AN ALTERNATIVE APPROACH TO THE ACTUAL BRAZILIAN MAIZE CROP ZONING

Maize produced under rainfed conditions are highly affected by the inherent annual and intra-seasonal weather variability, especially dry spells, that affect productivity. One of the simplest strategies with virtually no cost to mitigate this problem is the determination of a sowing window. The objectives of this study were to: a) use the results of maize yield simulated with a process-based model to establish sowing windows and, b) compare our results with the current methodology employed by the Brazilian Ministry of Agriculture (MAPA). The CSM-CERES-Maize model was used to simulate scenarios of weekly sowing dates, under rainfed conditions, for selected counties of Minas Gerais State, Brazil. For each sowing date it was determined the yield break by comparing the average yield of the current sowing date with the highest average yield obtained from all sowing dates. The use of a process-based model to simulate crop yield allows for the integration of many factors not considered in the current crop zoning approach used by MAPA. The proposed approach has advantages over the MAPA methodology in that it includes the possibility of determining the expected average yield and its amplitude.


348
Grown in almost every continent, maize economic importance is characterized by the various forms of its use, from grazing to biofuels.Worldwide, Brazil is the third largest maize grower, with the state of Minas Gerais ranking third in production within the country (Agrianual, 2013).
The average maize yield in Brazil is low with high variation within a region.In Minas Gerais, the average maize yield ranges from 6299 kg ha -1 in the Triângulo Mineiro region to 1898 kg ha -1 in the north region (IBGE, 2010).Possible causes of this low yield are related to the low level of technology employed by farmers, including sowing at inadequate time, water stress due to dry spells, and the use of cultivars with low adaptation to the region (Cruz et al., 2009;Forsthofer, 2004).
Minas Gerais agriculture is mainly rainfed.
The rainfall in the state, ranging from 650 mm in the north region to 2100 mm in the south/southwest regions, is related to topography and geography (Santana, 2004;Guimarães et al., 2010).It is expected, therefore, that the main factor affecting maize yield in the state is water.Among all climatic factors, rainfall, temperature and solar radiation, are the most important as they directly affect the production of dry matter and grain (Sans & Guimarães, 2006;Brachtvogel et al., 2009).Water availability and temperature are the most important factors to be considered to characterize the best growing seasons within a region (Wagner et al., 2013).
One strategy to minimize the risk of yield losses due to climatic conditions is sowing at the right time.The success of such approach is tightly related to the planning capability of the farmer as it is highly dependent on several factors associated to weather conditions (Sans & Guimarães, 2006).The Brazilian Ministry of Agriculture, Livestock and Supply (MAPA) provides farmers with an Agricultural Crop Zoning of Climate Risk (ZRC) tool, which is also used by policy makers and risk management agencies.The ZRC-MAPA is prepared annually with the objective of minimizing the risks associated to weather and to allow producers to identify the best sowing window for various crops with varied cropping season and growing on different soil types (Brasil, 2013).
The ZRC-MAPA approach is based only on a soil-water balance approach.Process-based simulation models that simulate growth, development, and yield of crops based on weather, soil, cultivar-specific coefficients, and management practices, are better tools to assist in the determination of sowing windows.These dynamic crop models describe daily changes in the inputs used to simulate the main crop physiological processes (Dallacort et al., 2006).Dynamic crop simulation models are useful tools when the effects on crops depend on complex interactions with soil, weather and with other factors related to the management of agro-ecosystems (Jones et al., 2006).The CERES (Crop Environment Resource Synthesis) consists of group of models developed by the Grassland Soil and Water Research Laboratory (Jones & Kiniry, 1986).Among the CERES models, CERES-Maize was developed for the maize crop and allows simulations of the growth and development of maize, water balance, N levels and also enables economic evaluations based on four input variables: soil, climate, crop management and genotypes (Soler, 2000).et al., 1998;Jones et al., 2003).The CSM-CERES-Maize is part of the Decision Support System for Agrotechnology Transfer, DSSAT (Jones et al., 2003;Hoogenboom et al., 2013), a software that includes models for 28 different crops.The cropping system simulation model (CSM) included in DSSAT v4.5 simulates growth and development of maize at a daily time step, from planting to maturity.Several studies have demonstrated the applicability of DSSAT as a tool to help making management decisions.For example, a study conducted by Singh & Srinivas (2007) in India aimed at determining best planting dates for irrigated and rainfed maize in a succession system with chickpeas.Soler et al. (2007) assessed the effect of different sowing dates for maize cultivars of different cycles growing on irrigated and rainfed fields for conditions in Manduri and Piracicaba counties, state of São Paulo, Brazil.
The objectives of this study were to use a simulation approach to determine best sowing windows for maize production and to compare those results with the current methodology employed by the ZRC-MAPA.

