cmip6_downscaling.methods.bcsd.tasks.spatial_anomalies = <Task: spatial_anomalies>[source]#

Returns spatial anomalies Calculate the seasonal cycle (12 timesteps) spatial anomaly associated with aggregating the fine_obs to a given coarsened scale and then reinterpolating it back to the original spatial resolution. The outputs of this function are dependent on three parameters: * a grid (as opposed to a specific GCM since some GCMs run on the same grid) * the time period which fine_obs (and by construct coarse_obs) cover * the variable We will save these anomalies to use them in the post-processing. We will add them to the spatially-interpolated coarse predictions to add the spatial heterogeneity back in. Conceptually, this step figures out, for example, how much colder a finer-scale pixel containing Mt. Rainier is compared to the coarse pixel where it exists. By saving those anomalies, we can then preserve the fact that “Mt Rainier is x degrees colder than the pixels around it” for the prediction. It is important to note that that spatial anomaly is the same for every month of the year and the same for every day. So, if in January a finescale pixel was on average 4 degrees colder than the neighboring pixel to the west, in every day in the prediction (historic or future) that pixel will also be 4 degrees colder.


UPath to observation dataset chunked in full_time.


UPath to observation dataset interpolated to gcm grid and chunked in full time.


Path to spatial anomalies dataset. (shape (nlat, nlon, 12))