Multi-source remote sensing daily-scale plot-level crop irrigation water demand forecasting method and system

Abstract

The invention discloses a multi-source remote sensing daily-scale plot-level crop irrigation water demand forecasting method and system, and relates to the technical field of agricultural water-saving irrigation, remote sensing data processing and WebGIS online services, and the method comprises the following steps: S1, data preprocessing; s2, modeling with reference to crop evapotranspiration deviation; s3, rainfall deviation modeling is carried out; s4, carrying out model migration and forecast correction; s5, performing net radiation correction and ET0 calculation; s6, crop evapotranspiration and irrigation water demand calculation; and S7, outputting and issuing a result. According to the multi-source remote sensing daily scale plot-level crop irrigation water demand forecasting method and system, remote sensing data, reanalysis / forecast meteorological data, crop parameters and a high-resolution crop distribution map layer are combined, irrigation water demands of different crops at different time periods can be dynamically calculated on the basis of correcting meteorological data deviation, and the irrigation efficiency is improved. And the result is visually published through the platform, so that scientific decision support is provided for agricultural irrigation scheduling.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical fields of agricultural water-saving irrigation, remote sensing data processing and WebGIS online service, in particular to a multi-source remote sensing daily scale plot-level crop irrigation water demand prediction method and system. BACKGROUND

[0002] Accurate estimation of farmland irrigation water demand is an important basis for water-saving irrigation and optimal allocation of water resources. Currently, the industry usually takes the reference crop evapotranspiration (ET0) as the driving force, combines the crop coefficient (K c ) to estimate the crop evapotranspiration (ET c ), and obtains the net irrigation water demand (NIR) after considering the effective precipitation (P e ) to guide irrigation decision-making. Although technologies such as Internet of Things sensing, remote sensing inversion and numerical weather prediction have been gradually introduced into this field, there are still significant challenges in large-scale and normalized business applications, which cannot meet the urgent needs of wheat, corn, rice and other major grain producing areas for daily scale and plot-level water demand services. Specifically, there are differences and systematic deviations in the variable definition, unit, time and space reference and resolution of multi-source data, which directly used for NIR calculation can easily lead to error amplification; at the same time, there are systematic errors in ET0 and precipitation and other key meteorological inputs, and static or regional average K c cannot reflect the phenology and regional differences, and the prediction end lacks a stable deviation correction and migration mechanism, resulting in insufficient accuracy of ET c and NIR estimation; in addition, existing platforms are mostly limited to regional pilots, lacking high-resolution NIR products covering the whole country, automatic updating and online publishing, and have not yet systematically integrated domestic prediction and observation data to adapt to China's monsoon climate and complex terrain conditions, which restricts cross-regional planning and localized business applications.

[0003] Therefore, the industry urgently needs to develop a technical solution that integrates multi-source data fusion, unified deviation correction and consistency control, dynamic K c based on phenology and effective precipitation estimation, prediction migration and physical rationality constraints, nationwide scale automatic product production and online publishing, and localized parameter regionalization calibration, so as to build a precision controllable, business-oriented and adaptive irrigation water demand estimation technical path for China's major producing areas, and provide reliable support for agricultural water-saving irrigation, precise water resource scheduling and drought prevention and control. SUMMARY

[0004] The purpose of the present application is to provide a multi-source remote sensing daily scale plot-level crop irrigation water demand prediction method and system to solve the problems raised in the background art.

[0005] To achieve the above-mentioned purpose, the present application provides a multi-source remote sensing daily scale plot-level crop irrigation water demand prediction method, comprising the following steps: S1, data preprocessing: obtain historical data and corresponding measured data of the target area, unify the unit and time reference, aggregate to daily scale and extract modeling features; S2, reference crop evapotranspiration deviation modeling: establish a reference crop evapotranspiration deviation model based on historical data, which is used to represent the prediction deviation of daily reference evapotranspiration ET0; S3, precipitation deviation modeling: establish a precipitation deviation model based on historical data, which is used to represent the prediction deviation of daily precipitation P; S4, model migration and prediction correction: migrate and apply the reference crop evapotranspiration deviation model and the precipitation deviation to the mesoscale numerical weather prediction product, and correct the daily prediction ET0 and P to obtain the corrected daily reference evapotranspiration ET 0,corr and the corrected daily precipitation P corr ; S5, net radiation correction and ET0 calculation: correct the daily net radiation R n according to the grading rules at the prediction end, and calculate and update the daily reference evapotranspiration ET0 using the Penman-Monteith formula; S6, crop evapotranspiration and irrigation water requirement calculation: generate the daily variation curve of crop coefficient K c using the crop spatial distribution map and crop growth period information, and calculate the crop evapotranspiration ET 0,corr based on the corrected daily reference evapotranspiration ET c ; Subtract the effective precipitation P c from the crop evapotranspiration ET e to obtain the net irrigation water requirement NIR; S7, result output and release: output ET c and NIR as raster products, and regularly online release and update.

[0006] Preferably, the reference crop evapotranspiration deviation model in S2 is trained using a random forest regression algorithm, and the training samples are composed of same-site and same-day paired data of historical phase ET0 and ground meteorological station measured data.

[0007] Preferably, the precipitation deviation model in S3 is constructed using quantile distribution mapping, and the mapping relationship of wet day cumulative distribution is established based on the same-site and same-day paired samples of historical phase precipitation and ground station measured precipitation data.

[0008] Preferably, S4 adopts a pixel-level linear bridging relationship, and the specific formula is: ; wherein, is the predicted ET0 or P of the prediction field; The bridging coefficient obtained by fitting for each pixel during the overlap period of reanalysis and forecast data, which is used to transfer the deviation relationship formed in the historical stage to the forecast end.

[0009] Preferably, the spatial resolution of the mesoscale numerical weather forecast product in S4 is 3 km, and it runs daily to form the daily-scale irrigation water demand forecast for the next day or the current day, and is updated at 08:00 Beijing time every day.

[0010] Preferably, the classification rule in S5 is specifically: When P≥10 mm on the current day, take R n =0; When 0<P<10 mm, take R n as the low quantile of the reanalysis R in the recent 7 days n ; When P = 0, take R n as the arithmetic mean of the reanalysis R in the recent 7 days n ;

[0011] Preferably, the effective precipitation P in S6 e has the following calculation formula: ; where [[ID=3,4]]is an empirical coefficient representing the effective utilization ratio of rainfall, which is selected within the range of [0.7, 0.85] according to soil infiltration conditions and rainfall characteristics, and varies with space and time.

[0012] Preferably, the spatial resolution of the crop spatial distribution map in S6 is 10 - 30 m, K c is assigned daily according to the phenological stage, and ET c and NIR are calculated at the crop pixel level, and a non-negativity constraint is imposed on NIR.

[0013] Preferably, the content published online in S7 specifically includes: publishing raster products as a geographic information service, and performing time slider browsing and regional statistics on the raster products. The statistical calibers include mean, total amount, and area-weighted indicators. [[ID=,45]]

[0014] A system for a multi-source remote sensing daily-scale plot-level crop irrigation water demand forecasting method, including: Deviation modeling module: including an ET0 deviation sub-module and a precipitation quantile mapping sub-module; Model migration and correction module: configured with pixel-level linear bridging coefficients a and b; Net radiation correction and ET0 calculation module: used to correct R n based on the classification rule; Effective rainfall and water demand calculation module: used to implement the calculation of P e [[ID=5,9]]and the non-negativity constraint of NIR; Results publishing module: used to publish raster products as geographic information services and implement time sliders and regional statistics.

[0015] Therefore, the present invention employs the above-mentioned multi-source remote sensing daily-scale plot-level crop irrigation water demand forecasting method and system, which has the following beneficial effects: (1) Multi-source data fusion ensures robust output. Even if there are missing input data such as precipitation and radiation or large deviations between different meteorological data sources, reasonable irrigation water demand estimation results can still be stably produced through deviation correction and data fusion.

[0016] (2) Significant advantages of high spatiotemporal resolution: This invention uses daily weather forecasts with a spatial resolution of 3km to drive the process, and combines it with spatially differentiated dynamic K c The coefficient and crop distribution data with an accuracy of 10-30m enable the estimated irrigation water demand to have fine spatial differences, making it closer to the actual situation of farmland.

[0017] (3) It can cover farmland across the country and has wide applicability. It can realize intelligent estimation of irrigation water demand with plot-level accuracy, provide accurate data support for agricultural irrigation scheduling, improve water resource management efficiency, and has significant application prospects and promotion value.

[0018] (4) Emphasis is placed on localization and accuracy improvement. Domestic numerical weather prediction, reanalysis data and remote sensing observation data are given priority as the main driving force. The parameterization scheme and assimilation system used are more in line with my country's meteorological observation network, topography and underlying surface characteristics. The established bias correction model and its linear mapping bridging are trained and applied based on the statistical characteristics of local Chinese data. This can effectively reduce cross-data source differences and regional systematic errors, significantly improve local adaptability and estimation accuracy in monsoon areas, complex terrain and areas with significant differences in underlying surface conditions, and at the same time ensure data availability and operational continuity.

[0019] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0020] Fig. 1 This is a schematic diagram of the overall system architecture of an embodiment of the multi-source remote sensing daily-scale plot-level crop irrigation water demand forecasting method and system of the present invention; Fig. 2 This is a flowchart of the deviation migration and correction of ET0 and P in an embodiment of a multi-source remote sensing daily-scale plot-level crop irrigation water demand forecasting method and system of the present invention. Fig. 3 This invention provides an embodiment of a multi-source remote sensing daily-scale plot-level crop irrigation water demand forecasting method and system. cFlowchart of curve dynamic driving and NIR calculation. Detailed Implementation

[0021] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0022] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0023] Example Please see Figs. 1-3 This invention provides a multi-source remote sensing method for daily-scale, plot-level crop irrigation water demand forecasting, with the following specific steps: S1. Demonstration Scenarios and Data Acquisition.

[0024] Historical reanalysis / forecast meteorological data and corresponding measured data for the target area are acquired, and the units and time bases are unified, aggregated to the daily scale, and modeling features are extracted.

[0025] This implementation uses May 8, 2025 as the demonstration date, selecting major wheat, maize, and rice producing areas in China as the application region. At 8:00 AM that day, forecast meteorological data (spatial resolution 3 km) for the region were acquired from the China Meteorological Administration's CMA-MESO mesoscale numerical weather prediction system. After bias correction processing, gridded forecast values ​​for reference crop evapotranspiration (ET0) and precipitation (P) for the day were obtained. Both ET0 and P are represented in a 3 km × 3 km grid format, with each grid cell containing reference evapotranspiration and precipitation elements for the entire day's weather forecast at that location.

[0026] S2, Modeling the deviation of crop evapotranspiration.

[0027] A model was developed based on historical periods to characterize the forecast bias of reference crop evapotranspiration (ET0).

[0028] The reference crop evapotranspiration deviation model is trained using the random forest regression algorithm. The training samples are composed of the paired data of historical ET0 and the measured data from ground meteorological stations on the same site and the same day.

[0029] The model input features include the estimated value of grid ET0, the equivalent variables of grid maximum / minimum temperature and relative humidity, as well as the sine and cosine encodings of year, month and day sequence within the year. The target variable is the measured ET0 - grid ET0, which is used to predict the forecast deviation of remote sensing ET0.

[0030] S3. Precipitation deviation modeling.

[0031] A model for characterizing the precipitation (P) forecast deviation is established based on the historical stage.

[0032] The precipitation deviation model is constructed by quantile distribution mapping. Based on the paired samples of historical precipitation and measured precipitation data from ground stations on the same site and the same day, the mapping relationship of the wet - day cumulative distribution is established.

[0033] Among them, wet days are mapped to the corresponding quantiles of historical station precipitation according to their cumulative frequencies in historical grid - point precipitation; for dry days, the grid - point precipitation is corrected to 0; for values outside the range of the empirical distribution of historical station wet - day precipitation, extrapolation is performed according to the nearest quantile.

[0034] S4. Model transfer and forecast correction.

[0035] The daily reference evapotranspiration (ET0) of the historical period is calculated using ERA5 - Land hourly re - analysis data, and the deviation of ET0 is corrected through a random forest machine learning model to integrate it with the measured data from ground meteorological stations to improve accuracy; Apply the above correction method to the weather forecast field: using CMA - MESO hourly numerical forecast data, through methods such as linear mapping or quantile mapping, transfer the established deviation correction model to the daily - scale forecast meteorological field with a 3 - km resolution, and correct the deviations of forecast elements such as ET0 and precipitation.

[0036] S5. Correction of daily net radiation (R n ) and calculation of ET0.

[0037] Considering the attenuation effect of precipitation on daily net radiation, perform scenario - based correction on the forecasted R n to improve the accuracy of ET0 estimation. The specific classification rules based on rainfall intensity are as follows: When the daily precipitation P reaches or exceeds 10 mm (i.e., heavy rain or rainstorm occurs), assume that the solar radiation during the day is extremely weak, and directly set R n to 0; When 0 < P < 10 mm (light rain), appropriately reduce the forecasted R n value, and set Rn Set as the low quantile level of recent sunny days (e.g., using the 10th percentile p10 of the daily net radiation in the most recent 7 days of ERA5-Land reanalysis data as the daily R...). n ); When P=0 (no rainfall), the recent average radiation level is used, and R is adjusted accordingly. n Set as the average of the daily net radiation values ​​of ERA5-Land over the most recent 7 days.

[0038] Based on the above-mentioned revised R n In addition to forecasted temperature, wind speed, humidity, and other factors, the daily reference crop evapotranspiration ET0 of each 3km grid cell is calculated using the Penman-Monteith formula. The corrected ET0 grid data serves as one of the basic inputs for subsequent irrigation calculations.

[0039] S6. Calculation of crop evapotranspiration and irrigation water requirements.

[0040] According to the crop coefficient (K) set by each province c The standard curve, combined with a 10–30m high-resolution crop planting distribution layer, determines the crop types for different plots, and a phenological-driven algorithm is used to dynamically correct the crop growth stage (updated daily). c (Value). Based on this, the actual crop evapotranspiration ET can be calculated. c That is, applying the corrected ET to each pixel. 0,corr Multiply by the K of the corresponding crop on that day c Get ET c .

[0041] S61. Precipitation correction and calculation of effective rainfall.

[0042] The forecast precipitation P is simultaneously subjected to bias correction to obtain a precipitation estimate Pcorr that better reflects local conditions. Based on the corrected precipitation, the effective rainfall P is calculated. e This deducts the effects of field runoff loss and deep seepage, and is used as a deduction from irrigation water demand. Effective rainfall refers to the portion of precipitation that can be effectively utilized by crops, and is calculated using an empirical formula: ; Among them, the empirical coefficient Indicates the effective utilization rate of rainfall. Coefficient The value can be determined empirically based on the regional soil infiltration characteristics and rainfall intensity, generally ranging from 0.7 to 0.85. For example, in farmland with moderate soil infiltration conditions, a value of 0.7 to 0.85 is often selected. This assumes that approximately 75% of rainfall can be retained in the root zone for crop use, with the remainder lost through runoff or deep infiltration. In this embodiment, appropriate soil and topographic conditions were selected for the demonstration area. kThe effective rainfall P in each cell at a 3km grid scale was calculated. e .

[0043] S62, Crop Distribution Data and K c Assign a value.

