CTCD UCL report for 2002 plans for 2003

1
CTCD UCLreport for 2002plans for 2003

• P. Lewis, P. Saich
• M. Disney, T. Quaife

2
Activities 2002

• Research strands
• 5 EO methodology development data analysis
• 6 Integration of EO into ecological models
• Specific projects
• Analysis of Hyper-spectral data
• 5.1 3D Dynamic scattering models using SHAC data
• UCL
• 5.2 Biophysical parameter recovery using
  CHRIS-PROBA
• UCL/Edinburgh
• 5.3 Photosynthetic activity timing in UK
• UCL/Sheffield (close involvement with 6.1)

3
Status and 2003 plans 1

• 5.1 Analysis of Hyper-spectral data
• (Disney 10/02)
• 5.1 3D Dynamic scattering models using SHAC data
• COMPLETE
• Closely-related work for ESA completed Jan 2003
• Milestone 2.5.1 short version report to be
  submitted to CTCD
• Overview follows
• 5.2 Biophysical parameter recovery using
  CHRIS-PROBA
• MODIFIED
• Biophysical parameter recovery from
  hyper-spectral and multi-angular data
• Original plan BOREAS/CHRIS-PROBA
• Updated plan follows

4
Status and 2003 plans 2

• 5.3 Photosynthetic activity timing in UK
• (Quaife 01/03)
• ONGOING
• Initial results available
• interactions with Sheffield
• links with Boston UMD
• Milestones
• 2.5.3 (9/02) Data preparation - much data
  collected, but ongoing activity
• 2.5.4 (12/02) Report and datasets on UK
  phenology - some datasets produced -
  validation in progress
• Further plans for 2003
• sub-pixel scale
• microwave phenology
• model/data preparation for data assimilation

5
3D Dynamic scattering models using SHAC data

• Project 5.1
• Mainly prior to PDRA employment
• Saich, Lewis, Disney ESA Contract
• AO/1-3679/00/NL/NB
• Development of Architectural Vegetation Growth
  Models for Remote Sensing Applications
• Peter Van Oevelen (WAU), Dirk Hoekman (WAU),
  Martin Vissers (WAU), Ian Woodhouse (Edinburgh)
• Results and approaches relevant to CTCD

5.1 3D Dynamic scattering models using SHAC data
6
3D Dynamic scattering models using SHAC data

• Motivation and Approach
• Use models of vegetation structural dynamics
• species/variety specific
• developing methods/understanding ( monitoring
  known targets)
• provide expectation of structure as function of
  time
• linkages between (generalised) biophysical
  parameters
• development of biomass, LAI etc.
• consistent framework for optical and microwave
• e.g. optical signal driven by LAI, but
  microwave more complex relationship with leaf
  dimensions thickness

5.1 3D Dynamic scattering models using SHAC data
7
3D Dynamic scattering models using SHAC data

• Use models of vegetation structural dynamics
• ESA contract involved wheat scots pine
• wheat work continuing through external funding
• SHIVA BNSC (NEWTON) with BAE systems UWS
• NERC NOT proposal under review

5.1 3D Dynamic scattering models using SHAC data
8
Development of Vegetation Growth Models for
Remote Sensing Applications

• the usefulness of vegetation growth models
  coupled to remote sensing data to yield
  architectural information in order to improve the
  retrieval of information for agricultural and
  forestry applications. AO 3679 Work Statement
• develop vegetation models
• growth, architecture
• for use with forward reflectance / scattering
  models
• optical, microwave
• improve retrievals using EO data
• optical, microwave, synergy

5.1 3D Dynamic scattering models using SHAC data
9
Principal developments / enhancements

• drat
• Efficient optical (ray-tracing) model
• casm
• Single-scattering microwave model (polarimetric
  and insar)
• ADEL-wheat
• New dynamic 3-d model for wheat
• TreeGrow
• 3-d tree model exploited by optical and microwave
  models

