Friday, 6 March 2015

Jellyfish are dish of the day at EnviroHack 2015... by Rachel Heaven

EnviroHack 2015 ©
Is open access data helping to solve environmental problems? Yes, but only with the innovative thinking and skills of people from a wide range of scientific and engineering disciplines. Rachel Heaven, Stephanie Bricker, Anubha Singh may have different expertise within BGS but one thing they all share is wanting to put data to it's very best use. Here they tell us more about their experience (and jellyfish) at this years NERC EnviroHack...

So it turns out that jellyfish can wreak havoc for power stations by getting sucked into and clogging up the water supply intake pipes – who knew? (Don’t believe me? The headline in National Geographic read Jellyfish Invasion Shuts Down Nuclear Reactor). A system to identify and track likely jellyfish hotspots using atmospheric and oceanographic data was just one of the great ideas we saw developed at NERC’s EnviroHack 2015 last weekend.

EnviroHack 2015 is a data jam. In essence you put a bunch of enthusiastic data scientists, designers and software engineers in front of a wealth of openly accessible environmental data (atmospheric, meteorological, hydrological, ecological, geological…), feeding them some pizzas (and perhaps a little beer), letting them cook for 2 days and then hopefully seeing some innovative application prototypes that could address important environmental challenges such as environmental change, resilience to environmental hazards and use of natural resources.


The event was hosted at Digital Catapult, in what is being branded the Knowledge Quarter in London. We were on a 9th floor office opposite the British Library, giving fabulous views over St Pancras Station, and across the London skyline. 

After admiring the view and a few talks from the event partners - which in addition to NERC and Digital Catapult included the Knowledge Transfer Network, Microsoft Research and Red Ninja Studios - we had a quick brainstorming session and then were encouraged to give lightning talks to pitch our ideas. Some of these had clearly been planned in advance and were well thought out with a good knowledge of the available data...others were loose ideas sparked off that morning that morphed and took shape as the project teams were formed around them.

The ideas that we pitched and worked on (food_chain and Energy Cast) definitely fell in the latter category, but the self-organising teams managed to do an amazing amount in a short time to present prototype solutions by the end of day 2.


The jellyfish hackers on team JellyStrike led by Centre for Ecology and Hydrology's (CEH) Tom Redd were catch of the day taking overall first prize and winning development support for the next 100 days, but - as at a school sports day - most teams managed to come away with at least one prize (best 2D visualisation, best mobile app etc).

Some of the other ideas were:
  • food_chain - connecting surplus produce e.g. from allotments, to food charities and businesses to reduce food waste.
  • Energy Cast - a weather-forecast and machine learning informed green energy app to help match domestic energy supply with energy consumption.
  • ShowMeTheNature - putting people in touch with green space, an app to rate the quality of the green space near to you using the metrics of your choice, with user-feedback.
  • CHARM - a self-learning data system to allow users to enrich environmental data by adding annotations, use cases and key metrics to make more sense of our data-rich world.

Why no geological data ? Well, independently of this event BGS had already started implementing ideas based on the same principles as EnviroHack. For example:
  • aligning our geochemical sampling datasets with those of CEH and British Oceanographic Data Centre (BODC) so that chemical pollutants can be tracked from source to sea
  • aggregating worldwide soil data in the BGS mySoil app
  • optimising the use of national scale datasets in Natural Capital Mapping (see the  CEH/BGS academic paper here)

The original idea for Energy Cast could have included various datasets available for assessment of Ground Source Heat Pump potential, but the hackers preferred the wow factor of presenting dynamic datasets (and who can blame them with an Oculus Rift virtual reality headset to compete for?!). It’s true many of BGS’s open datasets describing the subsurface are static, but they can provide the framework that determines the movement of fluids underground and are so important for many of today’s critical issues.  As an organisation we are increasingly dealing with time series monitoring data and outputs from process models, and we need to work on making all those open and accessible where appropriate so that other developers and data scientists can make use of them.

We also learnt a lot about how other people approach and solve software problems, got ideas about how we might apply machine learning, made a lot of good contacts and had a load of fun. 


What did the organisers learn ? That even though lots of datasets are available it still took them a lot of time and effort to find out what was out there and how to get hold of it, and even then the size and complexity of the data can inhibit its usage. BGS was held up as a great example of making data open and accessible but we still have far to go and it’s something we are continuously improving.

Rachel, Stephanie and Anuhba

No comments: