Fresh water stored by on-farm reservoirs (OFRs) is an important component of surface hydrology and is critical for meeting global irrigation needs. Farmers use OFRs to store water during the wet season and for crop irrigation during the dry season, yet their seasonal and inter-annual variability and downstream impacts are not quantified. Therefore, OFRs' sub-weekly surface area changes are critical to understanding their dynamics and mitigating their downstream impacts. However, prior to the recent increase in satellite imagery availability and improvement in sensors' spatial resolution, monitoring the OFRs' sub-weekly surface area changes across space and time was challenging because OFRs occur in high numbers (i.e. hundreds) and are small water bodies (< 50 ha). We propose a novel multi-sensor approach to monitor OFRs surface areas, developed based on 736 OFRs in eastern Arkansas, USA, which leverages the use of PlanetScope (PS), RapidEye (RE), Sentinel 2 (S2), and Sentinel 1 (S1). First, we estimate the uncertainties in surface area for each sensor by comparing the surface area estimates to a validation dataset, and by comparing RE, S2 and S1 to PS—the sensor with the highest spatial resolution (i.e. 3.125 m). Second, we use the uncertainties of each sensor with a data assimilation algorithm based on the Kalman filter to obtain sub-weekly surface area time series for all OFRs. Our results show the lowest uncertainties for PS, followed by RE, S2 and S1. These uncertainties varied according to the OFRs' size and shape complexities. The surface area estimates derived from the Kalman filter including only the optical sensors resulted in high agreement (r² > 0.95) and small uncertainties (4–8%) when compared to the validation dataset. We found higher uncertainties (5–14%) when adding S1 to the Kalman filter—this is related to the higher uncertainties found for S1 (~20%). The algorithm can assimilate optical and radar satellite data to increase the OFRs' surface area time series cadence allowing us to investigate sub-weekly surface area changes. The algorithm is not sensor-specific, and it accounts for the uncertainties in both the sensors observations and the resulting surface areas, which are key advantages when compared to other algorithms used to combine satellite data. By improving the surface area observations cadence and providing the surface area uncertainties, the approach presented in this study has the potential to enhance water conservation plans by allowing better assessment and management of the OFRs.

农场水库 (OFR) 储存的淡水是地表水文学的重要组成部分，对于满足全球灌溉需求至关重要。农民在雨季使用 OFR 储存水，在旱季使用 OFR 进行作物灌溉，但它们的季节性和年际变化以及下游影响尚未量化。因此，OFR 的亚周表面积变化对于了解其动态和减轻其下游影响至关重要。然而，在最近卫星图像可用性增加和传感器空间分辨率提高之前，监测 OFR 跨空间和时间的亚周表面积变化具有挑战性，因为 OF​​R 出现的数量很多（即数百个）并且是小型水体（< 50 公顷）。我们提出了一种新颖的多传感器方法来监测 OFR 表面积，该方法基于美国阿肯色州东部的 736 个 OFR 开发，利用了 PlanetScope (PS)、RapidEye (RE)、Sentinel 2 (S2) 和 Sentinel 1 ( S1)。首先，我们通过将表面积估计值与验证数据集进行比较，并将 RE、S2 和 S1 与 PS（具有最高空间分辨率（即 3.125 m）的传感器进行比较）来估计每个传感器的表面积不确定性。其次，我们使用每个传感器的不确定性和基于卡尔曼滤波器的数据同化算法来获得所有 OFR 的亚周表面积时间序列。我们的结果显示 PS 的不确定性最低，其次是 RE、S2 和 S1。这些不确定性根据 OFR 的大小和形状复杂性而变化。[R ² > 0.95) 和与验证数据集相比的小不确定性 (4–8%)。当将 S1 添加到卡尔曼滤波器时，我们发现更高的不确定性 (5–14%)——这与 S1 (~20%) 的更高不确定性有关。该算法可以同化光学和雷达卫星数据，以增加 OFR 的表面积时间序列节奏，使我们能够调查亚周表面积的变化。该算法不是特定于传感器的，它考虑了传感器观测和由此产生的表面积的不确定性，与用于组合卫星数据的其他算法相比，这是关键优势。通过改进表面积观测节奏并提供表面积不确定性，