A method and system for separating local and non-local precipitation effects of land use and cover change

Abstract

The application provides a land use and cover change local and non-local precipitation effect separation method and system, and relates to the technical field of climate change influence analysis and land-air interaction research. The application is based on observation and reanalysis data, identifies a processing grid that changes, and constructs a control sample library, and keeps the climate background consistent through boundary layer stability and seasonal stratification and circulation screening. A fan-shaped weighted window is established in the main water vapor transport direction to synthesize counterfactual precipitation time series, and precipitation feedback is extracted. In combination with a kernel weighted spatial weight matrix, non-local influence is represented, a hierarchical statistical model introducing a spatial correlation random field is established, local and non-local precipitation effects are quantitatively separated, and the scientificity and reliability of effect attribution are improved.

Description

Technical Field

[0001] This invention relates to the field of climate change impact analysis and land-atmosphere interaction research technology, and in particular to a method and system for separating the local and non-local precipitation effects of land use and cover change. Background Technology

[0002] Land use and land cover change (LUCC) is one of the major anthropogenic forcings affecting regional and even global climate systems. It influences boundary layer structure, vertical and horizontal transport of moisture and energy by altering land surface parameters such as albedo, evapotranspiration, roughness, soil moisture, and thermal properties, ultimately producing a feedback effect on precipitation. The impact of LUCC on precipitation is a typical scale-dependent process: 1) Local effects: LUCC directly affects the planetary boundary layer above it by altering surface heat and moisture fluxes, triggering precipitation feedback locally; 2) Non-local effects: LUCC transmits signals to distant regions by influencing the moisture, energy, and atmospheric circulation of the surrounding area, thereby affecting precipitation in those distant regions.

[0003] However, existing technologies suffer from four key limitations in separating and diagnosing these two effects: Existing Earth system model sensitivity experiments, such as the Decoupling Experiment or Coupled Experiment, can only calculate the total effect of LUCC on precipitation. While some techniques attempt to differentiate these effects by setting buffer zones or employing idealized experiments, these methods generally face the following difficulties:

[0004] (1) Ideal experiments are difficult to fully simulate the transport process in the real atmosphere, and it is difficult to define a physically clear and non-overlapping "local" boundary. The local effects calculated by the model often include non-local contributions from nearby areas, resulting in fuzzy decomposition results.

[0005] (2) To separate the effects of different scales, a large number of pattern sensitivity experiments need to be designed, which consumes huge computational resources and time and is not suitable for large-scale, multi-scenario rapid diagnosis.

[0006] (3) Lack of quantitative characterization of scale dependence and spatial heterogeneity: Existing physical diagnostic methods (such as water cycle diagnostics) mainly focus on vertical flux exchange between land and atmosphere, and it is difficult to effectively quantify the contribution of horizontal transport to nonlocal effects.

[0007] (4) Traditional statistical models typically treat land surface factors as independent point variables and lack the ability to embed spatial distance, direction, and scale effects into the model as weighting functions (i.e., spatial kernels), thus failing to statistically simulate and quantify nonlocal transport paths effectively. This results in a lack of reliable physical basis for the statistical diagnosis of nonlocal effects.

[0008] In summary, the main drawback of existing technologies lies in the key technical challenge of effectively and quantitatively separating and diagnosing the overlapping local and non-local impacts of LUCC on precipitation in the target area. Summary of the Invention

[0009] To overcome the shortcomings of existing technologies, the purpose of this invention is to provide a method and system for separating the local and non-local effects of land use and cover change on precipitation. This method utilizes observation and reanalysis data to accurately separate the local and non-local effects of land use and cover change on precipitation, significantly improving the scientific validity and reliability of effect attribution.

[0010] To achieve the above objectives, the present invention provides the following solution:

[0011] A method for separating the local and non-local precipitation effects of land use and cover change, comprising:

[0012] Based on observation and reanalysis data, treatment grids that have undergone land use and cover changes are identified, and control sample libraries that have not changed are selected within the same climate zone.

[0013] The spatiotemporal records corresponding to the processing raster are used as processing samples, and the processing samples and the control sample library are stratified according to boundary layer stability and season to obtain the stratified processing samples and control sample library.

[0014] Based on the wind direction, wind speed, temperature and specific humidity at the 850 hPa height level, the control sample library was screened to obtain a matching set that is consistent with the large-scale background of the processed raster.

