is also difficult to in-
tegrate sensor networks with existing IT infrastructures such
as the Internet. Thus, sensor networks often operate as sepa-
rate information silos, and the sensor resources and data can-
not be easily shared. In fact, with a large number of sensor
devices potentially deployed over a wide area, sensor net-
works are important distributed computing resources that can
be shared by different users and applications.
Grid computing is an established standards-based ap-
proach to solve large-scale problems through coordinated
sharing of distributed and heterogeneous resources in dy-
namic virtual organizations. Most existing developments in
grid computing focus on compute grids, which provide dis-
tributed computational resources for compute-intensive ap-
plications, and data grids, which provide seamless access to
large amounts of distributed data and storage resources.
This work is supported in part by Microsoft Research under
the SensorMap Request for Proposals (RFP) 2007.
Copyright is held by the author/owner(s).
November 6–9, 2007, Sydney, Australia.
ACM 1-59593-763-6/07/0011
Most recently, the concept of sensor grids [1] has received
increasing attention from the research community. Sensor
grids extend the grid computing paradigm to the sharing of
sensor resources in sensor networks. A sensor grid inte-
grates sensor networks with the computational and storage
resources in the conventional grid fabric. The vast amount
of data collected by the sensors can be stored, processed and
analyzed by the computational and data storage resources of
the grid. Sensor resources can be efficiently shared by differ-
ent users and applications through the resource sharing and
coordination capabilities of the grid.
The National Weather Study Project (NWSP) is a large-
scale community-based environmental initiative in Singa-
pore that aims to promote the awareness about weather pat-
terns, climate change, global warming and the environment.
In this project, hundreds of mini weather stations are de-
ployed in schools throughout Singapore. Each weather sta-
tion contains several sensors for measuring weather param-
eters such as temperature, rainfall, humidity, wind speed
and direction, etc. Since the geographical locations of these
school weather stations cover most parts of Singapore, the
microclimate weather data from these stations provide a
good profile of the weather patterns within Singapore.
We are designing and building the National Weather Sen-
sor Grid (NWSG) to support and to realize the full poten-
tial of the NWSP. The NWSG has several important fea-
tures. First, it connects the weather stations via the Inter-
net to automatically collect and aggregate weather data in
real-time. Second, the weather data are logically stored in a
Central Data Depository (CDD) which can be implemented
using distributed data storage resources. Third, the NWSG
integrates computational resources for the compute-intensive
processing of weather data. Fourth, the weather data can be
easily accessible and shared via the web through mash-ups,
blogs, and other user applications. We are developing tech-
niques and tools to efficiently publish, query, process, visu-
alize, archive and search the vast amount of weather data.
Finally, the NWSG should be scalable to handle hundreds
of weather stations, and also extensible to handle different
types of sensors besides weather stations.
At present, we have developed a prototype of the NWSG
with several connected weather stations. This prototype en-
ables us to improve the design of the sensor grid architecture.
It also provides several useful services for the users to access
and visualize the weather data.
Figure 1. Sensor Grid Architecture
2 System Architecture
Figure 1 shows the system architecture of the National
Weather Sensor Grid (NWSG). It consists of several major
Sensor Resource. In the NWSG, the sensor resources are
different types of weather stations. There are two types of
weather stations deployed in this project; namely, the Davis
Vantage Pro II and the WeatherHawk. Each weather sta-
tion integrates several sensors that directly measure weather
parameters like temperature, rainfall, humidity, wind speed
and direction, solar radiation, barometric pressure, etc. A
weather station or a group of weather stations are connected
to the sensor grid through a proxy system, which is currently
implemented as a desktop or laptop PC running a set of mid-
dleware services. The proxy may be wired or wireless.
The proxy will automatically retrieve weather data from
the weather stations and send the data to the sensor grid via
the Internet. The proxy performs some data processing tasks.
For example, it computes several indirect weather parame-
ters such as heat index and dew point using the direct mea-
surable weather parameters. It is also possible for the proxy
to control the operation of the weather station.
Computational/Data Storage Resource. A computa-
tional/data storage resource refers to any resource that per-
forms computational and/or data management tasks. These
resources are capable of efficiently handling large-scale
computational or data management jobs. They are man-
aged by middleware services that we are developing based
on standard grid protocols such as Globus.
In our current prototype, these resources consist of a sin-
gle database and several grid resource nodes. Although the
single database can handle the weather data for now, we plan
to extend it to a distributed database for better scalability as
more weather stations are connected to our system. With
the microclimate weather data collected throughout Singa-
pore, we can develop weather models and perform sophisti-
cated computations. The computational resources are there-
fore essential to perform such computational-intensive jobs
with relatively low cost.
Service Provider. The term service is used in the grid com-
puting context. Apart from conventional computational and
data processing jobs, the sensor grid can provide some useful
services to the users. The service provider makes use of the
weather stations, computational resources, and data storage
resources to provide the services. The users can subscribe to
these services through standard grid computing mechanisms.
Currently, we provide rain and temperature alert services
to users. Once a user has subscribed to these services, the
user will be notified if it is raining or if the temperature is
too high in certain locations. We also provide a weather dis-
play service which displays the weather conditions at user
specified locations in a flexible manner on web sites, blogs,
or other user applications.
3 Demonstration Highlights
We will demonstrate several features of our National
Weather Sensor Grid (NWSG) prototype. First, we will show
the distribution of the connected weather stations in Singa-
pore. We currently use two geo-centric web interfaces to
display and visualize the weather station information.
The first interface is Microsoft SensorMap [2], which is
specially designed for sensor information publishing. We
publish the weather station information using SensorMap’s
supported sensor types, and graphically display the weather
data using dynamically generated HTML pages. The sec-
ond interface is Google Earth. We generate kml script which
overlays the weather stations on the satellite map of Sin-
gapore, together with dynamically updated snapshots of the
weather data last captured from the weather stations.
We will demonstrate how a user submits a request for a
particular type of computational or data processing job, and
how the sensor grid completes the job and returns the results
to the user. Finally, we will show the weather alert services
and how users can subscribe to these services on the web.
4 References
[1] H. B. Lim, Y. M. Teo, P. Mukherjee, V. T. Lam, W. F.
Wong, and S. See, “Sensor grid: Integration of wire-
less sensor networks and the grid, Proc. of the 30th
IEEE Conference on Local Computer Networks (LCN
2005), pp. 91-98, Nov 2005.
[2] S. Nath, J. Liu, F. Zhao, “SensorMap for Wide-Area
Sensor Webs, IEEE Computer, Vol. 40, No. 7, pp.
90-93, Jul 2007.