Experimental Remote Sensing of Vegetation on the INEEL


This study encompasses several areas within the Birch Creek watershed of the INEEL. The Birch Creek watershed is an ecologically sensitive area and includes a portion of a national sagebrush steppe reserve which is pristine habitat for a number of native species. Precise inventories of ecologic biodiversity in this region assist in the assessment of natural diversity and ecological indicators. This study uses both passive and active remote sensing technologies to assess conservation targets on the INEEL. Specifically, the objectives of this study are to develop the use of hyperspectral remote sensing to monitor spotted knapweed (Centaurea maculosa, an invasive species), and airborne laser swath mapping (ALSM) to determine low-height vegetation canopy structure.

Although the application of hyperspectral remote sensing to vegetative mapping is relatively new, recent publications have demonstrated a degree of confidence in the ability of this technology to accurately model a landscape. This study uses HyMap hyperspectral data (HyVista, Inc.), which records incident solar radiation naturally reflected from the surface target
(e.g. passive remote sensing) using an airborne sensor. The high spatial and spectral resolution of this imagery differentiates electromagnetic absorption features that are commonly associated with vegetative targets. Spectral component analysis of these datasets allows for a detailed composition of the ecosystem to be assessed, thereby enabling a large area to be mapped in detail in a relatively short amount of time.

Airborne laser swath mapping (ALSM) is a relatively new and quickly growing field of active remote sensing which makes use of a scanning pulsed laser mounted aboard an airborne or satellite platform. Highly accurate timing instruments measure the pulse travel time, and when used in combination with a GPS and an inertia measurement unit (IMU), are able to determine the elevation of the surface from which the laser pulse is reflected. The ALSM data used in this study (Airborne 1) has a vertical accuracy of less than 10 cm (4 in.) and a horizontal accuracy of less than 1 m (3.2 ft).

Of the numerous studies of canopy structures using ALSM, the vast majority have targeted forests. This is due in part because many ALSM sensors have a vertical accuracy on the order of tens of centimeters and, as such, are well suited to canopies many meters high. In the case of some rangeland areas, however, the canopy heights (i.e. for grasslands) are of the same order as or slightly greater than the sensor accuracy, making it difficult to extract significant information about the vegetation. This study attempts to quantify the sensitivity limits and ultimately the usefulness of ALSM vegetation mapping within a rangeland setting.

Five ALSM areas and four hyperspectral lines were acquired in summer 2002 on the INEEL (Figure 9-7). The ALSM datasets range in size from 2 to 15 km2 (0.8 to 5.8 mi 2) and have post densities of approximately 1.2 m-2 (13 ft-2). Each ALSM data point measurement records time, X and Y coordinates, elevation, and intensity values for the both the first and last return of the laser pulse. The hyperspectral data sets are approximately 2 by 20 km (1.2 to 12.4 mi) each, with a 3.5 m (11.5 ft) spatial resolution and 126 spectral bands in the visible and infrared portions of the electromagnetic spectrum (wavelengths ranging from 450 nm to 2.48 Ám).


Hyperspectral Data Analysis - Using the hyperspectral data, the primary objective is to produce a remote sensing-based classified map of spotted knapweed in the Birch Creek watershed. The classified map is validated with field ground truth data. Additionally, this study aims to assess the spectral separability of the sagebrush with grasses (e.g. crested wheatgrass, fescue, and/or bunchgrass) in the study area. This project includes:

  • Assessing and rectifying the geometric precision of the hyperspectral data;
  • Validating and refining classifications for distribution of spotted knapweed in Birch Creek using field data; and
  • Exploring the potential to differentiate vegetative species from background (e.g. soil) in a semi-arid climate

ALSM Data Analysis - Using the ALSM data, the primary objective is to determine the heights of various types of rangeland vegetation to an accuracy of a few centimeters. This allows for the discrimination between rangeland grasses (both native and nonnative) from sagebrush, bitterbrush, and other types of low-lying vegetation. Such information is of value in monitoring the recovery of burned sagebrush stands as well as determining where invasive grasses (i.e. cheatgrass) are replacing native stands of sagebrush.

