Accurate estimation of terrestrial evapotranspiration (ET) is critical for understanding land-atmosphere interactions and supporting sustainable water resource management under climate change. However, large uncertainties persist in China due to complex surface heterogeneity, sparse ground observations, and frequent extreme events. This study presents the first comprehensive evaluation of evapotranspiration products over China using observations from 64 eddy covariance flux towers, assessing multiple daily and monthly datasets across diverse climate zones, land surface types, and five representative extreme climatic conditions, including high temperature (Temp), high vapor pressure deficit (VPD), high precipitation (Pre), high wind speed (WS), and drought. An explainable machine learning framework based on XGBoost was further developed to reduce ET uncertainties and identify dominant controlling factors, with independent validation conducted using additional flux sites. The results show that: 1) Under overall conditions, daily ET products exhibit larger uncertainties than monthly products, with correlation coefficients (r) generally ranging from 0.37 to 0.59 for daily estimates and from 0.61 to 0.85 for monthly estimates. Under extreme climatic conditions, daily ET accuracy declines sharply, with the mean correlation decreasing from 0.634 to 0.332, showing the strongest degradation under high VPD and relatively weaker sensitivity under high WS. Monthly ET products are less affected by extremes but still show notable performance deterioration during drought conditions. 2) In typical dense flux observations regions, none of the ET products capture spatial heterogeneity effectively, and estimation accuracy declines noticeably. 3) The XGBoost model significantly enhances ET estimation at both daily (r = 0.908) and monthly (r = 0.931) scales, particularly under extreme climatic conditions. ET products, solar radiation (Rs), and VPD are the primary contributors associated with higher estimated ET in the model. Independent validation across 12 sites confirmed robustness of XGBoost, demonstrating strong performance even under extreme climatic conditions. Overall, this study recommends integrating flux-tower-informed machine-learning fusion with existing ET products as an effective pathway to reduce ET uncertainty and enhance the reliability of regional water-cycle assessments and climate-impact analyses in complex climatic regions.

Thermal infrared image stitching in agricultural remote sensing often suffers from challenges such as low texture, temperature drift, and low geometric resolution. Furthermore, real-time responsiveness is highly demanded in agricultural applications, making it imperative to develop efficient and robust stitching methods. To address these challenges, we propose FTP-Stitch, a fast and robust thermal infrared image stitching method based on feature alignment and flight trajectory preservation under similarity transformation constraints. Specifically, building upon the similarity transformation model adopted in MegaStitch, we introduce a Flight Trajectory Preservation (FTP) term, which utilizes prior GNSS-based UAV flight paths to provide stable pixel shift estimation for each image, thereby accurately preserving their relative positions in physical space. Then, by jointly optimizing the feature alignment term and the FTP term, we construct a least-squares optimization model under similarity transformation constraints to solve the corresponding transformation matrices. To validate the effectiveness of the proposed method, we conduct extensive experiments on a thermal infrared remote sensing dataset that covers various crop types, flight altitudes, and environmental conditions. We evaluate the stitching results both subjectively and objectively in terms of alignment accuracy, visual naturalness, and stitching efficiency. Experimental results demonstrate that, compared with existing state-of-the-art thermal infrared stitching methods, the proposed method reduces the average RMSE by 19.7 % compared with OP-GSP and by 35.1 % compared with MegaStitch, while reducing the average stitching time by 78.0 % compared with OP-GSP. Meanwhile, superior performance is achieved in both geometric alignment and visual quality. This offers an effective solution for rapid large-scale thermal infrared image stitching in agricultural UAV remote sensing applications.

