ghislainv / forecasting-deforestation-mada Goto Github PK

Add random forest projection

Use 2010-2015 or 2010-2016 for observations

Check computation of deforested area obs vs. pred. Plot also the increase in correlation with decreasing resolution (from 30 m) to total area.

Plot fitted vs. observation. Add observation points and fitted lines for relationship between probability and factor (distance to roads for example).

Write a proper function to compute random effects only inside country. No random effects outside borders. In this function, compute an array with initial cell number (alla landscape) and corresponding cell number for iCAR. Then remove data (neighbors id and number of neighbors) for cells not in iCAR.