Using Seasonal Climate Forecasts To Improve Maize Production Decision Support in Zimbabwe 2011 Agricultural and Forest Meteorology

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Agricultural and Forest Meteorology 151 (2011) 17921799

Contents lists available at ScienceDirect

Agricultural and Forest Meteorology


journal homepage: www.elsevier.com/locate/agrformet

Using seasonal climate forecasts to improve maize production decision support


in Zimbabwe
N. Zinyengere a , T. Mhizha a, , E. Mashonjowa a , B. Chipindu a , S. Geerts b , D. Raes b
a
b

Department of Physics, University of Zimbabwe, Mount Pleasant Drive, Harare, Zimbabwe


Division of Soil and Water Management, K.U. Leuven University, Celestijnenlaan 200E, 3001 Leuven, Belgium

a r t i c l e

i n f o

Article history:
Received 7 September 2010
Received in revised form 19 July 2011
Accepted 22 July 2011
Keywords:
El Nino Southern Oscillation (ENSO)
Southern Oscillation Index (SOI)
RAINMAN
Weather forecasts
AquaCrop

a b s t r a c t
Maize production in marginal tropical regions is at great risk due to rainfall variability and climate change.
Climate change is set to increase the variability and uncertainty of inter-annual rainfall. Farmers who
depend on rainfed maize production for their livelihoods would therefore benet from improved climate
based forecasting of production likelihood. In this study we developed a simple maize production decision
support tool for Masvingo by using seasonal climate forecasts and a crop model to forecast maize yields
likelihood prior to the season. We follow up on earlier studies carried out in Zimbabwe which show that
the El Nino Southern Oscillation (ENSO) can be used to forecast rainfall and maize yields in Zimbabwe. An
ENSO based seasonal climate analysis tool (RAINMAN) was used to produce probabilistic monthly climate
forecasts for Masvingo corresponding to the phases of the Southern Oscillation Index (SOI). The climate
forecasts were used to run a crop model (AquaCrop) for a variety of scenarios relevant to maize production
(monthly rainfall, cultivar selection, planting date, and fertility level). The results of the simulations
were similar to those observed by Phillips et al. (1997) and formed the basis for the development of an
operational decision support tool. Simulated maize yields varied from 1.2 t/ha to 5.8 t/ha. The simulated
yields were higher than expected average yields in a marginal region like Masvingo especially under small
holder farming. The work suggested that optimal use of forecasts may lead to improved maize production
in Masvingo. The study set a platform for the development of operational climate based maize production
decision support tools in Zimbabwe.
2011 Elsevier B.V. All rights reserved.

1. Introduction
Maize is the staple food for Zimbabwe. It is the determinant
of household food security (Matarira et al., 2004). Maize is mainly
grown for subsistence by small holder farmers under rainfed conditions. Farmers who depend on rainfed agricultural production
to produce food have been known to experience perennial food
shortages. Hansen (2002) stated that rainfall variability is the single most important factor in agricultural production. Therefore,
to ensure food security, farming systems need to adjust to interannual rainfall variability. Seasonal climate forecasts are probably
the most useful tool in managing inter-annual rainfall variability
and in protecting farming communities from perennial food shortages. Climate change on the other hand is expected to increase
climate variability and exert more pressure on food security in rainfed farming system (Ogallo et al., 2000). Therefore, it is important

Corresponding author at: Department of Physics, University of Zimbabwe, PO


Box MP167, Mount Pleasant, Harare, Zimbabwe. Tel.: +263 712 798260;
fax: +263 4 307156.
E-mail addresses: mhizhat@yahoo.com, tmhizha@science.uz.ac.zw (T. Mhizha).
0168-1923/$ see front matter 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.agrformet.2011.07.015

to improve the usefulness of seasonal climate forecasts in order to


avoid current and future food shortages.
In Zimbabwe, the Zimbabwe Meteorological Service Department (ZMSD) is responsible for giving out seasonal forecasts. They
give out these forecasts towards and during each agricultural
season. Research by Phillips et al. (2001) suggests that farming
households receive these forecasts and apply them to some degree.
Patt and Gwata (2002) however asserted that one of the major challenges to the usefulness of the national forecast is its geographical
and temporal scale. They note that the ZMSD downscales regional
forecasts to Zimbabwe without adding much local specicity. The
ZMSD also forecast total rainfall for the season (OctoberMarch) or
for the rst (OctoberDecember) and second halves of the season
(JanuaryMarch). This information is not adequate. Hansen (2002)
notes that farmers need information that is relevant at the eld
scale, and that is expressed in terms of impacts and management
implications within the systems they operate.
Advances in the application of climate prediction to agriculture
suggest potential for improved production management strategies, enabling producers to better tailor management decisions
to the season (Hansen, 2002). Farmers can use site specic seasonal forecasts to mitigate unwanted impacts or take advantage

