landuse - diversity

[temospena]

          Regarding land use, it may be hard to filter what we really want. Most buildings only have tag building: yes`, without any type.

Originally posted by @temospena in https://github.com/U-Shift/SiteSelection/issues/6#issuecomment-1964611123

[temospena]

Ideas: Use census centroids + POI ? https://rstudio-pubs-static.s3.amazonaws.com/789901_9316efc2603c4d8c816e946b615d8a87.html

Check also this categories: https://observablehq.com/@targomo/exploring-the-15-minute-city

And entropy level viz: https://rpubs.com/aslvova/a-15-minute-city

[temospena]

Use flowers of proximity concept (applied to 15-min city) as categories of POI

(Benjamin Buttner - https://www.eiturbanmobility.eu/wp-content/uploads/2022/11/EIT-UrbanMobilityNext9_15-min-City_144dpi.pdf)

The 15-minutes.city use these tags:

[temospena]

Still regarding the 15min city concept, these places are for those who LIVE in the city. It excludes tourism-related POI, such as hotels. Should we include those? @valenca13

[temospena]

The public transportation stops should be used also in land use or exclusively as a separated input (for the very complex sites)? 🤔

[temospena]

@valenca13 here is a table of the amount of POIs extracted from OSM, with their type, to classify according to the literature in X groups. I would discard all the portuguese names, that are not part of the OSM tags convention.

[valenca13]

@temospena Vamos então usar as 6 grandes categorias mais os edifícios exclusivamente residenciais.

[temospena]

summary(landuse_grid$entropy)
   Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
 0.0000  0.0000  0.0000  0.1229  0.2188  0.8271    1750 

So, the max value for almada is 0.82. There are several 0, with only residential. @valenca13 do we need to normalize then the values before filtering the ones >0.5, or is ready to filter?

[temospena]

other thing, @valenca13 You confirm that /log(N), with N always = number_categories (N=7), and not N = number of categories existing in that cell ?

[valenca13]

@temospena we do not need to normalize the values, since it is already from 0 to 1. Not having a 1 value makes sense in terms of practice. You will never have the same number of different types of land use in a certain area. So the values are very low due to the number of categories. Let's try out like this. If it does not work, we aggregate the categories.

[valenca13]

@temospena N is always equal to 7, because we want to evaluate how many possible categories there are in a cell. If only 4 categories are present, then the indicator should take into consideration that there are 3 missing. Depending on the values, it may be coherent to aggregate the categories into the flower of proximity concept. Let's think a little bit.