Method and system for estimating unit area irrigation water consumption based on multi-source data fusion

Abstract

The invention discloses a unit area irrigation water consumption estimation method and system based on multi-source data fusion, and relates to the field of irrigation water consumption estimation. Multi-source heterogeneous data such as remote sensing images, meteorological data, soil humidity and crop growth information are fused, and observation and model errors are reduced through data assimilation; estimating crop evapotranspiration under an energy balance principle; estimating daily-scale irrigation water consumption by adopting a water balance method; a crop actual irrigation area identification model based on deep learning is introduced, actual irrigation areas of different crop types are effectively distinguished, and the influence of temporal and spatial changes on irrigation efficiency is dynamically reflected. According to the method, high-precision and rapid estimation of the comprehensive unit area irrigation water consumption of a single crop and multiple crops on different regional scales can be realized, the timeliness and accuracy of irrigation water quota evaluation are remarkably improved, and the method has strong adaptability to climate change, soil difference and crop diversity and is suitable for popularization and application. Powerful technical support is provided for precise irrigation and intelligent water conservancy management.

Description

Technical Field

[0001] This invention relates to the field of irrigation water consumption estimation technology, and more specifically to a method and system for estimating irrigation water consumption per unit area based on multi-source data fusion. Background Technology

[0002] Currently, irrigation water usage is a crucial technological foundation for achieving water conservation in farmland, improving water resource utilization efficiency, and ensuring sustainable agricultural development. However, current irrigation water monitoring and usage determination methods still have significant shortcomings in terms of data acquisition timeliness, estimation accuracy, and adaptability to complex environmental conditions. Especially in regions with diverse climate types, soil conditions, and crop planting structures, the combined influence of multiple factors such as meteorology, soil moisture, and crop growth makes it difficult for traditional irrigation water usage determination methods to balance accuracy and timeliness. Therefore, developing a new method that can integrate multi-source heterogeneous data, improve the accuracy of irrigation water usage estimation per unit area, and adapt to complex and changing environments is of great significance for promoting precision irrigation management, improving water resource allocation efficiency, and ensuring food security.

[0003] The Guizhou Provincial Institute of Water Resources Science has published a method for determining irrigation water quotas in karst mountainous areas (CN2024108089318). This method divides the region into zones based on natural geography, hydrology, meteorology, and irrigation conditions, selecting typical sample plots. It establishes hydrological year patterns based on precipitation conditions and performs rainfall frequency analysis, selecting the most unfavorable year. Combining the water requirements of the region's main characteristic crops throughout their entire growth period with the rainfall process of representative years, it uses water balance analysis to derive the irrigation quota for the entire growth period, and calculates the basic quota by incorporating the water utilization coefficient of the mountainous water distribution system. This method is suitable for specific complex landforms, but its data sources mainly rely on field monitoring and statistical data, lacking the fusion of multi-source remote sensing, meteorological, and soil information, and its dynamic updating capability is insufficient. Furthermore, it is not flexible enough in responding to rapid changes in meteorological and crop conditions, limiting its scope of application. The Pearl River Water Resources Research Institute has published a remote sensing-based method for estimating total agricultural irrigation water consumption in irrigation districts (CN2019111267540). This method identifies irrigation districts and crop distribution through remote sensing imagery and calculates total water consumption by combining planting structure, crop water quotas, and effective precipitation. While this method improves spatial coverage and accuracy to some extent, the water quota is a static input parameter and is not dynamically optimized for the irrigation water quota itself. Coefficients such as precipitation utilization rate are mostly empirical values ​​and fail to be adjusted based on soil moisture, crop growth stage, and real-time meteorological changes. Remote sensing is only used for area and crop type identification, lacking the fusion of multi-dimensional data such as evapotranspiration and soil moisture, making it difficult to accurately reflect the actual water consumption process. Overall, while existing technologies have some practical value in specific application scenarios, they generally suffer from the following shortcomings: limited data dimensions, insufficient integration of multi-source heterogeneous information, lack of dynamic update mechanisms, insufficient timeliness in responding to meteorological and crop changes, and limited adaptability in complex, multi-factor environments. These problems limit the widespread and efficient application of unit area irrigation water consumption estimation in modern precision irrigation and smart water management.

