Land water reserve change grid data processing method and storage medium

Abstract

The invention discloses a land water reserve change grid data processing method and a storage medium, and belongs to the field of data processing for a prediction purpose. In order to solve the problem of processing precision loss in a land water reserve change grid data downscaling processing process, the method comprises the following steps: firstly, acquiring two types of hydro meteorological grid data and multi-source original land water reserve change grid data of a research area; fusing the multi-source original data by applying a generalized triangular cap method and a least square method to obtain fused data; removing elements of first hydro meteorological grid data one by one from grid points in a research area, inputting the elements into an LSTM prediction model, and screening out key hydro meteorological elements with influence precision exceeding a threshold value through simulation of a residual error correction method; dividing grid point sub-regions according to key elements, training each sub-region based on an LSTM prediction model and a residual error correction method, and obtaining an optimal model and a corresponding data subset; and correcting the data containing the extreme point months by using the optimal model to obtain a final result. The method is used for a remote sensing system and has the advantage of high precision.

Description

Technical Field

[0001] The present invention relates to the field of data processing for prediction purposes, and in particular to a method for processing grid data of land water storage changes and a storage medium. Background Art

[0002] The prediction research on the change of terrestrial water storage can be divided into small and medium-scale areas and large-scale areas according to different spatial scales. Small and medium-scale areas include river basins, provincial or sub-national areas, and their prediction work on the change of terrestrial water storage focuses on the high-temporal and spatial resolution water storage changes in local areas such as the Yangtze River Basin and the Sichuan Basin. By capturing the hydrological dynamics in short time scales (from days to months), such as floods and droughts caused by extreme rainfall, it provides key basis for regional water resources management, ecological and environmental protection, and disaster warning. Large-scale areas include continents, countries, and even the global scope. The prediction of subtle changes in terrestrial water storage at this scale mainly reveals the long-term water storage change trend at the global or continental level, such as interannual changes and evolution trends over decades. At the same time, in-depth analysis of the impact of climate change and human activities (such as agricultural irrigation, reservoir scheduling, etc.) on water resources is carried out to support the formulation of global water security policies.

[0003] In order to convert low spatial resolution grid data of land water storage changes into high spatial resolution grid data of land water storage changes (i.e., spatial downscaling), the current downscaling methods mainly include: (1) statistical downscaling method, (2) physical model downscaling method, and (3) data assimilation method. Statistical downscaling method uses the statistical correlation between GRACE data and high-resolution hydrological and meteorological data to improve the resolution by using regression analysis or machine learning. However, such methods usually lack physical constraint mechanisms, resulting in poor generalization ability. Physical model downscaling method combines land surface models or numerical simulations with them, and uses high-resolution models as prior knowledge to improve spatial resolution. However, existing physical models have uncertainty and spatial heterogeneity problems. Data assimilation method uses filtering to fuse GRACE data with high-resolution hydrological models to enhance spatial resolution, but its filtering calculation cost is high and it is extremely sensitive to initial conditions and parameters.

[0004] To this end, this field first needs to find a simple spatial downscaling method that can handle complex multi-source data, accurately capture the long-term dependencies of different time series, have excellent time series prediction capabilities, and can automatically adjust learning content and make efficient predictions.

[0005] Secondly, traditional spatial downscaling methods generally face the problem of spatial heterogeneity. When dividing grid points, existing methods usually simply divide them according to administrative regions, such as dividing the Yangtze River into the upper reaches, middle reaches, and lower reaches of the Yangtze River. However, different regions have different impacts on terrestrial water storage. If the same prediction model is used, it is impossible to accurately consider the situation in all regions. Therefore, the simple grid point division method leads to poor spatial downscaling processing accuracy.

[0006] Finally, the existing technology has always ignored the impact of extreme points caused by mutation anomalies and nonlinear changes in the grid data of terrestrial water storage changes on the accuracy of spatial downscaling. Therefore, it has become more urgent to solve the problem of insufficient processing ability of spatial downscaling methods for nonlinear relationships.

[0007] Prior art 1: Lilu Cui, Yu Li, et al. Assessing the impact of 2022extreme drought on the Yangtze River basin using downscaled GRACE / GRACE-FOdata obtained by partitioned random forest algorithm. Vol. 3 Issue 3, 2024, International Journal of Remote Sensing.

[0008] Prior art 1 proposes a partitioned random forest downscaling method, which can effectively improve the accuracy compared with the traditional global random forest downscaling method. However, prior art 1 does not propose a more effective solution to the above technical problems. Summary of the invention

[0009] In order to alleviate or partially alleviate the above technical problems, the solution of the present invention is as follows:

[0010] A method for processing grid data of land water storage change, comprising:

[0011] Step S1: Acquire the first hydrological and meteorological grid data and the second hydrological and meteorological grid data in the study area and the original terrestrial water storage change grid data from multiple sources;

[0012] Step S2: using the generalized triangular hat method and the least squares method, the original land water storage change grid data from multiple sources are fused to obtain the fused land water storage change grid data;

[0013] Step S3: for any grid point in the study area, after excluding each hydrological and meteorological element in the first hydrological and meteorological grid data one by one, the LSTM prediction model is input, and the simulated terrestrial water storage change grid data is obtained by the residual correction method, and the hydrological and meteorological elements whose influence on the accuracy of the simulated terrestrial water storage change grid data is greater than the preset threshold are retained as key hydrological and meteorological elements;

[0014] Step S4: Divide the grid points with completely identical key hydrological and meteorological elements into a sub-region; for each sub-region, train and obtain an optimal LSTM prediction model based on the LSTM prediction model and the residual correction method, and use the optimal LSTM prediction model to obtain the terrestrial water storage change grid data for each sub-region, and form a subset of the first terrestrial water storage change grid data;

[0015] Step S5: For each sub-region and the month containing extreme points in the fused land water reserve change grid data, the optimal LSTM prediction model is used to obtain the land water reserve change grid correction data for the month, and the land water reserve change grid correction data for the month are used to replace the first land water reserve change grid data for the month to obtain the final land water reserve change grid data.

[0016] Furthermore, the spatial resolution of the first hydrological and meteorological grid data is smaller than the spatial resolution of the second hydrological and meteorological grid data.

[0017] Furthermore, the spatial resolution of the original land water storage change grid data from the multiple sources is lower than the spatial resolution of the final land water storage change grid data, and lower than the spatial resolution of the first land water storage change grid data.

[0018] Furthermore, the residual correction method comprises the following steps:

[0019] The grid prediction data of land water storage change output by the LSTM prediction model and the fused grid data of land water storage change are subtracted and then interpolated to obtain the second residual grid data as the correction data.

[0020] Furthermore, the residual correction method specifically includes the following steps:

[0021] The first hydrological and meteorological grid data and the fused land water storage change grid data are simultaneously used as inputs of the LSTM prediction model to obtain the land water storage change grid prediction data;

[0022] Subtracting the land water storage change grid prediction data from the fused land water storage change grid data to obtain the first residual grid data, and then using the Kriging interpolation method to interpolate the first residual grid data into the second residual grid data;

[0023] The second hydrological and meteorological grid data is used as the input of the LSTM prediction model to obtain the second land water storage change grid data, and the second land water storage change grid data is added to the second residual grid data to obtain the target land water storage change grid data.

[0024] Furthermore, when the fused grid data of land water storage changes are obtained, the optimal weight for fusing the data is obtained by the least squares method.

[0025] Furthermore, any type of hydrological and meteorological grid data input into the LSTM prediction model is first subjected to standardization processing.

[0026] Furthermore, the neurons in the LSTM prediction model use the Sigmoid activation function.

[0027] Furthermore, for different sub-regions, the optimal LSTM prediction model applied is different.

[0028] On the other hand, the present invention also discloses a storage medium storing a computer program / instruction, which, when executed or compiled by a processor, implements the steps of the method for processing grid data of land water storage changes as described in any of the preceding items.

