Selected Research Projects

Glacier National Park, Montana: 
Spectral-Field Analyses of Vegetation 
at the Alpine Treeline Ecotone

Research within Glacier National Park, Montana has been conducted over the years by a team of researchers lead by Stephen J. Walsh, University of North Carolina, George P. Malanson, University of Iowa, and David R. Butler, Southwest Texas State University. Research collaboration, initially begun doctoral their studies, now includes: Daniel G. Brown, Michigan State University, David M. Cairns, Texas A & M University, Thomas R. Allen, Old Dominion University, Stephen J. McGregor, University of North Carolina, and Ling Bian, SUNY-Buffalo.

View the Satellite Image Library
GNP Database

Our research into the spatial and biophysical character of mountain environments has focused on Glacier National Park, Montana, USA. The Park is characterized by pronounced topographic variability resulting in strong biophysical gradients that shape the corresponding vegetative landscape. The glacial history of the Park, its position astride the Continental Divide, its past geologic setting and present geomorphic processes, its spatial and biophysical complexity, and its relative inaccessibility have necessitated the development of an analytical framework that integrates satellite remotely-sensed data with biophysical coverages contained within a geographic information system for spatial analysis and modeling.

Our research within the Park has generally been organized around a set of primary initiatives:

• Development of a GIS to support spatial and biophysical studies of processes and feature distributions;
• Implementation of resource evaluation studies through the manipulation of the derived GIS databases, for the evaluation of natural hazards and snow-avalanche paths; assessment of lake turbidity levels and relationships to basin morphometry; identification of deltaic wetlands; and evaluation of forest fire potential;
• Examination of scale dependencies of plant biomass and topography and spatial analyses of derived patterns of the distributions of krummholz patches and associated alpine and subalpine features;
• Cartographic and quantitative modeling of disturbance factors and topoclimatic variables affecting the biophysical landscape at the alpine treeline ecotone.

GNP is a UNESCO-designated Biosphere Reserve of 405,000 ha, with peaks to an altitude of 3,100m, in northwestern Montana, USA. It was chosen as the study site for our research because of its ecotonal position at the eastern edge of the climatic region dominated by Pacific maritime air in the northwestern US, as well as for its ecotonal transition from mountains to the Great Plains. The Park also has been selected by the National Park Service to lead the park-oriented scientific investigations of responses to global climate change. 

Two mountain ranges, both with glaciers at the present time, occur within the Park: the Livingston Range and the Lewis Range. The Continental Divide follows the Livingston Range in the north and then shifts to the more easterly Lewis Range. The Park area was heavily glaciated during the Pleistocene, but a number of mountain crest and upland surfaces in the Lewis Range escaped glaciation. These gently sloping uplands are largely at or above treeline, with a variety of exposures, and are or have been subjected to periglacial mass-wasting and patterned ground formation in the recent past. Areas above treeline in the Livingston Range tend to be steeper and more geomorphically active, with little colluvial cover. The periglacial environment in the Park is currently encountered above approximately 1,950 - 2,000m, with current processes producing patterned ground above ca. 2,170m, and relict features occurring in the roughly 200, of transition between 1,950 and 2,170m. As a result of the its location, there are two distinct climatic regions within the Park., West of the Continental Divide, the climate is generally milder and more moist, akin to a Pacific maritime climatic pattern. East of the Divide, the climate is much harsher, with more severe low temperatures, stronger winds, and drier conditions. Snow avalanches include both wet-snow and dry-powder types, but historical data suggest that the former is much more common, especially on western slopes.

Tree species comprising the upper treeline ecotone reflect climatic differences: west of the Divide, upper treeline is comprised primarily of subalpine fir (Abies lasiocarpa), Engelmann spruce (Picea engelmannii), and subalpine larch (Larix lyallii); whereas east of the Divide, treeline is dominated by subalpine fir, whitebark pine (Pinus albicaulis), and limber pine (Pinus flexilis). Approximately two-thirds of the Park is forested. The subalpine forest grades into the alpine environment, characterized by the diminished vegetation density representative of tundra vegetation and bare rock, snow, and ice surfaces. On the eastern side of the Park, two treelines are evident: an upper temperature-related treeline, and a drought-related treeline at low elevations. The spatial extent of this vegetation grading, clearly evident at alpine treeline, is related to elevation, topographic orientation, topographic controls, and disturbances. The eastern two-thirds of the Park are dominated by unvegetated and more sparsely vegetated surfaces. These are interrupted by valleys and hanging-valleys of subalpine forest, brush, and herbaceous cover resulting primarily from the position of the Continental Divide and local orientations of the topography and landforms carved by glaciers. Snow-avalanche paths are striking features that extend from the alpine to the subalpine environment throughout the Park. Where vegetation disturbances exist, such as snow-avalanche paths and debris flows, the boundary between vegetated and unvegetated surfaces is distinct, and the transition zone from subalpine forest to alpine conditions is spatially irregular, depending on the pattern and type of disturbance. 

