Multi-source evapotranspiration data fusion method based on data source independence and dynamic weight

Abstract

The invention provides a multi-source evapotranspiration data fusion method based on data source independence and dynamic weight, and the method comprises the steps: carrying out the similarity calculation between multi-source actual evapotranspiration data sets, and obtaining a similarity calculation result; based on a similarity calculation result, clustering the multi-source actual evapotranspiration data set according to data source independence to obtain a clustering result; based on the clustering result, fusing the actual evapotranspiration data set in each cluster according to the dynamic weight to obtain a fusion result in the cluster; and fusing the actual evapotranspiration data set of each cluster based on the clustering internal fusion result to obtain a multi-source actual evapotranspiration data fusion result. According to the method, a dynamic weighting scheme of different actual evapotranspiration data sets is considered, adjustment is performed for years with non-overlapping vegetation types and actual evapotranspiration data set coverage ranges, and meanwhile, the problem of independence among the multi-source actual evapotranspiration data sets is considered, so that the estimation precision of the fusion data set on the actual evapotranspiration is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data fusion, and particularly relates to a multi-source evapotranspiration data fusion method based on data source independence and dynamic weight. BACKGROUND

[0002] Actual evapotranspiration (ET) is an important component of land water cycle, carbon cycle and energy balance. Due to the high complexity of the spatio-temporal dynamic change of actual evapotranspiration and the interactive influence of multiple biophysical factors, the spatial coverage and temporal continuity of ground observation data are limited, and there are large differences between the actual evapotranspiration grid data sets obtained by different estimation methods. The differences between these actual evapotranspiration data sets lead to a large uncertainty in the current research on the spatio-temporal pattern and interannual variation trend of global actual evapotranspiration. This uncertainty not only limits the scientific understanding of the mechanism of hydrological cycle responding to global change, but also poses a serious challenge to the sustainable assessment of freshwater resources, the optimization of agricultural irrigation strategies, the prediction and early warning of drought events, and the monitoring of ecosystem functions under the background of global change.

[0003] In order to reduce the uncertainty of actual evapotranspiration data, data fusion technology is used to compare, analyze and integrate different sources of actual evapotranspiration data sets to improve the accuracy of actual evapotranspiration estimation. The input data sets of the existing method are not independent of each other, which limits the ability to effectively capture the variation of regional actual evapotranspiration, resulting in low estimation accuracy of the fusion data set for actual evapotranspiration. SUMMARY

[0004] The present application provides a multi-source evapotranspiration data fusion method based on data source independence and dynamic weight to solve the defect that the estimation accuracy of the fusion data set for actual evapotranspiration is low in the prior art. The present application considers a dynamic weighting scheme for different actual evapotranspiration data sets, adjusts for different vegetation types and years in which the coverage range of the actual evapotranspiration data sets does not overlap, and considers the independence problem between multi-source actual evapotranspiration data sets, thereby improving the estimation accuracy of the fusion data set for actual evapotranspiration.

[0005] The present application provides a multi-source evapotranspiration data fusion method, comprising: calculating the similarity between multi-source actual evapotranspiration data sets to obtain a similarity calculation result; based on the similarity calculation result, clustering the multi-source actual evapotranspiration data sets according to data source independence to obtain a clustering result; based on the clustering result, fusing the actual evapotranspiration data sets within each cluster according to a dynamic weight to obtain a clustering internal fusion result; the dynamic weight is adjusted for different vegetation types and years in which the coverage range of the multi-source actual evapotranspiration data sets overlaps; based on the clustering internal fusion result, fusing the actual evapotranspiration data sets of each cluster to obtain a multi-source actual evapotranspiration data fusion result.

[0006] According to a multi-source evapotranspiration data fusion method provided by the present invention, the step of calculating the similarity between multi-source actual evapotranspiration datasets to obtain similarity calculation results includes: determining the time series of the multi-source actual evapotranspiration datasets; calculating a trend term for the multi-source actual evapotranspiration datasets to obtain a trend term; calculating a seasonality term based on the time series and the trend term to obtain a seasonal term; calculating a residual term based on the time series, the trend term, and the seasonal term, and using the obtained residual term as the similarity calculation result.

[0007] According to a multi-source evapotranspiration data fusion method provided by the present invention, the step of clustering the multi-source actual evapotranspiration dataset based on the similarity calculation result and according to the data source independence to obtain the clustering result includes: taking each data point in the multi-source actual evapotranspiration dataset as an independent cluster according to the residual term; calculating the sum of squared distances from each data point in the cluster to the cluster centroid for each data point in the cluster, and performing cluster merging according to the sum of squared distance calculation result to minimize the variance of the merged clusters; iteratively executing the step of calculating the sum of squared distances and the step of merging clusters until clustering is completed for the multi-source actual evapotranspiration datasets under all land cover types, and obtaining the clustering result.

[0008] According to a multi-source evapotranspiration data fusion method provided by the present invention, based on the clustering results, the actual evapotranspiration datasets within each cluster are fused according to dynamic weights to obtain a fusion result within the cluster. The method includes: determining the actual evapotranspiration data sequence corresponding to the actual evapotranspiration dataset in each cluster based on the site observation dataset; performing Bayesian model averaging analysis on the actual evapotranspiration data sequence using the site observation dataset as the observation value to obtain dynamic weights for the actual evapotranspiration data sequence; and fusing each actual evapotranspiration dataset within the cluster according to the actual evapotranspiration data sequence and the dynamic weights to obtain the fusion result within the cluster.

[0009] The multi-source evapotranspiration data fusion method provided by the present invention further includes: performing an accuracy evaluation on the fusion result of the multi-source actual evapotranspiration data.

[0010] According to the multi-source evapotranspiration data fusion method provided by the present invention, the accuracy evaluation indicators include correlation coefficient, mean absolute error, and root mean square error.

[0011] This invention also provides a multi-source evapotranspiration data fusion system, comprising: a similarity calculation module for calculating the similarity between multi-source actual evapotranspiration datasets to obtain similarity calculation results; a clustering module for clustering the multi-source actual evapotranspiration datasets based on the similarity calculation results and according to data source independence to obtain clustering results; an intra-cluster fusion module for fusing the actual evapotranspiration datasets within each cluster based on the clustering results and according to dynamic weights to obtain intra-cluster fusion results; wherein the dynamic weights are adjusted for different vegetation types and the years covered by the multi-source actual evapotranspiration datasets; and an inter-cluster fusion module for fusing the actual evapotranspiration datasets of each cluster based on the intra-cluster fusion results to obtain multi-source actual evapotranspiration data fusion results.

[0012] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the multi-source evapotranspiration data fusion method as described above.

[0013] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the multi-source evapotranspiration data fusion method as described above.

[0014] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the multi-source evapotranspiration data fusion method as described above.

[0015] This invention provides a multi-source evapotranspiration data fusion method based on data source independence and dynamic weighting. This invention considers the dynamic weighting scheme of different actual evapotranspiration datasets, adjusts the weighting for different vegetation types and years where the coverage of actual evapotranspiration datasets does not overlap, and considers the independence between multi-source actual evapotranspiration datasets, thereby improving the estimation accuracy of actual evapotranspiration by the fused dataset. Attached Figure Description

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

[0017] Figure 1 This is a flowchart illustrating a multi-source evapotranspiration data fusion method provided by the present invention.

[0018] Figure 2This is a schematic diagram showing the coverage years of each actual evapotranspiration dataset provided by this invention.

[0019] Figure 3 This is a flowchart of the BMA fusion solution provided by the present invention.

[0020] Figure 4 This is a diagram illustrating the principle of weight determination for each actual evapotranspiration dataset in years without common coverage.

[0021] Figure 5 This is a spatial distribution map of the accuracy assessment of the BMA-ET dataset (subplots a, b, and c show the spatial distribution of the correlation coefficient, mean absolute error, and root mean square error of BMA-ET and FLUXNET2015 flux site ET from 1991 to 2011, respectively).

[0022] Figure 6 This is a graph of the accuracy evaluation of the BMA-ET dataset based on external datasets (observation data are from subgraphs (a) FLUXNET2015 (2012–2015), (b) AmeriFlux (1994–2020), (c) ChinaFlux (2003–2010), and (d) ICOS (2003–2010)).

[0023] Figure 7 This is a schematic diagram of the structure of a multi-source evapotranspiration data fusion system provided by the present invention.

[0024] Figure 8 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0026] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating a multi-source evapotranspiration data fusion method provided by the present invention.

[0027] Please refer to Figure 2 , Figure 2 This is a schematic diagram showing the coverage years of the various actual evapotranspiration datasets provided by this invention.

