Allocating water resources can greatly contribute to the safe and stable operation of irrigation canal systems in order to improve the management level. This study aims to predict the hydraulic transition of the unsteady flow in irrigation canal systems caused by gate regulation. There was a hydraulic response of canal sections under different flow control conditions. Taking a trapezoidal branched experimental canal as the research object, the Preissmann four-point implicit finite difference and the chasing methods were used to discretize and solve the Saint-Venant equations. A one-dimensional unsteady flow mathematical model was established for the open channels with complex internal boundaries using MATLAB programming language. 10 test working conditions were designed to verify the model accuracy. For instance, the different canal head flow rates and gate regulation modes were set on the branched canal for the original physical model. The water depth variation at 12 measuring points along the canal was obtained using an ultrasonic measurement. The spatial step size of the numerical model was determined to be 0.25 m after grid independence verification. Combined with the physical experiments and HEC-RAS software simulations, the accuracy and precision of the simulation model and solution were systematically verified from three dimensions: the stable along- the - way distribution of canal water levels after gate regulation, the time - series water level changes at each measuring point, and the dynamic process of water levels around Gate I. Multiple working condition was then simulated to verify the reliability of the model. There were quantitative mechanisms and systematic interaction between gate regulation parameters and hydraulic response indicators. Once the incoming flow at the canal head was stable, particularly for every 1cm increase in gate opening (corresponding to a 5% relative variation in opening), the descending rate of the water level in front of the gate increased by approximately 6%, the maximum water level drop increased by about 0.93 cm, and the hydraulic response time prolonged by around 8.54%. Three parameters shared a significant positive correlation with the gate regulation amplitude. Although the maximum drop of the water level during regulation was related to the variation of the gate opening, the relative variation amplitude of the water level was outstandingly smaller than that of the opening. The water diversion was transmitted to raise the water level. A normal distribution was observed in the relationship between the water level variation amplitude and the water level rise stabilization time. An optimal regulation range was determined to minimize the canal water level fluctuation and the transition time. In addition, the water level at the canal outlet shared a trend of first rising and then falling when the gate opening increased. The optimal regulation range simultaneously enhanced the water level fluctuation and stabilization efficiency. Moreover, the diversion ratio was dominated by both the incoming flow and gate regulation. Especially under high-flow working conditions, there was a more significant intervention of gate regulation on the diversion ratio. The simulation model can provide rich data support for the gate group scheduling. Numerical simulation also shared the important practical value to explore the gate regulation on the water level amplitude and flow velocity. Effective scheduling strategies were formulated in the gate group. The finding can also provide the datasets and technical support on the joint scheduling of segmented gates.

Accurate measurement of irrigation water can enhance the water use efficiency of canal systems in smart agriculture. Conventional flow measurement cannot fully meet the large-scale production in recent years, due to the high cost and low efficiency in complex environments. While existing vision techniques have struggled with the illumination variations, they still lack the natural tracers. In this study, a non-contact approach was presented to measure the surface flow velocity in open channels, termed feature matching velocimetry (FMV), using monocular vision. The biodegradable artificial foam was introduced as a high-visibility and eco-friendly tracer in order to overcome the scarcity of discernible features in the concrete-lined irrigation channels. Feature recognizability was significantly enhanced on the water surface. The ORB features were extracted from the consecutive video frames and then matched using the FLANN algorithm. A spatial constraint filtering mechanism was implemented to eliminate the erroneous matches, thus improving matching accuracy. A zonal processing strategy was used to divide the field of view into three regions for independent analysis, in order to mitigate the adverse effects of uneven illumination (e.g., glare, reflections, and shadows). The optimal displacement was estimated for the surface velocity calculation. Gaussian curve fitting was applied to the histogram of the motion distances that accumulated from thousands of matched feature point pairs across multiple frames. The refined displacement was combined with the camera calibration parameters (intrinsic matrix, camera height Z above water) and frame rate. The precise inversion of the surface velocity was realized after the combination. The surface velocity coefficient (using η_s = 0.9 for concrete channels) was then employed to estimate the cross-sectional discharge, according to the measured center-region surface velocity and flow area. Comprehensive field validation was conducted on both