Project Portfolio
Featured Projects
  • First Detection of New Construction Starts

    Researcher: Jason Schatz, Ph.D.

    Uses changes in SAR backscatter (from persistently low to high when buildings are constructed) to detect probable new building construction. Current study area is in China.

    Slides FAQ

    Last Updated: 6/8/18

  • California Cropland Classification

    Researcher: Carly Beneke, M.S.

    The CA cropland classification is a field-level characterization of dominant crop types in the Central Valley and California coast - namely, the classification identifies almonds, annuals, grapes, olives, pistachios, walnuts, and other perennial tree crops. We use multispectral data to characterize the crop phenology and high resolution imagery to characterize the spatial structure of the field.

    FAQ

    Last Updated: 5/30/18

  • Off-Shore Wind Turbine Detection

    Researcher: Ryan Keisler, Ph.D.

    GIF shows synthetic aperture radar (SAR) from the Sentinel-1 satellites, 2016 vs 2018. The grids of points correspond to offshore wind turbines. Europe's offshore wind power is growing at a rapid pace ("Europe's total offshore wind capacity increased by 25 percent in 2017, according to WindEurope" from link). Renewable sources of energy such as wind power are key to reducing fossil fuel power generation and greenhouse gas emissions. In addition to the GIF there is also a side-by-side version.

    FAQ Video

    Last Updated: 5/30/18

  • Detecting Industrial Facilities by Flare Spotting

    Researcher: Jason Schatz, Ph.D.

    Similar to oil and gas detection, below, the first phase of this project uses spectral information from Landsat-8 and Sentinel-2 to detect flares and other high temperature point sources. The second phase uses these data as well as VIIRS Nightfire data to monitor activity at industrial facilities with open flares or buildings whose roofs reach very high temperatures. No studies have yet been done on industrial facilities.

    FAQ

    Last Updated: 6/8/18

  • Detecting Oil and Gas Facilities by Flares

    Researcher: Jason Schatz, Ph.D.

    The first phase of this project uses spectral information from Landsat-8 and Sentinel-2 to detect natural gas flares. The second phase uses these data as well as VIIRS Nightfire data to monitor natural gas flaring and estimate production. Current main study areas are the Permian (TX and NM) and Bakken (ND) formations.

    FAQ

    Last Updated: 6/8/18

  • Fire Event Captured with Sentinel 2 Data

    Researcher: Ryan Keisler, Ph.D.

    From our Sentinel-2 A/B dataset, these image combinations demonstrate how the shortware and near infrared bands in the sensor capture natural disasters. The infrared range penetrates the smoke from the fire, allowing us to view the flames.

    FAQ

    Last Updated: 06/6/18

  • Real-time Weather Data Layer

    Researcher: Ryan Keisler, Ph.D.

    Our real-time weather dataset, GOES, ingests 1-km resolution image of CONUS within a few minutes of its capture every 5 minutes.

    FAQ Video

    Last Updated: 6/1/18

  • National Water Mask from 1984 - 2015

    Researcher: Karen McKinnon, Ph.D.

    The water mask uses multispectral information from Landsat 4-8 and Sentinel 2 to estimate the probability of water on a pixel by pixel basis. The underlying algorithm is a boosted random forest. The model has been applied across the continental United States as well as at select dams around the world, and can be easily applied to other regions of interest. Around 150 TB of data were processed for this analysis.

    Last Updated: 5/29/18

  • Water Mask of Aswan Egypt, 2015

    Researcher: Karen McKinnon, Ph.D.

    The water mask uses multispectral information from Landsat 8 and Sentinel 2 to estimate the probability of water on a pixel by pixel basis. The underlying algorithm is a boosted random forest. This instance displays a portion of the Nile River just downstream of the Aswan Dam.

    FAQ

    Last Updated: 06/06/18

  • Palm Classification and Deforestation

    Researcher: Caitlin Kontgis, Ph.D.

    The rice paddy classification uses VV and VH backscatter data from Sentinel 1 SAR and the MapZen DEM as input feature data to train a random forest classifier. Annual mean, median, standard deviation, maximum and minimum values for backscatter data are computed separately for ascending & descending passes. We hand-labelled 380 points across Asia and labelled them as 'rice' or 'not-rice', and these were used to train and test the model. The final map is a classification of rice at the pixel-level.

    FAQ

    Last Updated: 06/06/18

  • Rice Paddy Classification over Asia

    Researcher: Caitlin Kontgis, Ph.D.

    The rice paddy classification uses VV and VH backscatter data from Sentinel 1 SAR and the MapZen DEM as input feature data to train a random forest classifier. We hand-labelled 380 points across Asia and labelled them as 'rice' or 'not-rice', and these were used to train and test the model. The final map is a classification of rice at the pixel-level.

    FAQ

    Last Updated: 06/06/18

  • Image Segmentation - Fields

    Researcher: Rick Chartrand, Ph.D.

    This internal layer is a segmentation of the land into regions of persistent, consistent usage meant to identify agricultural fields.he main step is looking for edges that persist through time, which has the additional effect of eliminating the influence of clouds (which produce non-persistent edges). The finding of edges amounts to computing the gradient (in the sense of calculus) for each image of an area, over a long stretch of time (years). The input data are from Landsat 7 and 8, Sentinel-2, and Rapid Eye.

    FAQ

    Last Updated: 5/30/18

  • Panama Canal Relief Visualization using SAR

    Researcher: Karla King

    The two bands produced by the Sentinel-1 vertically or horizontally oriented with respect to the direction the energy is traveling.

    FAQ

    Last Updated: 6/8/18

  • California Central Valley Relief Visualization

    Researcher: Karla King

    Central Valley of California as seen by the SRTM Digital Elevation Model with Landsat 8 imagery beneath. We are visualizing the slope of the DEM where white represents the most intense slope. You could also select altitude and aspect.

    FAQ

    Last Updated: 06/06/18

  • Ice Shelf Calving

    Researcher: Sam Skillman, Ph.D.

    Time Series of our Sentinel-1 product. Sentinel-1 consists of active SAR data that produces two bands: one vertically and one horizontally oriented with respect to the direction the energy is traveling.

    FAQ

    Last Updated: 6/8/18

  • Descartes Labs Surface Reflectance Product

    Researcher: Nathan Longbotham, Ph.D.

    Descartes Labs processes these data to surface reflectance using the USGS Landsat Surface Reflectance Code (LaSRC) which removes the effect of atmospheric interference. The visible bands (red, green, blue) are pansharpened to 15m. This Landsat 8 product is available from February 1, 2013 to April 21, 2017.

    FAQ

    Last Updated: 6/8/18

  • Descartes Labs Cloud Mask Product

    Researcher: Ryan Keisler, Ph.D.

    Descartes Labs Cloud and Cloud Shadow Masks for use in tandem with Landsat 8.

    FAQ

    Last Updated: 6/8/18

  • Descartes Labs Composite NDVI

    Researcher: Mike Warren, Ph.D.

    The DL Normalized Difference Vegetation Index is a layer that can be used to analyze whether the target being observed contains live green vegetation or not. Pure green represents high NDVI while red represents a lack of vegetative vigor.

    FAQ

    Last Updated: 6/6/18

  • Oil Tanks Observed from High Resolution Imagery

    Researcher: Ryan Keisler, Ph.D.

    Our 1 - 5 cm resolution image product depicts objects not visible in medium-resolution imagery. The cadence of capture ranges global from quarterly to annual. The imagery provider is the European company, AirBus.

    FAQ

    Last Updated: 5/30/18