Title: Enhanced photoperiod response modeling for improved biomass simulation in a Sudanian carbon accounti
1Enhanced photoperiod response modeling for
improved biomass simulation in a Sudanian carbon
accounting framework
in partnership with
with funding from
- P.S. Traoré, A. Folliard, M. Vaksmann, C. Porter,
M. Kouressy, J.W. Jones
2Overview
- The context
- C sequestration is only one (still small) part of
the livelihood portfolio - Climate risk will continue to reign in the poor
West African SAT - Local cereals, at the crossroads of tactical and
strategic opportunities ? a working base for a
win-win solution? - The problem
- Crop simulation models and local cereals
enhancements needed - Why improve predictive models in a data
assimilation framework anyway? - Methods
- Development (PP response), then growth
(partitioning) - DSSAT-Century
- Results (preliminary)
- Vegetative Phase Duration
- Biomass Production
- Next steps
3The context
4The place of sorghum in West Africa
Ntenimissa variety Guinea x Caudatum
hybrid Sudanian zone
Gadiaba variety Durra race Sahelian zone
- Major staple crop
- Mali 30 of cereal production
- With millet, 4th cereal worldwide
- More nutritive than maize, but tannins
- Losing ground to maize
5Millet and sorghum in a cotton-intensive year
(2003)
6Climate but what is so different about West
Africa?
High variability in both cases but
Sahel higher variations on decadal time steps
(low frequency)
does this mean relatively more risk for an annual
crop / farmer in SEA?
not necessarily because
Predictability is higher in SEA (both yearly and
in the long term)
SEA higher variations on yearly time steps (high
frequency)
(reproduced from IPCC, 2001)
7Climate but what is so different about West
Africa?
- Regional climate among the most variable in the
world (also most pronounced decadal change -0.3
rainfall over 20th century) - Largest tropical land mass with 6,000km east-west
extent ? high sensitivity to small surface
boundary forcings (yearly changes in land cover) - Regional climate modeling more complex reliance
on SST predictors not sufficient, weak ENSO
signal - Ability of GCMs to simulate observed interannual
Sahelian rainfall generally rather poor - Projections call for African climate warming,
esp. in semi-arid margins, but future changes in
rainfall less well defined in the Sahel
inconsistent projections, no or little change - Forecasting skill consistently lower over the
Sahel than for other regions of the globe,
especially at inter-annual time scales important
to agriculture (HF) - Total rainfall amounts have decreased, but no
significant change in LGP - Under SRES scenarii, precipitation may decrease
during the growing season and may increase at
other times of the year - Date of rains onset and distribution much more
critical to farmers than total amount, but rarely
in the set of predictands
Regional climate difficult to model
Regional climate (change) difficult to predict
8What would you do if you were an annual plant?
Sotuba (1239N, 755W)
Favorable rainfed cropping conditions
May-November
Decreasing daylengths
Daylength (h)
Rainfall (mm)
9What would you do if you were an annual plant?
- Limiting factor high rainfall variability
- Spatially along a N-S transect
- Temporally inter-annual
- Function of rains onset date
- Need to fit crop cycle to probable duration of
rains - Flexibility required from varieties to handle
climatic uncertainty - Photoperiod sensitivity in crops strategy to
manage climatic risk
10What would you do if you were an annual plant?
- Grouped flowering towards end of rainy season
- Minimize grain mold, insect bird damage (early
maturing varieties) - Avoid incomplete grain filling (late maturing
varieties)
x 3
Photoperiod sensitivity adaptation trait West
Africa highest PP sensitivity levels
worldwide Bonus for C sequestration large
biomass production
x 2
North
South
Dr. Hoogenboom (2m)
11The problem
12The problem
- Crop models and local cereals improvements are
needed
- Diagnostic
- underestimate photoperiod (PP) sensitivity
- do not parameterize PP sensitivity optimally
- underestimate vegetative phase duration
- do not partition biomass optimally
- underestimate vegetative biomass production
Cause (range of genetic coefficients
P2R) (choice of response curve, coefficients, DR
calculation approach) to be determined
- Why improve predictive models in a data
assimilation framework anyway? - Uncertainty reduction is critical for EnKF
performance and can be approached from either or
both sides (measurements, predictions) - It may be easier to reduce uncertainty in point
peak biomass estimates from models (as compared
to remote sensing)
13Data assimilation framework for C accounting
1. reduce uncertainties in measurements
(adaptation calibration, modification)
D
M
D
A
D
M
D
A
ATA
ATA
ODEL
ODEL
ATA
ATA
SSIMILATION
SSIMILATION
Biomass
Measured
Optimized
Measurement
Biomass
Biomass
ENSEMBLE
Estimation
Soil C
Measured
Soil Sampling
Soil C
KALMAN
DSSAT
Weather
Biomass
Simulated
FILTER
-
CENTURY
Management