[0004] Therefore, how to achieve high-precision and rapid estimation of irrigation water consumption per unit area for single crops and multiple crops in regions of different scales, thereby significantly improving the timeliness and accuracy of irrigation water quota assessment, is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a method and system for estimating irrigation water consumption per unit area based on multi-source data fusion. This method integrates heterogeneous data from multiple sources, including remote sensing imagery, meteorological data, soil moisture, and crop growth information. It utilizes data assimilation techniques to reduce observation and model errors, and estimates crop evapotranspiration with high accuracy under the energy balance principle. A water balance method is used to estimate daily-scale irrigation water consumption. Furthermore, a deep learning-based crop actual irrigation area identification model is introduced to effectively distinguish the actual irrigation area of ​​different crop types and dynamically reflect the impact of spatiotemporal changes on irrigation efficiency. Through these steps, the present invention can achieve high-precision and rapid estimation of irrigation water consumption per unit area for single crops and multiple crops in regions at different scales, significantly improving the timeliness and accuracy of irrigation water quota assessment. It also possesses strong adaptability to climate change, soil differences, and crop diversity, providing strong technical support for precision irrigation and smart water management.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: A method for estimating irrigation water consumption per unit area based on multi-source data fusion includes: Vegetation index, meteorological data, and soil moisture data were collected within the irrigation area; and the collected vegetation index, meteorological data, and soil moisture data were processed to obtain daily-scale vegetation index, daily-scale meteorological data, and daily-scale soil moisture. Statistical downscaling techniques were used to process daily vegetation indices, meteorological data, and soil moisture to give all data a uniform spatial resolution. Based on data with uniform spatial resolution, daily evapotranspiration is estimated using the SEBAL (Surface Energy Balance Algorithm for Land) model. Based on the obtained daily evapotranspiration, the daily irrigation water consumption per unit area was estimated using the water balance method, supplemented by data such as changes in runoff, infiltration, and soil water storage. The actual irrigated area was identified using deep learning and ensemble modeling techniques. Based on the calculated data, the irrigation amount per unit area for a single crop and the comprehensive irrigation amount per unit area for the crop are calculated.

[0007] Optionally, the collection of vegetation index, meteorological data, and soil moisture data within the irrigation area, and the processing of the collected vegetation index, meteorological data, and soil moisture data to obtain daily-scale vegetation index, daily-scale meteorological data, and daily-scale soil moisture, specifically involves: Vegetation indices: Using MODIS satellite sensors, regional-scale normalized vegetation index, enhanced vegetation index, green index, normalized water index, and soil-adjusted vegetation index are extracted to achieve large-scale dynamic monitoring of crop growth. Simultaneously, multispectral UAVs are used for aerial surveys during key crop growth periods to acquire hyperspectral images with a ground resolution of <5cm. Canopy spectral reflectance, spatial distribution of soil moisture, and visible irrigation traces are collected simultaneously. These high-precision data (collected data) are used to correct the satellite vegetation index through a scale transformation model to obtain the daily-scale vegetation index. Meteorological data: Three types of data sources are integrated. First, daily continuous observation data from meteorological stations within the irrigation area are integrated, including rainfall, surface temperature, air temperature, wind speed, and relative humidity parameters. Second, the ERA5 reanalysis dataset from the European Centre for Medium-Range Weather Forecasts (ECMWF) is introduced to supplement global meteorological elements. Based on the thermal infrared sensor of the Landsat-8 satellite, the split-window algorithm is used to retrieve surface temperature. Finally, the ensemble Kalman filter method is used to integrate multi-source daily meteorological data. Soil Moisture: To obtain the dynamics of root zone moisture, a multi-source satellite data assimilation and fusion method was adopted, combining the Noah-MP land surface process model with an ensemble Kalman filter data assimilation framework. Within this framework, soil moisture remote sensing data were first acquired from different satellite sensors, Sentinel-1, medium resolution imaging spectrometer, soil moisture and ocean salinity satellites, and interferometric synthetic aperture radar, covering soil moisture information at different spatial and temporal resolutions. The enhanced spatiotemporal adaptive reflectivity fusion model was used for data fusion. This model dynamically adjusts the model's state variables by combining different remote sensing data with the simulation results of the Noah-MP land surface process model, optimizing the estimation of soil moisture in real time, and thus obtaining the daily-scale soil moisture.

[0008] Optionally, the statistical downscaling technique is used to process the daily-scale vegetation index, daily-scale meteorological data, and daily-scale soil moisture to give all data a uniform spatial resolution, specifically including: A multi-source data downscaling framework was adopted to achieve spatial uniformity of data with different resolutions. The input data included 30-meter resolution vegetation indices generated by Landsat-8, 30-meter daily meteorological data fused by an enhanced spatiotemporal adaptive reflectance fusion model, and fused soil moisture data. The processing flow first performed feature engineering to extract the temporal gradient of vegetation indices and topographic adjustment factors. Then, interpolation and resampling methods were used to unify all input data to a spatial resolution of 30 meters. Finally, physical constraints were used to optimize soil moisture, generating soil moisture data with uniform spatial resolution, thus ensuring spatial consistency of data from different sources.

[0009] Optionally, the step of estimating diurnal evapotranspiration based on the SEBAL model using data with uniform spatial resolution to calculate crop water requirements specifically includes: All input data were fused from multiple sources and downscaled to a spatial resolution of 30 meters. In the SEBAL model calculation process, firstly, remote sensing data and meteorological data were used to calculate net radiation using the principle of surface energy balance. Secondly, surface heat flux and sensible heat flux were calculated by combining surface reflectance and solar radiation information. Finally, based on the net radiation, surface heat flux, and sensible heat flux obtained, the SEBAL model calculated diurnal evapotranspiration using the energy balance equation.

[0010] Optionally, the specific calculation formula for the SEBAL model is as follows: ; ; ; ; in, Where G is net radiation and G is surface heat flux. H For sensible heat flux, Evaporation rate It is the difference between the aerodynamic surface temperature and the reference altitude temperature. Where is the surface temperature, and a and b are empirical coefficients. It is air density. It is the specific heat of air. It is the aerodynamic drag that accounts for heat transfer between the aerodynamic surface and the reference altitude. For surface albedo, and For incident and outgoing longwave radiation, It is the surface thermal emissivity. It is the actual direct and diffuse solar radiation flux reaching the Earth's surface; The evapotranspiration of crops can be calculated using the formula above.