[0029] The technical solution of the present invention has one or more of the following beneficial technical effects:

[0030] (1) The grid data at the extreme values ​​in the time series are processed independently using a long short-term memory network to reduce the impact of nonlinear changes on the accuracy of spatial downscaling results.

[0031] (2) By predicting all input hydrological and meteorological elements based on the long short-term memory network, the key hydrological and meteorological elements of different grid points are determined, which effectively weakens the adverse effects of spatial heterogeneity on the accuracy of spatial downscaling processing.

[0032] (3) A spatial downscaling method based on long short-term memory network can capture long-term dependencies in time series, retain and output key information through gating mechanisms and memory units, and its processing method is simple, efficient, and gradient stable.

[0033] In addition, other beneficial effects of the present invention will be mentioned in the specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a flow chart of the grid data processing method of land water storage change. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] In order to clearly describe the technical solutions of the embodiments of the present invention, in the embodiments of the present invention, the words "first", "second", etc. are used to distinguish the same items or similar items with basically the same functions and effects. Those skilled in the art will understand that the words "first", "second", etc. do not limit the quantity and execution order. In addition, the "multiple" described in the present invention refers to at least 3.

[0037] Terminology explanation:

[0038] Gravity Recovery And Climate Experiment (GRACE): refers to a space Earth science mission that monitors global terrestrial water storage, ocean circulation, glacier melting and other geophysical phenomena by measuring changes in the Earth's gravity field, providing key GRACE data for studying climate change, water resource distribution and Earth system science. The GRACE mission measures changes in the Earth's gravity field through the GRACE satellite, and obtains data on changes in terrestrial water storage at a relatively low spatial resolution on a global scale.

[0039] Global Land Data Assimilation System (GLDAS): The main purpose of GLDAS is to provide estimates of land surface hydrological and meteorological variables with high temporal and spatial resolution on a global scale by integrating multiple observational data and model simulation results. It integrates satellite observation data, ground observation data and the output of numerical weather forecast models, and uses data assimilation technology to optimize model parameters and initial conditions, thereby generating more accurate and comprehensive land surface status information.

[0040] Long-Short Term Memory (LSTM): It is a special type of recurrent neural network in the field of artificial intelligence, which is good at and commonly used to process sequence data such as speech.

[0041] Three-Corned Hat (TCH): It is a method for estimating the error variance of each data set by analyzing the relationship between multiple data sets. It is used to quantify the uncertainty of multiple data sets (generally at least 3).

[0042] Study area: is the geographical area targeted or selected by the data processing method of the present invention.

[0043] Evapotranspiration: refers to the process by which water on the earth's surface enters the atmosphere through evaporation and plant transpiration. It is an important part of the water cycle. Evaporation is the process of converting liquid water into gaseous water, which mainly occurs on the water surface, soil surface, etc. Transpiration is the process by which plants release water from their bodies into the atmosphere through the stomata on the surface of their leaves.

[0044] Hydrometeorological grid data: a data format that organizes and stores hydrological and meteorological element information according to a certain spatial grid, which can be obtained from the China Meteorological Administration.

[0045] Land water storage change grid data: a data form used to describe the changes in the earth's land surface water storage over time and space, including information on the changes in the storage of various land water bodies (such as rivers, lakes, reservoirs, soil moisture, groundwater, etc.). Through the comprehensive monitoring and analysis of the changes in the storage of these different water bodies, the dynamic changes in land water storage in different regions and time scales can be reflected.

[0046] The method for processing grid data of land water storage change of the present invention comprises the following steps:

[0047] Step S1: Acquire the first hydrological and meteorological grid data and the second hydrological and meteorological grid data in the study area and the original terrestrial water storage change grid data from various sources.

[0048] In the present invention, the original terrestrial water storage change grid data from multiple sources may be terrestrial water storage change grid data released by multiple institutions known in the art. The present invention preferably uses at least 5 different sources of original terrestrial water storage change grid data. The original terrestrial water storage change grid data from multiple sources can be regarded as low spatial resolution terrestrial water storage change grid data in the present invention. The present invention is not limited to 5 specific original terrestrial water storage change grid data from different sources.

