Examining spectral reflectance features related to foliar nitrogen in forests: Implications for broad-scale nitrogen mapping
Introduction
The concentration of nitrogen (N) in foliage is linked to numerous biogeochemical, physiological and ecological processes and serves as a useful indicator of ecosystem metabolism. Foliar N has been repeatedly identified as a useful predictor of photosynthetic capacity, or Amax (Evans, 1989, Field and Mooney, 1986, Reich et al., 1999, Wright et al., 2004); it has been related to stand-level processes such as net primary production (NPP) and canopy light use efficiency (Green et al., 2003, Kergoat et al., 2008, Smith et al., 2002); it provides a widely used measure of herbivore forage quality and susceptibility to defoliation (Jefferies et al., 1994, Mattson, 1980, Peeters, 2002); and can provide direct input to ecosystem models (Ollinger and Smith, 2005, Wythers et al., 2005).
In addition to its influence on carbon assimilation, foliar %N is also tied to the availability of N in soils through mechanisms involving litter decay, net mineralization and plant N uptake (Merilä and Derome, 2008, Ollinger et al., 2002, Parton et al., 2007). This is important given the degree to which humans have perturbed the N cycle globally (e.g., Galloway et al., 2003), and the tendency for N to limit productivity in terrestrial ecosystems (Jandl et al., 2007, Vitousek and Howarth, 1991).
Despite its many important roles, foliar %N is rarely used as a driver in regional- to global-scale analyses. This is, in part, because we lack a reliable means of extending foliar N field measurements to broad-scale spatial patterns. At finer scales (~ 100–1000 km2), the capacity for foliar N estimation has been repeatedly demonstrated using high spectral resolution remote sensing instruments, or imaging spectrometers (e.g., Asner and Vitousek, 2005, Coops et al., 2003, Martin and Aber, 1997, McNeil et al., 2008, Ollinger and Smith, 2005, Smith et al., 2003, Townsend et al., 2003, Wessman et al., 1988), whose narrow bands can record spectral features that result from electron transitions in pigments and/or are associated with other biochemical constituents in foliage (e.g., Curran, 1989, Curran, Kupiec, and Smith, 1997, Kokaly and Clark, 1999). Martin, Plourde, Ollinger, Smith, and McNeil (2008) further demonstrated that relationships between field-measured %N and canopy spectral properties were strongly driven by NIR reflectance patterns, and were consistent enough across boreal, temperate and tropical forests, to allow development of a single, generalized partial least squares (PLS) equation. Still, application of these methods has been limited because presently available imaging spectrometers have swath widths in the range of 10 km or less and because the potential for similar approaches using other instruments has not been thoroughly tested.
There are at least two potential solutions to this problem. The first is development of a space-based imaging spectrometer capable of providing regional to global coverage. Although planning for such instruments is underway (e.g., HyspIRI; Chien et al., 2009, Council, 2007), it will likely be years before data become routinely available. A second possibility is to evaluate the degree to which foliar %N might also be estimated using spectral features available from existing sensors that provide broader spatial coverage. Potential for this approach was suggested by Ollinger et al. (2008) and Ollinger (2011) who observed that reflectance over broad portions of the NIR region was strongly correlated with measured %N in temperate and boreal forests, and by Martin et al. (2008) whose generalized partial least squares %N model was most heavily influenced by reflectance across the NIR plateau from 750 nm to 1250 nm. Although there have been other indications that broad-band spectral features contain information related to variability in canopy N (Gamon et al., 1995, Hollinger et al., 2010, Zhao et al., 2005), no studies to date have focused explicitly on how variability in spectral resolution, spatial resolution and sensor fidelity affect foliar N estimation capabilities.
In this study, we examined the influence of spectral resolution, spatial resolution and sensor fidelity on relationships between observed patterns of foliar %N and canopy reflectance. Sensor fidelity refers to the combination of signal-to-noise ratio (SNR), detector uniformity and stability of electronics in an imaging system that together affect the quality of spectra (e.g., Asner et al., 2007, Chen et al., 2012, Kokaly et al., 2009, Mouroulis and McKerns, 2000). Our analysis draws on a combination of remote sensing and coordinated field measurements from 155 plots within 13 forested research sites across North America. Field measurements were used to evaluate imaging spectrometer data from AVIRIS (high spectral resolution, high spatial resolution, high sensor fidelity); broad-scale sensor data from MODIS (moderate spectral resolution, coarse spatial resolution, high sensor fidelity) and broad-band data from Landsat 5 (coarse spectral resolution, moderate spatial resolution, moderate data fidelity). We also examined relationships between field-measured foliar %N and reflectance in visible and infrared wavelengths on their own, and in several commonly used vegetation indices, in order to evaluate their potential roles in future studies of %N estimation.