N. Zinyengere et al. / Agricultural and Forest Meteorology 151 (2011) 17921799

of favourable conditions thereby stabilizing yields. By providing


advance information with a sufcient lead time to adjust critical
agricultural decisions, seasonal forecasts have signicant potential
to contribute to the efciency of agricultural management and to
food and livelihood security (Apipattanavis et al., 2010).
Southern Oscillation (ENSO)
Phillips et al. (1997) studied El Nino
related maize yield variability in Zimbabwe, showing a signicant
relationship between ENSO and Zimbabwean maize yields. Studies
relating ENSO and crop yields have been carried out all over the
world with relative success (Cane et al., 1994; Hammer et al., 1996;
Fraisse et al., 2006; Shin et al., 2006; Meza, 2007; Baigorria et al.,
2008). However, despite perennial food shortages experienced in
Zimbabwe, few studies have attempted to build onto these earlier
studies and come up with tailor-made forecast tools for decision
making to cater for the management needs of small holder farmers
in the marginal areas of Zimbabwe.
The overarching goal of this paper is to build onto earlier studies
on ENSO based climate and maize yield forecasting in Zimbabwe
with the aim to develop a simple to use decision support tool for
maize production. We apply available seasonal climate forecasting
and crop growth simulation models. An ENSO-based seasonal climate forecast analysis tool, RAINMAN (Clewett, 1995) and a crop
production model, AquaCrop (Raes et al., 2009) are used to develop
a maize production decision support tool which can be used by
farmers in planning for an upcoming agricultural season. Although
the modeling ignores key constraints faced by small holder farmers,
it is hoped that the results of this study can be used in advanced participatory development of maize production management decision
support tools in Zimbabwe. We use the Masvingo climate station
as a case study.
2. Background
It is now well known that inter-annual variability in rainfall
in Zimbabwe is strongly correlated with the ENSO phenomenon
(Cane et al., 1994; Phillips et al., 1997, 2002; Patt and Gwata, 2002).
Although additional factors inuence the rainfall season, ENSO is
routinely used to forecast seasonal rainfall in Zimbabwe along with
other tools. Cane et al. (1994) noted that both seasonal rainfall over
Zimbabwe and the countrys maize yields, correlated strongly with
ENSO to the degree of more than 50% with the Nino-3 sea surface
temperature index.
Phillips et al. (1997) went further to investigate the value of
ENSO in the prediction of rainfall and maize yields in Zimbabwe
and thus maize production decisions making. Their conclusion
was that rainfall was strongly correlated to the phases of ENSO.
They also concluded that maize yields uctuated signicantly
with the phases of ENSO. Phillips et al. (1997) noted that precise seasonal climate predictions would be necessary in order
for forecasts to be useful for maize management decisions.
However, they also stressed the value of seasonal forecasts to
maize production decisions especially in more marginal areas
of Zimbabwe where low rainfall and high rainfall variability are
common.
Following the earlier studies, we sought to add weight to their
contentions which showed the potential of using ENSO to predict
seasonal climate and potential maize yields in Zimbabwe especially
in marginal maize production areas. We noted that despite the
potential of these studies, no work has been done yet to try and
operationalise it for the benet of small holder farmers in marginal
regions. We went on to apply a set of climate and crop growth
modeling tools which have not been used in Zimbabwe to come
up with a decision support tool based on the predictions of rainfall
and potential maize yields. Masvingo was selected for a case study
because it lies in a marginal region in Zimbabwe where low rainfall
and high variability are common.

1793

Observed climate predictors

Statistical
climate model

Analogue
years

Downscaled
dynamic model

Stochastic weather
generator

Crop model

Predicted crop yields


Fig. 1. Potential pathways to localized simulation-based predicted crop yields from
large scale observed climate predictors.
Source: Adapted from Hansen and Indeje (2004).

In this study, we used the Southern Oscillation Index (SOI). The


strength of the Southern Oscillation is measured by the difference
in air pressure between Darwin and Tahiti in the Pacic Ocean.
The SOI usually ranges from 30 to +30. When the SOI is strongly
positive or rising, the trade winds blow strongly across the warm
Pacic picking up plenty of moisture. Above average rainfall is likely
to be experienced in certain locations around the world including
Zimbabwe. When the SOI is strongly negative or falling, trade winds
are weak, and rainfall in Zimbabwe can be below average. A neutral
SOI is likely to result in normal rainfall (Clewett et al., 1992). This
relationship formed the basis from which maize yield predictions
and the maize management decision support tool were made.
3. Methodology
3.1. Linking the climate and crop models
In order to predict maize yields, we rst generated multiple climate scenarios consistent with the different phases of ENSO using
the ENSO-based climate analysis tool (RAINMAN). The climate scenarios were used as climate input for the crop models (AquaCrop)
simulations which produced probabilistic forecasts of maize. What
has been the standard approach for some time is to categorize the
observed predictor variables (e.g. ENSO phases), and use the predictor category to select sets of analog years from the observed station
time series as input to the crop model (Hansen and Indeje, 2004).
This method is referred to as the historical analogues approach.
RAINMAN uses a similar approach by conditioning the historical analogues with categorical indices based on the ENSO. Given
predicted ENSO conditions, RAINMAN then goes on to perform statistical tests to ascertain the probability of receiving a given rainfall
amount in the predicted ENSO phase. Fig. 1 shows the approaches
for linking dynamic crop simulation models with climate predictors.
3.2. The climate analysis tool (RAINMAN)
The seasonal climate forecast tool (RAINMAN) used for this
study aims to develop knowledge and skills for managing climate

