In this study, we assessed whether recent wildfires and spruce beetle outbreaks promoted upward range expansion of trembling aspen (Populus tremuloides) in the San Juan Mountains of southern Colorado. We assessed aerial imagery for presence/absence of aspen in the study area and ground truthed a subset of these points to compare the aerial imagery analysis with field records of aspen. From these data we determined the upper elevational limit of aspen in the study area before fire. We then conducted field surveys at and above the upper elevational limit of aspen in burned and unburned, but beetle killed areas and collected data on stand structure, surface cover, seedling density and microsite conditions associated with seedlings.   

Aerial Imagery Analysis

To determine the upper elevational limit of aspen in the study area prior to the 2013 wildfires, high-resolution aerial imagery (ArcGIS World Imagery Basemap January 2014 release; 0.3-m resolution) was assessed to identify aspen presence/absence from 7,494 circular plots placed along 520 randomly generated upslope transects that spanned from 3200 m elevation to the highest mountaintops in the study region. Pre-fire 2011 aerial imagery from the ArcGIS World Imagery Basemap (January 2014 release; 0.3-m resolution) was the primary source for assessment, but was supplemented with 2011 NAIP imagery (1-m resolution) if the primary source was obscured by presence of shadows, clouds, etc. To establish transects across the study area, we first generated a set of random points (n = 1,228) at 3200, 3300 and 3400 meters in elevation. Then, from each random point we established a transect that followed the terrain upslope, placing a circular plot (25-m radius) at horizontal increments of 60 m. We then selected a subset of these transects (n = 550) for sampling, as follows. To ensure that we sampled areas containing aspen, we evaluated all transects that intersected areas classified as aspen in the 2012 LANDFIRE existing vegetation type data (n = 181), which uses a combination of field data, satellite imagery, biophysical layers and regression trees to produce a predictive model of vegetation cover on the landscape. We then added an additional random sample of transects (n = 369), for a total of 550 transects assessed. Circular plots along each transect were visually examined using the aerial imagery to determine if aspen occurred anywhere within the 25-m radius circle. If so, this point was counted as an “aspen present” point. In the case where circular plots overlapped due to intersecting lines, only the first plot viewed was used for analysis. All imagery was collected during the summer months, when aspen leaves were present and could be easily visually distinguished from much darker conifer needles. Aspen is the only tall broadleaf species in this area and therefore we had high confidence that it was correctly visually classified. It was distinguished from broadleaf shrub species by the presence of shadows (indicating tree status) in the aerial imagery as well as landscape context (e.g. alpine areas with high willow cover were not classified as aspen).
We tested the accuracy of our ability to visually determine aspen presence by field sampling a subset of these plots (n = 400) for aspen presence (at least one aspen tree within the plot). In the summers of 2018 and 2019, a subset of plots (n = 400) were navigated to in the field using GPS coordinates, to test the accuracy of our aspen presence data derived from aerial imagery. Only plots outside of the fire perimeters that had not been disturbed (by fire or harvesting) since 2011 were used for field validation. Plots that were near roads and most easily accessible were selected. At each plot, we established a 25-m radius circular plot and recorded the presence or absence of canopy aspen within the plot.

Seedling Surveys

To assess whether aspen upslope expansion has occurred following fire across our study area, we field-sampled a total of 47 burned and 41 unburned sites across eight burned and seven unburned elevational transects. Although the unburned sites did not experience mortality from the West Fork Fire Complex, all were impacted by the spruce beetle outbreak of the 2000’s and therefore still experienced severe canopy mortality before our survey. From the aerial imagery analysis, we identified locations where aspen existed at or near its conservative upper limit (3506 m) prior to the fire. From the locations of each of these aspen stands, an elevational transect was established running upslope to the top of the mountain or a total increase of 240 m in elevation, whichever occurred first, maintaining the same aspect. The first site on each elevational transect was established immediately upslope and adjacent to the pre-fire aspen stand, though one of these first sites was determined to be hazardous and therefore excluded from analysis. Subsequent sites were established every 40 m of elevation gain on the transects, which translated to 82 – 1,226 m of horizontal distance on the ground. Minor adjustments to this protocol were necessary on a few sites to avoid hazardous conditions or to keep the transect in its intended burn category. Slope, aspect, elevation, latitude, and longitude were recorded for each site in the field. Linear distance to the nearest live adult aspen stand was calculated in ArcMap using post-fire (2016 & 2018) aerial imagery to find the nearest stands.

