Progress Towards Pan-Arctic Shrub Mapping using Spectral, Radar, and Stereo Metrics

Changes in tundra shrub cover and height have well-documented ecosystem effects on snow distribution, soil temperature, wildlife habitat, and other landscape properties [1]. Local-scale studies have documented shrub expansion in upland and riparian systems in many parts of the Arctic. Here we present research on modeling modern tundra shrub cover, canopy height, and leaf habit (deciduous, evergreen) for large study areas (10⁴ km² through 10⁶ km²) at 30 m resolution. We developed machine learning regression models to predict:

- Total cover of low (50-150 cm) and tall (>150 cm) shrubs [2];

- Total cover of deciduous vs. evergreen shrubs [2]; and

- Mean height of woody vegetation.

We developed a calibration and validation dataset from detailed field plot measurements, airborne LiDAR including the Land, Vegetation, and Ice Sensor (LVIS) , stereo models derived from commercial satellite imagery, and data products from Small UAS RGB cameras.

Suites of predictors included:

- Spectral predictors (Landsat and Sentinel 2);

- Polarimetric C-band Synthetic Aperture Radar (SAR) predictors (Sentinel 1); and

- Texture and height metrics derived from high-resolution commercial satellite imagery stereo pairs.

We present results from the North Slope and an intensive study site on the Yukon River Delta. The best performance was achieved with models that combined all 3 suites of predictors. These predictor datasets, and the workflow used to analyze them are suitable for developing a modern (circa 2015-2018) snapshot of shrub cover and canopy properties for the entire tundra biome (10⁷ km²). However, the radar and texture metrics are not available to support multi-temporal mapping for earlier epochs such as circa 2000. We discuss the feasibility and limitations of using globally available legacy data such as Landsat for mapping historical shrub cover.

- G.V. Frost, H.E. Epstein, D.A. Walker, G. Matyshak, and K. Ermokhina (2018). Seasonal and long-term changes to active-layer temperatures after tall shrub expansion and succession in Arctic tundra. Ecosystems 21:507-520. https://doi.org/10.1007/s10021-017-0165-5

- M.J. Macander, G.V. Frost, P.R. Nelson, and C.S. Swingley (2017). Regional Quantitative Cover Mapping of Tundra Plant Functional Types in Arctic Alaska. Remote Sensing, 9, 26 pp., http://dx.doi.org/10.3390/rs9101024