sap flow data analysis presentation - ict international

13/01/2014 1

Solutions for soil, plant & environmental monitoring

www.ictinternational.com

Sap Flow

Data Analysis

&

Presentation:

How to analyze sap flow data

Michael Forster

ICT International

13/01/2014 2



Tem

pera

ture

(°C

)

0

10

20

30

40

Rela

tive H

um

idity

(%)

20

40

60

80

100

E. cladocalyx

Time vs Relative Humidity

VP

D (

kP

a)

0

1

2

3

4

5

Rain

fall

(mm

)

0

20

40

60

80

100

J (

cm

3 c

m-2

day

-1)

0

20

40

60

80

100

J (

cm

3 c

m-2

day

-1)

0

20

40

60

80

100

2009 2010 2011

Dec Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Mar

J (

cm

3 c

m-2

day

-1)

0

20

40

60

80

100

(A)

(B)

(C)

(D) – E. cladocalyx

(E) – E. melliodora

(F) – E. polybractea

Summary Data

• Method:

• Collate entire data sets

• Daily averages or some other summary stat

• Typical variables:

• Temperature

• Relative Humidity

• VPD

• Solar Radiation

• Rainfall

• Wind Speed

• Soil Moisture

• Sap Flow

13/01/2014 3



Summary Data

Source: Fig 1. Ambrose et al.,2010

• Interpreting Data:

• Visual summary only

• Gives the reader an quick and

easy overview of conditions

during the study

13/01/2014 4



Representative Data

Source: Figs 3 and 4. Ambrose et al.,2010

13/01/2014 5



Summary Tables: Tree Characteristics

Source: Table 1. Ambrose et al.,2010 Source: Table 1. Pfautsch & Adams 2012

13/01/2014 6



Summary Tables: Descriptive Statistics

Source: Table 2. Ambrose et al.,2010

13/01/2014 7



Summary Tables: Descriptive Statistics

Species Summer ‘10 Autumn ‘10 Winter ‘10 Spring ‘10 Summer ‘11

E. cladocalyx 26.35 (±8.49) 17.02 (±6.59) 11.21 (±4.29) 17.25 (±4.33) 26.59 (±4.18)

E. melliodora 4.63 (±2.53) 2.67 (±1.43) 2.12 (±1.56) 4.59 (±2.52) 9.21 (±4.60)

E. polybractea 7.46 (±7.82) 4.76 (±5.01) 3.62 (±3.52) 8.16 (±7.57) 4.97 (±1.36)

Table 1. Summary of average daily water use (Q, L day-1) throughout the various seasons of the study period.

Values are total tree water use including multiple stems of E. melliodora and E. polybractea. Values are litres of

water (

SD).

13/01/2014 8



Time

Jan Feb Mar Apr

Cum

ula

tive

Q (

L d

ay-1

)

0

500

1000

1500

2000

2500

Summary Figures: Total or Cumulative

Amounts

Source: Fig. 1. Doronila & Forster in press.

E. cladocalyx

E. melliodora

E. polybractea

13/01/2014 9



Summary Figures: Total or Cumulative

Amounts

Ambient Elevated Ambient Elevated

Wet Soils Dry Soils

• Statistical Test:

• ANOVA with a Tukey’s

HSD post-hoc test

Source: Fig. 4. Zeppel et al. 2011

13/01/2014 10



Summary Figures: Treatment Effects

Source: Fig. 4. Ghuran et al., 2013 Source: Fig. 2. Pfautsch & Adams 2012

13/01/2014 11




1:1 Relationship

Measured

Relationship

Source: Unpublished, R. Duursma,

University of Western Sydney

13/01/2014 12




• E = Elevated CO2 Treatment

• A = Ambient CO2 Treatment:

• Horizontal line = 1:1 Relationship

• Interpretation:

• Day-time: Trees growing under ambient

CO2 have higher sap flow in both wet

and dry soils

• Night-time: Trees growing under elevated

CO2 have higher sap flow in wet soils but

lower sap flow in dry soils

Source: Fig. 3. Zeppel et al. 2011

13/01/2014 13



Sap Flow Versus Single Variable, e.g. VPD

Source: Fig. 3. Pfautsch et al. 2011

13/01/2014 14



Linear or Non-Linear Regression

Source: Fig 4.

Pfautsch & Adams, 2012

Simple, easy to use and interpret

Not rigorous!