[0011] Optionally, the calculation formula for estimating daily irrigation water consumption using the water balance method based on the obtained daily evapotranspiration is as follows: I t =ET t +R t +D t -P t +ΔS t ; In the formula, P t For precipitation, I t This refers to the daily irrigation volume. ET t Evaporation rate R t For runoff, D t This refers to the depth of leakage. Δ S t This refers to changes in soil water storage.

[0012] Optionally, the deep learning and ensemble model technology for identifying the actual irrigated area specifically includes: During the model training phase, the input data is first divided into different crop types, and a deep learning model is trained separately for each crop type; the U-Net model is used for fine segmentation of the irrigation area. In the ensemble model section, an ensemble learning method is used to fuse the outputs of multiple models, and the advantages of each model are combined by using a weighted average or voting mechanism. After training, feature variable importance analysis is performed, and tools are used to analyze in depth the contribution of each feature variable in the model to the identification of irrigation area. Output results: The model will output actual irrigated area data at the monthly scale, and all results will be spatially aggregated to the field scale.

[0013] Optionally, the calculation formula for the irrigation amount per unit area of ​​a single crop and the combined irrigation amount per unit area of ​​multiple crops is as follows: Single crop irrigation calculation: Based on crop type, growth stage, and climate zone, determine the basic irrigation amount W. base Dynamic correction factor: W adj =W base × ) ) ; in, ET actual For the actual remote sensing inversion ET , ET ref For reference to ET, S soil For real-time soil moisture, S threshold This is the irrigation trigger threshold; Using GIS spatial overlay analysis, the irrigation water consumption per unit area for different crops is weighted and averaged according to the planting area to generate the comprehensive irrigation water consumption per unit area for multiple crops at the county or irrigation district level: W comp = ; In the formula, W i For crops i Corrected irrigation water consumption per unit area A i The planting area.

[0014] A system for estimating irrigation water consumption per unit area based on multi-source data fusion includes: The data acquisition module collects vegetation index, meteorological data, and soil moisture data within the irrigation area; and processes the collected vegetation index, meteorological data, and soil moisture data to obtain daily-scale vegetation index, daily-scale meteorological data, and daily-scale soil moisture. The data downscaling module uses statistical downscaling techniques to process daily vegetation index, daily meteorological data, and daily soil moisture, so that all data have a uniform spatial resolution. The evapotranspiration estimation module, based on data with uniform spatial resolution, utilizes SEBA. L Model estimates of daily evapotranspiration; The irrigation water consumption estimation module, based on the obtained daily evapotranspiration, supplements runoff, infiltration and soil water storage change data, and uses the water balance method to estimate the daily irrigation water consumption per unit area. The irrigation area output module uses deep learning and ensemble model technology to identify the actual irrigation area. The irrigation output module calculates the irrigation amount per unit area for a single crop and the combined irrigation amount per unit area for multiple crops based on the calculated data.

[0015] As can be seen from the above technical solution, compared with the prior art, this invention discloses a method and system for estimating irrigation water consumption per unit area based on multi-source data fusion. It integrates heterogeneous data from multiple sources, such as remote sensing images, meteorological data, soil moisture, and crop growth information. Data assimilation techniques are used to reduce observation and model errors, and crop evapotranspiration is estimated with high precision under the energy balance principle. A water balance method is used to estimate daily irrigation water consumption. Furthermore, a deep learning-based crop irrigation area identification model is introduced to effectively distinguish the irrigation area of ​​different crop types and dynamically reflect the impact of spatiotemporal changes on irrigation efficiency. Through these steps, this invention can achieve high-precision and rapid estimation of irrigation water consumption per unit area for single crops and multiple crops combined, significantly improving the timeliness and accuracy of irrigation water quota assessment. It also possesses strong adaptability to climate change, soil differences, and crop diversity, providing strong technical support for precision irrigation and smart water management. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the equation process provided by the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] This invention discloses a method for estimating irrigation water consumption per unit area based on multi-source data fusion, such as... Figure 1 As shown, it includes: Vegetation index, meteorological data, and soil moisture data were collected within the irrigation area; and the collected vegetation index, meteorological data, and soil moisture data were processed to obtain daily-scale vegetation index, daily-scale meteorological data, and daily-scale soil moisture. Statistical downscaling techniques were used to process daily vegetation index, daily meteorological data, and daily soil moisture to give all data a uniform spatial resolution. Based on data with uniform spatial resolution, diurnal evapotranspiration is estimated using the SEBAL model. Based on the obtained daily evapotranspiration, supplemented by data on changes in runoff, infiltration, and soil water storage, the water balance method is used to estimate the daily irrigation water consumption per unit area. The actual irrigated area was identified using deep learning and ensemble modeling techniques. Based on the calculated data, the irrigation water volume per unit area for a single crop and the comprehensive irrigation water volume per unit area for multiple crops are calculated.

