Method and device for estimating irrigation water use in a field based on remote sensing and ground collaboration

Abstract

The application discloses a kind of based on remote sensing and ground cooperation's field irrigation water consumption estimation method and device, belong to remote sensing technical field.It includes: obtaining multi-source remote sensing data and extracting optical vegetation index, surface albedo, surface temperature and backscattering coefficient;Obtain meteorological data, soil volume water content measured value and daily actual evapotranspiration measured value;Based on optical vegetation index, surface albedo, surface temperature, meteorological data and daily actual evapotranspiration measured value, construct deep neural network and predict daily actual evapotranspiration estimate value;According to backscattering coefficient, optical vegetation index and soil volume water content measured value, estimate root zone soil water storage change amount;According to daily actual evapotranspiration estimate value, root zone soil water storage change amount and effective precipitation, calculate the total irrigation water consumption estimate value of complete irrigation cycle.The application realizes the irrigation water consumption estimation of high precision of field scale, large range, all-weather continuous without need of canal head metering facility.

Claims

1. A method for estimating field irrigation water consumption based on remote sensing and ground-based collaboration, characterized in that, The method includes: S1: Acquire optical remote sensing images, thermal infrared remote sensing images, and SAR images of the target study area for the complete irrigation cycle, and extract the optical vegetation index, surface albedo, surface temperature, and backscattering coefficient of each field in the target study area based on the optical remote sensing images, thermal infrared remote sensing images, and SAR images. S2: Obtain meteorological data, measured soil volumetric water content, and measured daily evapotranspiration for each field plot collected by the ground calibration station network; S3: Based on the optical vegetation index, surface albedo, surface temperature, meteorological data and measured daily actual evapotranspiration of each field, a deep neural network is constructed to predict the estimated daily actual evapotranspiration of each field. S4: The change in root zone soil water storage in each field is estimated based on the measured values ​​of backscattering coefficient, optical vegetation index and soil volumetric water content. S5: Based on the estimated daily actual evapotranspiration, changes in root zone soil water storage, and effective precipitation for each field, the estimated total irrigation water consumption for the complete irrigation cycle of each field is calculated using the water balance equation. The ground calibration station network includes a multi-depth soil volumetric water content sensor group, an automatic weather station, and an eddy covariance flux observation station. S2 includes: S21: Obtain the measured values ​​of soil volumetric water content at various depths from a multi-depth soil volumetric water content sensor group installed in representative fields in the target study area. The representative fields are selected according to soil type zoning and irrigation water source type, and the burial depth of the multi-depth soil volumetric water content sensor group covers the depth range of the crop root zone. S22: Obtain meteorological data collected by automatic weather stations installed in the representative fields; The meteorological data includes near-surface air temperature, wind speed, relative humidity, net radiation, and precipitation. S23: Obtain the measured daily actual evapotranspiration from a eddy covariance flux observation station installed in at least one representative field at the center of the target study area; S3 includes: S31: Construct a deep neural network; The inputs of the deep neural network include surface temperature, optical vegetation index, surface albedo, near-surface air temperature, wind speed, relative humidity, and net radiation. The outputs of the deep neural network include estimated daily actual evapotranspiration and intermediate variables for sensible heat flux estimation. S32: The deep neural network is trained using the following loss function: in, These are the total loss, data fitting loss, and physical constraint loss, respectively. These are the hyperparameters for physical constraint weights; and These are the estimated daily actual evapotranspiration and the measured daily actual evapotranspiration for the i-th training sample, respectively. , where n is the total number of training samples used to calculate the data fitting loss; These represent the intermediate variables for estimating net radiation, soil heat flux, sensible heat flux, and estimated daily actual evapotranspiration for the j-th training sample, respectively. m is the total number of training samples used to calculate the physical constraint loss. The latent heat of vaporization of water; S33: Input the surface temperature, optical vegetation index, surface albedo, near-surface air temperature, wind speed, relative humidity and net radiation of each field into the trained deep neural network to obtain the estimated value of the daily actual evapotranspiration of each field.

