1. School of Geography, University of Southampton, UK 2. Unité Mixte de Recherche Environnement...

Post on 13-Jan-2016

214 views 1 download

Tags:

Transcript of 1. School of Geography, University of Southampton, UK 2. Unité Mixte de Recherche Environnement...

1. School of Geography, University of Southampton, UK2. Unité Mixte de Recherche Environnement Méditerranéen et Modélisation des Agro-Hydrosystèmes, INRA, France3. Exploitation & Services Division, Industry Section, ESA-ESRIN

PHAVEOS - the PHenology And Vegetation EO Service.

Presented by: Thomas Lankester

18th June 2010

Lankester, T., Dash, J.1, Baret, F.2, Koetz, B.3 & Hubbard, S.

2

Objectives

Provide time series of a range of vegetation parameters, utilising the unique spectral, spatial and temporal resolution of the MERIS instrument

Make spatially and temporally continuous time series available through visualisation and download of maps and phenology curves for individual locations

Support the development of a validated baseline time series (2005 - ) in advance of the launch of Sentinels 2 and 3

3

Level 1b to Level 3 processing approach

Convert MERIS Level 1b data to Level 3 gridded maps, on a daily basis

geometric correction

radiometric correction

atmospheric correction

derive biophysical variable(s)

resample direct to target map grid (latlong, OSGB36, Irish Grid…)

4

Level 1b geometric accuracy issues

5

Level 1b geometric accuracy issues

6

Level 3 geometric accuracy

To preserve geometric fidelity, resampling into the target map grid is carried out in a single step

To conserve the scene statistics area weighted (flux-conserving) resampling is used

The blue grid represents the input (swath) data grid and the yellow grid the target map grid

7

Step 1 apply a cubic spline interpolation of the raw data to generate a continuous time series

0

0.5

1

1.5

2

2.5

0 50 100 150 200 250 300 350 400

Days

MT

CI

Cubic spline

Raw data

Level 4 processing - interpolation

8

Level 4 processing - smoothing

Step 2 smooth using a local weighted least squares regression (if no negative noise bias)

0

0.5

1

1.5

2

2.5

0 50 100 150 200 250 300 350 400

Days

MT

CI

rloess

Raw data

9

Level 4 processing – interpolation metrics

3

10

Level 4 processing – smoothing metric

11

Level 3 validation

Moving from Stage 1 to Stage 2(+) validation requires considerable product generation

PHAVEOS is utilising the ESA Grid Processing On-Demand (G-POD) environmentBased on MERIS FRS data from 2005 – present, will deliver a range of Level 3 and Level 4 time seriesLAI, fAPAR, fCover, MTCI, NDVI, 2G_RBi, ….

Provision of Level 3 products for MODIS match up sites (N. America)

Coverage of PAR@METER sites

OnLine Interactive Validation Exercise (OLIVE)

12

Level 4 validation

Land Surface Phenology product validation methods TBD.

issues of spatial disparity where Sentinel 2 could bridge the gap.

Access to Forestry Commission intense monitoring sites (leaf litter collections, phenocams)

Access to UK Phenology Network

Access to tropical (DRC) deforestation ground truth

13

Web Map Service dissemination concept

1

1.2

1.4

1.6

1.8

2

2.2

2.4

2.6

2.8

3

1-Ja

n

22-J

an

12-F

eb

5-M

ar

26-M

ar

16-A

pr

7-M

ay

28-M

ay

18-J

un9-

Jul

30-J

ul

20-A

ug

10-S

ep1-

Oct

22-O

ct

12-N

ov

3-Dec

24-D

ec

Days of Year

MT

CI

2004

2005

2006

53.756776 -1.774791

1

1.2

1.4

1.6

1.8

2

2.2

2.4

2.6

2.8

3

1-Ja

n

22-J

an

12-F

eb

5-M

ar

26-M

ar

16-A

pr

7-M

ay

28-M

ay

18-J

un9-

Jul

30-J

ul

20-A

ug

10-S

ep1-

Oct

22-O

ct

12-N

ov

3-Dec

24-D

ec

Days of Year

MT

CI

2004

2005

2006

54.000019 -1.837854

14

Phenology metrics – what is the point?

Why use, and validate / inter-compare, basic phenology statistics?

Loss of information from a continuous time series (are we hiding intra-annular information)

Why inter-compare on a handful of measures when full time series are available?

Extraction of metrics is sensitive to interaction of smoothing and metric extraction methods

Different users are interested in different aspects of time series (phenology curves)

Are simple metrics capturing a relevant reality?

15

Any questionsPhenological beauty is in the eye of the

beholder