Sugarcane variety trait modelling: evaluating and improving the APSIM-Sugar model for simulating crop performance under current and future climates across Brazil

Author(s)

Henrique Boriolo Dias

• Document type: Doctoral Thesis
• Press: Piracicaba.
• Institution: Universidade de São Paulo (USP). Escola Superior de Agricultura Luiz de Queiroz (ESALA/BC)
• Defense date: 2020-06-05

Examining board members

Paulo Cesar Sentelhas; Santiago Vianna Cuadra; Fábio Vale Scarpare

• Advisor: Paulo Cesar Sentelhas

Abstract

Brazil is the largest sugarcane producer in the world and the production systems across the country are very complex with large spatial and temporal variations in yields. This is a result of large Genotype × Environment × Management (G × E × M) interactions that ultimately affect crop performance. Moreover, likely changes in the climate conditions are expected in the future imposing additional challenges to sugarcane production. Process-based crop modelling is scientifically accepted as a way to understand those interactions. Upon that, this thesis aimed to integrate field experiments datasets with the APSIM-Sugar model for better understanding the G × E × M interactions in Brazilian cropping systems. The capability of the model was first evaluated with default settings and then modifications and traits were proposed to effectively predict G (varieties) differences. With APSIM-Sugar upgraded, an up-to-date projection of future climate impacts on sugarcane yields across Brazil was performed. The sugarcane dataset came from large field experiments carried out between 2012 and 2017 at two tropical sites in Brazil: Guadalupe, in the state of Piauí (PI), and São Romão, in the state of Minas Gerais (MG), where several varieties were cultivated under non-limiting (potential) conditions. These data were analysed and employed to characterise variety traits for use in APSIM-Sugar. Substantial changes were required to enable the model to simulate canopy development and stalk yield appropriately. The outcomes from the modelling studies showed that the model is now able, although yet empirically, to account for ageing processes, known as the reduced growth phenomenon, to get better yield predictions in high yielding environments, as well as the ability to distinguish between varieties when it comes to yield gains above ~ 150 t/ha (~ 40 t/ha in dry mass). The findings suggest it is reasonable to hypothesise that the APSIM-Sugar modifications and traits are plausible and are an important step for unravelling G × E × M interactions to improve the performance of the sugarcane industry. Another step of this thesis, before applying APSIM-Sugar, was to evaluate gridded weather datasets (NASA/POWER and other developed for Brazil referred to \'XAVIER\') as alternative climate inputs in the model. The results indicated these data sources should be employed with caution when the purpose is to simulate sugarcane yield because of the poor performance of gridded rainfall, at least for Center-South Brazil. Lastly, a climate change impact assessment was performed with APSIM-Sugar considering new features and traits for 10 Brazilian sites, being six under rainfed and four under fully-irrigated conditions. Five global climate models were selected to represent basic classes of climate changes: relatively cool/wet; cool/dry; middle; hot/wet; and hot/dry, for four future scenarios. These climate series were then used as input for simulations of typical cropping systems. The ensemble of future climate-crop model projections suggests that cane yields will mostly decline for both rainfed and fully-irrigated systems, however, the scenarios generated by climate models presented large uncertainty, requiring that these projections should be interpreted with caution. (AU)

• FAPESP's process: 16/11170-2 - Multi-model approach to sugarcane yield forecasting and assessment of crop performance under current and future climate conditions
• Grantee: Henrique Boriolo Dias
• Support Opportunities: Scholarships in Brazil - Doctorate