Soil moisture estimation is critical for environmental and agricultural sustainability, with its spatial and temporal variation playing a key role in drought monitoring and understanding climate change. The region of Prince Edward Island (PEI), Atlantic Canada's largest potato producer, is facing irregular precipitation patterns that stress crop water supplies. This study aims to estimate field-scale soil moisture utilizing satellite-based reflective and thermal infrared bands from Landsat-8 Operational Land Imager (OLI)/Thermal Infrared Sensor (TIRS) and Moderate-resolution Imaging Spectroradiometer (MODIS) over the cloud-based Google Earth Engine (GEE) platform. The GEE data catalog's pre-processed data endured to calculate various indicators for the agricultural seasons of 2021 and 2022 across three designated plots: A, B, and C. The indicators are land surface temperature (LST), normalized difference vegetation index (NDVI), normalized difference water index (NDWI), and normalized difference moisture index (NDMI). NDVI and LST were used to calculate the soil moisture index (SMI), representing the real-time soil moisture at the field scale. The soil moisture data was validated using in situ measurements. The analysis showed good Root Mean Square Error values of 1.43 % (Plot A), 2.12 % (Plot B), and 2.60 % (Plot C). A weak negative association between LST and NDVI was noticed in the study, with R² values of 0.25, 0.38 and 0.26 for Plots A, B and C, respectively. As the LST rises, vegetation declines due to the elevated temperatures in the study area. Second, a significant (p < 0.05) negative correlation (R2 =1) existed between SMI and LST in both the 2021 and 2022 seasons, showing a decrease in the top layer soil moisture with LST. The NDWI exhibited a significant inverse correlation with soil moisture, while NDMI and NDVI are effective predictors. Hence, based on the current study, optical and thermal remote sensing offers valuable insights into soil moisture dynamics and can be a good tool for irrigation control and water conservation.

中文翻译：

---

一种包容性的作物土壤湿度估算方法：利用 Google 地球引擎上的卫星热红外波段和植被指数


土壤水分估算对于环境和农业可持续性至关重要，其空间和时间变化在干旱监测和了解气候变化方面发挥着关键作用。爱德华王子岛 （PEI） 是加拿大大西洋地区最大的马铃薯生产地，该地区正面临不规则的降水模式，给作物供水带来压力。本研究旨在通过基于云的 Google Earth Engine （GEE） 平台，利用 Landsat-8 业务土地成像仪 （OLI）/热红外传感器 （TIRS） 和中分辨率成像光谱仪 （MODIS） 的卫星反射和热红外波段来估计田间土壤水分。GEE 数据目录的预处理数据经得起计算 2021 年和 2022 年农业季节的各种指标，涉及三个指定地块：A、B 和 C。这些指标是地表温度 （LST）、归一化差值植被指数 （NDVI）、归一化差值水分指数 （NDWI） 和归一化差值水分指数 （NDMI）。NDVI 和 LST 用于计算土壤水分指数 （SMI），表示田间尺度的实时土壤水分。土壤水分数据使用原位测量进行验证。分析显示良好的均方根误差值为 1.43 %（图 A）、2.12 %（图 B）和 2.60 %（图 C）。在研究中观察到 LST 和 NDVI 之间的微弱负相关，图 A、B 和 C 的 R² 值分别为 0.25、0.38 和 0.26。随着 LST 的升高，由于研究区域的温度升高，植被下降。其次，2021 年和 2022 年季节 SMI 和 LST 之间存在显著的 （p < 0.05） 负相关 （R2 =1），表明表层土壤水分与 LST 一起减少。 NDWI 与土壤水分呈显著负相关，而 NDMI 和 NDVI 是有效的预测因子。因此，根据目前的研究，光学和热遥感为土壤水分动态提供了有价值的见解，可以成为灌溉控制和节水的良好工具。