Italy, AGMIP Maize: analysis of the protocol with SARRA-H

Italy, AGMIP Maize

Italy, AGMIP Maize: analysis of the

: analysis of the

Italy, AGMIP Maize

Italy, AGMIP Maize: analysis of the

: analysis of the

protocol with SARRA

protocol with SARRA--H

H

AGRICU

AGRICULLTTURAL RESEARCHURAL RESEARCH FOR DEVELOPMENT

FOR DEVELOPMENT

www.cirad.fr www.cirad.fr

protocol with SARRA

protocol with SARRA--H

H

Christian BARON

Christian BARON

1

1

_,

_{, Myriam}

_{Myriam ADAM}

_ADAM

2

2

Christian BARON

Christian BARON

1

1

_,

_{, Myriam}

_{Myriam ADAM}

_ADAM

2

2

Contact: [email protected]

Contact: [email protected]

1

1

_{CIRAD, UMR TETIS, 500 rue J}

_{CIRAD, UMR TETIS, 500 rue J--F. Breton, Montpellier, F}

_{F. Breton, Montpellier, F--34093;}

_34093;

22

_{CIRAD, UMR AGAP/PAM Montpellier, France}

_{CIRAD, UMR AGAP/PAM Montpellier, France}

1

1

_{CIRAD, UMR TETIS, 500 rue J}

_{CIRAD, UMR TETIS, 500 rue J--F. Breton, Montpellier, F}

_{F. Breton, Montpellier, F--34093;}

_34093;

22

_{CIRAD, UMR AGAP/PAM Montpellier, France}

_{CIRAD, UMR AGAP/PAM Montpellier, France}

Calibration on 4 sites

USA, 167kg N.ha

-1

_,

_{no irrigation}

FR, 255kg N.ha

-1

_{, irrigation}

Sensitivity analysis on temperature, CO2 and N

Use of as default the management practices given in the experiments.

Main assumptions

FR, 255kg N.ha

-1

_{, irrigation}

Br, no N,

no irrigation

Tz, 61kg N.ha

-1

_{, irrigation}

(short rainy season)

Main assumptions

Consider the same initial soil water and sowing date as in the experimental dataset for the

scenario analysis (testing effect of temperature, CO2 and N).

No irrigation in USA and Brazil.

Questions

(short rainy season)

No irrigation in USA and Brazil.

Irrigation in France and Tanzania.

Questions

1. Why not using the soil water dynamics simulated within the model, rather than a fixed soil initial water?

2. By not irrigating, we might introduce water stress. What effect does it have on our simulations?

3. If we want to capture realistic management practices, should we not consider sowing date according to previous rain, to optimize crop emergence)?

3. If we want to capture realistic management practices, should we not consider sowing date according to previous rain, to optimize crop emergence)?

IMPACT ON YIELD

IMPACT ON YIELD

YIELD (t.ha-1)

Use the soil water dynamics from SARRA-H to test the impact initial soil water on _{YIELD (t.ha-1)}

USA BRAZIL FRANCE

Baseline* 7.5 (2.1) 7.4(1.4) 9.1(1.0)

Use the soil water dynamics from SARRA-H to test the impact initial soil water on

simulations for the different sites.

Use of automatic irrigation to determine the potential simulation of water stress in non

irrigated situation.

Simulating initial soil water non irrigated 5.4(2.9) 7.5(1.4) 2.8(1.2)

Simulating sowing date non irrigated 5.1(3.3) 7.5(1.3) 2.8(1.3)

irrigated situation.

Use of the potential sowing date routine from SARRA-H to test the impact of the sowing

date defined according to previous rain.

Simulating initial soil water irrigated 9.6(1.4) 8.1(0.9) 9.1(1.0)

Simulating sowing date irrigated

10.6

_(1.3) 8.4(0.8) 10.0(1.3) * irrigated for France; non irrigated for USA and Brazil

Use the soil water dynamics from SARRA-H to test the impact of initial soil water on

simulations for the different sites. * irrigated for France; non irrigated for USA and Brazil

Irrigation: Yield in the USA is impacted by water stress.

Table 1: Average grain yield (and standard deviation) from a time series simulation 1982-2010

450

Figure 1: Soil Water initialisation: baseline vs rainfall dynamic

Use the soil water dynamics from SARRA-H to test the impact of initial soil water on

simulations for the different sites.

Irrigation: Yield in the USA is impacted by water stress.