Material and Methods
The CSM-CERES-Maize model of DSSAT, version 4.5.1.013(Hoogenboom et al., 2013)  The genetic coefficients of the single-cross hybrid BRS1030 were previously determined from observed data obtained from maize field trials conducted under optimum conditions of growth and development (Santana et al., 2010).The adjusted coefficients were 263.8 for P1; 0.5 for P2; 1034 for P5; 648 for G2; 5.14 for G3; 44.22 for PHINT.
Because of the lack of specific soil information in the areas of interest (selected counties), the same approach used by the current ZRC-MAPA was used (Sans et al., 2001).The ZRC-MAPA criteria consists of a maize crop with a root system 0.50 m deep and three generic soil types with low (20 mm), medium (40 mm) and high (60 mm) water retention capacity.The values of the lower and upper limits of available water of each layer of existing DSSAT´s soil profiles data were adjusted to become similar, in terms of water retention, to the three soil types used in the ZRC-MAPA (Table 1).
The management conditions used in the simulations was obtained from the Embrapa Maize and Sorghum online recommendations for maize production (Cruz, 2009)   The seasonal analysis tool of DSSAT, which simulates yield for each sowing week and each year, was used.Additionally, the water and nutrient balance routine from the model was turned off to allow the simulation of maize potential yield, in which the crop is grown without any biotic or abiotic stresses.The model was set to perform weekly sowings, beginning on August 01 and extending for 52 weeks to July 24, generating 49 simulation scenarios at each county.
The highest average simulated yield was identified among all sowing dates and a yield break was calculated for each county according to (Amaral et al., 2009): where Y b is the yield break in percentage; Y w is the average yield of week "w" in kg ha -1 ; and Y max is the maximum average yield among all weeks in kg ha -1 .
An analysis of the yield break curve was performed to determine the sowing windows for different levels of risk that the decision maker would be willing to take.We arbitrarily assumed a risk level of 10%; i.e., the decision maker would tolerate a yield reduction of up to 10%.In the example showed in Figure 2, the sowing windows would be from October 24 to December 19, October 17 to December 26 and from October 10 to December 26, for soil type 1, type 2 and type 3, respectively.
The criteria used to determine whether a county was suitable or not for maize production was based on the relationship: where Y pr is the yield break relative to the potential yield in percentage; Y str is the average rainfed yield, for the sowing window, in kg ha -1 and Y pot is the average potential yield, for the same sowing window, in kg ha -1 .A county was considered suitable for maize production when Ypr was 60% or less.This procedure was performed for the three soil types in all counties.
The simulated yields from the 49 counties and for the three different soils were then interpolated by using a kriging procedure.The software gvSIG 1.11 (gvSIG, 2013) and Quantum GIS 1.9 (Quantum, 2013) were used to create the yield maps.
The yield break curve approach allowed determining the suitability for maize production and the establishment of a sowing window for each selected county as demonstrated for Uberaba, Minas Gerais, Brazil (Figure 2

