Climate Change and Thresholds of Ecosystem Change: Invasibility of Tundra in the Northern Rocky Mountains

Funded by USGS Biological Resources Division, 1999-2004


The aim of this research is to assess the sensitivity of alpine tundra to invasion by woody species from treeline in the northern Rocky Mountain region.  This will be accomplished by developing models of tree species establishment and growth that will reflect causal mechanisms, especially related to hydrology.  The models will be developed and validated at multiple spatial and temporal scales.  The results will allow the interpretation of past and ongoing changes at and above treeline in the wildlands of the western US.  The sensitivity of tundra to invasion is significant because considerable areas of natural resource with value for wildlife, recreation, and aesthetics exist just above the treeline ecotone.  The sensitivity of the ecotone and its use as an indicator of climatic change has been debated.  What the debates overlook, however, is that we should expect a highly nonlinear response because the ecotone is likely a balance of opposing positive feedbacks.  Such positive feedback switches are likely to have produced a system that can have a critical point and be subject to small or large periods of change with incremental climatic change.  Understanding these feedbacks and identifying thresholds of change is critical for managers who often assume linear responses to stress.

We will develop nested models of treeline response.  The core model will be a mechanistic tree establishment and growth model based on modifications to, and integration between, two existing types of models, biogeochemical cycling models (FOREST-BGC) and forest gap models (FORSKA).  We have a modified version of FOREST-BGC, ATE-BGC (Alpine Treeline Ecotone-BGC; Cairns 1995, 1997), that was modified specifically to include treeline ecotone processes such as needle abrasion and frost desiccation.  FORSKA will be modified to represent mat krummholz growth form as well as upright trees.  The two will be integrated by using ATE-BGC to identify the potential productivity of sites that will be used as site quality in FORSKA instead of empirically fitted curves as is usual in gap models.

The core model will be embedded in a meso-scale landscape model.  Sensitivity of tundra to invasion varies within the ecotone as a function of both meso and micro-scale conditions.  Meso-scale conditions are related to topographic and lithologic position and are relatively stable over time.  Since these locations change little over long time scales, during the historic period treeline will have been stable in some areas because any potential advance would be prevented by lithologic and related disturbance conditions (cliffs, rockfall, debris flow).  These conditions will be modeled using a topographic/cartographic modeling approach and will simply extend work that we have already done (Brown, 1994).  Where topography and lithology are not limiting, fine scale conditions influence fine-scale pattern within the ecotone and future changes in the position/pattern of the ecotone will mirror these conditions.  We will run the core model for selected slopes at a 10 m resolution; we will group 10 m cells into 30 m windows for validation using Landsat TM data.  New mixture-modeling interpretation of TM data will allow both more precise validation as well as more accurate extension of the simulation results.  From the core model we will also derive a parsimonious set of rules to create a very simple model, similar to a cellular automaton but with some stochasticity, for simulating the process at 1 m cell resolution.  This model will be validated using digital orthophotoquads and ground samples for a test slope.

Collaborative Research on Process and Pattern at Alpine Treeline

Funded by NSF, 1997-2000


We will assess the possible relations between pattern and process by using a spatially explicit simulation model that produces spatial patterns as the results of various simulated processes such as feedbacks and seed rain. We will compare the two dimensional patterns that we simulate with observed two dimensional patterns of alpine treeline. We will use patterns recorded by Landsat TM as our observations and we will compare several spatial metrics which have been used to assess the nature and degree of spatial organization of landscapes based upon different pattern descriptors. A spatially explicit model allows us to test for the effects of different positive feedback strength, seed rain, and environmental gradient on the abundance of a tree species at an ecotone. The goal of the remote sensing phase is to provide observations against which to compare the simulated patterns. An attempt will be made to control and simplify the environmental gradients on which the processes are modeled and on which resulting patterns are observed. For this reason, hillslopes will be selected which have relatively simple spatial patterns of environmental gradients, defined on the basis of multiple topographic variables describing sun and wind exposure potential, steepness, and topographic soil moisture potential . We will compare two calculations of site quality for these slopes so that a gradient can be input to the model: first, relative site quality will be calculated based on a mechanistic simulation of potential net primary productivity using a model designed for this environment; second, the spatial trend of a summary gradient will be modeled using a trend surface. The patterns observed on the landscape and simulated in the model runs will be quantified using common metrics of landscape pattern. Further, a hybrid or composite metric will be generated which best describes the dominant variation among the sampled units through PCA and multidimensional scaling of the pattern metric calculations. Thus, the overall uniqueness of each metric for describing spatial complexity will be interpretable statistically. The comparisons between observed and simulated patterns will be repeated for each of the selected pattern metrics. The specific influences of each of the hypothesized spatial processes will be examined through this multi-dimensional approach to spatial pattern analysis. The observed patterns for all hillslopes will be pooled to compare the statistical distributions of observed spatial patterns with each of the distributions of modeled patterns across all pattern metrics. We will then fit a model to predict the selected observed metrics from the simulation parameterizations. For example, we would examine ecotone fractal dimension as a function of productivity gradient steepness, seed rain, and feedback strength.

We expect to produce an improved understanding of the basic ecology of ecotones. While the processes at ecotones have been hypothesized, and patterns have been analyzed, we will provide a conceptual link between process and pattern.