At each site, belt transects measuring 50-m long by 8-m wide were established horizontally across the slope and searched for aspen juveniles (<1.4-m tall). The exception to this was at the pre-fire aspen stand sites (the lowest site on each elevational transect), where variable-width belt transects measuring 50 m in length were established. These belt transects were a minimum of 0.5-m wide and were widened by 0.5 m until at least 14 aspen juveniles were found, or a maximum width of 8 m was surveyed. This was done to account for the large number of aspen suckers at these pre-fire aspen-dominated sites.

For each aspen juvenile found at a site, we recorded whether it was a seedling or sucker by gently excavating the roots. Aspen suckers sprout from lateral roots that can usually be found within 10 cm of the soil surface and can extend up to 30 meters away from the parent stem (Day, 1944; USDA, NRCS, 2020). Therefore, the presence of a lateral root within 10 cm of the soil surface on a juvenile surrounded by other aspen juveniles indicated sucker status. Seedlings were characterized by multiple fine roots with no thick lateral root and generally occurred in isolation away from clumps of other aspen juveniles. In addition, one to five of the aspen seedlings found at the sites were excavated across each elevational transect and were confirmed to have a branching, fine root system characteristic of aspen established from seed. The height and diameter at soil surface were measured for each juvenile encountered, and any browse damage was noted. Microsite characteristics within 5 cm of the rooting location were recorded for each juvenile, including the presence or absence of a nurse object (defined here as rocks and logs > 5-cm diameter), the substrate type, the presence and depth of litter, and the presence or absence of vegetation. Canopy cover above the juvenile was measured with a spherical densiometer.

Across the 50-m length of the belt transect, the line point-intercept method was used to quantify the abundance of nurse objects, substrates, litter, understory vegetation, and canopy cover at the site. A pin flag was dropped every 0.5 m and all vegetation touching the pin flag (or touching the imaginary line ascending from the pin flag) was recorded according to functional group (grass, forb, shrub, tree) and status (live or dead). If multiple species of the same functional group touched the pin, the functional group was only recorded once for that point. Substrate type (soil, duff, boulder – rocks > 20-cm diameter and not easily lifted, or coarse woody debris (CWD) – wood > 20-cm diameter and not easily lifted), and presence of rocks (> 5-cm diameter and easily lifted), wood (> 5-mm diameter and easily lifted) and litter at each point was also recorded. The percent cover of all vegetation, substrate types and nurse objects at each site was calculated by dividing the number of points where the vegetation/substrate was recorded by the total number of points sampled (n = 100) and multiplying this by 100. A point classified as “bare soil” in the analysis had to have a substrate of soil and no litter, duff, moss, rock, or wood present at the point. Stand structure was additionally measured in three 100-m² circular plots (5.64-m radius) centered on the middle of the belt transect at 7 m, 25 m, and 43 m. The diameter at breast height (DBH), height, and status (live or dead) was recorded for each tree within the circular plot.  

Data Processing

All data processing and analysis was done in RStudio with the included R script. 

In the line-point-intercept data, some of the belt transects are missing surface cover codes (for example, each belt transect should have 100 points total, but some only have 98 or 99). This can be easily remedied when calculating percent cover by just dividing by the number of points that were sampled and multiplying by 100. 

For three belt transects (P231, P233, P235), the topcanopy code in the line-point-intercept (LPI) data was recorded incorrectly. Therefore, we used a linear regression between the % tree cover as calculated from LPI and the average canopy density as calculated from spherical densiometers in the three circular tree plots across the belt transect. We then used the model coefficients to calculate new % tree cover values for the three belt transects that were mis-recorded.  

In all data files, a value of "NA" indicates that data was not collected because it was not applicable, and a value of "null" indicates that the data should have been collected, but was not (i.e. missing data).   

A readme file has been created to assist in using this data (wildfire_catalyzes_aspen_range_expansion_readme.txt)