The growing interest in low-frequency SAR for soil parameter retrieval has led to the development of new active L-band satellites, that will provide novel surface soil moisture products and retrieval possibilities; however, due to data unavailability so far, limited applications have investigated the use of change detection models using L-band satellite SAR data. Since July 2020, high revisit time, high-resolution acquisitions by the Satélite Argentino de Observación COn Microondas (SAOCOM) Argentinian-Italian constellation have become accessible over Europe. Therefore, this research presents an investigation of the potential of multi-temporal L-band SAOCOM-1 for monitoring soil moisture variations underneath low and sparse agricultural vegetation. Moreover, it proposes a procedure for the mitigation of roughness contribution, by exploiting the entropy parameter derived from the dual-polarimetric decomposition. L-band sensitivity to soil moisture has been jointly evaluated in respect of Sentinel-1 C-band data by (1) comparing the temporal profiles of the backscattering coefficient, , at VV and VH polarization, with the support of decomposition parameters (entropy and ), NDVI and precipitation data; (2) regression analysis with in situ soil moisture measurements, obtained by the REMEDHUS network in the Douro River basin (Spain); 3) evaluating the soil moisture retrievals obtained at C- and L- band using a change detection method. Finally, the effectiveness of the roughness normalization procedure for SAOCOM data has been validated using in situ data. L-band co-polarized has proved to be the best configuration for soil moisture inversion, being relatively insensitive to vegetation, as demonstrated by decomposition results and trend interpretation. Overall, regressions detected an R 22% higher at L-band than C-band, with values up to 0.74 for VV (=0.32) and up to 0.47 for the VH band (=0.14). Co-polarized data obtained R on average 62.1% and 74.7% higher for SAOCOM and Sentinel-1. The retrieval models show an ubRMSD of 7.1% for SAOCOM data and 8.3% for Sentinel-1. The application of the proposed roughness normalization procedure to SAOCOM led to an ubRMSD of 6.7% improving the retrieved soil moisture trend by 7.9%. This exploratory analysis demonstrated SAOCOM data potential for soil moisture mapping and would serve as a foundation for more advanced retrieval procedures.

中文翻译：

---

在半干旱地区（西班牙杜罗河流域）使用变化检测方法反演表层土壤水分的 SAOCOM 和 Sentinel-1 数据的比较分析


人们对低频 SAR 用于土壤参数反演的兴趣日益浓厚，导致了新型有源 L 波段卫星的开发，这将提供新颖的表面土壤湿度产品和反演可能性；然而，由于迄今为止数据不可用，有限的应用研究了使用 L 波段卫星 SAR 数据的变化检测模型的使用。自 2020 年 7 月以来，阿根廷-意大利星座卫星阿根廷-意大利星座的高重访时间、高分辨率采集已经可以在欧洲上空进行访问。因此，本研究对多时相 L 波段 SAOCOM-1 用于监测低矮稀疏农业植被下土壤湿度变化的潜力进行了调查。此外，它提出了一种通过利用从双极化分解导出的熵参数来减轻粗糙度贡献的过程。 L 波段对土壤湿度的敏感性已针对 Sentinel-1 C 波段数据进行了联合评估，方法是：(1) 在分解参数（熵和 VH 偏振）的支持下比较后向散射系数 的时间分布。 )、NDVI 和降水数据； (2) 通过杜罗河流域（西班牙）的 REMEDHUS 网络获得的原位土壤湿度测量值进行回归分析； 3) 使用变化检测方法评估在 C 和 L 波段获得的土壤水分反演。最后，使用现场数据验证了 SAOCOM 数据粗糙度归一化程序的有效性。分解结果和趋势解释表明，L 波段共极化已被证明是土壤湿度反演的最佳配置，对植被相对不敏感。 总体而言，回归检测到 L 波段的 R 比 C 波段高 22%，VV 波段的值高达 0.74 (=0.32)，VH 波段的值高达 0.47 (=0.14)。 SAOCOM 和 Sentinel-1 的同极化数据获得的 R 平均高出 62.1% 和 74.7%。检索模型显示 SAOCOM 数据的 ubRMSD 为 7.1%，Sentinel-1 的 ubRMSD 为 8.3%。将所提出的粗糙度归一化程序应用于 SAOCOM 后，ubRMSD 提高了 6.7%，将反演的土壤湿度趋势提高了 7.9%。这项探索性分析证明了 SAOCOM 数据在土壤湿度测绘方面的潜力，并将作为更高级检索程序的基础。