Rain-gauge-free regional climate model precipitation bias correction method and device

Abstract

The present application relates to the technical field of climate change impact assessment, and discloses a rain-gauge-free regional climate model precipitation deviation correction method and device, which comprises the following steps: obtaining basic data of a rain-gauge-free target region and a rain-gauge-free reference region, constructing a regional similarity feature vector and dividing a similar climate subregion; determining a candidate reference region and a parameter migration weight of the target region according to the site coverage of the similar climate subregion and the comprehensive similarity; constructing a deviation correction parameter library in the reference region, and calculating a reliability score of the reference region; migrating the deviation correction parameters of the reference region to the target region, correcting the daily precipitation data of the historical period and the future period of the target region climate model, and verifying and optimizing the correction result. The present application can accurately correct the precipitation deviation of the climate model under the condition of rain-gauge-free observation data, and improve the applicability of the future precipitation data of the data-free region in hydrological simulation, urban waterlogging simulation, climate change impact assessment and engineering planning and design.

Description

Technical Field

[0001] This invention relates to the field of climate change impact assessment technology, specifically to a method and apparatus for correcting precipitation bias in regional climate models without rain gauges. Background Technology

[0002] Against the backdrop of climate change, the frequency, intensity, and spatiotemporal distribution of extreme precipitation events may all change. Climate model output data has become a crucial foundation for future precipitation change analysis, hydrological and water resource simulation, urban flooding risk assessment, drainage system adaptability evaluation, and engineering planning and design. However, global or regional climate models are affected by factors such as spatial resolution, physical process parameterization, topographic characterization capabilities, and the complexity of regional climate systems. Therefore, the daily precipitation data output by climate models typically exhibits systematic biases. Consequently, bias correction processing is usually required before applying climate model precipitation data to specific regional hydrological and meteorological analyses and engineering designs.

[0003] Existing methods for correcting precipitation bias in climate models largely rely on surface rain gauge data within the target area. However, in practical applications, some regions lack long-term, continuous surface rain gauge data, making it impossible to directly establish a bias correction relationship between climate model precipitation data and observed surface precipitation in the target area. To address the lack of local observation data in areas without rain gauges, existing techniques typically employ methods such as substituting data from nearby rain gauges, substituting multi-source grid precipitation data, extrapolating regional average correction parameters, or directly using climate model ensemble results. However, due to differences in topographic conditions, climate background, precipitation statistical characteristics, and climate model systematic biases, the bias correction results for precipitation data in areas without rain gauges are not accurate enough, and their stability and interpretability are poor. Summary of the Invention

[0004] This invention provides a method and apparatus for correcting precipitation bias in climate models in areas without rain gauges, in order to solve the problems of inaccurate correction of precipitation data in climate models in areas without rain gauges, as well as poor data stability and interpretability.

[0005] In a first aspect, the present invention provides a method for correcting precipitation bias in regional climate models without rain gauges, the method comprising: Basic data of the reference area with rain gauges and the target area without rain gauges were obtained respectively, and regional similarity feature vectors were constructed based on the basic data; Cluster analysis based on regional similarity feature vectors yields multiple similar climate zones. Rainfall station coverage is checked for each similar climate zone to determine candidate reference areas and parameter migration weights for each target area. Bias correction parameters are constructed, and based on parameter migration weights, the bias correction parameters of the candidate reference area are migrated to the target area to correct the historical precipitation and predicted precipitation of the climate model in the target area, thereby obtaining historical daily precipitation correction data and predicted daily precipitation correction data. Leave-one-out cross-validation is used to optimize the similarity feature weights, and the deviation correction results and uncertainty range of the target region are output.

[0006] The precipitation bias correction method for climate models in areas without rain gauges provided by this invention transfers the bias correction parameters of a reference area with rain gauges to a target area without rain gauges through similar climate zoning and parameter migration. This solves the problem that it is difficult to directly carry out precipitation bias correction for climate models in areas with little or no data. In the process of selecting candidate reference areas, climate similarity, spatial proximity, zoning differences, and the reliability of the reference area are considered simultaneously, which improves the rationality of parameter migration. Based on the bias correction parameters, the method accurately corrects the historical daily precipitation data and predicted daily precipitation data of the target area. Through leave-one-out cross-validation and uncertainty range output, the stability, interpretability, and engineering applicability of the correction results in areas without rain gauges are improved.

[0007] In one optional implementation, the basic data for the reference area includes: historical daily precipitation data, historical daily precipitation data for the first climate model, future daily precipitation data for the first climate model, reference geographic location data, reference topographic data, reference land use or underlying surface data, and reference climate background data. The basic data for the target area includes: historical daily precipitation data for the second climate model, future daily precipitation data for the second climate model, target geographic location data, target topographic data, target land use or underlying surface data, and target climate background data. Based on the basic data, a region similarity feature vector is constructed, including: The basic data is preprocessed to obtain a unified data source. The preprocessing includes: coordinate unification, time scale unification, spatial interpolation of climate model grid data, calendar format unification, missing value imputation, outlier removal, and precipitation unit unification. Geographic features, topographic features, underlying surface features, and precipitation statistics are extracted based on a unified data source; Each feature is standardized to construct a region similarity feature vector.

[0008] The precipitation bias correction method for regional climate models without rain gauges provided by this invention uses a unified basic data source that can be obtained in both the reference area and the target area to calculate similar feature vectors, ensuring the consistency of similarity comparison between the reference area with rain gauges and the target area without rain gauges. The method also standardizes each feature variable to eliminate differences in the dimensions of different feature variables, making comparison easier.

[0009] In one optional implementation, a rainfall station coverage check is performed on each similar climate zone to determine candidate reference areas and parameter migration weights for each target area, including: The number of reference areas with rain gauges in each similar climate zone is counted. If the number of reference areas is not less than the preset number, then the preset number of candidate reference areas are selected in the similar climate zone to which the target area belongs, based on the comprehensive similarity. If the number of reference areas is less than the preset number, all the reference areas with rain gauges in the similar climate zones to which the target area belongs will be used as candidate reference areas. Based on the comprehensive similarity, candidate reference areas will be selected from other similar climate zones to obtain the preset number of candidate reference areas. Based on the comprehensive similarity between each candidate reference region and the target region, the parameter migration weights of each candidate reference region are determined.

[0010] The precipitation bias correction method for climate models in areas without rain gauges provided by this invention prioritizes selecting parameter migration sources from reference areas with rain gauges that belong to the same zone as the target area without rain gauges. When the number of reference rain gauges in the same zone is insufficient or there is no reference area with rain gauges, reference areas are selected from other similar climate zones. Migration weights are allocated based on comprehensive similarity, which ensures that the climate characteristics of the reference area and the target area are highly matched, and that the number of reference samples is sufficient, thereby improving the rationality of parameter migration and the accuracy of correction results.

[0011] In one optional implementation, the parameter migration weights of each candidate reference region are determined based on the comprehensive similarity between each candidate reference region and the target region, including: The comprehensive distance between each candidate reference region and the target region is calculated based on the preset standardized feature variables, and the comprehensive similarity between each candidate reference region and the target region is determined based on the comprehensive distance. The initial parameter migration weights are obtained by normalizing the comprehensive similarity. Based on the historical daily precipitation data of each candidate reference area and the corresponding historical daily precipitation data of the climate model, bias correction parameters are established. The bias correction parameters are applied to the corresponding historical precipitation data of the climate model to obtain corrected historical precipitation data. The corrected historical precipitation data is compared with the corresponding historical precipitation data to obtain the parameter reliability score of the candidate reference area. The initial parameter migration weights are corrected by using the parameter reliability scores of each candidate reference region, resulting in the corrected parameter migration weights.

[0012] The precipitation bias correction method for regional climate models provided by this invention combines feature distance calculation with comprehensive similarity, and obtains initial weights through normalization. At the same time, it evaluates the reliability of reference area parameters based on measured and simulated precipitation data, and then corrects the weights using reliability scores. This method not only ensures regional similarity based on topographic and climatic features, but also selects reference samples with high data quality and good correction effects. This dual optimization of weight allocation reduces the errors caused by inferior parameters, making the migrated correction parameters more consistent with the actual situation of the target area, and significantly improving the reliability and accuracy of precipitation bias correction results.

