remote sensing for essential biodiversity variables (rs4ebv) · vrieling; marc paganini. aim and...

Post on 27-Jul-2020

0 Views

Category:

Documents

0 Downloads

Preview:

Click to see full reader

TRANSCRIPT

Remote Sensing for Essential Biodiversity Variables (RS4EBV)Brian O'Connor; Andrew Skidmore; Roshanak

Darvishzadeh; Tiejun Wang; Chris McOwen; Anton Vrieling; Marc Paganini

Aim and Objectives

• To develop methods to map and monitor EBVs from Sentinel-2 in support of biodiversity conservation, e.g. CBD Aichi Targets (5 and 15)

• Objectives:

– map direct EBVs at high temporal and spatial res

– develop a model for indirect EBV from in-situ and RS

– pilot the model and validate on 3 pilot sites across EU

Pilot sites

Lowland grasslandNorth Wyke Farm Platform,UK

SaltmarshSchiermonikoog Island,the Netherlands

Temperate ForestBavaria NP,Germany

Essential Biodiversity Variables• Direct EBVs:

– Leaf Area Index (LAI)

– Leaf chlorophyll

– Phenological metrics (SOS, EOS, Amp, Peak etc.)

• Indirect EBV:

– Functional Diversity (FD)

What is Functional Diversity

• Species richness has long been used by ecologists but is problematic in describing species function

• FD represents functional differences among the species in a community through analysis of their range of traits, e.g. leaf area for productivity

• Various FD metrics reported from an analysis of species trait values in functional trait space

The concept of functional space Functional space is a multidimensional euclidean space where axes are ecologically relevant traits Trait 1

Various metrics are computed to quantify FD (similar to PCA)

FD in practice

Same richness however community 2 (right) is clearly more functionally diverse

EBV Validation

• Phase 1: Fieldwork:

– 40 plots & 14 RGB cameras at Bavaria (July, 2015)

– 30 plots & 10 RGB cameras at Schiermonnikoog (July, 2015)

– 10 plots & 5 RGB cameras at North Wyke Farm Platform (June, 2015)

– 8 plots at Dartmoor National Park (June, 2015)

Botanical data

For each 20x20m plot:

– Floral species identification and % abundance

– In-situ traits:

• leaf chl, canopy height, LAI, fresh mass, dry mass..

– Generalised traits from the TRY database

The TRY database

Coverage of trait data

• How many and what traits can we reliably use?

• A balance is needed between traits for FD calculation and coverage of three pilot sites

– Traits selected based on their ecological relevance

– 85% threshold for coverage of trait values per species

Building a trait database

Assessment of species at sites

Compilation of species list

Checks on compatibility

Querying major databases

Inclusion of data in a species/trait matrix

Phase 1 satellite imageryImagery No. of images NL DE UK

Rapid Eye Total images acquired 9 9 9

Satisfactory quality 5 4 9

SPOT 5 Total images acquired 17 16 14

Satisfactory quality 9 10 4

Phenological analysis

Phenological model

Schiermonnikoog Bavaria

no observations!

Phenological results: SOS

Schiermonnikoog

Bavaria National ParkColor scale from:Vrieling et al. (2016): RSE 174, 44-55.

RT modelling for plant traits• Options considered:

INFORM (Atzberger 2000)

a physical based hybrid radiative transfer model for forests

PROSAIL (Verhoef, 1984, Jacquemoud & Baret, 1990)

PROSPECT leaf optical and SAILH models

Field measures in 2015 were used to parameterise the model: leaf dry matter, water and chlorophyll content, leaf area index, canopy diameter and stand height …

RT model inversionReflectance correctionValidation

INFORM results for LAI and Chl.

RGB imageRapid Eye for BNP

RGB image+ forest mask

5

10

15

20

25

30

35

40

45

50

55

LAI map 17/07/2015

Chlorophyll map 17/07/2015

Phenology metrics:

• SOS: start of season (20% of amplitude reached)

• PS: peak season (90% of amplitude reached)

• LG = PS-SOS

• AMP: amplitude between 31 August and 1 March

RS-EBV FRic FEve FDiv FDis

SOS -0.047 0.060 -0.111 -0.350

PS -0.240 -0.289 0.011 -0.476

LGS -0.221 -0.367 0.108 -0.184

max.NDVI 0.134 0.264 -0.504 -0.265

AMP 0.278 0.402 -0.266 0.051

RS-EBV FRic FEve FDiv FDis

SOS 0.819 0.771 0.589 0.058

PS 0.237 0.152 0.958 0.008

LGS 0.278 0.065 0.599 0.331

max.NDVI 0.513 0.193 0.009 0.156

AMP 0.169 0.042 0.189 0.791

Linking direct and indirect RS-EBVPearson’s r

P-value

Next steps• Apply RS algorithms to Sentinel 2 data in 2016

• Develop plot-FD-EBV relationships further

• Build (GLMM) model to predict FD from a combination of RS-EBVs as well as abiotic data (met, soil etc.)

• Apply the model at coarser spatial units, e.g. land cover, and scale up

Thanks for listening

brian.o’connor@unep-wcmc.org

Botanical diversity

BFNP

North WykeSchierm.

174

3026

11 18

1

Festuca rubra

>15% of European botanical species listed in TRY are present across the pilot sites

FD metrics

Functional spaceFunctional evenness: some parts of niche space are under utilised

Functional divergence : little niche differentiation

Functional dispersion: environmental filters determine community assemblages

Functional richness: resources potentially unused

top related