rectangular (0.5 m wide) and trapezoidal (0.3 m base, slope 1:1) concrete-lined channels under diverse conditions: seven flow rates (30-60 L/s), and three illumination scenarios (sunny, cloudy, nighttime with LED supplemental lighting), totaling 42 test conditions. A high-definition camera (60 frame/s) was used to record the foam-tagged flow. A propeller current meter was provided to measure the ground-truth velocity. The experimental results demonstrate that superior accuracy and robustness were achieved in measuring the flow velocity. There was a mean absolute error (MAE) of 0.026 m/s and a mean relative error (MRE) of 3.79%. The better performance was achieved over the spatiotemporal image velocimetry (STIV) under all test conditions, particularly under challenging illumination (e.g., reducing MRE to 3.00% compared to STIV's 16.09% in rectangular channels under cloudy conditions). The zonal processing strategy effectively balanced the uneven feature point distribution caused by illumination variations. In discharge estimation, the surface velocity coefficient (η_s=0.9) was calculated with excellent agreement with the electromagnetic flowmeter measurements, indicating MRE below 3.6% (3.59% for rectangular and 3.37% for trapezoidal channels) with the maximum relative errors below 7.2%. The FMV model with the artificial foam tracers also offered some advantages: low cost, non-intrusive measurement, strong resistance to illumination interference, suitability for all-weather operation, and ease of deployment using existing surveillance infrastructure. It effectively solved the limitations on the natural textures or specific lighting inherent in the vision-based techniques. This finding can provide a reliable, practical, and cost-effective technical solution to continuous and accurate flow monitoring in the small-to-medium-sized irrigation districts, thus facilitating their transition towards dynamic, precise, and intelligent water management. Future work will focus on the validation in the real-world irrigation canals and extension to more complex flow regimes.

Sediment deposition has posed a serious risk to the performance of emitter flow channels. This study aims to explore the influences of the emitter performance on sediment deposition, in order to reduce the risk of physical blockage occurrence. Different concentrations (1.8, 2.8, and 3.8 g/L) and particle sizes (0 ≤ d ＜ 0.054 mm, 0.054 mm ≤ d ＜ 0.075 mm, 0.075 mm ≤ d ＜ 0.1 mm) of sandy water were used in the emitter clogging test of intermittent drip irrigation. A comparison was also made on the particle size composition in the sediment output from the emitters that irrigated with the sandy water. A systematic analysis was then implemented for the sediment transport and siltation inside the emitters. A numerical simulation was also carried out to analyze the flow pattern in the emitter. The low-velocity or vortex zone was prone to clogging. An operation mode was finally recommended to improve the clogging-resistant performance of the emitters in the drip irrigation system, according to the flow pattern and the deposited sediments. The results show that: 1) There was a consistent effect of side-seam labyrinth emitters on the size of settled sediment, with a decrease in the specific gravity of clay and sand particles, and the increasing specific gravity of powder particles. The grain size composition of the input original sediment was compared with the output one. There was a decreased proportion of clay and sand particles with an increased proportion of silt particles. Since the water-following property of sand particles was weaker than that of clay particles and silt particles, there was a weak discharge from the emitter with water. Specifically, the proportions of clay particles and sand particles in the sediment decreased by 1.14-2.98 percentage points and 1.55-11.14 percentage points, respectively, whereas, the proportion of silt particles increased by 1.25-13.74 percentage points. 2) The low transport of large particle size was attributed to the small size of the labyrinth emitter flow channel. The clogging of the particle size was concentrated in the range of 0.054-0.1 mm particle size. Most sediment sizes were between 0.002 and 0.05 mm in the flow channel of the emitter, due to the agglomeration of sediment flocs. 3) Streamline in the mainstream area was moving forward in a wave-like manner. The flow velocity in the mainstream area was greater than that in the near-wall area, while the flow velocity at the central corner was large, and a vortex was generated in the upper and lower corners. The inner and outer water bodies in the centrifugal force and water pressure under the joint action of the inner water body flew to the lower wall of the lower runner unit, while the outer water body flew to the upper side wall of the upper runner unit, where the turbulence of the water flow diffusion rate was faster. A vortex area was formed in the inner and outer boundaries of the circulation in the labyrinth runner. Part of the particles was retained in the vortex area, thus causing the runner blockage. That was the main factor for the emitter clogging. After that, sand particles were very easy to settle and concentrate in the vortex center and the water surface. Some suggestions were proposed to reduce the number of right- and sharp-angled areas, in order to reduce the formation of low-velocity zones in the design optimization. The clogging of emitters can be prevented by the sand deposition. The long and narrow over-flow structures can also be avoided, according to the location of the clogging of sediments.