Accurate retrieval of volumetric soil moisture content (VSMC) from optical remote sensing observations remains challenging, primarily due to limited penetration depth and frequent signal contamination from vegetation cover. Physically based radiative transfer models face ill posed inversion and spectral redundancy, while traditional methods struggle to detect subtle physiological features under strong canopy influence. This study develops a domain adaptive framework combining Convolutional Neural Networks and Kolmogorov Arnold Networks to estimate upper root zone VSMC from canopy spectral data. The approach uses synthetic data produced by the SCOPE model for pre training. Adversarial domain adaptation helps extract domain invariant features, reducing redundancy and improving physical consistency. Long Short Term Memory and Multilayer Perceptron networks are integrated for the inversion process. Tested on winter wheat in the Yellow River Basin, the model outperformed standard CNN and conventional transfer learning methods, especially in deeper soil layers. Determination coefficients (R2) reached 0.637, 0.613, 0.705, and 0.792 at 10, 20, 30, and 40 cm depths, respectively. Compared with conventional machine learning models, it exhibits significant superiority in predicting soil moisture at 40 cm depth, reducing normalized relative error by over 38.0%. To reveal underlying mechanisms, this study employs mutual information and SHAP analysis. Results indicate that in shallow layers, competitive interplay between canopy structure information (800 to 850 nm) and the weak near infrared water vapor absorption band (855 to 999 nm) mitigates noise. In contrast, deep layers shift to a stable synergistic pattern dominated by the 800 nm band, capturing signals of physiological status. Integrating physical models with deep transfer learning is an effective approach to addressing challenges of root zone soil moisture inversion in vegetated areas. The elucidated spectral modeling mechanism provides a solid basis for developing physically interpretable models for smart agricultural water management.

Soil moisture content (SMC) is critical for crop growth and precision irrigation. Multi-angular remote sensing can capture directional reflectance variations that may provide additional information related to crop water status. However, the potential of hotspot-sensitive multi-angular observations for root-zone SMC estimation remains insufficiently explored, particularly in winter wheat. This study evaluated the potential of hotspot-sensitive multi-angular observations for root-zone SMC estimation using UAV observations and the LESS (LargE-Scale remote sensing data and image Simulation framework) radiative transfer model. Multispectral imagery of winter wheat was acquired under four irrigation treatments (T1–T4). The LESS model was parameterized using field-measured canopy characteristics to simulate bidirectional reflectance factors (BRFs) across the observation hemisphere. The simulated and observed datasets were used to analyze hotspot dynamics and evaluate the contribution of hotspot-sensitive observations to SMC estimation. Among the investigated bands, hotspot characteristics at 560 nm exhibited the strongest response to water stress. LESS-simulated BRFs showed greater angular variability (CV ≈ 0.30) than UAV-observed BRFs (CV ≈ 0.18). The highest SMC estimation performance was obtained in hotspot-sensitive backscattering directions, with an R2 ≈ 0.60 and an RMSE ≈ 0.022 m3 m−3. These findings indicate that hotspot-sensitive multi-angular observations provide useful complementary information for SMC estimation.

Canopy temperature (Tc) is an important indicator for characterizing crop water status and serves as the core variable for constructing the Crop Water Stress Index (CWSI). Timely and accurate diagnosis of crop water stress is of great significance for precision irrigation and yield improvement. Owing to its non-contact and high-efficiency characteristics, unmanned aerial vehicle (UAV) remote sensing has become an effective approach for high-spatiotemporal-resolution monitoring of crop water conditions. However, variations in observation geometry can introduce thermal directional effects in canopy temperature, thereby reducing the stability and reliability of CWSI estimation. In this study, multi-angular thermal infrared imagery acquired by a UAV platform was utilized to investigate the directional characteristics of winter wheat canopy temperature. A kernel-driven model was employed to separate the directional components of canopy temperature and retrieve isotropic temperature parameters that more closely represent the actual thermal status of the crop canopy. Based on these temperature parameters, three CWSI models were constructed and evaluated for crop water stress diagnosis. The results demonstrated that (1) winter wheat canopy temperature exhibited pronounced directional characteristics, and the observed temperature generally decreased with increasing relative azimuth angle between the viewing direction and solar incident direction; (2) after angular correction, the isotropic canopy temperature simulated by the kernel-driven model showed an improved correlation with soil moisture content at a depth of 30 cm (R2 = 0.54); and (3) when angular-corrected canopy temperature was used as the input variable for different CWSI models, the sensitivity of all models to crop water variation was substantially enhanced, resulting in improved discrimination among different irrigation treatments. Among the evaluated approaches, the empirical CWSI model achieved the best performance in diagnosing crop water stress variations (R2 = 0.73, RMSE = 1.59%). These findings provide a theoretical basis for UAV-based thermal infrared remote sensing of crop water status and offer technical support for precision irrigation management.

