Importance of Sampling Frequency in Reducing Uncertainty

Louise Barton1, Benjamin Wolf2, David Rowlings3, Clemens Scheer3, Ralf Kiese2, Peter Grace3, Katia Stefanova1, and Klaus Butterbach-Bahl2,4

1The University of Western Australia2Institute for Meteorology and Climate Research, Germany 3Queensland University of Technology, Australia 4International Livestock Research Institute, Kenya

Presenter’s travel costs were funded by OECD-Cooperative Research Program.

The various research projects have funded by the

• Australian Government’s Climate Change Research Program,

• Australian Research Council (DP0559791)

• Department of Agriculture & Food Western Australia.

• Grains & Research Development Corporation

• German Science Foundation

ACKNOWLEDGEMENTS

Jun-09 Aug-09 Oct-09 Dec-09 Feb-10 Apr-10 Jun-10

Dai

ly fl

ux (g

N h

a-1)

0

20

40

60

80

100

Source: Scheer et. al. 2013. Nutrient Cycling in Agroecosystems 95: 43–56.

Irrigated cropping, subtropical climate, Australia

NITROUS OXIDE FLUXES AND TEMPORAL VARIABILITY

“How often do I need to sample using manual chambers?”

Rain-fed cropping, Wongan Hills, south-western Australia

Investigate the effect of sample frequency on estimates of annual N2O fluxes by using data collected:

• On a sub-daily basis using automated chamber systems

• From a variety of climates and land-uses

OVERALL OBJECTIVE

Semi-arid

Sub-tropical

Tropical

LOCATION DATASETS

(Yrs,Treat)

CLIMATE LANDUSE ANNUAL N2O FLUX

(kg N ha-1)

Bellenden Kerr, Australia

One (1, 1) Tropical Forest 1.16

Cunderdin, Australia

Eight (4, 2) Semi-arid Cereal cropping

0.08–0.16

Höglwald, Germany

Two (1, 1) Temperate Plantation forest

0.58–2.46

Kingsthorpe, Australia

Three (1, 3) Sub-tropical Irrigated cereal-cottoncropping

2.61–2.93

Mooloolah Valley, Australia

Five (3, 2) Sub-tropical Forest, Pasture,Orchard

0.48–8.12

Wongan Hills, Australia

Eight (2, 4) Semi-arid Cereal cropping

0.03–0.07

Xilin, Inner Mongolia

One (1) Semi-arid Steppegrassland

0.21

TWENTY EIGHT DATASETS: RANGE IN ANNUAL FLUXES

LOCATION DATASETS

(Yrs,Treat)

CLIMATE LANDUSE ANNUAL N2O FLUX

(kg N ha-1)

DAILY FLUX CV(%)

EPISODICITY

Bellenden Kerr, Australia

One (1, 1) Tropical Forest 1.16 98 Moderate

Cunderdin, Australia

Eight (4, 2) Semi-arid Cereal cropping

0.08–0.16 173–428 High to Extreme

Höglwald, Germany

Two (1, 1) Temperate Plantation forest

0.58–2.46 169–179 High

Kingsthorpe, Australia

Three (1, 3) Sub-tropical Irrigated cereal-cottoncropping

2.61–2.93 181–235 High to Extreme

Mooloolah Valley, Australia

Five (3, 2) Sub-tropical Forest, Pasture,Orchard

0.48–8.12 78–172 Moderate to High

Wongan Hills, Australia

Eight (2, 4) Semi-arid Cereal cropping

0.03–0.07 380–913 Extreme

Xilin, Inner Mongolia

One (1) Semi-arid Steppegrassland

0.21 260 Extreme

TWENTY EIGHT DATASETS: RANGE IN ‘EPISODICITY’

CV, coefficient of variation

APPROACH

Daily fluxes by averaging sub-daily fluxes (removed diurnal variation)

Annual fluxes calculated for different sampling frequencies (5 intervals) using ‘Jack-Knife’ analysis

For each sampling frequency, annual flux compared with ‘best’ estimate (daily) Result expressed as % of ‘daily’ annual flux

For each data set, we calculated:

Frequency (day) Permutations Example

Daily (0) 1 All sample days

3-days/week (2) 7 Sun-Tue-Thu; Mon-Wed-Fri; Tue-Thu-Sat etc

Weekly (7) 7 Sun, Mon, Tue, Wed, Thu, Fri, Sat

Every 2nd week (14) 14 Sun (Week 1, 2), Mon (Week 1, 2) etc

Every 4th week (28) 28 Sun (Weeks 1–4), Mon (Week 1–4) etc

Jun-09 Aug-09 Oct-09 Dec-09 Feb-10 Apr-10 Jun-10

Dai

ly fl

ux (g

N h

a-1)

0

5

10

15

20 Plus limeNo lime

12%

9%

8%

Source: Barton et. al. 2013. Agriculture, Ecosystems and Environment 167: 23–32.

Cropping, semi-arid climate, Australia

Annual flux: 0.05–0.06 kg N ha-1

DAILY NITROUS OXIDE FLUX PROFILE: ‘Extremely’ episodic

Sampling interval (days)0 5 10 15 20 25 30

% D

evia

tion

of a

nnua

l flu

x

0

200

400

600

800

1000 No limePlus lime

Sampling interval (days)0 5 10 15 20 25 30

% D

evia

tion

of a

nnua

l flu

x

0

200

400

600

800

1000 No limePlus lime

12%

9%

SAMPLING FREQUENCY & ANNUAL FLUX : ‘Extremely’ episodic

Cropping, semi-arid climate, Australia %

of ‘

daily

’ ann

ual f

lux

Sampling interval (days)0 5 10 15 20 25 30

% D

evia

tion

of a

nnua

l flu

x

0

200

400

600

800

1000 No limePlus lime

12%

9%

8%

0 (Daily) 2 7 14 28% Best estimate 100 61–141 21–251 10–464 3–893

Annual Flux(kg N ha-1) 0.05–0.06 0.03–0.08 0.01–0.14 0.01–0.5 0–0.5

SAMPLING FREQUENCY & ANNUAL FLUX : ‘Extremely’ episodic

Sampling interval (days)0 5 10 15 20 25 30

% D

evia

tion

of a

nnua

l flu

x

0

200

400

600

800

1000 No limePlus lime

Cropping, semi-arid climate, Australia %

of ‘

daily

’ ann

ual f

lux

01-Jan 01-Mar 01-May 01-Jul 01-Sep 01-Nov 01-Jan

Dai

ly fl

ux (g

N h

a-1)

-20

0

20

40

60

80

100

12%

9%

8%

Source: Butterbach-Bahl et. al. 2002. Plant and Soil 240: 117–123.

Plantation forest, temperate climate, Germany

Annual flux: 2.46 kg N ha-1

DAILY NITROUS OXIDE FLUX PROFILE: ‘Highly’ episodic

9%

Sampling interval (days)0 5 10 15 20 25 30

% D

evia

tion

of a

nnua

l flu

x

40

60

80

100

120

140

160

180

SAMPLING FREQUENCY & ANNUAL FLUX : ‘Highly’ episodic

Plantation forest, temperate climate, Germany%

of ‘

daily

’ ann

ual f

lux

0 (Daily) 2 7 14 28% Best estimate 100 97–105 88–118 78–120 60–156

Annual Flux(kg N ha-1) 2.46 2.40–2.58 2.16–2.90 1.93–2.97 1.49–3.85

01-Apr 01-Jun 01-Aug 01-Oct 01-Dec 01-Feb 01-Apr

Daily

flux

(g N

ha-1

)

