Most species accounts include a probability of occurrence map that depicts the probability of encountering that species in a particular area of Denali. The data used to make the maps is derived from plant occurrences recorded during three extensive botanical studies: the landscape scale vegetation monitoring program, the floristic inventory, and the Denali soils map and ecological site inventory. Plant occurrences and ecological site data from these projects were combined and analyzed to create a probability-based model for the likelihood of encountering a species across the landscape of the study area.
Maps are based on a set of statistical analyses called generalized linear mixed models with multiple random effects, the results of which are then used in a fixed-rank kriging analysis to interpolate probability values across the entire study area. Our collaborator, statistician Dr. Jay Ver Hoef, performed these analyses. The input data into these statistical models were the physical plot parameters that affect plant distribution (slope, aspect, elevation, soil type, etc.) and the locations where each species occurs in these plots. Note that models are limited to a large rectangular area in the northeastern half of Denali — this area is the extent of the vegetation monitoring study region, beyond which data points are less dense.
While the spatial statistical analyses used are quite complex, the interpretation of the created maps is quite straightforward. Essentially – using a cold-to hot (blue to red) color scheme, maps show the probability that you will encounter a given species at a given location. Increasingly high probability of encountering the species is represented in yellow-to-orange-to-red hues, whereas diminishing probability of encountering a species is represented in green-to-light blue-to-deep blue hues. High values on the occurrence probability scale (yellow-orange-red) mean a species is more likely to occur.
Thus if you take a hike in areas on the map where the probability of occurrence is greater than 90% (orange-to-red) you are essentially certain to see that species sometime during your hike, whereas if you are in a dark blue area for a species you are probably not going to see the species no matter how far you hike that day!
The discrete area which our model predicts probability of occurrence is within a 200-m2 circle (16-m diameter), so imagine throwing a huge hula-hoop out to a random spot in the tundra – if our map predicts 70% probability of occurrence for a given area, this means that, we predict you will find the species within your giant hula hoop 7 times out of 10 throws (based on our analyses).
The probability of occurrence map shown here shows the model for Populus balsamifera, an early successional tree species. Note that the terraces and gravel bars in close proximity to streams and rivers are highlighted as the areas with increased probability of encountering this species. This result was arrived at through the statistical modeling procedures described above, but also accords with our experience in travelling and botanizing in Denali, where this species is by far the most common on gravel bars associated with large glacial rivers such as the McKinley River (conspicuously highlighted in the probability of occurrence model shown). Compare the modeled data to the documented locations map to examine the distribution of Populus balsamifera localities actually documented in plots.
Back to Understanding Data Presented…