SciELO - Scientific Electronic Library Online

 
vol.40 número2Managing water pressure for water savings in developing countriesA semi-quantitative survey of macroinvertebrates at selected sites to evaluate the ecosystem health of the Olifants River índice de autoresíndice de materiabúsqueda de artículos
Home Pagelista alfabética de revistas  

Servicios Personalizados

Articulo

Indicadores

Links relacionados

  • En proceso de indezaciónCitado por Google
  • En proceso de indezaciónSimilares en Google

Compartir


Water SA

versión On-line ISSN 1816-7950
versión impresa ISSN 0378-4738

Water SA vol.40 no.2 Pretoria abr. 2014

 

Use of the FAO AquaCrop model in developing sowing guidelines for rainfed maize in Zimbabwe

 

 

Teddious MhizhaI; Sam GeertsII; Eline VanuytrechtII; Amos MakarauIII; Dirk RaesII

IDepartment of Physics, University of Zimbabwe, Mount Pleasant Drive, Harare, Zimbabwe
IIDivision of Soil and Water Management, K.U. Leuven University, Celestijnenlaan 200E, 3001 Leuven, Belgium
IIIMeteorological Services Department, P.O. Box BE150, Belvedere, Harare, Zimbabwe

Correspondence

 

 


ABSTRACT

This paper presents a procedure in which the water-driven water productivity model AquaCrop was fine-tuned and validated for maize for the local conditions in Zimbabwe and then applied to develop sowing management options for decision support. Data from experiments of 2 seasons in Harare and from 5 other sites around Zimbabwe were used for the local calibration and validation of AquaCrop. Model parameters such as the reference harvest index (HIo); the canopy growth coefficient (CGC); early canopy decline and normalised biomass water productivity (WPb*) were adjusted during model calibration. Model performance was satisfactory after calibration with a Nash-Sutcliffe model efficiency parameter (EF = 0.81), RMSE = 15% and R2 = 0.86 upon validation. To develop sowing guidelines, historical climate series from 13 meteorological stations around Zimbabwe were used to simulate maize yield for 6 consecutive sowing dates determined according to criteria applicable in Zimbabwe. Three varieties and typical shallow and deep soil types were considered in the simulation scenarios. The simulated yield was analysed by an optimisation procedure to select the optimum sowing time that maximised long-term mean yield. Results showed that highest yields depended on the climate of the site (rainfall availability), variety (length of growing cycle) and soil depth (soil water storage capacity). The late variety gave higher mean yields for all sowing dates in the maize belt. Staggered sowing is recommended as a way of combating the effects of rainfall variability and as an answer to labour constraints.

Keywords: biomass water productivity, AquaCrop, maize sowing dates, crop modelling


 

 

INTRODUCTION

The global population is projected to continue on an upward trend (FAO, 1996; Mpande and Tawanda, 1998), more so in sub-Saharan Africa where food deficit is already a significant challenge (Pinstrup-Andersen et al., 1999). Competing demands for both freshwater and land use, such as from industry and municipalities, as well as environmental problems such as pollution, will limit future extension of both freshwater for irrigation and the cultivated land area. With limited room for expansion of both agricultural land and the irrigated portion of the arable land (Rockström and Baron, 2007), additional food production will have to come from intensification of production in rainfed farming systems. Rockström et al. (2003) showed that it is possible to at least double rainfed staple food production by producing more 'crop per drop' of rainwater. It is therefore necessary to explore ways of increasing water use efficiency in rainfed agricultural systems.

Climate variability has been identified as the major constraint to agricultural productivity in southern Africa, and hence reducing the risk associated with climate variability has a high potential for increasing productivity in Zimbabwe (Phillips et al., 1998). Despite commanding a large share of the annual grain output, rainfed production of maize in Zimbabwe is largely unstable (Mhizha, 2010). The fluctuations echo in the availability of food in the country, often with a telling effect on the economy as resources are channelled towards securing food to avert starvation, resources which would have otherwise gone to other economic sectors for development. The instability in rainfed production is largely credited to availability of rainwater, which itself shows wide variability in both total amounts and seasonal quality (Rockström and Barron, 2007). Rainfall variability, especially the less well defined onset of the rainy season has increased in the recent past possibly linked to climate change. The start and end of the rainy season defines the length of the rainy season which strongly determines the success or failure of rainfed crops. In addition, the quality of the growing season, as indicated by the length and severity of within-season dry spells, will also influence the yield gap and can often cause total crop failure (Geerts et al., 2006; Phillips et al., 1998). While agricultural water management has largely succeeded in maximising rainfall infiltration through soil and water conservation, the challenge of how to cope with dry-spells, short periods of water stress during crop growth, remains largely unsolved (Fox and Rockström, 2003). Because false planting dates requiring replanting are increasingly common in Zimbabwe (Raes et al., 2004), there is an increasing demand for sowing strategies that minimise risk of total crop failure, such as staggered planting.