[0049] In the present invention, the hydrological and meteorological grid data include but are not limited to the following hydrological and meteorological elements: precipitation, evapotranspiration, snow water equivalent, snow depth, vegetation canopy moisture, runoff, surface temperature and soil moisture data.

[0050] The present invention relates to first hydrological and meteorological grid data and second hydrological and meteorological grid data, wherein the spatial resolution of the first hydrological and meteorological grid data is lower than the spatial resolution of the second hydrological and meteorological grid data, that is, the second hydrological and meteorological grid data has a high spatial resolution, and the first hydrological and meteorological grid data has a low spatial resolution.

[0051] The meteorological grid data and the terrestrial water storage change grid data are divided into different grid points according to different locations in the study area. This is a conventional processing method known in the art and will not be described in detail in the present invention.

[0052] Step S2: Using the generalized triangular hat method and the least squares method, the original land water storage change grid data from multiple sources are fused to obtain the fused land water storage change grid data.

[0053] This step S2 can improve the data accuracy of the grid data of land water storage changes and serve as a basis for data processing in subsequent steps.

[0054] Regarding the generalized triangular hat method, at least the existing technology 2 can be referred to: Using the generalized triangular hat method to evaluate the uncertainty of GRACE inversion of water storage changes in the **** region. Yao Chaolong, Li Qiong, Luo Zhicai. Chinese Journal of Geophysics, 2019. 62(3): 883-897; DOI: 10.6038 / cjg2019L0454 Specifically, a feasible example of using the generalized triangular hat method and the least squares method to fuse the original land water storage change grid data from multiple sources is: (1) Obtain raw terrestrial water storage change grid data from multiple sources with time series properties from five different GRACE products.

[0057] (2) Calculate the covariance matrix C for five original terrestrial water storage change grid data x 1 , x 2 , x 3 , x 4 ,x 5 , calculate the covariance C between them ij , where i, j = 1, 2, 3, 4, 5, and i ≠ j, and construct a 5×5 covariance matrix C, where the diagonal elements C of the covariance matrix C ii is the variance of each original land water storage change grid data.

[0058] (3) Introduce a 5×5 noise covariance matrix R. Here, it is assumed that the noise is independent and identically distributed. The noise covariance matrix R is a diagonal matrix. The diagonal elements r of the noise covariance matrix R are ii Represents the noise variance of the i-th original land water storage change grid data, where i=1,2,3,4,5.

[0059] (4) Construct the relationship between the noise covariance matrix R and the covariance matrix C. Use the objective function f and the corresponding constraint function g to solve the objective function f to obtain the elements of the noise covariance matrix R. The diagonal elements r of the noise covariance matrix R are ii , as a measure of noise variance and uncertainty, where i=1,2,3,4,5.

[0060] (5) According to the diagonal element r of the noise covariance matrix R ii Calculate the fusion weight w i , fusion weight w i Inversely proportional to the noise variance, the fusion weight w i The specific calculation formula is: , where i=1,2,3,4,5.

[0061] (6) The fusion model in the present invention is , where y is the result of fusion, x i is the ith original land water storage change grid data, w i is the i-th fusion weight, is the error term. In the present invention, the single error , the sum of squared errors , that is, the least squares method is used here, where i=1,2,3,4,5.

[0062] (7) Find the error square sum S with respect to the fusion weight w i The partial derivative of , and set the partial derivative equal to zero, to obtain a set of constraint equations, solving this set of constraint equations can obtain the optimal weight that minimizes the sum of squared errors S , where i=1,2,3,4,5.

[0063] (8) The optimal weight obtained Substitute into the fusion model , you can get the fused land water storage change grid data with time series attributes , where i=1,2,3,4,5.

