Add conversion factors across dimensions (eg. from mass to volume) for specific flows

[fjuniorr]

See https://github.com/fjuniorr/flowmapper/issues/24#issuecomment-1833224902

[cmutel]

I don't think we need a general strategy for now - we can hard code the unit conversions we need for the real data we have. There aren't that many:

[   
    {
        'source': {'unit': 'Bq'}, 
        'target': {'unit': 'kBq'}, 
        'conversion_factor': 1e-3
    },
    {
        'source': {'unit': 'ton'}, 
        'target': {'unit': 'kg'}, 
        'conversion_factor': 1e-3
    },
    {
        'source': {'unit': 'g'}, 
        'target': {'unit': 'kg'}, 
        'conversion_factor': 1e3
    },
    {
        'source': {'unit': 'mg'}, 
        'target': {'unit': 'kg'}, 
        'conversion_factor': 1e-6
    },
]

[fjuniorr]

I'm working on this in #42 by creating a new get_conversion_factor that hardcodes the conversion factors that we have so far.

For the current matching flows in agribalyse-3.1.1-biosphere and industry-2.0-biosphere we are failing to generate conversions factors for:

SourceUnit	TargetUnit	SourceFlowName	TargetFlowName	ConversionFactor
kg	kBq	Lead-210/kg	Lead-210	?
kg	kBq	Radium-226/kg	Radium-226	?
kg	kBq	Radium-228/kg	Radium-228	?
kg	kBq	Ruthenium	Ruthenium-103	?
kg	m3	Gas, mine, off-gas, process, coal mining/kg	Gas, mine, off-gas, process, coal mining	?
m2a	m2*year	Occupation, *	Occupation, *	1
m3y	m3*year	Volume occupied, reservoir	Volume occupied, reservoir	1
ton	m3	Water	Water	1
g	m3	Water	Water	0,000001
kg	m3	Water	Water	0,001
l	m3	Water, *	Water, *	0,001

@cmutel should we add these as well? Could you help me fill this table and check the values?

[cmutel]

I will try, but the conversion factors aren't easily found. I will ask my network.

[fjuniorr]

Got it. I've added the conversion factors for occupation[^1] and water and renamed the issue to better reflect what is missing.

[^1]: For reference here is where I saw that m2a is equal to m2*year and m3y is equal to m3*year

[fjuniorr]

After https://github.com/fjuniorr/flowmapper/pull/51 I removed the following conversions factors across dimensions:

if s.name == 'Ammonia, as N' and t.name == 'Ammonia':
        result = 17 / 14
elif s.name.startswith('Water') and t.name.startswith('Water') and s.unit == 'ton' and t.unit == 'm3':
    result = 1.0
elif s.name.startswith('Water') and t.name.startswith('Water') and s.unit == 'g' and t.unit == 'm3':
    result = 1e-6
elif s.name.startswith('Water') and t.name.startswith('Water') and s.unit == 'kg' and t.unit == 'm3':
    result = 1e-3
elif s.name.startswith('Water') and t.name.startswith('Water') and s.unit == 'l' and t.unit == 'm3':

[fjuniorr]

I've not read the docs carefully but the pint package used for https://github.com/fjuniorr/flowmapper/issues/55 has support for conversions across dimensions using contexts.

[fjuniorr]

This question appears to be useful python - How to use transformations in pint package correctly? - Stack Overflow

[fjuniorr]

@cmutel now it's clearer to me that conversion_factor should be a method of Flow, not Unit. With this change we can properly use information from the flows to handle both conversions across dimensions and others (such as in https://github.com/fjuniorr/flowmapper-ci/issues/13 when to be modelled as pure elements the content of one substance in terms of other is not 1).