[0020] In a specific embodiment, multi-source data acquisition and fusion specifically includes: Vegetation Indices: Using MODIS satellite sensors (spatial resolution 250-1000m, revisit period 1 day), vegetation parameters such as regional-scale Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Green Index (GI), Normalized Difference Water Index (NDWI), and Soil-Adjusted Vegetation Index (SAVI) are extracted to achieve large-scale dynamic monitoring of crop growth. Simultaneously, to improve data accuracy, multispectral UAVs are used for aerial surveys during key crop growth periods, with flight altitude controlled at 50m to obtain hyperspectral images with a ground resolution <5cm. Canopy spectral reflectance, spatial distribution of soil moisture, and visible irrigation traces are collected simultaneously. These high-precision data are used to correct satellite vegetation indices through a scale transformation model, and combined with ground GPS positioning sampling points (horizontal accuracy ±3cm) to verify the accuracy of irrigation event identification (a confusion matrix can be used to calculate producer and user accuracy).

[0021] Meteorological data: The meteorological database is constructed by integrating three types of data sources. First, it integrates daily continuous observation data from meteorological stations within the irrigation area, including parameters such as rainfall (pre), surface temperature (lst), air temperature (temp), wind speed, and relative humidity. Second, it introduces the ERA5 reanalysis dataset from the European Centre for Medium-Range Weather Forecasts (ECMWF) (spatial resolution 0.25°×0.25°, temporal resolution 1 hour) to supplement global meteorological elements such as solar radiation and air pressure. Based on the thermal infrared sensor (TIRSBand 10-11) of the Landsat-8 satellite, the split-window algorithm is used to retrieve surface temperature (LST). Finally, the ensemble Kalman filter (EnKF) method is used to integrate multi-source daily meteorological data: based on the observation data of meteorological stations in the irrigation area, it integrates the daily products (spatial resolution 0.25°) from the ERA5 reanalysis dataset from the ECMWF, and coordinates with the daily surface temperature products (30m resolution) retrieved from the Landsat-8 satellite.

[0022] Soil Moisture: To obtain dynamics of root zone moisture, a multi-source satellite data assimilation and fusion method was adopted, combining the Noah-MP land surface process model with the Ensemble Kalman Filter (EnKF) data assimilation framework. Within this framework, soil moisture remote sensing data acquired from different satellite sensors (such as the Soil Moisture Active Passive Satellite (SMAP), Sentinel-1, Moderate Resolution Imaging Spectroradiometer (MODIS), and Soil Moisture and Ocean Salinity Satellite (SMOS)) and Interferometric Synthetic Aperture Radar (InSAR) were preprocessed, covering different spatial resolutions (e.g., 10 km, 1 km) and temporal resolutions (e.g., 1 day or less). These data contain soil moisture information at different levels, and each data source has different spatial and temporal resolutions; therefore, standardization and spatial interpolation processing are required to ensure the consistency of the different data.

[0023] Next, the enhanced spatiotemporal adaptive reflectivity fusion model (ESTARFM) was used for data fusion. ESTARFM dynamically adjusts the model's state variables (such as the depth distribution of soil moisture) by combining different remote sensing data with simulation results from the Noah-MP land surface process model, thereby optimizing soil moisture estimation in real time. The specific steps include: first, using the Noah-MP model to make a preliminary prediction of soil moisture; then, ESTARFM updates the model based on real-time satellite remote sensing data, correcting the prediction results to make the final output soil moisture data more accurate. By combining low-spatial-resolution remote sensing imagery with high-spatial-resolution data, ESTARFM effectively reduces model errors and improves estimation accuracy.

[0024] To further improve the accuracy and reliability of the model, historical meteorological data and soil type information will be used to assist the data assimilation process. Specifically, precipitation, temperature, and humidity data from meteorological stations will be used as model input to better reflect changes in soil moisture. In addition, the spatial distribution of soil types will also be incorporated into the assimilation framework to account for the impact of different soil types on water retention and transport.

[0025] Spatial Statistical Downscaling Method: This invention employs a multi-source data downscaling framework to achieve spatial uniformity of data from different resolutions. Input data includes 30-meter resolution vegetation indices generated by Landsat-8, 30-meter diurnal meteorological data (temperature, precipitation, etc.) fused with ESTARFM, and fused soil moisture data. The processing flow first performs feature engineering to extract the temporal gradient of vegetation indices and topographic adjustment factors (such as elevation and slope). Then, interpolation and resampling methods are used to unify all data to a 30-meter spatial resolution. Finally, physical constraints are used to optimize soil moisture, generating soil moisture data with uniform spatial resolution. This method ensures spatial consistency of data from different sources, providing support for precision irrigation management.

[0026] The purpose of this section is: This research addresses the challenges of inconsistent scales, varying accuracy, and significant differences in spatiotemporal resolution among multi-source data sources, including vegetation indices, meteorological data, soil moisture, and environmental information such as elevation. It leverages data from multiple optical remote sensing satellites (Landsat / MODIS / Sentinel-2 MSI), microwave remote sensing satellites (Sentinel-1A), meteorological data, statistical data, and sensor data. These data are transformed into daily-scale, uniformly spatially resolved vegetation index, meteorological, and soil moisture data using data fusion (ESTARFM), data assimilation (Noah-MP+EnKF), and spatial statistical downscaling methods. This section resolves the issues of scale inconsistency, accuracy discrepancies, and significant spatiotemporal resolution differences among multi-source heterogeneous data, providing a unified and reliable data foundation for subsequent high-precision irrigation water estimation.