[0064] Step S3: For any grid point in the study area, after excluding each hydrometeorological element in the first hydrometeorological grid data one by one, the LSTM prediction model is input, and the simulated terrestrial water storage change grid data is obtained through the residual correction method. The hydrometeorological elements whose impact on the accuracy of the simulated terrestrial water storage change grid data is greater than the preset threshold are retained as key hydrometeorological elements.

[0065] In the present invention, for any grid point, all key hydrological and meteorological elements constitute the hydrological and meteorological element combination of any grid point.

[0066] For example, in the present invention, if there are 8 hydrometeorological elements in the hydrometeorological grid data, after excluding one hydrometeorological element each time, there are still 7 hydrometeorological elements left. For example, after excluding precipitation at a certain time, the hydrometeorological elements also include: evapotranspiration, snow water equivalent, snow depth, vegetation canopy moisture, runoff, surface temperature and soil moisture data. After excluding evapotranspiration the next time, the hydrometeorological elements also include: precipitation, snow water equivalent, snow depth, vegetation canopy moisture, runoff, surface temperature and soil moisture data.

[0067] After excluding the hydrometeorological elements in the hydrometeorological grid data one by one and obtaining the simulated land water storage change grid data through LSTM and residual correction method, if a hydrometeorological element is a key hydrometeorological element of the grid point and is excluded by a certain process, the data accuracy between the obtained simulated land water storage change grid data and the previously obtained simulated land water storage change grid data will exceed the preset threshold; and if a hydrometeorological element is not a key hydrometeorological element of the grid point and is excluded by a certain process, the data accuracy between the obtained simulated land water storage change grid data and the previously obtained simulated land water storage change grid data will not exceed the preset threshold, so this method can identify whether it is a key hydrometeorological element of the grid point. In other words, if a meteorological element is a key hydrometeorological element of the grid point and is excluded by a certain process, the accuracy of the obtained simulated land water storage change grid data will fluctuate significantly, thereby determining that the excluded hydrometeorological element belongs to the key hydrometeorological element.

[0068] For example, the precision fluctuation value can be obtained by comparing with the simulated land water storage change grid data before excluding a certain hydrological and meteorological element. Further, the precision fluctuation value can be measured by the root mean square error.

[0069] In addition, the present invention will provide the specific implementation of the above residual correction method and LSTM prediction model later.

[0070] Step S4: Divide the grid points with exactly the same key hydrological and meteorological elements into a sub-region; for each sub-region, train and obtain the optimal LSTM prediction model based on the LSTM prediction model and the residual correction method, and use the optimal LSTM prediction model to obtain the terrestrial water storage change grid data for each sub-region, and constitute a subset of the first terrestrial water storage change grid data.

[0071] In other words, the first land water storage change grid data can be obtained by merging the land water storage change grid data of all sub-regions obtained according to the optimal LSTM prediction model. Compared with the original land water storage change grid data from various sources, the first land water storage change grid data is high-resolution land water storage change grid data.

[0072] In the present invention, the optimal LSTM prediction models corresponding to and applied to different sub-regions are different.

[0073] Furthermore, in step S4, the fused land water storage change grid data for 12 consecutive months may be uniformly processed to obtain the first land water storage change grid data after downscaling.

[0074] The inventors found that, similar to the common practice in the prior art, if the first terrestrial water storage change grid data obtained is simply used as the final downscaled data, the influence of extreme points caused by sudden abnormalities and nonlinear changes in the terrestrial water storage change grid data on the accuracy of spatial downscaling will be ignored.

[0075] Further, step S5: for each sub-region and the month containing extreme points in the fused land water reserve change grid data, the optimal LSTM prediction model is used to obtain the land water reserve change grid correction data about the month, and the land water reserve change grid correction data about the month is used to replace the first land water reserve change grid data about the month to obtain the final land water reserve change grid data.

[0076] The final land water storage change grid data is land water storage change grid data with high spatial resolution, and is the processing result that the present invention ultimately hopes to obtain.

