July 3, 2016 Spring Water/non-water extent in the Sacramento Valley of California 1983-2015 ## Contact: danica.schaffer.smith@duke.edu ## Citation: Schaffer-Smith et al. 2017. Three decades of Landsat-derived spring surface water dynamics in an agricultural wetland mosaic; Implications for migratory shorebirds. Remote Sensing of Environment, 193: 180-192. doi: 10.1016/j.rse.2017.02.016 ## Description: This collection contains the following datasets: 1) Classifications of water and non-water for individual Landsat image dates. Water: value = 1 Non-water: value = 2 NoData: clouds, shadows, urban, elevations >70m, areas outside of the Central Valley Joint Venture Boundary Each filename contains the original Landsat sceneID, which stored information about the sensor, the path/row, year date (YYYYDDD), and base station that downloaded the data from the satellite. See the Landsat surface reflectance documentation for more detail about the naming conventions 2) Inputs: 2a) Python scripts used to produce water/non-water classifications 2b) Landsat scene list containing the SceneIDs for all processed images ## Image Processing Methods: All available Landsat surface reflectance images from spring (Feb - May) 1983-2015 for the Sacramento Valley (path/row 044/033) were downloaded from . An optimized spring threshold to separate water from non-water was applied to the mid-infrared band for each image. Values <0.69 were classified as water, while values >0.69 were classified as non-water. See Schaffer-Smith et al. for more detail regarding the threshold optimization approach. A series of masking and clipping operations were performed to produce the final maps to exclude clouds, shadows, urban areas, and steep topography. Cloud and shadow regions are identified in the cfmask band of the surface reflectance dataset (Zhu et al. 2015). Values = 2, 4 were reclassified to NoData, while all others were classified as 1. urban regions were identified from the USDA National Agricultural Statistics Service (USDA NASS) cropland data layer (2014, Value = 23, 24, 25). Pixels in steep mountainous areas(>70m) elevation were also excluded based on the 10-m National Elevation Dataset (USGS). Finally, regions outside of the Central Valley Joint Venture (CVJV) boundary, which is based on watersheds, were masked out (Ducks Unlimited 2014). For SLC-off Landsat 7 images, which have artifacts due to slide line corrector failure, additional processing was required. See Schaffer-Smith et al. for more information about this issue. Inverse distance weighted (IDW) interpolation was applied to the thresholded water/non-water map, guided by county land use survey boundaries from the California Department of Water Resources (2016). The cfmask band of each SLC-off Landsat 7 surface reflectance dataset is also affected by these artifacts. To fill gaps in the cloud and shadow mask before final masking and clipping steps, IDW interpolation was also applied to the cloud and shadow mask. For more detail regarding image processing and analysis methods, see Schaffer-Smith et al. ## References: California Department of Water Resources. 2016. County land use survey data. Accessed from Ducks Unlimited. 2014. Central Valley Joint Venture boundary. Accessed from Schaffer-Smith et al. 2017. Three decades of Landsat-derived spring surface water dynamics in an agricultural wetland mosaic; Implications for migratory shorebirds. Remote Sensing of Environment, 193: 180-192. doi: 10.1016/j.rse.2017.02.016 USDA. 2014. USDA National Agricultural Statistics Service cropland data layers. USGS. 10m seamless National Elevation Dataset. Accessed from Zhu, Z.*, Wang, S.*, and Woodcock, C.E. 2015. Improvement and expansion of the Fmask algorithm: cloud, cloud shadow, and snow detection for Landsats 4-7, 8, and Sentinel 2 images. Remote Sensing of Environment, 159:269-277. doi: 10.1016/j.rse.2014.12.014 ## Acknowledgements: This research was supported by a NASA Earth and Space Science Fellowship and an NSF Geography and Spatial Sciences Doctoral Dissertation Improvement Grant.