We integrated 250-m enhanced Moderate Resolution Imaging Spectroradiometer (eMODIS) Normalized Difference Vegetation Index (NDVI) with land cover, biogeophysical (e.g., soils, topography) and climate data into regression-tree software (Cubist®). We integrated this data to create a time series of spatially explicit predictions of herbaceous annual vegetation cover in sagebrush ecosystems, with an emphasis on annual grasses. Annual grass cover in sagebrush ecosystems is highly variable year-to-year because it is strongly dependent on highly variable weather patterns, particularly precipitation timing and totals. Annual grass cover also reflects past disturbances and management decisions. We produced 17 consecutive years (2000 – 2016) of herbaceous annual vegetation cover data in 1% cover increments, which allows extensive analyses of annual grass dynamics. Annual grasses, especially cheatgrass (Bromus tectorum), result in grass-fire cycles that endanger human-built structures, reduce air quality, and compromise hunting resources by destroying wildlife habitat. These sagebrush ecosystems, whose character and composition are dramatically altered, are critical for water quality and the survival of sagebrush-dependent wildlife. The geographic coverage of the study area includes the Great Basin, the Snake River Plain, the state of Wyoming, and contiguous areas. We applied a mask to areas above 2250 m elevation because annual grasses are unlikely to exist at substantial cover at or above this threshold (Boyte et al. 2016. doi: 10.1016/j.rama.2016.03.002). To target likely sagebrush ecosystems, the same mask also covered pixels classified as something other than shrub or grassland/herbaceous by the National Land Cover Database(NLCD). For the study area, the annual herbaceous cover mean for the 17-year period equaled 7.31% with a standard deviation of 7.03%. The study area's mean pixel value for the study period ranged from 0-96%, although individual years' pixel values reached 100%. Training data (n = 33,746) were harvested from 30-meter spatially explicit 2001 cheatgrass (Peterson, E.B. 2005. doi:10.1080/01431160500127815), 2006 annual grass (Peterson, E. B. 2007. A map of annual grasses in the Owyhee Uplands, Spring 2006, derived from multitemporal Landsat 5 TM imagery. Report for the U.S.D.I. Bureau of Land Management, Nevada State Office, Reno, by the Nevada Natural Heritage Program, Carson City, Nevada), and 2013 - 2015 herbaceous annual (data acquired from the USGS sagebrush ecosysem team at the Center for Earth Resources Observation and Science) datasets. These datasets were spatially averaged to 250 meters to match the resolution of the remotely sensed data.
References:
Boyte, S.P, B.K. Wylie, and D.J. Major. 2016. Cheatgrass percent cover change: Comparing recent estimates to climate change-driven predictions in the northern Great Basin. Rangeland Ecology and Management 69:265-279.
https://dx.doi.org/10.1016/j.rama.2016.03.002.
Gu, Y. B.K. Wylie, S.P. Boyte, J. Picotte, D.M. Howard, K. Smith, K.J. Nelson. 2016. An optimal sample data usage strategy to minimize overfitting and underfitting effects in regression tree models based on remotely sensed data. Remote Sensing. https://dx.doi.org/10.3390/rs8110943.
Peterson, E.B. 2005. Estimating cover of an invasive grass (Bromus tectorum) using tobit regression and phenology derived from two dates of Landsat ETM+ data. International Journal of Remote Sensing 26:2491-2507. http://dx.doi.org/10.1080/01431160500127815.
Peterson, E.B. 2007. 2007. A map of annual grasses in the Owyhee Uplands, Spring 2006, derived from multitemporal Landsat 5 TM inagery. Report prepared for: USDOI, Bureau of Land Management, Nevada State Office, Reno, NV. Nevada Natural Heritage program website -- http://heritage.nv.gov/gis; last accessed October 12, 2017.
