Advices for large jurisdictions#

Some advices for large jurisdictions#

It could be necessary to change the default parameters of the plugin for large jurisdictions (e.g. country scale).

  1. Use a larger Max. distance to forest egde (m), e.g. 6000 m in the Benchmark and MW models.

  2. If you want to save computation time and look at the most important results, only:

    • Calibrate the models on the calibration period, checking calib. period in the box Fit model to data.

    • Predict for the validation period, checking t2 validation in the box Predict the deforestation risk.

  3. For FAR models:

    • Use a larger Spatial cell size (km) for sampling, e.g. 10 km to avoid estimating a too large number of parameters for spatial random effects.

    • Use a larger Spatial cell size interpolation (km), e.g. 1 km.

  4. For the MW models, use only one Window sizes (# pixels), e.g. 21. This should correspond to about 630 m for a resolution of 30 m.

  5. For the validation step, use only one Coarse grid cell sizes (# pixels) and increase its size, using e.g. 300. This should correspond to about 9 km for a resolution of 30 m.