Last week

Validation

I finished my first pass on occurrence validation from the Dahal et al validation standard for AOHs. This was enabled thanks to the tidying work I did on Yirgacheffe last week. Unfortunately it requires me to re-run my AOH pipeline as it relies on new data being propegated when we get the species data from the IUCN Redlist, so I kicked that off at the end of the week to run over the weekend.

Hybrid habitat maps

I've continued to push on the new hybrid habitat map that we've been looking at in the LIFE team. This is an extension of the habitat map Tom Ball made for the Nature Food paper, with some method improvements. When I last mentioned it we had something that looks like it works at a course grain view (10km per pixel and larger), but I'm very unhappy with below that due to the naive statistical distribution of how we applied the course grain Global Agro-Ecological Zones and HYDE anthropogenic land use estimates data on top of the fine grain Jung habitat layers.

To solve that I've been looking into the Woodman et al paper that proposes a way to do better clustering of change pixels when merging course and fine grain maps. I've got a version of their LandScaleR R Package working with a variation of my pipeline, and I'm starting with exploring that - again, I have some first data generated, but I need to do analysis on it. Unfortunately being R based it's a lot slower than my pipeline which uses Yirgacheffe to leverage parallelisation on bigger machines, so I'm currently working on just Scotland before I try do a global run. It was a reminder that all the effort I pour into Yirgacheffe does actually mean something.

STAR

My implementation of the STAR metric has fallen a little out of sync of the LIFE implementation in terms of using newer features in Yirgacheffe to simplify the code and in terms of using the AOH code via pip rather than submodule. I want to use both STAR and LIFE to test the validation code, so I spent a little time catching STAR up. I think it's about done, but I need to do a full run just to shake down all the changes, and that sits in the queue behind all the other number crunching I've had machines doing this last week.

PROPL

We had to respin the camera ready for the paper due to a bug in the ACM Article latex package (I guess this?). They also asked for some other tidying bits, so I swept those up and resubmitted it.

I failed to make progress on the presentation a I nerd sniped myself when trying to find nice sample data for the talk. My reasoning was I should get data I could also use for the documentation, so is permissively licensed, and there went a whole afternoon playing around with Sentinel 2 data that I'd never looked at before.

Active Travel Hackday

I wrote up the Active Travel Hackday in a blog post.

FOC on Esolangs

I enjoyed the latest episode of the Feeling of Computing podcast which covered a survey paper on Esoteric Programming Languages. I guess I have a soft spot for things like this as I managed to get an entire section on Intercal into my PhD dissertation. At the time I clearly didn't get the joke, beyond understanding it was deliberately obtuse, but still, I feel a sense of achievement of having built my career based on some small part of taking an esoteric language seriously.

This week

• PROPL presentation MUST BE DONE
• Look more at the hybrid maps
• Look more at validation
• Do a run of STAR