Conclusions

The development of a GIS was an essential element of the spatial and biophysical research conducted within the Park. The GIS facilitated the integration of remotely-sensed information with core GIS coverages; HIS coverages and attribute fields derived through manipulation of the core coverages; calculated statistical coverages and spatial measures of feature distances and distributions; and spatially normalized and pattern-defined indices of spatial structure. The spatial operations provided through the GIS permitted not only the verification of remote sensing inputs to the GIS by comparison to existing GIS coverages and field measurements recorded in the database, but also the validation of GIS coverages by comparison to remote sensing inputs and to other GIS coverages. The spatial perspective afforded by the GIS also facilitated the generation of new hypotheses and the explanation of established research questions. Spatial operations were essential to the manipulation of the GIS database to support spatial and biophysical queries and for the display of research results as three-dimensional graphics and feature drapes along with statistical findings presented as models, spatial deviations from models, and measures of the spatial pattern, spatial ordering, and spatial structure of diverse phenomena.

The research into spatial dependence and spatial analysis suggests the interrelationships of scale, pattern, and process, as well as the importance of fractal and semivariance analyses to the examination of the characteristic scale that controls much of the primary biophysical variation within the study area. 

Our research on alpine treeline provides information on the relationships between multiple and interacting environmental factors and patterns of vegetation. The effects of spatial interactions will be included in subsequent models to test the importance of spatial autocorrelation for describing the pattern of alpine treeline components. Because the spatial pattern of vegetation at an ecotone is an importance component of its response to environmental change or disturbance, the determination of the range of patterns and their related processes is important in order to be able to identify which patterns represent potential fronts for the invasion of vegetation as a consequence of global climate change. Models that explore the relationship between the location and character of alpine treeline, together with sets of derived topoclimatic variables that describe the differences between predicted and observed patterns of alpine treeline, were used to develop local disturbance variables to complement the more regional indicators of the treeline ecotone.
 
 

Leaf Area Index (LAI), the area of leaves per unit ground area, and leaf area distribution are fundamental measures of canopy structure. Together, they indicate the amount of foliage present and its distribution, both critical vegetation parameters that strongly influence the rates of evapotranspiration and photosynthesis. Spatial variation in the canopy structure over the landscape provides a quantitative measure of ecological heterogeneity. High resolution satellites are capable of monitoring canopy structure over time and space following the establishment of correlations between LAI measurements of in-situ conditions and satellite spectral responses (Peterson and Running 1989). Regional variations in LAI were found to be linearly related to site water balance, above-ground net primary production, and stand volume growth of western conifer forests (McLeod and Running 1988, Nemani and Running 1989). The ability to estimate and map LAI on landscape scales from satellite remote sensing has given ecologists and bgiogeographers an opportunity to extend their theories of water and carbon exchange from leaf level to landscape scales (Running et al. 1987). 

Spectral indices are empirical regression-based relationships between remotely-sensed radiance values and LAI. These spectral indices have been shown to have validity over a wide range of scenes and sensors. Canopy closure is a key spatial variable governing the scene brightness in conifer canopies, because it controls the fraction of understory and overstory visible to the sensor (Spanner et al. 1990). Nemani et al (1987) has included the middle-infrared spectral wavelengths, together with the red-visible and near-infrared wavlengths, in an adjusted version of the Normalized Difference Vegetation Index for that estimating canopy closure from satellite data. These spectral regions are sensitive to the moisture content of the leaf, plant pigmentation, and the chlorophyll content of the leaf respectively. This research will compare the utility of estimating LAI with the adjusted version of NDVI for the alpine treeline ecotone in Glacier National Park, Montana, the site of previous research and in-situ data collection (Walsh 19xx), by directly comparing LAI with reflectance measured at forest stands using regression techniques. 
 
