赵军, 刘坚, 胡飞鹏,等.基于时空融合模型和TVDI的土壤水分监测方法研究[J].河南理工大学学报（自然科学版）,2025,44(3):112-118.

ZHAO J, LIU J, HU F P, et al.Study on soil moisture monitoring method based on Spatiotemporal Image-fusion Model and TVDI[J]. Journal of Henan Polytechnic University(Natural Science) , 2025, 44(3): 112-118.

基于时空融合模型和TVDI的土壤水分监测方法研究

赵军, 刘坚, 胡飞鹏, 托瑞, 孙紫云

西北师范大学 地理与环境科学学院，甘肃 兰州 730070

摘要: 目的 针对以往利用温度植被干旱指数（temperature vegetation dryness index，TVDI）监测土壤水分时多采用MODIS或Landsat TIRS/TIRS-2等单一遥感数据，导致在时空分辨率方面存在数据缺陷问题，  方法 提出一种基于多源遥感数据时空融合模型（spatiotemporal image-fusion model，STI-FM）和TVDI的土壤水分监测方法，探究该方法的土壤水分监测效果和STI-FM融合多源遥感地表温度在试验区的适用性。  结果 结果表明：（1）STI-FM融合MODIS或Landsat TIRS/TIRS-2的地表温度数据质量较好；运用STI-FM推测的地表温度数据与遥感监测的地表温度数据具有较强的相关性，得到R²为0.89，均方根误差（root mean squared error，RMSE）为2.85 K，即STI-FM适用于本文试验区。（2）与传统TVDI相比，利用STI-FM改进地表温度数据并结合优选植被指数的TVDI对试验区土壤水分监测精度提高效果更显著，与原始NDVI-LST特征空间计算的TVDI相比，6，7月份TVDI与土壤水分的相关系数分别提高了0.07和0.14。（3）与再分析土壤水分资料ERA5-Land的土壤水分数据相比，改进后的TVDI土壤水分监测结果精度更高，6，7月份TVDI与土壤水分的相关系数分别提高了0.1和0.44。  结论 综上所述，基于STI-FM的TVDI可以提高土壤水分监测效果，在提高监测精度的同时，监测土壤水分数据较其他土壤水分监测方法具有更高的空间分辨率，为TVDI土壤水分监测提供了一种新思路。

关键词:土壤水分;TVDI;时空融合模型;时空融合;适用性

doi: 10.16186/j.cnki.1673-9787.2023120040

基金项目:国家自然科学基金资助项目（42161072）

收稿日期:2023/12/14

修回日期:2024/02/23

出版日期:2025-04-18

Study on soil moisture monitoring method based on Spatiotemporal Image-fusion Model and TVDI

ZHAO Jun, LIU Jian, HU Feipeng, TUO Rui, SUN Ziyun

College of Geography and Environmental Science， Northwest Normal University， Lanzhou 730070， Gansu， China

Abstract: Objectives To address the limitations of using single-source remote sensing data （e.g.， MODIS or Landsat TIRS/TIRS-2） in monitoring soil moisture through the Temperature Vegetation Dryness Index （TVDI）， which often leads to insufficient spatiotemporal resolution，  Methods a soil moisture monitoring approach was proposed by integrating the spatiotemporal image-fusion model （STI-FM） with TVDI. This study evaluated the monitoring performance of this method and assessed the applicability of STI-FM in fusing multi-source remote sensing land surface temperature （LST） data within the study area.  Results （1）The STI-FM model demonstrates superior performance in fusing MODIS and Landsat TIRS/TIRS-2 land surface temperature （LST） data， achieving strong correlation （R²=0.89） and low error （RMSE=2.85 K） between the predicted and satellite-observed LST values， confirming its applicability in the study area. （2） Compared to traditional TVDI， the modified TVDI combining STI-FM-enhanced LST data and optimized vegetation indices significantly improved soil moisture monitoring accuracy. The correlation coefficients between TVDI and soil moisture increased by 0.07 （June） and 0.14 （July） compared to original NDVI-LST feature space calculations. （3） The improved TVDI achieved higher accuracy than ERA5-Land reanalysis soil moisture data. Correlation coefficients between TVDI and soil moisture surpassed ERA5-Land by 0.10 （June） and 0.44 （July）.  Conclusions The STI-FM-enhanced TVDI framework significantly improves soil moisture monitoring by simultaneously enhancing estimation accuracy and providing higher spatial resolution compared to conventional methods， offering a novel paradigm for TVDI-based moisture retrieval.

Key words: soil moisture; TVDI; spatiotemporal image-fusion model; spatio-temporal fusion; applicability