Conventions

Atlite uses the following conventions which are applied for processing geo-spatial and temporal data.

Grid coordinates

According to the xarray conventions, grid coordinates are ordered such that both x and y are ascending.

The coordinates represent points in the center of the corresponding grid cells. Given a cutout, the geographical data of the grid is given by atlite.Cutout.grid, which returns a GeoPandas dataframe with coordinates and geometries. The coordinates are the geographical centroids of the geometries, the grid cells. When initializing a cutout with e.g.

>>> cutout = atlite.Cutout('example', module='era5', x=slice(5,10), y=slice(30,35), time='2013')

the cutout is built with centroids starting at \(5^\circ\) for x (longitude) and \(30^\circ\) for y (latitude). That means the effective covered area by the cutout spans from longitude \(4.875^\circ\) to \(10.125^\circ\) and from latitude \(29.875^\circ\) to \(35.125^\circ\), given the default resolution of \(0.25^\circ\times0.25^\circ\) per grid cell. This information is accessible via the the extent of the cutout:

>>> cutout.extent
array([ 4.875, 10.125, 29.875, 35.125])

Time Points

Following the ERA5 convention, the time-index of time-dependent data refers to the end of the time-span over which was averaged. So, given a time resolution of 1 hour (=averaging window), the time-index 12:00 refers to the time-averaged values from 11:00 to 12:00. For datasets other than ERA5 this convention is not necessarily fulfilled. For example, the SARAH data refers to instantaneous data-points, i.e. data for the time-index 12:00 refers to the momentaneous value of the variable at 12:00. In the implementation, we try to consider this circumstance in order to appropriately align the datasets in order to merge them.