1794

N. Zinyengere et al. / Agricultural and Forest Meteorology 151 (2011) 17921799

variability in agriculture by analysing effects of ENSO on rainfall


to derive probability-based seasonal climate forecasts (Clewett,
1995). RAINMAN allows users to input a set of historical rainfall
data from any rainfall station for analysis. RAINMAN categorises
the historical rainfall data into the 5 phases of the SOI in which the
rainfall amount was received (i.e., positive, negative, falling, rising
and neutral). It performs probabilistic prediction of rainfall at a seasonal lead time based on discrete categories or phases of the SOI.
Users can specify the rainfall period they would like to forecast
e.g. the entire season (OctoberApril) or a specic month within
the season, e.g., November, January, March etc. RAINMAN allows
the user to input the exact SOI value of the months they intend
to use as predictor months, e.g., July (4.5) and August (10.2). The
tool uses this input information to perform statistical tests which
determine the relationship between the SOI and rainfall for the
location thereby producing the probability of occurrence of various rainfall amounts e.g. given input predictor months condition
i.e. July (4.5) and August (10.2), there is 20% chance of receiving at
least (783 mm), 50% chance of receiving at least (625 mm) and 80%
chance of receiving at least (459 mm). The major statistical tests
used in RAINMAN are the KruskalWallis (KW), and the LEPS (linear error in probability space) skill score test and the probability
score (p) (Clewett, 1995). RAINMAN version 4.1 was used in this
study.
3.3. The crop model (AquaCrop)
AquaCrop is a dynamic water-driven simulation model that
requires a relatively low number of parameters and input data to
simulate the yield response to water of most of the major eld and
vegetable crops cultivated worldwide. Its parameters are explicit
and mostly intuitive and the model maintains sufcient balance
between accuracy, simplicity and robustness (Steduto et al., 2009).
The models ability to use few parameters while maintaining accuracy makes it attractive for marginal locations in Zimbabwe where
some information may be unavailable.
The model simulates the three major components which require
management i.e. the soil, plant and atmosphere. AquaCrop models
the soil, with its water balance; the plant, with its development,
growth and yield processes; and the atmosphere, with its thermal
regime, rainfall, evaporative demand and carbon dioxide concentration. Additionally, some management aspects are explicitly
considered (e.g., irrigation, fertilization, etc.) as they will affect
the soil water balance, crop development and therefore nal yield
(Raes et al., 2009). Required model inputs include the climate
(minimum and maximum temperature, precipitation and reference evapotranspiration (ETo), crop characteristics, planting date,
eld management (fertility), and soil properties. AquaCrop computes plant growth and development processes on a daily basis in
a specic location, from planting date to maturity date.
AquaCrops determination of crop yield (Y) is water based. Relative evapotranspiration (ET) is pivotal in determining yield. ET is
separated into crop transpiration (Tr) and soil evaporation (E), so as
to separate E to avoid the confusing effect of non-productive water
use. The core function of AquaCrop is expressed by the equation
Steduto et al. (2009):
B = WP Tr (Steduto et al., 2009)
where B is the nal biomass. The nal yield (Y) is a function of
nal biomass (B) and harvestable biomass expressed as HI (harvest
index). WP is the water productivity (biomass per unit of cumulative transpiration), which tends to be constant for a given climatic
condition (Steduto et al., 2009).
AquaCrop has been validated for maize in varying locations
and was declared applicable to a wide range of conditions and

non-specic to a given maize cultivar (Heng et al., 2009; Hsiao et al.,


2009). The model allows for the use of monthly or 10 daily weather
data despite quantifying plant physiological response on a daily
time-step (Raes et al., 2009). These characteristics ensured that we
could carry out simulations for Masvingo despite a lack of daily
weather data and maize yield data. AquaCrop version 3.0 was used
in this study.
3.4. Of the case study
We applied the seasonal forecast-crop model framework in
Masvingo district. Masvingo (longitude 30 52 E; latitude
20 04 S) is located in south-east Zimbabwe. It is located in natural
region 4 (NR4) under the farming regions classication by Vincent
and Thomas (1960). NR4 covers 37% of the land area in Zimbabwe
where rainfall is inherently variable and unreliable (Matarira et al.,
2004). Masvingo has an altitude of 1100 m above sea level. Mean
annual rainfall in NR4 ranges between 400 and 650 mm, with
Masvingo averaging 641 mm (Chenje et al., 1998). Mean monthly
maximum temperatures range from 25 to 29 C (Chenje et al.,
1998). Masvingo has a predominantly semi-arid climate. Although
the area is semi arid and marginal for maize production, a large
proportion of the population still grows maize for food and subsistence.
Ochse et al. (1961) suggested that in very general terms, optimum rainfall conditions for maize are a little rain at the start of
the growth period, soaking rains every 45 days from the end of
the rst month up to about 3 weeks after owering and a gradual
tapering off of rain until harvest. Consequently we opted for generating monthly rainfall totals as opposed to seasonal totals in order
to account for the distribution of rainfall during the maize growing
period from sowing to maturity. We used monthly rainfall totals
generated for each SOI phase through RAINMAN. We produced
multiple realisations of crop yields based on a variety of agrometeorological scenarios which we used to develop a decision support
tool for Masvingo. In the coming sections we detail the major steps
taken to generate rainfall series for the 5 SOI phases and to simulate
maize yields from various agrometeorological scenarios relevant to
small holder maize production management.
3.5. Generation of weather series
Historical mean monthly minimum and maximum temperature
values were used as input parameters to AquaCrop. Meza et al.
(2003) showed that the ENSO effect on temperature is not a marked
as that of rainfall. We concluded that historical mean monthly
temperatures would sufce for the purposes of this study. Daily reference evapotranspiration (ETo) for the months within the growing
period (October to April) was calculated from mean monthly minimum and maximum temperatures using a software package (ETo
Calculator) based on the procedures described by Allen et al. (1998).
Mean monthly rainfall records for Masvingo from
1899/19001989/1990 were available within RAINMAN. These
were assumed to be accurate since RAINMAN was developed with
the input of the Zimbabwe Meteorological Services Department
(ZMSD). The data was appended up to the 2006/2007 season
using data from ZMSD. The rainfall data was used to generate
probabilistic monthly rainfall totals for Masvingo conditioned on
the ve phases of the SOI. 15 realisations of monthly weather
series were generated for running AquaCrop. The weather series
represented three probabilities of occurrence in each SOI phase
i.e. 20% (rainfall (mm) likely to occur once in every 5 years), 50%
(rainfall (mm) likely to be received once every 2 years) and 80%
(rainfall (mm) likely to be received in 4 out of every 5 years).
The 20% probability of occurrence was dened as above normal
category. The 50% probability of occurrence was dened as the

