- The forest alpine tundra ecotone (FTE, also known as alpine treeline or subalpine parkland), is a conspicuous feature of mountain landscapes throughout the world. Climate change-driven increases in temperature are believed to result in FTE movement and tree invasion of subalpine meadows, which have been documented throughout the Northern Hemisphere across a wide range of geographic locations, climatic regimes, forest types, land
use histories, and disturbance regimes. Climate-driven FTE movement may have numerous ecological effects such as: positive temperature feedbacks, increased net primary productivity
and carbon storage, and declines of plant populations and species. The magnitude of these ecological effects is highly uncertain, but will be largely determined by the rates and spatial
extent of FTE movement and meadow invasion. FTE movement and meadow invasion are often considered at global or regional spatial scales in relation to climate, yet they are
fundamentally driven by tree regeneration processes that are influenced by a variety of climatic and biophysical factors at micro site, landscape, and regional scales. Much of the
research on the FTE has not taken a landscape approach incorporating multi-scale processes. For example, species distribution models used to project climate change effects on future species distributions and plant biodiversity in mountainous landscapes rely on species distribution data that is often sparse and incomplete across FTE landscapes.
This dissertation attempts to overcome many of the limitations in FTE research by taking a landscape approach to develop a greater understanding of past spatiotemporal patterns of tree invasion, current spatial patterns of vegetation composition and structure, and potential future patterns of climate-driven tree invasion in the FTE. The setting for this research is Jefferson Park, a 260 ha subalpine parkland landscape in the Oregon High Cascades, USA.
This study uses field plots, remotely sensed imagery, airborne Light Detection and Ranging (LiDAR), and simulation modeling to: 1) predictively map current fine-scale species distributions, vegetation structure, and tree ages; 2) reconstruct patterns of tree invasion over the last fifty years in subalpine meadows in relation to climatic conditions, landforms, microtopography, and seed dispersal limitations; and 3) develop a statistical model that projects future patterns of tree invasion into subalpine meadows under different climate scenarios in Jefferson Park.
In chapter two, I generated fine-scale spatially-explicit predictions of current vegetation composition, structure, and tree ages in the Jefferson Park study area. Objectives
of this chapter were threefold: 1) to characterize spatial patterns of tree ages, vegetation composition, and vegetation structure in a FTE landscape in the Oregon Cascades using
predictive mapping; 2) determine how vegetation composition and structure were associated with gradients of environmental factors derived from multispectral satellite imagery and Light Detection and Ranging (LiDAR) data; and 3) determine if predictive mapping
characterizations of tree age, vegetation composition, and vegetation structure were improved by the inclusion of LiDAR data. Predictive mapping of vegetation attributes was accomplished using gradient analysis with nearest neighbor imputation; integrating field plots, multispectral SPOT 5 satellite imagery, and LiDAR data. Vegetation composition was best
described by SPOT 5 imagery and LiDAR-derived topography, while vegetation structure was best described by LiDAR-derived vegetation heights. Predictions of species occurrence were
most accurate for tree species, moderate for shrub species and vegetation groups, and highly variable for graminoid species. Tree age, which was the most accurately predicted vegetation
structure variable, indicates the study area was largely un-forested in 1600, gradually invaded by trees from 1600 to the 1920's, and rapidly invaded from the 1920's to 1980. Predictive
mapping of vegetation structure variables such as basal area and stand density were subject to large amounts of error, possibly resulting from scale incompatibilities between vegetation
patterns and plot size, and/or heterogeneous FTE landscapes where forest structure does not develop along consistent trajectories with stand age. This study suggests integrating multispectral satellite imagery, LiDAR data, and field plots can accurately predict fine-scale spatial characterizations of species distributions and tree invasion in the FTE. This study also
indicates that sample design can influence spatial patterns of model uncertainty, which needs to be considered if predictive mapping of vegetation and sensitive ecosystems is a component
of inventory and monitoring programs.
In chapter three, I focused on quantifying spatiotemporal patterns of subalpine parkland tree invasion in Jefferson Park over the past five decades in relation multi-scale climatic and biophysical controls. LiDAR data provided previously unavailable fine-scale spatial characterizations of microtopography and vegetation structure. I utilized LiDAR, georeferenced
field plots, and tree establishment reconstructions to quantify spatiotemporal patterns of tree invasion in relation to late season snow persistence, landform types, fine-scale
topographic variability, distances from potential seed sources, and climate variation within 130 ha of the subalpine parkland landscape of Jefferson Park. Tree occurrence (i.e. tree
presence in 2 m plots and grid cells) occurred in 7.75% of study area meadows in 1950 and increased to 34.7% in 2007. Landform types and finer-scale patterns of topography and vegetation structure influenced summer snow depth, which influenced temporal and spatial patterns of tree establishment. Tree invasion rates were higher on debris flow landforms, which had lower summer snow depth than glacial landforms, suggesting potentially rapid
treeline responses to disturbance events. Tree invasion rates were strongly associated with reduced annual snow fall on glacial landforms, but not on debris flows. Tree establishment
was spatially constrained to micro sites with high topographic positions and close proximity to overstory canopy, site conditions associated with low summer snow depth. Seed source
limitations placed an additional species-specific spatial constraint on where trees invaded meadows. Climate and topography had an interactive effect, with trees establishing on higher
topographic positions during both high snow/low temperature and low snow/high temperature periods, but had greater than expected establishment on lower topographic positions during
low snow/high temperature periods. Within the context of larger landform types, topography and proximity to overstory trees constrained where trees established in the meadows, even
during climate periods with higher temperatures and lower snowfall. Results of this study suggest large scale climate-driven models of vegetation change may overestimate treeline
movement and meadow invasion, because they do not account for biophysical controls limiting tree establishment at multiple spatial scales.
In chapter four, I used field data and analyses from chapter 3 to parameterize a spatially and temporally explicit statistical model of fine-scale tree invasion within 130 ha of the Jefferson Park study area. The model incorporated both the climatic and biophysical controls found in chapter 3 to influence tree invasion. The model was used in two ways: (1) to spatially project patterns of tree invasion from 1950 to 2007 in response to historical climate; and (2) to project future tree invasion of the study area from 2007 to 2064 under six different annual snowfall scenarios. Modeling addressed the following questions: (1) Can fine-scale (2 m pixel size) patterns of historical tree invasion be accurately predicted? (2) How sensitive is future tree invasion (and therefore meadow persistence) to different future snowfall scenarios? (3) Are non-climatic factors such as landforms and biotic interactions associated with different
spatial patterns of tree invasion? From 1950 to 2007, simulated historical meadow area declined from 82% to 65% of the study area. Model outputs of historical area, spatial distributions, and spatial clustering of tree invasion generally agreed with independent validation, and suggest biotic interactions due to young tree establishment facilitation are important on glacial landforms but not debris flows. Simulations of future scenarios indicated meadow declined to 36 to 43% of the study area by 2064. Projected meadow area declined with reduced annual snow fall, but not under prolonged high and low snow fall periods. Meadows persisted under all future scenarios in 2064. This model suggests subalpine meadows may significantly decline under climate warming, but will still persist in 2064. Micro sites and recruitment limitation may be equally or more important factors than climate change in influencing subalpine landscape change, suggesting local high-elevation persistence of subalpine meadows under future climate warming.