Dry bean crop productivity simulation for soil and climatic conditions of Tangará da Serra, MT-Brazil

The Brazilian national growing of dry bean (Phaseolus vulgaris) currently comes in three annual harvests, which are the wet season, sowing between October and December, the Dry Season, sowing between February and May, and finally the Winter Season, sowing in June to August. The objective of this study was to determine the optimal sowing date for each of the three different sowing seasons wet, dry and winter of dry bean to the Tangará da Serrá region using a crop simulation software called Decision Support System for Agrotechnology Transfer (DSSAT). The DSSAT is comprised of crop simulation models, in which the CROPGRO-Drybean model was used to simulate the dry bean growth, development, and yield. The model was calibrated using the dry bean cultivar ‘BRS Esplendor’, planted on 15 December 2011 in Tangará da Serra, located in the Mato Grosso state of Brazil. The weather variables (maximum and minimum temperature, solar radiation and precipitation), phenological and soil variables were recorded during the season and used in the model calibration to ensure a satisfactory simulation. Following the calibration, simulations were performed for six sowing dates in each of the three seasons. Of the three growing seasons simulated, the wet season had the best grain yields for the dry bean ‘BRS Esplendor’, the sowing date of December 1 st had the highest yields of 3.3 t ha -1 . The dry season had the second high simulated yields, and the highest yield into this growing season was 3.0 t ha -1 . In the dry season, grain yield decreased as late sowing date occurred, and the lowest simulated yield was 0.1 t ha -1 . Finally, the winter season had the lowest simulated yields among the three growing season, with a maximum yield of 0.5 t ha -1 . The CROPGRO-Drybean model had a high sensitivity to rainfall events, and drought periods during the reproductive stage of dry bean was the weather parameter that most affected grain yield. The winter season had lower yields than the wet and dry season in consequence of low rainfall events during the simulated crop cycles, the soil moisture was highly affected by precipitation, which directly affected the leaf area index and crop yield in all sowing dates.


INTRODUCTION
In Brazil, the dry bean (Phaseolus vulgaris) is cultivated by small and big farmers around the country, the leguminosae is considerate a subsistence crop that requires low technological development as well as the use of low seeds quality (Leite et al., 2009).
The Brazilian national growing of dry bean currently comes in three annual harvests, which are wet season, sowing between October and December, the dry Season, sowing between February and May, and finally the winter season, sowing in June to August (Vieira et al., 1995).
The Mato Grosso state stands out in 2014, as the third highest national yield of dry beans, with 23.4, 82.5 and 36.2 thousands of tons for the wet season, dry season and winter season, respectively. However, the crop cultivation is undergoing by a significant variation in relation to cultivated area, consequences of weather characteristics and market behavior (CONAB, 2014).
Weather characteristics such as temperature and solar radiation are important environmental factors, which affect the crop development. The dry bean has an ideal temperature of 21°C, which determine the crop development (Fancelli, 2009). While the solar radiation lead processes photosynthetic and photo-morphogenetic (Kunz et al., 2007). On the other hand, the stress hydric in non-irrigated agricultural areas is the main factor that has been affecting the dry bean. The high temporal variability may cause an excess or deficit hydric, and in both cases, it is directly influencing the dry bean development and production (Dallacort et al., 2011b).
Current weather variability has led growers and researchers to make decisions on best management practices based on simulation techniques. In this context, the Decision Support System for Agrotechnology Transfer (DSSAT) is increasingly used, and the CROPGRO-Drybean model (Hoogenboom et al., 1994), which is one of several crop development models present at the DSSAT, have been used extensively to evaluate effects of irrigation requirements (Heinemann and Hoogenboom, 2000), sowing dates (Dallacort et al., 2005;Lima Filho, Coelho Filho and Heinemann, 2013) and yield simulation (Dallacort et al., 2011a;Oliveira et al., 2012;Meireles et al., 2002Meireles et al., , 2003. The difference between the daily water uptake by plants and the crop transpiration is the factor that most penalize crop development and yield in the CROPGRO-Drybean. The water stress will affect the crop development through two different physiological factors, the photosynthesis, a less sensitive factor, and the cell elongation, which is highly affected by drought (Hoogenboom et al., 1994). Dallacort et al. (2005) reported the use of the CROPGRO-Drybean model to determine the optimum planting dates for south of Brazil. The model strongly penalized the grain yield when the crop was submitted to water stress. Authors concluded that accumulative rainfall had a direct influence on leaf area index (LAI), biomass dry weight and yield.
The dry bean aptitude to Mato Grosso state (Marco et al. 2012) and the performance of the CROPGRO-Drybean model to simulate water stress factors (Heinemann and Hoogenboom, 2000;Dallacort et al., da Silva et al. 5101 2005) are already known. However, there is a lack of information on the influence of weather patterns on dry bean planting dates for the Mato Grosso state, therefore, the necessity of crop development stages and yield predictions to the region, which will help growers to better strategy best management practices, is required. The objective of this study was to determine the optimal planting date for each of the three different sowing seasons wet, dry and winter of dry bean to the Tangará da Serrá region, located in the Mato Grosso state, using the CROPGRO-Drybean simulation model.