[0028] The present invention provides a multi-source evapotranspiration data fusion method based on data source independence and dynamic weights. It is based on a dynamic weight scheme and considers the independence between actual evapotranspiration datasets, thereby achieving the fusion of multi-source actual evapotranspiration datasets.

[0029] This invention uses 30 sets of actual evapotranspiration datasets as examples to illustrate the implementation scheme for actual evapotranspiration fusion. PML (Penman-Monteith-Leuning) is an evapotranspiration dataset based on the Penman-Monteith-Leuning model; GLEAM (Global Land Evaporation Amsterdam Model) is the Amsterdam global land evapotranspiration model dataset; GLASS (Global Land Surface Satellite Evapotranspiration) is a global land surface evapotranspiration product based on satellite remote sensing; PLSH (Process-based Land Surface Evapotranspiration / Heat Fluxes) is a dataset of evapotranspiration and heat flux based on land surface processes; FLUXCOM (FLUXNET-based Empirical Upscaling of Terrestrial Carbon and Water Fluxes) is a global carbon, water, and energy flux dataset based on the FLUXNET flux observation network; MTE (Model Tree Ensemble) is a global evapotranspiration dataset based on the model tree ensemble method; ERA5-Land (ECMWF Reanalysis v5-Land) is a fifth-generation land surface reanalysis dataset; MERRA-Land (Modern Era Retrospective-Analysis for Research and...) Application-Land is a modern retrospective analysis and research application-land surface dataset; GLDAS (NASA Global Land Data Assimilation System) is a global land data assimilation system, with GLDAS-CLSM, GLDAS-NOAH, and GLDAS-VIC estimating evapotranspiration based on the Catchment, NOAH, and VIC variable infiltration capacity models, respectively; TRENDY (Trends in net landatmosphere carbon exchanges) is a multi-model ensemble dataset of land-atmosphere net carbon exchange trends, where CABLE-POP is a land surface process model dataset integrating a population dynamics module, CLASSIC is a land surface model dataset developed by the Centre for Large Climate Simulation, and CLM5.The datasets listed are: 0 (based on the NCAR land surface processes model), DLEM (based on the dynamic land ecosystem model), E3SM (based on the high-resolution Earth system model), EDv3 (based on the third edition of the Ecosystem Succession Model), IBIS (based on the Integrated Biosphere Simulator), ISBA-CTRIP (based on the coupled land surface processes and river transport modules), JSBACH (based on land surface biogeochemical processes), LPJ-GUESS (based on the Lund-Potsdam-Jena Universal Ecosystem Simulator), LPJmL (based on the Lund-Potsdam-Jena Managed Land Model), LPX-Bern (based on the Bern Land Precipitation and Exchange Model), OCN (a model dataset focused on simulating the global carbon cycle), ORCHIDEE (based on the carbon and hydrological organization model in dynamic ecosystems), SDGVM (based on the Sheffield Dynamic Global Vegetation Model), VISIT (based on the integrated vegetation and trace gas simulator), and YIBs (based on the Yale Interactive Terrestrial Biosphere Model).

[0030] This invention provides a method for fusing multi-source evapotranspiration data, comprising: 101: Similarity calculation is performed between multi-source actual evapotranspiration datasets to obtain similarity calculation results.

[0031] As a preferred embodiment, similarity calculation is performed between multi-source actual evapotranspiration datasets to obtain similarity calculation results, including: determining the time series of the multi-source actual evapotranspiration datasets; calculating the trend term of the multi-source actual evapotranspiration datasets to obtain the trend term; calculating the seasonality term based on the time series and the trend term to obtain the seasonality term; calculating the residual term based on the time series, the trend term, and the seasonality term, and using the obtained residual term as the similarity calculation result.

[0032] In this embodiment, the time series ET(t) for each evapotranspiration dataset satisfies: , in, Indicates the time step; For trend items, It is a seasonal item. This is the residual term.

[0033] Trend Item Calculated using the central moving average: , in, j For index variables (loop variables) For the first j Evapotranspiration data for each locationh It is the window half-width; for monthly data, it is usually taken as... .

[0034] For the start and end of a time series h The formula for calculating the number of points is: , , The calculation process for the seasonality term is as follows: Based on the time series and the trend term, first calculate the detrended series: , For monthly data, calculate the average anomaly for each month: , in, Indicates the month. It is the number of years. It is the first y Year m Monthly time index.

[0035] Adjust the seasonality factor to 0: , Based on the trend and seasonality terms, the time series fit values ​​for each evapotranspiration dataset are: , in, It is time t The corresponding month.