N. Zinyengere et al. / Agricultural and Forest Meteorology 151 (2011) 17921799

1795

Table 1
Showing R2 , t-statistics, the error statistics mean square error (MSE), root mean square error (RMSE), mean absolute error (MAE), and the percentage variance of the MAE
from the mean rainfall (brackets).
Month

Mean rainfall (mm)

R2

MSE (mm)

RMSE (mm)

MAE (mm)

October
November
December
January
February
March
April

25
80
145
144
119
82
22

0.76
0.70
0.37
0.68
0.67
0.48
0.81

444
1180
11113
3935
4941
7064
764

21
34
105
63
70
84
28

9.6 (36)
21 (26)
57 (39)
38 (26)
32 (27)
34 (42)
15 (68)

3.6. Maize yield simulations


The generated rainfall series, mean monthly maximum and
minimum temperature and ETo were used as weather input to
AquaCrop to simulate maize yields (tons/hectare). AquaCrop is considered applicable to a wide range of conditions and non-specic
to a given maize cultivar (Heng et al., 2009). AquaCrop was therefore suitable for Masvingo since maize yield data was not readily
available for model validation. Representative soil conditions for
Masvingo were obtained from Sithole (2003). The soil texture characteristics at Masvingo were considered to be a moderately deep
coarse loam (Phillips et al., 1997). Since the length of season for
Masvingo is sometimes as short as 95 days (dry years) and as long
as 145 days (wet years) (Sithole, 2003), three maize cultivars were
used during the simulation. 100-day (early maturing), 125-day and
140-day (Late maturing) maize cultivars were used. The length of
growth stages of maize were based on calendar days and not growing degree days. Conservative characteristics applied to all maize
cultivars were: biomass WP of 29 g m2 ; HI of 36%; a maximum
rooting depth of 1.2 m and planting density of 37 037 plants/ha.
Three planting dates (early (29 October), median (16 November),
and late (7 December)) were used to run the simulations. The planting dates were based on the optimal planting dates generated by
Sithole (2003) using a depth criterion. Two levels of fertilizer application were used: a optimal level which represents a soil which
is not nitrogen decient throughout the season and a poor level
which represents a nitrogen decient soil, more representative of
resource poor farming communities in Zimbabwe.

(MAE), Loague and Green (1991). There is no absolute criterion for


a good value of the error statistics mentioned above. It depends
on the units in which the variable is measured and on the degree of
forecasting accuracy, as measured in those units, which is sought
in a particular application (Legates and McCabe, 1999).

4. Results and discussion


4.1. Validation of RAINMAN and generation of weather series
Historical observed monthly rainfall totals were compared to
monthly rainfall totals generated by RAINMAN for the period
1991/19922006/2007. Table 1 and Fig. 2 show that RAINMAN
has a signicantly useful ability in predicting the variation of rainfall per month during the period 1991/19922006/2007. RAINMAN
explained more than two thirds of the variation of monthly rainfall
in October, November, January, February and April as shown by the
coefcients of determination which range from 0.67 to 0.81. The
relationship was marginally signicant for the months of December (0.37) and March (0.48). For the month of December, RAINMAN
failed to make good predictions when very rainfall amounts were
recorded. RAINMAN had difculty predicting highly anomalous
rainfall amounts experienced in December of 1992/1993 (376 mm),
2001/2002 (330 mm) and 2005/2006 (457 mm) as shown by the
scatter in Fig. 2. A highly anomalous rainfall amount of 413 mm
received in March 2003 resulted in only 48% of rainfall variation
being explained. This was a result of cyclone Japheth which the
tool is not designed to predict. April rainfall predictions showed
a very signicant relationship to the observed rainfall (R2 = 0.87).

500

predicted total monthly rainfall (mm)

normal category while the 80% probability of occurrence was


dened as the below normal category.
Although maize yield simulations were performed for all the
15 realisations of monthly rainfall series, it was determined that
the 50% category was more relevant and useful for maize production management at Masvingo. Hence, for the purposes of this
study only 5 realisations of monthly weather were considered
i.e. one series per SOI category. Rainfall data from 1899/1900 to
1990/1991 was used to train the model; this was termed the estimation period. 1991/19922006/2007 (validation period) rainfall
data was used to validate the performance of RAINMAN. RAINMAN
was used to generate rainfall totals for each month of the agricultural season (October, November, December, January, February,
March and April) for the validation period given SOI conditions for
July and August. RAINMAN simulations were compared with actual
observed monthly totals for the same period. The combined data
from 1899/19902006/2007 was used to generate the rainfall series
which formed the input into AquaCrop.