[0044] High-resolution distribution data of major crops (wheat, corn, and rice) within the demonstration area were acquired, and a crop distribution mask raster (resolution 10–30 m) was constructed. This crop mask represents land use using a fine-grid raster, where each cell identifies the corresponding crop type or non-arable land. First, the crop mask was filtered to extract the set of cells marked as wheat, corn, or rice; other non-target crop or non-farmland areas were not included in subsequent irrigation demand calculations. After establishing the spatial extent of the target crops, a corresponding K was assigned to each cell based on the crop type and the current date. c The crop coefficient reflects the relative water requirement of a specific crop at different growth stages. It is determined in advance, based on agricultural trials or the FAO irrigation manual, by setting daily K values ​​for the entire growth period of wheat, corn, and rice in different planting provinces. c Reference table. Based on the demonstration date (May 8th), and considering the crop phenological stage, the K value for each target crop is taken as the corresponding date. c The value is assigned to the corresponding pixel. For example, if May 8th is in the late grain-filling stage of winter wheat, and the growth cycle is K... c If the value is 1.1, then the K value of all wheat pixels on that day is... c If the value is 1.1 or the corn is in a harvested state, no calculation is performed; if the corn is in the seedling stage, then K... c It may be around 0.4; K during the rice seedling stage c Approximately 0.8, and so on. Through the above matching, each crop pixel ranging from 10 to 30 meters obtained its crop coefficient K for that day. c The input data now forms a hierarchical configuration: the weather grid (3km) provides ET0 and P for each unit. e Fine-scale crop grids (10–30 m) provide the crop species and Kelvin for each pixel. c property.

[0045] S63, crop evapotranspiration ET c calculate.

[0046] Meteorological data and crop masks are spatially overlaid and calculated using a raster grid. Spatially, each high-resolution crop pixel belongs to a specific 3km meteorological grid cell. Based on this, the actual crop evapotranspiration (ET) is calculated for each crop pixel. c The calculation formula is: ; The crop evapotranspiration at a specific growth stage is obtained by multiplying the reference evapotranspiration, corrected by the bias model, by the crop coefficient. Specifically, the ET of the 3km grid to which the pixel belongs is used in the calculation. 0,corr The value, and the K of that pixel c Multiply by each other to obtain the daily crop evapotranspiration ET for that pixel. c (Unit: mm). This step is performed in batches on all target crop pixels in a GIS raster environment to generate ET. c High-resolution spatial distribution data (10–30 m resolution raster).

[0047] S64. Calculation of Net Irrigation Requirement (NIR).

[0048] Net Irrigation Requirement (NIR) is calculated pixel-by-pixel by combining crop evapotranspiration and effective rainfall. NIR is defined as the portion of crop water requirement that needs to be met by irrigation, and is obtained by subtracting effective rainfall from crop evapotranspiration. For each crop pixel, its ET (Emergency Threshold) is taken. c Subtract the effective rainfall P of the 3km grid where the pixel is located e The initial irrigation requirement was calculated as follows: ; If the difference is positive, it indicates that part of the crop's evapotranspiration demand cannot be met by the rainfall on that day and needs to be provided by irrigation. This difference is the required net irrigation water depth (unit: mm). If the difference is negative or zero, it indicates that the rainfall on that day has fully met or exceeded the crop's water consumption, and irrigation is not required. In this case, NIR is considered to be 0. To ensure the physical reasonableness of the results, this embodiment applies ET to all occurrences. c <P e The NIR values ​​of all pixels are uniformly corrected to 0, and negative values ​​are no longer retained. After this step of calculation, a raster map of the net irrigation water requirement of wheat, corn, and rice crops in the demonstration area is obtained.

[0049] S7. Results entry into the database and online publication.

[0050] The results are exported in GeoTIFF raster format and loaded into the QGIS database as intermediate data. They are then visualized and published through the platform (using GeoServer map service combined with the Leaflet front-end). The platform supports multi-layer overlay browsing, time slider selection, and statistical analysis by administrative region or custom area, thereby enabling the sharing and publication of irrigation water demand information.

[0051] The calculated Net Irrigation Requirement (NIR) raster product is imported into the open-source GIS software QGIS in GeoTIFF format. Consistency checks are performed on spatial reference, pixel resolution, numerical range, and invalid value encoding to ensure data quality and long-term preservation. After successful verification, it is registered as raster data storage in GeoServer and published as a standard map service (WMS / WMTS). Appropriate styles (SLD) and tile caching are configured to improve online rendering and access performance. The front-end loads the above services based on Leaflet, enabling overlay display of NIR data with thematic layers such as weather forecasts, time dimension switching, and regional statistical display. End users can obtain the spatial distribution of NIR at the field unit scale via web pages or mobile devices for irrigation scheduling and water management decisions. This step does not change the raster data values; it only completes the quality control, service publishing, and sharing of the results.