5.1 3D Dynamic scattering models using SHAC data
10
Datasets

• Principal datasets - SHAC June 2000 (two weeks
  apart)
• Thetford
• E-SAR, X-VV and L-polarimetric Insar, Hymap
• Tree heights, densities, dbh, canopy depth,
  biomass
• AirSAR (1991), SIR-C (1994)
• (others in the study though they are not relevant
  here)

5.1 3D Dynamic scattering models using SHAC data
11
L-HH, VV, HV (? ? / sin?)
Hymap (17 June 00)
5.1 3D Dynamic scattering models using SHAC data
12
Tree model TreeGrow
5.1 3D Dynamic scattering models using SHAC data
13
TreeGrow

• Parameterised using destructive tree measurements
• for Scots Pine, Norwegian Spruce
• parameterisation of structure into (e.g.)
  cylindrical elements
• x,y,z position, length, radius, status (live /
  dead), parent/child
• can be up to (e.g.) 250 000 elements

5.1 3D Dynamic scattering models using SHAC data
14
TreeGrow comparison with data

• Close agreement in height (and no reduction with
  age!) with measured Thetford data
• Different thinning in Finnish stands?

• Close agreement between modelled and measured dbh
  (Thetford)
• Different thinning in Finnish stands?

5.1 3D Dynamic scattering models using SHAC data
15
Forest stand reflectance

• Temporal trend reasonable
• Magnitude ok in NIR, but not in visible

5.1 3D Dynamic scattering models using SHAC data
16
Comparison with L- and P-band AirSAR data for
Thetford (1991)
5.1 3D Dynamic scattering models using SHAC data
17
Comparison with interferometric heights for
Thetford (2000)


5.1 3D Dynamic scattering models using SHAC data
18
Forest parameter retrievals

• Four stands selected from HyMAP data
• 10 years of age
• Better fit in visible than for 5 year old stands
  (less visible soil?)
• Bottom left case very much poorer
• Is this in fact 10 year old stand?
• Table shows large range of retrieved ages.
• Indicative of variability of stand reflectance
• Need better angular sampling (for .e.g) to pin
  this down

5.1 3D Dynamic scattering models using SHAC data
19
Forest parameter retrievals

• Stands of 20 to 34 years of age
• Not a terribly good fit, except for high
  threshold in lower left case
• 34 years old fits better in water absorption AND
  visible
• Known issues with absorption spectra in LIBERTY
  needle reflectance model
• Understory is important assumed Lambertian soil
  spectra here
• BUT, in reality

5.1 3D Dynamic scattering models using SHAC data
20
Forest parameter retrievals

• Stands 42, 53 and 74 years of age
• Better fits
• See less and less understory??
• For 53 74 years old, outside range of ages
  simulated
• Outside LUT here, so extrapolation not
  interpolation valid?
• BUT parameters vary little from 45-50 years on,
  so maybe.

5.1 3D Dynamic scattering models using SHAC data
21
Conclusions

• Shown that the models can be linked in a sensible
  way
• 3d tree model, optical reflectance, microwave
  scattering
• And that they can match data
• Though could do with better (more appropriate?)
  datasets (eg multi-angular, longer wavelength
  microwave
• And that they can sometimes be simplified
• Eg parametric representation of optical model
• Inversions less clear
• though its unclear whether this is a data problem
  or a model problem
• Route forward
• New datasets
• generalising the description of tree structure
  (ie less complex?)