[0077] In the present invention, the specific implementation method of the residual improvement method is as follows:

[0078] (1) The first hydrological and meteorological grid data and the fused land water storage change grid data are simultaneously used as the input of the LSTM prediction model to obtain the land water storage change grid prediction data.

[0079] (2) Subtract the predicted land water storage change grid data from the fused land water storage change grid data to obtain the first residual grid data, and then use the Kriging interpolation method to interpolate the first residual grid data into the second residual grid data.

[0080] In comparison, the spatial resolutions of the second residual grid data and the land water storage change grid prediction data in the present invention are both low-resolution grid data, and the second residual grid data are high-resolution grid data.

[0081] (3) Use the second hydro-meteorological grid data as the input of the LSTM prediction model to obtain the second terrestrial water storage change grid data, and add the second terrestrial water storage change grid data to the second residual grid data to obtain the target terrestrial water storage change grid data.

[0082] In the present invention, a feasible embodiment of the Kriging interpolation method is as follows:

[0083] (1) Determine the longitude coordinates and latitude coordinates of the known low-spatial-resolution terrestrial water storage change grid data within the study area.

[0084] (2) Calculate the variogram to describe the spatial variability of the regionalized variable, where the variogram can be expressed as:

[0085] ;

[0086] where the variogram value represents the number of sample pairs at a spacing of h, N(h) represents the number of sample pairs at a spacing of h, and are the observed values at the known location points and the location point respectively, and i is the sequence number value.

[0087] (3) Determine the interpolation points, set as a regular grid of 0.1°×0.1°, where "°" represents the unit in the coordinate system composed of longitude and latitude.

[0088] (4) For each interpolation point, establish the Kriging equations according to the spatial positions of the known points and the variogram . Specifically, the Kriging equations are:

[0089]

[0090] where is the weight coefficient of the jth known location point , is the Lagrange multiplier, and are both known location points, is the interpolation point, and are the variogram values between the known location points and the known location point respectively, and the variogram value between the known location point and the interpolation point . By solving the above Kriging equations, the weight coefficient of any known location point is obtained. , where i, j are serial numbers and n is a positive integer.

[0091] (5) According to the calculated weight coefficients of each known position point, the observed values ​​of the known position points are weighted summed to obtain the estimated value of the interpolation point. The interpolation formula is: ,in is the interpolation point The estimated value of is the jth known position point The observed value of is the jth known position point The weight coefficient of .

[0092] In addition, as a specific embodiment, the implementation scheme of the LSTM prediction model in the present invention can be implemented by the following method:

[0093] (1) Determine the high spatial resolution, monthly-scale hydrological and meteorological grid data of the study area, including precipitation, evapotranspiration, snow water equivalent, snow depth, vegetation canopy moisture, runoff, surface temperature, and soil moisture, and standardize them using the Z-score function. The standardization formula is: , where x is the hydrological and meteorological grid data, is the mean, is the standard deviation, It is the standardized hydrological and meteorological grid data. The input of the LSTM prediction model is the high-resolution, monthly-scale grid data of land water storage changes, and the output of the LSTM prediction model is the grid data.

[0094] (2) By way of example, the LSTM prediction model in the present invention comprises 1 input layer, 3 hidden layers and 1 output layer, wherein the hidden layer has 128 neurons (or cells).

[0095] (3) In the network layer of the LSTM prediction model, the learning mechanism is controlled by the forget gate, input gate, memory unit, and output gate.

[0096] In the present invention, the forget gate determines the cell state vector C at the previous time step t-1 t-1 How much information needs to be forgotten? The formula is: , where f t is the activation value, is the Sigmoid activation function, W f is the weight matrix of the forget gate, h t-1 represents the hidden state vector at the previous time step t-1, x t represents the input vector at time step t, b fis the bias vector of the forget gate, [h t-1 , x t ] means to set h t-1 and x t The concatenated vectors, t-1 and t are both time steps.