Results and Discussion
Our results showed large maize yield variation as a consequence of the climatic diversity in the state of Minas Gerais, characterized by humid climate in the south to semi-arid climate in the north and northeastern regions (Figure 3).
The average yield of maize, for the sowing window, ranged from 750 kg ha -1 , in Espinosa, to 10,523 kg ha -1 , in Maria da Fé.Regardless of the soil-water retention capacity, the regions North, Jequitinhonha and Vale do Rio Doce, which are drier and warmer, presented the lowest yield, while the south region, which is wetter and has milder temperatures, showed the highest.Yields below 1,000 kg ha -1 were considerably larger in the regions North, Jequitinhonha, Vale do Mucuri and Vale do Rio Doce, especially in soils with poor water holding capacity (soil type 1).For soils with medium or high water holding capacity (soils type 2 and type 3), the amplitude of the simulated yield was higher due to better crop response to nitrogen fertilization in years with more favorable climatic conditions (Figure 3).This yield variability observed in our results is the consequence of the interactions between crop, soil and climatic conditions, especially water.Crop productivity is the result of the interaction of various factors, especially those related to soil attributes.However, Bergamaschi et al. (2004), indicate that water is the main factor that affects maize yield worldwide.
There is a close relationship between the average simulated maize yield and elevation and latitude of the sites (Figure 4).The higher the latitude and altitude, the greater the simulated maize yields.This is because for conditions in Minas Gerais, rainfall and air temperature are related to both, elevation and latitude (Guimarães et al., 2010;Santana, 2004).
When comparing the average simulated yield of the sowing window, for soil type two, with the average maize yield estimated by IBGE, for the 2003 to 2011 period and for the same counties, considerable differences can be observed (Figure 5).These gaps indicate that the average maize  2).The soil water retention capacity had an important role in defining the start and length of the sowing window, especially in cities with low rainfall.
Shorter sowing windows were also associated with lower yields (Figure 3 and Table 2).It is possible that the sowing date used by farmers is not the most appropriate for each site, contributing to lower than expected yields indicated by the estimates made by IBGE (Figure 5).Sowing out of the appropriate period is the major cause of low maize yield in Brazil (Forsthofer, 2004).
The methodology used to define the suitability of a certain county for dryland maize production proved consistent (Table 2).Most of the counties that were considered unsuitable for maize production are located in the North and Jequitinhonha regions, which receive lower amounts of precipitation and, in some cases, are associated with high temperatures.Obviously, the smaller the water holding capacity of the soils, the greater the number of counties considered unsuitable for rainfed maize production.
When comparing the sowing window and the suitability of the county for maize production (Tables 2 and 3), it can be observed that the proposed methodology is more restrictive than the ZRC-MAPA approach.According to the proposed methodology, and regardless of soil type, the counties of Espinoza, Janauba and Pedra Azul are not suitable for rainfed maize production, while when using the ZRC-MAPA approach those same counties are only suitable for maize production if the soil is of type three.If the soil has low water holding capacity (soil type 1), 34 counties are considered non suitable for rainfed maize production, against nine counties considered non suitable by the ZRC-MAPA approach.For soils with average water holding capacity (soil type 2), nine counties are considered non suitable by the proposed method while all counties are considered suitable by the ZRC-MAPA approach.For the ZRC-MAPA approach (Table 3), the beginning of the sowing window in all counties and for the three soil types, is October 01, while by the proposed methodology (Table 2), the starting of the sowing window ranged from September 19, in Passa Quatro and Maria da Fé, to November 21, in Formoso.
Additionally, by the proposed methodology, the duration of the sowing window varied from eight days in Juramento, Januaria and Carbonita to 107 days, in Juiz de Fora.
For the ZRC-MAPA approach the expected duration of the sowing window is 31, 20, 20 and 92 days, respectively for Juramento, Januaria, Carbonita and Juiz de Fora.In general, in drier regions the proposed methodology was more restrictive than the ZRC-MAPA approach, establishing shorter sowing windows.
The discrepancy in the results obtained with the proposed methodology and the ZRC-MAPA is due to the use of different approaches to determine the sowing window and the suitability of the different counties for maize production.The ZRC-MAPA is based solely on crop water requirements satisfaction, determined through a simple soil water balance (Sans et al., 2001).

Conclusões
The use of process-based simulation model allowed the establishment of sowing windows as well as the determination of the expected average maize yield for various counties in the state of Minas Gerais, Brazil.
Our results indicate that there is a potential to increase rainfed maize yield for conditions in the state of Minas Gerais.
The proposed methodology turned out to be more restrictive than the approach currently used by the ZRC-MAPA.
The study highlights the needs for better monitoring of environmental conditions in the state of Minas Gerais, a necessary step to enable the widespread application of the proposed methodology.
The cropping season of maize in the CERES-Maize is divided into various phases (germination, emergence, end of juvenile phase, floral induction, silking, beginning of grain filling, and harvest maturity), while development is influenced by the thermal sum or thermal time, expressed in degree-days (DD), which is calculated based on the minimum and maximum daily temperatures.The thermal time required to progress from one stage of development to another is a user input and can be defined as: P1 -Thermal time from seedling emergence to the end of the juvenile phase (expressed in DD above a base temperature of 8 o C), during which the plant is not responsive to changes in photoperiod; P2 -Extent to which development (expressed in days) is delayed for each hour increase in photoperiod above the longest photoperiod at which development proceeds at a maximum rate (which is considered to be 12.5 hours); P5 -Thermal time from silking to physiological maturity (expressed in DD above a base temperature of 8 o C); G2 -Maximum possible number of kernels per plant; G3 -Kernel filling rate during the linear grain filling stage and under optimum conditions (mg day -1 ); PHINT -Phylochron interval, the interval in DD between successive leaf tip appearances (Ritchie . The model was set for 0.90 m row spacing and a plant population of 68,000 plants ha -1 .Fertilization consisted of 500 kg ha -1 of 8-28-16 + Zn applied at sowing and 200 kg ha -1 of nitrogen as urea + 70 kg ha -1 of K 2 O in, as potassium chloride, side-dressed at 40 days after sowing (DAS).

FIGURE 1 .
FIGURE 1. Location of weather stations of the state of Minas Gerais, Brazil, used in the study.
).Our results were compared with the suitability and sowing windows established by the ZRC-MAPA.Additionally, the average expected maize yield was determined for each soil type and county.Results for soil type 2, the most common in Brazil, were compared to the average estimated yield provided by the Brazilian Institute of Geography and Statistics (IBGE), for the period 2003 to 2011.

FIGURE 2 .FIGURE 4 .FIGURE 5 .
FIGURE 2. Maize yield break for different sowing dates in a rainfed production system for conditions in Uberaba, Minas Gerais, Brazil.

TABLE 1 .
Attributes of the three generic soil profiles.

TABLE 3 .
Sowing window and number of days of the sowing window for different counties of Minas Gerais state, Brazil, as established by the ZRC-MAPA.