5.1 3D Dynamic scattering models using SHAC data
22
Publications

• 3 IGARSS papers
• P. Lewis, P. Saich, M. Disney Modelling the
  radiometric response of a dynamic, 3D structural
  model of wheat in the optical and microwave
  domains
• M. Disney, P. Saich and P. Lewis Modelling the
  radiometric response of a dynamic, 3D structural
  model of Scots Pine in the optical and microwave
  domains
• P. Saich, P. Lewis, M. Disney Biophysical
  parameter retrieval from forest and crop canopies
  in the optical and microwave domains using 3D
  models of canopy structure
• Full Journal publications to be prepared 2003

5.1 3D Dynamic scattering models using SHAC data
23
5.2 Biophysical parameter recovery using
CHRIS-PROBA
Disney

• Original plan BOREAS/CHRIS-PROBA
• Problems with CHRIS-PROBA
• pointing accuracy
• now OK (5 overlapping images)
• Logistical problems wrt BOREAS
• other groups fieldwork plans uncertain
• So milestone 2.5.2 delayed

5.2 Biophysical parameter recovery from
hyper-spectral and multi-angular data
24
5.2 Biophysical parameter recovery from
hyper-spectral and multi-angular data
Disney

• Expanded work package
• group activities involving testbed optical
  instruments
• CHRIS-PROBA, HyMAP, Hyperion, etc.
• Defined test sites
• (Thetford)
• San Rossore
• Harwood

5.2 Biophysical parameter recovery from
hyper-spectral and multi-angular data
25
San Rossore
Disney

• CarboEurope site
• Maritime Pine
• liaison with Italian groups
• Data
• CHRIS-PROBA
• overflights
• April, June, July
• Fieldwork
• June/July UCL/Edinburgh
• NERC ARSF 2004?

5.2 Biophysical parameter recovery from
hyper-spectral and multi-angular data
26
San Rossore
Disney
CHRIS-PROBA San Rossore
5.2 Biophysical parameter recovery from
hyper-spectral and multi-angular data
27
Harwood
Disney

• CarboEurope site
• Sitka spruce and Lodgepole pine
• Sitka spruce stands
• Unplanted control, 5- 40 years old
• Close liaison with Edinburgh
• Detailed stand information from FR

5.2 Biophysical parameter recovery from
hyper-spectral and multi-angular data
28
Harwood
Disney

• Data
• NERC ARSF (NERC ARSF/EPFS awards)
• overflights fieldwork
• April, Sept.
• ATM, CASI, air photos, LiDAR
• Helicopter flights
• April-Sept
• with mounted spectroradiometer
• Purchase Hyperion?
• EO-1 hyperspectral (30m resolution 400-2500nm)

5.2 Biophysical parameter recovery from
hyper-spectral and multi-angular data
29
Harwood San Rossore
Disney

• Activities
• continuing activities to explore biophysical
  information extraction from optical data
• specifically
• high resolution / quality / information content
  data collection
• optical model development and testing
• links to other CTCD projects
• 3D models
• simpler models
• LUE/PRI with Edinburgh
• inversion strategies

5.2 Biophysical parameter recovery from
hyper-spectral and multi-angular data
30
Harwood San Rossore
Disney

• optical model development and testing
• links to other CTCD projects
• build LUTs / emulators for sensitivity use in
  assimilation?
• 3D models
• detailed representations
• extend from 5.1
• explore detailed sensitivities (e.g. PRI)
• explored development with CIRAD (expensive),
  following up with Lignum group (Finland),
  possibly Strathclyde (Stats modelling science -
  Eric Renshaw?)
• simpler models
• e.g. layered turbid medium
• impacts of structural representation

5.2 Biophysical parameter recovery from
hyper-spectral and multi-angular data
31
Harwood San Rossore
Disney

• Achievements 2002
• Oct 2002 start
• NERC ARSF/EPFS submitted approved
• To do 2003
• data collection large effort
• revised Milestone 2.5.2 (Harwood and San Rossore)
• Dec 2003
• Canopy radiation models over test sites
• Oct 2003
• 3D model development and testing? Dec 2003?
• but may need to delay
• PRI other Sensitivity studies
• Dec 2003 - Edinburgh link to Sheffield stats?

