Crop yield predictions using seasonal climate forecasts Simone M. S. Costa and Caio A. S. Coelho Instituto Nacional de Pesquisas Espaciais

1
Crop yield predictions using seasonal climate
forecastsSimone M. S. Costa and Caio A. S.
CoelhoInstituto Nacional de Pesquisas Espaciais
INPE, São Paulo - SP
This study aims to investigate the potential of
using monthly mean climate forecasts from the
European Centre for Medium-range Weather Forecast
(ECMWF) model for producing maize yield
predictions in RS in 5 months in advance.
SEASONAL WEATHER DATA INTO CROP MODEL
STUDY AREA
Brazil is the 3rd main maize producer in the
entire world after USA and China, and RS State is
the 2nd greatest producer in Brazil (IBGE, 2006).
Maize yields interseasonal variability is high
due to irregular rainfall during the cropping
seasons.
A stochastic weather generator was used to
disaggregate the 11 ensemble members of monthly
mean rainfall into daily rainfall. Then the
disaggregated daily rainfall was used as input
data to a process-based crop model to predict
maize crop yield.
Weather generator estimates the rainfall
occurrence based on a 1st-order Markov chain and
the amount on a gamma distribution fit to 11-yrs
of daily observed rainfall.
Crop model - General Large Area Model, GLAM is a
process-based model for annual crops and requires
daily data of solar radiation, temperature and
rainfall.
Data - Rainfall - ECMWF forecasts (bias
corrected). Radiation Temp. - observed
climatology (INMET, FEPAGRO).
Figure 1 - RS state map showing the main maize
producer region and the 11 municipalities.
RESULTS
ECMWF couple seasonal forecast model does show
skill on monthly mean rainfall forecasts during
the maize crop cycle (Fig.2), suggesting the
potential use this information for crop
prediction.
There is a generally good agreement between the
simulated and the predicted yield, particularly
for the last 10yrs. For most years the obs. yield
is within 95 prediction interval, indicating
good reliability of yield predictions.
A reasonable agreement is noticed between the
observed and disaggregated histograms (Fig.3),
indicating that the used weather generator can
reproduce the observed daily rainfall
distribution accordingly.
Grain yield prediction for indiv. Municipality
produced six months in advance for 16 years
Correlation Between ECMWF monthly Forecasts and
Observed Rainfall Anomalies (1981-2005), Issue
Sep.
Daily rainfall histogram for Santa Rosa county
Figure 4 - Grain yield prediction produced 5
months in advance for 3 municipalites (3, 5 and
7, Fig1). Black line is the ensemble mean grain
yield (i.e. mean of the 11 dots). Dashed lines
indicate the 95 prediction interval.
Figure 3 Daily rainfall histogram for Sept. to
Feb. (1989 2005) based on observed rainfall and
disaggregated rainfall for two of the 11 ECMWF
ensemble members.
Figure 2 High positive correlation is noticed
over nearly all South America in September,
indicating good association between observed and
forecast anomalies.
Preliminary results show promising usefulness of
monthly mean rainfall forecasts produced by ECMWF
model for predict maize yield for RS in 5 months
in advance.
This work was supported by the EUROBRISA network
project (F/00 144/AT) kindly funded by the
Leverhulme Trust. The dynamical ensemble forecast
data were kindly provided by ECMWF as part of the
EUROSIP project. Three forecasting centres are
the partners in EUROSIP (ECMWF, the UK Met Office
and Meteo-France). The authors thank Vicent Moron
for making the weather generator software
available at IRI website. CASC thanks the FAPESP
(process 2005/05210-7 and 2006/02497-6) for
funding part of the EUROBRISA project.