[0027] In a specific embodiment, the energy balance principle method for crop water requirement includes: The SEBAL model was used to estimate crop water requirements, with the goal of generating diurnal evapotranspiration (ET) data. The SEBAL model accurately estimates evapotranspiration based on the physical principles of surface energy balance, thereby calculating crop water requirements. In the preliminary steps, all input data (including vegetation indices, meteorological data, soil moisture, etc.) underwent multi-source data fusion and downscaling to be unified to a 30-meter spatial resolution to fully utilize high-resolution remote sensing imagery. This data assimilation process ensured the consistency and high accuracy of the input data, making the final estimated crop water requirements more reliable and accurate. During the SEBAL model calculation, remote sensing data (such as Landsat-8 reflectance and surface temperature) and meteorological data (such as precipitation and air temperature) were first used to calculate net radiation (EEG) based on the surface energy balance principle. Net radiation is the total radiant energy received by the Earth's surface. After reflection, absorption, and transmission, a portion of it is converted into surface heat. Subsequently, combining information such as surface reflectivity and solar radiation, the surface heat flux (G) and sensible heat flux (H) are calculated. Sensible heat flux is the process of heat exchange between the surface and the air, while surface heat flux is the main source of surface temperature changes. Next, the SEBAL model uses this data to accurately calculate evapotranspiration (H) through the energy balance equation. This refers to the water requirement of crops. The specific calculation formula is: ; ; ; ; in, Where G is net radiation and G is surface heat flux. H For sensible heat flux, Evaporation rate It is the difference between the aerodynamic surface temperature and the reference altitude temperature. Where is the surface temperature, and a and b are empirical coefficients. It is air density. It is the specific heat of air. It is the aerodynamic drag that accounts for heat transfer between the aerodynamic surface and the reference altitude. For surface albedo, and For incident and outgoing longwave radiation, It is the surface thermal emissivity. It is the actual direct and diffuse solar radiation flux reaching the Earth's surface; The evapotranspiration of crops can be calculated using the formula above.

[0028] Ultimately, the SEBAL model outputs evapotranspiration data with a spatial resolution of 30 meters, generated over a diurnal time span. The evapotranspiration calculation for each pixel is precisely assigned and analyzed based on the specific climate and crop type within the region. This method, by accurately simulating the surface energy balance, provides a high-precision estimate of crop water requirements, offering strong data support for precision irrigation decisions, water resource allocation, and agricultural water management. This result not only effectively improves water resource utilization efficiency but also provides a scientific basis for optimizing agricultural irrigation strategies, promoting the implementation of precision agriculture and sustainable water resource management.

[0029] The purpose of this section is: Using the high-precision multi-source fusion input dataset generated in the first step, the SEBAL (Surface Energy Balance Algorithm for Land) method is employed to accurately estimate crop evapotranspiration through surface energy balance calculations, utilizing remote sensing data (such as surface temperature and reflectivity) and meteorological data. This method addresses the shortcomings of traditional empirical formulas, such as insufficient estimation accuracy and inability to dynamically reflect changes in meteorological and surface conditions, thus achieving accurate calculation of daily crop water requirements.

[0030] In one specific embodiment, irrigation water estimation based on the water balance method includes: Daily irrigation water consumption was estimated using the water balance method. Input data included 30-meter resolution daily actual evapotranspiration (AET) data estimated in the second step based on the energy balance principle, combined with multi-source auxiliary data such as interpolated data from local meteorological stations, runoff data, infiltration data, and soil moisture data (30-meter resolution) obtained in the aforementioned downscaling step. This data was used to calculate daily irrigation water consumption using the water balance equation, as shown in the following formula: I t =ET t +R t +D t -P t +ΔS t ; In the formula, P t For precipitation, I t This refers to the daily irrigation volume. ET t Evaporation rate R t For runoff, D t This refers to the depth of leakage. Δ S t This refers to changes in soil water storage.

[0031] To enhance the physical plausibility of the calculation results, this invention introduces dual constraints: first, evapotranspiration (ET) data is used as an energy balance constraint to ensure crop water demand mechanisms; second, soil moisture data is used as a mass conservation constraint, combined with threshold filtering for outliers (e.g., limiting the overestimation of seepage when soil moisture reaches saturation). These two constraints help ensure that irrigation water consumption estimates conform to the actual laws of the hydrological cycle and further improve the accuracy of the results.

[0032] The specific implementation steps include three parts: First, in the preprocessing stage, all data are aligned to a 30-meter grid, and areas affected by cloud cover are removed to ensure spatial consistency of the data; second, in the core calculation stage, an iterative optimization algorithm (based on gradient descent) is used to solve the water balance equation, and constraint terms are applied to correct the output after each iteration; finally, in the post-processing stage, the results are extrapolated to the field scale through spatial aggregation, and the time series is smoothed by a 7-day moving average filter to remove noise and ensure the stability and reliability of the results.

