Fig 1.
Spatial coverage of weather stations with precipitation records.
Remotely sensed precipitation estimates are not available above 60° latitude, indicated by the dashed line. The figure uses public domain spatial data from Natural Earth (http://www.naturalearthdata.com/) and public domain location data from the International Tree Ring Databank (https://www.ncei.noaa.gov/products/paleoclimatology/tree-ring).
Table 1.
Global interpolated precipitation products evaluated in this study.
Datasets were generated by the University of East Anglia Climatic Research Unit (CRU), the University of Delaware Terrestrial Precipitation (UDEL), the Global Precipitation Climatology Centre (GPCC), the European Centre for Medium-Range Weather Forecasts Reanalysis v5 (ERA5), the Swiss Federal Institute for Forest, Snow and Landscape Research (CHELSA), the Center for Hydrometeorology and Remote Sensing at the University of California (PERSIANN), and the Department of Civil and Environmental Engineering at Princeton University (MSWEP).
Fig 2.
Spatial coverage of tree ring chronologies and remotely sensed vegetation greenness.
Remotely sensed enhanced vegetation index (EVI) coverage is restricted to pixels with a dominant growing season, allowing for an annual area under the curve estimate from EVI data as a proxy for vegetation productivity, equivalent to tree ring widths. The figure uses public domain spatial data from Natural Earth (http://www.naturalearthdata.com/) and an original spatial layer developed from open-access EVI2 data (https://lpdaac.usgs.gov/products/mcd12q2v006/).
Fig 3.
Temporal data coverage of climate products, remotely sensed EVI data, and tree ring chronologies.
Tree ring chronologies are ordered by end year and truncated at 1900s to match climate data (total of 4422 sites). Shades of grey in lower panel represent the percentiles of the dataset. Only temporally pairwise-complete data was used for climate product comparisons.
Fig 4.
Best regional interpolated precipitation products for a recent 2000–2017 period against remotely sensed vegetation greenness.
The comparisons are based on the strongest correlation with Enhanced Vegetation Index (EVI) annual area under the curve values. The map by ecoregions represents the best performing precipitation product, using the Condorcet winner method, where R2 values of individual EVI pixels are used equivalently to ranked ballots. The figure uses public domain data from Natural Earth (http://www.naturalearthdata.com/) and original results generated in this study.
Fig 5.
Best regional interpolated precipitation products in a long term (1901–2017) evaluation against tree ring records.
Trends over time in validation statistics for the three global interpolated precipitation products that extend back to the 1900. The lines represent variance explained in tree ring with by precipitation using 20-year moving windows.