 

Indirect measures of canopy structure are based on the close coupling between radiation penetration and canopy structure. The measurement of the fraction of sky visible through the canopy (Gap Fraction) at various angles is an approach that is used in remote sensing studies (Ross and Myneni 1990). The LAI Plant Canopy Analyzer, to be used in this research, employs mathematical techniques for deducing foliage density and angle distribution from measurements of gap fraction (Wells and Norman 1991). The device consists of an optical sensor and control box. The sensor incorporates fisheye optics to project a hemispheric image onto five silicon detectors, arranged in concentric rings. The sensors also contain an optical filter to restrict sensed radiation to wavelengths below 490 nm, in order to minimize the contribution of radiation that had been scattered by foliage. The control box records the sensor's data and performs the necessary calculations for determining LAI and the mean inclination angle of the foliage. 

The technique combines a measurement of sky brightness from a leveled sensor above the canopy with a second measurement taken beneath the canopy with the sensor again viewing skyward. The ratio of each ring's signal (below to above) is then assumed to be equivalent to the canopy's gap fraction at the ring's viewing angle. In practice, multiple below-canopy measurements are taken to achieve a suitable spatial average for the measurement site. Multiple below-canopy measurements are combined by averaging the logarithms of the computed gap fraction after Lang and Xiang (1986). Above canopy measurements in tall forest canopies can be made through (a) measurements in large nearby clearings linked to below canopy measurements in the target canopy or (b) a unit and data logger attached to a tower or to an adjacent tree for above canopy measurements. Gap fraction is converted to LAI and mean inclination angle in the control box using a method similar to Lang (1987).
 
 

LAI measurements will be collected in three specific situations: (a) using the plant community classification, developed by Townsend and Walsh (1997) and derived from a Landsat TM time-series for the study area, LAI measures will be calculated at approximately 30 sites within each community type. A distribution of derived LAI measures will be developed and statistically summarized; (b) using GPS devices, field plots established in the study area by Townsend and Peet will be relocated and LAI measures will be derived and additional floristic information collected (e.g. tree ring data, tree height, canopy size) where appropriate; and (c) additional sites will be defined for LAI and floristic information for targeted species and at random locations. In addition to the LAI at point locations, a continuous surface of LAI measures will be derived through use of satellite spectral responses, from sites not used to initialize the vegetation simulations. Canopy LAI will be determined by calibrating the Landsat TM data with the field-measured LAI, following the methods of Nemani et al (1993).
 
 

LAI measurements will be collected in three specific situations: (a) using the plant community classification, developed by Townsend and Walsh (1997a, 1998b) and derived from a Landsat TM time-series for the study area, LAI measures will be calculated at approximately 30 sites within each community type. A distribution of derived LAI measures will be developed and statistically summarized; (b) using GPS devices, field plots established in the study area by Townsend (1997) and Rice and Peet (1997) will be relocated and LAI measures will be derived and additional floristic information collected (e.g. tree ring data, tree height, canopy size) where appropriate; and (c) additional sites will be defined for LAI and floristic information for targeted species and at random locations. In addition to the LAI at point locations, a continuous surface of LAI measures will be derived through use of satellite spectral responses, from sites not used to initialize the vegetation simulations. Canopy LAI will be determined by calibrating the multi-temporal Landsat TM data with the field-measured LAI, following the methods of Nemani et al (1993). 
 
 

The Spatial Analysis Lab, Department of Geography, University of North Carolina, directed by Stephen J. Walsh, has recently been expanded through the development of a Spectral Field Lab. Available equipment (see facilities statement for UNC-CH) includes: Portable Spectroradiometers (measures the spectral properties of leaves by measuring their reflectance and transmittance across a range of selected wavelengths, visible through the NIR); Plant Canopy Analyzer (calculates LAI, foliage density, and other canopy structure attributes from radiation measurements, as well as the fraction of the sky visible from beneath the canopy); Portable Laser Leaf Area Meter (measures reductions in leaf area associated with drought, insect damage, or phenological variations in plant stages); and Data Loggers (stores downloaded climate data at stations and other in-situ data collected in the field). Such equipment will be used in this project for the calculation of LAI and light extinction coefficients for community types and sample sites.