Where it Started: Three Little Birds. (Monitoring: Air/Sun/Water)
I created the ‘Good Air Canary’ for RS DesignSpark back in 2021, with expert coding and electronic help from Pete Milne - which connected over WiFi to a personal Air Quality Sensor array (ESDK), we created in 2020. The yellow automata bird was to make the science bits more accessible, using the ‘canary in the coal mine’ metaphor. Video in action, here.
Following this, we wanted to explore using APIs (the digital things which make many apps work with 3rd party data). We were aware that the Met Office API was pretty well developed, but not often used beyond creating home ‘Weather Stations’. Strangely UV Radiation (ie sunburn risk) was not very prominent, perhaps as most people (myself included) don’t realise that even on a cloudy day, 80% of UV radiation passes through! So a year later, we created The UV Budgie, to continue the ‘bird theme’, and to recycle the form factor, as we explored API and IoT tech.
Image: From silliness, occasionally comes useful insights and innovation. The IoT bird evolution: Good Air Canary, The UV Budgie, The Flood Falcon, all using IoT hardware and automata.
I mention this to say to folks at the early stages of their career, that there is a running joke of ‘How long did it take you to do x?’, answer, ‘Only 10 years!’... and even though the Flood Alert was developed quite rapidly, (mostly thanks to Pete Milne on cracking the API code!), in contrast to the ESDK and Canary/Budgie, it was rapid because it was benefitting from the knowledge accumulated over a decade of experience.
Design has ‘compound interest’ it seems! So keep working, even if it’s a bit weird/whimsical.
So the third bird, the Flood Falcon, was created more just to have a bit of a laugh between Pete and myself and finish the ‘Trilogy’ in style. (Though in hindsight, the UV Budgie should be Red, and the Flood Falcon should be Blue, but such is life). We kinda knew deep down that this was not the right form factor, and even in the early days, I was thinking it should be much more simple, perhaps something like an Alexa. But at this stage, we were doing this to define the concept enough to ‘pitch’ back to RS DesignSpark, so it was our free time we were fooling around with! And who says your client doesn’t like a laugh!?
“Great Artists Steal”
So says the Picasso quote. The full version was that “good artists copy, great artists steal”, and it’s worth ‘unpacking’ perhaps the more interesting insight that is rarely included with this quote, rendering it rather arrogant and seemingly justifying plagiarism…
Although Picasso certainly did ‘steal’ (google african art and cubism, and his initial denial of it), what anyone who has tried to ‘copy’ something they are fascinated by - is that unless you are a professional art forger, and literally intend to copy every stroke, your work actually ends up being more of a ‘homage’ to the original source. So it’s not that one shouldn’t be honourable and cite ones sources / influences (e.g. like academic papers), but it’s more that even when you try to ‘copy’ it’s rarely a ‘clone’, it’s a product of two or more ‘parents’, and takes on it’s own life...
There was another phrase I learned whilst working at LEGO, which was “Never before seen in Play” - this loosely guiding mantra (from when I was a Tech Scout) was not to find something never before seen in the world, but just not in play - indeed, as I was to learn after time, it actually helped to ‘borrow’ something for another industry, be it medical, automotive, aerospace, etc… and to reappropriate it into the world of play. In many ways this was the founding insight for much of my work now as a consultant - ‘fortune favours the prepared mind’ (Louis Pasteur), is a good way to view using inspiration, citing, and not plagiarising.
Functionally I instinctively felt the Flood Alert device should be like some sort of ‘Flood Alexa - with a screen’, (at the time I was unaware that the Alexa Show was available). So I began searching for inspiration, and found CryptoInk on Hackaday.
Images: CRYPTOiNK by John Loefller, using eInk display and elegantly simple API crypto data.
(With permission from John).
As you can see from the Good Air Canary, I was already a fan of eInk displays, but what really captured my imagination - for the end user - was that it was on a Fridge!