[0015] With the main water vapor transport direction as the axis, a fan-shaped weighted window is constructed upstream of the processing grid. Spatial weights are calculated according to distance and direction. Precipitation sequences under conditions of no land use and cover change are synthesized within the matching set to obtain a counterfactual precipitation time series.

[0016] The difference between the actual precipitation time series and the counterfactual precipitation time series of the processed grid is used to characterize the precipitation feedback of land use and cover change, and a precipitation feedback sequence is obtained.

[0017] Centered on the processing grid, the land use and cover change intensity of the surrounding grids are kernel-weighted, and a spatial weight matrix is ​​constructed by combining distance attenuation and direction sensitivity rules to obtain the weighted non-local influence item;

[0018] Using the precipitation feedback sequence as the dependent variable, the intensity of local land use and cover change, the non-local impact term, and control variables are incorporated into a hierarchical statistical model. A spatially correlated random field is introduced into the error term to perform parameter inference. The contributions and confidence intervals of local and non-local effects are output to obtain the effect separation results.

[0019] Preferably, the control sample library is constructed within the same climate-ecological zone as the treatment grid, and is matched one by one according to the corresponding time of the treatment sample under the condition of boundary layer stability and seasonal stratification. Only samples that are in the same category as the treatment sample in terms of seasonality and boundary layer stability and have not changed are retained as candidate control samples.

[0020] Preferably, the step of screening the control sample library is constrained by the combined state of wind direction, wind speed, temperature and specific humidity at the 850 hPa height layer, forming the matching set within the control sample library; wherein, the matching set is consistent with the method of stratifying by boundary layer stability and season, and samples that are inconsistent with the processed samples in different time periods or stability are removed.

[0021] Preferably, the fan-shaped weighted window expands upstream of the main water vapor transport direction with the processing grid as the center, and a predetermined fan-shaped angle and radius are set and updated periodically according to the main water vapor transport direction.

[0022] Preferably, the spatial weight matrix determined by the sector weighted window is jointly determined by the distance decay rule and the direction sensitivity rule, and the weights within the same time period are normalized.

[0023] Preferably, the counterfactual precipitation time series is obtained by selecting precipitation data from the control sample at the same time as the processed sample in the matching set, weighting and synthesizing them according to the spatial weight matrix determined by the sector weighting window, and using the counterfactual precipitation time series as the precipitation baseline under the condition of no land use and cover change, and differentiating it with the actual precipitation time series of the processed grid to form the precipitation feedback sequence.

[0024] Preferably, the calculation method for the non-local impact term includes:

[0025] A spatial neighborhood is determined with the processing grid as the center; the orientation of the spatial neighborhood is set in the same way as the sector-shaped weighted window.

[0026] Within the spatial neighborhood, a spatial weight matrix for nonlocal influence terms is constructed based on distance decay rules and direction sensitivity rules, and the spatial weight matrix is ​​normalized according to time sequence.

[0027] The nonlocal impact term is obtained by applying kernel weighting to the land use and cover change intensity of the surrounding grid using the spatial weight matrix.

[0028] Preferably, the hierarchical statistical model uses the precipitation feedback sequence as the dependent variable and the intensity of local land use and cover change, the non-local impact term, and the control variables as independent variables; the control variables include circulation elements used to characterize the large-scale background and other meteorological elements consistent with the reanalysis data.

[0029] Preferably, the spatially correlated random field is introduced into the model error term to characterize the unmodeled regional transport effect, and the spatial and temporal range of parameter inference is limited to be consistent with the processing grid and the precipitation feedback sequence.

[0030] A system for separating the local and non-local precipitation effects of land use and cover change includes:

[0031] Sample construction and stratification units are used to identify treatment grids that have undergone land use and cover change based on observation and reanalysis data, and to select control sample libraries that have not changed within the same climate zone.

[0032] The circulation background matching unit is used to take the spatiotemporal records corresponding to the processing grid as processing samples, and to stratify the processing samples and the control sample library according to the boundary layer stability and the season, so as to obtain the stratified processing samples and control sample library.

[0033] The upstream weighted counterfactual construction unit is used to screen the control sample library based on the wind direction, wind speed, temperature and specific humidity at the 850 hPa height layer to obtain a matching set that is consistent with the large-scale background of the processed raster.