Section snippets
Study sites and field data collection
Our analysis used an existing collection of data from several previous investigations (Ollinger, 2011, Ollinger et al., 2008, Martin et al., 2008) that included thirteen North American research sites representing temperate and boreal evergreen needleleaf and deciduous broadleaf and mixed forests, spanning a range of ages. Site descriptions and sampling dates are given in Table 1.
At each site, samples of sunlit canopy foliage were collected from eight to twenty ~ 20 × 20 m plots, generally located
Spectral resolution
The multi-site PLS regression of field-measured, whole-canopy %N with AVIRIS reflectance at the nominal 10 nm spectral resolution (i.e., the benchmark model) resulted in a strong, highly significant relationship (r2 = 0.86, PRESS RMSE = 0.21 and RMSE = 0.19; Table 3a, Fig. 1a) and produced a predicted versus observed trend that did not differ significantly from the 1:1 line. Predicted and observed %N values were lowest in evergreen-dominated forests, highest in deciduous forests and intermediate in
Discussion
The purpose of this study was to evaluate how observed relationships between spectral reflectance and forest canopy %N vary as a function of sensor properties, with the goals of furthering our understanding of factors affecting reflectance and advancing approaches to estimating spatial patterns in %N. As with any empirically based method of remote sensing, results should be applied only within the domain of the data used to generate them and not, for example, extended to tropical forests,
Conclusions
Results from regressions of AVIRIS reflectance on canopy N from 13 forested sites in North America suggest that general methods for %N estimation over broad spatial scales should be possible with future high-fidelity global imaging spectrometers as well as high-fidelity broad-band sensors. We saw little loss of accuracy when AVIRIS spectra were degraded to coarse bandwidths, suggesting that some of the synergy captured with narrowband data is also captured by broad-band sensors. The strength of
Acknowledgments
Funding and AVIRIS data were provided by NASA in support of the following grants: NASA Carbon Cycle Science Award #NNX08AG14G and #NNX14AJ18G, and NASA Terrestrial Ecology Award #NNX11AB88G. Portions of this research were also supported through the Harvard Forest and Hubbard Brook Long Term Ecological Research programs (NSF 1237491 and 1114804, respectively) and the USDA Forest Service, Northeastern States Research Cooperative (Award #15-DG-11242307-053). We also acknowledge the Canadian Carbon
References (96)
- et al.
Variability in leaf and litter optical properties: Implications for BRDF model inversions using AVHRR, MODIS and MISR
Remote Sensing of Environment
(1998) - et al.
Hyperspectral discrimination of tropical rain forest tree species at leaf to crown scales
Remote Sensing of Environment
(2005) Remote sensing of foliar chemistry
Remote Sensing of Environment
(1989)- et al.
Estimating corn leaf chlorophyll concentration from leaf and canopy reflectance
Remote Sensing of Environment
(2000) - et al.
Endmember selection for multiple endmember spectral mixture analysis using endmember average RMSE
Remote Sensing of Environment
(2003) - et al.
Potential of airborne hyperspectral remote sensing to detect nitrogen deficiency and weed infestation in corn
Computers and Electronics in Agriculture
(2003) - et al.
Imaging spectroscopy and the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS)
Remote Sensing of Environment
(1998) - et al.
Identification of invasive vegetation using hyperspectral remote sensing in the California Delta ecosystem
Remote Sensing of Environment
(2008) - et al.
The next Landsat satellite: The Landsat Data Continuity Mission
Remote Sensing of Environment
(2012) - et al.