As climate warming intensifies extreme weather, efficient agricultural water resource management is crucial. The unit crop water footprint (CWF), defined as the total volume of freshwater consumed and contaminated per unit harvested crop yield, serves as a key metric for evaluating agricultural freshwater appropriation. However, its accurate quantification at regional scales remains challenging due to scale effects, spatial heterogeneity, and limitations inherent in existing approaches. This study quantifies the regional winter wheat CWF in Shijiazhuang, a representative winter wheat-dominated agricultural region of the Huang–Huai–Hai Plain in North China, using a parameter optimization-based data assimilation framework that integrates multi-source optical and radar remote sensing with a process-based crop growth model. Specifically, the CERES-wheat model is constrained through multi-variable parameter optimization using time series of leaf area index and surface soil moisture, with parameters optimized via the shuffled complex evolution-University of Arizona algorithm, to improve grid-scale estimates of winter wheat CWF. This approach overcomes the limitations of single-method frameworks, enhances spatial resolution, and captures regional heterogeneity in water footprint patterns. Results indicated that the winter wheat CWF ranged from 0.52 to 1.21 m ^3 kg ^−1 , with higher values observed in the western and central parts of the study area and lower values in the northeastern and southeastern regions. Blue and green water footprints exhibited contrasting spatial distributions, with blue water predominating in the southern areas and green water dominating in the northern areas. This study reveals the spatial patterns and heterogeneity of the regional winter wheat water footprint and provides a robust methodological basis for advancing the understanding of agricultural water use during crop growth.

Accurate quantification of evapotranspiration (ET) is crucial for agricultural water management and climate change adaptation, especially in global warming and extreme climate events. Despite the availability of various ET products, their applicability across different scales and climatic conditions has not been comprehensively verified. This study evaluates nine ET products at grid, basin, and site scales in China from 2003 to 2014 under varying climatic conditions, including extreme temperatures, vapor pressure deficit (VPD), and drought. The main results are as follows: (1) At the grid scale, all products except the MODIS/Terra Net Evapotranspiration 8-Day L4 Global 500m SIN Grid (MOD16A2) product showed high consistency, with the Global Land Evaporation Amsterdam Model V4.2a (GLEAM) product exhibiting the highest comparability. The three-cornered hat (TCH) method revealed that GLEAM and the Synthesized Global Actual Evapotranspiration Dataset (Syn) had low uncertainties in multiple basins, while the Reliability Ensemble Averaging (REA) product and Penman–Monteith–Leuning Evapotranspiration V2 (PMLv2) product had the smallest uncertainties in the Songhua River and Hai River Basins. (2) At the basin scale, ET products were closely aligned with water-balance-based ET (WB-ET), with GLEAM achieving the smallest root mean square error (RMSE) (22.94 mm/month). (3) At the site scale, accuracy decreased significantly under extreme climatic conditions, with the coefficient of determination (R2) dropping from about 0.60 to below 0.30 and the mean absolute error (MAE) increasing by 110.30% (extreme high temperatures) and 101.40% (extreme high VPD). Drought conditions caused slight instability in ET estimations, with MAE increasing by approximately 12.00–40.00%. (4) Finally, using a small number of daily ET products as inputs for machine learning models, such as random forest (RF), greatly improved ET estimation, with R2 reaching 0.91 overall and 0.81 under extreme conditions. GLEAM was the most important product for RF in ET estimation. This study provides essential guidance for selecting and improving ET products to enhance agricultural water-use efficiency and sustainable irrigation.

Crop water deficit indicators such as crop water stress index (CWSI), actual crop evapotranspiration (ET), and stomatal conductance (gs) are widely utilized for soil water content (SWC) monitoring. However, time-lag effects between canopy temperature (Tc) and environmental factors can influence their correlation with SWC, thereby complicating the identification of the most reliable diagnostic indicator. This study conducted a two-year field experiment on winter wheat under four irrigation levels (80-95 %, 65-80 %, 50-65 %, and 40-50 % field capacity). Time-lag cross-correlation, time-lag mutual information, grey time-lag correlation analysis, time-lag Almon, and time-lag partial least squares (PLS) were applied to calculate the time-lag parameters. These timelag parameters were subsequently used to correct the correlations between CWSI, ET, gs, and SWC. The indicator with the strongest correlation to SWC was selected and then predicted using four machine learning models. Results demonstrated that time-lag correction significantly enhanced the correlation between SWC and theoretical CWSI, empirical CWSI, gs, and ET, with increases of 0.15, 0.33, 0.11, and 0.21, respectively; Time-lag mutual information exhibited the highest effectiveness in correcting time-lag effects; The sudden decline in gs and the peak advancement in severe water stress treatments led to abrupt changes in time-lag parameters; The Convolutional Neural Network-Bidirectional Long Short-Term Memory-Adaptive Boosting model achieved the highest accuracy in predicting gs corrected by time-lag mutual information from 8:00-15:00 (R2=0.96). These results provided a theoretical foundation for accurately assessing soil moisture conditions in agricultural fields and contributed to advancing water conservation techniques in arid farmland.