[0034] The precipitation feedback extraction unit is used to construct a fan-shaped weighted window upstream of the processing grid with the main water vapor transport direction as the axis, calculate the spatial weight according to distance and direction, and synthesize the precipitation sequence under the condition of no land use and cover change in the matching set to obtain the counterfactual precipitation time series.

[0035] The non-local impact characterization unit is used to characterize the precipitation feedback of land use and cover change by the difference between the actual precipitation time series and the counterfactual precipitation time series of the processed grid, and to obtain the precipitation feedback sequence.

[0036] The hierarchical statistical modeling unit is used to perform kernel weighting on the land use and cover change intensity of the surrounding grids with the processing grid as the center, and construct a spatial weight matrix by combining distance attenuation and direction sensitivity rules to obtain the weighted non-local influence item.

[0037] The effect separation and output unit is used to take the precipitation feedback sequence as the dependent variable, incorporate the intensity of local land use and cover change, the non-local impact term and control variables into the hierarchical statistical model, introduce a spatially correlated random field into the error term, perform parameter inference, output the contribution and confidence interval of local and non-local effects, and obtain the effect separation result.

[0038] The present invention discloses the following technical effects:

[0039] This invention achieves the separate analysis of precipitation effects of land use and cover change through a sampled design based on observational and reanalysis data, without relying on sensitivity tests and strongly assumed boundary conditions found in traditional numerical models. Existing technologies typically require artificial setting of the underlying surface or the construction of dual-scheme simulations, which suffers from limited spatial resolution, inconsistent circulation feedback, and insufficient sample representativeness. This invention, by directly identifying changed processing grids and a control sample library within the same climate zone within the observational data domain, ensures that precipitation differences originate from real LUCC processes, significantly improving the physical rationality and repeatability of effect identification.

[0040] This invention introduces dual stratification conditions of boundary layer stability and season during the sample matching stage, and further utilizes wind direction, wind speed, temperature, and specific humidity at the 850 hPa height to constrain the screening of the control sample library, ensuring that the processed samples and control samples remain consistent against the large-scale circulation background. Compared with previous methods that only match based on geographical proximity or statistical similarity, this design effectively eliminates background interference from climate anomaly years, avoids misjudging large-scale climate fluctuations as LUCC effects, and thus improves the scientific rigor and robustness of precipitation response attribution.

[0041] This invention constructs a fan-shaped weighted window upstream of the processing grid, using the main water vapor transport direction as the axis. A spatial weight matrix is ​​generated by combining distance attenuation and direction-sensitive rules, which is used to synthesize precipitation sequences under conditions without land use and cover change within the matching set. This counterfactual precipitation time series fully considers the water vapor transport characteristics and non-uniform spatial effects of the upstream region, thus making the obtained "LUCC-free" baseline closer to the actual physical process. Compared with traditional interpolation or spatial averaging methods, this weighting strategy reduces systematic bias caused by spatial configuration differences, providing a reliable benchmark for subsequent precipitation feedback extraction.

[0042] This invention centers on a processing grid and constructs a spatial weight matrix based on distance attenuation and direction sensitivity rules. It then applies kernel weighting to the LUCC intensity of surrounding grids, forming a weighted nonlocal impact term. This design incorporates the cumulative and directional effects of land change in the surrounding area into the same framework, enabling the model to simultaneously identify both local direct responses and nonlocal precipitation feedback caused by water vapor transport. This overcomes the limitation of existing studies that can only qualitatively analyze the "underlying surface influence range," providing a quantitative approach to the cascading precipitation effects of regional LUCC.

[0043] This invention uses precipitation feedback sequences as the dependent variable and constructs a hierarchical statistical model by combining local LUCC intensity, non-local impact terms, and control variables, while introducing a spatially correlated random field into the error term. This model maintains physical consistency while possessing statistical adaptability, enabling the separation of local and non-local effects, identification of controlling factors, and quantification of confidence intervals within the same framework. Compared to traditional regression or fixed-effects models, the hierarchical structure of this invention can achieve multi-scale information fusion under spatially heterogeneous conditions, improving the robustness of effect estimation and the interpretability of results, and providing a highly reliable data foundation for the monitoring and decision support of LUCC climate effects. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 A flowchart of the method provided in an embodiment of the present invention;

[0046] Figure 2 A schematic diagram illustrating the possible distribution of the target area and surrounding area provided for embodiments of the present invention;

[0047] Figure 3 A spatial schematic diagram of the regression coefficients of local and non-local precipitation effects provided for embodiments of the present invention;

[0048] Figure 4 A spatial schematic diagram of the separation results of local and non-local precipitation effects provided in an embodiment of the present invention. Detailed Implementation

[0049] 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, and 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.