Being in a well-frequented place like a Fridge Door seemed to make instant sense to me. There is an irony that at the time of discovery, this tech product was designed for what one can fairly reasonably assume are the ‘tech elite’, being into APIs, Crypto, and cool DIY Hardware - so my sentiment was to ‘steal’ the insight that a Fridge is someone that my Mum or Dad, for example, who are not especially techy, would be more likely to check this than a smartphone app.
My parents didn’t use a smartphone and I know this is not uncommon, and indeed, with the previous interview with Heather Shepherd, from Flood Forum, she certainly felt that this was a group of people who might appreciate an intimidating bespoke tech device, rather than another App on their phone. This is not to say Pete Milne or I were ‘anti-App’, or ‘anti-tech’, but more this was an ‘inclusive design’ provocation. Will this be needed in 2060, probably not, but in 2022 - it felt like a good number of people still wanted ‘Internet of Things’ to be more friendly and accessible.
So I got to work and started making a cardboard prototype. You can see the references…
Images: Prototyping process showing the ‘changing display screens’ in rolled-up paper! Magnetic too!
Pleasingly, the ‘roller paper screen’ got a smile or a laugh every time, for being rather ‘nifty’ as a mechanism. Having run many prototyping courses (GSA, D&TA, How.Do) in the UK and abroad, for students through to corporations and agencies, I am still bemused at the beguiling nature of cardboard prototypes! Check out tips online (free), at YouTube.
Although very rudimentary (ie no actual working electronics), you can see it in action here, this was a great way to test this with the users and councillors below.
Images: These are some of the local Councillors / Flood Forum, and local residents (pixelated) - and although I have not got any photos of the cardboard prototype discussions taking place, as sometimes you have to respect one has to ‘read the room’ and sometimes photos just don’t feel the right vibe.
The evening with local residents and counsellors was quite charged, ranging from those who were calmly concerned, to some who were irate and despondent. I did, however, manage to speak to a few residents and councillors after the session, one-to-one, and this was where I got some nice insights:
- Not everyone loves an App.
- Not everyone has a smartphone (or checks it daily). It may seem implausible, but digital exclusion is real even if not often discussed in our circles.
- Getting a Met Office email sounded tricky / would not do. Want ‘plug and play’ solution.
- It being more ‘homely’ was better. (Varying likes on Fridge/Stand/Wall/etc.)
- ‘Flood Alexa’ became a ‘shorthand’ in conversation, even for those who didn’t have Alexa, they appreciated the metaphor for what the device essentially was offering.
- Most were unsure how the ‘machine learning’ bit worked, but understood that things like Google Maps took advantage of other drivers’ data if there was an accident, so this was analogous, and would ‘help everyone through 1 person’s data’.
- Simple interface of ‘traffic lights’ was popular.
- One asked about sensing outdoors - can it detect rain, barometric pressure, etc?
- Must not be too expensive (under £50/£60 ideally was a general sentiment).
Met Office Email Alerts vs API
As discovered in my chats with local residents, and with Flood Forum staff, ‘not everyone loves an App’ - and this insight is at the core of why I wanted to build a Flood Alert IoT device that is ‘stand-alone’ and not just an App - though I must say, it is very logical to assume that this project could also create an App as well as the physical device. However, for some people, having a more bespoke solution is key - and having it’s location be part of the motivation to check it is also part of getting the message early.
I happen to have signed up to the Met Office / Environmental Agency’s email alerts on Flood Warnings, as part of this project…
The process is that you’d follow the link for your postcode, and select one or more areas that relate to you. Interestingly none are offered for Wood Street, so one has to take a ‘best guess’ as to which of the offered [river/fluvial] datasets are most applicable. Proximity seems a reasonable assumption, but then again, it may not be. And this is my point on the lack of *local* data for your given area. Anyway, something is better than nothing, so do sign up…
Images: From this you will see the Map, and be asked for your email address (or you can choose text), assuming you check either of these.
