Spectroscopic Detection of Nitrogen Concentrations in Sagebrush

Spectroscopic Detection of Nitrogen Concentrations in Sagebrush: Implications for Hyperspectral Remote Sensing

 

Interactions between plants and incident solar radiation generate unique reflection and absorption features across the electromagnetic spectrum.  Imaging spectroscopy can resolve detailed spectral signatures, portions of which may be linked to plant biochemistry, including N concentration.  Foliar N concentrations are strongly related to carbon uptake/ cycling and rates of net photosynthesis and primary production.  Spectroscopy lab experiments and empirical studies link foliar N concentrations to narrow absorption regions in the visible and near-infrared.  Signal enhancement methods (e.g., derivative transformation, continuum removal, band ratioing) have been used to estimate N concentrations in dry leaves and hyperspectral remote sensing studies have successfully mapped canopy N concentrations for forested ecosystems.  The extent to which similar approaches can be extended to sparsely vegetated systems has been largely unexplored.  Known challenges are related to the dampening or masking of features of interest by water absorption, soil reflectance, open canopy, and leaf angles.

 

Objectives

  • Relate sagebrush leaf and shrub N concentrations to corresponding spectral responses

  • Examine differences between sagebrush N concentration and spectral response at leaf and shrub scales

  • Identify transformed bandwidth intervals most closely related to N concentrations 

  • Examine the strength at which narrow absorption features are expressed using a field radiometer versus an unmanned aerial vehicle-based sensor platform (PIKA, 400 – 900 nm); evaluate results in the context of extending sagebrush N concentration to a larger scale project.

 

Accomplishments through 2009

 

From May to October 2009, the following measurements were collected in the field for each of 35 individual, spatially isolated sagebrush (Wyoming and basin big sage [Artemesia tridentata ssp. wyomingensis and A. t. ssp. tridentata]):

 

  • Absolute canopy reflectance measurements (350-2500 nm) using a field spectroradiometer (ASD)

  • Photographs oriented orthogonal to the ground (for calculating proportional leaf area)

  • Sagebrush green leaf samples (2 stems per shrub with representative leaf forms) for analysis of N content.

The following measurements were collected in the lab for green leaf samples:

  • Single sided leaf mass/unit area measurements

  • Leaf level N concentrations of oven dried ground foliage

  • Absolute reflectance spectra for dried and ground leaf samples using a benchtop spectroradiometer with contact probe attachment.

 

Spectroradiometer data collected in the lab for dry leaves and in the field for shrub canopy were processed and transformed to enhance absorption features of interest.  The transformed spectral data were then related to corresponding leaf and shrub N concentrations to identify wavelength intervals with strong predictive potential.

 

Hyperspectral imagery was acquired concurrent with data collection using a PIKA sensor (5 nm bandwidth, 23 cm2 pixel resolution) mounted on an unmanned aerial vehicle (UAV).  The sensor and UAV are owned and operated by the INL.  Preliminary results were presented at The Art, Science and Applications of Reflectance Spectroscopy Symposium sponsored by Geoscience and Remote Sensing Society (GRSS), a member organization of the Institute of Electrical and Electronic Engineers (IEEE), on February 24, 2010, Boulder, Colorado.

 

Results

We plotted correlation coefficients against wavelengths for both sagebrush dry leaf and shrub canopy ASD data (Figure 1 and 2).  The dry leaf data indicates potentially strong predictors of sagebrush N concentration near 520 nm, 1725 nm, 2209 nm and 2307 nm.  The shrub canopy data indicates wavelengths near 530 nm and 2250 nm are potentially strong predictors of sagebrush N concentration.  However, noise in the short wave infrared makes it difficult to isolate additional wavelengths of interest.

 

Figure 1. The ASD-based dried sagebrush leaf correlogram indicates that wavelengths near 520 nm, 1725 nm, 2209 nm & 2307 nm are potentially strong predictors of sagebrush N concentration.

 

Figure 2. The ASD-based sagebrush canopy correlogram indicates that wavelengths near 530 nm & 2250 nm are potentially strong predictors of sagebrush N concentration. Noise in the short wave infrafred makes it difficult to isolate additional wavelengths of interest.

 

Plans for Continuation

Additional processing techniques will be applied to the shrub canopy spectral data in an effort to reduce the noise in the SWIR.  One such technique that will be applied is spectral resampling to wider bandwidths, which smoothes noise out over large intervals and can be configured to simulate remote sensor specifications.  The spectral data will also be analyzed using chemometric software and multivariate statistics as an approach to isolating wavelengths of predictive interest.  The PIKA hyperspectral imagery will be closely examined to determine the feasibility of extending this project to the landscape scale.  Potential for fusion with LiDAR data may also be evaluated.  The results of this study will be submitted for publication in the journal Remote Sensing of Environment in early summer 2010.

 

Publications, Theses and Reports

Lab-Directed Research and Development (LDRD) Fiscal Year 2009 Final Report

The Influence of Precipitation, Vegetation