[0050] The purpose of this invention is to provide a method and system for separating the local and non-local precipitation effects of land use and cover change. By introducing circulation constraint matching, upstream weighted synthesis and spatial hierarchical modeling, the invention achieves quantitative identification and uncertainty assessment of precipitation response caused by land use and cover change, thereby improving the accuracy and applicability of land-atmosphere coupling process analysis.

[0051] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0052] Figure 1 The method flowchart provided in the embodiments of the present invention is as follows: Figure 1 As shown, the present invention provides a method for separating the local and non-local precipitation effects of land use and cover change, comprising:

[0053] Step 100: Based on observation and reanalysis data, identify the treatment grids where land use and cover changes have occurred, and select control sample libraries that have not changed within the same climate zone;

[0054] Step 200: Use the spatiotemporal records corresponding to the processed raster as the processed samples, and stratify the processed samples and the control sample library according to the boundary layer stability and the season to obtain the stratified processed samples and control sample library.

[0055] Step 300: Based on the wind direction, wind speed, temperature and specific humidity at the 850 hPa height level, the control sample library is screened to obtain a matching set that is consistent with the large-scale background of the processed raster.

[0056] Step 400: Construct a fan-shaped weighted window upstream of the processing grid with the main water vapor transport direction as the axis, calculate the spatial weight according to distance and direction, and synthesize the precipitation sequence under the condition of no land use and cover change in the matching set to obtain the counterfactual precipitation time series;

[0057] Step 500: Characterize the precipitation feedback of land use and cover change by the difference between the actual precipitation time series and the counterfactual precipitation time series of the processed raster, and obtain the precipitation feedback series;

[0058] Step 600: Centered on the processing grid, apply kernel weighting to the land use and cover change intensity of the surrounding grids, and construct a spatial weight matrix by combining distance decay and direction sensitivity rules to obtain the weighted nonlocal impact term;

[0059] Step 700: Using the precipitation feedback sequence as the dependent variable, incorporate the intensity of local land use and cover change, non-local impact terms, and control variables into the hierarchical statistical model. Introduce a spatially correlated random field into the error term, perform parameter inference, output the contributions and confidence intervals of local and non-local effects, and obtain the effect separation results.

[0060] Specifically, the technical approach in this embodiment is as follows:

[0061] Step 1: Quantify precipitation feedback.

[0062] To overcome the shortcomings of traditional model sensitivity test methods, such as high computational cost, limited spatial resolution, and difficulty in distinguishing climate background effects, this invention proposes a Reanalysis Counterfactual Method (RCM) based on reanalysis data. This method quantitatively obtains the feedback effect of land use / cover change (LUCC) on precipitation without relying on numerical sensitivity tests. The core idea is to directly estimate the amount of precipitation change caused by LUCC by constructing comparable samples between a "LUCC-containing scenario" and a "LUCC-free counterfactual scenario."

[0063] (1) Determination of treatment and control areas:

[0064] First, within a given study period (e.g., 1950 to present), treatment rasters exhibiting significant LUCC (Local Area Conversion) are identified based on land cover data, such as forest-to-farmland conversion, grassland-to-urban conversion, or significant expansion of irrigated areas. Simultaneously, within the same eco-climate zone (based on aridity index and land type), areas without significant LUCC are selected as a control sample library. This zoning helps ensure consistency in climate and ecological background among samples, thereby reducing systematic bias.

[0065] (2) Stability and seasonal stratification:

[0066] Considering the influence of atmospheric boundary layer state on precipitation formation, the samples are stratified using the Monin–Obukhov length L or the apparent Richardson number Rib:

[0067] Unstable layer (U): <-0.1 ;

[0068] Neutral layer (N): ≤-0.1 ;

[0069] Stable layer (S): >0.1;

[0070] The precipitation difference is calculated separately during the growing season and the non-growing season, and finally weighted and summarized according to the frequency of occurrence of each layer to improve signal stability.

