xavier draye, université catholique de louvain, belgium

EPSO: The European Plant Science OrganisationEPSO Workshop on Plant PhenotypingNovember 02-03, 2009Forschungszentrum Jülich, Germany

Forschungszentrum Jülich, GermanyICG-3: PhytosphereJülich Plant Phenotyping Centre (JPPC)Website: http://www.jppc.de

http://www.plantphenomics.com/phenotyping2009

Xavier Draye, Université Catholique de Louvain, Belgium

Model-assisted phenotyping of root system construction and function

Model-assisted phenotyping of root system architecture and function

Xavier DrayeGuillaume LobetMathieu Javaux

Crop Physiology and Plant BreedingSoil and Water ResourcesUniversité catholique de Louvain, Belgium

Juelich, 2 november 2008

Outline

Part 1. Phenotyping of root system architecture

Trends: throughput, dynamic, modeling

Part 2. Functional phenotyping (water capture)

Getting plant biology and soil hydrodynamics together

Part 1. RSA phenotyping platforms

Evolution in the level of phenotyping details

-2000:Root mass, deep root mass, root volume

2000-today:Number and length of root axesLateral root length and densitySurrogates for growth rate

Root axes: LAUZLateral roots: conical aperture

Root anglesDirect growth rate measurements

future:3D entering the sceneMore dynamics

Current demand: capturing genetic variability with high throughput

Which level of detail?

Maximum leaf length (mm) Maximum axis length (mm) Axis diameter (mm)

Number of tillers Number of axes LAUZ (mm)

Leaf dry weight (g) Total length of axes (mm) LR density (mm-1)

HT [#10, 6-19%]NT [#17, 4-43%]SDM [#16, 4-24%]

MAxL [#8, 7-15%]NAx [#12, 6-37%]TAxL [#11, 6-19%]

AxØ [#1, 15%]LAUZ [#3, 10-22%]LaL [#2, 14-17%]

LaØ [#1, 12%]LaDens [#1, 10%]LaLR [#2, 13-29%]

Detailed phenotyping is worth the effort

0

5

10

15

20

25

0 50 100 150 2000

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

LengthGrowth

Dreaming of throughput and dynamics (Arabidopsis… again)

X. Draye & G. Beemster, unpublished

Root imaging in rhizotrons

Soil-grown plants, transparent tubes

High resolution images

Length & area → diameter

(e.g. Lemnatec©, Traitmill©)

↓

↓

© Lemnatec

New in the field: statistical modeling approach

Flat filter-paper substrate

Root scan (every 2 days)

Length-diameter histograms (WinRHIZO)

Statistical model

Growth parameters estimates

↓

↓

↓

↓

Another one: explicit architectural modeling

RootTyp (PlantSoil 258:103)

Another one: explicit architectural modeling

Strategy:

1. Time-dense phenotyping2. Target a small set of

parameters (distribution)3. Simulate realistic root systems (many)4. Calculate… calculate… calculate

- Age-specific information- Hydraulic architecture- …

Towards automated phenotyping

BenefitsHigh resolutionTime-lapseRoot-background separation

IssuesRoot joiningOverlap in dense regionsRegistration not possible

StrategyFocus on few rootsSkeletonize → Vector-based[SmartRoot software]

ImprovementsBetter separation from the topInteractive image mining

Part 2. Functional phenotyping

Water demand

Distribution ofroot uptake activity

Distribution ofsoil resources

0

50

100

150

200

250

0 1 2 3

RSA

Working hypotheses in root water capture

Spatial 1D approach

Root length density (RLD) and root depthBasis for irrigation scheduling, crop growth models

Typical working hypotheses (observations):

Water uptake usually proportional to RLD in wet soilsDeeper roots lead to improved access to water under drought

Motivations for a 3D approach

RSA is closer to underlying biological processes than RLDRSA is the backbone of hydraulic architectureOpportunities to address fine-scale phenomenonsUnderstanding/sorting of key parameters or mechanismsCan be ultimately reduced to lower dimensions

Modeling approach of the soil/root hydraulic architecture

θ(x, y, z, t)ψ(x, y, z, t)Ks(x, y, z, t)

{ node: (x, y, z, d, age, parent) }ψx(node, t)Kr(node, t)Ka(node, t)

ψs

ψs

ψs

ψs

ψs ψx ψx

ψx

ψx

ψx

ψx

Kx

Kx

Kx

Kx

Kx

Kr

Kr Kr

Kr

Kr

Ks

Ks

Ks

Ks

Ks

ψs

ψs

Ks

Hypotheses: Negligible osmotic gradientNegligible capacity

Boundary conditions:Flux or water potential at the collarWater potential of the deep soil

Modeling approach of the soil/root hydraulic architecture

Light transmission experiments

NIR estimation of surface θ

0h 2h 4h 6h 8h

© Lemnatec

Experimental dataset

0

50

100

150

200

250

0 0,5 1 1,5 2 2,5

0

0,1

0,2

0,3

0,4

0,5

0 2 4 6

Transpiration

Soil hydraulics (θ - ψ)3D structure

Time series of soil water content maps

Root hydraulics (Kr, Kx)

Typical output of the model: from ∆θ to sink term

Root properties

Soil structureand properties

Inverse modeling: Estimating unknown parameters

( )ji

n

j

n

ijiji

nnerr

j i

∑∑= =

−= 1 1

2

,, θ̂θ

θ

100 x Errθ [-]

Inverse modeling provides estimates of biological parameters

Thank you for your attention...

QTL analysis : Sophie de Dorlodot, René Civava, Pierre Faux, Jean-François Dumasy,Charlotte de Mey, Bill Thomas, Brian Forster

Aeroponics : Tristan Lavigne, Aurélie babé, Beata Orman, Imaging : Eric Aussems, Olivier MonnartModeling : Loïc PagèsSupporters : Achim Walter, François Tardieu, Tom Beeckman

Gerrit Beemster, Malcolm Bennett,…

Funding Fonds National de la Recherche ScientifiqueUniversité catholique de LouvainCommunauté française de BelgiqueBelgian Scientific Policy