Adapt matching frames to include only relevant control cells

[Jo-Schie]

I notice that our current matching frames include all cells irrespective if they are PAs and leakage cells. This has to be adapted asap.

Fyi @yotaae and @melvinhlwong

[Jo-Schie]

The following routine is suggested to be implemented in the grid creation scripts.

• [x] Load supported areas
• [x] Match treatment year
• [x] Load grid subsets (gridcells within buffer zones)
• [x] get centroids of the grid polygons (point data)
• [x] subset supported areas for treatment year
• [x] project supported areas layer
• [x] create 50 km buffers around supported areas
• [x] reproject layer to match CRS of grid centroids
• [x] intersect supported areas with grid polygons
• [x] exclude supported grid polygons from data
• [x] intersect remaining gridpoints with year specific buffered supported areas
• [x] delete those grids from the remaining treatment cells.

Output:

• spatial grid dataset that contains columns contols_2000 to controls_2020 with 1/0 dummy.

[yotaae]

Hi @Jo-Schie @melvinhlwong,

I have checked the map including treated cells from both the fishnet dataset (which includes the geom data) and the new matching frames. Specifically, I have checked the cells for the year 2007.

Quick info regarding the colors in the screenshot below:

• yellow lines: treated PAs (first year in 2007) according to KfW data
• red cells: treated cells according to fishnet (matchingyear == 2007)
• blue cells: treated cells according to matching frame 2007

Issue: It seems that the matching frame is missing some treated cells -> if all cells in treated PAs were included in the matching frames, the red cells would not be visible (the matching-frame-layer is above the fishnet-layer). The fishnet layer, however, seems to include the correct treated cells.

Note: Treatment in the fishnet data equals the "intersect" variable -- cells on the border are thus included.

[yotaae]

Help for debugging:

Table with number of treatment cells in the fishnet:

[Jo-Schie]

First impression. There are somehow some areas, that are not part of the whole matched polygon IDs, even before creating the matching frames..

The reasons seems to be a problem in the intersection of the Grid with WDPA already earlier in the 01_grid_creation.R script.

Nevertheless, this does not explain the absence of treatment cells as in yotas comment. So I guess I fix the matching frames creation first and then I will go for this initial issue of the intersection.

[Jo-Schie]

@yotaae

These are the cells that are now returned as treatment in 2007

So I can confirm, that at leat for 2007 everything is fine. Please check on your side as well for eventual bugs. Did you maybe plot the cells after matching?

Please note, that your comparison table is not really good to cover the issue, because a lot of the poly_ids in the years are somehow dupliacates. Please test again, if the problem persists.

[Jo-Schie]

First impression. There are somehow some areas, that are not part of the whole matched polygon IDs, even before creating the matching frames..

The reasons seems to be a problem in the intersection of the Grid with WDPA already earlier in the 01_grid_creation.R script.

Nevertheless, this does not explain the absence of treatment cells as in yotas comment. So I guess I fix the matching frames creation first and then I will go for this initial issue of the intersection.

This issue seems not to persist anymore when applying the relevant filters to exclude marine areas, proposed areas, point polygons and UNESCO-MAB Biosphere Reserve.

wdpa_LA <-
  wdpa_LA %>%
  filter(DESIG_ENG != "UNESCO-MAB Biosphere Reserve") %>%
  filter(STATUS != "Proposed") %>%
  filter(GEOMETRY_TYPE != "POINT")

Hence I see this issue as resolved. See also this map, which shows the filtered areas:

[Jo-Schie]

I also checked the following years and they returned correct results: 2009, 2010, 2012, 2013 and 2016. The others I did not control because partly they have a very high number of polygons and it is not possible to render the resulting data in mapView but I see no indication why there should be problems left. Data checks are now included in the script 03_db_creation.R which allow to control the results after creating the matching frames.

[Jo-Schie]

I created now the routine to only include leakage zones for the respective years. The results look like this (example year 2011, snippet is from Colombia I guess)

After reimport and a check on your side @yotaae, we can close this issue.