Images: Showing that you’ll need to accept T&Cs which are pretty fair - they are not promising infallibility, or that you can sue them for any effects of their information sent (or not sent) to you. And indeed, for this project, I’d have to say the same!
Images: All done - you’ll get an Alert like this, above.
Although this is certainly a good effort of the Environmental Agency, one can see why residents are not totally satisfied with this, as phrases like “It is currently unclear which areas will experience the heaviest rain” are of course the logical and legal thing to write, but when I spoke to residents, this leaves them feeling unsure how to react.
Furthermore, when it advises “We expect river levels that receive the heaviest rain to respond by rising quickly. We recommend you stay aware of local weather conditions and water levels. Consider putting your flood plan into action. You should avoid walking, cycling or driving through flood water.”
Images: These are my email alerts from the past 2 years. This is not an exhaustive or exacting dataset, but it makes the point that it’s not just Autumn or Spring that one gets alerts. And indeed, none of these alerts resulted in my being flooded, or (fingers crossed) not really seeming at risk of being flooded, and yet the alerts came.
This nonetheless corresponds with the issue that residents are suffering from the equivalent of ‘Boy Who Cried Wolf’ - and tend to ignore the alerts if they do get them, as they are often over-cautious. To be fair to the Environmental Agency, when they were previously more conservative (I’m told anecdotally before I started this project) a few years ago - their alerts were only if a really serious flood was imminent, as so this meant they inevitably failed to alert people one time - and a fairly bad (but not catastrophic) flood occurred, so this explains why they are not wanting to under-estimate again. So it’s an impossible task to make this a ‘Goldilocks’ or ‘just right’ degree of alert, as every borough, road and home is different. This again, is why ‘hyper-local’ is needed, to dial-in the accuracy of the forecast.
Who Needs a Flood Alert and How Should it Work?
Building off the interviews, and combining with the ‘insight’ of CryptoInk placing an IoT device in a prominent position in someone’s home, I tried to visualise the generic user experience. Aside from my terrible renderings, I should stress that this is not about saying only women or elderly people might need this, but that it evokes the right assumptions. It’s nonetheless a balance not to overly stereotype people, and indeed, as OXO Good Grips is a prime example - sometimes you design for a niche group, e.g. Elderly - but it ends up appealing to broader ages. This was certainly my hope, but for my sins, this is the first embodiment of my implementation of the Flood Alert, based on discussions with real people and from insights derived from reports, councils, charities and more.
The key insight here is not that the Flood Alert is in a prominent position, and that it’s evidently not an App, and that by lazy stereotype, the old lady is presumed not to be that tech savvy, but the dialogue being that the Flood Alert is first warning of trouble, but secondly, it is triggering an action for the Lady to call a friend to help out, perhaps with sandbags, flood gates, etc. One cannot hold back torrential flooding, but for minor flash-floods (as described by David from Thames Water) this is a critical moment to intervene in a relatively easy to avert mini-flood in your home.
Images: The dialogue here is basically implying that the Lady has called her more capable friend, and has explained that from her experience, she knows that when the water is pooling up like this, with no sign of stopping, in a short matter of time, it’ll cause flooding. This is a paraphrasing of what many people are quite astute about.
From the residents I spoke to, to the Flood Forum, the same observation was consistent - residents are good at predicting when things are ‘getting worse’. The Flood Alert now takes on a different role, and it’s intention is to *record* the individual’s personal assessment of the rainfall. If it feels ‘too close for comfort’, rightly or wrongly if it does flood, they should press the Orange Button. If it actually floods, they should press the Red Button. This would then log the situation, such that predictive models might be used in future, in ways the user does not know.
Here’s some archetypal examples of micro-flooding, which might be possibly averted:
- Dry then Very Wet.