450
400
350
300
250
200
150
100
50
0

100

200

300

400

500

3.7. Data analysis

Observed total montly rainfall (mm)


We employed standard descriptive measures of goodness-of-t
to evaluate the accuracy of seasonal rainfall predictions made by
RAINMAN. The goodness of t measures include, mean square error
(MSE), root mean square error (RMSE), and mean absolute error

Fig. 2. Observed and simulated total monthly rainfall (mm) for the 7 months (October diamonds: November shadowed squares: December triangles: January
dash: February crossed squares: March circles: April squares) within the maize
growing period in Masvingo for the validation period (1991/19922006/2007).

1796

N. Zinyengere et al. / Agricultural and Forest Meteorology 151 (2011) 17921799

ily on seed and fertilizer inputs through government subsidies and


nongovernmental organizations (NGOs). Whereas optimum fertilization application rates vary from one place to the other according
to soil type, rainfall distribution, topography, etc., poor farmers
end up applying available fertilizer amounts as opposed to suitable
amounts. The risk associated with the use of expensive fertilizer
in low and unpredictable rainfall environments also hinder most
farmers from full investments in optimal fertilizer application.
Planting date was however, signicant for the maize cultivars under
optimal fertility especially for the 140-day cultivar for a falling, negative and neutral SOI phase. Under these conditions, yields varied
by as much as 1.9 t/ha. For the 100-day and 125-day maize cultivars, planting date was important given a neutral SOI. Yields varied
by as much as 1.3 t/ha. It is clear that the typical average yields
experienced in marginal areas of Zimbabwe (12 t/ha) are common
when poor fertilizer application is experienced despite the climate
conditions.
Fig. 3. Monthly rainfall generated using RAINMAN by SOI phase (falling continuous line; negative squares; neutral triangles; rising dots; and positive stars)
and 50% probability of occurrence for the growing season at Masvingo.

However, a MAE of 68% of mean rainfall (Table 1) shows that April


predictions are not as accurate as the relationship may imply.
t-distribution signicance tests carried out for the correlation
coefcients on the relationship between observed and predicted
rainfall accepted Ho (t < t=0.025 (2.49)) for all months thereby
showing that a signicant relationship exists between RAINMAN predictions and observed rainfall for all months. It was
concluded that RAINMAN shows a signicant predictive ability in making monthly predictions of rainfall for the period of
1991/19922006/2007.
The mean monthly maximum and minimum temperatures were
calculated using data obtained from the ZMSD for a period of at least
30 years. ETo values were obtained using the FAO ETo calculator
using the minimum and maximum temperatures in Table 2 for a
semi-arid location and moderate wind speed.
Fig. 3 shows the predictions of rainfall made by RAINMAN using
the SOI phase system for the months October to April for the 50%
probability of occurrence category of rainfall. The rainfall series
shown in Fig. 3 are the 5 weather series used as rainfall input in
crop yield simulations. The rainfall amounts generated by RAINMAN as shown by Fig. 3 show the expected trend of lower total
monthly rainfall amounts for a falling and negative SOI index and
higher rainfall amounts for the rising and positive SOI index for the
growing season relative to the neutral phase. As expected, rainfall
peaks for the season were found between December and February
for all SOI phases and the 50% probability of occurrences.
4.2. Crop simulations and decision criterion
In order to achieve a clear picture of the dynamics between
maize growth and agrometeorological factors at Masvingo, maize
production simulations were carried out bearing probabilistic rainfall predictions by RAINMAN based on the SOI phase, fertility levels
and optimal planting dates. Grain yields were obtained for these
scenarios for 3 maize cultivars as shown in Table 3.
Maize yields are typically around 12 t/ha on smallholder farmers elds in marginal regions like Masvingo (Phillips et al., 2002)
and 48 t/ha on average for the whole of Zimbabwe (SeedCo, 2005).
Simulated maize yields ranged from 1.2 to 5.8 t/ha. Sowing date
was found to have no particular impact on the maize yields for all
the SOI phases under poor fertility. This showed that for Masvingo,
farmers are best advised to apply optimal fertilizer levels in order
to achieve good yields. This however presents another challenge as
most resource poor farming communities in Zimbabwe rely heav-

4.2.1. SOI impacts by maize cultivar


The 100-day maize cultivar experienced highest grain yields of
4.4 t/ha. 125-day maize cultivars experienced maximum yields of
5.3 t/ha while the 140-day maize cultivar experienced the highest
grain yields of 5.8 t/ha. The highest grain yields for each cultivar
were experienced during a rising or positive SOI phase. The rising
and positive SOI correlates with above normal rainfall in Zimbabwe.
The lowest yields for each maize cultivar were experienced during
the falling or negative SOI phases which correlate with low rainfall
in Zimbabwe.
4.2.2. SOI impacts by planting date
Planting late (December 7) during the rising and positive SOI
phases attains highest yields for all the maize cultivars (100 day
4.4 t/ha; 125 day 5.3 t/ha; 140 day 5.8 t/ha). This suggests that
during seasons when above normal rains are predicted (positive
and rising SOI), moisture would likely be available for the entire
season thereby allowing for late planting. By planting late, farmers avoid the potential false starts which may occur early in the
season. This ensures that crops do not suffer moisture stress in the
initial and crucial stages of development. The 100 day maize cultivar obtains optimum yields for all SOI phases when planted late.
This results from the crop making full use of the high rainfall common in December, January and February since the entire growth
period of the cultivar will span this period. Planting early is most
benecial when a neutral SOI is predicted. Neutral SOIs correlate
with normal rains in Zimbabwe. The 125-day maize cultivar is most
productive for this phase when early planting is exercised.
4.2.3. SOI impacts by fertility level
Maize yields were depressed under poor fertility for all maize
cultivars for a rising, positive and neutral SOI phase. This showed
that despite predicted good rainfall conditions, poor fertility levels
would greatly compromise yields. Maximum predicted yields of
5.8 t/ha were obtained for optimal fertility and a positive and rising
SOI for the 140-day maize cultivar (Table 3). The 140-day maize
cultivar yielded the maximum yields for near optimal fertility levels while the 100-day cultivar did better under poor fertility levels
for all SOI phases, yielding up to 2.4 t/ha (Table 3). The short season
variety is therefore a better option for farmers in marginal small
holder farming areas where soil fertility levels are low and fertilizer application is inadequate. Minimum grain yield was obtained
for the 140-day maize cultivar under poor fertility. Longer season
varieties are unsuitable for areas were soil fertility is poor.
4.2.4. Decision support tool
Fig. 4 shows the decision support tool for maize production at
Masvingo. The tool was developed based on the maize yield pre-