Leaf Area Index (LAI) is one of the most crucial influencing parameters on crop photosynthesis during growth status. There is a high demand for the timely and accurate monitoring of maize LAI in the high crop productivity. The purpose of this research was to integrate the Unmanned Aerial Vehicle (UAV) spectral, thermal, and morphological parameters of crop canopy, in order to improve the accuracy of maize LAI estimation at various growth stages. Moreover, an attempt was also made to create the LAI inversion maps for summer maize. The optimal estimation model was achieved to facilitate precise water and nitrogen management. Firstly, the images of maize canopies were collected at multiple growth stages under different water and nitrogen treatments using multispectral and thermal infrared sensors carried by UAVs. The parameters of maize growth (LAI and crop height) were simultaneously measured in field experiments in 2022-2023. Secondly, the LAI estimation models were developed using canopy spectral, thermal infrared data, canopy morphological parameters, and their combinations. Partial Least Squares Regression (PLS), Backpropagation Neural Network (BP), and Random Forest (RF) were also selected using various machine learning algorithms. Finally, the LAI inversion maps at the in-situ scale were created using the optimal RF estimation model. It was found that the maize LAI responded significantly to the water and water-nitrogen treatments, suitable for crop growth monitoring. Among them, the LAI values were ranged from 0.61 to 4.92 (0.89 to 5.94) under the N6 (N6W2) treatment; The LAI values were ranged from 0.49 to 4.37 (0.65 to 4.08) under the N1 (N1W0) treatment. Vegetation index (VIs), normalized relative temperature (NRCT), and morphological parameters (CM) of the maize canopy showed stable correlations with the LAI. However, there were some limitations on a single source of information, in order to monitor the LAI of maize at multiple growth stages. The LAI estimation model with the spectral data shared the R² between 0.36 and 0.61, and RMSE between 0.09 and 0.57. The LAI estimation models with the thermal infrared data shared the R² between 0.25 and 0.48 and RMSE between 0.11 and 0.62. The LAI estimation models with the canopy morphology (plant height or plant height+canopy coverage) shared the R² between 0.31-0.58 and 0.50-0.67 and RMSE between 0.06-0.54 and 0.08-0.51. In contrast, the multiple sources of data (including canopy spectra, thermal infrared data, and canopy morphological parameters) were integrated to significantly improve the accuracy of LAI estimates, especially at the later stages of maize growth. The underestimation was avoided from a single source of information. Compared with the NRCT+CM and VIs+CM, the R² was improved by 15.19 % to 19.01 % and 3.87 % to 12.57 %, respectively, and RMSE was reduced by 24.93 % to 28.13 % and 9.51 % to 19.94 %, respectively. The RF model showed the highest estimation accuracy with R² = 0.814-0.867, RMSE = 0.065-0.276, and MAE = 0.048-0.207 among the three machine learning models. The inverse maps of in situ LAI with the RF model accurately reflected the water and nitrogen status of the crop. The ranges of LAI values for the different growth stages under the N1 treatment were 0.36-4.97 under the N1 treatment; The range of LAI values under the N6 treatment was 0.54-6.27. The range of LAI at different growth stages under the W0 treatment was 0.31-4.73, and the range of LAI at different growth stages under the W3 treatment was 0.39-5.84. The finding can provide a feasible technological pathway to monitor the growth status of the crop using the unmanned platform. The water and nitrogen management were optimized for the high application potential. This finding can provide a feasible technical pathway for the UAV-based monitoring of crop growth.