[0052] S8, Spatial Resolution and Processing Flow Summary.

[0053] In summary, this specific implementation method establishes a complete computational chain from weather forecasting to field irrigation demand, and achieves seamless processing of multi-source, multi-scale data. First, a 3km resolution forecast weather grid provides daily reference evapotranspiration and precipitation at the regional scale. ET is then calculated using bias correction and empirical formulas. 0,corr and P e (3km resolution). Then, a crop distribution mask with a resolution of 10–30m is introduced, and high-resolution land use information is overlaid with coarse-resolution meteorological elements: for each pixel in the mask marked as the target crop, the ET0 value and P of its corresponding 3km grid are extracted. e The value is obtained, and the corresponding K value for that crop on that day is read. c Value. Based on these three parameters, raster operations are performed on a per-pixel basis to calculate the ET value of that pixel. c =ET0×K c And net irrigation water requirement NIR=ET c -P e After processing all pixels sequentially, a high-resolution NIR result layer consistent with the crop mask raster is output, where the value of each pixel represents the net irrigation water requirement (mm) for the crop at that location on the demonstration day. Through the above layered calculation process, the organic combination of regional-scale weather forecasting and local field-scale irrigation decision-making is achieved, and the resulting NIR spatial distribution can provide a scientific basis for refined irrigation scheduling.

[0054] This invention also provides a system for multi-source remote sensing daily-scale plot-level crop irrigation water demand forecasting, comprising: Deviation modeling module: includes ET0 deviation submodule and precipitation level mapping submodule; Model transfer and correction module: configured with cell-level linear bridging coefficients a and b; Net radiation correction and ETO calculation module: used for calculating R based on hierarchical rules. n Make corrections; Effective rainfall and water demand calculation module: used to implement P e Calculation and NIR nonnegativity constraint; Results publishing module: Used to publish raster products as geographic information services and implement time slider and regional statistics functions.

[0055] Therefore, this invention employs the aforementioned multi-source remote sensing daily-scale plot-level crop irrigation water demand forecasting method and system. It utilizes multi-source fusion of ERA5-Land reanalysis data, NCEI station data, and CMA-MESO meteorological forecast data. Machine learning and quantile mapping are used to construct a deviation model between ETO and precipitation (P), which is then transferred to the forecast field. This is combined with plot-level crop layers and dynamic crop coefficients (K... c This invention estimates the net irrigation water requirement (NIR) of crops daily using a curve and builds an online map service to publish rasterized results periodically. It solves problems such as inconsistent data source definitions, insufficient accuracy in irrigation water requirement estimation, lack of nationwide online products, and inadequate localization adaptation. It enables high spatiotemporal resolution dynamic prediction of NIR at the crop scale, significantly improving the timeliness and accuracy of regional irrigation management.

[0056] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A multi-source remote sensing method for daily-scale plot-level crop irrigation water demand forecasting, characterized in that, Includes the following steps: S1. Data preprocessing: Obtain historical data and corresponding measured data of the target area, unify the units and time base, aggregate to the daily scale and extract modeling features; S2. Reference crop evapotranspiration deviation modeling: A reference crop evapotranspiration deviation model is established based on historical data to characterize the forecast deviation of daily reference evapotranspiration ET0; S3. Precipitation deviation modeling: A precipitation deviation model is established based on historical data to characterize the forecast deviation of daily precipitation P; S4. Model Transfer and Forecast Correction: The reference crop evapotranspiration bias model and precipitation bias are transferred and applied to mesoscale numerical weather prediction products to correct the biases of the daily forecasts ET0 and P, obtaining the corrected daily reference evapotranspiration ET. 0,corr With the corrected daily precipitation P corr ; S5. Net Radiation Correction and ETO Calculation: The daily net radiation R is adjusted at the forecast end according to the tiered rules. n The daily reference evapotranspiration ET0 was adjusted and calculated and updated using the Penman-Monteith formula. S6. Calculation of Crop Evapotranspiration and Irrigation Water Requirement: Using crop spatial distribution maps and crop growth period information, generate the crop coefficient K. c The daily variation curve is based on the corrected daily reference evapotranspiration ET. 0,corr Calculate crop evapotranspiration ET c ; Crop evapotranspiration ET c Deduct effective precipitation P e The net irrigation water requirement (NIR) is obtained. S7. Results Output and Publication: ET c The NIR output is a raster product, and it is regularly published and updated online.

