Modeled annual seed production estimates for Pinus edulis 1900-2024
Data files
Feb 10, 2025 version files 71.03 MB
-
mastcast_data_package_(2).zip
71 MB
-
pinyon_mast_metadata_revised.xml
27.21 KB
-
README.md
2.99 KB
Abstract
These data are gridded estimates of annual seed production (or mast seeding intensity) in a conifer species of the southwestern United States. The extent of these predictions cover the range of Pinus edulis in the U.S. states of Colorado, New Mexico, Arizona, and Utah between 1900-2024. Also included is a map of the variance in bootstrapped predictions for the covered range, and the unpublished training data used in model validation. All other data sources are publically available and free to use.
This data is associated with the publication Mast Hindcasting reveals pervasive effects of extreme drought on a foundational conifer species, in New Phytologist, 2025.
Description of the data and file structure
Data descriptions can be found in the associated metadata.
Validation Data contains a csv file, containing testing data used in model validation. These data were collected for this study and have not been previously published elsewhere. These data are associated with this data release by relation to the publication and associated analyses but should be considered auxiliary to the primary components.
The columns in the Validation Data are as follows:
site - unique ID
year - year of a given cone scar as observed on a given branch. This must be adjusted +1 to reflect cone production in the year of maturation.
cwd - 30 year average annual climatic water deficit, a measurement of cumulative drought stress, derived from TerraClimate
mon - 30 year average annual proportion of monsoon precipitation falling between July-Sep, derived from PRISM
cone_total - the number of cone scars counted multiplied by the number of cone bearing branches
fert_vpd - z-score scaled vapor pressure deficit during the year of cone pollination (year of cone scar observation, -1 from maturation)
aug_vpd - z-score scaled vapor pressure deficit during the year of cone initiation (-1 year of cone scar observation, -2 from maturation)
x - Longitude (EPSG:4326)
y - Latitude (EPSG:4326)
scaled_cones - a relativized measure of annual cone production at a given site compared to all years measured. These values are standardized between 0-1 to reflect low cone years (closer to zero) and high cone years (closer to one). Predictor variable used in most/all models.
____________________________________
Raw predictions - the raw model output and associated bootstrapped confidence intervals
Prediction percentiles - transformed model predictions, constrained to 0-1 representing the relative ranking of a given mast year at a given pixel relative to all other years at that same location.
Raw predictions includes a file - model out, which is the primary datafile associated with this release. The folder “bootstrapped CI” contains the 5th and 95th percentiles of bootstrapped predictions, as well as the confidence interval width for each year, and the mean CI width for all years.
Prediction percentiles includes a similar file - scaled, which is the secondary datafile associated with this release. Each raster layer contains values between 0-1. These values represent the rank-order mast year relative to all other years at a given pixel. The bootstrapped CI’s are identical to those above, but again, transformed to values between 0-1.
Email awion@usgs.gov or andreas.wion@gmail.com with data quality issues or questions.
These data were collected with the purpose of validating an established forecast model of cone production in this species. The model was parameterized with existing data on seed production from Wion et al. 2020 (Ecography), and these known relationships between seed production and weather were extrapolated across the range of this species and hindcasted into the past. These data may be used to understand where climatic conditions were most likely favorable to mast seed production in this species for a given year, and where those mast years are likely to have occurred.