13/01/2014 15




• Method:

• Data from non-rain days, or all data

• Allocate data to logical categories (e.g. season,

pre- and post-treatment)

• Sap flow on y-axis, variable on x-axis

• To lessen variance, average data, e.g. if

measuring at 15 min intervals, use hourly or 2-

hourly averages


• Linear Regression

• Non-Linear Regression (logarithmic)

• Use whichever gives highest R2 value

13/01/2014 16




• Interpretation:

• A significant linear relationship means no, or

very little, stomatal closure to variable

• A significant non-linear relationship means

stomatal closure to variable

• No relationship means that variable is not

influencing sap flow

13/01/2014 17




• Example References:

• Many studies use this technique

• Pfautsch & Adams 2012. Oecologia

• Rosado et al., 2012. Agric. For. Meteor

13/01/2014 18



How Much Data to Use?

VPD (kPa)

0 1 2 3 4 5J (

cm

3 c

m-2

day

-1)

-20

0

20

40

60

80

100

Summary variable or averages All data points

13/01/2014 19



How Much Data to Use?

Summary variable or averages All data points

• You’ve collected the data so you

should use it??

• Assumptions on which data to

average or summarise may not be

logically or biologically valid

• Usually much greater variability in the

dataset

• “Cleaner”

• Easier to visualise

• Less variability

• Therefore can achieve a higher R2

13/01/2014 20



Summarising or Averaging Variables

• Hourly averages in 6-hour windows:

• e.g. Pfautsch et al. (2011).

• Windows = 10am – 4pm; and 12am – 6am

• Data measured every 10 minutes then averaged into 30 minute bins:

• e.g. Forster (2012).

• Data measured every 30 minutes then averaged into 2 hour bins:

• e.g. Duursma et al. (2011).

• Sum of diurnal data versus some maximum measurement.

• e.g. Pfautsch & Adams (2012): Sum of nightly sap flow versus nightly VPDmax

13/01/2014 21



Sap Flow Versus Many Variable

Solar Radiation (W/m2)

0 200 400 600 800

JS (

cm

3/h

r)

0

40

80

120

160

200

Solar Radiation

VPD (kPa)

0.0 0.5 1.0 1.5 2.0 2.5

JS (

cm

3/h

r)

0

40

80

120

160

200

VPD

Soil Water Content (m3/m

3)

0.090 0.095 0.100 0.105 0.110 0.115 0.120

JS (

cm

3/h

r)

0

40

80

120

160

200

Soil Water Content

Soil Water Potential (kPa)

-300 -250 -200 -150 -100 -50 0

JS (

cm

3/h

r)

0

40

80

120

160

200

Soil Water Potential

13/01/2014 22



Partial Regression

Source: Table 1, Forster 2012

Relatively simple and easy to interpret

Shows the response of one predictor while controlling for other

related predictors

13/01/2014 23



Partial Regression

• Method:

• Data from non-rain days, or all data

• Sap flow is the dependent variable, there can be

multiple independent variables but only choose

variables which are meaningful

• To lessen variance, average data, e.g. if

measuring at 15 min intervals, use hourly or 2-

hourly averages

13/01/2014 24



Partial Regression

• Method (cont.):

• A statistical package, e.g. SPSS, is needed:

• ANALYSE > REGRESSION > LINEAR - then

insert your dependent and independent

variables and make sure that "Method:" is set to

"Enter". Then click on STATISTICS > make

sure that the "part and partial correlations

box is ticked" > press continue. Now click

PLOTS > tick the "produce partial plots" box >

CONTINUE. Now click OK and the model will

run.

13/01/2014 25



Partial Regression

• Method (cont.):

• Find the results in the Partial Correlation table

(in SPSS this is in the coefficients table)

• Partial R2 = the square of the partial correlation

multiplied by 100

• VPD Partial corr. = 0.682

• Partial R2 = (0.682 * 0.682) * 100 = 46.518%

13/01/2014 26



Partial Regression Plot Source: Adapted from

Table 1, Forster 2012

Tree Sap Flow

-100 -80 -60 -40 -20 0 20 40 60 80 100

Gal

l S

ap F

low

-3

-2

-1

0

1

2

3

Solar Radiation

-400 -200 0 200 400 600

Gal

l S

ap F

low

-3

-2

-1

0

1

2

3

Temperature

-10 -8 -6 -4 -2 0 2 4 6 8

Gal

l S

ap F

low

-3

-2

-1

0

1

2

3

VPD

-0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0

Gal

l S

ap F

low

-3

-2

-1

0

1

2

3

13/01/2014 27



Partial Regression

• Interpretation:

• The variance explained is a clear quantitative

indicator of which variable is most important

while taking into account all other variables

• A univariate linear regression analysis may find

a significant relationship between 2 variables

(e.g. solar radiation and VPD), but a multivariate

linear regression analysis will tell you which is

more important

• The partial slope or correlation indicates

whether the relationship is positive or negative

13/01/2014 28



Partial Regression


• Taylor & Eamus, 2008. Tree Phys.