2. The method for estimating field irrigation water consumption based on remote sensing and ground coordination according to claim 1, characterized in that, S1 includes: S11: Extract the optical vegetation index and surface albedo of each pixel in the target study area based on the optical remote sensing image. The optical vegetation index is either the normalized vegetation index or the enhanced vegetation index. S12: Extract the surface temperature of each pixel in the target study area based on the thermal infrared remote sensing image; S13: Extract the backscattering coefficient of each pixel in the target study area based on the SAR image; S14: Resample the pixels of the optical remote sensing image, thermal infrared remote sensing image and SAR image to the same spatial resolution; S15: Obtain the field vector boundary of each field, and perform weighted averaging of the optical vegetation index, surface albedo, surface temperature and backscattering coefficient of each pixel contained within the field vector boundary to obtain the optical vegetation index, surface albedo, surface temperature and backscattering coefficient of each field.

3. The method for estimating field irrigation water consumption based on remote sensing and ground coordination according to claim 2, characterized in that, S4 includes: S41: Decompose the backscattering coefficient of each field into the scattering contribution of the vegetation layer and the scattering contribution of the soil layer, and calculate the scattering contribution of the soil layer. in, These are the backscattering coefficient, the scattering contribution from the vegetation layer, and the scattering contribution from the soil layer, respectively. The vegetation layer scattering contribution and the vegetation layer two-way attenuation factor are obtained by parameterizing the optical vegetation index. S42: Pair the soil layer scattering contribution with the measured value of soil volumetric water content in the soil surface layer, train a random forest regression model to invert the soil volumetric water content in the surface layer, and perform linear deviation calibration using the measured value of soil volumetric water content in the soil surface layer. in, This represents the inverted surface soil volumetric water content. This is the calibrated topsoil volumetric water content. , Here, t represents the calibration coefficient, and t represents time. S43: Based on the measured values ​​of soil volumetric water content at various depths, establish a statistical transfer function between the surface soil and the soil water content of each layer in the root zone; in, For soil depth, For soil depth The corresponding soil volumetric water content, For field attribute vectors, Represents a nonlinear mapping function; S44: Two moments before and after the irrigation event and The change in soil water storage in the root zone is obtained by integrating within the depth range of the crop root zone. in, This represents the change in soil water storage in the root zone. This refers to the depth of the crop root zone.

4. The method for estimating field irrigation water consumption based on remote sensing and ground coordination according to claim 3, characterized in that, S5 includes: S51: Divide the complete irrigation cycle into several consecutive time periods; S52: Calculate the effective precipitation in each time period; in, For effective precipitation, The effective utilization coefficient of precipitation is based on soil permeability parameters. Ground slope and rainfall intensity Let P be the cumulative value of precipitation during the specified time period; S53: Calculate the uncut estimate of net irrigation water consumption for the specified period based on the field-scale water balance equation. ; in, This is the cumulative value of the estimated daily actual evapotranspiration during the specified period; S54: Set reasonable physical constraints to truncate the time period and obtain the net irrigation water consumption for that time period. ; S55: The net irrigation water consumption of each time period is summed up to obtain an estimated total irrigation water consumption.

5. The method for estimating field irrigation water consumption based on remote sensing and ground coordination according to claim 4, characterized in that, The method further includes: S6: Based on the measured irrigation metering records within the target study area, perform systematic error correction on the net irrigation water consumption and output a 90% confidence interval; Wherein, S6 includes: S61: Select calibration plots with measured irrigation records from the target study area as the training sample set, and train them using the quantile loss function. , and The spatial random forest quantile regression model with three quantile levels is denoted as follows: , and ; The quantile loss function for: in, This represents the output value of the spatial random forest quantile regression model. To measure the actual amount of irrigation water, The target quantile level; S62: Calculate the median model residuals on the calibration plots. For fields within the target study area without actual irrigation metering records, the residuals of each calibration field are interpolated to obtain the residual field for the target study area. ; in, The feature vector of the field. The spatial coordinates of the field; S63: The net irrigation water consumption is systematically corrected using the following formula, and a 90% confidence interval is output: in, This is the net irrigation water consumption value after system error correction. These are the lower and upper bounds of the 90% confidence interval, respectively.