[0013] In one optional implementation, the bias correction parameters include: wet day frequency, wet day correction threshold, and precipitation intensity correction coefficients at different quantile levels. The bias correction parameters are constructed, and based on parameter migration weights, the bias correction parameters of the candidate reference area are migrated to the target area, including: Based on historical daily precipitation data of the candidate reference area and the corresponding historical daily precipitation data of the climate model, the measured wet day frequency, wet day correction threshold, and precipitation intensity correction coefficients at different quantile levels of the candidate reference area are calculated. Based on the parameter migration weights of each candidate reference region, the measured wet day frequencies of each candidate reference region are weighted and fused to obtain the wet day frequency of the target region. Based on the wet day frequency of the target region and the empirical quantile function of the historical daily precipitation series of the climate model in the target region, the wet day correction threshold of the target region is calculated. Based on the parameter migration weights of each candidate reference region, the precipitation intensity correction coefficients of each candidate reference region at different quantile levels are weighted and fused to obtain the precipitation intensity correction coefficient of the target region.

[0014] The precipitation bias correction method for regional climate models without rain gauges provided by this invention constructs complete correction parameters from three dimensions: wet day frequency, correction threshold, and quantile intensity coefficient. Then, it combines migration weights to fuse parameters from multiple reference regions to obtain target region-specific parameters, simultaneously correcting two types of biases: precipitation frequency and rainfall amount. The parameter division is detailed and highly targeted. Furthermore, the target region threshold is calculated based on its own precipitation characteristics, rather than directly copying reference region values, making the correction rules more consistent with local precipitation patterns and effectively improving the overall correction accuracy and applicability.

[0015] In one optional implementation, historical precipitation data from the climate model of the target area are corrected to obtain corrected historical daily precipitation data, including: Compare the historical daily precipitation data of the target area climate model. If the historical daily precipitation data is less than the wet day correction threshold of the target area, then the historical daily precipitation data of the corresponding period will be corrected to rainless days. If the historical daily precipitation is not less than the wet day correction threshold of the target area, then calculate the quantile level of the historical daily precipitation in the historical wet day distribution of the climate model of the target area, and calculate the precipitation intensity correction coefficient for the corresponding period based on the quantile level. The daily precipitation for the corresponding period is corrected using the precipitation intensity correction factor to obtain historical daily precipitation correction data.

[0016] The precipitation bias correction method for regional climate models without rain gauges provided by this invention accurately distinguishes between periods of rainfall and periods without rainfall by using wet day correction thresholds. Then, it combines precipitation distribution quantile matching with corresponding correction coefficients to complete the historical precipitation correction in layers. This method can effectively correct the precipitation frequency bias simulated by the climate model and correct the precipitation intensity bias for different rainfall levels. The correction logic fits the precipitation distribution characteristics, and the step-by-step processing makes the error correction more refined, significantly improving the accuracy of historical precipitation correction data.

[0017] In one optional implementation, the predicted precipitation from the climate model for the target area is corrected to obtain corrected daily precipitation data, including: The wet day frequency of the target area in the future period is calculated based on the original wet day frequency of the climate model in the future period and the original wet day frequency of the climate model in the historical period. The target wet day frequency is obtained by using a preset wet day frequency range as a constraint. Based on the target wet day frequency and the empirical quantile function of the future daily precipitation sequence of the climate model in the target area, the future wet day correction threshold of the target area is determined, and the predicted daily precipitation of the climate model in the target area is compared with the future wet day correction threshold of the target area. If the predicted daily precipitation is less than the future wet day correction threshold for the target area, the predicted daily precipitation for the corresponding period will be corrected to a rainless day. If the predicted daily precipitation is not less than the future wet day correction threshold for the target area, then the quantile level of the predicted daily precipitation in the future wet day precipitation distribution of the climate model in the target area is calculated, and the corresponding precipitation intensity correction coefficient for the target area is called according to the quantile level to correct the predicted daily precipitation, thus obtaining the corrected predicted daily precipitation data; or, the quantile change factor of the climate model in the target area for future periods relative to historical periods at the corresponding quantile level is calculated, and the predicted daily precipitation is corrected by combining the quantile change factor and the precipitation intensity correction coefficient for the target area, thus obtaining the corrected predicted daily precipitation data.

[0018] The precipitation bias correction method for regional climate models without rain gauges provided by this invention combines historical and future original wet day frequencies to extrapolate future rainfall frequencies and sets numerical range constraints to avoid extreme anomalies in the results. It divides rainy days and rainless days by thresholds and then corrects rainfall by combining quantile variation factors and correction coefficients. This method not only preserves the precipitation evolution trend predicted by the climate model, but also corrects frequency and intensity biases in a stratified manner, making the predicted precipitation correction results more consistent with objective laws and improving the rationality and accuracy of future precipitation data.

[0019] Secondly, the present invention provides a precipitation bias correction device for regional climate models without rain gauges, the device comprising: The similarity feature vector construction module is used to obtain basic data of the reference area with rain gauges and the target area without rain gauges, and to construct regional similarity feature vectors based on the basic data. The candidate reference area and parameter migration weight calculation module is used to perform cluster analysis based on regional similarity feature vectors to obtain multiple similar climate zones, perform rain gauge station coverage checks on each similar climate zone, and determine the candidate reference area and parameter migration weight for each target area. The parameter migration and target area data correction module is used to construct bias correction parameters and, based on the parameter migration weights, migrate the bias correction parameters of the candidate reference area to the target area, correct the historical precipitation and predicted precipitation of the climate model in the target area, and obtain historical daily precipitation correction data and predicted daily precipitation correction data. The result verification module is used to optimize the similarity feature weights using leave-one-out cross-validation and output the deviation correction results and uncertainty range of the target region.

[0020] Thirdly, the present invention provides an electronic device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the method described in the first aspect or any corresponding embodiment thereof.

[0021] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to perform the method described in the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0022] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of an application scenario according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the first step in the precipitation bias correction method for regional climate models without rain gauges according to an embodiment of the present invention. Figure 3 This is a schematic diagram of the second process of the precipitation bias correction method for regional climate models without rain gauges according to an embodiment of the present invention. Figure 4 This is a structural block diagram of a precipitation bias correction device for a regional climate model with no rain gauge, according to an embodiment of the present invention. Figure 5 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0026] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0027] As an optional application scenario of this invention, such as Figure 1 As shown, the precipitation bias correction system for this climate model may include at least one terminal device and at least one server. Figure 1 The system is illustrated in the example, which includes a computer 101, a mobile terminal 102, and a server 103, and the terminal devices such as the computer 101 and the mobile terminal 102 are connected to the server 103 through a network 110.

[0028] Specifically, the terminal device can be a smartphone, tablet, laptop, PDA, desktop computer, game console, smart TV, smart wearable device, in-vehicle terminal, VR (Virtual Reality) device, AR (Augmented Reality) device, etc. Server 103 can be a standalone physical server, a server cluster, a distributed system, or a cloud server providing cloud services. Network 110 can be a wired or wireless network, examples of which include, but are not limited to, the Internet, corporate intranet, local area network, wide area network, mobile communication network, and combinations thereof.

[0029] Existing methods for correcting precipitation biases in climate models largely rely on surface rain gauge data within the target area. These methods correct historical and future precipitation data by establishing statistical relationships between historical and measured precipitation data from the climate model. Common methods include linear proportional correction, distribution mapping, quantile mapping, quantile increment mapping, empirical distribution correction, and trend-preserving bias correction. These methods are well-suited for areas with long-term measured rainfall data, but their fundamental premise is the existence of surface rain gauge data within the target area that can be used to establish correction relationships.

[0030] In practical applications, some mountainous watersheds, new urban areas, planned construction areas, remote areas, or areas with scarce data lack long-term continuous ground rain gauge data, making it impossible to directly establish a bias correction relationship between climate model precipitation data and measured ground precipitation in the target area. For such areas without rain gauges, directly using the original climate model precipitation data to analyze future precipitation changes can easily transmit the biases of the climate model itself to subsequent stages such as hydrological simulation, flood risk assessment, urban flooding simulation, drainage system adaptability evaluation, and engineering planning and design, thereby affecting the reliability of the calculation results and the scientific nature of engineering decisions.