[0033] The verification process of this invention will be conducted in the Dengkou Irrigation District of Inner Mongolia and the Yuzhuang Irrigation District of Henan Province, focusing on comparing irrigation data measured by smart water meters. The error between the model's output irrigation water volume and the measured data will be evaluated through spatial alignment and temporal consistency checks, using RMSE, MAE, and R² indicators for error analysis. Furthermore, a ten-fold cross-validation method will be used to assess the model's stability and generalization ability. The accuracy of the soil water storage change estimation in the water balance equation will be verified further by comparing with measured soil moisture and meteorological data. Finally, spatial aggregation and field-scale applications will be performed, using a 7-day moving average filter to smooth the time series, ensuring the model's accuracy and practical operability.

[0034] This section, based on the daily evapotranspiration data from the second step, combines runoff, infiltration, and precipitation data to conduct a water balance analysis. By simulating soil moisture changes and hydrological processes, it accurately calculates daily irrigation water consumption. This method addresses the problem that existing irrigation water estimation methods fail to adequately consider changes in soil water storage and water balance, thus improving the scientific rigor and accuracy of water consumption estimation.

[0035] In one specific embodiment, monitoring of actual irrigated area includes: This invention employs deep learning and ensemble modeling techniques to identify actual irrigated areas. The goal is to generate high-precision irrigation area monitoring data using a multi-source data fusion method. Input data includes a previously generated high-resolution multi-source dataset, comprising vegetation index (NDVI), evapotranspiration (ET), surface temperature, meteorological data (such as air temperature and precipitation), soil moisture, spectrum, drought index, SAR data, elevation, slope, land use information, and phenological characteristics. All data has undergone downscaling in previous steps to a spatial resolution of 30 meters to ensure data consistency and high accuracy.

[0036] During the model training phase, the input data is first categorized into different crop types (such as rice paddies and corn), and a separate deep learning model is trained for each crop type. The U-Net model is used for fine segmentation of the irrigation area; U-Net, with its superior image segmentation capabilities, can identify subtle irrigation area boundaries. To further improve the model's accuracy and robustness, traditional machine learning methods such as XGBoost and Random Forest are combined. These methods effectively handle complex relationships and nonlinear features in the input data, providing powerful feature fusion capabilities.

[0037] In the ensemble model section, an ensemble learning approach is used to fuse the outputs of multiple models (such as U-Net and XGBoost / Random Forest). Weighted averaging or voting mechanisms are employed to combine the strengths of each model, thereby improving the overall model's accuracy and stability. Furthermore, combining deep learning models with traditional machine learning methods enables better extraction of feature information in complex irrigation regions.

[0038] After training, feature variable importance analysis is performed. Tools such as SHAP (Shapley Additive Explanations) are used to analyze in depth the contribution of each feature variable in the model to the identification of irrigation area, so as to improve the interpretability and transparency of the model.

[0039] Model validation will be conducted in the Dengkou Irrigation District of Inner Mongolia and the Yuzhuang Irrigation District of Henan Province. These two regions have different climatic conditions, soil types, and irrigation management patterns, providing broad applicability for model validation. Validation data will include high-resolution actual irrigated area data obtained through UAV data, used for comparison with the model output. In addition, statistical data (such as irrigation management records) will also serve as validation data sources. A 10-fold cross-validation method will be used to evaluate the model's performance on different datasets, ensuring its stability and generalization ability. UAV data will be used as residual input to further optimize model training and enhance model accuracy.

[0040] Output results: Finally, the model will output the actual irrigated area data on a monthly scale. All results will be spatially aggregated to the field scale, and the time series will be smoothed by a 7-day moving average filter to reduce noise.

[0041] This section utilizes the high-precision multi-source fusion input dataset generated in the first step, selecting vegetation index (NDVI), evapotranspiration, surface temperature, meteorological data, soil moisture, spectrum, drought index, SAR data, elevation, slope, land use information, and phenological characteristics as feature variables. Deep learning and ensemble models such as U-Net, XGBoost, and random forest are trained for different crops (e.g., paddy fields). Combined with an irrigated farmland edge monitoring and segmentation model, the actual irrigated area is identified on a monthly scale. This method solves the problems of existing irrigated area statistics, such as reliance on manual surveys, long update cycles, and insufficient accuracy, achieving large-scale, high-frequency, and high-precision dynamic monitoring of irrigated area.

[0042] In one specific embodiment, irrigation water consumption monitoring specifically includes: Single-crop irrigation water calculation: Based on crop type, growth stage, and climate zone, determine the basic irrigation water volume W. base Dynamic correction factor: W adj =W base × ) ) ; in, ET actual For the actual remote sensing inversion ET , ET ref For reference to ET, S soil For real-time soil moisture, S threshold This is the irrigation trigger threshold; Using GIS spatial overlay analysis, the irrigation water consumption of different crops is weighted and averaged according to the planting area to generate the comprehensive irrigation water consumption of multiple crops at the county or irrigation district level: W comp = ; In the formula, W i For crops i Corrected irrigation water volume A i The planting area.

