Look into better integration with GHS-POP timeseries

[cjacobscrisioni]

There are a number of discrepancies between GHS-POP and 2BURP projections. Key issues to look into (at a later stage):

• discretisation increases noise and differences with GHS-POP. The reason to do this is expected leakage to low population fractions. What are the modelling implications of not making population discrete?

• discretisation also seems to cause sizeable displacement of population (from dispersed population to clusters). Can we reduce displacement when creating a discrete population grid? For example by combining suitbility with a potential of the rounded-down population fractions.

• To my best knowledge, population allocation in non-builtup areas is allowed. Are we allocating population in non-builtup?

• Conversion from lat/lon to equal area also causes noise. What are the costs of modelling Mollweide equal area instead of lat/lon?

[jipclaassens]

punt 3: attribute<Person> MaxAllowed_Population (CompactedDomain) := MaxAllowed_Population_Density * NewState_BuiltUp/BuiltUp_Residential_Area;

[jipclaassens]

punt 1: effect van discretisering maakt niet veel uit voor dou definities. Communicatie Marcello.

[jipclaassens]

punt 2: ondervangen door een nth_element op de initial pop map toe te passen en na allocatie opnieuw.

[jipclaassens]

punt 4:

108