[0036] Calculate the residuals of evapotranspiration values ​​for each actual evapotranspiration dataset: .

[0037] 102: Based on the similarity calculation results, cluster the multi-source actual evapotranspiration dataset according to the data source independence to obtain the clustering results.

[0038] As a preferred embodiment, based on the similarity calculation results, the multi-source actual evapotranspiration dataset is clustered according to the data source independence to obtain the clustering results. This includes: treating each data point in the multi-source actual evapotranspiration dataset as an independent cluster based on the residual term; calculating the sum of squared distances from each data point in the cluster to the cluster centroid for each data point in the cluster, and merging the clusters based on the sum of squared distances to minimize the variance of the merged clusters; iteratively executing the steps of calculating the sum of squared distances and merging the clusters until clustering is completed for the multi-source actual evapotranspiration datasets under all land cover types, thus obtaining the clustering results.

[0039] In this embodiment, 30 sets of actual evapotranspiration datasets are hierarchically clustered according to vegetation type. The Ward method merges the two clusters with the smallest increase in intra-cluster variance at each step. Partitioning is performed according to land cover type (in this embodiment, the land cover data product MCD12Q1 version 6.1 is selected). Taking deciduous broad-leaved forest as an example, the operation steps are as follows: First, based on the residual term, each data point in the multi-source actual evapotranspiration dataset is treated as an independent cluster: , in, For the first i The residual sequence of a set of actual evapotranspiration datasets.

[0040] Define cluster Center of mass: , in, For the first i The residual sequence of the actual evapotranspiration dataset, Cluster The number of data points in the data. For the first The set of all data points in a cluster.

[0041] Cluster For each data point in the cluster, calculate the squared distance from it to the cluster centroid. The formula for calculating the sum of squares within the cluster (sum of squared errors) is as follows: , Increase after merging clusters A and B: , in, Let A be the centroid of cluster A. Let B be the centroid of cluster B. This represents the number of data points in cluster A. This indicates the number of data points in cluster B.

[0042] The merging method is determined based on the variance within each cluster, with the goal of minimizing the increase in variance with each merge. After merging two clusters, the total variance of the new cluster is minimized.

[0043] Repeat the cluster merging process until all data points for the deciduous broadleaf forest are merged into one cluster.

[0044] For other vegetation types, repeat the steps described in this embodiment to obtain the clustering results of the actual evapotranspiration dataset for all land cover types.

[0045] 103: Based on the clustering results, the actual evapotranspiration datasets within each cluster are fused according to the dynamic weights to obtain the fusion results within the clusters; the dynamic weights are adjusted for different vegetation types and the years covered by the multi-source actual evapotranspiration datasets.

[0046] As a preferred embodiment, based on the clustering results, the actual evapotranspiration datasets within each cluster are fused according to dynamic weights to obtain the cluster fusion result, including: determining the actual evapotranspiration data sequence corresponding to the actual evapotranspiration dataset in each cluster based on the site observation dataset; performing Bayesian model averaging analysis on the actual evapotranspiration data sequence using the site observation dataset as the observation value to obtain the dynamic weights for the actual evapotranspiration data sequence; and fusing the actual evapotranspiration datasets within each cluster according to the actual evapotranspiration data sequence and the dynamic weights to obtain the cluster fusion result.

[0047] Please refer to Figure 3 , Figure 3 The flowchart of the BMA fusion scheme provided by the present invention is shown.

[0048] In this embodiment, based on land cover type zoning, taking deciduous broad-leaved forest as an example, 60% of the deciduous broad-leaved forest (DBF) sites are selected to participate in BMA (Bayesian Model Averaging) fusion, and the remaining 40% of the deciduous broad-leaved forest sites are used for validation.

[0049] Since the coverage period of flux stations is generally short, this invention splices together the time series of all stations in the deciduous broad-leaved forest to obtain a station observation dataset with a longer time span. ).

[0050] For each actual evapotranspiration dataset in cluster 1 (there are a total of 9 actual evapotranspiration datasets in cluster 1), the time series of the actual evapotranspiration estimates for the locations of the aforementioned sites are concatenated. This step is performed on all actual evapotranspiration datasets to obtain 9 datasets with... Time series of the same length .

