EnviroSensing Monthly telecons
back to EnviroSensing Cluster main page
Telecons are on the fourth Tuesday of every month at 4:00pm ET. Click on 'join'
- December 23, 2014, 4:00 PM EST. No telecon - Happy Holidays
- January 27,2015, 4:00 PM EST. Rick Susfalk, Division of Hydrologic Sciences, Desert Research Institute
Notes and recordings from past telecons
To watch the recordings make sure you allow popups for https://esipfed.webex.com
Today we had Dave Johnson from LiCOR Environmental speak with us about eddy covariance. We had a large group of attendees.
- The problem they are solving is that the data coming in with eddy covariance is high resolution and large, more than 300 million records per year.
- They were able to put much of the processing and QC logic into the SMART system, which is on the instruments in the field.
- An eddy covariance system can "see" about 100 x its height above the canopy.
- We were interested in the possibilities for real-time data qc, and also how the information can be transferred between the instruments in the field (i.e. cell modem, line, etc.)
Today we had William Jeffries from Heat Seek NYC talk to us about his platform for civilian monitoring of home temperatures in apartments using low-cost sensors and wireless mesh networks taking hourly readings.
Attending the call were Fox and Don from Andrews, Josh at UM, Jason in Fairbanks, Becky from Onset, Ryan from UT and Scotty from UNV as well as some listen-in callers.
- Heat seek NYC has nodes containing up to 1000 low-power, low-cost thermometers which are mapped to custom printed circuit boards. The software stack is Ruby on Rails with a Post-Gres backend.
- HSNYC focuses on sensor network's ability to affect and effect policy, and so far has seen that the government views sensor networks as a cool solution to a lot of difficult regulatory challenges. They provide reliable information that officials and citizens can use with printable, court-friendly PDF's.
- Funding is by kickstarter and was started by the NYC Big Apps campaign, a giant Civic Hacking Competition.
- They have been incorporated as a non profit in NY since 2014
- First sensors were Twine Sensors off the shelf temperature sensors; not so good
- Now they make their own temperature sensors, using a push system instead of a pull system. (if there's a problem with the system just don't get pull out of it.)
- We asked about LCD and the reason for not implementing is that many tenants aren't actually as interested in the data as the response from the policy makers
- Some people will abuse the system so they use tamper evident tape and photos of installation to protect ; landlords have financial incentives and there can be intentional lack of repairs
- Local cache-ing of the sensor occurs; sensors cache at the hub, relay server that sends back
- There are several levels of caching, and also well as flash memory on the XP radios
- A key question was is there a long-term storage that is a local cache? We don't have a solution either
- We had interest in a company called H20degree
- For QC they do indoor and outdoor temperature comparison, comparing the sensors to one another and to store bought sensors before and after installation...
- Frequency for radios/caching is hourly waking up point data
- We are all interested in a low-or-no power source/ transmitter, maybe based on raspberri pi
- need a high peak power envelope-- transmission needs to deal with being fairly far apart... implementation needs to come online and transmit an awesome signal for a short time and go back to sleep.
Note on Webex recording: I am going to check into the recording- my system claims to have been recording but it is nowhere to be found (Fox)
Rick Susfalk from the Desert Research Institute presented about Acuity Data Portal. Notes from meeting taken by Fox Peterson, please edit as you see fit.
We had in attendance 8 persons.
Acuity Portal System
- Started in 2006
- originally VDV data solution
- improvements to web-interface; sits on top of VDV as Acuity server
- Acuity is a continuous monitoring of key client-driven data
- it includes sensors and data logging deployment and maintenance, telemetry, data storage and analysis, automated airing, web portal for data access
- individualized web presence tailored to client needs
- not a single tool, but instead integrates commercial, open source, and proprietary hardware and tools
- customizable project specific descriptions
- common tools used to provide rapid, cost-effective deployment of individualized portals
- physical infrastructure is shared amongst smaller clients for cost-saving or it can be segregated for larger clients.
- access is controlled down to the variable level-- "we can define who gets to see what"-- for example, public can not see some features
- one view could be "pre-defined graphs" without logging in, but if you want to download the data you must log in at your permissions level
- priority, very important
- DOD certification and accredidation (possible link below)
- have hired a security professional
- https:// everywhere for good protocol
- customized thresholding and data-freshness
- trending alerts, for example, know if battery will go bad
- stochastic and numerical modeling
- scoring incoming data for QA/QC processing
- web-based GUI
- users and managers can create, edit, and modify alerts online
- groups can be created so that you can schedule management and alerts
- also offers localized redundant alerting
- two-way communication with the campbell data loggers (cr1000)
- more about getting the data to the data managers for more in-depth QA/QC than about providing that part of the tooling
Data graphing features
- Pre-determined graphs for basic users
- Data selector for more advanced users
- "we don't know what the users want to see so we give them the tools to do it" (good idea!)
- anything you can change in Excel you can change in their graphs on the website -
- vista data vision (http://www.vistadatavision.com/)
- another vdv link: (http://www.vistadatavision.com/features/responsive/)
- relates your parameters to the network and what other sensors are doing
- current system is getting more flexible
- metadata is still largely user responsibility
Flight plan - safety tool
- field personnel are data
- users put in the travel time for safety
- buddy system, alerted right before you return, then calls boss etc. Many levels hierarchy
- Portals that are monitoring things
- ability for data refreshment
- colors for indication, ex., data would not be gray if there was lots of new data
- users can change the settings on the data logger
- scrolling, scaling, plotting, etc. via interaction with the user
- can save your own graphs
graphs and alerts
- many parameters
- you can save!