[0043] This section, based on the obtained irrigation water consumption data for different crops and the actual irrigated area, combined with the water requirements of crops, accurately estimates the irrigation water consumption for a single crop. It also integrates the irrigation needs of multiple crops with the regional water resource distribution to calculate the comprehensive irrigation water consumption for the entire irrigation district. This achieves accurate assessment of the water demand of crops at different growth stages and provides a basis for irrigation scheduling decisions. This method solves the problems of lacking real-time updates and difficulty in dynamically adapting to climate and crop changes in existing irrigation water consumption calculations, thus improving the scientific and rational nature of water resource allocation.

[0044] An irrigation water consumption estimation system based on multi-source data fusion includes: The data acquisition module collects vegetation index, meteorological data, and soil moisture data within the irrigation area; and processes the collected vegetation index, meteorological data, and soil moisture data to obtain daily-scale vegetation index, daily-scale meteorological data, and daily-scale soil moisture. The data downscaling module uses statistical downscaling techniques to process daily vegetation index, daily meteorological data, and daily soil moisture, so that all data have a uniform spatial resolution. The evapotranspiration estimation module estimates daily evapotranspiration based on the SEBAL model, using data with uniform spatial resolution. The irrigation water consumption estimation module, based on the obtained daily evapotranspiration, supplements runoff, infiltration and soil water storage change data, and uses the water balance method to estimate the daily irrigation water consumption per unit area. The irrigation area output module uses deep learning and ensemble model technology to identify the actual irrigation area. The irrigation output module calculates the irrigation amount for a single crop and the total irrigation amount for the crop based on the calculated data.

[0045] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0046] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for estimating irrigation water consumption per unit area based on multi-source data fusion, characterized in that, include: Collect vegetation index, meteorological data, and soil moisture data within the irrigated area; The collected vegetation index, meteorological data, and soil moisture data were processed to obtain daily-scale vegetation index, daily-scale meteorological data, and daily-scale soil moisture. Statistical downscaling techniques were used to process daily vegetation index, daily meteorological data, and daily soil moisture to give all data a uniform spatial resolution. Based on data with uniform spatial resolution, diurnal evapotranspiration is estimated using the SEBAL model. Based on the obtained daily evapotranspiration, supplemented by data on changes in runoff, infiltration, and soil water storage, the water balance method is used to estimate the daily irrigation water consumption per unit area. The actual irrigated area was identified using deep learning and ensemble modeling techniques. Based on the calculated data, the irrigation water volume per unit area for a single crop and the comprehensive irrigation water volume per unit area for multiple crops are calculated.

2. The method for estimating irrigation water consumption per unit area based on multi-source data fusion according to claim 1, characterized in that, The collection of vegetation index, meteorological data, and soil moisture data within the irrigation area; and the processing of the collected vegetation index, meteorological data, and soil moisture data to obtain daily-scale vegetation index, daily-scale meteorological data, and daily-scale soil moisture are specifically as follows: Vegetation indices: Using MODIS satellite sensors, regional-scale normalized vegetation index, enhanced vegetation index, green index, normalized water index, and soil-adjusted vegetation index are extracted to achieve large-scale dynamic monitoring of crop growth. Simultaneously, during the key growth stages of crops, multispectral UAVs are used for aerial surveys to acquire hyperspectral images with a ground resolution of <5cm. Canopy spectral reflectance, spatial distribution of soil moisture, and visible irrigation traces are collected simultaneously. The collected data are then corrected using a scale transformation model to obtain the daily-scale vegetation index. Meteorological data: The data sources are integrated. First, the daily continuous observation data of meteorological stations in the irrigation area are integrated, including rainfall, surface temperature, air temperature, wind speed and relative humidity parameters. Second, the reanalysis dataset of the weather forecast center is introduced to supplement global meteorological elements. The surface temperature is retrieved by applying the split window algorithm based on the thermal infrared sensor of the Landsat-8 satellite. Finally, the multi-source daily meteorological data is integrated using the ensemble Kalman filter method. Soil Moisture: To obtain the dynamics of root zone moisture, a multi-source satellite data assimilation and fusion method was adopted, combining the Noah-MP land surface process model with an ensemble Kalman filter data assimilation framework. Within this framework, soil moisture remote sensing data were first acquired from different satellite sensors, Sentinel-1, medium resolution imaging spectrometer, soil moisture and ocean salinity satellites, and interferometric synthetic aperture radar, covering soil moisture information at different spatial and temporal resolutions. The enhanced spatiotemporal adaptive reflectivity fusion model was used for data fusion. This model dynamically adjusts the model's state variables by combining different remote sensing data with the simulation results of the Noah-MP land surface process model, optimizing the estimation of soil moisture in real time, and thus obtaining the daily-scale soil moisture.