Canopy chlorophyll content (CCC) is a vital indicator of crop growth. Timely, efficient and non-destructive estimation of CCC is crucial for field management. However, current methods for constructing CCC inversion models based on orthophoto information face challenges, such as limited accuracy and inadequate information retrieval. To address these problems, this study adopted multi-angle image information. During the main growth period of winter wheat (February to May), CCC measurements and high-resolution multi-angle multispectral remote sensing images were collected. The angle-insensitive spectral indices were screened by correlation and deviation analysis. Using these spectral indices and measured CCC date, four CCC estimation models were constructed: Partial Least Squares Regression (PLSR), Random Forest (RF), Extreme Gradient Boosting (XGBOOST) and Particle Swarm Optimization-Backpropagation Neural Network (PSO-BP). The results were as follows: (1) Two-band spectral indices (SI) exhibited a stronger correlation with CCC and higher angle applicability. In particular, the near infrared band was the least affected by angle variations; (2) Among 13 observation angles, the overall performance of the four algorithms were basically consistent: the observation angles at the backward of the Solar Principal Plane (SPP) consistently outperformed others, with the optimal observation angle identified at the hotspot (VAA2-45 degrees); (3) The PSO-BP model demonstrated the best performance and overall stability, achieving the highest accuracy at the optimal angle (R2 = 0.91, RMSE = 0.18 mg/g). These results provide valuable insights into selecting optimal observation angles and constructing CCC estimation models using UAVbased multi-angle multispectral remote sensing images.

Existing remote sensing approaches for estimating root zone soil moisture are limited by their dependence on initial conditions, sensitivity to model parameters, and high computational costs. This study proposes a lightweight model for predicting root zone soil moisture at the irrigation district scale. The model is developed based on the soil water balance equation and incorporates multi-source remote sensing data. A random forest algorithm is employed as the core predictive framework. The model is validated in the Jiefangzha Irrigation District. Results show: (1) The model achieves satisfactory accuracy, with site-level R values of 0.43-0.72 and RMSE of 0.007-0.01; for scattered locations, R values range from 0.53 to 0.66 and RMSE from 0.005 to 0.01; (2) Downscaling methods effectively resolve spatial scale mismatches, allowing substitution of water balance equation features and high-resolution simulations. Downscaling errors range from 12.56 %-16.60 % for RH, 3.18-3.61 mm for PET, 0.03-0.05 for kNDVI, 1.76-4.74 degrees C for LST, and 0.08-0.11 m(3) /m(3) for SSM; (3) Annual average AWF in 2018 and 2019 remains stable at similar to 0.12, with daily variations mainly from late July to early September; (4) Initial soil moisture has minor impact on long-term simulations, with convergence after similar to 40 days; (5) The relative importance of influencing factors is: AWFt-1, SMAPt-1, RH, kNDVI, SMAPt, precipitation, and PET. The proposed model reduces sensitivity and computational burden, enabling accurate root zone soil moisture prediction at the irrigation district scale.

