Hyperspectral remote sensing can advance the study of agricultural crops and inform decision making on issues of food and water security. Spaceborne hyperspectral data have recently become more widely available with the launch of two new instruments: German DESIS (Deutsches Zentrum für Luft- und Raumfahrt [DLR] Earth Sensing Imaging Spectrometer) and Italian PRISMA (PRecursore IperSpettrale della Missione Applicativa). Advances in cloud computing and machine learning/deep learning (ML/DL) processing enable analysis of these massive hyperspectral datasets over global extents and decadal timespans. However, ML/DL classification is dependent on the existence and accessibility of high-quality reference datasets. To create comprehensive datasets for training, testing, and validating ML/DL classification algorithms, USGS researchers are compiling a Global Hyperspectral Imaging Spectral-library of Agricultural crops (GHISA). GHISA consists of spectral signatures of major crops of the world, including wheat, rice, corn, soybeans, barley, and cotton, acquired across a variety of sensors, vegetation developmental stages, and growing conditions.
GHISA can be used to classify crop types and their growth stages, estimate crop characteristics, and assess the relative advantages of different sensors for agricultural research. To avoid the curse of high data dimensionality, in which the number of bands is potentially close to or greater than the number of observations, GHISA can also be used to identify optimal hyperspectral narrow bands (OHNBs), which can inform band selection in future missions such as National Air and Space Administration’s Surface Biology and Geology satellite or Landsat Next. Hyperspectral vegetation indices (HVIs) using OHNBs can enable estimation of various crop biophysical and biochemical characteristics like pigment content, plant stress, and plant biomass/yield. See the GHISA website for details on this research. GHISA datasets are currently available for the conterminous U.S., Central Asia, and the Central Valley in California.
Spectral libraries of study crops. A) Spectra from Deutsches Zentrum für Luft- und Raumfahrt (DLR) Earth Sensing Imaging Spectrometer (DESIS), acquired in June; b) spectra from PRecursore IperSpettrale della Missione Applicativa (PRISMA), acquired in June; and c) DESIS spectra acquired in August. All spectra are averaged by crop type; the number of spectra (N) used to calculate the average is the same for all three plots. The associated manuscript is available here.
Disclaimer: Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.