[0071] (3) Circulation similarity screening:

[0072] To ensure that the treatment and control areas are under similar large-scale climatic conditions, wind direction, wind speed, and temperature and humidity characteristics were used for screening. A target grid was established. With candidate grid points In time The 850 hPa circulation parameters are respectively wind direction Wind speed ,temperature , specific moisture A match can only be included in the matching set if the following conditions are met:

[0073] (1)

[0074] (4) Upstream weight and spatial matching:

[0075] Considering the nonlocal transport characteristics of precipitation, a weighted sector window is constructed upstream of the target grid (60° angle, distance 200–800 km) with the principal component direction Ѳ of the water vapor flux F=qV as the principal axis. The weights of candidate grid points are defined as follows:

[0076] (2)

[0077] in, The distance between grid points It is the direction angle. The feature transmission scale (default 400km). This is the direction set parameter (default 4). This indicates the recirculation share of candidate grid points.

[0078] (5) Constructing counterfactual precipitation sequences:

[0079] For each processing grid point Based on the precipitation and major climate element sequences of the period T0 (e.g., 10 years) before the occurrence of LUCC, the linear weights are solved within the comparison set that satisfies equation (1) and weight equation (2). :

[0080] (3)

[0081] in This is the regularization coefficient (default 0.1). The upstream weights are normalized.

[0082] The obtained synthetic control precipitation was:

[0083] (4)

[0084] This represents the counterfactual precipitation sequence for that grid point under the "no LUCC scenario".

[0085] (6) Calculate the LUCC precipitation feedback:

[0086] The precipitation feedback induced by LUCC is defined as the difference between actual precipitation and counterfactual precipitation:

[0087] (5)

[0088] In the formula, Let represent the LUCC precipitation feedback effect of the i-th grid at time t.

[0089] Step 2: Construct and quantify the nonlocal transmission structure.

[0090] (1) Define nonlocal LUCC intensity:

[0091] Quantization target grid Surrounding non-local grids The intensity of LUCC is defined as the nonlocal land surface change factor. .

[0092] (2) Constructing a spatial transmission distance weighting function ( ):

[0093] Based on the land-atmosphere coupling theory and physical transmission characteristics, a weighting function is constructed to describe the attenuation of land surface signals with distance. ,in It is the target grid With surrounding grid Distance between:

[0094] (6)

[0095] in, The characteristic transport scale reflects the effective range of influence of land-atmosphere interaction.

[0096] (3) Constructing the spatial transmission direction weighting function ( ):

[0097] Based on land-atmosphere coupling theory and physical transmission characteristics, a weighting function describing the direction of signal transmission over the land surface is constructed. ,in It is a grid With grid Direction between:

[0098] (7)

[0099] in, : grid Relative to the dominant direction The included angle; : Direction sensitivity parameter, controls the rate at which the weight decays with angular deviation. The smaller the value, the more concentrated the weight is in the dominant direction; when hour, =1 (maximum weight); with As the weight increases, the weight exhibits a Gaussian (bell-shaped) decay.

[0100] (4) Construct the spatial weight matrix ( ):

[0101] By combining the distance weights and the orientation weights, the final nonlocal space weight matrix elements are constructed:

[0102] (8)

[0103] Step 3: Construct a Bayesian hierarchical model

[0104] (1) Construct a Bayesian statistical model based on spatial hierarchy effects:

[0105] Rainfall feedback Local LUCC variation as the dependent variable and spatially weighted nonlocal LUCC variation Using spatial hierarchy effects as independent variables, a Bayesian statistical model is constructed:

[0106] (9)

[0107] in, Represents the target grid The precipitation feedback effect For local LUCC intensity ; Indicates the passage through the space core Weighted nonlocal LUCC impact term ; This represents control variables (e.g., background climate field, key heat and humidity factors, etc.). , , The regression coefficients to be solved represent the contribution strengths of local effects, non-local effects, and control variables, respectively. This represents the spatially correlated random effects term caused by large-scale climate or circulation background. This indicates an independent noise term.

[0108] Random items It satisfies the spatial Gaussian process (GP) distribution:

[0109] (10)

[0110] in, Let be the spatial covariance function. Represents the variance of a random field. This indicates the characteristic correlation scale. This term can characterize the spatial propagation effect of indirect LUCC perturbations (such as energy transfer, convection triggering).