Judicious management decision making, such as planting dates and fertiliser application rates, can contribute to increased yields under rainfed conditions. Management decision support for rainfed farming systems is a challenge for resource-poor communities such as subsistence farmers in Sub-Saharan Africa. Optimum management practices, such as planting date, cultivar selection, fertilisation, or water and pesticide application, can be assessed through validated models for making seasonal or within-season decisions (Boote et al., 1996). Simulation experiments can be of significant use in exploring different management options for decision support in cases where field experiments are scarce. Accurate modelling of crop response to water plays an important role in development of guidelines for improving water use efficiency in agriculture (Geerts et al., 2009a). There are many models that simulate the growth and development of maize, such as CERES-Maize (Jones et al., 1987) and Hybrid-Maize (Yang et al., 2004), but most are often applicable only to the fields for which they are calibrated and require a number of parameters next to impossible to collect in rainfed field conditions. AquaCrop, on the other hand, although based on complex crop physiological processes, uses a relatively small number of explicit and mostly intuitive parameters and attempts to balance simplicity, accuracy and robustness (Steduto et al., 2009; Raes et al., 2009a). AquaCrop can in this regard be considered suitable for application in resource-challenged communities where extensive input data may not be available.

AquaCrop is a crop water productivity model (Steduto et al., 2009; Raes et al., 2009a) broadly tested for simulating maize yield response to water (Hsiao et al., 2009; Heng et al., 2009). The model was validated for a wide range of environmental conditions, namely, extraordinarily high evapotran-spiration and wind speed in Bushland, Texas, rainy weather and sandy soil in Gainesville, Florida, and semiarid conditions in Zaragoza, Spain (Heng et al., 2009). Many papers have reported the application of AquaCrop in simulating various management scenarios for many crops including maize (Hsiao et al., 2009, Heng et al., 2009, Stricevic et al., 2011), quinoa (Geerts et al., 2009a), cotton (Garcia-Vila et al., 2009), sunflower (Todorovic et al., 2009, Stricevic et al., 2011) and sugar beet (Stricevic et al., 2011) with success. Against such background, Aquacrop is expected to be potentially suitable for simulating maize yield response to water availability in the semi-arid conditions of Zimbabwe, although reported studies are lacking. To account for the unpredictability of rainfall in the season, farmers often aim at minimising risk, which often means settling for low inputs and low but stable yields (Phillips et al., 1998). One possible strategy is to sow several varieties of a single crop on several planting dates at the start of the rainy season. The varieties differ in their length of the growing cycle and their potential yield. Depending on the length of the rainy season, the total amount of rainfall received during the season and the frequency, length and period of dry spells, it is expected that at least one variety on one of the planting dates will give good yields. As such the farmer is guaranteed an income each year. However the guidelines for this sowing strategy have not been clearly laid down in literature for Zimbabwe's agro-ecological zones. Also the sowing strategy needs to be evaluated for effectiveness in reducing risk of crop failure.

Raes et al. (2004) evaluated the performance of 3 criteria that can be used to determine first planting dates for maize in Zimbabwe. The study concluded that the 3 criteria had different failure rates of 1 in 2 years, 2 in 5 years and 1 in 4 years, depending on the severity of the criterion. The analysis was limited to the establishment stage (first 30 days after sowing) and the rest of the growing season was not assessed. The aim of this paper is to apply AquaCrop to develop decision support guidelines for sowing maize under rainfed conditions for the semi-arid tropical climate of Zimbabwe. A combination of field trials and model simulation experiments were used in this study to analyse response of maize yield to long-term variation in rainfall at climate stations around Zimbabwe. The results are then applied to guide a sowing strategy that minimises year to year variation in rainfed maize yields of smallholder farmers. Field experiments at a research station in Harare and 5 other locations in areas around Zimbabwe where maize is commonly grown provided data for model evaluation. The first part of the paper presents the local calibration of AquaCrop for maize by adjusting for a lower soil fertility level. The second part presents the application of the validated AquaCrop to simulate maize yield response to a long climate data series in order to develop optimised staggered sowing guidelines for rainfed maize in Zimbabwe.

