Monthly average precipitation data for the
Rocky Mountain region of the US.
General information about the format.
Large spatial prediction problems and nonstationary random fields
For all these data sets the precipitation measurements have been
transformed by the square root and centered by the 30 year monthly mean
The main challenge for these data is to extrapolate to a fine
regular grid and quantify the uncertainty in the estimates.
For a spatial analysis of these data and some
context for why it is important see
Nychka, D., Wikle, C., Royle, J.A. (1999)
the technical report in
or just the
There is also a shorter submitted manuscript that gives less explanation
about 2-d wavelet bases:
1963 August monthly precipitation at 800 sites
(Not all sites are reporting in every month and missing
values are coded as NA.)
- download the data,
- get site locations as a
845 X 2 table of longitudes and latitudes.
These locations are in the same order as the rows of
the data table.
- get Cross-validation site
logical variable. This is a vector where
0 = site was used for fitting surface,
1 = site was reserved for cross-validation of
There are 45 total CV sites (some are
missing values) that are used in the analysis of Nychka et al.