If the Weather is forecast to have say 7 days of heat, with no water, the ground becomes very hard [for this location]. It means if it rains hard one day, the water just ‘flows’, rather than soaking in - this is a useful risk-analysis pattern to consider in preparing someone for the severity of a flood, that is not immediately obvious to many as a secondary factor. - Too Wet - Tipping Point.
Conversely, it may have been raining gently for 3 days in a row, on-and-off. The ground is soaking it up, but [for the given location] it is now ‘at capacity’, so even a relatively modest shower now results in flooding. So again the predictive model is looking at mm of rain fallen over successive days, and evaluating the probability of it suddenly getting bad. - Just My Luck!
Many residents described the huge inconvenience of being at work, and away from home when they had a mini-flood, which with a sandbag or two could have just kept those few inches of the surge at bay, and averted an issue. Similarly, one may go to bed, miss an Environmental Agency alert (which can come at any time it seems day or night), and so one really needs the ‘should I put sandbags out before I leave the house tomorrow’ estimation, not the ‘flooding in 5 hours’ - which may sound like ample time, but is not if you’re working a 8 hour day plus 1-1.5hr commute as many folks are in London. - Out of the Blue.
Last but not least, is of course when you get simply a lot of rain. However, as illustrated above, heavy rain can come in the form of wintery downpours, but also summer thunderstorms - and all in between - my point being in the UK, weather is incredibly hard to predict! (Having also lived in California, it’s hard to explain if you’ve not lived here!). So this will always be difficult, even for the most ‘hyper-local’ dataset as by definition it is especially hard to predict, but perhaps we can predict it better with collective data input.
Let’s not forget for my 1 square mile area, there are about 5 river sensors and zero surface water sensors, and of course nothing at the street-level of citizen input - and this is what might make it start to be more accurate. This is the hopeful goal of incorporating more data on top of the Environmental Agency data, but also the Met Office (weather) data.
In summary, what the Flood Alert’s predictive model might be able to do is add a layer of predictive pattern recognition to otherwise very good, but not localised government data.
Flood Alert is “Standing on the Shoulders of Giants”
I want to stress this is not saying the Environmental Agency or Met Office are ‘bad’, or trying to criticise them, but rather that their data is perhaps let’s suppose 90% accurate, but that last 10% is incredibly hard to ‘dial-in’ to a local level - and this is what the Flood Alert is trying to operate on.
Image: Just one example of the excellent data from the Met Office. Image Credits: Met Office: (link)
I personally hope it may take us from (for example) 90% accurate, to 95% accurate, but I’m also realistic that this will never be 100% accurate. This is where adding some ‘AI’ should not be overstated, as the AI is really a humble ‘predictive engine’, but I’m not wanting to claim this is Deep Mind level of sophistication.
What WeatherTech Can Learn from FinTech
It's worth pointing out that if I were able to create a model that was this accurate, I’d probably have a FinTech company knocking down my door long before Gov.uk contacted me!
This may seem a flippant comment, but within this is a serious point - I in fact do fully intend to ‘steal’ again from other predictive models…
Images: Above is an example of CMC Markets - a day trading (FinTech) platform.
I happened to work with them a few years back and was blown away by the incredible mathematics used to predict trends in economic shifts, but also was chastened by the very real truth that although you may be right 90% of the time, like flooding, that 10% might wipe you right out.
If you observe what the day trader success rates are, they are not even close to 90%, and for many people don’t even get to 50% - ie loss-making overall. So this is also a sobering statistic that I need to be very focussed on a realistic goal for predicting weather and flood risk, that does not pretend that weather is any less turbulent and unpredictable than the stock markets and global political forces.
Images: Another example of Met Office analysis - looking not dissimilar to FinTech! (link).