[0111] (2) Model parameter solution and uncertainty estimation:

[0112] Estimating the parameter vector using Bayesian inference method .

[0113] The prior distribution is set as follows:

[0114] (11)

[0115] Posterior sampling is performed using Markov Chain Monte Carlo (MCMC) or Integrated Nested Laplace Approximation (INLA). The resulting parameters not only provide effect estimates but also 95% confidence intervals (posterior confidence intervals) to quantify the uncertainty of the diagnostic outcome.

[0116] (3) Effect separation and quantitative diagnosis:

[0117] After obtaining parameter estimates, the contribution of local precipitation effect ( ) and nonlocal precipitation effects ( They are represented as follows:

[0118] (12)

[0119] (13)

[0120] In the formula, Characterizing the direct contribution of local LUCC to precipitation feedback, It includes a combined nonlocal contribution from spatial kernel transmission and random field propagation.

[0121] By comparison and The spatial distribution of the data can determine the main control area and influence range of land-atmosphere coupling, enabling spatial identification and quantitative separation of LUCC precipitation feedback.

[0122] like Figure 2 The diagram shown is a schematic representation of the possible distribution of the target area and its surrounding area according to an embodiment of the present invention. Figure 2 The central black square represents the processing grid where land use and cover change occur, surrounded by corresponding surrounding grid areas, used to describe the spatial relationship between upstream, lateral, and downstream grid points under the fan-shaped weighted window. Through this spatial structure illustration, this invention establishes a fan-shaped weighted window upstream of the processing grid and determines distance attenuation and direction sensitivity rules based on the main water vapor transport direction, thereby constructing a spatial weight matrix for counterfactual precipitation time series synthesis and nonlocal impact term calculation, achieving a physical spatial mapping of local and nonlocal effects.

[0123] like Figure 3 The diagram shown illustrates the spatial distribution of regression coefficients for local and non-local precipitation effects according to an embodiment of the present invention. The upper diagram represents the spatial pattern of the local effect regression coefficients, and the lower diagram represents the spatial pattern of the non-local effect regression coefficients. Figure 3 As can be seen, the hierarchical statistical model established in this invention can simultaneously estimate the intensity of the local and non-local responses of land use and cover change to precipitation in different spatial units at global or regional scales. The two responses show significant differences in numerical distribution and spatial consistency, which verifies the feasibility and robustness of the method of this invention in separating the coupling effects of land and atmosphere processes.

[0124] like Figure 4 The figure shown is a spatial distribution diagram of the separation results of local and non-local precipitation effects provided by an embodiment of the present invention. The upper figure shows the distribution of local precipitation effects, and the lower figure shows the distribution of non-local precipitation effects. By performing hierarchical modeling and inference on the precipitation feedback sequence obtained by differencing the counterfactual precipitation time series and the actual precipitation time series, the present invention achieves spatial saliency identification of precipitation effects. Figure 4 It is evident that the local and non-local effects in different regions exhibit significant spatial differences. The local effects are prominent in areas with strong land surface changes, while the non-local effects are more pronounced in downstream areas that are significantly affected by water vapor transport. This fully demonstrates that the method of the present invention can effectively characterize the multi-scale influence structure of LUCC on precipitation.

[0125] Corresponding to the above, this embodiment also provides a system for separating the local and non-local precipitation effects of land use and cover change, including:

[0126] Sample construction and stratification units are used to identify treatment grids that have undergone land use and cover change based on observation and reanalysis data, and to select control sample libraries that have not changed within the same climate zone.

[0127] The circulation background matching unit is used to take the spatiotemporal records corresponding to the processing grid as processing samples, and to stratify the processing samples and the control sample library according to the boundary layer stability and the season, so as to obtain the stratified processing samples and control sample library.

[0128] The upstream weighted counterfactual construction unit is used to screen the control sample library based on the wind direction, wind speed, temperature and specific humidity at the 850 hPa height layer to obtain a matching set that is consistent with the large-scale background of the processed raster.

[0129] The precipitation feedback extraction unit is used to construct a fan-shaped weighted window upstream of the processing grid with the main water vapor transport direction as the axis, calculate the spatial weight according to distance and direction, and synthesize the precipitation sequence under the condition of no land use and cover change in the matching set to obtain the counterfactual precipitation time series.