N. Zinyengere et al. / Agricultural and Forest Meteorology 151 (2011) 17921799

1797

Table 2
Mean monthly maximum and minimum temperatures and ETo for the growing season at Masvingo.

Mean monthly maximum


temperature ( C)
Mean monthly minimum
temperature ( C)
ETo (mm/day)

October

November

December

January

February

March

April

29

29.3

28.5

28.6

27.7

27.6

26.1

14.5

16.2

16.9

17.3

16.9

15.7

12.5

5.5

5.7

5.5

5.4

5.0

4.7

4.0

Table 3
Simulated maize yields by SOI phase for the 3 maize cultivars and sowing dates for Masvingo for poor and near optimal fertility and 50% probability of rainfall occurrence
for each phase.
Fertility level

Mean yields (t/ha)


Early (29 October)

Falling
Negative
SOI phase

Neutral
Positive
Rising

Poor
Near optimal
Poor
Near optimal
Poor
Near optimal
Poor
Near optimal
Poor
Near optimal

Mid (16 November)

Late (7 December)

100-day

125-day

140-day

100-day

125-day

140-day

100-day

125-day

140-day

2.2
3.7
2.1
3.9
2.2
4.1
2.2
4.1
2.2
4.1

1.8
3.1
1.8
4.6
1.9
4.9
1.9
5
1.8
5

1.8
2.3
1.7
3.5
1.8
4.2
1.8
5.4
1.8
5.4

2.2
3.9
2.1
4.1
2.3
4.2
2.3
4.3
2.3
4.3

1.8
3.5
1.9
4.9
1.9
4.3
2
5.2
1.9
5.1

1.8
2.7
1.8
4.5
1.9
3.1
2
5.8
2
5.7

1.9
4
2.3
4.2
2.3
3.6
2.4
4.4
2.2
4.4

1.9
3.2
1.9
4.9
1.9
3.6
2
5.3
1.9
5.2

1.9
2
1.9
4
1.9
2.3
2
5.8
2
5.8

dictions by AquaCrop with the ENSO based rainfall predictions by


RAINMAN. The tool is meant to help maize producers to make decisions for maize production management based on the phase of the
SOI prevailing in July and August of each season. The effects of a
strong ENSO have been known to prevail for as long as 9 months
(Clewett, 1995), thereby making the SOI suitable for Masvingo
where the effective crop growth period is no more than 5 months

(Sithole, 2003). In essence, the tool condenses the science involved


in ENSO based maize yield predictions into a simple to use applicable instrument which farmers can use.
It was found that the 50% probability of occurrence (rainfall
likely to be received once every 2 years given the prevailing ENSO
condition) produced the best option for farmers who sought to
estimate their yields and tailor their management options with a

Fig. 4. Decision criteria for maize production at Masvingo, showing expected grain yields based on SOI phases (normal (50%) probability of occurrence), optimal fertility,
maize cultivar and planting date.