In my favour, is that there is much less subjectivity in the problem I’m trying to crack here. Weather does not have quite the same level of capriciousness as humans, as one would surely agree that, unlike the financial sector, the Meteorological sector has steadily improved its scientific models, as last I checked the weather reports were way better than when I was a kid! Whereas 2000… 2008… 2015/16… you get the idea. But as a technologist, I’m intrigued by what I can learn from my contacts in both worlds, and beyond. One thing is for sure AI models will improve, be it screening for cancer in X-Ray images, or predicting global fluctuations in finance, agriculture, or indeed, weather…
Future Ideation: Expansion Capabilities in Hardware and Software
Although the project certainly needs to deliver on the main initial goals of getting the device working, and having some rudimentary ‘predictive’ or ‘pattern recognition’ system, it’s also interesting to explore how this may also interact with other tech - namely computer vision.
Image: The example illustrated here is exploring the combination of predictive models which combine a. Met Office (weather), with b. Env Agency (river/fluvial), and c. ‘hyper-local’ data - as described, but also d. with visual machine learning (camera) data, to look at the road over long periods of time.
This is of course ‘getting ahead of oneself’, but I find in ideation this can mean that you nonetheless prepare for things as best you can. This in fact meant that one very small and easy-to-make change was done to the 3d Print and circuit board integration which means that things like this *could* be done, with a simple auxiliary port, rather than having to start all over again.
Anyway, on with the actual build! See next chapter…
Exploring the initial Flood Alert Offering, with Flood Forum’s - Edward Flaherty
Many thanks to Edward for his time and enthusiasm to help explore this concept. At this stage, I had some basic sketches and ideas, and the ‘Flood Alexa’ notion was something we could connect on, and ideate from. Ideas eventually become nebulous, and a composite of many direct and indirect suggestions and insights. So many thanks for such a great discussion. Video and Highlights below.
Edward Flaherty - Flood Forum - https://youtu.be/yDGkNO-QWWA
Complexity of Flood Risk Management: Both Edward and I acknowledge the intricate nature of flood risk management, highlighting the confusion around responsibilities, jargon, and lack of accessible information for the public.
Gaps in Existing Warning Systems: Current flood warning systems primarily focus on river levels, neglecting surface water flooding, a significant issue in urban areas like London. Existing systems also struggle to incorporate real-time, localised data.
Potential of Citizen Science: Both parties express enthusiasm for citizen science as a valuable tool to bridge these gaps. The flood alert system, with its ability to collect granular, localised data, is seen as a positive step towards addressing this need.
Importance of Community Engagement and Trust: Building trust within the community is crucial for the project's success. This involves clear communication, transparency about data usage, and demonstrating the system's value through tangible examples and case studies.
Addressing Vulnerability: The interview emphasises the need to consider vulnerable individuals who may lack resources, support networks, or technical literacy. The project's aim is to empower and equip all individuals, regardless of their circumstances, to respond effectively to flood risks.
Traditional flood warning systems, often managed by the Environment Agency, primarily focus on river flooding and may not adequately address surface water flooding, which falls under the purview of local authorities.
The new Flood Alert system uses a physical device with a simple interface allowing residents to report localised flooding events, contributing real-time data that enhances traditional forecasting models. However, data privacy is a paramount concern and the system is designed to anonymize and obfuscate personal information, ensuring ethical data handling.
There is a recognised need to carefully onboard individuals who may be unfamiliar or uncomfortable with technology, ensuring they understand the system's purpose, benefits, and data privacy measures.
NB - Summary Drafted by NotebookML [AI] from Youtube Transcript. Edited by Jude Pullen.
Blog Series Contents:
Prologue - The Case for 'Hyper-Localisation' of Civic Data
Research & Development:
Part 1: Filling the Local Data Gap
Part 2: Civic Services & User Experience Research
Part 3: Ideation of Flood Alert Concept
Part 4: Prototyping Back-Story
Part 5: Citizen Science Learnings
Open Source Build Guide:
Part 6: Build Guide for 3D Printed Assembly
Part 7: DIY Decals for 3D Prints
Part 8: Code & Data Guide
Future Ambitions:
Part 9: Project Reboot with Machine Learning
Comments