ModelBased Monitoring for Early Warning Flood Detection

1
Model-Based Monitoring for Early Warning Flood
Detection

• Elizabeth A. Basha, Computer Science and
  Artificial Intelligence Laboratory, Massachusetts
  Institute of Technology
• Daniela Rus, Computer Science and Artificial
  Intelligence Laboratory, Massachusetts Institute
  of Technology
• Sai Ravela, Earth Atmospheric and Planetary
  Science Massachusetts Institute of Technology

2
Outline

• Motivation
• Previous Work
• Prediction Model
• Sensor Network Architecture
• Installation and Results
• Conclusion
• ProsCons

3
Motivation

• River flooding detection
• Deployment target rural and developing countries
• Requirements
• Withstanding hardware to river flooding and
  storms
• Monitor and communicate over 10000km2 basin
• Predict flooding autonomously
• Limit costs allowing feasible implementation in
  development country

4
Introduction

• Flood Prediction Algorithm is based on a
  regression model.
• Nearly as good as that used by hydrology
  researchers

5
Previous work (1/2)

• Sensor network for environmental monitoring
• Redwood tree (air temperature, humidity, solar
  radiation).
• Off-line data analysis
• Light intensity
• Communication via Zigbee
• James reserve (humidity, rain, wind)
• Deployment in Bangladesh rice paddy to measure
  nitrate, calcium and phosphate
• Volcano
• Seismic and acoustic data

6
Previous work (2/2)

• None above envision system requirements
• Minimalistic number of sensor available
• Real-time need of data
• Computational autonomy
• Complexity necessary to perform prediction

7
Sensor networks for flood detection

• Castillo-Effen
• Suggested an architecture but unclear on basin
  characteristics and no hardware detail
• Hughes
• Gumstix sensor nodes, linux OS
• Tested in the lab but no field test
• Planned deployment of 13 nodes along 1km
  riverside without flood prediction model.

8
Operational systems for flood detection

• US Emergency Alert System
• Volunteer and limited technology
• MIKE 11-based flood forecasting system

9
Computational requirements

• SAC-SMA
• Modeling different methods of rainfall surface
  runoff to determine how much water will enter the
  river
• Complex equations to establish the model
• Not easily running on sensor network

10
Prediction Model

• Rainfall-runoff model
• Computational burden, difficult to customized for
  individual basin
• Statistic model
• Based on observed records
• Intrinsically self-calibrated, real-time
• Used in other areas such as hurricane intensity
  forecasting
• Linear regression models assume a linear equation
  can describe system behavior
• Weighting the past N records of relevant inputs
  at time T to produce prediction at Tt
• Past prediction errors are allowed

11
Flood prediction algorithm
12
Test data and setup

• Use 7 years of rainfall, temperature and river
  flow data for Blue River in Oklahoma
• Compare data to DMPI
• 3 criteria for the quality of algorithm
• Modified correlation coefficient
• False positive and negative

13
Model Calibration

• Training window 1/3/6/9/12 months
• Optimal values of inputs Sweep the order for
  each input of past prediction
• Pick the best input values with high MCC and low
  false positive/negative
• Other approaches climatology, persistence
• 1/24 hours prediction

14
Sensor network architecture (1/2)

• Monitor events over large geographic regions of
  10000 km2
• Provide real-time communication of measurements
  covering a wide variety of variables contributing
  to the event occurrence
• Survive long-term element exposure, the potential
  devastating event of interest, and minimal
  maintenance
• Recover from node losses
• Minimize costs
• Predict the event of interest using a distributed
  model driven by data collected
• Distribute among nodes the significant
  computation needed for the prediction

15
Sensor network architecture (2/2)

• 2-tier communication network
• Long-range communication node transmits on the
  order of 25 km using 144 MHz radio
• Low power sensing node operates at 900 MHz
• Office and communication nodes for UI

16
Base system

• Base system
• ARM7TDMI-S microcontroller core for LPC2148 from
  NXP
• Using photovoltaic charging of lithium-polymer
  battery at 3.7V

17
Base system hardware

• An ARM7TDMI-S microcontroller core
• Extend to 8 serial ports by adding Xilinx
  CoolRunner-II CPDL
• Mini-SD card and FRAM supply data and
  configuration storage
• Running software package developed in C using
  WinARM

18
Communication

• AC4790 900MHz modules operate at 76.5 kb/s
• Modem uses MX614 Bell 202 integrated circuit

19
Sensing node

• Measuring rainfall, air temperature, water
  pressure
• Log data
• Compute data statistic over each hour
• Analyze data for potential sensor failures

20
Communication node

• Computation of prediction
• Maintain a record of all values and examine data
  correction
• Request data if encountering prediction model
  uncertainty

21
Office and community node

• Maintained by governmental agencies
• Display malfunctioning nodes
• Provide data and prediction regarding the event
  of interests
• Community nodes provide a simpler UI and do not
  supply detailed information regarding node status
  and private data

22
Installation and results

• Test the flood prediction algorithm
• using a large set of physical river flow data
• Demonstrate long-term data collection of river
  flow data with a sensor network
• Test the networking capabilities of 2-tier sensor
  network in a rural setting

23
Blue River testing

• Use a large data set to test prediction algorithm
• 7 years of data measured from 1 river flow and 6
  rainfall sensors and a weather station
• Autocorrelation at 24 hours 0.627

24
Blue River testing
25
Dover field test

• 5 weeks data
• collection

26
Honduras field tests

• Collaboration with FSAR to install the system and
  understand deployment issues
• Radio antennas need line-of-sight high in the air
• Possible water measuring
• system

27
Conclusion

• Described a complete architecture for predictive
  environmental sensor networks over large
  geographic areas
• Nodes-limited and cost constraints
• Implementation of flood prediction algorithm and
  evaluation

28
ProsCons

• Pros
• A complete study
• Use off-the-shelf devices
• Detailed hardware description
• Cons
• No real flooding occurred during evaluation
• Energy consumption problem