A global temperature change comprehensive evaluation method and system, device and medium based on multi-dimensional time series analysis and pattern fusion

Abstract

The present application relates to the technical field of earth observation and climate change analysis, and particularly relates to a global temperature change comprehensive evaluation method and system based on multi-dimensional time series analysis and model fusion, equipment and medium, adopting the scheme provided by the present application, mainly including obtaining the ground temperature data of different times of the current pixel, obtaining target indexes based on the ground temperature data, the target indexes including average temperature, maximum temperature, minimum temperature and temperature difference, and extracting multi-dimensional features of the target indexes, innovatively integrating three key dimensions of trend direction and rate (linear analysis), dominance (trend signal-to-noise ratio analysis) and dynamic characteristics (nonlinear mutation analysis). All classification criteria are based on data-driven statistics, not subjective thresholds, ensuring the objectivity and repeatability of classification.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of earth observation and climate change analysis, and in particular, relates to a global temperature change comprehensive evaluation method and system based on multi-dimensional time series analysis and model fusion, equipment and medium. BACKGROUND

[0002] At present, accurate, comprehensive and fine monitoring and evaluation of global land surface temperature change in the past decades is the scientific basis and key prerequisite for deeply understanding the evolution law of climate system, predicting future climate change trend and formulating effective global and regional climate change adaptation and mitigation strategies.

[0003] With the development of satellite remote sensing technology and climate data reanalysis system, long-term, global land and sea surface, spatiotemporal continuous land surface temperature data sets can be obtained, but the existing technology has the following problems:

[0004] Limitation of linear hypothesis: the existing method generally assumes that temperature change is linear and stationary. However, the real climate system change process is complex and nonlinear, which may contain acceleration, deceleration, mutation or trend reversal. Simple linear trend analysis will average these important dynamic details, which may lead to misjudgment of the change rate and ignore the key turning points in the change process.

[0005] Single information dimension: the existing technology mainly focuses on the change direction and average rate (i.e. linear trend), and ignores the other two important dimensions of the change process: one is the volatility of the change (i.e. the strength contrast of long-term trend to interannual fluctuation), and the other is the nonlinear dynamic characteristics of the change (such as whether there is a mutation). This leads to one-sidedness and incompleteness of the cognition of temperature change.

[0006] Coarse classification system: the change classification based on the existing technology is usually limited to a few categories such as significant warming, significant cooling and no significant change, and the classification system is too coarse. In addition, the information of average temperature, maximum / minimum temperature and diurnal temperature difference cannot be effectively fused, which cannot finely describe the specific mode of temperature change in different regions and its diurnal difference, such as asymmetric change of night warming faster than day. SUMMARY

[0007] The purpose of the present application is to provide a global temperature change comprehensive evaluation method and system based on multi-dimensional time series analysis and model fusion, equipment and medium to solve the above problems in the prior art.

[0008] The present application is realized by the following technical scheme:

[0009] In a first aspect, the present application provides a global temperature change comprehensive evaluation method based on multi-dimensional time series analysis and pattern fusion, comprising:

[0010] Obtaining ground temperature data of the current pixel at different times, obtaining target indicators based on the ground temperature data, the target indicators including average temperature, maximum temperature, minimum temperature and temperature difference, and extracting multi-dimensional features of the target indicators;

[0011] Establishing a plurality of temperature change patterns, setting the judgment range of the temperature change patterns with respect to a plurality of multi-dimensional features, and establishing a feature reference table based on the temperature change patterns and the judgment range of the corresponding multi-dimensional features;

[0012] Obtaining a plurality of target multi-dimensional features of the target indicators of the current pixel, indexing the target multi-dimensional features in the feature reference table to obtain a plurality of temperature change patterns of different target indicators of the current pixel;

[0013] Based on the plurality of temperature change patterns of the target indicators, outputting the final evaluation result.

[0014] Preferably, the multi-dimensional feature extraction of the target indicators comprises:

[0015] Performing linear trend analysis, volatility analysis and nonlinear change detection analysis on each target indicator to obtain overall slope, significance p value, trend signal-to-noise ratio, breakpoint number, breakpoint probability, breakpoint slope before and after, and mean jump amplitude.

[0016] Preferably, the linear trend analysis comprises:

[0017] Obtaining a time series composed of data at different times of a certain target indicator, calculating the slope between any two different time points in the time series;

[0018] Assembling all the slopes into a slope set, sorting the slope set in ascending order, obtaining the median of the sorted slope set as the overall slope;

[0019] Using the non-parametric Mann-Kendall rank test method to evaluate the statistical significance of the monotonic trend without assuming that the time series data follows a specific distribution, and obtaining the significance p value.

[0020] Preferably, the volatility analysis comprises:

[0021] Using the STL decomposition method to decompose the time series of each temperature indicator of the current pixel into three independent components: trend item, seasonal item and residual item;

[0022] The standard deviation of the residual term and the accumulated change of the trend term are obtained, and the trend signal-to-noise ratio is calculated based on the standard deviation and the accumulated change.