How strongly can forest management influence soil carbon sequestration
Geoderma
(2007)
Spectroscopic determination of leaf biochemistry using band-depth analysis of absorption features and stepwise multiple linear regression
Remote Sensing of Environment
Characterizing canopy biochemistry from imaging spectroscopy and its application to ecosystem studies
Remote Sensing of Environment
A generalizable method for remote sensing of canopy nitrogen across a wide range of forest ecosystems
Remote Sensing of Environment
Modification of light-acclimation of Pinus sylvestris shoot architecture by site fertility
Agricultural and Forest Meteorology
Changes in hyperspectral reflectance signatures of lettuce leaves in response to macronutrient deficiencies
Advances in Space Research
Large area forest classification and biophysical parameter estimations using the 5-Scale canopy reflectance model in Multiple-Forward-Mode
Remote Sensing of Environment
Using AVIRIS to assess hemlock abundance and early decline in the Catskills, New York
Remote Sensing of Environment
Geostatistical interpolation of SLC-off Landsat ETM plus images
ISPRS Journal of Photogrammetry and Remote Sensing
Landsat-8: Science and product vision for terrestrial global change research
Remote Sensing of Environment
First operational BRDF, albedo nadir reflectance products from MODIS
Remote Sensing of Environment
Assessing canopy PRI for water stress detection with diurnal airborne imagery
Remote Sensing of Environment
Characterization of seasonal variation in forest canopy in a temperate deciduous broadleaf forest, using daily MODIS data
Remote Sensing of Environment
Predicting grain protein content of winter wheat using remote sensing data based on nitrogen status and water stress
International Journal of Applied Earth Observation and Geoinformation
Foliage-height profiles and succession in northern hardwood forests
Ecology
A method for estimating foliage-height profiles in broad-leaved forests
Journal of Ecology
Remote analysis of biological invasion and biogeochemical change
Proceedings of the National Academy of Sciences of the USA
Carnegie Airborne Observatory: In-flight fusion of hyperspectral imaging and waveform light detection and ranging (wLIDAR) for three-dimensional studies of ecosystems
Journal of Applied Remote Sensing
Determination of carbon fraction and nitrogen concentration in tree foliage by near infrared reflectance: A comparison of statistical methods
Canadian Journal of Forest Research
Computation of signal-to-noise ratio of airborne hyperspectral imaging spectrometer
Onboard science processing concepts for the HyspIRI mission
IEEE Intelligent Systems
Evaluating variations of physiology-based hyperspectral features along a soil water gradient in a Eucalyptus grandis plantation
International Journal of Remote Sensing
Leaf angle responds to nitrogen supply in eucalypt seedlings. Is it a photoprotective mechanism?
Tree Physiology
Using vegetation reflectance variability for species level classification of hyperspectral data
International Journal of Remote Sensing
Interactions among nitrogen, carbon, plant shape, and photosynthesis
The American Naturalist
Prediction of eucalypt foliage nitrogen content from satellite-derived hyperspectral data
IEEE Transactions on Geoscience and Remote Sensing
Remote sensing the biochemical composition of a slash pine canopy
IEEE Transactions on Geoscience and Remote Sensing
Precision and accuracy of EO-1 Advanced Land Imager (ALI) data for semiarid vegetation studies
IEEE Transactions on Geoscience and Remote Sensing
Photosynthesis and nitrogen relationships in leaves of C3 plants
Oecologia
Nitrogen detection with hyperspectral normalized ratio indices across multiple plant species
International Journal of Remote Sensing
The photosynthesis-nitrogen relationship in wild plants
Adapting the use of hyperspectral imagery in ocean process studies
The nitrogen cascade
Bioscience
Relationships between NDVI, canopy structure and photosynthesis in three Californian vegetation types
Ecological Applications
Earth Resources Observation and Science (EROS) Center
Remote sensing detection of nutrient uptake in vineyards using narrow-band hyperspectral imagery
Vitis
Processing Hyperion and ALI for forest classification
IEEE Transactions on Geoscience and Remote Sensing
Foliar morphology and canopy nitrogen as predictors of light-use efficiency in terrestrial vegetation
Agricultural and Forest Meteorology
Cited by (59)
An improved light use efficiency model by considering canopy nitrogen concentrations and multiple environmental factors
2023, Agricultural and Forest MeteorologyEstimating leaf nitrogen content by coupling a nitrogen allocation model with canopy reflectance
2022, Remote Sensing of EnvironmentDetermination of foliar traits in an ecologically distinct conifer species in Maine using Sentinel-2 imagery and site variables: Assessing the effect of leaf trait expression and upscaling approach on prediction accuracy
2022, ISPRS Journal of Photogrammetry and Remote SensingField-scale estimation of sugarcane leaf nitrogen content using vegetation indices and spectral bands of Sentinel-2: Application of random forest and support vector regression
2022, Computers and Electronics in AgricultureNitrogen variability assessment of pasture fields under an integrated crop-livestock system using UAV, PlanetScope, and Sentinel-2 data
2022, Computers and Electronics in Agriculture