5.2 Biophysical parameter recovery from
hyper-spectral and multi-angular data
32
Publications

• Participate in SPECTRA meeting at ESTEC (October
  2003)
• Follow up with full Journal publication
• IGARSS paper
• G. Thackrah P. Lewis T. Quaife and M. Barnsley
  An initial analysis of CHRIS-on-board-PROBA data
  for the purposes of biophysical parameter mapping
  over a variety of land cover types.
• Publication in preparation
• linked to NERC NOT project, but very relevant to
  CTCD

5.2 Biophysical parameter recovery from
hyper-spectral and multi-angular data
33
5.3 Photosynthetic timing
Quaife

• AIM
• To provide vegetation canopy timing information
  to drive and validate vegetation growth models.
• Objectives
• Comparison of products from various sources.
• Delivery of 1km phenology maps of the UK.
• Uncertainty analysis of delivered product.

5.3 Photosynthetic timing
34
Quaife
Milestones 2002

• Milestone 2.5.3, September 2002 Data
  preparation complete. achieved
• ongoing MODIS
• Milestone 2.5.4, December 2002 Report and
  datasets on UK vegetation timing data for last 10
  years.
• in preparation

5.3 Photosynthetic timing
35
Quaife
Acquired EO data for the UK

• AVHRR FASIR data, 8km (82-99)
• SPOT-4 VGT S10 products, 1km (99-01)
• MODIS phenology product (01)
• Zhang - Boston University
• MODIS 500m surface reflectance (01, incomplete)
• download from GSFC (Roy)

5.3 Photosynthetic timing
36
MODIS Phenology 2001
greenup
maturity
DOY 0
DOY 365
senescence
dormancy
37
Quaife
Use Gaussian curves to represent individual plant
canopies within the pixel.
VGT S10
5.3 Photosynthetic timing
38
Quaife
VGT S10
Extract canopy timing information
39
Quaife
Intercomparison of EO products
MAX NDVI
a) AVHRR FASIR
b) SPOT-4 VGT
80
255
Day of year (1999)
5.3 Photosynthetic timing
40
Quaife
Relevance to the CTCD
Plant canopy timing information provides a direct
point of comparison with the outputs of the
GVMs. Additionally the information may form the
basis of a data assimilation scheme to couple EO
data with the models.
Links within CTCD
Discussions about the nature and the utility of
the phenology data are ongoing between UCL and
most of the other CTCD groups. Sheffield in
particular have already started trying to
incorporate the data into the SDGVM.
5.3 Photosynthetic timing
41
Quaife
Priorities for 2003

• Processing of 2001 MODIS data.
• Most accurate dataset
• use to compare with others (VGT, AVHRR)
• establish uncertainty
• Milestones interim report May 2003
• full report December 2003
• Processing of 2001 ERS data.
• 14 ERS scenes East Anglia 2000
• Examine and explain ERS phenology signal
• (with Saich)
• Milestone report June 2003

5.3 Photosynthetic timing
42
Quaife
Priorities for 2003

• Unmixing of the phenological signal.
• sub-pixel effects
• examine unmixing of signal
• use land cover
• assume signal locally constant per land cover
  class
• Milestone August 2003 sub-pixel processed data
  for E. Anglia
• Planning for assimilation scheme
• work with modellers on radiation model
  interfacing
• Milestone April 2004 models and documentation

5.3 Photosynthetic timing
43
Publications

• IGARSS paper
• T. Quaife, P. Lewis, M. Disney and M. Lomas,
  Intercomparison of phenological measures derived
  from medium and coarse resolution earth
  observation and implications for assimilation
  into vegetation growth models.
• Follow up with Journal paper 2003

5.3 Photosynthetic timing
44
Summary

• Most work completed 2002
• staff issues - delays on some reports
• Important interactions established
• esp. with Sheffield and Edinburgh at present
• also Italian teams, CLASSIC, Boston, GSFC
• Publishable results being achieved
• Clear plan for 2003

5.3 Photosynthetic timing