Sugarcane Productivity Simulation Under Different Scenarios by DSSAT/CANEGRO Model in the Western São Paulo

Sugarcane (Saccharum officinarum L.) is one of the most important crops in Brazil and its growth and development can be simulated through process-based models. The current study evaluated a model based on the decision support system for the transfer of Agrotechnology DSSAT/CANEGRO to simulate the sugarcane crop productivity in the western region of São Paulo. The DSSAT/CANEGRO model was calibrated using published yield parameters from a selection of five Brazilian sugarcane cultivars, while sugarcane yield data (tons of stems per hectare) from commercial land were used as benchmark data. Other modeling inputs were derived from the primary regional cultivar. The root mean square error (RMSE), Willmott agreement index (d), and mean absolute error (MAE) were used as performance metrics. The DSSAT/CANEGRO model resulted in a good RMSE performance. The productivity estimates were better for the cultivars SP791010 and RB835486, with RMSE equal to 2.27 and 4.48 Mg ha-1, respectively. The comparison between model-based estimates and observed data produced d values in the range from 0.86 to 0.99, and MAE values in the range of 1.84 to 4.22 Mg ha-1.


Introduction
The cultivation of sugarcane (Saccharum officinarum) is among the most important crops in Brazilian agribusiness. Brazil is the world's largest sugarcane producer and the second largest producer of ethanol. The areas under production continue to gradually increase, although at a slower pace in the midwest states in Brazil and southeast regions. Since 2008, industrial units of sugarcane processing facilities were installed in the west of São Paulo state, which facilitated developing additional sugarcane fields (CONAB, 2018). This region has some edaphoclimatic characteristics that are different from the other sugarcane regions of the state, such as sandy soils with low water retention capacity, high temperatures, heavy rains and long periods without rain (summer), which promote plant water stress.
There are different models for estimating growth and evaluating the development of process-based cultures that can facilitate monitoring and contribute to activities related to productivity forecasting, as well as assist in understanding those mechanisms that are directly involved in the different responses of culture to the environmental conditions (Marin et al., 2011;Nassif et al., 2012).
According to Marin et al. (2011), currently there are several models that can be used for sugarcane growth simulations, such as: AUSCANE (Jones et al., 1988), QCANE (Liu & Kingston, 1995), APSIM (Keating et al., 1999), and CASUPRO (Villegas et al., 2005). One of the main and most used models is the DSSAT/CANEGRO (Inman-Bamber, 1991;Singels & Bezuidenhout, 2002) is also one of the main simulation models of growth of the sugarcane currently in use (Nassif et al., 2012). The DSSAT/CANEGRO model is based on the Ceres-Maize model (Jones et al., 1986), which was developed to model the most important physiological processes related to sugar production processes in South Africa (Inman-Bamber, 1991).
The DSSAT/CANEGRO model is being used in different regions of the world to analyze the different sugarcane production systems (Inman-Bamber, 1991;Marin et al., 2011;Singels & Bezuidenhout, 2002;Singels et al., 2008;Nassif et al., 2012). In Brazil, Marin et al. (2011) calibrated the DSSAT/CANEGRO model for two cultivars in the production systems of the center-south of Brazil.
Thus, the aim of this study is to estimate the sugarcane productivity under conditions in the western portion of the state of São Paulo. The following specific objectives will be developed: (i) to evaluate the DSSAT model under different climatic and soil conditions for sugarcane production; (ii) to evaluate the performance of the DSSAT model using data reported by the sugarcane mills, and (iii) to evaluate the sugarcane productivity estimates in the western portion of the state of São Paulo, Brazil.

Model Description
Sugarcane productivity simulations in western portion of São Paulo state were carried out with the DSSAT/CANEGRO version 4.5 to model the most relevant sugarcane physiological processes, whereas the Weatherman subroutine to analyze the climatic data.
The DSSAT/CANEGRO model requires water balance information and daily meteorological data (i.e., solar radiation, maximum and minimum temperatures, and precipitation). The sugarcane growth modeling includes phenology, canopy development, accumulation of biomass and sucrose, partitioning, root growth, water stress and lodging data (Singels et al., 2008). The model also requires soil physics data (i.e., field capacity, permanent wilting point, water saturation and soil depth) at the entrance of the process to adjust the water balance (Nassif et al., 2012).
The varieties used to perform the simulations (RB835486, SP791011, RB931530, and RB93509) were selected based on the sugar mill productivity data for the last 15 years. The RB867515 cultivar parameters were used to calibrate the model. These cultivars were selected due to their representativeness in planting sugarcane fields in the studied region.
The soil profile characterizations were classified according to the Pedological Map of São Paulo state presented by Rossi (2017). The most representative soils of the Presidente Prudente-SP microregion were used in the simulation: Argilossos and Latossolos according to Brazilian System of Soil Classification (SiBCS) (Santos et al., 2013), which are equivalent to Ultisols and Typic Hapludox subgroups, respectively, according to U.S. Soil Taxonomy (Soil Survey Staff, 2019).

Method of Acquisition, Selection and Transformation of Climatic Data
The metadata used for weather stations are shown in Table 1. The region of western São Paulo state on the borders with Paraná and Mato Grosso do Sul state, has a tropical climate, type CWa according to the Köppen climate classification, characterized by hot and rainy summers, and cold and dry winters. The average annual precipitation is 1,308 mm, with a maximum of 2,049 mm in 2009. January has the highest average rainfall (212 mm), according to data recorded by the meteorological station of Presidente Prudente from 1969 to 2013. Severe drought events were observed, demonstrating again the great randomness and complexity of the atmospheric system, with the year 2001 being classified as unusual with relation to climate normals. La Niña's (a cold phase oscillation quasiperiodic of climate pattern that arises across the tropical Pacific Ocean on the coast of Peru and Ecuador every five years) (Gómez-Aguilar, 2020) years are no exception to this characteristic, even though they tend to be drier years. El Niño (describe the warm oceanic phase of climate pattern) (Gómez-Aguilar, 2020) years are characterized in most cases by the presence of extreme events in the region, such as intense rains (Berezuk & Neto, 2006) (Figure 2b).  The RB867515 cultivar was previously calibrated for the DSSAT/Canegro and APSIM/Sugar models using inputs from six different regions of Brazil (Marin et al., 2013). The values of the parameters of each cultivar used in the simulation for the DSSAT/CANEGRO model are summarized in Table 3.

Simulation Scenarios
The scenarios were based on the maturation cycle of the cultivars used: early, medium and late; three harvest seasons June 15th (early), August 15th (medium) and September 15th (late). Ten years of planting were simulated for each combination of climate and soil. Thus, for each location, the five varieties and two soils were considered, totaling 10 scenarios per region (Table 4). The planting date in all cases was on the 15th of June.  (Santos et al., 2013) and its equivalent according to the closest Soil Survey Staff (2019) (in parentheses).

Evaluation of Performance of the Models
In this study, the agreement index (d), mean error (ME), mean absolute error (MAE), and root mean square error (RMSE) were used as performance statistical metrics (Willmott et al., 1985), with Equations (6 to 9) as follows: Where, Yi and Y are the estimated and observed sugarcane yield, in Mg ha -1 , respectively; Ym are the average of estimated and observed sugarcane yield, in Mg ha -1 ; and n is the number of observations.

Meteorological Conditions
Weather conditions during sugarcane period scenarios from January 1969 to December 2013 are shown in Figure  2a. The values of, ETo was lower than precipitation (P), with ETo equal to 200 mm in the summer humid season and 25 mm in winter (Figure 2b).       Vol. 12, No. 7;2020