Coupled Model Intercomparison Project Phase 5 (CMIP5) University of Wisconsin-Madison Probabilistic Downscaling Dataset Dataset for Machine Learning
Install DagsHub:
pip install dagshub
To stream this data directly on DagsHub
from dagshub.streaming import DagsHubFilesystem
fs = DagsHubFilesystem(".", repo_url="https://dagshub.com/DagsHub-Datasets/noaa-uwpd-cmip5-dataset")
fs.listdir("s3://noaa-nws-uwpd-cmip5-pds")
Description
The University of Wisconsin Probabilistic Downscaling (UWPD) is a statistically downscaled dataset based on the Coupled Model Intercomparison Project Phase 5 (CMIP5) climate models. UWPD consists of three variables, daily precipitation and maximum and minimum temperature. The spatial resolution is 0.1°x0.1° degree resolution for the United States and southern Canada east of the Rocky Mountains.
The downscaling methodology is not deterministic. Instead, to properly capture unexplained variability and extreme events, the methodology predicts a spatially and temporally varying Probability Density Function (PDF) for each variable. Statistics such as the mean, mean PDF and annual maximum statistics can be calculated directly from the daily PDF and these statistics are included in the dataset. In addition, “standard”, “raw” data is created by randomly sampling from the PDFs to create a “realization” of the local scale given the large-scale from the climate model. There are 3 realizations for temperature and 14 realizations for precipitation.
The directory structure of the data is as follows
[cmip_version]/[scenario]/[climate_model]/[ensemble_member]/
The realizations are as follows
prcp_[realization_number]_[year].nc
temp_[realization_number]_[year].nc
The time mean files averaged over certain year bounds are as follows
prcp_mean_[year_bound_1]_[year_bound_2].nc
temp_mean_[year_bound_1]_[year_bound_2].nc
The time-mean Cumulative Distribution Function (CDF) files are as follows
prcp_cdf_[year_bound_1]_[year_bound_2].nc
temp_cdf_[year_bound_1]_[year_bound_2].nc
The CDF of the annual maximum precipitation is given for each year in the record prcp_annual_max_cdf_[start_year_of_scenario]_[end_year_of_scenario].nc
Additional information
Documentation
Update frequency
Periodically, as new data becomes available or when corrections are implemented.
Managed by
License
Open Data. There are no restrictions on the use of this data.