Spatio-Temporal Dynamics of Alpine Treeline Ecotones in the Western United States Under Climate Change

Human-mediated climate change over the past century has had significant impacts on global ecosystems and biodiversity including accelerating redistribution of the geographic ranges of species. In mountainous regions, the impact of climate change can be observed on the abrupt transition zone from closed-canopy montane forests to treeless alpine tundra areas at higher elevations, which is commonly referred to as "alpine treeline ecotone" (ATE). Globally, warming climate is expected to drive ATEs upslope, which could lead to negative impacts on local biodiversity and changes in ecosystem functioning. However, existing ATE studies rely primarily on field-based data which are difficult and time consuming to collect. The objectives of this study are: 1) to define an automated ATE detection metric using easily accessible remote sensing datasets, and 2) to apply the developed metric to monitor the spatio-temporal dynamics of ATEs during last decades in the western United States. We first defined three characteristics of ATEs with respect to vegetation coverage and elevation, which included: a) abrupt spatial gradient in normalized difference vegetation index (NDVI), b) intermediary values of NDVI, and c) spatial co-variation of elevation and NDVI. According to these characteristics, three identification components were constructed with LANDSAT and AW3D30 elevation datasets at 30-m spatial resolution. We then developed an ATE index (ATEI) ranging from 0 to 1 with a binomial logistic regression model of the identification components at 141 sampled LANDSAT pixels manually classified with high-resolution remotely sensed imagery in Google Earth. The average model accuracy was around 0.713 ( 0.111) and the average Kappa coefficient was approximately 0.426 ( 0.221) based on a 100-time repeated 10-fold cross-validation. Additionally, we calculated annual ATEIs from 1984 to 2011 in the western U.S. and found the ATEI-estimated elevations were highly correlated (Pearson's r = 0.98) with a published set of field-collected ATE elevations at 22 sampling sites across the region. The 30-m annual ATEIs were then aggregated to the USGS sub-watershed (SWD) level, and the change of ATE elevation over time was estimated for each SWD using iterated re-weighted least squares regression. We then filtered out SWDs with substantial disturbance (e.g., fire or pests) detected by the "LandTrendr" algorithm. After that, the remaining SWDs were grouped into three categories based on the estimated elevation change over time: decreasing (< -2 m/decade), no change (between -2 and 2 m/decade), and increasing (> 2 m/decade). We found that the ATE had a median increase of 7.09 (5th-95th percentile: 2.46-37.03) m/decade in 51.1% of the SWDs, a median decrease of 6.79 (2.42-35.09) m/decade in 25.2%, and was stable in the remaining 23.7%. In summary, over half of the ATEs in the western U.S. have shifted upslope at a median rate of around 7 m/decade since the 1980s, which improves our understanding of the spatio-temporal variation of geographic ranges of plant species in mountainous areas under the changing climate. We also expect the ATE detection metric developed in this study to be useful in monitoring the dynamics of other ecosystem boundaries.