KIT – University of the State of Baden-Wuerttemberg and
National Research Center of the Helmholtz Association
•
Challenges
:
• Detect rain events in the incoming
raw CML data in real-time
• Robustly distinguish data with
rain-induced fluctuations
from
data with
artifacts
•
Status
• adaptive rolling statistics work
okay, except for some artifacts
•
Challenges
:
• Build and run a dynamic web-visualization
• Include live radar data
• Legal aspects of showing results on
zoomable maps on the web
•
Status
• We have a working CML-vs-radar rainfall
map web-viz
• The web-viz is built using bokeh-server
•
Outlook
• 5-minute real-time rainfall CML-rainfall
• Real-time merging with DWD radar data
Real-time acquisition and processing of data from a country-wide network of
commercial microwave links in Germany: Current status and challenges
Christian Chwala (1,2), Gerhard Smiatek (1), Maximilian Graf (1), Julius Polz (1), Harald Kunstmann (1,2)
(1) Karlsruhe Institute of Technology (IMK-IFU), Institute of Meteorology and Climate Research, Garmisch-Partenkirchen,Germany (christian.chwala@kit.edu),
(2) Institute for Geography, Regional Climate and Hydrology, University of Augsburg,Augsburg, Germany
Acknowledgment: This work is funded via the DFG project IMAP and the BMBF project HoWa-innovativ
• Rainfall observations by rain gauges and weather
radars are prone to errors
• Attenuation data from commercial microwave link
(CMLs) networks provides additional rainfall information
• CMLs are used to provide a large part of the backhaul
of cellular networks and hence are available almost
everywhere where cellular networks exist
• We compare hourly
rainfall of 4000 CMLs vs
RADOLAN for one year
• Performance is good for
the majority of CMLs
• Solid- or mixed-precip
periods are problematic
Data acquisition
Real-time data processing
Visualization and delivery
Motivation
Long-term offline performance
Real-time online CML-derived rainfall
•
Challenges
:
• Get access to data
• Continuously acquire and forward data
• Robustly handle failures
•
Status
• We operate a open-source real-time CML
DAQ system (
pySNMPdaq
)
• We continuously get data for 4000 CMLs in
real-time in Germany
•
Outlook
• Extend DAQ to more CMLs
• Make system more robust
•
Outlook
:
• We are testing machine learning
and neuronal network approaches
