harnessing a real time sensor network for illicit and ...network for illicit and accidental ......
TRANSCRIPT
Harnessing a Real Time Sensor Network for Illicit and Accidental
Discharge Detection: Case Study Using Charlotte-Mecklenburg’s Sensor Network and the R Programming
Language
Caroline Burgett and Joel Nipper
!=
SSO Historical Frequency
Yearly SSOs vs 5-Year Moving Average
Available dataMonitoring Networks
USGS Rain Gages
USGS Stage and
YSI WQ Sondes
Webcams
Flow Gages
Historical Workflows
• Work with vendor for online visualizations (available to public)
• Threshold alerts
• Manual dailyish data review
• (Recently) increased interactivity with visualizations
• (More recently) have vendor add USGS flow data
AssessIF NO RAIN AND
TURBIDITY SPIKETHEN
SEND ALERT
Hydromet Cloud(Web Host)
dataRetrieval
RSelenium
USGS-NWIS
NTS
Approach Overview
riem
METAR
Do Nothing
taskscheduleR
gmailr
ELSE DO NOTHING
First Implementations(Flow / Stage Only)First Implementations(Flow / Stage Only)
First Implementations (Hydro. andWQ Sonde)
leaflet
plotly
shiny
Next Level Analysis:Looking For Signatures
Approach: SSO Signatures
Approach: SSO Signatures
Approach: SSO Signatures
Approach: SSO Signatures
Approach: DW Signatures
Results / Lessons Learned
• Don’t dismiss sensor anomalies as malfunctions or natural conditions
• Automate
• Site Selection
• Refinement of signatures, widespread deployment
• Migration of WQ sonde to Contrail® (OneRain Inc)
• Wet weather alerts
Next Steps
• dataRetrieval, RSelenium, shiny, ggplot,
dplyr, leaflet, riem, maptools, plotly, gmailr
• and RStudio
Acknowledgements
• Colleagues, Program Management
• USGS