publiceerbare, herbruikbare Python toolkit voor SDM-DL (app. note; pyPI package)
toepassing van toolkit op use cases (2+ research papers)
formuleren van vervolg(-aanvraag)
[x] use cases:
HS voor Observation.org data om beeldherkenning mee te schalen.
idem voor planten met/zonder secundaire houtigheid
idem voor planten met ectomycorrhiza, arbusculaire mycorrhiza, of iets anders
idem voor gedomesticeerde versus wilde planten
[x] toolkit:
consumeren van occurrences van diverse bronnen (GBIF/Observation.org/derden). In scope: transformeren van data in diverse formats (e.g. TSV/CSV/DarwinCore/JSON). Out of scope: opschonen van data, wat b.v. kan met CoordinateCleaner etc.
consumeren van GIS data van diverse bronnen (bioclim/envirem/derden/map queries). In scope: importeren van diverse formats (e.g. TIFF/BIL); croppen; herprojecteren; bufferen. Out of scope: complexere GIS operaties.
consumeren van andere variabelen (metadata van waarnemingen)
configureren/runnen/evalueren van model training. In scope: prepareren van input vectors (e.g. co-occurrences ja/nee); configureren/parameteriseren van DNN topologie; epochs. Evaluatie van AUC's en variable contributions.
projecteren van modelvoorspellingen. In scope: genereren van habitat suitability (HS) kaarten/data sets voor huidige verspreiding. Out of scope: IPCC klimaatvoorspellingen.