Evaluation of vegetation indices and imaging spectroscopy to estimate foliar nitrogen across disparate biomes

Abstract

AbstractThe nitrogen content in plant foliar tissues (foliar N) regulates photosynthetic capacity and has a major impact on global biogeochemical cycles. Despite its importance, a robust, time, and cost‐effective methodology to estimate variation in foliar N concentration across globally represented terrestrial systems does not exist. Although advances in remote sensing data have enabled landscape‐scale foliar N predictions, improved accuracy is needed to effectively capture variation in foliar N across ecosystems. Airborne remote sensing imagery was analyzed in conjunction with ground‐sampled foliar chemistry data (n = 692), provided by the NEON, to predict foliar N at sites across the United States covering a variety of plant communities and climate types. We developed indices from novel two‐band combinations that predicted foliar N more accurately than existing indices (≈8% improvement across all sites and a 45% improvement in arid sites). Compared with two‐band indices, we increased accuracy and decreased bias of foliar N predictions by using full‐spectrum reflectance information and partial least squares regression (PLSR) models (R2 = 0.638; root mean square error = 0.440). Significant wavelengths included red edge (720–765 nm), near infrared (NIR) reflectance at 1125 nm, and shortwave infrared (SWIR) reflectance at 2050 and 2095 nm, which are regions indicative of foliar traits such as growth type (e.g., leaf area index with NIR) and photosynthetic parameters (e.g., chlorophyll and Rubisco with red and SWIR reflectance, respectively). With the confluence of rapid increases in computing power, several forthcoming or recently launched hyperspectral missions, and the development of large‐scale environmental research observatories worldwide, we have an exciting opportunity to estimate foliar N across larger spatial areas covering more diverse biomes than ever before. We anticipate that these predictions will prove to be invaluable in helping to constrain biogeochemical model uncertainties across a global range of terrestrial ecosystems.