Simulating seed reserve mobilization and seedling growth of rice in DSRICE1

The simulation of seed reserve mobilization and seedling growth of rice in the model DSRICE1 was analyzed using published data. Early plant DM patterns are characterized by a decline in total (kernel+seedling) DM during heterotrophy, and then an increase into exponential growth after CO₂ assimilation (autotrophy) begins. Data for a tropical japonica variety were used for qualitative comparisons of observed and predicted kernel and total DM. The published version of DSRICE1 was sequentially modified to (1) subtract respiration costs from mobilized reserves, (2) use a constant mobilization rate, (3) use the same partitioning fractions (PF) for both reserves and assimilates, and (4) assess the effects of leaching losses. After the first modification, the model reproduced observed DM patterns, and setting a constant mobilization rate allowed complete use of reserves. Using consistent PF values for both assimilates and reserves was justified because it simplified the model and had only slight effects on predictions. Predicted mobilization efficiencies (ME; g seedling g⁻¹ seed) were greater than measured values, but were improved by accounting for leaching losses. In sensitivity analyses, six seed-related parameters had significant effects on seedling DM at 14 days after seeding (DAS), but none significantly affected mature DM or yields. In other tests, the effects of four traits on seedling growth were assessed over their likely empirical ranges. The start of CO₂ assimilation (DAS) had the greatest effect, followed by the start of mobilization and seed mass. Mobilization rate had the smallest effect over its likely range. Finally, simulated and measured early PF values were significantly different, particularly for root partitioning. Predicted DM to 18 DAS using three sets of PF values also differed, although more for total DM per plant than for some plant parts. Using the measured PF values will likely improve the model. It was also demonstrated that simulating reserve mobilization allows dynamic responses to variation in environmental and cultural inputs that may have critical effects on later growth. Thus, the calibration of parameters such as initial DM and leaf area required in other models may be avoided. Most important, using a more process-based approach in DSRICE1 facilitates (or forces) realistic linkages between seedling growth and environment by management interactions, which can improve analyses of crop and weed establishment and management strategies.