Near-real-time Herbaceous Annual Cover in the Sagebrush Ecosystem, USA, July 2019
Dates
Publication Date
2019-07-10
Time Period
2019-06-24
Citation
Boyte, S.P., and Wylie, B.K., 2019, Near-real-time Herbaceous Annual Cover in the Sagebrush Ecosystem, USA, July 2019: U.S. Geological Survey data release, https://doi.org/10.5066/P96PVZIF.
Summary
This dataset provides a near-real-time estimate of 2019 herbaceous annual cover with an emphasis on annual grass (Boyte and Wylie. 2016. Near-real-time cheatgrass percent cover in the Northern Great Basin, USA, 2015. Rangelands 38:278-284.) This estimate was based on remotely sensed enhanced Moderate Resolution Imaging Spectroradiometer (eMODIS) Normalized Difference Vegetation Index (NDVI) data gathered through June 24, 2019. This is the second iteration of an early estimate of herbaceous annual cover for 2019 over the same geographic area. The previous dataset used eMODIS NDVI data gathered through April 28, 2019 (https://doi.org/10.5066/P9ZEK5M1). The pixel values for this most recent estimate ranged from 0 to100% with an overall [...]
Summary
This dataset provides a near-real-time estimate of 2019 herbaceous annual cover with an emphasis on annual grass (Boyte and Wylie. 2016. Near-real-time cheatgrass percent cover in the Northern Great Basin, USA, 2015. Rangelands 38:278-284.) This estimate was based on remotely sensed enhanced Moderate Resolution Imaging Spectroradiometer (eMODIS) Normalized Difference Vegetation Index (NDVI) data gathered through June 24, 2019. This is the second iteration of an early estimate of herbaceous annual cover for 2019 over the same geographic area. The previous dataset used eMODIS NDVI data gathered through April 28, 2019 (https://doi.org/10.5066/P9ZEK5M1). The pixel values for this most recent estimate ranged from 0 to100% with an overall mean value of 8.24% and a standard deviation of +/-9.39. The model's test mean error rate (n = 1664), based on nine different randomizations, equaled 5.2% with a standard deviation of +/- 0.09. Overall statistics between the May and June datasets were similar. However, some individual pixel differences can be considerable and are attributed to changing conditions on the ground that are reflected in the satellite data. These changes can influence how the models relate the dependent variable to the independent variables. Both datasets were generated by integrating ground-truth measurements of annual herbaceous percent cover with 250-m spatial resolution eMODIS NDVI satellite derived data and geophysical variables into regression-tree software. The geographic coverage 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 above this threshold. To target likely sagebrush ecosystems, the mask also covered pixels classified as something other than shrub or grassland/herbaceous by the 2011 National Land Cover Dataset (NLCD). The model was not trained on any masked pixels. Cheatgrass (Bromus tectorum) is the most common annual grass in the study area, but red brome (Bromus rubens), medusahead (Taeniatherum caput-medusae), and ventenata (Ventenata dubia) are also problematic. They grow from seed, usually in spring, mature quickly, produce seed, and die. After dying, these annual grasses contribute fine fuels that facilitate fire ignition and spread throughout sagebrush ecosystems. These fires remove sagebrush stands. Increasing fire frequencies, land management practices, and development have all contributed to the fragmentation of the once expansive sagebrush ecosystems. These ecosystems are critical for water quality, reduced fire threats, and the survival of sagebrush-dependent wildlife
These data were developed to provide land managers and researchers with near-real-time estimates of spatially explicit percent cover predictions of herbaceous annual vegetation cover in the study area. Appropriate use of the data should be defined by the user; however, this data comes with caveats. First, these estimates should be viewed as relative abundances. Second, each pixel in the dataset represent 250-meters and can include a geolocation error of up to 125 meters. Comparing this dataset to similar datasets with different spatial resolutions or different dates can lead to substantial differences between datasets.