Global navigation satellite system (GNSS) interferometric reflectometry (GNSS-IR) technology can realize continuous and dynamic monitoring of soil moisture at areas. The GNSS signal is highly susceptible to the influence of external factors, causing anomalous terms in the process of characteristic parameters extraction. In this study, an integrated outlier detection method is proposed. This method first detects outlier based on Inter-Quartile Range, Grubbs test and Hampel filter outlier detection methods for the characteristic parameters data, and derives the outlier location by complementing the detection results with each other and corrected by rainfall data. Then, four kinds of outlier repair methods, namely, moving average, locally weighted primary linear regression, locally weighted quadratic linear regression, and robust locally weighted quadratic linear regression, were used to repair and filter out the optimal data according to the location of the outlier. Finally, three machine learning methods, namely, Bagging Tree, Support Vector Machine (SVM), and Gaussian Process Regression (GPR), were combined to build a soil water content inversion model and evaluate the accuracy. The results show that the proposed comprehensive outlier detection method in this study can accurately detect the location of outlier. The correlation between the characteristic parameters and in-situ soil moisture can be effectively improved after remediation treatment. The outlier repair helped to improve the inversion accuracy of the soil moisture inversion model, and the R2 of the soil moisture inversion model increased by 13.21 % to 27.08 % (mean 18.25 %), the RMSE decreased by 12.97 % to 23.61 % (mean 18.16 %). The comprehensive outlier detection and repair method proposed in this study can provide reference for the quality control of input data before the establishment of GNSSIR model, and effectively improve the inversion accuracy of GNSS-IR soil moisture inversion model. (c) 2024 COSPAR. Published by Elsevier B.V. All rights are reserved, including those for text and data mining, AI training, and similar technologies.

Timely and accurate assessment of crop water status using unmanned aerial vehicle (UAV) imagery is helpful for precision irrigation and field management. The aim of this study is to investigate the application potential of continuous wavelet transform (CWT)-based hyperspectral combined with thermal infrared image data for the estimation of leaf water content (LWC) in winter wheat. This study evaluates the performance of convolutional neural networks (CNN) for feature extraction and long short-term memory (LSTM) networks for sequential data processing in LWC estimation. A UAV platform carrying hyperspectral and thermal infrared sensors was used to collect high spatial resolution images of winter wheat under different water treatments over two years. The LWC was collected simultaneously. The original (OR) and CWT-transformed canopy spectral and textural features, as well as canopy temperature indicators, were extracted from the UAV-based images. On this basis, the LWC estimation model was established using the CNN and LSTM model. The results showed that the combination of thermal features with spectral and texture features significantly improves model performance compared to models built on a single data. The CWT-transformed spectral features improved LWC estimation compared to the original spectrum, with the third scale (CWT3) yielding the best results. Moreover, the CWT-transformed texture at multi-decomposition scales proved to be effective for estimating LWC. Compared to other models, the LSTM model (T-STCWT3-LSTM), built by thermal feature fusion with CWT3-based spectral and texture features, achieved the best LWC estimation results, with R2 of 0.827 and 0.836, RMSE of 2.575 % and 1.822 %, and MAE of 2.041 % and 1.434 % for 2022 and 2023, respectively. In addition, the robustness of the T-STCWT3-LSTM model was successfully verified at different growth stages. Overall, the CWT technique and multi-feature fusion approach provide a valuable technical reference for real-time crop water status monitoring, supporting improved precision irrigation practices and sustainable crop management.

This study aims to evaluate crop water status by fusing multiple features from the unmanned aerial vehicle (UAV)-based canopy images with model updating strategy. A UAV platform carrying multispectral and thermal infrared cameras was used to collect high spatial resolution images of winter wheat and summer maize under different water treatments over two years. The plant water content (PWC) and above-ground biomass (AGB), which represent crop water status, were collected simultaneously. The vegetation indices (VIs), texture features, and canopy thermal indicators were extracted from UAV-based images to estimate PWC and AGB based on CNNLSTM-Attention (CLA) model. The results showed that combining spectral, textural, and thermal features with the CLA model significantly improved estimation accuracy. Specifically, multi-feature fusion achieved the best performance in winter wheat, with MAE of 1.80 % and 1.23 %, and RMSE of 2.13 % and 1.57 % for PWC in 2022 and 2023, respectively. For AGB, the corresponding MAE values were 1.12 t/hm2 and 1.04 t/hm2, and RMSE values were 1.41 t/hm2 and 1.31 t/hm2. In addition, the model updating strategy successfully verified the robustness of the estimation model for winter wheat across different years, and the application of the CLA model to summer maize demonstrated its effective transferability. In summary, this method can improve the estimation accuracy of PWC and AGB, thereby achieving efficient evaluation of crop water status.