02468

1012141618

12%

9%

8%

Source: Kiese et. al. 2003. Global Biogeochemical Cycles 17: 1043

Annual flux: 1.2 kg N ha-1

DAILY NITROUS OXIDE FLUX PROFILE: ‘Moderately’ episodic

Rainforest, tropical climate, Australia

Sampling interval (days)0 5 10 15 20 25 30

% D

evia

tion

of a

nnua

l flu

x

50

60

70

80

90

100

110

120

130

9%

8%

SAMPLING FREQUENCY & ANNUAL FLUX: ‘Moderately’ episodic

Rainforest, tropical climate, Australia

0 (Daily) 2 7 14 28% Best estimate 100 100–102 98–103 72–107 57–123

Annual Flux(kg N ha-1) 1.2 1.1–1.2 1.1–1.2 0.8–1.2 0.7–1.4

% o

f ‘da

ily’ a

nnua

l flu

x

8%

RECOMMENDED SAMPLING FREQUENCY

Sampling interval (d)0 7 14 21 28

Num

ber o

f dat

aset

s

0

5

10

15

20

25Within 10% Within 20% Within 30% Within 40% Within 50%

Source: Barton et. al. 2015. Sci. Rep. 5:15912 | DOI: 10.1038/srep15912

• 71% of data sets = daily measurements• 25% of data sets = 3 daily measurements per week

8%

Source: Barton et. al. 2015. Sci. Rep. 5:15912 | DOI: 10.1038/srep15912

RECOMMENDED SAMPLING FREQUENCY

Sampling interval (d)0 7 14 21 28

Num

ber o

f dat

aset

s

0

5

10

15

20

25Within 10% Within 20% Within 30% Within 40% Within 50%

Corn-soybean, Iowa, USA“At relatively frequent sampling intensities (i.e., once every 3d) N2O–N flux estimates were within ±10% of the expected value”

Parkin (2008)Cotton, Shanxi Province, China“Sampling daily (at 9am) to every two days caused a deviation of up to 7.3% from annual flux estimated from sub-daily measurements.”

Lui et al. (2010)Pasture, Otago, New Zealand“… gas samples collected three times a week between 10:00-12:00h provided zero bias in calculating cumulative emissions when compared with those based on frequent, 2-hourly, flux measurements.”

Van der Weeden et al. 2013Sugar Cane, Queensland, Australia“Weekly sampling with biweekly sampling for one week after >20 mm of rainfall was the recommended sampling regime. It resulted in no extreme (>20%) deviations from the ‘actuals’, had a high probability of estimating the annual cumulative emissions within 10% precision …”

Reeves et al. (2016)

OTHER STUDIES

8%

RECOMMENDED SAMPLING STRATEGY:‘Jackknife’ and ‘Informed’

LOCATION DATASETS

(Yrs,Treat)

CLIMATE LANDUSE ANNUAL N2O FLUX(kg N ha-1)

EPISODICITY JACKNIFE(sampling days)

INFORMED(sampling days)

Bellenden Kerr, Australia

One (1, 1) Tropical Forest 1.16 Moderate 52 156

Mooloolah Valley, Australia

Five (3, 2) Sub-tropical Forest, Pasture,Orchard

0.48–8.12 Moderate to High

156–365 Not determined

Höglwald, Germany

Two (1, 1) Temperate Plantation forest

0.58–2.46 High 156 83

Xilin, Inner Mongolia

One (1) Semi-arid Steppegrassland

0.21 Extreme 156 Not determined

Cunderdin, Australia

Eight (4, 2) Semi-arid Cereal cropping

0.08–0.16 High to Extreme

156–365 Not determined

Kingsthorpe, Australia

Three (1, 3) Sub-tropical Irrigated cereal-cottoncropping

2.61–2.93 High to Extreme

365 Not determined

Wongan Hills, Australia

Eight (2, 4) Semi-arid Cereal cropping

0.03–0.07 Extreme 365 60

æ Nitrous oxide emissions generally need to be measured daily to accurately estimate (within 10%) annual fluxes in a variety of land-uses and climates• 71% of data sets = daily measurements• 25% of data sets = 3 daily measurements per week

æ For study sites where temporal variability has not previously been characterised (or when variability is expected to change from year-to-year) recommend daily sampling with manual chambers

æ Recommended sampling frequency increases with increasing “episodicity”

æ Data from automated chambers should be continuously used to develop guidelines for manual chamber sampling frequency

CONCLUSIONS

0 5 10 15 20 25 30

% D

evia

tion

of E

F

-500

0

500

1000 Year 1Year 2 Year 3 12%

9%

Emission Factor: 0.01-0.06%

SAMPLING FREQUENCY & EMISSION FACTOR

Cropping, semi-arid climate, Australia

12%

9%

Relationship between coefficient of variation of the daily N2O flux and the deviation (range) from the ‘best estimate’ annual N2O flux: 4-weekly sampling interval