Model characteristics
The CROPGRO-Dry bean is a cropping system model from the Decision Support System for Agro-technology Transfer (DSSAT) (Jones et al., 2003). The model was developed by (Hoogenboom et al., 1994) and it simulates the common bean crop growth, development, and yield, as well as weather, genotype and soil properties (Meireles et al., 2003). The minimum data set required to run the model are the plant genetic coefficients, soil characteristics, weather data and crop management data. The genetic coefficients are comprised by three files: .ECO, which characterize the ecotype, genetic coefficients that differ cultivars of determinate and indeterminate growth, .SPE, which characterize the species, genetic coefficients that determine the photosynthesis, nitrogen uptake capacity, phenology, growth, and senescence, and finally that file. CUL, which characterize the cultivar, such as photoperiod, photosynthetic rate, leaf area index (LAI), grain mass, trefoil maximum area, mean of grain per pod, period between emergence and first flower, first flower and first pod, first flower and first grain, first grain and maturation and first flower and end of leaves expansion.
The CROPGRO-Drybean uses physical soil characteristics as field capacity, permanent wilting point and saturation to calculate the soil water balance for the soil layer based on the water from irrigation, precipitation, and drainage. Furthermore, the model estimates the soil water evaporation (Es), plant transpiration (Ep) and crop evapotranspiration (ETc) in mm day -1 , using the orientated-model of the soil water balance developed by Ritchie (1985).
The weather data required is maximum and minimum temperature, rainfall and solar radiation, which are stored in two files: Station, WTH and station CLI. While the soil characteristics data, such as chemical analysis and physical-hydric analysis stored in the file SOIL.SOL. Finally, the crop management data for fertilizer applications, irrigation events, tillage and sowing dates are separately stored in the X file.

Experimental procedures
The field trial was carried out during 2011/2012 dry bean season in the experimental field of UNEMAT (Mato Grosso State University) at Tangará da Serrá, Mato Grosso state, latitude 14º 39' 55'' S, longitude 57º 25' 05'' W and altitude of 321.5 m. The research area has the soil classified as Oxisol, with 1 % of slope. The soil has a *Corresponding author: E-mail: biscaia@ufl.edu.
Author(s) agree that this article remain permanently open access under the terms of the Creative Commons Attribution License 4.0 International License Table 1. Soil chemical analyses for 0-30 cm soil depth at the experimental area.  clay texture, with the proportion of particle fractions of 67%, 7 and 26% of clay, silt and sand, respectively. The soil physical parameters, such as field capacity (0.301 cm 3 cm -3 ), permanent wilting point (0.239 cm 3 cm -3 ) and bulk density (1.09 g.cm -3 ), and the soil chemical analyses (Table 1) were determined by soil samples.