The project scope includes:

  • Georeferencing the ALSM data to an accuracy of better than 0.5 m (1.6 ft);
  • Validating the calculated ground surface models;
  • Determining ALSM reflection/penetration rates into the rangeland canopy and relating such measurements to canopy cross-sections, surface area, and/or biomass;
  • Validating discrete vegetation height data points;
  • Validating surface roughness characteristics over various length scales; and
  • Correlating vegetation height and surface roughness to recovery characteristics in recently burned areas.

An additional objective of this study, to be investigated during summer 2004, is to fuse the ALSM data with hyperspectral data. In areas where the sensors' data overlap, each dataset will be classified independently to map the distribution of both brush and grass groundcover types. Following co-registration, comparative analysis of these datasets in combination with field validation will be performed. For example, the study will address whether the ALSM data is applicable solely to mapping structure (e.g. vegetation heights) or whether it can also provide species discrimination (based on vegetation canopy structure) comparable to the hyperspectral data. The study will also compare the cost and processing feasibility of each dataset.


The hyperspectral classification data analysis has been completed for the distribution of spotted knapweed on INEEL (Figure 9-8). Field validation for the classification is ongoing (summer 2004). Several theoretical issues have been explored during data analysis, including hyperspectral processing methods, geometric precision, GPS data integration, field spectra modeling and training, and atmospheric influences. More information on these techniques may be found in Mundt (2003).

Computational algorithms have been developed to analyze the ALSM data for the INEEL. These algorithms use an iterative method to separate the ALSM data into bare ground and vegetation categories. This allows the underlying ground surface to be modeled and subsequently subtracted from the vegetation data, resulting in the vegetation heights. The length scales on which these algorithms operate are determined by a surface fractal analysis.

In the interest of preserving the accuracy of the ALSM data, the raw, irregularly spaced data points are used in the analysis in lieu of interpolating the data into a regularly spaced digital elevation model (DEM). Because the datasets are quite large (several millions of data points), much effort has been invested in developing computationally efficient and numerically precise algorithms. These efforts were presented in a poster titled "Detection and characterization of rangeland vegetation using airborne laser swath mapping" at the Fall 2003 meeting of the American Geophysical Union (Streutker 2003).


Hyperspectral Data Analysis - At this time, field data indicates a high accuracy potential, with approximately 80 percent of classified pixels falling within known spotted knapweed occurrences in the remote sensing-derived map. While these numbers are preliminary and additional field data needs to be collected, the initial results are encouraging for the differentiation of relatively sparse vegetation in a semi-arid ecosystem.

ALSM Data Analysis - Initial maps of vegetation height have been produced for the areas under investigation in the INEEL, as well the other areas in eastern Idaho for which data was collected (e.g. U.S. Sheep Experiment Station, Dubois). The calculated vegetation heights range from a few centimeters for grasslands, several tens of centimeters for sagebrush, and several meters for trees. In areas of recent fires, the fire boundaries are clearly delineated within the vegetation height data. The vegetation height data is also used to calculate surface roughness, which can, in turn, be utilized in discriminating between different vegetation communities (i.e. dense sagebrush stands versus grasslands).

Plans for Continuation

Concurrent with the validation of the spotted knapweed study and validation of the vegetation heights in the ALSM-derived maps, field data will be collected to map the biodiversity where the ALSM and hyperspectral data overlay. This will include detailed GPS measurements of vegetation distribution and structure.

Investigators and Affiliations

Nancy F. Glenn, Assistant Research Professor, Department of Geosciences, Idaho State University, Boise, ID

David R. Streutker, Postdoctoral Researcher, Department of Geosciences, Idaho State University, Boise, ID

Jacob T. Mundt, Research Associate, Department of Geosciences, Idaho State University, Boise, ID

Funding Sources
ISU-INEEL - Partnership for Integrated Environmental Analysis Education Outreach Program


Home | Background | Surveillance |  Land Management | Education | Research | Risk Assessment | Publications |  Links | Feedback | Opportunities