[0051] Will As observed values, for 9 actual evapotranspiration data sequences BMA analysis was performed to obtain the weights of each actual evapotranspiration dataset under the deciduous broad-leaved forest vegetation type, denoted as . .in, This weight is applied to all deciduous broad-leaved forest grid points in each actual evapotranspiration dataset.

[0052] By fusing the actual evapotranspiration datasets from various sources, we obtain the fused evapotranspiration data for cluster 1 within the deciduous broadleaf forest type: .

[0053] For other clusters, repeat the steps described above in this implementation to obtain the actual evapotranspiration fusion dataset under all clusters in the deciduous broad-leaved forest vegetation.

[0054] The remaining 40% of deciduous broad-leaved forest sites were used to validate and evaluate the actual evapotranspiration fusion dataset under deciduous broad-leaved forest types. The accuracy.

[0055] For other vegetation types, repeat the above steps to obtain the actual evapotranspiration fusion dataset for all land cover types.

[0056] Spatially stitch together the actual evapotranspiration fusion datasets under each vegetation type to obtain the global actual evapotranspiration fusion dataset (internal fusion result of clustering).

[0057] Please refer to Figure 4 , Figure 4 A schematic diagram illustrating the principle of determining the weights of each actual evapotranspiration dataset in years without common coverage.

[0058] BMA analysis was conducted from 1991 to 2011 (i.e., the period covered by both site data and 30 actual evapotranspiration datasets). Through the steps described above, the weights of each actual evapotranspiration dataset under each vegetation type were obtained, and these weights were applied to all periods from 1982 to 2011. For the years with shared coverage of the actual evapotranspiration datasets (1982–2011), the weights were calculated following the steps described above. For the non-shared coverage years of 1980–1981 and 2012–2020, the weights for each year were obtained by performing BMA analysis on all actual evapotranspiration datasets covering that year. For example, there were 26 actual evapotranspiration datasets covering 1980; the weights obtained from BMA analysis on the data from these 26 datasets from 1980 to 2011 were used as the weights for these 26 datasets in 1980.

[0059] This invention utilizes a BMA method based on the Markov Chain Monte Carlo (MCMC) algorithm to calculate the weights of various real-world evapotranspiration datasets, providing a data foundation for the fusion research of real-world evapotranspiration datasets. To explain the BMA method, let... Indicates from K Prediction sets obtained from different datasets It is the target quantity. In the BMA method, each set member predicts... Its conditional probability density function Related, whenf k When it is the best prediction in the set, This can be interpreted as Given f k The conditional probability density function at time t. The BMA prediction model can be expressed as: , In the formula, Indicates prediction k The posterior probability for the best prediction is represented by the weights of each dataset. Maximum likelihood estimation can be used to estimate the posterior probability of the dataset. Solve the problem. Weights greater than 0 and summing to 1 are considered to reflect the relative contribution of a single data point to predictive ability during training.

[0060] 104: Based on the fusion results within clusters, the actual evapotranspiration datasets of each cluster are fused to obtain the fusion results of multi-source actual evapotranspiration data.

[0061] In this embodiment, based on the fusion results within clusters, the BMA fusion method is used to fuse the actual evapotranspiration datasets of each cluster, thereby obtaining the final global actual evapotranspiration fusion dataset BMA-ET (multi-source actual evapotranspiration data fusion result).

[0062] As a preferred embodiment, it also includes: performing an accuracy assessment on the fusion results of multi-source actual evapotranspiration data.

[0063] As a preferred embodiment, the metrics for accuracy evaluation include correlation coefficient, mean absolute error, and root mean square error.

[0064] In this embodiment, the following accuracy evaluation indicators are mainly used when evaluating the accuracy of the fusion results of multi-source actual evapotranspiration data: Correlation coefficient: , Mean absolute error: , Root mean square error: , in, For the sample size, For the first i Estimated values ​​based on actual evapotranspiration datasets. These are actual evapotranspiration observations from various stations. and They are respectively and The average value.

[0065] Please refer to Figure 5 , Figure 5 Spatial distribution map for the accuracy assessment of the BMA-ET dataset (subplots a, b, and c show the spatial distribution of the correlation coefficient, mean absolute error, and root mean square error of BMA-ET and FLUXNET2015 flux site ET from 1991 to 2011, respectively).