[0031] To address the lack of local observational data in areas without rain gauges, existing techniques typically employ methods such as substituting data from nearby rain gauges, substituting data from multi-source grid precipitation, extrapolating regional average correction parameters, or directly using ensemble results from climate models. However, these methods still have certain shortcomings: First, nearby rain gauges may not have similar topographic, climatic, and precipitation statistical characteristics to the target area, and simply using the nearest station may lead to unreasonable parameter migration; second, multi-source grid precipitation, satellite precipitation, or reanalysis precipitation inherently contain errors, and directly replacing ground observational data to establish correction relationships may introduce new data source biases; third, regional average correction parameters ignore the differences in precipitation and climate between different regions, making it difficult to reflect the local characteristics of the target area; fourth, existing methods do not adequately consider the quality of reference area data, sample length, station reliability, and migration uncertainties, resulting in limited stability and interpretability of correction results for areas without rain gauges.

[0032] To address the aforementioned issues, this invention provides a method for correcting precipitation bias in climate models in areas without rain gauges. Without relying on local surface rain gauge data in the target area, it fully utilizes observational data and climate model data from reference areas with rain gauges. Combining geographical, topographical, and climatic background factors with precipitation statistical characteristics, it identifies reference areas similar to the target area without rain gauges. The bias correction parameters of these reference areas are then weighted and transferred according to similarity and reliability. This allows for the reasonable correction of historical and future precipitation data in climate models in areas without rain gauges, providing reliable precipitation input for climate change impact assessments, hydrological model-driven applications, urban flooding simulations, drainage system adaptability analysis, and future engineering planning and design in areas with little or no data.

[0033] According to an embodiment of the present invention, an embodiment of a method for correcting precipitation bias in regional climate models without rain gauges is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0034] This embodiment provides a method for correcting precipitation bias in regional climate models without rain gauges, which can be used in the aforementioned computer system. Figure 2 This is a flowchart of a precipitation bias correction method for regional climate models without rain gauges according to an embodiment of the present invention, as shown below. Figure 2 As shown, the process includes the following steps: Step S201: Obtain basic data for the reference area with rain gauges and the target area without rain gauges, and construct a regional similarity feature vector based on the basic data.

[0035] Specifically, historical daily precipitation data of the reference area with rain gauges, historical daily precipitation data of the climate model, and future daily precipitation data of the climate model are obtained within the study area, as well as historical and future daily precipitation data of the climate model for the target area without rain gauges. At the same time, geographical location data, topographic data, land use or underlying surface data, and climate background data of the reference area and the target area are obtained, and the above basic data are processed to unify coordinates, time scales, spatial interpolation, calendar format, imputation of missing values, removal of outliers, and unification of precipitation units.

[0036] Based on a unified data source available in both the reference and target regions, geographical, topographic, underlying surface, and precipitation statistical features are extracted to construct a regional similarity feature vector. Among these, surface-observed precipitation data from the reference region with rain gauges are primarily used for subsequent bias correction parameter extraction and are not the sole data source for similarity feature calculation. Furthermore, each feature variable is standardized to construct the regional similarity feature vector. The precipitation statistical features used in the similarity feature construction are preferably calculated from the same data source among multi-source fused grid precipitation data, satellite precipitation data, reanalysis precipitation data, or historical precipitation data from climate models, to ensure consistency in similarity comparison between the reference region with rain gauges and the target region without rain gauges.

[0037] Step S202: Perform cluster analysis based on regional similarity feature vectors to obtain multiple similar climate zones. Perform rainfall station coverage checks on each similar climate zone to determine the candidate reference area and parameter migration weights for each target area.

[0038] Specifically, based on the standardized regional similarity feature vector, cluster analysis is performed on the reference area with rain gauges and the target area without rain gauges to divide the region into multiple similar climate zones. Regions within the same similar climate zone are similar in terms of geography, topography, climate background and precipitation statistics.

[0039] The number of reference areas with rain gauges within each similar climate zone is counted. For any target area without rain gauges, reference areas with rain gauges within its own similar climate zone are selected as candidate reference areas. When the number of reference areas with rain gauges within the same similar climate zone is insufficient or there are no reference areas with rain gauges, candidate reference areas are selected from other similar climate zones based on comprehensive similarity. A comprehensive similarity is constructed based on feature distance, spatial distance, and zone differences to determine the candidate reference areas and their parameter migration weights. If there are no reference areas with rain gauges within a similar climate zone, reference areas with rain gauges are selected from other similar climate zones that have the closest climate characteristics to that zone.

[0040] Step S203: Construct bias correction parameters, and based on parameter migration weights, migrate the bias correction parameters of the candidate reference area to the target area, correct the historical precipitation and predicted precipitation of the climate model in the target area, and obtain historical daily precipitation correction data and predicted daily precipitation correction data.

[0041] Specifically, bias correction parameters are established for each reference area of ​​a rain gauge station. For any reference area of ​​a rain gauge station, based on the historical daily precipitation sequence of the reference area and the historical daily precipitation sequence of the corresponding climate model, the frequency of observed wet days, the wet day correction threshold of the climate model, and the precipitation intensity correction coefficient at different quantile levels are calculated by month to form a bias correction parameter library for the reference area of ​​the rain gauge station.

[0042] For any reference area with rain gauges, bias correction parameters are established using all historical data of the reference area. These bias correction parameters are then applied to the historical precipitation data of the corresponding climate model for that reference area to obtain the corrected historical precipitation series. The corrected historical precipitation series is then compared with the historical precipitation series observed by surface rain gauges in the reference area. One or more of the following are calculated: monthly precipitation bias, wet day frequency bias, extreme quantile bias, normalized root mean square error, data integrity index, and sample length index. A reliability score for the reference area is then constructed, and this reliability score is used to correct the parameter migration weights.

[0043] Based on the corrected parameter migration weights, the wet day frequency parameters and quantile intensity correction coefficients of the reference area with rain gauges are migrated to the target area without rain gauges. The wet day correction threshold of the target area without rain gauges is not directly adopted from the reference area threshold, but is determined by back-calculation from the historical daily precipitation series of the target area's own climate model based on the migrated target wet day frequency.

[0044] Using the wet day frequency parameters, wet day correction threshold, and quantile intensity correction coefficient obtained from migration, bias corrections are performed on the historical daily precipitation data of the climate model in the target area of ​​the rainless station. For historical daily precipitation data of the climate model that are below the wet day correction threshold, they are corrected to rainless days. For historical daily precipitation data of the climate model that are not below the wet day correction threshold, the corresponding quantile intensity correction coefficient is applied for correction based on their quantile level in the historical wet day precipitation distribution of the climate model in the target area.

[0045] Using the bias correction parameters obtained from migration, the daily precipitation data of the climate model for the target area of ​​the rainless station are corrected for the future period. In the correction process for the future period, the frequency of wet days in the future is determined based on the wet day frequency change characteristics of the climate model in the future period relative to the historical period, and upper and lower limits are imposed on the frequency of wet days in the future. Furthermore, based on the quantile level of the daily precipitation in the future period in the future wet day precipitation distribution of the climate model, the corresponding migration correction coefficient is called to correct the future precipitation, or the trend-preserving method is used to superimpose the quantile change signal of the climate model in the future relative to the historical period.

[0046] Step S204: Use leave-one-out cross-validation to optimize the similarity feature weights and output the deviation correction results and uncertainty range of the target region.

[0047] Specifically, physical constraints and consistency corrections are applied to the corrected historical and future daily precipitation sequences for the target area of ​​the rainless monitoring station. These physical constraints and consistency corrections include: non-negativity constraints on precipitation and smoothing constraints on quantile correction coefficients. The non-negativity constraint on precipitation is expressed as follows:

[0048] in, Indicates the target area without rain gauges In the month , No. Daily precipitation after correction.

[0049] When the quantile correction coefficients show abrupt changes between adjacent quantile levels, for Perform smoothing:

[0050] in, This represents the quantile correction coefficients after smoothing; Indicates the target area without rain gauges In the month quantile level Quantile correction coefficients obtained from down-migrating; and These represent the quantile correction coefficients corresponding to adjacent lower quantile levels and adjacent higher quantile levels, respectively.