[0097] In the present invention, the input gate is used to update the decision and create candidate values. The update decision formula is , where i t Indicates the update degree of each cell state element, is the Sigmoid activation function, W i is the weight matrix of the input gate, h t-1 represents the hidden state vector at the previous time step t-1, x t represents the input vector at time step t, b i is the bias vector of the input gate, [h t-1 , x t ] means to set h t-1 and x t The concatenated vector, i is the serial number value, t-1 and t are both time steps.

[0098] Furthermore, the formula for creating candidate values ​​in the present invention is: , where the current cell state vector C t is the cell state vector at time step t, tanh is the hyperbolic tangent activation function, and W C is the weight matrix used to create candidate values, h t-1 represents the hidden state vector at the previous time step t-1, x t represents the input vector at time step t, b C is the corresponding bias vector, [h t-1 , x t ] means to set h t-1 and x t The memory unit updates the current cell state vector C according to the output of the input gate and the output of the forget gate. t , the update formula is ,in represents element-by-element multiplication, t-1 and t are both time steps, f t is the activation value, i t Indicates the update degree of each cell state element, C t-1 is the cell state vector at the previous time step t-1.

[0099] In the present invention, the output gate determines the current cell state vector C t How much information needs to be output to the hidden state output vector h at time step t? t, which contains the output decision vector and the hidden state output vector. The formula for the output decision is ,in, is the output decision vector, is the Sigmoid activation function, W o is the weight matrix of the output gate, h t-1 represents the hidden state vector at the previous time step t-1, x t represents the input vector at time step t, b o is the bias vector of the output gate, [h t-1 , x t ] means to set h t-1 and x t The concatenated vectors, t-1 and t are both time steps; in addition, the formula for the hidden state output vector is , where h t is the hidden state output vector, is the output decision vector, represents element-by-element multiplication, tanh is the hyperbolic tangent activation function, C t is the current cell state vector.

[0100] (4) By inputting standardized hydrological and meteorological grid data The LSTM prediction model is trained, and then the second hydrological and meteorological grid data with high spatial resolution is used as the input of the LSTM prediction model, and the terrestrial water storage change grid prediction data is output.

[0101] More specific implementation methods and details about the LSTM prediction model itself are common knowledge in the art and will not be elaborated in the present invention.

[0102] In addition, the present invention also discloses a storage medium storing a computer program / instruction, which implements the steps of the method for processing grid data of land water storage changes as described in any one of the above items when executed or compiled by a processor.

[0103] Table 1: Comparison of relevant indicators of the present invention and other solutions

[0104]

[0105] Table 1 shows the comparison of relevant indicators of an actual embodiment of the present invention with other solutions. Table 1 overall reflects the accuracy of different downscaling processing results, where the closer the correlation coefficient is to 1, the better, the closer the Nash coefficient is to 1, the better, the closer the consistency coefficient is to 1, the better, the smaller the root mean square error is, the better, and the smaller the mean absolute error is, the better, where mm is millimeter. It is worth mentioning that the present invention is also compared with the solution based only on LSTM.

[0106] Compared with the LSTM scheme, the correlation coefficient, Nash coefficient, consistency coefficient, root mean square error, and mean absolute error of the present invention are respectively increased by 1.01%, 4.04%, 3.09%, 67.75%, and 71.23%. Compared with the gradient boosting scheme, the correlation coefficient, Nash coefficient, consistency coefficient, root mean square error, and mean absolute error of the present invention are respectively increased by 51.51%, 74.75%, 156.41%, 91.26%, and 91.90%. Compared with the random forest scheme, the correlation coefficient, Nash coefficient, consistency coefficient, root mean square error, and mean absolute error of the present invention are respectively increased by 2.04%, 19.28%, 12.36%, 81.95%, and 83.52%.

[0107] In summary, the present invention is not only ahead of traditional solutions in multiple indicators, but also has achieved higher improvements compared to solutions that rely only on LSTM, especially in terms of root mean square error and mean absolute error. In other words, compared with solutions that rely only on LSTM, the sub-region partitioning method and the data correction method for the relevant months where the extreme points are located in the present invention have effectively improved the prediction accuracy of different spatial locations and the accuracy of the downscaling processing results, respectively, making the present invention superior to other solutions overall.

