Uncertainty in sugarcane (Saccharum officinarum (L.)) growth modeling is caused by a poor description of events such as artificial leaf stripping and natural storms. Data assimilation methods provide an effective strategy to integrate various measurements into crop models, rendering them less susceptible to model uncertainties. The objective of this study is to investigate the ability of three data assimilation methods to cope with interfered leaf area index (LAI), estimate stalk yield, and diagnose water stress. Through a two-year sugarcane field experiment, the data contribution to sugarcane soil-plant-atmosphere continuum simulation was quantified by assimilating soil water content (SWC) and LAI observations. Results demonstrated the importance of choosing appropriate data assimilation strategies. The forcing method failed to accurately simulate the temporal evolution of the soil moisture profile due to soil parameters not being updated, and the calibration method led to underestimated LAI before defoliation and overestimated LAI after defoliation. The EnKF method performed the best in estimating soil water content, LAI development, and sugarcane yield, as it reconciled simulation and observation uncertainties. The relative contribution of LAI and SWC data to yield estimation was dependent on water stress level. A reasonable diagnosis of water stress and subsequent impacts relies on the choice of data assimilation method. The forcing method is prone to producing high spatial variability of daily dry matter increase, and overestimated LAI from the calibration method after defoliation may lead to exacerbated and biased water stress estimates.
Read full abstract