[0051] If the change in quantile correction coefficients before and after smoothing is small, the original quantile correction coefficients can be retained; if the difference between adjacent quantile correction coefficients exceeds the preset abrupt change threshold, the smoothed quantile correction coefficients will be used in subsequent precipitation correction calculations.

[0052] To improve the reliability of parameter migration, leave-one-out cross-validation is used in the reference area of ​​rain gauges to optimize the similarity feature weights and reference area selection parameters.

[0053] Within the set of reference areas with rain gauges, one of these reference areas is temporarily considered as a target area without rain gauges, and the remaining reference areas with rain gauges are used for parameter migration and precipitation correction. For any given reference area with rain gauges... This area is temporarily considered a target area without rain gauges and is removed from the set of reference areas with rain gauges.

[0054] Utilizing the remaining reference area A deviation correction parameter database was established using all historical data, and the region was analyzed. Parameter migration and precipitation correction were performed to obtain the correction results under simulated no-station conditions:

[0055] Then, combine it with the region By comparing with the actual ground-observed precipitation sequences, the leave-one-out verification error was calculated:

[0056] in, Indicates the region Verification error when areas are temporarily considered as areas without rain gauges; , , , These represent the monthly precipitation deviation, wet day frequency deviation, extreme quantile deviation, and normalized root mean square error, respectively. Let the weights be represented, and satisfy the following conditions:

[0057] Repeat the above process for all reference areas with rain gauges to obtain the total verification error:

[0058] With the goal of minimizing the total validation error, the similarity feature weights, the balance coefficient between feature similarity and spatial proximity, and the number of reference regions are optimized:

[0059] By using leave-one-out cross-validation, the parameter migration process in areas without rain gauges can be simulated without dividing the period into rate period and validation period, thereby evaluating and optimizing the applicability of this method in areas without rain gauges.

[0060] Output the corrected historical daily precipitation series of the climate model for the target area with no rain gauges. Future daily precipitation sequence It also outputs the source of parameter migration in the target area, the similarity weight of the reference area, the migration correction parameters, the statistical characteristics of precipitation after correction, and the range of uncertainty.

[0061] The ensemble correction results are generated based on multiple reference regions:

[0062] in, Indicates that only the reference area is used. The correction results obtained from the migration parameters.

[0063] The set mean is calculated as follows:

[0064] in, Indicates the number of reference areas. Indicates the reference area with rain gauges. Towards the target area of ​​no rain gauge Similarity weights when adjusting for migration bias. The set standard deviation is calculated as follows:

[0065] This yields the uncertainty range of the correction results for the target area of ​​the rainless weather station. The uncertainty range can be expressed as the mean, standard deviation, 5th percentile, 95th percentile, or maximum-minimum range.

[0066] The precipitation bias correction method for climate models in areas without rain gauges provided in this embodiment transfers the bias correction parameters of the reference area with rain gauges to the target area without rain gauges through similar climate zoning and parameter migration. This solves the problem that it is difficult to directly carry out precipitation bias correction for climate models in areas with no or little data. In the process of selecting candidate reference areas, climate similarity, spatial proximity, zoning differences, and the reliability of the reference area are considered simultaneously, which improves the rationality of parameter migration. Based on the bias correction parameters, the method accurately corrects the historical daily precipitation data and predicted daily precipitation data of the target area. Through leave-one-out cross-validation and uncertainty range output, the stability, interpretability, and engineering applicability of the correction results in areas without rain gauges are improved.

[0067] This embodiment provides a method for correcting precipitation bias in regional climate models without rain gauges, which can be used in the aforementioned computer system. Figure 3 This is a flowchart of a precipitation bias correction method for regional climate models without rain gauges according to an embodiment of the present invention, as shown below. Figure 3 As shown, the process includes the following steps: Step S301: Obtain basic data for the reference area with rain gauges and the target area without rain gauges, and construct a regional similarity feature vector based on the basic data.

[0068] Specifically, the basic data for the reference area includes: historical daily precipitation data, historical daily precipitation data for the first climate model, future daily precipitation data for the first climate model, reference geographical location data, reference topographic data, reference land use or underlying surface data, and reference climate background data. The basic data for the target area includes: historical daily precipitation data for the second climate model, future daily precipitation data for the second climate model, target geographical location data, target topographic data, target land use or underlying surface data, and target climate background data. Step S301 above includes: Step S3011: Preprocess the basic data to obtain a unified data source. The preprocessing includes: coordinate unification, time scale unification, spatial interpolation of climate model grid data, calendar format unification, missing value imputation, outlier removal, and precipitation unit unification.

[0069] Specifically, the reference area for rain gauges is denoted as: ,in, This indicates a set of reference areas for rain gauges; Indicates the first A reference area or reference rain gauge station; This indicates the number of reference areas. The target area for no-rain gauges is denoted as: ,in, This represents the set of target areas with no rainfall. Indicates the first Target areas, target grids, or target small watersheds for no-rainfall monitoring stations; Indicates the number of target areas.

[0070] The basic data includes geographical location data such as longitude, latitude, distance from the sea, distance from mountains, or distance from river networks; topographic data includes elevation, slope, aspect, and topographic relief; and climate background data includes multi-year average precipitation, proportion of precipitation during the flood season, monthly precipitation distribution, frequency of wet days, consecutive dry days, consecutive wet days, extreme precipitation percentile, or extreme precipitation index. The preprocessing of the basic data includes: coordinate unification, time scale unification, spatial interpolation of climate model grid data, calendar format unification, missing value imputation, outlier removal, and precipitation unit unification. These preprocessing techniques are mature and existing, and will not be elaborated upon here.

[0071] Step S3012: Extract geographical features, topographic features, underlying surface features, and precipitation statistical features based on a unified data source.

[0072] Specifically, similarity feature vectors are constructed for the reference area with rain gauges and the target area without rain gauges, respectively. For any region Its similarity feature vector is represented as:

[0073] in, Indicates the region Similarity feature vectors; Indicates the first One characteristic variable; This indicates the number of feature variables. Similarity feature vectors include:

[0074] in, Indicates longitude; Indicates latitude; Indicates average elevation; Indicates the average slope; This represents the average annual precipitation over many years. This indicates the percentage of rainfall during the flood season; Indicates the frequency of wet days; and These represent the 90th and 95th percentile values ​​of daily precipitation, respectively. Indicates the maximum number of consecutive dry days; This indicates the maximum number of consecutive wet days.

[0075] To ensure consistency in similarity comparisons between reference areas with rain gauges and target areas without rain gauges, similarity features are preferably calculated using a unified data source available in both types of areas. Specifically, geographic features are calculated from spatial location data, topographic features from digital elevation models, land use or underlying surface features from unified land use data, and precipitation statistical features are preferably calculated from the same data source among multi-source fused grid precipitation data, satellite precipitation data, reanalysis precipitation data, or historical precipitation data from climate models. For reference areas with rain gauges, historical observation data from surface rain gauges are primarily used for subsequent bias correction parameter extraction, and not as the sole data source for similarity feature calculations.

[0076] Step S3013: Standardize each feature and construct a region similarity feature vector.

[0077] Specifically, the sample mean and standard deviation of each characteristic variable are calculated using a set of reference areas with rain gauges, and the same set of means and standard deviations are used to standardize both the reference areas with rain gauges and the target areas without rain gauges.

[0078] No. One characteristic variable The sample mean and standard deviation are calculated as follows:

[0079]

[0080] The standardized feature variables are calculated as follows:

[0081] in, and They represent the first and second halves of the calculation based on the reference area set of rain gauges. The mean and standard deviation of each characteristic variable; This indicates a reference area with any rain gauge station or a target area without a rain gauge station. This represents the smallest positive number that prevents the denominator from being zero. The standardized eigenvector is represented as: .

[0082] The precipitation bias correction method for climate models in areas without rain gauges provided in this embodiment uses a unified basic data source that can be obtained in both the reference area and the target area to calculate similar feature vectors, ensuring the consistency of similarity comparison between the reference area with rain gauges and the target area without rain gauges. The method also standardizes each feature variable to eliminate differences in the dimensions of different feature variables, making comparison easier.

[0083] Step S302: Perform cluster analysis based on regional similarity feature vectors to obtain multiple similar climate zones. Perform rainfall station coverage checks on each similar climate zone to determine the candidate reference area and parameter migration weights for each target area.