• Forster, 2012. Fungal Ecology.

13/01/2014 29



Generalized Additive Model

Source: Fig 3.

Duursma et al. 2011

13/01/2014 30



• Method:

• Type of regression analysis of sap flow (y-axis)

against an environmental variable, e.g. VPD or

solar radiation, (x-axis)

• See Wood (2006) for more details

• Dendrometers (DBL60 Band Dendrometer)


• Smoothed regression with 95% C.I.


• Duursma et al., 2011, Tree Phys.

Generalized Additive Model

13/01/2014 31



Source: Fig 3.

Forster 2012

Normalising Data

• Sap flow data can be of

different magnitudes

• Typical comparison: roots

versus trunk; small versus

large tree

• Display data on different axes

or normalise data

13/01/2014 32



Normalising Data

Source: Fig 4.

Eller et al. 2013

• Method:

• Sap velocity data

• Find the maximum value (careful with

erroneous data, including spikes or noise)

• Calculate percentage based on this

maximum value


• Visual Inspection

• Repeated Measures ANOVA

13/01/2014 33



Normalising Data

Source: Fig 4.

Eller et al. 2013

• Interpretation:

• If direction and magnitude are the same

then the hydraulic behaviour or

architecture is the same

• Example Reference:

• Eller et al., (2013) New Phytologist

13/01/2014 34


www.ictinternational.com From: Caldwell, et al. 1998

Only HRM and HFD can measure reverse sap flow

Hydraulic Redistribution – Reverse Flow

13/01/2014 35



Hydraulic Redistribution

Figure from Oliveira et al..2005

Typically, visual inspection of data only

13/01/2014 36




Source: Fig. 6. Oliveira et al. (2005)

Typically, visual inspection of data only – average of night time values

Flow to

Trunk

Flow to

Lower

Soil

Dry Season Wet Season

13/01/2014 37



Hydraulic Redistribution & Schematics

A complicated story…

Distal Sensor on Lateral Root

Proximal Sensor on Main Root

Source: Fig. 7. Bleby et al. (2010)

Distal Sensor on Lateral Root

Proximal Sensor on Main Root

13/01/2014 38




… can be simplified with a schematic

13/01/2014 39



Source: Fig. 2 Nadezhdina et al. (2009)


13/01/2014 40



Source: Fig. 2 Nadezhdina et al. (2009)


13/01/2014 41



Source: Figs. 5, Nadezhdina et al. (2009)

Hydraulic Redistribution & Xylem Depth

Northern Root = Black Line

Southern Root = Grey Line

Typically, average data are presented…

13/01/2014 42




… but with Sap Flow Tool, you can show hydraulic redistribution pattern over

the entire xylem depth (radial profile)

Day 1 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Day 9 Day 10 Day 11 Day 12

Irrigation

Event

Bark

Heartwood

Green =

Reverse

Flow

13/01/2014 43




Sap Flow Tool Demonstration

13/01/2014 44




• Notes:

• You can use either heat velocity, sap velocity or

sap flow data

• CRITICAL: YOU MUST ENSURE YOU HAVE

MEASURED ZERO FLOW ACCURATELY

13/01/2014 45



Circumferential Data – 2D Figure

Source: Fig 5. Cermak et al. 2004

13/01/2014 46




4 sensors installed at

cardinal points

Divide tree into 4

segments

Find maximum for

each segment

North

East

South

West

13/01/2014 47




Source: Fig 5. Cermak et al. 2004

13/01/2014 48



Fre

qu

ency

of

J S P

eak

Tim

e (%

)

0

4

8

12

16

20

24

28

Hours

8 10 12 14 16 18 20

0

4

8

12

16

20

24

28

0

4

8

12

16

20

24

28 E. cladocalyx

E. melliodora

E. polybractea

JS Peak Time

• Method:

• Data from non-rain days

• Usually summer data

• You can focus on a single month

• Sort data into hour or half-hourly bins

• Determine a frequency for each bin


• Comparing single bin: ANOVA

• Comparing all data: Repeated Measures

ANOVA

Source: Fig 4. Doronila & Forster in press

13/01/2014 49



JS Peak Time Demonstration

JS Peak Time

13/01/2014 50



Fre

qu

ency

of

J S P

eak

Tim

e (%

)

0

4

8

12

16

20

24

28

Hours

8 10 12 14 16 18 20

0

4

8

12

16

20

24

28

0

4

8

12

16

20

24

28 E. cladocalyx

E. melliodora

E. polybractea

JS Peak Time

• Interpretation:

• Uni-modal data means there is an optimal time

for stomatal opening

• Multi-modal data means stomata are sensitive

to VPD

• Early peak means stomata close early in day.