Soil depth (cm) pH ---------------------cmolc/dm³ ---------------------
The studied region has two well-defined weather seasons, which is a dry season from May to September and a wet season October to April (Dallacort et al., 2011b). The minimum weather data required by the model was collect in a meteorological station from UNEMAT located in site, and six years data set, from 2005 to 2010 used to model simulation. The 2011 data set was used to model calibration. During the experimental periods and simulations, rainfall events were the only source of water ( Figure 1) for all dry bean harvest seasons (wet, dry and winter season).
The experimental purpose was to collect data to calibrate the CROPGRO-Drybean model; therefore, the crop management practices followed growers activities and local crop recommendation. No irrigation events occurred at any period of crop development. The dry bean seeds (var. 'BRS Esplendor') were planted on December 15 th , 2011. Seeds were planted in-row spacing of 0.45 m and 12 plants per linear meter. The total experimental area was four replications of 6 m of length by 6 dry bean rows.
The fertilizer application was split into three applications: Initially, 240 kg ha -1 of the N, P, K formulated 5-25-15 was applied before planting at 20 cm of soil depth. The second application of 50 kg ha -1 of urea occurs 15 days after plant emergence, and the last application was 28 days after plant emergence with 50 kg ha -1 of urea. The grain harvest occurs in the four center rows of each replication, the average yield used to model calibration.
The data of genetic coefficients (Table 2) required by the model were collected through frequent field inspections. The genetic coefficients collected were critical day length (CSDL), response inclination regarding development for the photophase with time (1/h) (PPSEN), period between plant emergence and the appearance of the first flower in photothermal days (EM-FL), period between the appearance of the first flower and the first pod in photothermal days (FL-SH), period between the appearance of the first flower and the start of seed formation in photothermal days (FL-SD), period between the start of seed formation and physiological maturity in photothermal days (SD-PM), period between the appearance of the first flower and the end of leaf expansion (FL-LF), maximum leaf photosynthesis rate at an optimal temperature rate of 30°C (LFMAX), specific leaf area under standard growth conditions in cm 2 (SLAVARN), maximum size of completely expanded leaf in cm 2 (SIZLF), maximum fraction of the daily growth that is partitioned between the seed plots the pod (XFRT), maximum weight per seed in g (WTPSD), duration of the grain swelling period in the pods, under standard growth conditions in photothermal days (SFDUR); mean seeds per pod (SDPDV), and time necessary for the cultivar to reach ideal pod conditions in photothermal days (PODUR). During the model calibration the genetic coefficients were adjusted following the methodology proposed by Boote (1999).

Simulations
The simulations were performed to Tangará  The simulated grain yields in response to the planting dates were analyzed in a cumulative probability distribution for each studied growing season to determine the best planting dates for each season (wet, dry and winter season).

RESULTS
Highest average yield (2.3 t ha -1 ) were simulated for the planting dates of the wet season in the six years studied, followed by the dry season (1.3 t.ha -1 ). Lowest grain yields were simulated for the winter season (average of 0.2 t.ha -1 ) ( Table 3). The average rainfall for each harvest season studied; during the 6 years were 678, 742 and 60 mm for the wet season, dry season and winter season, respectively. The lower average of rainfall presented by the wet season compared to the dry season is explained by an atypical wet season in 2010, in which the total rainfall amount was 212.9 mm during all season.
The wet season had in 2007 the highest total rain accumulation during the full season, with 1,131 mm well distributed during all season (Figure 2), the average dry bean yield from all planting dates was 3.0 t.ha , respectively. After the flowering stage, the soil moisture decreased because of the reduction in rainfall events, the averages were 0.26 and 0.28 cm 3 cm -3 , respectively. The comparison between planting dates within the wet season, simulations indicated that when dry bean was planted in 01/12 yields increased, the average yield for this planting date was 2.9 t.ha -1 ( Figure 5). The planting date of 01/10 presented the lowest simulated yields, and the average was 1.6 t.ha -1 ( Figure 5). The cumulative probability analysis for the wet season ( Figure 6) indicated that higher grain yield could be achieved with the planting date of December 1 st (01/12), and yield decreases as early as planting occurs in the season. Lowest yields will most likely occur with the planting date of October 1 st . Simulated data of 2010, when the cumulative rainfall was 593 mm concentrated during the late season, had grain yields of 0.2, 0.1, 0.4, 1.0, 3.3 and 2.6 t.ha -1 for 01/10, 15/10, 01/11, 15/11, 01, 12 and 15/12, respectively.
The winter season is unfeasible (Figure 5), because of the low precipitation during this year period (Figure 1). Simulated yields were not higher than 0.5 tha -1 , and the planting date that showed highest yield at the cumulative probability of 75% was 15/04 ( Figure 6).