[0130] The non-local impact characterization unit is used to characterize the precipitation feedback of land use and cover change by the difference between the actual precipitation time series and the counterfactual precipitation time series of the processed grid, and to obtain the precipitation feedback sequence.

[0131] The hierarchical statistical modeling unit is used to perform kernel weighting on the land use and cover change intensity of the surrounding grids with the processing grid as the center, and construct a spatial weight matrix by combining distance attenuation and direction sensitivity rules to obtain the weighted non-local influence item.

[0132] The effect separation and output unit is used to take the precipitation feedback sequence as the dependent variable, incorporate the intensity of local land use and cover change, the non-local impact term and control variables into the hierarchical statistical model, introduce a spatially correlated random field into the error term, perform parameter inference, output the contribution and confidence interval of local and non-local effects, and obtain the effect separation result.

[0133] Compared with the traditional LUCC precipitation feedback analysis method that relies on model sensitivity experiments, this invention has the following advantages:

[0134] (1) Innovative computational framework and significantly reduced cost: This invention proposes a counterfactual inference method (RCM) based on reanalysis data, which can directly obtain quantitative precipitation feedback caused by LUCC from observation and reanalysis data without having to conduct complex model sensitivity experiments again, greatly reducing computational costs and improving spatial resolution and operability.

[0135] (2) Spatial structure optimization and enhanced physical interpretability: By constructing a dual-core spatial weight function that combines distance attenuation and prevailing wind direction, the physical characterization of the spatial transmission of land surface disturbance signals and their upstream impact is realized. This can effectively capture the long-range effects of non-local LUCC on precipitation in the target area, and improve the physical consistency and interpretability of the model.

[0136] (3) Upgrading the statistical model and improving the accuracy of effect identification: A Bayesian hierarchical model based on Gaussian process was introduced, and a spatially correlated random field was embedded in the traditional regression framework. This achieved probabilistic separation and uncertainty quantification of local and non-local precipitation effects, significantly enhancing the robustness and scientific credibility of the results.

[0137] (4) Strong comprehensive applicability and intuitive visualization of results: This invention can be applied to different scales, different climate zones and multi-source data (remote sensing, reanalysis or model output). The obtained LUCC precipitation feedback results have clear spatial visualization characteristics, which facilitates practical applications in scientific research, climate assessment and regional planning.

[0138] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0139] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for separating the local and non-local precipitation effects of land use and cover change, characterized in that, include: Based on observation and reanalysis data, treatment grids that have undergone land use and cover changes are identified, and control sample libraries that have not changed are selected within the same climate zone. The spatiotemporal records corresponding to the processing raster are used as processing samples, and the processing samples and the control sample library are stratified according to boundary layer stability and season to obtain the stratified processing samples and control sample library. Based on the wind direction, wind speed, temperature and specific humidity at the 850 hPa height level, the control sample library was screened to obtain a matching set that is consistent with the large-scale background of the processed raster. With the main water vapor transport direction as the axis, a fan-shaped weighted window is constructed upstream of the processing grid. Spatial weights are calculated according to distance and direction. Precipitation sequences under conditions of no land use and cover change are synthesized within the matching set to obtain a counterfactual precipitation time series. The difference between the actual precipitation time series and the counterfactual precipitation time series of the processed grid is used to characterize the precipitation feedback of land use and cover change, and a precipitation feedback sequence is obtained. Centered on the processing grid, the land use and cover change intensity of the surrounding grids are kernel-weighted, and a spatial weight matrix is ​​constructed by combining distance attenuation and direction sensitivity rules to obtain the weighted non-local influence item; Using the precipitation feedback sequence as the dependent variable, the intensity of local land use and cover change, the non-local impact term, and control variables are incorporated into a hierarchical statistical model. A spatially correlated random field is introduced into the error term to perform parameter inference. The contributions and confidence intervals of local and non-local effects are output to obtain the effect separation results.

2. The method for separating the local and non-local precipitation effects of land use and cover change according to claim 1, characterized in that, The control sample library is constructed within the same climate-ecological partition as the treatment raster, and is matched one by one according to the corresponding time of the treatment sample under the condition of boundary layer stability and seasonal stratification. Only samples that are in the same category as the treatment sample in terms of seasonality and boundary layer stability and have not changed are retained as candidate control samples.