With the advancement of precision agriculture, Unmanned Aerial Vehicle (UAV)-based remote sensing has been increasingly employed for monitoring crop water and nutrient status due to its high flexibility, fine spatial resolution, and rapid data acquisition capabilities. This review systematically examines recent research progress and key technological pathways in UAV-based remote sensing for crop water and nutrient monitoring. It provides an in-depth analysis of UAV platforms, sensor configurations, and their suitability across diverse agricultural applications. The review also highlights critical data processing steps—including radiometric correction, image stitching, segmentation, and data fusion—and compares three major modeling approaches for parameter inversion: vegetation index-based, data-driven, and physically based methods. Representative application cases across various crops and spatiotemporal scales are summarized. Furthermore, the review explores factors affecting monitoring performance, such as crop growth stages, spatial resolution, illumination and meteorological conditions, and model generalization. Despite significant advancements, current limitations include insufficient sensor versatility, labor-intensive data processing chains, and limited model scalability. Finally, the review outlines future directions, including the integration of edge intelligence, hybrid physical–data modeling, and multi-source, three-dimensional collaborative sensing. This work aims to provide theoretical insights and technical support for advancing UAV-based remote sensing in precision agriculture.

Optical satellites with increasing spatiotemporal resolution show significant potential and advantages in soil moisture (SM) retrieval at field scales. However, the accuracy and applicability of SM estimation based on optical methods are relatively poor in complex heterogeneous regions such as crop-soil mixed. To overcome this limitation, we acquired unmanned aerial vehicle (UAV) spectral data, fractional vegetation cover (FVC), and in-situ SM data at four crop growth stages (emergence, jointing, tasseling, and maturity). We proposed a linear spectral reconstruction (LSR) method, which has the particular advantage of improving optical satellite data on a larger scale with only limited auxiliary data, including representative UAV data, FVC, and in-situ SM data. The LSR method was tested on Landsat-8 satellite spectral data, including reflectance (RGB, NIR, SWIR1 and SWIR2) and multiple spectral indices. The grey correlation degree was employed to screen out the reflectance and spectral index strongly correlated with SM, and the back-propagation neural network was adopted to construct SM inversion models at different crop growth stages. The results showed that the LSR method can significantly enhance the correlations of reflectance and spectral index with SM, thereby improving the SM estimation accuracy, particularly at the jointing and tasseling stages with medium vegetation cover. In the reconstruction of the reflectance, SWIR1 was best, SWIR2 and RGB were overall comparable, and NIR was relatively worst. The reconstruction of the non-ratio-based indices outperformed that of the ratio-based indices overall. Compared with SM inversion models constructed with original satellite spectral data, those using the reconstructed satellite spectral data showed higher accuracy. At the four crop growth stages, the coefficient of determination (R2) of the validation set of inversion models improved from 0.438 to 0.593 (an increase of 35.4%) on average, and the mean absolute error (MAE) decreased from 2.32% to 2.00% (a decrease of 13.8%). The LSR method can provide a theoretical reference to enhance the accuracy of SM estimation based on optical methods at varying vegetation covers.

Leaf stomatal conductance (Gs) is an important indicator for measuring crop water stress. Influenced by variation of environmental conditions and growth stages of crops, achieving the reliable and accurate Gs estimation by UAV image is of challenge. Therefore, this study aimed to explore the potential of Gs estimation of winter wheat by UAV-based multispectral imagery based on coupling meteorological factors with the PROSAIL model. Firstly, we set up field experiments with different moisture treatments, acquired the canopy images of winter wheat at different fertility stages using the UAV equipped with a multispectral camera, and acquired meteorological factors (MFs) synchronously. Next, we collected leaf chlorophyll content (Cab), leaf area index (LAI), canopy chlorophyll content (CCC) and Gs. Then, we used PROSAIL model and machine learning models to estimated Gs from UAV-based multispectral images, and the estimation results of Gs at different growth stages were evaluated by coupling MFs. The results showed that, (1) the PROSAIL model successfully retrieved Cab, LAI, and CCC from UAV-based multispectral images, with rRMSE of 0.109, 0.136, and 0.191 respectively, (2) the Cab, LAI and CCC retrieved by PROSAIL model performed well to estimate Gs, with rRMSE of 0.166, 0.150 and 0.130, respectively, (3) the coupling of meteorological factors with the retrieved Cab, LAI, and CCC further enhanced the estimation accuracy of Gs, which is comparable to the results obtained with machine learning models, importantly. The proposed method also enhanced the robustness of estimating Gs at different growth stages. In conclusion, the potential of the Gs estimation with UAV-based multispectral images was proved through the PROSAIL model coupled with meteorological factors, which also provided a technical reference and idea for the assessment of crop water stress.

