B055-05
Improving the spatio-temporal characterization of grass flowering in Australian rainfed grasslands using digital time-lapse photography and landscape phenology from Sentinel-2 and MODIS.

Thursday, 10 December 2020: 16:18
Virtual
HA Nguyen, University of Technology Sydney, Ultimo, NSW, Australia, Alfredo R Huete, University of Technology Sydney, Faculty of Science, Ultimo, NSW, Australia, Elizabeth E Ebert, Bureau of Meteorology, Melbourne, VIC, Australia, Paul Beggs, Macquarie University, Sydney, Australia, Kathryn Emmerson, CSIRO Marine and Atmospheric Research, Aspendale, Australia, Jeremy Silver, The University of Melbourne, Melbourne, Australia and Janet Davies, Queensland University of Technology, School of Biomedical Sciences, Brisbane, Australia
Abstract:
Grasslands constitute a key component of the terrestrial biosphere and are fundamental to the meat and dairy industries. Flowering is the interim step between maximum plant height/ green biomass and plant reproductivity, from which pollen production and emission directly affect biodiversity, biological invasion, agricultural intensification and thunderstorm asthma events. However, this phenophase has so far been difficult to observe via ground or remote sensing.

We aim to improve the spatio-temporal characterization of grass flowering in eastern Australia rainfed grasslands for the year 2019 by combining atmospheric (transport) modeling, citizen science, digital time-lapse photography (phenocam) and satellite remote sensing of grassland. We used pollen sampling as indicator of high flowering activity and selected high pollen days in the pollen-monitoring season of 2019 to perform simulations of wind trajectories around pollen samplers in Queensland, New South Wales and Victoria. We then obtained Sentinel-2 satellite data along these trajectories and extracted the Red edge position (REP) for each Sentinel pixel. Sentinel-2 REP temporal profiles were compared to actual flowering observations of collocated phenocams in order to determine the best parameters from Sentinel-2 that approximate the first date of grass flowering. We also explore and compare the capabilities of Sentinel-2 red edge against greenness and yellowness indices from Sentinel-2 and the Moderate Resolution Imaging Spectroradiometer (MODIS) in capturing the transition from vegetative development to reproductive development and senescence in grassland.

Our study is a novel attempt to extend previous studies on hyperspectral reflectances of maize, which showed that the red edge position well indicated the onset and progression of maize pollen release, to grass flowering using Sentinel-2 red-edge bands. As projections of grassland growth and productivity from model inter-comparison studies diverge greatly under future climate change scenarios, characterization the spatio-temporal patterns of grass flowering will contribute to understanding of a key phenophase which is otherwise unobservable from space but which is critical to agriculture, biodiversity, ecology, geo-health and climate change applications.

<< Previous Abstract | Next Abstract >>