Reaching out and engagement

[mattjbr123]

Continuing from NERC-CEH/rse_group#27
(I'm sure there's a nice way to merge/migrate discussions, but having a link between will do for now I guess!)

[mattjbr123]

Had an idea for engagement:

"
I realised recently I don't know anything about SQL and database type stuff and would be interested in picking up at least a basics level knowledge on it so I don't completely blank out when people mention it. If others in the group/org know something about it I'd be up for an informal show-and-tell-and-ask-annoying-questions type session, that I'd be happy to reciprocate for other things. Could be something we invite a wider audience to too (though not necessarily initially whilst we trial it)
"

I feel like there'll be some people in the Hydrological Data Science Network that know a fair bit about SQL, I'll maybe reach out there as a starting point

[metazool]

Continuing from NERC-CEH/rse_group#27 (I'm sure there's a nice way to merge/migrate discussions, but having a link between will do for now I guess!)

It looks possible to transfer them between projects but only within the same Github Organisation - we could move the old group project across and recreate it as a stub there, it feels like a hard commitment while we're still trialling this setup...

[metazool]

https://volcan01010.github.io/RSECon24/index.html - this was a lovely talk by @volcan01010 at RSECon, "SQLite for Everyone". It has a quick SQL intro section and examples of how SQLite integrates usefully with different tooling.

Groups doing spatial data analysis work will likely have built up PostGIS/ Postgres expertise, too.

Datasette has an interesting stack built on sqlite. Now I've seen it has a plugin for searching vectors of neural network embeddings, I want to find a place for it again!

[volcan01010]

Datasette is so good. @metazool - the volcanic ash advisory analysis tool that Peter wrote for BGS volcanologists finally broke. I rewrote it last week to parse the files and write them to a Datasette instance. They are just looking at the raw tables for now, but it will be so easy to build analytics or dashboards on top.

[metazool]

Q&A in Wallingford about how to work with/through the RSE group as a vector for new collaborations across disciplines

At the moment the answer is "engage through Hub resources" but the forum here should also be part of that answer, how does that look in practise?

[metazool]

https://github.com/unifhy-org/unifhy - dedicated GH org, Hydrology work

https://unifhy-org.github.io/unifhy/

"it would be a far better tool if it were easier to add new science to it"

[mattjbr123]

Hehe, yes I completely agree!

There's two aspects to it:

• Adding science to existing components in the framework (for which read "areas of the earth system that can currently be modelled with UnifHy") (currently "openwater", "subsurface", "surface" for all hydrological modelling and their equivalents for nutrients modelling)
• Adding completely new components (new supported areas of the earth system e.g. atmosphere, ocean, or in my case, nutrient modelling)

It took me a long time to add nutrients to it, which falls under the second item above, and next on my list was using that experience to develop a way of automating this process, or at least making it a lot easier. However I then applied for the RSE role and my attention shifted elsewhere xD It is something I would love to take up again though.

My suspicion however is that

"it would be a far better tool if it were easier to add new science to it"
means more that first item - adding existing models/science to existing UnifHy components. Currently it is quite a technical approach to do this, though I think the instructions on how to do this are comprehensive. A lot of thought is also required on how to factorise the existing models to fit the framework, and once factorised, how to adapt the code to work with how UnifHy integrates the components together (it does this in a specific way that can change the outcome of your component's code if you do not account for it). On these aspects the documentation is lacking, I think.

All that is not mentioning how finnicky UnifHy is when accepting new driving data, which was also on my list to fix xD

So that's my view on things. I would definitely be interested to know in more detail what potential users think could be done to improve the experience. I could potentially then feed this back into further work on the project.