Check for offshore electricity emissions

[lewisjared]

The problem

Decide if the electricity layer should be land-masked or not.

From Peter's comment here: #43 (comment)

Definition of "done"

• Decide if the electricity layer should be land-masked or not.

Additional context

[lewisjared]

The night time lights layer is causing 12.5% of stationary, transport and industrial emissions to be placed offshore. The discrepancy test fails for these layers if they are land masked. This seems very high so it should be investigated.

[lewisjared]

I'm not sure about the details, but perhaps instead of masking we need to apply land area weighting? Alternatively the landmark could be expanded by x km to provide a buffer for offshore facilities

[prayner]

is this with the national domain or NSW?

[prayner]

I think if masking puts 12.5% of nightlights offshore something has a problem

[lewisjared]

National domain

[prayner]

This is less of a problem than it appears ... using percentage errors
makes things look worse.
I ran the code then calculated totals without (1st number) and with
(second) the landmask. Fields which are masked already of course show
no difference.

OCH4_AGRICULTURE 0.16581610706220368 0.16581610706220368
OCH4_LULUCF 0.496588772438125 0.496588772438125
OCH4_WASTE 0.438749315468585 0.438749315468585
OCH4_LIVESTOCK 2.0541241120884504 2.0541241120884504
OCH4_INDUSTRIAL 0.0027976190476189915 0.0024463792167483014
OCH4_STATIONARY 0.05175595238095241 0.04525801550984454
OCH4_TRANSPORT 0.011190476190476162 0.00978551686699338
OCH4_ELECTRICITY 0.010087008862274222 0.010087008862274222
OCH4_FUGITIVE 1.149471726190475 0.9856538901068982
OCH4_TERMITE 0.47830002347239997 0.47830002347239997
OCH4_FIRE 0.33801258175785165 0.33801258175785165
OCH4_WETLANDS 9.725745624000517 9.725745624000517
OCH4_TOTAL 14.922639367772161 14.750567395663124

The fields which are showing impacts of landmasking are very small
indeed. The difference in the total is about 1% and that's almost all
contributed by fugitives which we don't want to mask. Electricity is
tiny but seems to correspond to a gridpoint which is misbehaving in
early inversion runs so I'll still look at that. The only thing I'm
worried about is whether this behaviour comes from a deeper error e.g.
nightlight regridding going wrong. I'll show a regridded nightlight
map to someone competent and if it's ok will close.