[0084] Specifically, step S302 includes: Step S3021: Count the number of reference areas with rain gauges in each similar climate zone. If the number of reference areas is not less than the preset number, select the preset number of candidate reference areas in the similar climate zone to which the target area belongs, based on the comprehensive similarity.

[0085] Specifically, a clustering method is used to divide the reference area with rain gauges and the target area without rain gauges into several similar climate zones:

[0086] in, Represents a set of similar climate zones; Indicates the first Several similar climate zones; This indicates the number of similar climate zones.

[0087] Clustering methods can include K-means clustering, hierarchical clustering, spectral clustering, self-organizing maps, or geographically constrained clustering. The clustering objective function is:

[0088] in, Represent the clustering objective function; Indicates the first Characteristic centers of similar climate zones; Indicates that it belongs to a partition The region object; Indicates the region The squared Euclidean distance between the standardized eigenvector and the center of its similar climate zone.

[0089] By using similar climate zones, parameter migration sources can be preferentially selected from reference areas with rain gauges that belong to the same zone as the target area without rain gauges. The number of reference areas with rain gauges within each similar climate zone is counted, and candidate reference areas for the target area without rain gauges are determined based on station coverage.

[0090] For the Similar climate zones The set of reference areas for rain gauges within it is represented as follows:

[0091] The number of reference areas with rain gauges within it is:

[0092] in, Indicates the first Number of rain gauge reference areas available for parameter migration within similar climate zones.

[0093] For any target area without rain gauge Let its similar climate zone be . The preset number of reference areas is .according to and The relationship is used to determine the candidate reference region. At that time, only in the target area Belonging to similar climate zones A reference area with rain gauges is selected within the area. Specifically, the target area is calculated. With partitions The overall similarity between the reference areas of each rain gauge is calculated, and the top-ranked stations are selected according to their similarity from highest to lowest. One reference region is selected as a candidate reference region.

[0094] Step S3022: If the number of reference areas is less than the preset number, all the reference areas with rain gauges in the similar climate zones to which the target area belongs are taken as candidate reference areas, and candidate reference areas are selected from other similar climate zones according to the comprehensive similarity to obtain the preset number of candidate reference areas.

[0095] Specifically, when At that time, partition All rain gauge reference areas within the region were included as candidate reference areas. Reference areas were then selected from other similar climate zones based on overall similarity, ranked from highest to lowest, until the number of candidate reference areas reached a certain threshold. .

[0096] when When, it indicates the target area. There is no reference area with rain gauges within the same similar climate zone. Therefore, the target zone is calculated. Other similar climate zones Distance between partition centers:

[0097] in, Indicating similar climate zones and Distance between partition centers; Indicates partition Standardized feature centers; Indicates partition Standardized feature centers. According to Determine the target partition from smallest to largest The most similar adjacent climate zones were identified, and candidate reference regions were selected from the rain gauge reference areas within these similar climate zones. These regions were then sorted by overall similarity from highest to lowest, and the top-ranked regions were selected. The target area is a reference area with rain gauges. Candidate reference regions.

[0098] Step S3023: Determine the parameter migration weights of each candidate reference region based on the comprehensive similarity between each candidate reference region and the target region.

[0099] Specifically, after determining the candidate reference regions, the comprehensive similarity is normalized, and the parameter migration weight of each candidate reference region is determined by combining the reliability score of each candidate reference region.

[0100] If the total number of rain gauge stations within the entire study area is [number], then [the total number of] reference areas is [number]. Less than the number of preset reference areas Then, all available rain gauge reference areas will be used as candidate reference areas, and the number of actual reference areas will be adjusted as follows: .

[0101] If there are no reference areas with rain gauges within the entire study area, i.e. If no data is available, a bias correction parameter library cannot be established based on ground-based measured rain gauge data. In this case, the reference area should be expanded to include rain gauge data from adjacent watersheds, adjacent cities, higher-level administrative regions, or regions with similar climates. If no rain gauge data is still available, this method cannot perform bias correction based on ground-based observation parameter migration; instead, scenario correction and uncertainty assessment can only be performed using multi-source grid precipitation, satellite precipitation, reanalysis precipitation, or climate model ensemble results.

[0102] The precipitation bias correction method for climate models in areas without rain gauges provided in this embodiment prioritizes selecting parameter migration sources from reference areas with rain gauges that belong to the same zone as the target area without rain gauges. When the number of reference rain gauges in the same zone is insufficient or there is no reference area with rain gauges, reference areas are selected from other similar climate zones. Migration weights are allocated based on comprehensive similarity, which ensures that the climate characteristics of the reference area and the target area are highly matched, and that the number of reference samples is sufficient, thereby improving the rationality of parameter migration and the accuracy of correction results.

[0103] In some optional implementations, step S3023 above includes: Step a1: Calculate the comprehensive distance between each candidate reference region and the target region based on the preset standardized feature variables, and determine the comprehensive similarity between each candidate reference region and the target region based on the comprehensive distance.

[0104] Specifically, calculating the comprehensive similarity requires considering feature similarity, spatial proximity, and cross-climate migration simultaneously. For any target area without rainfall stations... and any reference area with rain gauge First, calculate the weighted feature distance (feature similarity) between the two:

[0105] in, Indicates the target area With reference area Weighted feature distance between them; Indicates the target area The One standardized feature variable; Indicates reference area The One standardized feature variable; Indicates the first The weights of the feature variables, satisfying:

[0106] No. Weights of each feature variable The appropriate method can be determined using equal weighting, grouped equal weighting, expert weighting, or leave-one-out cross-validation. When prior experience is lacking or the validation sample is insufficient, equal weighting can be used. Preferably, the weights of each feature variable are determined by leave-one-out cross-validation optimization method. That is, one reference area is selected sequentially from the set of reference areas with rain gauges as the target area for simulating no-rain gauges. The remaining reference areas are used for parameter migration and precipitation correction. The weights of each feature variable are optimized with the goal of minimizing the error between the correction result and the measured precipitation in the reference area.

[0107] Further, calculate the target region. With reference area Spatial distance (spatial proximity):

[0108] To simultaneously consider feature similarity, spatial proximity, and cross-climate zone migration effects, a comprehensive distance is constructed:

[0109] in, Indicates the target area With reference area The combined distance between them; Indicates the target area The corresponding similar climate zone number; Indicates reference area The corresponding similar climate zone number; This represents the distance between the centers of similar climate zones to which the target area and the reference area belong. ; Let represent the weights of feature similarity, spatial proximity, and climate zone differences, respectively, and satisfy the following: .

[0110] Convert the overall distance to an overall similarity:

[0111] in, Indicates the target area With reference area The overall similarity between them. The smaller the overall distance, the greater the overall similarity.

[0112] Step a2: Normalize the overall similarity to obtain the initial parameter migration weights.

[0113] Specifically, after determining the candidate reference regions, the comprehensive similarity is normalized to obtain the initial parameter migration weights:

[0114] in, Indicates the reference area with rain gauges. Towards the target area of ​​no rain gauge The similarity weights used in migration bias correction parameters, and satisfying the following:

[0115] Step a3: Based on the historical daily precipitation data of each candidate reference area and the corresponding historical daily precipitation data of the climate model, establish bias correction parameters, apply the bias correction parameters to the corresponding historical precipitation data of the climate model to obtain corrected historical precipitation data, compare the corrected historical precipitation data with the corresponding historical precipitation data to obtain the parameter reliability score of the candidate reference area.

[0116] Specifically, deviation correction parameters are established for each reference area of ​​a rainfall station. For any reference area of ​​a rainfall station... Based on the historical daily precipitation sequences observed in the reference area and the corresponding historical daily precipitation sequences from climate models, the observed wet day frequency, the climate model wet day correction threshold, and the precipitation intensity correction coefficients at different quantile levels are calculated for the reference area, thus forming a bias correction parameter library for all reference areas with rain gauges. A target area without rain gauges... Selected Instead of selecting a single reference area from the rain gauge reference areas to establish parameters, deviation correction parameters are established separately for each rain gauge reference area, and then weighted migration is performed based on similarity weight and reference area reliability score.

[0117] Step a4: Use the parameter reliability scores of each candidate reference region to correct the initial parameter migration weights and obtain the corrected parameter migration weights.