Transpiration is the dominant process driving water loss in crops, significantly influencing their growth, development, and yield. Efficient monitoring of transpiration rate (Tr) is crucial for evaluating crop physiological status and optimizing water management strategies. The three-temperature (3T) model has potential for rapid estimation of transpiration rates, but its application to low-altitude remote sensing has not yet been further investigated. To evaluate the performance of 3T model based on land surface temperature (LST) and canopy temperature (TC) in estimating transpiration rate, this study utilized an unmanned aerial vehicle (UAV) equipped with a thermal infrared (TIR) camera to capture TIR images of summer maize during the nodulation-irrigation stage under four different moisture treatments, from which LST was extracted. The Gaussian Hidden Markov Random Field (GHMRF) model was applied to segment the TIR images, facilitating the extraction of TC. Finally, an improved 3T model incorporating fractional vegetation coverage (FVC) was proposed. The findings of the study demonstrate that: (1) The GHMRF model offers an effective approach for TIR image segmentation. The mechanism of thermal TIR segmentation implemented by the GHMRF model is explored. The results indicate that when the potential energy function parameter β value is 0.1, the optimal performance is provided. (2) The feasibility of utilizing UAV-based TIR remote sensing in conjunction with the 3T model for estimating Tr has been demonstrated, showing a significant correlation between the measured and the estimated transpiration rate (Tr-3TC), derived from TC data obtained through the segmentation and processing of TIR imagery. The correlation coefficients (r) were 0.946 in 2022 and 0.872 in 2023. (3) The improved 3T model has demonstrated its ability to enhance the estimation accuracy of crop Tr rapidly and effectively, exhibiting a robust correlation with Tr-3TC. The correlation coefficients for the two observed years are 0.991 and 0.989, respectively, while the model maintains low RMSE of 0.756 mmol H2O m−2 s−1 and 0.555 mmol H2O m−2 s−1 for the respective years, indicating strong interannual stability.

Accurate soil moisture data with detailed spatial and temporal resolutions are essential for hydrological modeling, precision agriculture, and climate research. Nonetheless, the intrinsic trade-off between spatial and temporal resolution in remote sensing limits the accessibility of soil moisture products at granular scales. This study presents a spatiotemporal fusion algorithm utilizing Fourier transform (STFFT), integrated with Random Forest (RF), the Water Cloud Model (WCM), and the radiative transfer model (PROSAIL) to create a comprehensive framework for downscaling surface soil moisture (SSM). Employing Sentinel-1 and Sentinel-2 datasets, we downscaled Soil Moisture Active and Passive (SMAP) soil moisture products to generate daily Soil Surface Moisture (SSM) maps at a 20-meter spatial resolution for the study area. The findings indicate that STFFT is more adept at accommodating SSM data marked by significant heterogeneity and scale discrepancies compared to traditional spatiotemporal fusion algorithms. Furthermore, STFFT exhibits computational efficiency and is independent of reference image selection. The amalgamation of RF with WCM and PROSAIL adeptly elucidates the intricate correlations between remote sensing variables and soil moisture; the suggested framework attains precise soil moisture mapping, evidenced by an average correlation coefficient (R) of 0.892 and a root mean square error (RMSE) of 0.034 m3/m3 across diverse land cover types. Compared to benchmark methods that produce an average R of 0.753 and an RMSE of 0.043 m3/m3, STFFT demonstrates markedly enhanced accuracy and robustness, particularly in heterogeneous terrains. This study introduces an improved methodology for producing fine-scale soil moisture products characterized by enhanced spatiotemporal continuity and reliability.

