Last week

Tropical Moist Forest Evaluation Methodology Implementation

This last week had a lot of TMFEMI in it, but I did more herding and data review than actual coding, so no pretty pictures from me this week, that’ll be with Patrick and Robin. The TL;DR is that we might finally have results from our implementation that match the one Tom has in Google Earth Engine.

We now need to work out both what’s needed in terms of tidying up the work we’ve done, and then aligning it with the next stages.

IUCN work

I did find time to tick off from my list generating more test data for IUCN, generating rasters for a dozen or so seasonal species’ AoH - before the code worked on resident species, but now I’ve added migratory species support to the code.

The difficultly then came from trying to compare results between our code and theirs. My results are not pixel perfect matches for those, and nor do I expect them to be: at the most basic we both have to rasterise a vector, and different line drawing algorithms may place pixels in slightly different places; add to that then map projection changes, floating point rounding, and so forth, there’s bound to be a little variance.

In the TMFEMI, and in this instance, one results to looking at the rasters to see if they match, but I feel we need some other metric that is meaningful: the AoH between the two is within x percent or such. I feel I can’t be the first to hit this, so want to ask the rest of the team this (oh, hi reader!) if they have any pointers.

I did make some small tweaks to Yirgacheffe to let me specify TIFF options that make rasters easier to preview on a Mac: it turns out for 8 bit data in a TIFF you can say you’re using fewer bits, and then viewers will scale up the max to be equivalent of white, which means what would otherwise be a black preview makes sense in Finder, making it easy to quickly rough check a large batch of data:

In the IUCN TIFFs they were using this as they had only three values, so they set this NBITS setting to 2, and I wanted to do the same. Took me a while to figure out where in GDAL you can prod this in, but it’s there now :) A small thing, but super helpful when doing work like this.

DIFC

I read the two DIFC papers (the original and the HiStar OS one) for the reading group ran by Patrick and Ryan, and that was super useful. I now have a vague understanding of DIFC, enough that I hopefully will understand it actually next time I read the papers :)

Two two linked papers both actually have better versions where the authors had more pages to provide more practical examples, so if you’re interested I’d recommend you read this longer original paper and this longer HiStar paper.

This week

• I failed to make progress on Biodivisity fact checking, so need to catch up with Ali and see what’s needed there, or if that’s now moot.
• Write up what clean up is needed for the TMFEMI pipeline
• Get IUCN peeps a list of the input data we need for our pipeline so they can give us all 40k species
• Start learning how to implement a pipeline in OCurrent, so I can do the AoH one, if not myself, at least not leave it all to Patrick.