niels schütze and oleksandr mialyk tu dresden, university of … · 5 irrigation strategies x 32...

Niels Schütze and Oleksandr MialykTU Dresden, University of Twente

The value of information for the management ofdeficit irrigation systemsEGU General Assembly 2020, 6. Mai 2020

Key messages• the value of information about irrigation control strategies, future climatedevelopment, soil characteristics, and initial moisture conditions isevaluated for an arid and a semi-arid irrigation site• as a stochastic framework the Deficit Irrigation Toolbox (DIT) is used• value of information =̂ costs for compensation of uncertainty in units ofadditional water requirements• results are:

– costs of climate uncertainty (variability) are higher for the semi-arid site– costs of soil uncertainty (variability) are higher for the arid site, and evenhigher for wet initial conditions– costs of soil uncertainty (variability) can be reduced by proper soil analysis– but highest added value of information have: knowledge about theproper deficit irrigation strategy and information about initial conditionsManagement of deficit irrigation systemsTU Dresden, University of Twente // Niels Schütze and Oleksandr MialykEGU General Assembly 2020, 6. Mai 2020 Folie 2 von 29

OutlineIntroductionControlled deficit irrigationDIT modeling frameworkApplication to different climate regionsResultsImpact of climate and soil variabilityImpact of deficit irrigation strategyConclusions and Outlook

Management of deficit irrigation systemsTU Dresden, University of Twente // Niels Schütze and Oleksandr MialykEGU General Assembly 2020, 6. Mai 2020 Folie 3 von 29

Introduction

Management of deficit irrigation systemsTU Dresden, University of Twente // Niels Schütze and Oleksandr MialykEGU General Assembly 2020, May 6, 2020 Slide 4 of 29

Focus on controlled deficit irrigation (CDI) forimproving water productivityFocus on water productivity

• More crop per drop: Improvement of Water Productivity (WP)– Improvement of crop growth– Efficient and sustainable irrigation– Preservation of farmland through better cultivation practices

• WP = gain over expenses– (Marketable) yield over total irrigation amount or evapotranspiration

• Typical ranges of WPET

Maize: Wheat: Rice: 0.3 – 1.7 kg m-3 0.6 – 1.5 kg m-3 0.4 – 1.4 kg m-3


Theoretical framework of CDI:The crop water production function (CWPF)

18

Focus on water productivitysources of improvements in water productivity –deficit irrigation strategy -

norm

alized

norm

alized


What is unknown?1) future weather development Focus on water productivityconsidering climate variability and food security

19

Variability in yield for rainfed, supplemental and full irrigation for maize (Montpellier, France)If the site is known:optimized simulations of 500 possible growing seasons for different irrigation water amounts available.Management of deficit irrigation systemsTU Dresden, University of Twente // Niels Schütze and Oleksandr MialykEGU General Assembly 2020, May 6, 2020 Slide 8 of 29

What is unknown?2) soil variability and / or 3) initial soil moisture

0.8 0.6 0.4 0.2 0

0.8

0.6

0.4

0.2

0 0.8

0.6

0.4

0.2

0

sand

siltclay

S lS saL

LsiL

Si

L

cL

saCL

siCL

saCsiC

C

sand

siltclay

triangular rangesynthetic samples

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0

0.920.920.92

0.92

0.92

1.13

1.13

1.13

1.13

1.13

1.351.351.35

1.35

1.571.571.57

1.57

1.791.79

1.79

22.222.442.652.872.87

sand

siltclay

soil texture log (Ks)If nothing about the soil is known:sampling of 500 texture realizations and translated saturated conductivity values by the ROSETTA-PTF.


Evaluation of the value of information using:The Deficit irrigation toolbox (DIT) – http://bit.ly/TUD-DIT

Crop models• Cropwat (FAO)• Aquacrop OS∗

(FAO)• Daisy• Apsim• . . .

Irrigation strategies (full and deficit)• no irrigation• full irrigation• open-loop control (fixed)• closed-loop (feedback) control• open-loop control (calendar)• . . .


DIT uncertainty framework:parameters, initial and boundary conditions

Climate variabilityusing a weathergenerator, e.g.LARS-WG• based on climatestations, e.g. NOAA• generatesrealisations ofhistoric and futuregrowing seasons• and meteorologicalforecasts

Soil variability• using a soil texturegenerator

Initial conditions• θ0 = PWP• θ0 = FC• θ0 = PWP . . . θs


DIT modeling framework v1.0Features• probabilistic framework(SCWPF)• parallel computation(works on HPC)• visualization tools forperformance analysis• manual, examples,tutorial

Software development• uses softwareengineering tools

– Unit tests– FusionForge– Code revision control• provided as open source

– http://bit.ly/TUD-DIT


Application to different climate regions


Considered climate regionsArid site: Seeb, Oman Semi-arid site: Montpellier, France

Temperature (100 years)

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan0

10

20

30

40C°

Growing seasonAverage TmaxMeanAverage Tmin

Climatic water balance (100 years)


75

150

225

300

mm Growing season

ET 10% quantileET 50% quantileET 90% quantileP 10% quantileP 50% quantileP 90% quantile

Temperature (100 years)


10

20

30

40

C°

Growing seasonAverage TmaxMeanAverage Tmin

Climatic water balance (100 years)


75

150

225

300

mm

Growing seasonET 10% quantileET 50% quantileET 90% quantileP 10% quantileP 50% quantileP 90% quantile