DISCUSSION
The main weather variable that influenced plant development and grain yield was the rainfall. The high yields of the wet season (Table 3) can be explained by the rainfall events typically occurring during the season period ( Figure 1). The crop development variables affected by water stress are reported in the comparison between the highest precipitation year for the wet season (planting date of 01/12/2010) and the typical year of the winter season (planting date of 06/01/2009).
The planting dates of wet season had plenty rainfall events associated with a good rain distribution, which increased soil moisture and dry bean yield. However, the low and non-uniform rain distribution observed in the winter season resulted in a soil drought during crop flowering and maturation (Figure 3). Nascimento (2004) reported that a reduction in soil available water of 40% during the reproductive stage could reduce pod number, pod size and number of grain per pods. In addition, Guimaraes et al. (2011) andBastos et al. (2011) showed an average yields loss of 58% when irrigated dry bean plants were compared to non-irrigated plants.
The water stress in the first soil layer presented during the vegetative stage, when leaf number is produced, may reduce LAI and biomass accumulation (Nascimento,     2004; Miorini et al., 2011). In the present study, wet and winter seasons had a water stress in the initial stages of crop development for the planting dates of 01/12/2010 and 06/01/2009, respectively, which decreased LAI ( Figure 4). However, the planting date of the wet season (01/12/2010) had the highest grain yield, explained by the increasing in rainfall events during the reproductive stages (Guimaraes et al., 1996). The planting date of 06/01/2009 had a low precipitation during all crop development, therefore lower yield.
Overall, most of the low simulated yields were consequence of soil water stress between flowering and pod maturation, growth stages that water shortage can strongly reduce grain yield (Araujo, 1996;Dallacort et al., 2010). In addition, planting dates of the winter season received rainfall amounts smaller than the dry bean water demand recommendation of 300 to 600 mm during all crop development. The dry bean daily water consumption is from 3 to 4 mm per day, requiring 100 mm monthly (Fancelli, 2009).
In the wet season, the grain yield increased as late as planting occurs, which is consequence of a welldistributed rainfall events in the late of the season, according to the six years studied. The planting date of December 1 st had the highest average yield for all cumulative probabilities. At 75% of cumulative probability, the grain yield was 2.5 t ha -1 . The regular rainfall distribution during the crop season of the planting date of wet season supply the dry bean water demand (Fancelli, 2009) mainly at the reproductive stages, when water stress most penalize grain yield (Guimaraes et al., 2011).
The dry season has an opposite weather pattern than the wet season, high precipitation and better rain distribution is presented in the early moment of the dry season. However, plentiful soil moisture content only during the early season can reduce grain yield. The soil water stress at reproductive stages will decrease nutrients uptake and grain yield (Nascimento, 2004). Therefore, late planting dates were most affected by a water stress after flowering. The grain yield of 1.7 and 1.8 t ha -1 at 75% of cumulative probability was simulated for the early planting dates of January 15 th and February 1 st , respectively, planting dates that had rainfall events well distributed over all crop development.
The drought periods during the plant development of all planting dates from the winter season affected the nutrient uptake and biomass accumulation Carlesso, 1998, Fiegenbaum et al. 1991). The effects of drought start when plants evapotranspiration is higher than water absorption by the root system (Vieira et al., 2006). Irrigation practices are an option to supply plants water requirement and consequently increase grain yields for the winter season.

Conclusion
Despite several available models, like the DEMANDAsis, which determine best management practices through the soil water balance, the CLIGEN, which simulate agricultural managements based on weather parameters, and several other crop models. The CROPGRO-Drybean model demonstrated to be an excellent tool to help research and growers to increase dry bean yields. The CROPGRO-Drybean model showed high sensitivity to precipitation events, in which high rainfall events well distributed over the crop development increased the dry bean grain yield. The drought stress during the reproductive stages for all seasons was the environmental variable that most affect dry bean productivity simulation.
The wet season had the highest simulated grain yields, consequence of high rainfall events well distributed over the season. Furthermore, as late planting occurs in the wet season higher were the probability to achieve high yields. The planting date of December 1 st provided the highest simulated grain yield within the wet season. In the dry season, planting dates of January 15 th and February 1 st are the best planting date for growers achieve higher yields, those planting dates had the highest likelihood to attend the crop water demand through rainfall events. Finally, the winter season requires irrigation practices for all simulated planting dates to increment dry bean grain yield

Conflicts of Interests
The authors have not declared any conflict of interests.