[achubaty]

prepInputs_SCANFI_LCC_FAO() now reads the land cover of the forest-land years under LandR's SCANFI remap, like year's own land cover.

Defect. Since #274 the remap is applied per download, inside .withSCANFIAccess() and prepInputs_SCANFI_structure(), instead of globally when LandR loads. In prepInputs_SCANFI_LCC_FAO() only the read of year's land cover (R/maps.R:488-495) is wrapped. The forest-land years are read afterwards, by the lccFor() closure passed to .forestLandFor() (R/maps.R:531-539, called from R/forestLand.R:191-196 when forestLandFrom is "both", the default, or "lccYears"). Those reads ask Google Drive for the restricted SCANFI files directly. Before #274 the global remap covered them. For a 2020 year this means the land cover of 1985, 1995, 2005, 2015 and 2025:

• a user without access to SCANFI's Drive folder fails;
• a user with access downloads each national raster in full (2.4-3 GB each) instead of reading the study-area window from the mirror.

Biomass_borealDataPrep reaches this whenever it builds rstLCC itself, i.e. when rstLCC is not supplied, because it calls prepInputs_SCANFI_LCC_FAO() with the default forestLandFrom.

Fix. The lccFor() read goes through .withSCANFIAccess(), with the same remap and access-error handling as year's land cover. V3, which does not use the remap, is unchanged (enabled = dataVersion != "V3").

Test. In test-scanfiMirror.R, a mocked prepInputs() records the URLs it receives while the remap is set. The test checks that each forest-land year's URL is among them and that the global option is still NULL afterwards. It fails for all five years without the fix and passes with it.

Checks: devtools::test() on R 4.6.1 gives 225 tests, 0 failed, 12 skipped (slow, network or interactive). No version bump here, following the recent PRs.

🤖 Generated with Claude Code