Detecting changes in understorey and canopy vegetation cycles in West Central Alberta using a fusion of Landsat and MODIS

Abstract

Aims To model regional vegetation cycles through data fusion methods for creating a 30‐m daily vegetation product from 2000 to 2018 and to analyze annual vegetation trends over this time period. Location The Yellowhead Bear Management Area, a 31,180‐km2 area in west central Alberta, Canada. Methods In this paper, we use Dynamic Time Warping (DTW) as a data fusion technique to combine Landsat 5, 7 and 8 satellite data and Moderate Resolution Image Spectroradiometer (MODIS) Aqua and Terra imagery, to quantify daily vegetation using Enhanced Vegetation Index at a 30‐m resolution, for the years 2000–2018. We validated this approach, entitled DRIVE (Daily Remote Inference of VEgetation), using imagery acquired from a network of ground cameras. Results When DRIVE was compared to start and end of season dates (SOS and EOS respectively) derived from ground cameras, correlations were r = 0.73 at SOS and r = 0.85 at EOS with a mean absolute error of 7.17 days at SOS and 10.76 days at EOS. Results showed that DRIVE accurately increased spatial and temporal resolution of remote‐sensing data. We demonstrated that SOS is advancing at a maximum rate of 0.78 days per year temporally over the 18‐year time period for varying elevation gradients and land cover classes over the region. Conclusions With DRIVE, we demonstrate the utility of DTW in quantifying vegetation cycles over a large heterogeneous region and determining how changing climate is affecting regional vegetation. DRIVE may prove to be an important method to determine how carbon sequestration is varying within fine‐scale individual plant communities in response to changing climate and likely will be beneficial to wildlife movement and habitat selection studies examining the varying response of wildlife species to changing vegetation cycles under shifting climatic conditions.

Simultaneous monitoring of vegetation dynamics and wildlife activity with camera traps to assess habitat change

“…2 and 6), observed differences underscore the consequences of choosing appropriate metrics and scales of inference in heterogeneous landscapes (Dronova et al., 2021). Our findings are consistent with known discrepancies between satellite and ground‐level sensing of understory vegetation patterns (McClelland et al., 2019; Tuanmu et al., 2010; Zhao et al., 2020), but also demonstrate a way forward through CTs to acknowledge these differences when making ecological inferences.…”

“…Careful sampling design will help capture environmental signals of interest and move CT surveys away from post hoc analyses of by‐catch vegetation data, while automated extraction can help scale up such efforts. Another promising line of research is the development of integrated canopy and understory metrics, such as with satellites and CTs, to characterize habitat conditions across spatiotemporal scales (e.g., Baumann et al., 2017; McClelland et al., 2019). In the face of increasing habitat disturbance in the Anthropocene and the need for effective wildlife conservation, we strongly advocate for the use of long‐term camera trapping surveys in disturbed landscapes to investigate wildlife and habitat patterns to understand the processes underlying ecological interactions and habitat change.…”

Comparative Quality and Trend of Remotely Sensed Phenology and Productivity Metrics across the Western United States

“…Variation in spatial footprints of PhenoCam data and resolution of LSP datasets could also add noise to the comparison of LSP and near-surface datasets. New fine-resolution datasets, such as daily 30 m EVI products developed using fusion between MODIS and Landsat [79,80] or combinations of Landsat and Sentinal-2 imagery to increase temporal resolution and enable 30 m phenology retrievals [81], could better match the scale of reference data and user analysis objectives.…”

Changing spring snow cover dynamics and early season forage availability affect the behavior of a large carnivore

“…Inter‐annual variation of early season forage species targeted by grizzly bears postden emergence was assessed using daily phenology data derived from an approach known as DRIVE (Daily Remote Inference of VEgetation; McClelland, Coops, Berman et al, 2020). DRIVE uses Dynamic Time Warping to combine the daily spatial resolution of Moderate Resolution Image Spectroradiometer (MODIS) imagery from Aqua and Terra satellites and the 30 m resolution of Thematic Mapper (ETM, ETM+ and OLI) imagery from Landsat 5, 7 and 8 satellite, to create a 30 m resolution daily phenology fusion product, from which daily Enhanced Vegetation Index values are derived at a 30 m spatial resolution over the entire Yellowhead BMA from 2000 to 2018.…”