 

MATERIALS AND METHODS

Study area

The study was conducted in the so-called maize belt of Zimbabwe and its surroundings (Fig. 1). Vincent and Thomas (1960) divided Zimbabwe into 5 main natural regions primarily on the basis of rainfall, but also considering other factors such as soil type, altitude and land use. Rainfall patterns and crop production progressively deteriorate from Region I to V. Annual mean rainfall is highest in Natural Region I which covers approximately 2% of the land area. It is a specialised and diversified farming region with plantation forestry, fruit and intensive livestock production. Natural Region II, covering 15% of the land area, receives lower rainfall than Region I, but is nevertheless suitable for intensive farming based on crop or livestock production. Rainfed maize production has the highest potential in Regions IIa and IIb (Fig. 1) (Eicher, 1995; Burt et al., 2001; Philips et al., 2002) because rainfall in Regions III to V is too low and erratic for the reliable production of rainfed maize. Natural Region II is traditionally referred to as the maize belt of Zimbabwe.

 

 

Field trials and observations

The data for calibration and validation of AquaCrop were obtained from field experiments conducted at 6 locations in 5 seasons (Table 1).

 

 

Intensive measurements for model calibration data collection were carried out at Thornpark (Table 1) during the 2006 to 2007 and 2007 to 2008 farming seasons, while the other sites were extensively monitored to give datasets for the validation process (Table 2). At Thornpark, land preparation each year was by ploughing and harrowing in October before the first rains. All sowing was wet sowing after rainfall, except in 2006 when supplementary irrigation was applied at sowing to allow for very early sowing before the first rains. All plots measured 10 x 10 m and plants were spaced 90 cm between rows and 30 cm in row. One seed was sown per sowing station giving a target plant density of about 37 000 plants/ha. At sowing, a basal fertiliser of compound D (nitrogen; phosphorus; potassium, (NPK): 7%; 16%; 5%, respectively) was applied at a rate of 300 kg/ha. Basal dressing was applied in the sowing holes and covered together with the seed. Top dressing was applied at 5 weeks using ammonium nitrate (34.5% N) granules at a rate of 300 kg/ha (about 104 kg/ha N). This application rate, although it is recommended for the rainfed conditions in Zimbabwe, is rather low when compared to the optimal rate of 200 kg/ha N recommended by FAO (2010). Spot application without covering with soil was used for the top dressing. Management at the other sites is considered similar to Thornpark although no detailed records are available.

 

 

The parameters observed for the trials at Thornpark farm included canopy cover (CC), soil water content (SWC), maximum effective rooting depth (Zr) above ground biomass (B) and grain yield (Y). Weather data, including daily rainfall, were observed on site by means of an automated weather station. Data collected at the other locations consisted of grain yield, daily rainfall and reference evapotranspiration (ETo) estimated from weather data of the weather station nearest the trial site via the FAO Penman-Monteith equation (Allen et al., 1998).

Canopy cover (CC) was estimated using the meter-stick method (Armbrust, 1990), in which the proportion of the ground shaded by the crop canopy under clear skies and within 2 hours of solar noon (Local time = GMT + 2 h) is expressed as CC in percentage. The restriction of time was imposed to reduce bias caused by effects of the solar elevation angle on the size of the shaded area.

Soil water content (SWC) was measured fortnightly at 10 cm depth intervals up to 1 m gravimetrically. Bulk densities were determined by using an undisturbed soil sampling kit consisting an auger and cylindrical cores of known volume (100 cm3), then weighing the soil after oven drying at 105°C for 24 h. Observed root zone water content was calculated by considering measured SWC for the soil depth equal to the effective rooting depth. The estimated maximum effective rooting depth (Zr ) was used for this purpose.

Maximum effective rooting depth (Zr) was estimated by visual inspection of 1.5 m deep pits dug in the plots to expose the roots at physiological maturity. Washing of the profile with water facilitated clarity in identifying the roots and the lowest level where roots of the maize crop could be observed was considered the maximum effective rooting depth. Soil water retention characteristics were derived from a soil baseline study at the Thornpark site and literature (Mhizha, 2010).

