nk027 / deforestation_mt Goto Github PK

The SPEI is added, but variables derived are currently limited to the following (per year):

• mean
• sd
• quantiles 0.2, 0.5 & 0.8
• iqr

Municipality borders differ wildly in the early 2000s (even new municipalities were added) and slightly after 2005.

When limiting oneself to a subset after 2005 these border changes may be negligible. Adjustments are small and due to e.g. changes of river courses.

The reference models are now computed in a Bayesian fashion. Analysis of these will need to be adjusted accordingly.

Some municipios report no crop production leading to an issue when calculating yields / prices. This could be counteracted by aggregating or subsetting.

• Automatise tables
• Fix fonts used

Even after projecting the shapefile (long-lat areas are only estimates) the calculated area is partly significantly (>10% ) smaller than the area implied by the amount of contained pixels.

This is due to the fact that while MOD13Q1 is supposed to have a pixel size of 250m^2 it actually seems to be 231.6564m^2 (according to the raster package and QGIS). After accounting for that the areas match very closely (+/- 1%).

• The municipality plot could maybe use some playing around.
• Maybe add more RMSE results?
• More alternative models?
• Moran's Plot?
• Local Moran's I?
• More plots of the LUC maps?

Latest revision changed that (nicer scale), still need to mention it in the paper.

None found, as of 1b40558!

To calculate the areas (precisely) of the municipalities the shapefile needs to be projected.

• Distinguish supplementary material and online appendix
• Add data plots
• Use tmap for municipality plot

Need to re-run calculations for that.

IIASA data on crop yields and travel time is included, but currently limited to the mean variables.

A possible issue is a rather large number of NAs for the yield data.