[0066] This embodiment uses the FLUXNET2015 dataset (1991–2011) as a reference to verify the accuracy of the fused dataset BMA-ET. The verification results show that BMA-ET and FLUXNET2015 have a correlation coefficient greater than 0.6 at over 70% of the flux stations, and the stations with high correlation coefficients are mainly distributed in the mid-to-high latitude regions of the Northern Hemisphere. 72% of the stations have a mean absolute error of less than 30 mm, while nearly half of the stations have a root mean square error of less than 30 mm.

[0067] Please refer to Figure 6 , Figure 6 The image shows the accuracy evaluation of the BMA-ET dataset based on external datasets (the observation data are from subplots (a) FLUXNET2015 (2012–2015), (b) AmeriFlux (1994–2020), (c) ChinaFlux (2003–2010), and (d) ICOS (2003–2010)).

[0068] This embodiment further uses independent data sources to validate the fused dataset BMA-ET, specifically including AmeriFlux, ChinaFlux, and ICOS flux sites. In addition, this embodiment uses FLUXNET2015 data from 2012–2015 to evaluate the accuracy of the BMA-ET dataset. Using FLUXNET2015 as the reference observation data, the correlation coefficient between BMA-ET and FLUXNET2015 site evapotranspiration from 2012–2015 is 0.58. The BMA-ET dataset performs better than other external datasets, with correlation coefficients of 0.61, 0.72, and 0.74 between BMA-ET and site evapotranspiration observations from AmeriFlux, ChinaFlux, and ICOS, respectively.

[0069] The multi-source evapotranspiration data fusion system provided by the present invention is described below. The multi-source evapotranspiration data fusion system described below can be referred to in correspondence with the multi-source evapotranspiration data fusion method described above.

[0070] Please refer to Figure 7 , Figure 7 This is a schematic diagram of the structure of a multi-source evapotranspiration data fusion system provided by the present invention.

[0071] This invention also provides a multi-source evapotranspiration data fusion system, comprising: a similarity calculation module 701, used to calculate the similarity between multi-source actual evapotranspiration datasets to obtain similarity calculation results; a clustering module 702, used to cluster the multi-source actual evapotranspiration datasets according to the data source independence based on the similarity calculation results to obtain clustering results; an intra-cluster fusion module 703, used to fuse the actual evapotranspiration datasets within each cluster according to dynamic weights based on the clustering results to obtain intra-cluster fusion results; the dynamic weights are adjusted for different vegetation types and the years covered by the multi-source actual evapotranspiration datasets; and an inter-cluster fusion module 704, used to fuse the actual evapotranspiration datasets of each cluster based on the intra-cluster fusion results to obtain multi-source actual evapotranspiration data fusion results.

[0072] Figure 8 An example is a schematic diagram of the structure of an electronic device, such as... Figure 8 As shown, the electronic device may include a processor 801, a communications interface 802, a memory 803, and a communication bus 804. The processor 801, communications interface 802, and memory 803 communicate with each other via the communication bus 804. The processor 801 can call logical instructions in the memory 803 to execute a multi-source evapotranspiration data fusion method. This method includes: calculating the similarity between multi-source actual evapotranspiration datasets to obtain similarity calculation results; clustering the multi-source actual evapotranspiration datasets according to data source independence based on the similarity calculation results to obtain clustering results; fusing the actual evapotranspiration datasets within each cluster according to dynamic weights to obtain intra-cluster fusion results; the dynamic weights are adjusted for different vegetation types and the years covering the multi-source actual evapotranspiration datasets; and fusing the actual evapotranspiration datasets of each cluster based on the intra-cluster fusion results to obtain a multi-source actual evapotranspiration data fusion result.

[0073] Furthermore, the logical instructions in the aforementioned memory 803 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0074] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the multi-source evapotranspiration data fusion method provided by the above methods. The method includes: calculating the similarity between multi-source actual evapotranspiration datasets to obtain similarity calculation results; clustering the multi-source actual evapotranspiration datasets according to data source independence based on the similarity calculation results to obtain clustering results; fusing the actual evapotranspiration datasets within each cluster according to dynamic weights based on the clustering results to obtain intra-cluster fusion results; the dynamic weights are adjusted for different vegetation types and the years covered by the multi-source actual evapotranspiration datasets; and fusing the actual evapotranspiration datasets of each cluster based on the intra-cluster fusion results to obtain multi-source actual evapotranspiration data fusion results.