Does early stomatal closure leads to less CO2?

Source: Fig 4. Doronila & Forster in press

13/01/2014 51



JS Peak Time

• Related Measurements:

• Stomatal conductance (SC-1 Leaf Porometer)

• Photosynthetic Rate (CI-340 from CID)

• Dendrometers (DBL60 Band Dendrometer)


• Doronila & Forster, in press, Int. J. Phyt. Rem.

• Du et al. 2011. Agric. For. Meteor.

Source: Fig 5.Du et al. 2011

13/01/2014 52



Time of JS,max

08:00:00 10:00:00 12:00:00 14:00:00 16:00:00 18:00:00

VP

Dm

ax (

kP

a)

0

1

2

3

4

5

6

7

8

JS Peak Time and VPDmax

E. cladocalyx

E. melliodora

E. polybractea

Source: Fig 3.

Doronila & Forster in press

13/01/2014 53




Time of JS,max

08:00:00 10:00:00 12:00:00 14:00:00 16:00:00 18:00:00

VP

Dm

ax (

kP

a)

0

1

2

3

4

5

6

7

8 • Method:

• Data from non-rain days

• Usually summer data

• You can focus on a single month

• VPDmax = maximum VPD for a diurnal period

• JS,max = peak sap flow for same diurnal period


• Linear or Non-Linear Regression

13/01/2014 54



Time of JS,max

08:00:00 10:00:00 12:00:00 14:00:00 16:00:00 18:00:00

VP

Dm

ax (

kP

a)

0

1

2

3

4

5

6

7

8



• High VPD day

• Mid-summer of Heat-wave

• Early closure of stomata

• Peak JS early in the day

13/01/2014 55



Time of JS,max

08:00:00 10:00:00 12:00:00 14:00:00 16:00:00 18:00:00

VP

Dm

ax (

kP

a)

0

1

2

3

4

5

6

7

8



• High VPD day

• Mid-summer of Heat-wave

• Late closure of stomata

• Peak JS late in the day

• Unlikely scenario

13/01/2014 56



Time of JS,max

08:00:00 10:00:00 12:00:00 14:00:00 16:00:00 18:00:00

VP

Dm

ax (

kP

a)

0

1

2

3

4

5

6

7

8



• “Normal” VPD day

• Typical, sunny day

• Little to no stomatal closure

• Peak JS around midday

13/01/2014 57



Time of JS,max

08:00:00 10:00:00 12:00:00 14:00:00 16:00:00 18:00:00

VP

Dm

ax (

kP

a)

0

1

2

3

4

5

6

7

8



• Low VPD day

• Cool, cloudy day

• Late opening of stomata

• Peak JS late in the day

13/01/2014 58




Time of JS,max

08:00:00 10:00:00 12:00:00 14:00:00 16:00:00 18:00:00

VP

Dm

ax (

kP

a)

0

1

2

3

4

5

6

7

8 • Related Measurements:

• PSY1 Stem Psychrometer

• SC-1 Leaf Porometer

• SMM Soil Moisture Meter

• Example Reference:

• Doronila & Forster, in press, Int. J. Phyt. Rem.

13/01/2014 59



JS and Optimal VPD or Temperature

10

20

30

40

50

AB A AB CD CD BC CD D CD

Tem

pera

ture

(

C)

EC2 EC5 EC7 EM5 EM7 EM9 EP3 EP5 EP8

0

1

2

3

4

5

6

7

AB A ABC DE DE BCD DE E CDE

VP

D (

kP

a)


(a)

(b)

• Reference:

• Doronila & Forster, in press, Int. J. Phyt Rem.

• Definition:

• The optimal VPD or temperature for the

maximum rate of sap velocity

• At what value of VPD or temperature is a plant

the most happiest

13/01/2014 60




10

20

30

40

50


Tem

pera

ture

(

C)


0

1

2

3

4

5

6

7


VP

D (

kP

a)


(a)

(b)

• E. cladocalyx:

• VPD: 2.6 kPa

• Temperature: 26.2

C

• E. melliodora:

• VPD: 2.1 kPa


C

• E. polybractea:

• VPD: 2.0 kPa


C

13/01/2014 61




10

20

30

40

50


Tem

pera

ture

(

C)


0

1

2

3

4

5

6

7


VP

D (

kP

a)


(a)

(b)

• Method:

• Data collection should be done over an

extensive time period in order to capture

varying VPD and temperature

• Other environmental variables, particularly soil

moisture, are assumed to be optimal

• Take JS data and sort from highest to lowest

• Remove the lowest 95% of values

• Keep highest 5% of values for analysis

• Note: 5% is an arbitrary value

13/01/2014 62



Data Manipulation Demonstration


13/01/2014 63




10

20

30

40

50

AB A AB CD CD BC CD D CDT

em

pera

ture

(

C)


0

1

2

3

4

5

6

7


VP

D (

kP

a)


(a)

(b)• Statistical Test: One-Way ANOVA with a Tukey’s HSD post-hoc test

13/01/2014 64



JS, VPD and Hysteresis

Source: Fig 3.