6. A field irrigation water consumption estimation device based on remote sensing and ground coordination, characterized in that, The device includes: The multi-source remote sensing data acquisition module is used to acquire optical remote sensing images, thermal infrared remote sensing images, and SAR images of the target study area for the complete irrigation cycle, and to extract the optical vegetation index, surface albedo, surface temperature, and backscattering coefficient of each field in the target study area based on the optical remote sensing images, thermal infrared remote sensing images, and SAR images. The ground calibration data acquisition module is used to acquire meteorological data, measured values ​​of soil volumetric water content, and measured values ​​of daily actual evapotranspiration for each field plot collected by the ground calibration station network. The actual evapotranspiration inversion module is used to construct a deep neural network based on the optical vegetation index, surface albedo, surface temperature, meteorological data and measured daily actual evapotranspiration of each field, and to predict the estimated daily actual evapotranspiration of each field. The module for calculating the change in root zone soil water storage is used to estimate the change in root zone soil water storage for each field based on the measured values ​​of backscattering coefficient, optical vegetation index, and soil volumetric water content. The irrigation water consumption calculation module is used to calculate the estimated total irrigation water consumption for each field during the complete irrigation cycle based on the water balance equation, according to the estimated daily actual evapotranspiration, changes in root zone soil water storage, and effective precipitation. The ground calibration station network includes a multi-depth soil volumetric water content sensor group, an automatic weather station, and an eddy covariance flux observation station. The ground calibration data acquisition module includes: The moisture content data acquisition unit is used to acquire the measured values ​​of soil volumetric moisture content at various depths collected by a multi-depth soil volumetric moisture content sensor group installed in representative fields in the target study area. The representative fields are selected according to soil type zoning and irrigation water source type, and the burial depth of the multi-depth soil volumetric water content sensor group covers the depth range of the crop root zone. A meteorological data acquisition unit is used to acquire meteorological data collected by automatic weather stations installed in the representative field. The meteorological data includes near-surface air temperature, wind speed, relative humidity, net radiation, and precipitation. The evapotranspiration data acquisition unit is used to acquire the actual daily evapotranspiration measured by the eddy covariance flux observation station installed in at least one representative field at the center of the target study area. The actual evapotranspiration inversion module includes: Network building blocks, used to build deep neural networks; The inputs of the deep neural network include surface temperature, optical vegetation index, surface albedo, near-surface air temperature, wind speed, relative humidity, and net radiation. The outputs of the deep neural network include estimated daily actual evapotranspiration and intermediate variables for sensible heat flux estimation. The training unit is used to train the deep neural network using the following loss function: in, These are the total loss, data fitting loss, and physical constraint loss, respectively. These are the hyperparameters for physical constraint weights; and These are the estimated daily actual evapotranspiration and the measured daily actual evapotranspiration for the i-th training sample, respectively. , where n is the total number of training samples used to calculate the data fitting loss; These represent the intermediate variables for estimating net radiation, soil heat flux, sensible heat flux, and estimated daily actual evapotranspiration for the j-th training sample, respectively. m is the total number of training samples used to calculate the physical constraint loss. The latent heat of vaporization of water; The inference unit is used to input the surface temperature, optical vegetation index, surface albedo, near-surface air temperature, wind speed, relative humidity, and net radiation of each field into the trained deep neural network to obtain the estimated daily actual evapotranspiration of each field.

7. A computer-readable storage medium for estimating field irrigation water consumption based on remote sensing and ground coordination, characterized in that, It includes a memory for storing processor-executable instructions, which, when executed by the processor, implement the steps of the field irrigation water estimation method based on remote sensing and ground coordination as described in any one of claims 1-5.

8. A device for estimating field irrigation water consumption based on remote sensing and ground coordination, characterized in that, It includes at least one processor and a memory storing computer-executable instructions, wherein the processor executes the instructions to implement the steps of the field irrigation water estimation method based on remote sensing and ground coordination as described in any one of claims 1-5.

Images