Soil C
Simulated
Crop/Soil C
Soil Properties
Model
Parameters
Optimized Parameter Estimation
Optimized Soil
-
C Estimation
2. reduce uncertainties in predictions
(adaptation calibration, modification)
14Modeling current approaches
P1
P2
P3
P4
P5
P6
P0
Panicle initiation
Sowing
Flag leaf
End juvenile phase
Flowering
Maturity
Harvest
Start grain filling
Emergence
15Modeling current approaches
Modeling approaches will differ depending on how
they handle temperature photoperiod interactions
during the PIP
Juvenile phase Fixed duration No PI possible T
control
Photoperiod induced phase (PIP) Durationf(P,T)
Ends at PI P control
P1
P2
P3
P4
P5
P6
P0
Panicle initiation
Sowing
Flag leaf
End juvenile phase
Flowering
Maturity
Harvest
Start grain filling
Emergence
16Methods
17PP response options
- Response curves thermal time to PI as a
function of photoperiod - Purpose model the delay imposed by non-optimal P
on plant development (how it slows down its speed
or development rate) - Linear rice (Vergara Chang, 1985), other
SD/LD crops (Major Kiniry, 1991) sorghum
(Ritchie Alagarswamy, 1989) - Hyperbolic (Franquin, 1976 Hadley, 1983 Hammer,
1989 Brisson, 2002)
PI will eventually occur
PI may not occur
- Consequences for qualitative plants
18DR calculation options
- Even more important is the procedure for
calculating development rates (DR) - DR inverse of phase duration
- Case 1 cumulative photo-thermal ratios
- Case 2 threshold on thermal time requirements
- Physiological interpretation
- Plant progresses every day towards flowering
with a variable rate function of T and P
Requires that daylength conditions be met for
flowering to take place
19Experimental design
- Typical Guinea cultivar CSM388, avg. cycle
duration 130 days, P1413C.days (Vaksmann al.,
1996) - Calibration 1996 planting date experiment in
Sotuba (1239N), June-August, PI dates observed
by dissections every 5 days - Genetic coefficients screening ranges and
increments - Validation 1994 planting date experiment in
Sotuba (1239N), Cinzana (1315N) and Koporo
(1414N), February-September, FL expansion dates
observed and translated into PI dates
Flag Leaf Sowing date June 20
Flag Leaf Sowing date July 20
20Results
21Results (PP)
1996 experimental observations used for
calibration. All durations computed from
emergence
22Results (PP)
Model calibration. Best estimate of genetic
coefficients for the 4 model types
23Results (PP)
Scatterplots of calculated emergence-flag leaf
expansion durations (EFLcalc) against
observations from the 1994 experiment (EFLobs)
Cumulative Threshold
R20.89
R20.41
Linear Hyperbolic
R20.13
R20.97
24Results (PP)
- Predictions of EFL as a function of planting
dates for the 4 approaches, as compared to 6
observations (EFLobs) from the 1994 experiment in
Sotuba, Mali
25Growth quantitative, development qualitative ?
- Growing Degree Days appropriate to describe
quantitative processes such as plant growth, but - Photo-thermal time concept appears inappropriate
for simulation of plant progress towards
flowering ( plant development) - Short Day plants rather decreasing day
- Threshold-hyperbolic approach may be more
consistent with crop physiology as it associates - cumulative (temperature)
- processes
- and that better reflect
- trigger (photoperiod)
- events
- quantitative plant
- growth
- and
- qualitative plant development
- Need to incorporate more knowledge of plant
physiology genetics into phenological crop
models (shifts in hormone balances rather than
florigen concept, )
26Prospects
- Implementation in CERES-Sorghum is
straightforward replace 1 parameter, - re-write 3 lines of code
Source code RATEIN 1.0/102.0 IF (TWILEN
.GT. P2O) THEN RATEIN 1.0/(102.0P2R(TWILEN-P
2O)) ENDIF SIND SIND RATEINPDTT
Modifications RATEIN 1.0/P1 IF (TWILEN .GT.
P2O) THEN RATEIN (TWILEN-PBASE)/(P2O-PBASE) EN
DIF SIND RATEINSUMDTT
27Impact on biomass estimates
- Sotuba 1996 planting date experiment Sorghum
growth analysis
28Next steps
29Next steps
- DSSAT-Century development OK, now look at
growth ? biomass partitioning in sorghum and
millet (thesis M. Kouressy) - DSSAT-Century compute new set of genetic
coefficients for database cultivars - Using DSSAT-Century GIS interface, spatially
simulate biomass production for the Oumarbougou
study area (2004?)
30Conclusions
- Improved DSSAT-Century PP response code has been
included in a beta version and simulates plant
development correctly - The modified function is more consistent with
short-day plant physiology and has more universal
applicability in theory in practice, change of
1 genetic coefficient requires re-computation of
crop genetic sets in DSSAT-Century - The impact on simulation of VPD using a modified
PP response algorithm is negligible whenever the
crop cycle is below 120 days (ie under most
current normal planting conditions) no impact
on biomass expected under these conditions - Starting work on biomass partitioning will
probably have a more significant impact on
enhancing biomass estimation capability under
normal planting conditions