Introduction - The STICS model (Brisson et al., 2003) has been mainly developed and tested in temperate regions, making its adaptation, or tropicalization, a crucial step for robust use in tropical agroecosystems. Sugarcane, a perennial crop of major economic importance, provides a challenging test case due to its long cycle, strong dependence on nitrogen (N) inputs, and the diversity of soil types found in tropical islands such as La Réunion. This contribution synthesizes results from three complementary studies aiming to tropicalize STICS for sugarcane-based systems: (i) calibration and robustness assessment of sugarcane crop growth across contrasted environments (Kebalo et al., 2025); (ii) parameterization of nitrogen mineralization and fertilization response in volcanic and tropical soils; (iii) calibration and evaluation of nitrogen losses (NH₃ volatilization, N₂O emissions, and NO₃⁻ leaching) under tropical conditions. Together, these studies provide a consolidated framework to improve tropical applications of STICS and to support sustainable management of sugarcane production systems. Material and methods - The first study focused on calibrating and validating STICS simulations of sugarcane growth under optimal nutrient conditions. Parameterization used crop cycles from four countries, including measurements of below- and aboveground variables (Chevalier et al., 2025 ; Jones et al., 2019). Validation relied on 62 additional R570 cycles and 158 cycles from 30 international varieties (Christina et al., 2020). Structural parameters governing phenology, biomass allocation, and canopy development were adjusted, and performance was evaluated on aboveground biomass, LAI, and harvest traits. The climatic domain of validity was assessed by analysing model errors along gradients in temperature and water availability. The second study aimed to improve the representation of nitrogen availability in andic and non-andic tropical soils in La Reunion. Ten sites (four andic, six non-andic) with fertilized and unfertilized treatments were used (Chevalier et al., 2025). For andic soils, clay content was adjusted on unfertilized plots to match observed mineralization and linked to allophane content. For non-andic soils, fine silt was incorporated into the clay fraction to establish a separate calibration curve. The orgeng parameter (N immobilization) was optimized using 10 observations from fertilized plots. Validation was conducted with 22 datasets from four sites (TERO project, eRcane). The third study calibrated nitrogen losses using a multi-year dataset of 12 sugarcane cycles with mineral and organic fertilization (poultry manure, pig slurry, sewage sludge), including measurements of NH₃ volatilization, N leaching, and N₂O emissions (SOERE-PRO, Poultney, 2021). Validation used the multi-site TERO database (4 sites, 6–11 cycles) with observations of yield, N uptake, and soil C and N stocks. Calibration was sequential: (i) plant N uptake, (ii) NH₃ volatilization through adjustment of maximum volatilization rate and pH threshold, and (iii) N₂O emissions and nitrate leaching via optimization of the daily nitrification rate adapted to tropical soils. Results and discussions -Sugarcane growth: robust simulation across tropical environments. STICS satisfactorily simulated biomass accumulation, intercepted radiation, and biomass partitioning across parameterization and validation datasets. Despite calibration on Réunion varieties only, the model demonstrated strong robustness across international varieties, maintaining reasonable accuracy for biomass at harvest. Error analysis across climatic gradients revealed a decrease in accuracy in extreme conditions. This provides a roadmap for future improvements, such as refining temperature responses, water-stress functions, and variety-specific traits. Improved representation of N mineralization and fertilization response in volcanic soils. The adjusted relationships linking clay content (or fine silt correction) to soil allophane successfully reproduced observed N mineralization dynamics in both andic and non-andic soils. This constitutes an important step toward representing the unique physicochemical properties of volcanic soils, particularly their high capacity to stabilize organic matter. Calibration of N immobilization improved the simulation of N availability following fertilization. Validation across 22 observations showed satisfactory agreement for N uptake and biomass response, demonstrating the parameterization's transferability across different sites and soil types. Calibration of N losses under tropical conditions. Sequential calibration improved the representation of gaseous and leaching losses. Adjusting volatilization parameters allowed the model to reproduce the magnitude of NH₃ volatilization across mineral and organic fertilization scenarios. Optimizing the maximum daily fraction of ammonium that can be nitrified, which is typically higher in warm, moist tropical soils, significantly improved the model's ability to simulate measured N₂O emissions and NO₃⁻ leaching fluxes. Validation with multi-site TERO data demonstrated that the calibrated model can reproduce yields, N uptake, and soil N and C trends across a wide range of organic fertilization systems. Conclusion - This work presents a comprehensive effort to tropicalize the STICS model for sugarcane-based systems. The three studies collectively show that: i) STICS can robustly simulate sugarcane biomass production across varieties and climatic conditions, although leaf dynamics remains a priority for improvement; ii) Volcanic soil characteristics, especially allophane content and associated mineralization behavior, can be effectively integrated to improve predictions of N availability and fertilization response; iii) Calibration of N volatilization, nitrification, N₂O emissions, and nitrate leaching allows the model to capture the major nitrogen loss pathways typical of tropical systems. Together, these advances significantly enhance STICS's capacity to evaluate agronomic and environmental performance in tropical sugarcane systems.
Kebalo et al. (Thu,) studied this question.