Above-ground biomass (B) samples were collected by cutting the maize at a stubble height of 5 cm and oven-drying at 80°C for 48 h. The sampling was at 2-weekly intervals from approximately the end of the establishment phase of the crop (30 days after sowing) to biological maturity.

At harvest, the grain yield (Y) was weighed and its moisture content measured by a crop moisture meter. The fresh weight was standardised by calculating the equivalent mass at standard moisture content of 12.5% using Eq. (1).

where:

mstd is the grain mass (kg) at 12.5% moisture content

m is the measured mass of grain (kg) at M% moisture content wet basis at harvest.

The harvest index was calculated as the ratio of standardised grain yield to the dry above-ground biomass at harvest.

The crop characteristic variables monitored and the sites of their observation are described in Table 2. The data fields for calibration and validation are identified in this table.

AquaCrop

For detailed description of AquaCrop parameterisation refer to Raes et al. (2009a), Raes et al. (2009b) and Steduto et al. (2009). The AquaCrop input parameters for maize were reported by Hsiao et al. (2009) and validated by Heng et al. (2009) to be either conservative or cultivar specific. Conservative crop parameters are considered constant for all maize cultivars (Hsiao et al., 2009; Heng et al., 2009) while cultivar-specific parameters, on the other hand, may need fine tuning to be applicable to specific local cultivar characteristics. The conservative crop parameters describe the crop development, transpiration, biomass accumulation and grain yield production for optimal environmental conditions. These processes are modified when water stress exists, in which case the modelled processes are adjusted in proportion with the level of stress through the various water stress coefficients (Raes et al., 2009b).

In AquaCrop the effect of soil fertility needs to be calibrated by means of a set of soil fertility stress coefficients (Raes et al., 2009b). The calibration corrects for the effect of soil fertility stress on: (i) canopy development (CGC), (ii) rate of canopy decline once maximum canopy cover (CCx) is reached (early canopy decline), (iii) biomass water productivity (WPb) and (iv) reference harvest index (HIo). As the fertilisation level was below optimum, some of the conservative crop parameters (Table 3) were adjusted for the local 'near optimal' soil fertility levels of the field trials. The AquaCrop model parameters calibrated for this study are presented in Table 3.

 

 

The calibration procedure followed (Mhizha, 2010) involved adjusting model parameters for canopy development (CGC), early canopy decline, biomass accumulation (WPb) and grain yield production (HIo) using measured data on canopy cover (CC), biomass and grain yield respectively. The procedure consisted in using specific observed variables as the reference variables (Table 4) in the calibration and adjusting only those parameters (degrees of freedom) that are known to influence the reference variables the most.

 

 

The match between simulated and observed reference variables (Table 4) was assessed using goodness of fit tests comprising: the Nash-Sutcliffe model efficiency coefficient (EF) (Nash and Sutcliffe, 1970; Wglarczyk, 1998; Krause et al., 2005) that was evaluated to assess the predictive power of the model; the root mean square error (RMSE) and the coefficient of determination (R2) (Loague and Green, 1991) that were evaluated to assess the error in the model estimates and the correlation between modelled and observed variables, respectively. RMSE was minimised while EF and R2 were maximised in the termination criteria of the calibration. In the validation process, separate data fields (Table 2) were used as observed reference variables in similar goodness of fit tests without any further changes to the calibrated parameters.

Development of sowing guidelines

This section describes how AquaCrop was applied to simulate the response of maize yield to scenarios of sowing date and variety over a long series of climate data at 13 stations in Zimbabwe. The yield response data were used to develop guidelines for sowing maize under rainfed conditions in the study areas.

Climate data

Historical climate data for the study area (Table 5) were used as the input for the climate environment in AquaCrop. Each climate file comprised of daily rainfall data, daily (or monthly) maximum and minimum air temperatures and daily (or monthly) reference evapotranspiration (ET0) data. The default CO2 file within AquaCrop was used to adjust the normalised biomass water productivity (WPb*) to the CO2 concentration of the simulated year (Raes et al., 2009a).

Crop and soil characteristics

In addition to climate input, 3 varieties differing only in growth cycle length (Table 6), with the rest of the crop parameters as calibrated in the previous section, were considered for each site. The soil type at each site (Table 5) was considered for the respective sites. Typical shallow (0.6 m) and deep (1.2 m) soil depths were considered making 2 soil files for each site. Overall there were 3 crop files and 2 soil files for each climate site.