[0075] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the multi-source evapotranspiration data fusion method provided by the above methods. This method includes: calculating similarity among multi-source actual evapotranspiration datasets to obtain similarity calculation results; clustering the multi-source actual evapotranspiration datasets according to data source independence based on the similarity calculation results to obtain clustering results; fusing the actual evapotranspiration datasets within each cluster according to dynamic weights based on the clustering results to obtain intra-cluster fusion results; wherein the dynamic weights are adjusted for different vegetation types and the years covering the multi-source actual evapotranspiration datasets; and fusing the actual evapotranspiration datasets of each cluster based on the intra-cluster fusion results to obtain multi-source actual evapotranspiration data fusion results.

[0076] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0077] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0078] 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 the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for fusing multi-source evapotranspiration data, characterized in that, include: Similarity calculations were performed on the data sets of actual evapotranspiration from multiple sources to obtain the similarity calculation results; Based on the similarity calculation results, the multi-source actual evapotranspiration dataset is clustered according to the data source independence to obtain the clustering results; Based on the clustering results, the actual evapotranspiration datasets within each cluster are fused according to dynamic weights to obtain the cluster fusion results; The dynamic weights are adjusted for different vegetation types and the years covered by the multi-source actual evapotranspiration dataset; Based on the fusion results within the clusters, the actual evapotranspiration datasets of each cluster are fused to obtain the fusion results of multi-source actual evapotranspiration data.

2. The multi-source evapotranspiration data fusion method according to claim 1, characterized in that, The similarity calculation among the multi-source actual evapotranspiration datasets, to obtain the similarity calculation results, includes: Determine the time series of the multi-source actual evapotranspiration dataset; The trend term is calculated by performing trend term calculation on the multi-source actual evapotranspiration dataset; The seasonality term is calculated based on the time series and the trend term to obtain the seasonality term. The residual term is calculated based on the time series, the trend term, and the seasonality term, and the resulting residual term is used as the similarity calculation result.

3. The multi-source evapotranspiration data fusion method according to claim 2, characterized in that, Based on the similarity calculation results, the multi-source actual evapotranspiration dataset is clustered according to data source independence to obtain clustering results, including: Based on the residual term, each data point in the multi-source actual evapotranspiration dataset is treated as an independent cluster; For each data point in a cluster, calculate the sum of squared distances from the data point to the cluster centroid, and merge clusters based on the results of the sum of squared distances to minimize the variance of the merged clusters; The steps of calculating the sum of squared distances and merging clusters are performed iteratively until clustering is completed for all multi-source actual evapotranspiration datasets under all land cover types, and the clustering results are obtained.

4. The multi-source evapotranspiration data fusion method according to claim 3, characterized in that, Based on the clustering results, the actual evapotranspiration datasets within each cluster are fused according to dynamic weights to obtain the cluster fusion results, including: Based on the site observation dataset, determine the actual evapotranspiration data sequence corresponding to the actual evapotranspiration dataset in each cluster; Using the site observation dataset as the observation value, a Bayesian model averaging analysis is performed on the actual evapotranspiration data sequence to obtain the dynamic weights for the actual evapotranspiration data sequence. Based on the actual evapotranspiration data sequence and the dynamic weights, the actual evapotranspiration datasets within the cluster are fused to obtain the fusion result within the cluster.

5. The multi-source evapotranspiration data fusion method according to any one of claims 1 to 4, characterized in that, Also includes: The accuracy of the fusion results of the multi-source actual evapotranspiration data is evaluated.

6. The multi-source evapotranspiration data fusion method according to claim 5, characterized in that, The accuracy assessment metrics include correlation coefficient, mean absolute error, and root mean square error.

7. A multi-source evapotranspiration data fusion system, characterized in that, include: The similarity calculation module is used to calculate the similarity between multi-source actual evapotranspiration datasets and obtain the similarity calculation results; The clustering module is used to cluster the multi-source actual evapotranspiration dataset based on the similarity calculation results and according to the data source independence to obtain the clustering results. The intra-cluster fusion module is used to fuse the actual evapotranspiration datasets within each cluster based on the clustering results and according to dynamic weights, to obtain the intra-cluster fusion result; The dynamic weights are adjusted for different vegetation types and the years covered by the multi-source actual evapotranspiration dataset; The inter-cluster fusion module is used to fuse the actual evapotranspiration datasets of each cluster based on the fusion results within the clusters, so as to obtain the fusion results of multi-source actual evapotranspiration data.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the multi-source evapotranspiration data fusion method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the multi-source evapotranspiration data fusion method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the multi-source evapotranspiration data fusion method as described in any one of claims 1 to 6.

Images