DPastCliM-NA dataset: a high-resolution, two-millennium monthly downscaled climate dataset for North America

The DPastCliM-NA (Downscaled Past Climate reconstruction at a Monthly timestep for North America) dataset provides monthly downscaled reconstructions of past climate for North America, obtained via a Principal Component Regression (PCR) based approach. 

Temporal coverage: 0–2014 CE
Timestep: monthly
Spatial coverage: North America (lon=[-175°,-50°], lat=[5°,72°]) 
Variables: surface air temperature (2 m; tas) and precipitation (pr)
Spatial resolution: 0.2° × 0.2° (NA_020 grid)
ESM origin: MPI-ESM1-2-LR (Mauritsen et al., 2019)
PMIP4 experiments: past2k and historical  
Observational datasets: GHCN-m (Menne et al., 2018; Applequist et al., 2024)

Citation instructions

Please cite both the present dataset and the associated reference paper (<doi>).

Citation examples:

Guaita, P.R., Crippa, P. A high-resolution, two-millennium monthly downscaled climate dataset for North America. Sci Data (2026). https://doi.org/10.1038/s41597-026-07572-x

Guaita, P. R., & Crippa, P. (2025). DPastCliM-NA dataset: a high-resolution, two-millennium monthly climate reconstruction for North America (x.x.x) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.17245353

DPastCliM-NA timetep note

Please note that time t is in expressed in months since 0 CE (included), i.e.:

year=floor((t-1)/12)

month=t-12*floor((t-1)/12)

(e.g. t=15 corresponds to March 1 CE)

Files format and description

1. <variable name "tas | pr">_metadata_table_<grid ID "NA_020">.csv: station metadata table with: ID, latitude ([-90,90]), longitude ([-180,180]), elevation (metres), and flags for calibration (flag_cal), model_selection (flag_ms), testing (flag_test).
2. <variable name "tas | pr">_<ESM model "MPI-ESM-1-2-LR">_<PMIP4 experiment>_statreal_PCRdownscaled_<grid ID "NA_020">.nc: netCDF-4 file containing the DPastCliM-NA station-wise realizations. The first main output is the reconstructed timeseries in every calibration stations ('Ods_hat_mat') a (station x timestep) matrix, with rows in the same order as indicated in the metadata table. NaN rows indicate stations that were not reconstructed (testing-only stations). The second main output is the 95% Prediction Intervals associated with every reconstructed time-series ('PI_mat') a (station x timestep x bound) (lower/upper) matrix.
3. <variable name "tas | pr">_<ESM model "MPI-ESM-1-2-LR">_<PMIP4 experiment>_PCRdownscaled_<grid ID "NA_020">.zip: archive containing multiple netCDF-4 files with monthly timestep gridded fields (from realizations). Each file covers a 100-year period; suffixes (e.g., 1-1200) indicate the range of timesteps included.
4. <variable name "tas | pr">_<ESM model "MPI-ESM-1-2-LR">_<PMIP4 experiment>_PI_PCRdownscaled_<grid ID "NA_020">.zip: same as (3), but containing the corresponding gridded prediction intervals.

Usage advices

• The gridded products (*_PCRdownscaled_* and *_PI_PCRdownscaled_*) were produced using a computationally efficient method (natural neighbor interpolation) from the station-wise values. These products are intended as a ready-made gridded product for continental-scale assessments, but should be used with caution in regions with sparse calibration coverage or complex topography (e.g. high latitudes, Mesoamerica).
• The station-wise products (*_statreal_PCRdownscaled_*) allow users to perform custom gridding or spatial aggregation, enabling applications that require finer control or alternative interpolation methods.
• The DPastCliM-NA products inherit the long-term variability and trends of the parent model (MPI-ESM1-2-LR). Users should account for this when comparing against independent reconstructions or other model ensembles.
• The DPastCliM-NA products are not designed for direct temporal alignment with proxy-based reconstructions. Although the underlying ESM simulations capture broad climatic patterns, differences in event timing, intensity, and persistence are to be expected.
• Each downscaled time series (and gridded field) in DPastCliM-NA is accompanied by a 95% Prediction Interval (PI), which quantifies the uncertainty associated with the reconstructed monthly climate values. The lower and upper PI bounds indicate the range within which the true climate variable (temperature or precipitation) is expected to lie, given the uncertainty in the regression model and residual variability.