For sustainable food production in the Mekong Delta, local information on irrigation status is essential for allocating water resources efficiently at the community level. ALOS-2 PALSAR-2 L-band SAR can be used to detect submerged and nonsubmerged soil covered by rice plants at a spatial resolution compatible with field observation but a low temporal resolution. In this study, a new multiscale data assimilation technique is developed to estimate the spatiotemporal dynamics of field water levels at a temporal resolution appropriate to inform decision-making on when to initiate irrigation. The method employs an irrigation model based on parameters representing farmers' irrigation practices. ALOS-2 PALSAR-2 data are used to derive the hydrological parameters of the model, including an irrigation parameter representing how deep the field water level dropped until the next round of irrigation was initiated. We developed observation operators for the soil submersion status under vegetation and spatial submersion percentages instead of assimilating soil moisture products. The study uses ALOS-2 PALSAR-2 data (25–100 m spatial resolution, approximately every 42 days) over the Mekong Delta and experimental data collected in situ for model initialization and validation. The estimated irrigation parameter controlling the maximum depth of field water level showed spatio-temporal consistency with the ground-observed value (RMSE = 4.24 cm). The values of the parameters also showed spatial consistency with respect to DEM data; he paddies with low irrigation model parameter values were prone to be located in low-elevation zones (<2 m), whereas high model parameters values tended to be located in paddies in the high-elevation zones (>3 m).The results show promising applications using L-band SAR observations for monitoring paddy field water level, for irrigation practices and for estimation of the water consumption and of methane emissions.

对于湄公河三角洲的可持续粮食生产，当地的灌溉状况信息对于在社区层面有效分配水资源至关重要。ALOS-2 PALSAR-2 L 波段 SAR 可用于检测水稻植物覆盖的淹没和非淹没土壤，其空间分辨率与田间观测兼容，但时间分辨率较低。在这项研究中，开发了一种新的多尺度数据同化技术，以适当的时间分辨率估计田间水位的时空动态，从而为何时开始灌溉的决策提供信息。该方法采用基于代表农民灌溉实践的参数的灌溉模型。ALOS-2 PALSAR-2 数据用于推导模型的水文参数，包括一个灌溉参数，表示在下一轮灌溉开始之前田间水位下降的深度。我们开发了植被和空间淹没百分比下的土壤淹没状态观测算子，而不是同化土壤水分产品。该研究使用湄公河三角洲上空的 ALOS-2 PALSAR-2 数据（25–100 m 空间分辨率，大约每 42 天一次）和现场收集的实验数据用于模型初始化和验证。控制最大田间水位深度的估计灌溉参数与地面观测值（RMSE = 4.24 cm）显示出时空一致性。参数值也显示出与 DEM 数据相关的空间一致性；