... | @@ -254,7 +254,7 @@ If cross-validation is chosen, leave-one-out cross-validation will be applied. T |
... | @@ -254,7 +254,7 @@ If cross-validation is chosen, leave-one-out cross-validation will be applied. T |
|
|
|
|
|
The Downscaling module performs downscaling on the anomalies making use of the functions in the [CSDownscale package](https://earth.bsc.es/gitlab/es/csdownscale). It accepts the output of the Anomalies module as input and also requires the recipe. The module applies the selected downscaling method to the hindcast anomalies using observed anomalies as the reference and returns the downscaled data and its metadata as an s2dv_cube object.
|
|
The Downscaling module performs downscaling on the anomalies making use of the functions in the [CSDownscale package](https://earth.bsc.es/gitlab/es/csdownscale). It accepts the output of the Anomalies module as input and also requires the recipe. The module applies the selected downscaling method to the hindcast anomalies using observed anomalies as the reference and returns the downscaled data and its metadata as an s2dv_cube object.
|
|
|
|
|
|
Additionally, forecast data can be downscaled. In this case, while hindcast and observational data are used to train the model, the relationship obtained from the training period is utilized to downscale the forecast data.
|
|
Additionally, forecast data can be downscaled. In this case, while the hindcast and observational data are used to train the model, the relationship obtained from the training period is utilized to downscale the forecast data.
|
|
|
|
|
|
The output of the main function, **Downscaling()**, is a list containing the downscaled hindcast (forecast) and observations, named **hcst** (**fcst**) and **obs**.
|
|
The output of the main function, **Downscaling()**, is a list containing the downscaled hindcast (forecast) and observations, named **hcst** (**fcst**) and **obs**.
|
|
|
|
|
... | @@ -272,7 +272,7 @@ This specification is a mandatory requirement and must be defined in the recipe |
... | @@ -272,7 +272,7 @@ This specification is a mandatory requirement and must be defined in the recipe |
|
|
|
|
|
When selecting the downscaling method `'intbc'`, both interpolation and bias correction methods should be specified; for 'intlr,' both interpolation and linear interpolation methods are required; and for 'logreg,' both interpolation and logistic regression methods should be provided. Leave-one-out cross-validation is always applied for all the methods in the module.
|
|
When selecting the downscaling method `'intbc'`, both interpolation and bias correction methods should be specified; for 'intlr,' both interpolation and linear interpolation methods are required; and for 'logreg,' both interpolation and logistic regression methods should be provided. Leave-one-out cross-validation is always applied for all the methods in the module.
|
|
|
|
|
|
For the analogs method, downscaling can also be applied by using a large scale variable as the predictor. In this scenario, the function identifies the day in the observation data that closely resembles the large-scale pattern of interest in the model. When it identifies the date of the best analog, the function extracts the corresponding local-scale variable for that day from the observation of the local scale variable. The used local-scale and large-scale variables can be retrieved from independent regions. If this approach is desired to be used, in addition to local-scale observations (obs), observation, and hindcast (forecast) data of the large-scale variable should also be provided. For example, when downscaling is performed via the large-scale variable, the s2dv_cube objects that need to be provided within the list object (i.e., "data" in this example) are as follows: data\$obs, data\$obsL, data\$hcstL (data\$fcstL; in case forecast downscaling is aimed).
|
|
For the analogs method, downscaling can also be applied by using a large scale variable as the predictor. In this scenario, the function identifies the day in the observation data that closely resembles the large-scale pattern of interest in the model. When it identifies the date of the best analog, the function extracts the corresponding local-scale variable for that day from the observation of the local scale variable. The used local-scale and large-scale variables can be retrieved from independent regions. If this approach is desired to be used, in addition to local-scale observations (obs), observation and hindcast (forecast) data of the large-scale variable should also be provided. For example, when downscaling is performed via the large-scale variable, the s2dv_cube objects that need to be provided within the list object (i.e., "data" in this example) are as follows: data\$obs, data\$obsL, data\$hcstL (data\$fcstL; in case forecast downscaling is aimed).
|
|
|
|
|
|
Another option in the recipe is **Workflow:Downscaling:target_grid**. This argument is a character vector indicating the target grid (i.e., to which grid system the dataset will be downscaled). It can be the path to a netCDF file or a grid string or grid description file accepted by CDO.
|
|
Another option in the recipe is **Workflow:Downscaling:target_grid**. This argument is a character vector indicating the target grid (i.e., to which grid system the dataset will be downscaled). It can be the path to a netCDF file or a grid string or grid description file accepted by CDO.
|
|
|
|
|
... | | ... | |