[0118] Specifically, to avoid the adverse impact of low-quality reference areas on the correction results for target areas without rainfall, a reliability evaluation is performed on the deviation correction parameters for each reference area with rainfall. For any reference area with rainfall... The bias correction parameters are established directly using all historical data of the reference area, and then applied to the historical precipitation data of the corresponding climate model for that reference area to obtain the corrected historical precipitation series. The corrected historical precipitation series is then compared with the historical precipitation series observed by surface rain gauges in the reference area to calculate the parameter reliability score for the reference area.

[0119] Reliability evaluation indicators include one or more of the following: monthly precipitation deviation, wet day frequency deviation, extreme quantile deviation, normalized root mean square error, data integrity indicators, and sample length indicators.

[0120] (1) For months The formula for calculating the monthly precipitation deviation is:

[0121] in, Indicates reference area In the month The deviation in monthly precipitation; Indicates reference area In the Daily revised precipitation data from climate models; Indicates reference area In the Daily observed precipitation; This represents all months in historical data. A set of dates; .

[0122] Reference area The comprehensive monthly precipitation deviation is calculated as follows:

[0123] (2) The frequency deviation of wet days is calculated as follows:

[0124] in, This indicates the wet day frequency in the corrected historical precipitation sequence; This indicates the wet day frequency of the observed precipitation sequence.

[0125] (3) The extreme quantile deviation is calculated as follows:

[0126] in, This represents the 95th percentile precipitation in the corrected historical precipitation sequence. It represents the 95th percentile precipitation in the observed precipitation sequence.

[0127] (4) The normalized root mean square error is calculated as follows:

[0128] in, Indicates reference area Normalized root mean square error; Indicates the number of days in historical data; This represents the average observed precipitation.

[0129] (5) The data completeness index is calculated as follows:

[0130] in, Indicates reference area Data completeness indicators; Indicates the number of valid observation days; This indicates that the theory should include the number of observation days.

[0131] (6) The sample length index is calculated as follows:

[0132] in, Indicates reference area The sample length index; Indicates reference area The number of effective observation years; This indicates the preset baseline lifespan, with 30 years being the preferred option.

[0133] Based on the above evaluation indicators, a reliability score for the reference area is constructed:

[0134] in, Indicates reference area Reliability score; Let represent the weights of monthly precipitation deviation, wet day frequency deviation, extreme quantile deviation, and normalized root mean square error, respectively, and satisfy the following:

[0135] A higher reference area reliability score indicates better quality and more sufficient sample length of the observation data in that reference area, and a higher consistency between the bias correction parameters established based on that reference area and the statistical characteristics of observed precipitation.

[0136] Furthermore, by combining the comprehensive similarity score with the reliability score of the reference region, the corrected parameter migration weights are obtained:

[0137] in, This represents the parameter migration weights that simultaneously consider both similarity and the reliability of the reference region. Indicates the target area With reference area Similarity weights between them.

[0138] The precipitation bias correction method for regional climate models provided in this embodiment combines feature distance calculation with comprehensive similarity, and obtains initial weights after normalization. At the same time, it evaluates the reliability of reference area parameters based on measured and simulated precipitation data, and then uses reliability scores to correct the weights. This method not only ensures regional similarity based on topographic and climatic features, but also selects reference samples with high data quality and good correction effects. This dual optimization of weight allocation reduces the errors caused by poor parameters, making the migrated correction parameters more consistent with the actual situation of the target area, and significantly improving the reliability and accuracy of precipitation bias correction results.

[0139] Step S303: Construct bias correction parameters, and based on parameter migration weights, migrate the bias correction parameters of the candidate reference area to the target area, correct the historical precipitation and predicted precipitation of the climate model in the target area, and obtain historical daily precipitation correction data and predicted daily precipitation correction data.

[0140] Specifically, the deviation correction parameters include: wet day frequency, wet day correction threshold, and precipitation intensity correction coefficients at different quantile levels. Step S303 above includes: Step S3031: Based on the historical observed daily precipitation data of the candidate reference area and the corresponding historical daily precipitation data of the climate model, calculate the measured wet day frequency, wet day correction threshold, and precipitation intensity correction coefficients at different quantile levels for the candidate reference area.

[0141] Specifically, deviation correction parameters are established separately for each month. For each month... Reference area The observed daily precipitation sequence is denoted as: Historical daily precipitation sequences for climate models are denoted as: First, determine the wet day threshold, and set the observed wet day discrimination threshold as follows: Preferably: .

[0142] Reference area In the month The observed wet day frequency is calculated as follows:

[0143] in, Indicates the frequency of observed wet days; Indicates the number of observed wet days; This indicates the total number of days observed.

[0144] Based on the observed frequency of wet days, the wet day correction thresholds for historical periods of the climate model in this reference area are deduced:

[0145] in, Indicates reference area In the month The wet day correction threshold for climate models; Empirical quantile functions representing historical daily precipitation sequences in climate models. Within the reference region. In this study, the wet day discrimination threshold in the historical precipitation series of the climate model is determined based on the observed wet day frequency, so that the proportion of wet days in the historical precipitation series of the climate model that are greater than the threshold is consistent with the observed wet day frequency.

[0146] Further, extract wet day intensity samples:

[0147]

[0148] Then, at several quantile levels The following calculates the quantile correction factor for the reference region:

[0149] in, Indicates reference area ,month quantile level The following is a correction factor for precipitation intensity; This indicates the quantile level of the observed wet day precipitation sample. The quantile value below; This indicates the historical wet day precipitation samples from climate models at the quantile level. The quantile value below.

[0150] The set of quantile levels can be represented as:

[0151] in, One or more values ​​can be selected from 0.05, 0.10, 0.25, 0.50, 0.75, 0.90, 0.95, and 0.99.

[0152] Therefore, each reference region forms the following set of parameters:

[0153] in, Indicates reference area The set of deviation correction parameters.

[0154] Step S3032: Based on the parameter migration weights of each candidate reference region, the measured wet day frequencies of each candidate reference region are weighted and fused to obtain the wet day frequency of the target region; based on the wet day frequency of the target region and the empirical quantile function of the historical daily precipitation sequence of the climate model in the target region, the wet day correction threshold of the target region is calculated; based on the parameter migration weights of each candidate reference region, the precipitation intensity correction coefficients of each candidate reference region at different quantile levels are weighted and fused to obtain the precipitation intensity correction coefficient of the target region.

[0155] Specifically, for the target area without rain gauges ,month The target wet day frequency parameter is calculated as follows:

[0156] in, Indicates the month of the target area obtained through migration. Wet day frequency parameters.

[0157] The wet day correction threshold for historical periods in the target region's climate model is calculated as follows:

[0158] in, Indicates the target area In the month The wet day correction threshold; The empirical quantile function represents the historical daily precipitation sequence of the climate model for the target region.

[0159] It should be noted that the wet day correction threshold for the target area is not directly adopted from the reference area threshold, but is determined by back-calculation from the historical daily precipitation series of the target area's own climate model based on the target wet day frequency obtained from migration.

[0160] For months quantile level The precipitation intensity correction factor for the target area is calculated as follows:

[0161] in, This represents the correction coefficient for precipitation intensity in the target area of ​​the no-rain gauge after migration. Therefore, the set of migration bias correction parameters for the target area of ​​the no-rain gauge is obtained: .

[0162] The precipitation bias correction method for regional climate models without rain gauges provided in this embodiment constructs complete correction parameters from three dimensions: wet day frequency, correction threshold, and quantile intensity coefficient. Then, it combines migration weights to fuse parameters from multiple reference regions to obtain target region-specific parameters, simultaneously correcting two types of biases: precipitation frequency and rainfall amount. The parameter division is detailed and highly targeted. Furthermore, the target region threshold is calculated based on its own precipitation characteristics, rather than directly copying reference region values, making the correction rules more consistent with local precipitation patterns and effectively improving the overall correction accuracy and applicability.

[0163] In some optional implementations, historical precipitation data from climate models of the target area are corrected to obtain corrected historical daily precipitation data, including: Step S3033: Compare the historical daily precipitation of the target area climate model. If the historical daily precipitation is less than the wet day correction threshold of the target area, then correct the historical daily precipitation of the corresponding period to a rainless day.

[0164] Specifically, for the target area ,month Climate model historical period Daily precipitation is .

