ESTIMATING EARTH’S BIOMASS USING SATELLITE REMOTE SENSING
1
Marco Lavalle Jet Propulsion Laboratory, California Institute of Technology
objec&ves and mo&va&on
2
The overall goal of this work is to design a standard method for monitoring Earth’s ecosystems using remote sensing data collected by forthcoming radar missions, such as the ESA/Sen@nel-‐1, the Japanese ALOS-‐2 and the proposed US DESDynI satellite missions. Par@cularly, we aim at measuring canopy height and tree ver@cal structure from space. This informa@on is urgently needed to quan@fy the worldwide biomass and its varia@on over @me in order to understand the global carbon budget and climate changes. The problem of es@ma@ng biomass is currently unsolved. Here, we demonstrate a strategy that represents – we believe – an important step forward towards the solu@on of the problem.
method polarimetric space interferometer
physical model (RMoG model)
t = t1
In our approach, measurements are made using a repeat-‐pass space interferometer, i.e. a synthe@c aperture radar (SAR) that observes the same por@on of the vegetated Earth surface from two different orbital posi@ons at different @mes.
field data
Pol-‐InSAR data t = t2
valida@on signal processing and image forma@on
Electromagne@c waves are coherently transmi[ed and received with horizontal and ver@cal polariza@on states. Polarimetry “sees” different sca[ering mechanisms within vegeta@on while interferometry associates a 3D ver@cal loca@on to them. This synergy of polarimetry and interferometry (Pol-‐InSAR) is useful to monitor semi-‐ transparent media such as forests from space.
model-‐based LS inversion
Canopy height and ver@cal structure are extracted from Pol-‐InSAR observa@ons by inver@ng our physical model with a least-‐square op@miza@on algorithm. Biomass spa@al distribu@on is derived by applying ecological allometric equa@ons to the es@mated canopy height. Valida@on is performed against op@cal or field data.
allometric conversion
3
Lavalle, M., and Hensley, S., “Demonstra@on of repeat-‐pass Pol-‐InSAR using UAVSAR: The RMoG model”, proceedings of IEEE IGARSS 2012, July 2012. canopy height
biomass
physical model Key element of our method is a forward physical model that correctly relates the polarimetric-‐ interferometric degree of coherence measured by the interferometer to important vegeta@on characteris@cs. We derived the model based on electromagne@c and signal processing theories and physical assump@ons. The model is termed random-‐mo@on-‐over-‐ground (RMoG) and accounts for structural (e.g. canopy density) and dynamic (e.g. wind effects) characteris@cs of forests. The RMoG model func@on fRMoG can be wri[en in closed form as a func@on of 6 unknown bio-‐physical parameters canopy height ! $ wave ex@nc@on # & ground topography γ ( p) = fRMoG # ground-‐to-‐canopy ra@o & # & canopy sca[erers mo@on # ground sca[erers mo@on & " % where γ ( p) is the interferometric coherence measured at an arbitrary polariza@on state p.
structure func@on temporal func@on (first-‐ order approxima@on)
temporal coherence at ground-‐level temporal-‐structural coherence in the canopy layer
interferometric coherence
Lavalle, M., Simard, M., and Hensley, S., “A temporal decorrela@on model for polarimetric radar interferometers”, IEEE Trans. on Geoscience and Remote Sensing, 2011. ground sca[erers mo@on canopy sca[erers mo@on auxiliary variables
4
ground-‐to-‐canopy sca[ering ra@o
results and impact Magnitude (brightness) and phase (color) of the canopy-‐dominated coherence (top) and ground-‐dominated coherence (bo[om).
The RMoG model inversion has been tested on L-‐band UAVSAR (Uninhabited Aerial Vehicle Synthe@c Aperture Radar) data collected by the Jet Propulsion Laboratory over Harvard Forest (MA). On the leb, the canopy-‐ dominated coherence, the ground-‐dominated coherence and the es@mated canopy height map are shown. The RMoG height map has been compared against sparse lidar data for valida@on, revealing very good agreement. The es@mated mo@on of the ground sca[ering elements is considerably smaller than the mo@on of the canopy sca[ering elements. These results demonstrate the validity of the RMoG approach and provide the worldwide ecosystem science community with a new tool for monitoring and understanding environmental changes on Earth. lidar data
ground sca[erers mo@on
RMoG Canopy height es@mated from repeat-‐pass Pol-‐InSAR UAVSAR data using the RMoG model. The inversion algorithm selects only the pixels where the solu@on is believed to be correct and masks the remaining pixels (white areas).
0 m
ignoring sca[erers mo@on
30 m
This research was conducted at Jet Propulsion Laboratory, California Ins@tute of Technology, under contract with the Na@onal Aeronau@cs and Space Administra@on.
canopy sca[erers mo@on