Experimental setup:32 combinations of varying levels of information in

Climate variability2 levels of informationfrom 2 climate stations:Seeb, Montpellier• climate series of 100generated years• average year (perfectknowledge)

Soil variability2 levels of informationfrom2 texture samplings:soil triangle,USDA class: sandy loam• 100 soil samples• cendroid of eachsample (perfectknowledge)

Initial conditions2 cases of initialcondition:• dry: θ0 = PWP• wet: θ0 = FC


Experimental setup:varying knowledge / information about 5 irrigation strategies

Crop model• Aquacrop OS∗

(FAO)

Irrigation strategies• no irrigation• full irrigation• closed-loop control (decision table)• open-loop control (long term calendar)• open-loop control (perfect knowledge)

• 5 irrigation strategies x 32 scenarios → 160 combinations of informationManagement of deficit irrigation systemsTU Dresden, University of Twente // Niels Schütze and Oleksandr MialykEGU General Assembly 2020, May 6, 2020 Slide 17 of 29

Results


Example plot:costs of uncertainty (additional water requirements) =̂ value of information

0 200 400 600 800 1,0000

2

4

6

8

10

12

14

16

Irrigation volume [mm]

Dry

yield[t/ha]

DI: 5% to 95% envelope FI: ensembleFI: averageDI: 90% quantileDI: medianDI: 10% quantile

∆ costs of uncertainty


Climate variabilityEach season optimized (perfect knowledge of soil characteristics)

Montpellier (semi-arid) Seeb (arid)dry wet dry wet

0 200 400 600 800 1 ,0000

2

4

6

8

10

12

14

16


Dry

yield

[t/h

a]

0 200 400 600 800 1 ,0000

2

4

6

8

10

12

14

16


Dry

yield

[t/h

a]

0 200 400 600 800 1 ,0000

2

4

6

8

10

12

14

16


Dry

yield

[t/h

a]

0 200 400 600 800 1 ,0000

2

4

6

8

10

12

14

16


Dry

yield

[t/h

a]


Soil variability: Montpellier (semi-arid)Each season optimized (perfect knowledge of weather development)

no information sandy loamdry wet dry wet

0 200 400 600 800 1 ,0000

2

4

6

8

10

12

14

16


Dry

yield

[t/h

a]

0 200 400 600 800 1 ,0000

2

4

6

8

10

12

14

16


Dry

yield

[t/h

a]

0 200 400 600 800 1 ,0000

2

4

6

8

10

12

14

16


Dry

yield

[t/h

a]

0 200 400 600 800 1 ,0000

2

4

6

8

10

12

14

16


Dry

yield

[t/h

a]


Soil variability: Seeb (arid)Each season optimized (perfect knowledge of weather development)

no information sandy loamdry wet dry wet

0 200 400 600 800 1 ,0000

2

4

6

8

10

12

14

16


Dry

yield

[t/h

a]

0 200 400 600 800 1 ,0000

2

4

6

8

10

12

14

16


Dry

yield

[t/h

a]

0 200 400 600 800 1 ,0000

2

4

6

8

10

12

14

16


Dry

yield

[t/h

a]

0 200 400 600 800 1 ,0000

2

4

6

8

10

12

14

16


Dry

yield

[t/h

a]


Example plotDifferent irrigation strategies and costs of uncertainty in [mm] for 0.9 quantile

0 200 400 600 800 1 ,0000

2

4

6

8

10

12

14

16


Dry

yield[t/ha]

FI: ensembleFI: averageDI: decision tableDI: irrigation calendarDI: perfect knowledge

∆

decision table irrigation calendarCosts in [mm]: 178 208


Yield for maize for a semi-arid site: FranceCosts for dry conditions in [mm]

Montpellier (semi-arid) Seeb (arid)climate variability soil variability climate variability soil variability

0 200 400 600 800 1 ,0000

2

4

6

8

10

12

14

16


Dry

yield

[t/h

a]

0 200 400 600 800 1 ,0000

2

4

6

8

10

12

14

16


Dry

yield

[t/h

a]

0 200 400 600 800 1 ,0000

2

4

6

8

10

12

14

16


Dry

yield

[t/h

a]

0 200 400 600 800 1 ,0000

2

4

6

8

10

12

14

16


Dry

yield

[t/h

a]

178 180 76 43 30 10 88 40


An Example for Seeb for mixed variabilityStochastic Crop Water Production Function for wet conditions

0 200 400 600 800 1 ,0000

2

4

6

8

10

12

14

16


Dry

yield

[t/h

a]

0 200 400 600 800 1 ,0000

2

4

6

8

10

12

14

16


Dry

yield

[t/h

a]

irrigation calendar simple deficit irrigationManagement of deficit irrigation systemsTU Dresden, University of Twente // Niels Schütze and Oleksandr MialykEGU General Assembly 2020, May 6, 2020 Slide 27 of 29

Conclusions and Outlook


Conclusions• value of information =̂ costs for compensation of uncertainty in units ofadditional water requirements• costs of climate uncertainty (variability) are higher for the semi-arid sitecompared to arid climate conditions• costs of soil uncertainty (variability) are higher for the arid site, and evenhigher for wet initial conditions• costs of soil uncertainty (variability) can be reduced by proper soil analysis• but highest added value of information have: knowledge aboutdeficit irrigation strategy and information about initial conditions


niels schütze and oleksandr mialyk tu dresden, university of … · 5 irrigation strategies x 32...

Documents