Estimation of Wheat Biophysical Parameters by Inverting Remote Sensing Data using 3D Models of Plant

1
Estimation of Wheat Biophysical Parameters by
Inverting Remote Sensing Data using 3D Models of
Plant Dynamics at Optical and Microwave
Wavelengths

• P. Lewis, P. Saich, J. Hillier, J. Watt, M.
  Disney
• (University College London, UK),
• B. Andrieu und C. Fournier
• (INRA Versailles-Grignon, France),
• T. Macklin, J. Bodley,
• (BAE Systems, UK)

2
The Remote Sensing Problem

• Estimate a relevant set of parameters of area
  viewed from remote measurements of scattered
  radiation
• Parameter set depends on task, but value is in
  application to large areas
• e.g. for crops provide estimates of
• cover type (inventories)
• crop status (yield estimates, controlling farm
  inputs)
• Models to use depend on levels of prior knowledge
• e.g. detailed (know planting date, variety,
  density)
• e.g. general (know crop is winter wheat,
  typical planting date)
• EO data sources
• Ground, airborne or spaceborne optical/thermal/mic
  rowave
• Different characteristics and capabilities
• e.g. microwave largely unaffected by cloud cover,
  but lower information content than optical

3
Aims of study

• Develop models to predict scattered radiation
  from (winter wheat) crop
• Optical and microwave
• Use to model scenarios for new sensors
• Develop inversion strategies
• Estimate parameterisation of crop from EO data
• Funding
• European Space Agency (initial work, completed)
• BNSC (SHIVA project, near completion)
• NERC (follow on to SHIVA, 3 years)

4
ESA study

• Tool building
• ADEL-wheat predict development of plant
  structure
• Function of integrated thermal time
• drat optical modelling tool (MCRT)
• Developed methods for efficient modelling of
  hyperspectral data and describing impacts of
  structure
• CASM microwave modelling tool
• Operation
• ADEL-wheat predicts struture, input leaf and soil
  material properties (pigments, water) to
  predict EO signal
• Demonstrated
• Some ability to model signal
• comparisons with EO data (single date data for
  optical)
• Even though different variety to that use to
  build ADEL-wheat

5
Main Research Questions Arising

• Need to test ADEL-wheat more and develop further
• No concept of senescence
• too many leaves
• No tiller mortality
• No concept of development of material scattering
  properties
• Leaf pigments, water content
• Dont know how model operates/differs for wider
  conditions
• Other varieties
• Other plant densities
• Need to test against wider EO datasets

6
SHIVA study

• Availability of comprehensive EO dataset
• 2 years (only 1 used)
• 3 times per year
• Plots with different varieties and treatments
• airborne optical data (visible/NIR)
• airborne polarimetric SAR (X,C,L band)
• Field measurements at overflight time
• Generalised canopy parameters
• LAI green leaf number tiller number density
  height

7
2003 plots
 

 
8
Modifications to ADEL-wheat

• ADEL-wheat parameterised model of plant number
  density, essentially controlling
• tiller number density
• Leaf number and dimensions and development
• Given limitations of current ADEL-wheat
• Attempt to use ground data from 2002 to
  calibrate model
• No measurements of key ADEL information
• e.g. number of leaves on main stem
• e.g. number of tillers per plant
• Many biophysical parameters measured have large
  uncertainties
• E.g. tiller number density
• No simple relationships observed
• Decided to
• Enforce mean leaf number per tiller
• Parameterise with simple tiller number density
  behaviour

9
tiller number density observed and modelled
Ntillers/plant enforced constant, but general
behaviour of Ntillers/plant reasonable
10
Enforcement of tiller number density

• ADEL-wheat develops tillers
• If density over threshold, latest tillers
  chopped
• Threshold is set (model parameter)
• But include empirical model of decrease for later
  stages
• Decrease to half maximum value between 2000 and
  2500odays
• No real evidence for this parameterisation
• Tiller number density effectively becomes main
  driving variable for structure
• (other than thermal time)
• Model seen as interim measure
• Prefer to use physiological model

11
Need to modify leaf number empirical model
mean Nleaves/tiller Calibrate relationship on
2002 data
12
But not good relationship for all of 2003 data
So, either empirical model is poor, or thermal
time not only control on late-planted crop
development (or both!)
13
Using this model, variation seen in peak leaf
number density as function of plant number
density
14
Measured (Delta-T) and modelled PAI Discrepancies
for late crop (2003) Over-estimate of PAI post
2000odays? effect of plant number density (via
leaf number density and leaf area)
15
ADEL-wheat plant height different to measured -
varietal differences? - no real impact on optical
simulations, but can have effect at microwave
16

• Mean leaf length
• impact of plant number density on modelled
  values
• - no clear pattern from measured data
• reasonable match of measured / modelled
• - But early season discrepancy

17
Summary ADEL-wheat

• Two pragmatic modifications made to ADEL-wheat
• Fix tiller number density
• Impose mean leaf number per tiller function
• Tiller number density then becomes main driving
  variable
• Relatively small residual effects of plant number
  density
• Through variations in leaf number density and
  leaf dimensions
• Despite lack of data for modifications and
  application to different varieties
• Many canopy parameters provide reasonable match
  to simulated
• Not plant height
• Issues with late planted crop

18
Optical Simulations
A) 1500odays
B) 2000odays
19
Develop approximate representation of canopy
scattering
Derive parameters using MCRT (error lt 0.01)
20
Allows expression of sensitivity
21
Sensitivity to (Thermal) time
22
Comparisons with measured EO data

• Experiments ongoing (SHIVA)
• Preliminary results of forward modelling
• Assume approximate thermal time known
• LUT inversion of leaf and soil optical properties
  for assumed structure (/- 100odays)
• Soil brightness Chlorophyll Leaf dry matter
  Senescence
• In summary
• Able to simulate early-planted crop well
• Large error in simulating late-planted crop
• ADEL-wheat does not predict structure of this
  crop well at present, as seen

23

• 1st pass forward modelling
• Still need to further investigate atmospheric
  effects
• But ability to reconstruct measured signal
  generally promising

24
Where Next?

• Approach shows much promise
• Allows consistent simulation of optical and
  microwave
• Used only first pass ADEL-wheat model
• With heuristics to impose sensible behaviour
• Under new funding (2004-2007)
• Investigate ADEL-wheat parameterisation in more
  detail
• Field trials, INRA Grignon, 2004, 2005
• Examine mainly
• Senescence
• Varietal differences
• Leaf optical properties?
• Take canopy-scale measurements during experiment
• Canopy cover
• Spectroradiometric data (400-2500nm)
• Test whether model operates correctly at this
  scale
• Information on tillering?
• From Wageningen experiments?