1798

N. Zinyengere et al. / Agricultural and Forest Meteorology 151 (2011) 17921799

reasonable level of accuracy. Only optimal fertility levels were considered since yields were mostly depressed for poor fertility levels
despite rainfall amount, maize cultivar or planting date (Table 3).
The decision support tool shows the optimal grain yields which
can be obtained for each maize cultivar and optimal planting date,
given prevailing ENSO conditions for July and August. The use of
ENSO conditions prior to the beginning of the rainy season allows
for farmers to make adjustments to their maize production management based on the prevailing ENSO conditions and the potential
implications for the coming farming season. The decision support
tool quanties the impact of the prevailing ENSO conditions on
maize yield with emphasis on the management options which
farmers have i.e. cultivar selection, planting date and fertilizer
application.
In applying the decision support tool farmers would note that
when the SOI phase is rising or positive, yields as high as 5.8 t/ha can
be attained if the 140-day maize cultivar is planted late (7 December) with sufcient fertilizer application. They will also note that,
all the maize cultivars attain their highest possible yields if planted
late for a rising and positive SOI phase. 100-day cultivars can yield
as much 4.4 t/ha despite predicted poor rainfall conditions (falling
and negative SOI) if planted late. 125-day maize can yield up to
5.2 t/ha for a rising SOI and 5.3 t/ha for a positive SOI. For a neutral SOI phase, the best possible yields of as much as 5 t/ha can be
obtained when a 125-day maize cultivar is planted early (29 October). If the 140-day and 100-day cultivars are planted on median
(16 November) and late planting dates, respectively, the best yields
attainable for the neutral SOI are 4 t/ha. The 125-day maize cultivar
also attains highest yields for a negative SOI phase if planted on the
median planting date (16 November). The 100-day maize cultivar
is best planted late (7 December) for all the SOI phases if maximum
yields are to be obtained. The 100-day cultivar is the most productive given a falling SOI phase. The tool allows farmers to adjust
management decisions in terms of cultivar selection and planting
date in response to the likely effect of prevailing SOI conditions on
maize yields.
4.2.5. Recommendations for implementation
Seasonal climate forecasts are routinely released in Zimbabwe
with the expectation that the information will improve crop production management. However, they are not always fully utilized.
This is partly because climate forecasts do not necessarily reach
small holder farmers on time and in the formats particularly relevant and useful to them. Credibility and interpretation challenges
can be overcome through target community education and participatory involvement of the user community as partners in all
stages of interpreting and operationalising the climate information
for decision making. In this way, farmers will feel more in control,
giving them greater condence with the information and choices
they have to make.
In order for this study and others of its nature to be implementable, the relevant constraints have to be surmounted. The
implementation of these results and the decision support tool is
unlikely to succeed without the full involvement of the user communities. We recommend that before such tools are availed to user
communities, they should be tested further with the full involvement of the user community over a long period of time. On-site
trials can be performed through the establishment of climate eld
schools with full participation of farmers and local extension ofcers. This will ensure that all relevant stakeholders acquire the
requisite knowledge to apply such tools and to develop condence
in the usefulness of the tools through years of observation. Ideally, a
consultative process involving all stakeholders should be initiated
prior to any research or trials. The consultative process will ascertain the kind of information to constitute the decision support tool
for each location or region. This can be achieved through interviews,

Focus Group Discussions (FGDs) and farmer eld schools. In an ideal


world, adjusting production management to seasonal climate forecasts based on the participatory involvement of user communities
should stabilize yields and minimize food shortages.
5. Conclusions
The study showed that climate (SOI phases), maize cultivar,
planting date and fertility affect maize yields considerably. It is
therefore important for farmers to put all these agrometeorological factors into consideration as they plan for an upcoming
agricultural season in order to optimize their yields. The study conrmed earlier conclusions of the potential of ENSO in predicting
maize yields in advance of the season. SOI phases are therefore
potentially useful in planning activities for maize production at
Masvingo. Although the simulated maize yields were not validated, they closely resemble average maize yields for Zimbabwe of
48 t/ha (SeedCo, 2005) under good management (optimal fertility)
and 0.42.3 t/ha (FAOSTAT, 2007) under communal fertility levels
(poor fertility). It can therefore be concluded that the integrated
ENSO-AquaCrop platform is capable of simulating maize yields to
a satisfactory degree of accuracy with only monthly climatic data
for the various agrometeorological scenarios.
Although the ENSO-AquaCrop platform was able to simulate
reasonable crop yields using monthly seasonal forecast rainfall data
and mean monthly temperature and ETo data, there is considerable room for improvement if daily or 10-day climate data can
be obtained. The shorter rainfall periods will help to account for
intra-seasonal rainfall variations which are known to affect maize
growth, e.g., dry spells. Later versions of RAINMAN can be used
to generate the daily forecasts for rainfall and temperature. Decision support tools similar to the one produced in this study can
be improved and developed for other locations in Zimbabwe. The
tools applicability can be improved with participatory involvement of farmers so as to tailor the tools according to the specic
needs of farmers and the systems in which they operate. We recommend that if climate forecast based production management
is to be adopted by small holder farmers in Zimbabwe, appropriate market and policy intervention needs to support participatory
involvement of farming communities in developing, interpreting
and operationalising climate based decision support tools. Further
studies may also include other crops which are vital for food security in Zimbabwe.
References
Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop evapotranspiration: guidelines for computing crop water requirements. Irrigation and Drainage Paper
no.56. FAO, Rome, Italy, 300 pp.
Apipattanavis, S., Bert, F., Posdesta, G., Rajagopalan, B., 2010. Linking weather generators and crop models for assessment of climate forecast outcomes. Agric. Forest
Meteorol. 150, 166174.
Baigorria, G.A., Jones, J.W., OBrien, J.J., 2008. Potential predictability of crop yield
using an ensemble climate forecast by a regional circulation model. Agric. Forest
Meteorol. 148, 13531361.
Cane, M.A., Eshel, G., Buckland, R.W., 1994. Forecasting Zimbabwean maize yield
using eastern equatorial Pacic sea surface temperature. Nature 370, 204205.
Chenje, M., Sola, L., Paleczny, D., 1998. The State of Zimbabwes Environment. Ministry of tourism and environment, Harare.
Clewett, J.F., Owens, D.T., Clarkson, N.M., Partridge, I.J., 1992. RAINMAN: using El
and Australias rainfall history for better management today. In: HarnessNino
ing Information for a Smarter Agriculture. Proceedings of Australian Institute of
Agricultural Science, National Conference, Launceston, Tasmania.
Clewett, J.F., 1995. Seasonal Climate Forecasts and Decision Support Seasonal Climate Forecasts and Decision Support Systems for Drought Prone Agriculture: A
Case Study Based on the Development and Application of the RAINMAN Climate
Analysis Software. Agroclim, Queensland.
Food and Agricultural Organisation of the United Nations (FAO) Statistical Databases
& Data-sets (FAOSTAT), 2007. Available at: http://faostat.fao.org/.
Fraisse, C.W., Breuer, N.E., Zierden, D., Bellow, J.G., Paz, J., Cabrera, V.E., Garcia, A.G.Y.,
Ingram, K.T., Hatch, U., Hoogenboom, G., Jones, J.W., OBrien, J.J., 2006. Agclimate:

N. Zinyengere et al. / Agricultural and Forest Meteorology 151 (2011) 17921799


a climate forecast information system for agricultural risk management in the
southeastern USA. Comput. Electron. Agric. 53, 1327.
Hammer, G.L., Holzwoth, D.P., Stone, R., 1996. The value of skill in seasonal climate
forecasting to wheat crop management in a region with high climatic variability.
Aust. J. Agric. Res. 47 (5), 717737.
Hansen, J.W., 2002. Realizing the potential benets of climate prediction to agriculture: issues, approaches, challenges. Agric. Syst. 74, 309330.
Hansen, J.W., Indeje, M., 2004. Linking Dynamic Seasonal Climate Forecasts with
Crop Simulation for Maize Yield Prediction in Semi-arid Kenya. International
Research Institute for Climate Prediction, Palisades, NY.
Heng, L.K., Evett, S.R., Howell, T.A., Hsiao, T.C., 2009. Calibration and testing of FAO
AquaCrop model for maize in several locations. Agron. J. 101, 488498.
Hsiao, T.C., Heng, L.K., Steduto, P., Raes, D., Fereres, E., 2009. AquaCrop model
parameterization and testing for maize. Agron. J. 101, 448459.
Legates, D.R., McCabe, G.J., 1999. Evaluating the use of Goodness of Fit measures in
hydrologic and hydroclimatic model validation. Water Resour. Res. 35, 233241.
Loague, K., Green, R.E., 1991. Statistical and graphical methods for evaluating solute
transport models: overview and application. J. Contam. Hydrol. 7, 5173.
Matarira, S.H., Makadho, J.C., Mukanahana-Sangarwe, M., 2004. In: Zhakata, W.
(Ed.), Vulnerability and Adaptation of Maize Production to Climate Change in
Zimbabwe. Ministry of Environment and Tourism, Harare.
Meza, F.J., Wilks, D.S., Riha, S.J., Stedinger, J.R., 2003. Value of perfect forecasts of
sea surface temperature anomalies for selected rain-fed agricultural locations
of Chile. Agric. Forest Meteorol. 116, 117135.
Meza, F.J. 2007. Use of ENSO driven climatic information for optimum irrigation
under drought conditions: Preliminary assessment based on model results. In
Climate Prediction and Agriculture: Advances and Challenges. J. Hansen and
M.V.K. Sivakumar (Eds.) Springer 306.
Ogallo, L.A., Boulahya, M.S., Keane, T., 2000. Applications of seasonal to interannual climate prediction in agricultural planning and operations. Agric. Forest
Meteorol. 103, 159166.
Ochse, J.J., Soule, M.J., Dijkman, M.J., Wehlburg, C., 1961. Tropical and Subtropical
Agriculture, vols. 1 and 2. McMillan, NewYork, p. 1446.

1799

Patt, A., Gwata, C., 2002. Effective seasonal climate forecast applications: examining
constraints for subsistence farmers in Zimbabwe. Global Environ. Change 12,
185195.
Phillips, J.G., Cane, M.A., Rosenzweig, C., 1997. ENSO, seasonal rainfall patterns,
and simulated maize yield variability in Zimbabwe. Agric. For. Meteorol. 90,
3950.
Phillips, J., Makaudze, E., Unganai, L., 2001. Current and potential use of climate
forecasts for resource-poor farmers in Zimbabwe. Impacts of El Ni*no and Climate Variability on Agriculture, vol. 63. American Society of Agronomy Special
Publication Series, pp. 87100.
Phillips, J.G., Dianne, D., Unganai, L., Chimeli, A., 2002. Implications of farmlevel response to seasonal climate forecasts for aggregate grain production in
Zimbabwe. Agric. Syst. 74, 351369.
Raes, D., Steduto, P., Hsiao, T.C., Fereres, E., 2009. AquaCrop-The FAO crop model for
predicting yield response to water: II. Main algorithms and software description.
Agron. J. 101, 438447.
SeedCo, 2005. SeedCo Agronomy Manual. SeedCo Limited, Harare,
http://www.seedco.co.zw/productmanuals.cfm.
Shin, D.W., Lim, Y.K., Cocke, S., LaRow, T.E., OBrien, J.J., 2006. Assessing Crop Yield
Simulations with Various Seasonal Climate Data. Center for Ocean-Atmospheric
Prediction Studies Florida State University, Tallahasse.
Sithole, A., 2003. Optimal planting dates for rainfed maize and sorghum in agroecological zones of Zimbabwe. Msc Thesis. Department of Physics, University of
Zimbabwe, Harare.
Steduto, P., Hsiao, T.C., Raes, D., Fereres, E., 2009. AquaCrop the FAO crop model to
simulate yield response to water: I. Concepts and underlying principles. Agron.
J. 101, 426437.
Vincent, V., Thomas, R.G., 1960. An Agricultural Survey of Southern Rhodesia Part
1. Agro-ecological Survey. Federation of Rhodesia and Nyasaland, Government
Printers, Salisbury.

You might also like