3. The method for estimating irrigation water consumption per unit area based on multi-source data fusion according to claim 2, characterized in that, The statistical downscaling technique is used to process the daily-scale vegetation index, daily-scale meteorological data, and daily-scale soil moisture to give all data a uniform spatial resolution. Specifically, this includes: A multi-source data downscaling framework was adopted to achieve spatial uniformity of data with different resolutions. The input data included 30-meter resolution vegetation indices generated by Landsat-8, 30-meter daily meteorological data fused by an enhanced spatiotemporal adaptive reflectance fusion model, and fused soil moisture data. The processing flow first performed feature engineering to extract the temporal gradient of vegetation indices and topographic adjustment factors. Then, interpolation and resampling methods were used to unify all input data to a spatial resolution of 30 meters. Finally, physical constraints were used to optimize soil moisture, generating soil moisture data with uniform spatial resolution, thus ensuring spatial consistency of data from different sources.

4. The method for estimating irrigation water consumption per unit area based on multi-source data fusion according to claim 1, characterized in that, The estimation of diurnal evapotranspiration based on the SEBAL model, using data with uniform spatial resolution, specifically includes: All input data were fused from multiple sources and downscaled to a spatial resolution of 30 meters. In the SEBAL model calculation process, firstly, remote sensing data and meteorological data were used to calculate net radiation using the principle of surface energy balance. Secondly, surface heat flux and sensible heat flux were calculated by combining surface reflectance and solar radiation information. Finally, based on the net radiation, surface heat flux, and sensible heat flux obtained, the SEBAL model calculated diurnal evapotranspiration using the energy balance equation.

5. The method for estimating irrigation water consumption per unit area based on multi-source data fusion according to claim 1, characterized in that, The specific calculation formula for the SEBAL model is as follows: ； ； ； ； in, Where G is net radiation and G is surface heat flux. H For sensible heat flux, Evaporation rate It is the difference between the aerodynamic surface temperature and the reference altitude temperature. Where is the surface temperature, and a and b are empirical coefficients. It is air density. It is the specific heat of air. It is the aerodynamic drag that accounts for heat transfer between the aerodynamic surface and the reference altitude. For surface albedo, and For incident and outgoing longwave radiation, It is the surface thermal emissivity. It is the actual direct and diffuse solar radiation flux reaching the Earth's surface; The evapotranspiration of crops can be calculated using the formula above.

6. The method for estimating irrigation water consumption per unit area based on multi-source data fusion according to claim 1, characterized in that, The specific formula for estimating daily irrigation water consumption using the water balance method based on the obtained daily evapotranspiration is as follows: I t =ET t +R t +D t -P t +ΔS t ； In the formula, P t For precipitation, I t This refers to the daily irrigation volume. ET t Evaporation rate R t For runoff, D t This refers to the depth of leakage. ΔS t This refers to changes in soil water storage.

7. The method for estimating irrigation water consumption per unit area based on multi-source data fusion according to claim 1, characterized in that, The deep learning and ensemble model technology for identifying the actual irrigated area specifically includes: During the model training phase, the input data is first divided into different crop types, and a deep learning model is trained separately for each crop type; the U-Net model is used for fine segmentation of the irrigation area. In the ensemble model section, an ensemble learning method is used to fuse the outputs of multiple models, and the advantages of each model are combined by using a weighted average or voting mechanism. After training, feature variable importance analysis is performed, and tools are used to analyze in depth the contribution of each feature variable in the model to the identification of irrigation area. Output results: The model will output actual irrigated area data at the monthly scale, and all results will be spatially aggregated to the field scale.

8. The method for estimating irrigation water consumption per unit area based on multi-source data fusion according to claim 1, characterized in that, The formulas for calculating the irrigation water volume per unit area for a single crop and the combined irrigation water volume per unit area for multiple crops are as follows: Calculation of irrigation water per unit area for a single crop: Based on crop type, growth stage, and climate zone, determine the basic irrigation water per unit area W. base Dynamic correction factor: ； in, ET actual For the actual remote sensing inversion ET , ET ref For reference to ET, S soil For real-time soil moisture, S threshold This is the irrigation trigger threshold; Using GIS spatial overlay analysis, the irrigation amount per unit area for different crops is weighted and averaged according to the planting area to generate the comprehensive irrigation amount per unit area for multiple crops at the county or irrigation district level: W comp = ； In the formula, W i For crops i The correction amount, A i The planting area.

9. A system for estimating irrigation water consumption per unit area based on multi-source data fusion, characterized in that, The method for estimating irrigation water consumption per unit area based on multi-source data fusion according to any one of claims 1-8 includes: The data acquisition module collects vegetation index, meteorological data, and soil moisture data within the irrigation area; and processes the collected vegetation index, meteorological data, and soil moisture data to obtain daily-scale vegetation index, daily-scale meteorological data, and daily-scale soil moisture. The data downscaling module uses statistical downscaling techniques to process daily vegetation index, daily meteorological data, and daily soil moisture, so that all data have a uniform spatial resolution. The evapotranspiration estimation module estimates daily evapotranspiration based on the SEBAL model, using data with uniform spatial resolution. The irrigation water consumption estimation module, based on the obtained daily evapotranspiration, supplements runoff, infiltration and soil water storage change data, and uses the water balance method to estimate the daily irrigation water consumption per unit area. The irrigation area output module uses deep learning and ensemble model technology to identify the actual irrigation area. The irrigation output module calculates the irrigation volume per unit area for a single crop and the comprehensive irrigation volume per unit area for multiple crops in different scale regions based on the calculated data.

Images