Cross-validate sewageTreated with van Puijenbroek (2019)

[Katsivelisp]

Do the sums of Primary, Secondary, Tertiary (and Quaternary) correspond to van Puijenbroek (2019)? https://www.sciencedirect.com/science/article/pii/S0301479718311824?via%3Dihub

Also check future projections: https://www.sciencedirect.com/science/article/pii/S0301479723017097

[microstijn]

Our baseline is already the 2019 isn't it?

I noticed a bug in how we feed these values to GloWPa.

We use in treatment.csv the four treatment fractions.

FractionPrimarytreatment
FractionSecondarytreatment
FractionTertiarytreatment
FractionQuaternarytreatment

For Uganda these are

0.001, 0, 0, and 0

These are then summed by "combine.py" into sewageTreated (rur and urb), which are both 0.001.

However, Glowpa reads this value as the share of collected sewage that gets treated, eg, we provide treated / total pop, but we need to provide treated / sewered pop.

In the Ugandan case:

sewr = 0.5994 %
treated = 0.1000 %

We provide 0.1 / 100 = 0.01 sewage treated
But it should be 0.1 / 0.5994 = 0.167 sewage treated.

Solution here is to propagate and use the sewr value from the "treatment_fractions/data/original/SSP1.xlsx" and its ilk.

We also have "Countries with all missing data were added based on regional characteristics.", which results in a problem if we use this method. For example Chad and Cameroon have really low sewered %'s, but then get assigned a regional mean of treated 1.795 %. Then treated / sewered > 100 %. There are also some European countries with treated > sewered.

In the imputation case I think we should report them as missing, not use the imputed values. For the european cases I think we can set, if treated > sewered, the resulting ratio to 1.0.