Mapping a keystone shrub species, huckleberry ( Vaccinium membranaceum ), using seasonal colour change in the Rocky Mountains

• Main Authors: Shores, Carolyn R., Mikle, Nate, Graves, Tab A.
• Format: Dataset
• Language: unknown
• Published: Taylor & Francis 2019
• Subjects: 59999 Environmental Sciences not elsewhere classified; FOS Earth and related environmental sciences; Ecology; FOS Biological sciences; 69999 Biological Sciences not elsewhere classified; 80699 Information Systems not elsewhere classified; FOS Computer and information sciences; Plant Biology; Ursus arctos
• Online Access: https://dx.doi.org/10.6084/m9.figshare.8058923.v1; https://tandf.figshare.com/articles/Mapping_a_keystone_shrub_species_huckleberry_i_Vaccinium_membranaceum_i_using_seasonal_colour_change_in_the_Rocky_Mountains/8058923/1

Black huckleberries ( Vaccinium membranaceum ) provide a critical food resource to many wildlife species, including apex omnivores such as the grizzly bear ( Ursus arctos ), and play an important socioeconomic role for many communities in western North America, especially indigenous peoples. Remote sensing imagery offers the potential for accurate landscape-level mapping of huckleberries because the shrub changes colour seasonally. We developed two methods, for local and regional scales, to map a shrub species using leaf seasonal colour change from remote sensing imagery. We assessed accuracy with ground-based vegetation surveys. The high-resolution supervised random forest classification from one-meter resolution National Agricultural Imagery Program (NAIP) imagery achieved an overall accuracy of 75.31% (kappa = 0.26). The approach using multi-temporal 30-meter Landsat imagery similarly had an overall accuracy of 79.19% (kappa = .31). We found underprediction error was related to higher forest cover and a lack of visible colour change on the ground in some plots. Where forest cover was low, both models performed better. In areas with <10% forest cover, the high-resolution classification achieved an accuracy of 80.73% (kappa = 0.48), while the Landsat model had an accuracy of 82.55% (kappa = 0.47). Based on the fine-scale predictions, we found that 94% of huckleberry shrubs identified in our study area of Glacier National Park, Montana, USA are over 100 meters from human recreation trails. This information could be combined with productivity and phenology information to estimate the timing and availability of food resources for wildlife and to provide managers with a tool to identify and manage huckleberries. The development of the multi-temporal Landsat models sets the stage for assessment of impacts of disturbance at regional scales on this ecologically, culturally, and economically important shrub species. Our approach to map huckleberries is straightforward, efficient and accessible to wildlife and environmental managers and researchers in diverse fields.