3. The method for separating the local and non-local precipitation effects of land use and cover change according to claim 1, characterized in that, The screening process for the control sample library is constrained by the combined state of wind direction, wind speed, temperature, and specific humidity at the 850 hPa altitude layer, forming the matching set within the control sample library. The matching set is consistent with the method of stratifying by boundary layer stability and season, and samples that are inconsistent with the processed samples in terms of time period or stability are removed.

4. The method for separating the local and non-local precipitation effects of land use and cover change according to claim 1, characterized in that, The fan-shaped weighted window expands upstream along the main water vapor transport direction with the processing grid as the center, and sets a predetermined fan angle and radius, and updates it periodically according to the main water vapor transport direction.

5. The method for separating the local and non-local precipitation effects of land use and cover change according to claim 1, characterized in that, The spatial weight matrix determined by the sector-weighted window is jointly determined by the distance decay rule and the direction sensitivity rule, and the weights within the same time period are normalized.

6. The method for separating the local and non-local precipitation effects of land use and cover change according to claim 1, characterized in that, The counterfactual precipitation time series is obtained by selecting precipitation data from the control sample at the same time as the processed sample in the matching set, weighting and synthesizing them according to the spatial weight matrix determined by the sector weighting window, and using the counterfactual precipitation time series as the precipitation baseline under the condition of no land use and cover change, and differentiating it with the actual precipitation time series of the processed grid to form the precipitation feedback sequence.

7. The method for separating the local and non-local precipitation effects of land use and cover change according to claim 1, characterized in that, The calculation methods for the nonlocal impact item include: A spatial neighborhood is determined with the processing grid as the center; the orientation of the spatial neighborhood is set in the same way as the sector-shaped weighted window. Within the spatial neighborhood, a spatial weight matrix for nonlocal influence terms is constructed based on distance decay rules and direction sensitivity rules, and the spatial weight matrix is ​​normalized according to time sequence. The nonlocal impact term is obtained by applying kernel weighting to the land use and cover change intensity of the surrounding grid using the spatial weight matrix.

8. The method for separating the local and non-local precipitation effects of land use and cover change according to claim 1, characterized in that, The hierarchical statistical model uses the precipitation feedback sequence as the dependent variable and the intensity of local land use and cover change, the non-local impact term, and the control variables as independent variables. The control variables include circulation elements used to characterize the large-scale background and other meteorological elements consistent with the reanalysis data.

9. The method for separating the local and non-local precipitation effects of land use and cover change according to claim 1, characterized in that, The spatially correlated random field is introduced into the model error term to characterize the unmodeled regional transport effect, and the spatial and temporal range of parameter inference is limited to be consistent with the processed grid and the precipitation feedback sequence.

10. A system for separating the local and non-local precipitation effects of land use and cover change, characterized in that, include: Sample construction and stratification units are used to identify treatment grids that have undergone land use and cover change based on observation and reanalysis data, and to select control sample libraries that have not changed within the same climate zone. The circulation background matching unit is used to take the spatiotemporal records corresponding to the processing grid as processing samples, and to stratify the processing samples and the control sample library according to the boundary layer stability and the season, so as to obtain the stratified processing samples and control sample library. The upstream weighted counterfactual construction unit is used to screen the control sample library based on the wind direction, wind speed, temperature and specific humidity at the 850 hPa height layer to obtain a matching set that is consistent with the large-scale background of the processed raster. The precipitation feedback extraction unit is used to construct a fan-shaped weighted window upstream of the processing grid with the main water vapor transport direction as the axis, calculate the spatial weight according to distance and direction, and synthesize the precipitation sequence under the condition of no land use and cover change in the matching set to obtain the counterfactual precipitation time series. The non-local impact characterization unit is used to characterize the precipitation feedback of land use and cover change by the difference between the actual precipitation time series and the counterfactual precipitation time series of the processed grid, and to obtain the precipitation feedback sequence. The hierarchical statistical modeling unit is used to perform kernel weighting on the land use and cover change intensity of the surrounding grids with the processing grid as the center, and construct a spatial weight matrix by combining distance attenuation and direction sensitivity rules to obtain the weighted non-local influence item. The effect separation and output unit is used to take the precipitation feedback sequence as the dependent variable, incorporate the intensity of local land use and cover change, the non-local impact term and control variables into the hierarchical statistical model, introduce a spatially correlated random field into the error term, perform parameter inference, output the contribution and confidence interval of local and non-local effects, and obtain the effect separation result.

Images