- email, sms, phone
- default settings for users
- lots of personnel management tools in this in general
- cross-station "truly alarm or not" if station1 has a value but station2 has a different one, don't alarm sorts of rules
- lists/user groups appear to be very important with this tool
- sensor and triggers: customize one or more parameters that you are bringing into your database
real-time updates on loggers
- ex. 10 minute data, user comes in and makes a change, the information is saved to the database and then is presented to all other users
- the person will request a change and say what that change is
- when there are different levels of connectivity ie. analog phone modems, before the data has the chance to work its way back into the system there is a lot of validation being done
- everglades heat pulse flow meters
- uses google maps
- extends beyond the vdv, more than 1 .dat file
- integrate multiple .dat into many tables
- managed by the data managers at DRI
- workflow :
logger net --> vdv --> acuity, ok, let's give access to the DM for all these variables, click on it ok now the manager can see it --> generate an excel file with tables for all this metadata --> enter the data into the excel files--> send back to acuity --> injests, runs queries, back to db-- > metadata in bulk, quickly
- we asked if the system ends before the qa/qc process begins, answer: qa/qc is done at the DRI, near real time QA though
- direct the managers to the future data problems
- manual decision making
Scotty asked about the duration (long and short term) of projects and how affects funding. most is funded by long-term projects; this is why they do the stats and numerical methods in the future
Amber asked about pricing; pricing is by hour to get up, then a price for maintaining the system for the duration of the project 5-10 .dat files, only 8 hours of person time at DRI to make a portal
Jordan Read presented the SensorQC R package
Wade Sheldon presented the GCE Data Toolbox – a short summary follows:
- Community-oriented environmental software package
- Lightweight, portable, file-based data management system implemented in MATLAB
- generalized technical analysis framework, useful for automatic processing, and it's a good compromise using either programmed-in or file-based operations
- Generalized tabular data model
- Metadata, data, robust API, GUI library, support files, MATLAB databases
- Benefits and costs: platform independent, sharing both code and data seamlessly across the systems, version independent as far as MATLAB goes, and is now "free and open source" software. There is a growing community of users in LTER.
Toolbox data model
- Data model is meant to be a self-describing environmental data set-- the metadata is associated with the data, create date and edit date and such are maintained, and its lineage.
- Quality control criteria- can apply custom function or one already in the toolbox
- Data arrays, corresponding arrays of qualifier flags -- similar to a relational database table but with more associated metadata
Toolbox function library
- The software library is referred to as a "toolbox"
- a growing level of analytical functions, transformations, aggregation tools
- GUI functions to simplify the usage
- indexing and search support tools, and data harvest management tools
- Command line API but there is also a large and growing set of graphical form interfaces and you can start the toolbox without even using the command line
Data management framework
- Data management cycle - designed to help an LTER site do all of its data management tasks
- Data and metadata can be imported into the framework and a very mature set of predefined import filters exist: csv, space- and tab-delimited and generic parsers. Also, specialized parsers are available for Sea-Bird CTD, sondes, Campbell, Hobo, Schlumberger, OSIL, etc.
- Live connections i.e. Data Turbine, ClimDB, SQL DB's, access to the MATLAB data toolbox
- Can import data from NWIS, NOAA, NCDC, etc.
- Can set evaluation rules, conditions, evaluations, etc.
- Automated QC on import but can do interactive analysis and revision
- All steps are automatically documented, so you can generate an anomalies report by variable and date range which lets you communicate more to the users of the data
- Fox Peterson (Andrews LTER) reported on QA/QC methods they are applying to historic climate records (~13 million data points for each of 6 sites).
The challenge was that most automated approaches still produced too many flagged data that needed to be manually checked. Multiple statistical methods were tested based on long-term historical data. The method they selected was to use a moving window of data from the same hour over 30 days and test for 4 standard deviations in that window; E.g., use all data for 1 pm for days 30 - 60 of the year, compute four standard deviations, and set the range for the midpoint day (45) at the 1pm hour to that range.
- Josh Cole reported on his system, which is in development and he will be able to share scripts with the group.
- Brief discussion of displaying results using web tools.
- Great Basin site discussed the variability in their data, which "has no normal"-- how could we perform qa/qc based on statistics and ranges in this case?
- Discussion of bringing Wade Sheldon to call next time / usefulness of the toolbox for data managers
- Discussion of using Pandas package- does anyone have experience, can we get them on?
- Discussion of the trade off between large data stores, computational strength, and power. Good solutions?
- ESIP email had some student opportunities which may be of interest
- Overall, it was considered helpful if people were willing to share scripts. Discussion of a GIT repository for the group, or possibly just use the Wiki.
Suggestions for future discussion topics
- Citizen Science contributions to environmental monitoring
- 'open' sensors - non-commercial sensors made in-house, technology, use, best practices
- Latest sensor technologies
- Efficient data processing approaches
- Online data visualizations
- New collaborations to develop new algorithms for better data processing
- Sensor system management tools (communicating field events and associating them with data)