← Input Data Preparation
SeasFire Cube v3 provides a list of 59 climate variables on a regular grid that are related to wildfires in an xArray dataset stored in .zarr format. As described previously in the Data section, a subset of 9 climate variables was chosen. Most of them have time, latitude and longitude as their coordinates. The variable lsm is the only variable having latitude and longitude as its coordinates; this is because the Land-Sea Mask is a binary map of water bodies.
In order to have a dataset with time, latitude and longitude coordinates, the variable lsm must be adjusted. The variable lsm has latitude and longitude as coordinates, since it is a binary map of the globe in which pixels are equal to 1 in land areas and 0 in sea areas. To add the time coordinate to the variable lsm, the map was repeated for each timestamp in the dataset.
Once the dataset has been created, it is necessary to choose a scaler to scale data. Tipically, the data are scaled using Standard or MinMax scalers. Two different approaches can be used to achieve this, one based on maps and one based on arrays containing individual values of the reference statistics for the chosen scaler:
- Standard Scaler: Mean and Standard Deviation maps or arrays;
- MinMax Scaler: Maximum and Minimum maps or arrays.
The climatic variables considered not only change over time, but also geographically. Although it is more intuitive to proceed with the first approach and use maps, the second approach is used because the variations that the climatic variables can undergo in the time frame considered are not significant and, consequently, it is conceivable to use a single representative value for the Mean, Standard Deviation, Maximum and Minimum statistics.