Remotely sensed crown nutrient concentrations modulate forest reproduction across the contiguous United States
Data files
May 22, 2024 version files 31.96 KB
Abstract
Global forests are increasingly lost to climate change, disturbance, and human management. Evaluating forests' capacities to regenerate and colonize new habitats has to start with the seed production of individual trees and how it depends on nutrient access. Studies on the linkage between reproduction and foliar nutrients are limited to a few locations and few species, due to the large investment needed for field measurements on both variables. We synthesized tree fecundity estimates from the Masting Inference and Forecasting (MASTIF) network with crown nutrient concentrations from hyperspectral remote sensing at the National Ecological Observatory Network (NEON) across the United States. We evaluated the relationships between seed production and foliar nutrients for 56,544 tree-years from 26 species at individual and community scales. We found a prevalent association between high foliar phosphorous (P) concentration and low individual seed production (ISP) at the continental scale. With-species coefficients to nitrogen (N), potassium (K), calcium (Ca), and magnesium (Mg) are related to species differences in nutrient demand, with distinct biogeographic patterns. Community seed production (CSP) decreased four orders of magnitude from the lowest to the highest foliar P. This first study on hyperspectral imagery indicates promise for future monitoring of reproductive potential. The fact that both ISP and CSP decline at high foliar P levels has immediate applications in improving forest demographic and regeneration models by providing more realistic nutrient effects at multiple scales.
README: Data for: Remotely sensed crown nutrient concentrations affect forest reproduction across the United States
The nutrient data are derived from hyperspectral remote sensing, where each column contains a specific nutrient concentration from the Partial Least Squares Regression (PLSR) model. Each row is a NEON plot. The fecundity data are derived from the Masting Inference and Forecasting (MASTIF) network, where each column contains a specific species and each row is a NEON site. The code provides instructions on how PLSR can be used and how a generalized Joint Attribute Model (GJAM) is fitted to the two data. The code can be executed on R studio.
Description of the data and file structure
There are two CSV files, where nutrient variables (i.e., the predictors, unit mg/g) and fecundity estimation (i.e., the response, unit g per m^2) are saved in each file.
The five nutrient predictors include Nitrogen (N), Phosphorus (P), Potassium (K), Magnesium (Mg), and Calcium (Ca).
The response file lists the species in the first row.
plot.ID is the plot identity designation.
Data was derived from the following sources:
- EcoSIS Library
- Masting Inference and Forecasting Network
The associated code is available on Zenodo at the link provided in the Related Works section of this dataset submission.
Methods
The fecundity data from the 13 NEON sites are part of the Masting Inference and Forecasting (MASTIF) network. MASTIF includes two types of raw data, seed traps (ST) and crop counts (CC). ST data were the number of seeds from seed traps associated with individual trees from the mapped stands at the 13 NEON sites. Data compilation, modeling, and computation are open-access in the R package MASTIF, with more details provided in Clark et al., 2019, 2021.
We generated wall-to-wall foliar nutrient maps following Wang et al., 2020 by applying the Partial Least Squares Regression (PLSR) coefficients to the hyperspectral surface reflectance data. The coefficients are available at https://ecosml.org/package/github/EnSpec/NEON-Trait-Models. Five nutrients, including Nitrogen (N), Phosphorus (P), Potassium (K), Magnesium (Mg), and Calcium (Ca), were analyzed based on their important roles in plant reproduction. The unit of the nutrient is mg per g.