[0108] In order to better illustrate the present invention, numerous specific details are provided in the above specific embodiments. It should be understood by those skilled in the art that the present invention can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present invention.

[0109] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for processing grid data of land water storage changes, characterized in that: include: Step S1: Acquire the first hydrological and meteorological grid data and the second hydrological and meteorological grid data in the study area and the original terrestrial water storage change grid data from multiple sources; Step S2: using the generalized triangular hat method and the least squares method, the original land water storage change grid data from multiple sources are integrated to obtain the integrated land water storage change grid data; Step S3: for any grid point in the study area, after excluding each hydrological and meteorological element in the first hydrological and meteorological grid data one by one, the LSTM prediction model is input, and the simulated terrestrial water storage change grid data is obtained by the residual correction method, and the hydrological and meteorological elements whose influence on the accuracy of the simulated terrestrial water storage change grid data is greater than the preset threshold are retained as key hydrological and meteorological elements; Step S4: Divide the grid points with completely identical key hydrological and meteorological elements into a sub-region; for each sub-region, train and obtain an optimal LSTM prediction model based on the LSTM prediction model and the residual correction method, and use the optimal LSTM prediction model to obtain the terrestrial water storage change grid data for each sub-region, and form a subset of the first terrestrial water storage change grid data; Step S5: For each sub-region and the month containing extreme points in the fused land water reserve change grid data, the optimal LSTM prediction model is used to obtain the land water reserve change grid correction data for the month, and the land water reserve change grid correction data for the month are used to replace the first land water reserve change grid data for the month to obtain the final land water reserve change grid data.

2. The method for processing grid data of land water storage change according to claim 1, characterized in that: The spatial resolution of the first hydrological and meteorological grid data is smaller than the spatial resolution of the second hydrological and meteorological grid data.

3. The method for processing grid data of land water storage change according to claim 2, characterized in that: The spatial resolution of the original terrestrial water storage change grid data from the multiple sources is lower than the spatial resolution of the final terrestrial water storage change grid data, and lower than the spatial resolution of the first terrestrial water storage change grid data.

4. The method for processing grid data of land water storage change according to claim 3, characterized in that: The residual correction method comprises the following steps: The grid prediction data of land water storage change output by the LSTM prediction model and the fused grid data of land water storage change are subtracted and then interpolated to obtain the second residual grid data as the correction data.

5. The method for processing grid data of land water storage change according to claim 4, characterized in that: The residual correction method specifically comprises the following steps: The first hydrological and meteorological grid data and the fused land water storage change grid data are simultaneously used as inputs of the LSTM prediction model to obtain the land water storage change grid prediction data; Subtracting the land water storage change grid prediction data from the fused land water storage change grid data to obtain the first residual grid data, and then using the Kriging interpolation method to interpolate the first residual grid data into the second residual grid data; The second hydrological and meteorological grid data is used as the input of the LSTM prediction model to obtain the second land water storage change grid data, and the second land water storage change grid data is added to the second residual grid data to obtain the target land water storage change grid data.

6. The method for processing grid data of land water storage change according to claim 5, characterized in that: When the fused grid data of land water storage changes are obtained, the optimal weight for fusion data is obtained by the least squares method.

7. The method for processing grid data of land water storage change according to claim 6, characterized in that: For any type of hydrological and meteorological grid data input into the LSTM prediction model, standardization is first performed.

8. The method for processing grid data of land water storage change according to claim 7, characterized in that: The neurons in the LSTM prediction model use the Sigmoid activation function.

9. The method for processing grid data of land water storage change according to claim 8, characterized in that: For different sub-regions, the optimal LSTM prediction model applied is different.

10. A storage medium having a computer program / instruction stored therein, characterized in that: When the computer program / instruction is executed by a processor or compiled and executed, the steps of the method for processing grid data of land water storage changes as described in any one of claims 1 to 9 are implemented.

Images