This child data release includes hyperspectral and RGB images acquired from an Unmanned Aircraft System (UAS) during an experiment performed at the USGS Columbia Environmental Research Center, near Columbia, Missouri, on April 2, 2019. The purpose of the experiment was to assess the feasibility of inferring concentrations of a visible dye (Rhodamine WT) tracer from various types of remotely sensed data in water with varying levels of turbidity. Whereas previous research on remote sensing of tracer dye concentrations has focused on clear-flowing streams, the Missouri River is much more turbid and the reflectance signal associated with the sediment-laden water could obscure that related to the presence and amount of dye. This experiment thus provided an initial test of the potential to map dye concentrations from remotely sensed data in more turbid rivers like the Missouri, where tracer studies involving the release of a visible dye can provide insight regarding the dispersal of endangered sturgeon larvae. The experiment involved manipulating the turbidity and Rhodamine WT dye concentration in two water tanks, acquring hyperspectral and RGB images, and attempting to infer dye concentrations from the images for varying levels of turbidity. Hyperspectral imagery (HSI) was collected with a Headwall Nano-Hyperspec (Headwall Photonics, Bolton, MA), a pushbroom sensor that measures reflectance from 400 - 1000 nm in the VNIR (visual and near-infrared). Sensor calibration was performed by collecting a dark reference with the lens cap on, and a white reference with a 25.4 cm x 25.4 cm Labsphere Spectralon? (Labsphere, INC, North Sutton, NH) calibrated diffuse reference target that reflects 99% of light in accordance with the National Institute of Standards and Technology. The sensor was mounted to a DJI Ronin-MX gimbal (DJI, Shenzhen, China) affixed to a DJI M600 Pro unmanned aerial vehicle (UAV) and flown 30 m above the tanks, yielding a ground sampling distance of 2 cm. The gimbal provides stability for the payload which aids in post-processing of the HSI. The UAV repeated a flight plan over the two tanks to create image data cubes. The resulting HSI was radiometrically corrected in the Headwall HyperspecIII SpectralView software package to convert raw digital numbers to radiance. The same software was used to orthorectify the images, which applies latitude and longitude GPS information to the cubes using data from an Xsens MTi-G-710 inertial measurement unit (IMU; Xsens, Enschede, The Netherlands). The white reference was included in each scene and used to make atmospheric corrections in ENVI (Harris Geospatial Solutions, Inc., Broomfield, CO) to convert radiance to relative reflectance. Timestamps from HSI were then compared to time stamps from the field spectra in a related data release to select only the data cubes that were nearest in time to when the field spectra were recorded.
The RGB images were acquired using the built-in 12 megapixel camera on a DJI Mavic Pro UAV with an on-board GPS that collected position data during the flights. Images were acquired on a two-second interval while the UAV hovered in the same position. The original RGB images were used directly without further pre-processing. Time stamps for the images were used to link them to turbidity and concentration measurements made in situ in each tank during the experiment. The image data are compiled in a set of zip files, two for the hyperspectral images and one for the RGB images, and a text file listing the time stamps and file names for both types of images. The hyperspectral and RGB images selected for analysis were based on the time interval during which field spectra were recorded. The RGB image closest in time to each of the hyperspectral images used was selected from the list of available RGB images.