Soil moisture is a key indicator for studying the exchange of energy and matter in the soil-plant-atmosphere circulation system. With the rapid development of GNSS base station networks, the application of GNSS-IR technology in the field of soil moisture monitoring is further promoted, based on the characteristics of GNSS satellites and microwaves (L-band). The global navigation satellite system (GNSS) interferometric reflectometry (GNSS-IR) technology can realize high-precision, continuous, dynamic, and real-time monitoring of soil moisture. However, the conventional GNSS-IR monitoring technology cannot fully consider the differences and complementarity of satellite signals in different frequency bands owing to the use of single-frequency band data processing. In this study, the Helmert variance component estimation (HVCE) fusion method is proposed to perform weight fusion processing of the L1 and L2 frequency band characteristic parameters data of GNSS satellites. This method is based on the basic principle of posterior variance weighting, which can satisfactorily ensure the rationality of weight determination. To verify the effectiveness of the proposed method, four models, namely linear, multilayer perceptron (MLP), support vector machine (SVM), and Gaussian process regression (GPR), were used to construct the soil moisture inversion models before and after fusion. Then, the accuracies were compared with those of the entropy fusion method. The results show that the HVCE fusion method can greatly improve the correlation between characteristic parameters and in-situ soil moisture, and the amplitude and phase of nine satellites coefficient of determination (R2) increased by 53.70 % and 59.70 % on average than before fusion, which is 56.86 % higher than that of entropy method on average. Second, the HVCE fusion method can effectively improve the modeling accuracy. The GPR model achieves the highest accuracy, with its correlation coefficient (R) and root mean square error (RMSE) values improving by 14.68 % and 21.43 %, respectively. The SVM model is slightly less accurate than the GPR model, with its R and RMSE improving by 15.23 % and 15.35 %, respectively. The MLP model ranks third among the four models, whereas the linear model has the worst performance. In addition, compared with the conventional entropy fusion method, the proposed HVCE fusion method improves the R and RMSE accuracy by 19.50 % and 19.56 %, respectively. Thus, the proposed method can provide an effective solution to the problem of reasonable weighting fusion of GNSS-IR dual-frequency data. This study can be a reference for the application area of multi-source data fusion in hydrogeodesy and water resources monitoring.

Nitrogen is a fundamental component for building amino acids and proteins, playing a crucial role in the growth and development of plants. Leaf nitrogen concentration (LNC) serves as a key indicator for assessing plant growth and development. Monitoring LNC provides insights into the absorption and utilization of nitrogen from the soil, offering valuable information for rational nutrient management. This, in turn, contributes to optimizing nutrient supply, enhancing crop yields, and minimizing adverse environmental impacts. Efficient and non-destructive estimation of crop LNC is of paramount importance for on-field crop management. Spectral technology, with its advantages of repeatability and high-throughput observations, provides a feasible method for obtaining LNC data. This study explores the responsiveness of spectral parameters to soybean LNC at different vertical scales, aiming to refine nitrogen management in soybeans. This research collected hyperspectral reflectance data and LNC data from different leaf layers of soybeans. Three types of spectral parameters, nitrogen-sensitive empirical spectral indices, randomly combined dual-band spectral indices, and “three-edge” parameters, were calculated. Four optimal spectral index selection strategies were constructed based on the correlation coefficients between the spectral parameters and LNC for each leaf layer. These strategies included empirical spectral index combinations (Combination 1), randomly combined dual-band spectral index combinations (Combination 2), “three-edge” parameter combinations (Combination 3), and a mixed combination (Combination 4). Subsequently, these four combinations were used as input variables to build LNC estimation models for soybeans at different vertical scales using partial least squares regression (PLSR), random forest (RF), and a backpropagation neural network (BPNN). The results demonstrated that the correlation coefficients between the LNC and spectral parameters reached the highest values in the upper soybean leaves, with most parameters showing significant correlations with the LNC (p < 0.05). Notably, the reciprocal difference index (VI6) exhibited the highest correlation with the upper-layer LNC at 0.732, with a wavelength combination of 841 nm and 842 nm. In constructing the LNC estimation models for soybeans at different leaf layers, the accuracy of the models gradually improved with the increasing height of the soybean plants. The upper layer exhibited the best estimation performance, with a validation set coefficient of determination (R2) that was higher by 9.9% to 16.0% compared to other layers. RF demonstrated the highest accuracy in estimating the upper-layer LNC, with a validation set R2 higher by 6.2% to 8.8% compared to other models. The RMSE was lower by 2.1% to 7.0%, and the MRE was lower by 4.7% to 5.6% compared to other models. Among different input combinations, Combination 4 achieved the highest accuracy, with a validation set R2 higher by 2.3% to 13.7%. In conclusion, by employing Combination 4 as the input, the RF model achieved the optimal estimation results for the upper-layer LNC, with a validation set R2 of 0.856, RMSE of 0.551, and MRE of 10.405%. The findings of this study provide technical support for remote sensing monitoring of soybean LNCs at different spatial scales.