[0165] If the historical daily precipitation is less than the wet day correction threshold for the target area:

[0166] in, Indicates the target area In the month The wet day correction threshold.

[0167] It is then corrected to a rainless day: .

[0168] Step S3034: If the historical daily precipitation is not less than the wet day correction threshold of the target area, calculate the quantile level of the historical daily precipitation in the historical wet day distribution of the climate model of the target area, and calculate the precipitation intensity correction coefficient for the corresponding period based on the quantile level.

[0169] Specifically, if:

[0170] Then calculate the quantile level of the daily precipitation in the historical wet day distribution of the climate model for the target area:

[0171] in, This represents the historical wet day precipitation empirical distribution function of the climate model for the target region.

[0172] According to quantile level The corresponding precipitation intensity correction coefficient is obtained through linear interpolation:

[0173] in, This represents the interpolation function at quantile levels.

[0174] Step S3035: Correct the daily precipitation for the corresponding period using the precipitation intensity correction coefficient to obtain historical daily precipitation correction data.

[0175] Specifically, the corrected calculation of daily precipitation for the historical period is as follows:

[0176] in, This represents the historical daily precipitation in the target area after correction based on the climate model, without rain gauges.

[0177] The precipitation bias correction method for regional climate models without rain gauges provided in this embodiment accurately distinguishes between periods of rainfall and periods without rainfall by using wet day correction thresholds. Then, it combines precipitation distribution quantile matching with corresponding correction coefficients to complete the historical precipitation correction in layers. This method can effectively correct the precipitation frequency bias simulated by the climate model and correct the precipitation intensity bias for different rainfall levels. The correction logic fits the precipitation distribution characteristics, and the step-by-step processing makes the error correction more refined, significantly improving the accuracy of historical precipitation correction data.

[0178] In some optional implementations, the predicted precipitation from the climate model for the target area is corrected to obtain corrected daily precipitation data, including: Step S3036: Calculate the wet day frequency of the target area in the future period based on the original wet day frequency of the climate model in the future period and the original wet day frequency of the climate model in the historical period, and use the preset wet day frequency range for constraint to obtain the target wet day frequency.

[0179] Specifically, the bias correction parameters obtained from the migration are used to correct the daily precipitation data of the climate model for the target area without rain gauges for future periods.

[0180] For the target area ,month Climate model future time period Daily precipitation is:

[0181] To preserve the future wet day frequency variation signal of the climate model relative to historical periods, the future wet day frequency variation coefficient of the climate model in the target area is calculated:

[0182] in, Indicates the month of the target region The coefficient of variation of wet diurnal frequency; This indicates the original wet day frequency for future periods in the climate model; This represents the original wet day frequency for historical periods in the climate model.

[0183] The target wet day frequency for the future period is calculated as follows:

[0184] To avoid future wet day frequency corrections exceeding a reasonable range, upper and lower limits are imposed on the future wet day frequency of the target area. These limits can be determined based on the precipitation climate characteristics of the study area, the statistical range of wet day frequency in similar reference areas, or engineering experience. When reasonable upper and lower limits for the wet day frequency in each month cannot be determined, it is recommended to set the lower limit to 0.01 and the upper limit to 0.99.

[0185]

[0186] in, Indicates the lower limit of wet day frequency; This represents the upper limit of the frequency of wet days. Preferably, when there is a lack of clear upper and lower limits, we take... .

[0187] Step S3037: Based on the target wet day frequency and the empirical quantile function of the future daily precipitation sequence of the climate model in the target area, determine the future wet day correction threshold for the target area, and compare the predicted daily precipitation of the climate model in the target area with the future wet day correction threshold for the target area.

[0188] Specifically, the future wet day correction threshold is determined based on the future target wet day frequency:

[0189] in, Indicates the wet day correction threshold for future periods; This represents the empirical quantile function of the future precipitation sequence in the target region's climate model.

[0190] Step S3038: If the predicted daily precipitation is less than the future wet day correction threshold for the target area, then the predicted daily precipitation for the corresponding period is corrected to a rainless day.

[0191] Specifically, if Then it is corrected to a rainless day: .

[0192] Step S3039: If the predicted daily precipitation is not less than the future wet day correction threshold of the target area, calculate the quantile level of the predicted daily precipitation in the future wet day precipitation distribution of the climate model in the target area, and correct the predicted daily precipitation by calling the corresponding precipitation intensity correction coefficient of the target area according to the quantile level, so as to obtain the predicted daily precipitation correction data; or, calculate the quantile change factor of the climate model of the target area in the future period relative to the historical period at the corresponding quantile level, and correct the predicted daily precipitation by combining the quantile change factor and the precipitation intensity correction coefficient of the target area, so as to obtain the predicted daily precipitation correction data.

[0193] Specifically, if Then calculate the quantile level of the future precipitation on that day in the future wet day distribution of the climate model:

[0194] in, This represents the empirical distribution function of future wet day precipitation in the target region's climate model.

[0195] Furthermore, construct the quantile change factor for future periods relative to historical periods:

[0196] in, Indicates the month of the target region At the quantile level The future—historical quantile change factor; This indicates the future wet day precipitation level in the climate model; This represents the historical wet day precipitation level in the climate model.

[0197] Calculate the corrected future daily precipitation based on the migration correction coefficient and the future change factor:

[0198] Alternatively, in a trend-maintaining implementation, it can also be calculated as follows:

[0199] The latter form represents first constructing historical baseline quantile values ​​for the target region using migration correction parameters, and then superimposing the future quantile change signal from the climate model relative to the history.

[0200] The precipitation bias correction method for regional climate models without rain gauges provided in this embodiment combines historical and future original wet day frequencies to extrapolate future rainfall frequencies and sets numerical range constraints to avoid extreme anomalies in the results. It divides rainy days and rainless days by thresholds and then corrects rainfall by combining quantile variation factors and correction coefficients. This method not only preserves the precipitation evolution trend predicted by the climate model, but also corrects frequency and intensity biases in a stratified manner, making the predicted precipitation correction results more consistent with objective laws and improving the rationality and accuracy of future precipitation data.

[0201] Step S304 involves optimizing the similarity feature weights using leave-one-out cross-validation, and outputting the bias correction result and uncertainty range for the target region. For details, please refer to [link to relevant documentation]. Figure 2 Step S204 of the illustrated embodiment will not be described again here.

[0202] This embodiment also provides a precipitation bias correction device for a climate model in a rainless area. This device is used to implement the above embodiments and preferred embodiments, and details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0203] This embodiment provides a precipitation bias correction device for a climate model in an area without rain gauges, such as... Figure 4 As shown, it includes: The similarity feature vector construction module 401 is used to obtain the basic data of the reference area with rain gauges and the target area without rain gauges, and to construct the regional similarity feature vector based on the basic data.

[0204] The candidate reference area and parameter migration weight calculation module 402 is used to perform cluster analysis based on regional similarity feature vectors to obtain multiple similar climate zones, perform rainfall station coverage checks on each similar climate zone, and determine the candidate reference area and parameter migration weight for each target area.

[0205] The parameter migration and target area data correction module 403 is used to construct bias correction parameters and, based on the parameter migration weights, migrate the bias correction parameters of the candidate reference area to the target area, correct the historical precipitation and predicted precipitation of the climate model in the target area, and obtain historical daily precipitation correction data and predicted daily precipitation correction data.

[0206] The result verification module 404 is used to optimize the similarity feature weights using leave-one-out cross-validation and output the deviation correction result and uncertainty range of the target region.

[0207] The precipitation bias correction device for regional climate models with no rain gauges provided in this embodiment of the invention can execute the precipitation bias correction method for regional climate models with no rain gauges provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the above modules and units are the same as in the corresponding embodiments described above, and will not be repeated here.

[0208] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0209] The following is a detailed reference. Figure 5 The diagram illustrates a structural schematic suitable for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 501, which can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) 502 or a program loaded from memory 508 into random access memory (RAM) 503. The RAM 503 also stores various programs and data required for the operation of the electronic device. The processor 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0210] Typically, the following devices can be connected to I / O interface 505: input devices 506 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 507 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 508 including, for example, magnetic tapes, hard disks, etc.; and communication devices 509. Communication device 509 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 5 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0211] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a memory 508, or installed from a ROM 502. When the computer program is executed by the processor 501, it performs the functions defined in the precipitation bias correction method for regional climate models without rain gauges according to embodiments of the present invention.

