Description Usage Arguments Value Multiple runs/files Volcanic forcing
View source: R/get_predictors.R
get.predictors
builds basis functions as predictors for
the quantile regression in the quantile mapping climate
projection method of Haugen et al. 2018. The basis
functions are constructed such that the resultant quantile
estimate is smoothly varying and depends on:
the seasonal cycle (dependent only on month-of-year), estimated by a periodic b-spline
the long-term trend (dependent only on the year index), estimated by a natural cubic spline
an interaction between the seasonal cycle and long-term trend (to capture changes in the seasonal cycle)
The degrees of freedom for each of these basis functions are set through the
parameters df.x
, df.t
, and df.xt
, respectively, or through a 3 x 1
vector dfs
.
1 2 3 4 |
n_files |
number of runs (if |
dfs |
degrees of freedom - a |
df.x, df.t, df.xt |
alternatively, input the dfs separately,
with seasonal |
year.range |
|
get.volc |
whether to add a volcanic degree of freedom |
lat |
only necessary if |
save.predictors |
if true, the basis functions are saved
to an .RData file (def: |
save.fn |
if |
a numerical matrix
get.predictors
is specifically designed to allow for basis
functions for quantile regression across multiple data sets that
span the same time period. In the context of the quantile mapping
climate projection method, these multiple data sets are multiple
runs of the same model over the same time period. In this case, the
basis vectors are repeated end-to-end n_files
times.
The get.volc
input additionally allows for the addition
of a historical volcanic CO2 forcing as a predictor for the
quantile regression. This volcanic forcing is taken from the
XXXXXX hindcast from CCSM4XXXX. If get.volc=TRUE
, then a
latitude must be specified through the lat
parameter.
Since the volcanic forcing is only available 1850-2008, this
is discouraged for values beyond 2008...
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.