Plant Phenotyping in SPICY. Chris Glasbey Biomathematics and Statistics Scotland (BioSS)

Transcription

1 Plant Phenotyping in SPICY Chris Glasbey Biomathematics and Statistics Scotland (BioSS)

2 Aim of SPICY: Develop tools to predict phenotypic response of a genotype under a range of environmental conditions This requires: Characterisation of a large number of genotypes by Manual measurements Automatic phenotyping tools

3 Measurements in phenotyping experiments Total fresh fruit yield (kg m -2 ) Total dry fruit yield (kg m -2 ) Fruit dry matter content (kg kg -1 ) Total dry matter (kg m -2 ) Fraction to the fruits (kg kg -1 ) LUE (g MJ -1 ) Fraction of light intercepted (MJ m -2 ) Sink strength of fruits No of fruits Photosynthetic rate (g m -2 s -1 ) LAI (m m -2 ) Potential fruit weight (g) Fruit growth duration (d) No of flowers Abortion rate No of internodes

4 Plan of talk 1. Manual phenotyping 2. Phenotyping by fluorescence 3. Phenotyping by image analysis

5 1. Manual phenotyping Anja Dieleman, Wageningen UR Greenhouse Horticulture Juan José Magan, Estación Experimental de la Fundación Cajamar

6 Manual phenotyping Start of the experiment: Destructive harvest Leaf area DW leaves DW stem Stem length before 1 th branching No of internodes before 1 th branching

7 Weekly: No. of fruits on plants FW/DW harvested fruits 2-weekly: No. of internodes 4-weekly Stem length

8 Fruit growth duration Potential fruit weight

9 Final destructive harvest: DW leaves DW stem DW fruits No. of fruits Leaf area Stem length

10 Manual phenotyping Disadvantages: - Slow and expensive - Variation between observers - Sometimes destructive

11 2. Phenotyping by fluorescence Attila Barocsi & Sandor Lenk Budapest University of Technology and Economics

12 Principles of the technique N.R. Baker and E. Rosenqvist, Applications of chlorophyll fluorescence can improve crop production strategies: an examination of future possibilities, J Exp Botany, 55, (2004)

13 Principles of the technique Fluorescence response: to low dose F 0 : minimum, PSII open (dark adapted)

14 Principles of the technique Fluorescence response: to step illumination

15 Principles of the technique Fluorescence response: to transient pulses F m : max, PSII closed (dark adapted), F m : in light

16 Principles of the technique Fluorescence response: to far red excitation F 0 : PSI selectively opens

17 Fluorescence sensor developed: Short measurement cycle Portable 2 operating sensors per system Wireless data acquisition

18 Fq'/Fv' Fv'/Fm' Fq'/Fm' Fv' F0' Fq Fm' Fv/Fm Fv Fm F0 Results: Heritability of estimated parameters H 690nm H 730nm H PAM F Fm Fv Fv/Fm Fm' 60s s s Fq 60s s s F0' 60s s s Fv' 60s s s Fq'/Fm' 60s s s Fv'/Fm' 60s s s Fq'/Fv' 60s s s NPQ 60s High heritabilities at short times faster measurements

19 3. Phenotyping by image analysis Chris Glasbey, Graham Horgan, Yu Song Biomathematics and Statistics Scotland (BioSS) Gerie van der Heijden, Gerrit Polder Biometris, Wageningen UR

20 Phenotyping by Image Analysis Scanalyzer 3D, LemnaTec Sorting Anthurium cuttings, Wageningen UR Most image analysis systems for automatic phenotyping bring the plants to the camera.

21 Commercial tomato plants in Almeria, Spain Pepper plants in our experiments But for some crops, like pepper and tomato, this is not feasible bring the cameras to the plants!

22 SPYSEE equipment

23 SPYSEE 4* IR, Colour, Range (ToF) cameras

24 We aim to : Replace manual by automatic measurements Find new features, which are not possible or too difficult for manual measurement Two approaches: A. 3D B. Statistical

25 A. 3D approach 3D information can be recovered from stereo pairs, because Depth = constant / disparity

26 Stereo pair + ToF range image detailed range image

27 Foreground leaves

28 Leaf in 3D automatic measurement of size, orientation, etc

29 Individual leaf size (cm 2 ) Automatic Manual Validation trial (11 genotypes, 55 leaves): Correlation = 98% RMSE = 9.50cm 2

30 QTL analysis of automatically measured leaf sizes for 151 genotypes Leaf size had a heritability of 0.70, three QTLs were found, together explaining 29% of the variation.

31 Leaf orientation: Angle between the leaf and the vertical axis.

32 Leaf orientation: Angle between the leaf and the vertical axis.

33 QTL analysis of automatically measured leaf orientation for 151 genotypes Heritability was 0.56, and one QTL explained 11% of the total variation

34 B. Statistical approach Plant height estimated, from locations of green pixels

35 Correlation 93% between automatic and manual plant heights

36 Total leaf area is a measure of how much solar radiation the plant can intercept

37 Frequency Colour distribution Counts how many pixels in the image have each red intensity Intensity

38 Frequency Colour distribution Intensity

39 Frequency Colour histograms Another example Intensity

40 Principal component regression Call: lm(formula = sep.leafarea ~ pr1$x[, 1:6], na.action = na.exclude) Number crunching to link colour histograms to manually measured total leaf area Coefficients: Estimate Std. Error t value Pr(> t ) (Intercept) 4.899e e+01 Complex but standard <2e-16 methodology *** PC e e <2e-16 *** PC e e <2e-16 *** PC e e <2e-16 *** PC e e PC e e PC e e * --- Residual standard error: on 209 degrees of freedom (1334 observations deleted due to missingness) Multiple R-squared: , Adjusted R-squared: F-statistic: on 6 and 209 DF, p-value: < 2.2e-16

41 Predicted projecte leaf area Prediction vs manual Correlation 80% Total leaf area

42 Coefficient Regression coefficients Weight of each colour intensity count in predicting the leaf area index Intensity

43 Multivariate histograms Count the number of times each combination of the three colour components occurs Too many possibilities, so use bins of length 8 per component, leading to 16 3 = 4096 variables Again do Principal Components regression

44 Predicted projecte leaf area Multivariate histograms Correlation 83% Total leaf area

45 QTL analysis of automatically measured total leaf area for 151 genotypes The heritability of total leaf area was 0.55, and 20% of the variation was explained by QTLs 2 QTLs agree with 2 of 3 found from manual measurements

46 Work in progress: Automatically find fruits Measure plant development Image Fruit Probability

47 31 Aug 2 Sep 5 Sep 8 Sep 9 Sep

48 Summary Manual phenotyping is slow and expensive Two approaches to automatic phenotyping have been explored in SPICY: 1. Fluorescence 2. Image analysis Results are promising but more work is needed