[0212] Figure 5The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0213] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the precipitation deviation correction method for regional climate models without rain gauges shown in the above embodiments is implemented.

[0214] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for correcting precipitation bias in regional climate models without rain gauges, characterized in that, The method includes: Basic data of the reference area with rain gauges and the target area without rain gauges are obtained respectively, and regional similarity feature vectors are constructed based on the basic data; Cluster analysis is performed based on the regional similarity feature vectors to obtain multiple similar climate zones. Rainfall station coverage is checked in each similar climate zone to determine candidate reference zones and parameter migration weights for each target region. This includes: counting the number of reference zones with rainfall stations in each similar climate zone; if the number of reference zones is not less than a preset number, then a preset number of candidate reference zones are selected in the similar climate zone to which the target region belongs, based on comprehensive similarity; if the number of reference zones is less than the preset number, then all reference zones with rainfall stations in the similar climate zone to which the target region belongs are selected as candidate reference zones, and candidate reference zones are supplemented from other similar climate zones based on comprehensive similarity to obtain a preset number of candidate reference zones; and the parameter migration weights of each candidate reference zone are determined based on the comprehensive similarity between each candidate reference zone and the target region. Bias correction parameters are constructed, and based on the parameter migration weights, the bias correction parameters of the candidate reference area are migrated to the target area to correct the historical precipitation and predicted precipitation of the climate model in the target area, thereby obtaining historical daily precipitation correction data and predicted daily precipitation correction data. Leave-one-out cross-validation is used to optimize the similarity feature weights, and the deviation correction results and uncertainty range of the target region are output.

2. The method according to claim 1, characterized in that, The basic data for the reference area include: historical daily precipitation data, historical daily precipitation data for the first climate model, future daily precipitation data for the first climate model, reference geographical location data, reference topographic data, reference land use or underlying surface data, and reference climate background data. The basic data for the target area include: historical daily precipitation data for the second climate model, future daily precipitation data for the second climate model, target geographical location data, target topographic data, target land use or underlying surface data, and target climate background data. Based on the aforementioned basic data, a region similarity feature vector is constructed, including: The basic data is preprocessed to obtain a unified data source. The preprocessing includes: coordinate unification, time scale unification, spatial interpolation of climate model grid data, calendar format unification, missing value imputation, outlier removal, and precipitation unit unification. Geographic features, topographic features, underlying surface features, and precipitation statistics are extracted based on the unified data source. Each feature is standardized to construct a region similarity feature vector.

3. The method according to claim 1, characterized in that, Based on the comprehensive similarity between each candidate reference region and the target region, the parameter migration weights for each candidate reference region are determined, including: The comprehensive distance between each candidate reference region and the target region is calculated based on preset standardized feature variables, and the comprehensive similarity between each candidate reference region and the target region is determined based on the comprehensive distance. The initial parameter migration weights are obtained by normalizing the comprehensive similarity. Based on the historical daily precipitation data of each candidate reference area and the corresponding historical daily precipitation data of the climate model, a bias correction parameter is established. The bias correction parameter is applied to the corresponding historical precipitation data of the climate model to obtain the corrected historical precipitation data. The corrected historical precipitation data is compared with the corresponding historical precipitation data to obtain the parameter reliability score of the candidate reference area. The initial parameter migration weights are corrected by using the parameter reliability scores of each candidate reference region to obtain the corrected parameter migration weights.

4. The method according to claim 1, characterized in that, The deviation correction parameters include: wet day frequency, wet day correction threshold, and precipitation intensity correction coefficients at different quantile levels. The deviation correction parameters are constructed, and based on the parameter migration weights, the deviation correction parameters of the candidate reference region are migrated to the target region, including: Based on historical daily precipitation data of the candidate reference area and the corresponding historical daily precipitation data of the climate model, the measured wet day frequency, wet day correction threshold, and precipitation intensity correction coefficients at different quantile levels of the candidate reference area are calculated. Based on the parameter migration weights of each candidate reference region, the measured wet day frequencies of each candidate reference region are weighted and fused to obtain the wet day frequency of the target region. Based on the wet day frequency of the target region and the empirical quantile function of the historical daily precipitation series of the climate model in the target region, the wet day correction threshold of the target region is calculated. Based on the parameter migration weights of each candidate reference region, the precipitation intensity correction coefficients of each candidate reference region at different quantile levels are weighted and fused to obtain the precipitation intensity correction coefficient of the target region.

5. The method according to claim 4, characterized in that, Historical precipitation data from climate models in the target area were corrected to obtain corrected historical daily precipitation data, including: Compare the historical daily precipitation data of the climate model in the target area. If the historical daily precipitation data is less than the wet day correction threshold of the target area, then the historical daily precipitation data for the corresponding period is corrected to rainless days. If the historical daily precipitation is not less than the wet day correction threshold of the target area, then the quantile level of the historical daily precipitation in the historical wet day distribution of the climate model of the target area is calculated, and the precipitation intensity correction coefficient for the corresponding period is calculated based on the quantile level. The daily precipitation for the corresponding period is corrected using the precipitation intensity correction coefficient to obtain historical daily precipitation correction data.

6. The method according to claim 4, characterized in that, The predicted precipitation from the climate model for the target area is corrected to obtain corrected daily precipitation data, including: The wet day frequency of the target area in the future period is calculated based on the original wet day frequency of the climate model in the future period and the original wet day frequency of the climate model in the historical period. The target wet day frequency is obtained by using a preset wet day frequency range as a constraint. Based on the target wet day frequency and the empirical quantile function of the future daily precipitation sequence of the climate model in the target area, the future wet day correction threshold of the target area is determined, and the predicted daily precipitation of the climate model in the target area is compared with the future wet day correction threshold of the target area. If the predicted daily precipitation is less than the future wet day correction threshold for the target area, the predicted daily precipitation for the corresponding period will be corrected to a rainless day. If the predicted daily precipitation is not less than the future wet day correction threshold for the target area, then the quantile level of the predicted daily precipitation in the future wet day precipitation distribution of the climate model in the target area is calculated, and the predicted daily precipitation is corrected according to the precipitation intensity correction coefficient corresponding to the target area based on the quantile level, to obtain the predicted daily precipitation correction data; or, the quantile change factor of the climate model in the target area for future periods relative to historical periods at the corresponding quantile level is calculated, and the predicted daily precipitation is corrected by combining the quantile change factor and the precipitation intensity correction coefficient of the target area, to obtain the predicted daily precipitation correction data.

7. A precipitation bias correction device for a regional climate model without rain gauges, characterized in that, The device includes: The similarity feature vector construction module is used to obtain basic data of the reference area with rain gauges and the target area without rain gauges, and construct regional similarity feature vectors based on the basic data. The candidate reference region and parameter migration weight calculation module is used to perform cluster analysis based on the region similarity feature vector to obtain multiple similar climate zones, perform rainfall station coverage checks on each similar climate zone, and determine the candidate reference regions and parameter migration weights for each target region. This includes: counting the number of reference regions with rainfall stations in each similar climate zone; if the number of reference regions is not less than a preset number, then selecting a preset number of candidate reference regions in the similar climate zone to which the target region belongs, based on comprehensive similarity; if the number of reference regions is less than the preset number, then all reference regions with rainfall stations in the similar climate zone to which the target region belongs are used as candidate reference regions, and candidate reference regions are supplemented from other similar climate zones based on comprehensive similarity to obtain a preset number of candidate reference regions; and determining the parameter migration weights of each candidate reference region based on the comprehensive similarity between each candidate reference region and the target region. The parameter migration and target area data correction module is used to construct bias correction parameters and, based on the parameter migration weights, migrate the bias correction parameters of the candidate reference area to the target area, correct the historical precipitation and predicted precipitation of the climate model in the target area, and obtain historical daily precipitation correction data and predicted daily precipitation correction data. The result verification module is used to optimize the similarity feature weights using leave-one-out cross-validation and output the deviation correction results and uncertainty range of the target region.

8. An electronic device, characterized in that, include: A memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, the processor executing the computer instructions to perform the method of any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a computer to perform the method of any one of claims 1 to 6.

Images