Pfautsch & Adams 2012

13/01/2014 65



Hysteresis Definition

• The lagging of an effect behind its cause

• “Wetting” and “Drying” curves differ

• When relating one variable to another, you

must declare whether you are on the

wetting or drying curve

Source: http://jan.ucc.nau.edu/~doetqp-p/courses/env302/lec18/LEC18.html

13/01/2014 66


www.ictinternational.com VPD

0.0 0.5 1.0 1.5 2.0 2.5 3.0

JS (

cm

3/h

r)

0

5

10

15

20

25

30

8am

9am

10am

11am12pm1pm

2pm

3pm4pm

5pm

6pm

7pm8pm

Hysteresis Definition

13/01/2014 67



• Method:

• Data from selective days, e.g. prior, during and

post-treament

• VPD on x-axis

• JS = on y-axis


• Usually visual inspection

• Curve fitting procedure?


VPD

0.0 0.5 1.0 1.5 2.0 2.5 3.0

JS (

cm

3/h

r)

0

5

10

15

20

25

30

8am

9am

10am

11am12pm1pm

2pm

3pm4pm

5pm

6pm

7pm8pm

13/01/2014 68




• Different shape curves for

different treatments


VPD

0.0 0.5 1.0 1.5 2.0 2.5 3.0

JS (

cm

3/h

r)

0

5

10

15

20

25

30

IRRIGATED

DROUGHT

13/01/2014 69



Source: Fig 3.

Pfautsch & Adams 2012


Early

Summer

Mid

Summer Heatwave Recovery

13/01/2014 70



Nocturnal Sap Flow

• What is the definition of “night”?

• If you have access to a solar radiation sensor, night is defined as values >1

or > 2 W/m2

• Arbitrary designation of night – e.g. Midnight to 6am (Pfautsch et al. 2011)

• Arbitrary designation of night – 1 hour post-sunset to 1 hour prior-sunrise

where sunset and sunrise times are collated from some official website or

database

13/01/2014 71



Nocturnal Sap Flow

- Nocturnal sap velocity as a percentage of

maximum day-time rate (e.g. Rosado et al.,

2012)

• Corrects for differences in plant or stem size

• Calculations can be based on heat velocity or

sap velocity data

• Over what time period are measurements

taken? e.g. is maximum day-time rate

calculated over a 12 month, 1 season, 1 week

period??

• Is it biologically or physiologically meaningful?

Source: Fig 3.

Rosado et al. 2012

13/01/2014 72



Nocturnal Sap Flow

- Nocturnal sap velocity as a percentage of dry,

day-time summer rates measured at noon (e.g.

Dawson et al., 2007)

• Corrects for differences in plant or stem size

• Calculations can be based on heat velocity or

sap velocity data

• Assuming this period is the highest velocity the

plant will exhibit

• Is it biologically or physiologically meaningful?

Source: Fig 1.

Dawson et al. 2007

13/01/2014 73



Nocturnal Sap Flow

- Nocturnal sap flow as a proportion of diurnal sap flow (e.g.

Pfautsch et al., 2011)

• i.e. ΣQnight divided by ΣQday

• No correction for differences in plant or stem size

• Calculations must be based on corrected, sap flow data

• No consensus on how many days/nights need to be sampled

• Integrating the entire day-time and night-time period

13/01/2014 74



Nocturnal Sap Flow

- Nocturnal sap flow as a proportion of total daily sap flow (e.g.

Forster, submitted)

• i.e. ΣQnight divided by ΣQtotal

• No correction for differences in plant or stem size

• Calculations must be based on corrected, sap flow data

• No consensus on how many days/nights need to be sampled

• Integrating entire night-time with total diurnal period

13/01/2014 75





ICT International Pty Ltd Solutions for soil, plant and environmental monitoring


[email protected]

Phone: 61 2 6772 6770

Fax: 61 2 6772 7616

PO Box 503, Armidale, NSW, Australia, 2350

INTERNATIONAL

http://www.ictinternational.com/

mailto:[email protected]



sap flow data analysis presentation - ict international

Documents