search

CSIRO-enviro-informatics / loci-excelerator

A reapportionment tool for the Loc-I project
1 stars 1 forks source link

Handling cases where there are 'spatial holes' #22

Open jyucsiro opened 5 years ago

jyucsiro commented 5 years ago

There are cases where there will be spatial holes between datasets eg. ASGS MB and GF CC. due to rivers and water bodies.

How do we handle these cases and communicate their implications?

jyucsiro commented 5 years ago

Example of ASGS 2016 SA1 and GF CC v2.1 in northern part of Tasmania

image

shaneseaton commented 5 years ago

Related to #24

benjaminleighton commented 5 years ago

The difference between this and https://github.com/CSIRO-enviro-informatics/loci-excelerator/issues/24 is this is not going to go away but otherwise they are similar. We should be able to bulk report out cases where we are not going to get mass balance across a particular linkset path.

Some options:

  1. we could use phonebook to highlight every say SA1 to CC where this occurs, and if satisfied that they are all rivers we could link to a general explanation every time a mass balance error occurs for a particular case.

  2. we could provide an option to either "distribute to missing target areas" or "distribute to known target areas" presumably only one of these is happening now @shaneseaton. e.g for "distribute to missing target errors" if you distributing from an asgs SA1 with an id of 1234 with an area of 1 to a series of cc 1 2 3 and 4 each with an area 1/5 then you could have an output "1234-Unknown" which would get the 1/5 area unaccounted for.

benjaminleighton commented 5 years ago

Thinking more we could probably precompute these "Unknown" areas at linkset build time. That would be handy for a number of reasons: Reporting linkset coverage being the main one, and removing the need for mass balance calculations at runtime being another. This relates to a broader issue of putting fractional distributions into linkset relationships but that is another (github) issue potentially.