Sowing date: in potential growth conditions, defining an potential sowing date slightly

increase yield. The potential sowing date is on average 1 month before the defined sowing date (as in the experiment):

350 400 450

Figure 2: Main water balance component for (a) USA; (b) Brazil; and (c)France date (as in the experiment):

USA 1st April instead of 5 May; France 1stApril instead of 25April, and Brazil 15 September instead of 23 October

250 300 350 w a te r (m m ) _observedUSA, soil water 12 14

Figure 2: Main water balance component for (a) USA; (b) Brazil; and (c)France

150 200 S o il w a te r (m m 8 10 50 100 France Dyn France Baseline USA Dyn France, observed soil water 4 6 0 -220 -200 -180 -160 -140 -120 -100 -80 -60 -40 -20 0 20 40 60 80 100 120 140 160 180

Number of days after sowing

USA Dyn USA Baseline

WATER BALANCE

WATER BALANCE

(a) USA _{Initial soil water Irrigation Transpiration Drainage}

2 4

(a) USA _{Initial soil water Irrigation Transpiration Drainage}

Baseline no irrigation 205 (0) 0 357(53) 43(27) Simulating initial soil water non irrigated 103(70) 0 280(105) 43(27)

0

1981 1986 1991 1996 2001 2006

Rainfed Oct, Irrig and Sim Sowing date Rainfed Oct and Irrig

Simulated sowing date from Oct past Year

Simulating sowing date non irrigated 79(53) 0 296(117) 43(27) Simulating initial soil water irrigated 231₍₁₃₃₎ 413(43) 50(26) Simulating sowing date irrigated 317₍₁₇₇₎ 460(49) 96(24)

Simulated sowing date from Oct past Year Rainfed dynamic from Oct past year

Baseline (No Irrigation,constant: soil water init 205 mm, sowing date May 5th)

Discussions

Discussions

Simulating sowing date irrigated 317₍₁₇₇₎ 460(49) 96(24)

(b) BRAZIL _{Initial soil water Irrigation Transpiration Drainage}

Baseline non irrigated 0 414(49) 82(34) Simulating initial soil water non irrigated 84(53) 0 417(49) 82(34)

Discussions

Discussions

US: not potential production, experiences water stress.
Fr, Br: enough rain or irrigation.

Simulating initial soil water non irrigated 84(53) 0 417(49) 82(34) Simulating sowing date non irrigated 76(43) 0 421(46) 77(32) Simulating initial soil water irrigated 86(34) 442(29) 86(34)

Fr, Br: enough rain or irrigation.

• Water dynamics

If we want to capture the water dynamics and its impact of crop yield, would not it better to simulate the soil water dynamics considering the rain before the cropping season (without

(c) France Initial soil water Irrigation Transpiration Drainage

Baseline _irrigated 151(20) 288₍₆₀₎ 416(36) 11(7) Simulating initial soil water irrigated 86(34) 442(29) 86(34)

Simulating sowing date irrigated 81(31) 456(33) 81(31)

simulate the soil water dynamics considering the rain before the cropping season (without

irrigation)?

If so, how to decouple the water stress effect on the temperature effect tested in the scenario?

Baseline irrigated 151(20) 288₍₆₀₎ 416(36) 11(7) Simulating initial soil water non irrigated 149(23) 0 228(44) 5(5)

Simulating sowing date non irrigated 149(23) 0 234(48) 5(5) Simulating initial soil water irrigated 282₍₆₄₎ 417(35) 11(7)

C h ri st ia n B A R O N , O ct o b er 2 0 1 2 C h ri st ia n B A R O N , O ct o b er 2 0 1 2 scenario? • Sowing date

If we represent the inter-annual variability by considering rain (without irrigation), it seems important to simulate variable sowing date corresponding to the right initial conditions (Fig. 2)

Table 2: Main water balance component for (a) USA; (b) Brazil; and (c) France

Simulating initial soil water irrigated 282₍₆₄₎ 417(35) 11(7) Simulating sowing date irrigated 292₍₆₉₎ 445(42) 11(7)

C h ri st ia n B A R O N , O ct o b er 2 0 1 2 C h ri st ia n B A R O N , O ct o b er 2 0 1 2

important to simulate variable sowing date corresponding to the right initial conditions (Fig. 2)

If so, we avoid a sowing date when there is water stress; we observe a slight gain in yield.

USA: we show the importance of irrigation to

simulate potential growth conditions.

Brazil: rain seems to be enough; less than 100mm

C h ri st ia n B A R O N , O ct o b er 2 0 1 2 C h ri st ia n B A R O N , O ct o b er 2 0 1 2

• Should we really consider fixed management practices, as defined in the

experimentation (irrigation vs. no irrigation, sowing date) if we want to

Brazil: rain seems to be enough; less than 100mm

of irrigation is needed for potential growth conditions.

France: With automatic irrigation, we reproduce

C ir a d ~ U M R T E T IS : C ir a d ~ U M R T E T IS : C h ri st ia n B A R O N , O ct o b er 2 0 1 2 C h ri st ia n B A R O N , O ct o b er 2 0 1 2

experimentation (irrigation vs. no irrigation, sowing date) if we want to capture annual variability?

• Tanzania: not shown in this analysis as maize appears to be mostly planted

during the long rainy season in Morogoro (Tumbo et al.*, to discuss with

France: With automatic irrigation, we reproduce

the average irrigation needed, as done in the experiment used for calibration.

C ir a d ~ U M R T E T IS : C ir a d ~ U M R T E T IS :

during the long rainy season in Morogoro (Tumbo et al.*, to discuss with

experts), while our data are on the short rainy season. How well do we want to represent farmers’ practices?