2. The multi-source remote sensing daily-scale plot-level crop irrigation water demand forecasting method according to claim 1, characterized in that: The reference crop evapotranspiration deviation model in S2 is trained using a random forest regression algorithm. The training samples consist of paired data from historical ETO and measured data from ground meteorological stations on the same location and day.

3. The multi-source remote sensing daily-scale plot-level crop irrigation water demand forecasting method according to claim 1, characterized in that: The precipitation deviation model in S3 is constructed using quantile distribution mapping, and a mapping relationship for the cumulative distribution of wet days is established based on paired samples of historical precipitation and measured precipitation data from ground stations on the same site and day.

4. The multi-source remote sensing daily-scale plot-level crop irrigation water demand forecasting method according to claim 1, characterized in that, The S4 adopts a cell-level linear bridging relationship, and the specific formula is as follows: y = a × x + b ； in, x For the forecast field, ET0 or P; a , b The bridging coefficients, obtained by pixel-wise fitting during the overlap period between reanalysis and forecast data, are used to transfer the bias relationships formed in historical stages to the forecast.

5. The multi-source remote sensing daily-scale plot-level crop irrigation water demand forecasting method according to claim 1, characterized in that: The mesoscale numerical weather forecast product in S4 has a spatial resolution of 3km. It operates daily to generate daily-scale irrigation water demand forecasts for the next day or the current day, and is updated daily at 08:00 Beijing time.

6. The multi-source remote sensing daily-scale plot-level crop irrigation water demand forecasting method according to claim 1, characterized in that: The hierarchical rules in S5 are as follows: When P ≥ 10 mm on that day, take R. n =0; When 0 < P < 10 mm, take R n as the lower quantile of the reanalysis of R n in the most recent 7 days; When P=0, take R n For the recent 7-day reanalysis of R n The arithmetic mean of .

7. The multi-source remote sensing daily-scale plot-level crop irrigation water demand forecasting method according to claim 1, characterized in that, The effective precipitation P in S6 e The calculation formula is: ； in, k The coefficient is an empirical coefficient representing the effective utilization rate of rainfall. It is selected within the range of [0.7, 0.85] based on soil infiltration conditions and rainfall characteristics, and varies with space and time.

8. The multi-source remote sensing daily-scale plot-level crop irrigation water demand forecasting method according to claim 1, characterized in that: The spatial resolution of the crop spatial distribution map in S6 is 10-30m, K c ET is assigned a value daily according to phenological stages. c The NIR is calculated at the crop pixel level, and a nonnegativity constraint is applied to the NIR.

9. The multi-source remote sensing daily-scale plot-level crop irrigation water demand forecasting method according to claim 1, characterized in that, The content published online in S7 specifically includes: publishing raster products using geographic information services, browsing raster products with time sliders and regional statistics, and the statistical caliber includes mean, total and area-weighted indicators.

10. A system for forecasting crop irrigation water demand at the daily scale using a multi-source remote sensing method for plot-level crop irrigation as described in any one of claims 1-9, characterized in that, include: Deviation modeling module: includes ET0 deviation submodule and precipitation level mapping submodule; Model transfer and correction module: configured with cell-level linear bridging coefficients a and b; Net radiation correction and ETO calculation module: used for calculating R based on hierarchical rules. n Make corrections; Effective rainfall and water demand calculation module: used to implement P e Calculation and NIR nonnegativity constraint; Results publishing module: used to publish raster products as geographic information services and implement time sliders and regional statistics.

Images