[0023] Preferably, the nonlinear change detection analysis includes:

[0024] A Bayesian ensemble algorithm Rbeast is used to detect structural mutation points of each time series, the time series is modeled as a segmented function, and the connection point between each segmented function is a mutation point, and the number of breakpoints, breakpoint probability and slope before and after the breakpoint are obtained.

[0025] Preferably, the output of the final evaluation result based on the temperature change mode of the target index includes:

[0026] The number of temperature change modes belonging to the cooling type and the heating type based on the target index is obtained to obtain the temperature change trend of the current pixel.

[0027] A plurality of different priorities of the influence degree of the temperature change mode on the climate system are set, and the priority of the current pixel is obtained based on the temperature change mode of the target index of the current pixel.

[0028] The final judgment result is output based on the temperature change trend and the priority of the current pixel.

[0029] Preferably, the final judgment result includes fluctuation dominance, stable heating, stable cooling, acceleration / deceleration heating, acceleration / deceleration cooling, jump heating, jump cooling, trend reversal, and day / night trend reversal.

[0030] In a second aspect, the present application also provides a global temperature change comprehensive evaluation system based on multi-dimensional time series analysis and pattern fusion, which is used to execute the global temperature change comprehensive evaluation method based on multi-dimensional time series analysis and pattern fusion described above, and includes:

[0031] A data acquisition module is configured to acquire ground temperature data of a current pixel at different times, acquire target indexes based on the ground temperature data, the target indexes including average temperature, maximum temperature, minimum temperature and temperature difference, and extract multi-dimensional features of the target indexes;

[0032] A comparison module is configured to establish a plurality of temperature change modes, set a judgment range of the temperature change mode with respect to a plurality of multi-dimensional features, establish a feature comparison table based on the temperature change mode and the judgment range of the corresponding multi-dimensional feature, acquire a plurality of target multi-dimensional features of the target indexes of the current pixel, index the target multi-dimensional features in the feature comparison table to obtain a plurality of temperature change modes of different target indexes of the current pixel.

[0033] An output module is configured to output a final evaluation result based on a plurality of temperature change modes of the target index.

[0034] In a third aspect, the present application also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned global temperature change comprehensive evaluation method based on multi-dimensional time series analysis and pattern fusion when executing the computer program.

[0035] In a fourth aspect, the present application also provides a computer readable storage medium, wherein a computer program is stored in the computer readable storage medium, and the computer program implements the above-mentioned global temperature change comprehensive evaluation method based on multi-dimensional time series analysis and pattern fusion when executed by a processor.

[0036] The technical scheme of the present application has at least the following advantages and beneficial effects:

[0037] By using the scheme provided by the present application, the current pixel different time surface temperature data is obtained, the target index is obtained based on the surface temperature data, the target index includes the average temperature, the highest temperature, the lowest temperature and the temperature difference, and the multi-dimensional feature of the target index is extracted, and the trend direction and rate (linear analysis), the dominance (trend signal-to-noise ratio analysis) and the kinetic characteristics (nonlinear mutation analysis) three key dimensions are innovatively integrated. All classification criteria are based on data-driven statistics, not subjective thresholds, which ensures the objectivity and repeatability of classification.

[0038] Refinement of the classification system: a refined classification system including 14 single-index modes and 9 comprehensive modes is proposed, which can clearly distinguish different dynamic processes such as "stable", "accelerate" and "mutation", far beyond the rough classification of the prior art, greatly improving the depth of cognition of temperature change process.

[0039] Intelligent fusion of multi-index information: the "majority voting + priority" fusion rule used in the present application is the core advantage of the present application. It skillfully balances the robustness and sensitivity of the evaluation results. On the one hand, through majority voting, the overall direction of regional temperature change is grasped, avoiding the interference of single index noise; on the other hand, through priority setting, it can sensitively capture the most severe, most scientific and risk alerting dynamic events (such as mutation or reversal) occurring on any key index, thereby providing a more comprehensive evaluation close to the complexity of the real climate system.

[0040] The finally output fine and comprehensive global temperature change pattern map of the application carries three-dimensional core information of change direction, dynamics and diurnal asymmetry in 9 categories, which is intuitive and easy to understand. This not only helps to deepen the scientific understanding of climate change facts, processes and mechanisms, but also provides strong spatiotemporal data and methodological support for climate change impact assessment, regional adaptability strategy formulation and related environmental and ecological policy scientific decision-making. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some of the embodiments of the application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor.

[0042] Fig. 1 It is a schematic diagram of the overall process of the application.

[0043] Fig. 2 It is a feature comparison table about temperature change pattern of the application. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solutions and advantages of the embodiments of the application more clear, the following will combine the drawings in the embodiments of the application to clearly and completely describe the technical solutions in the embodiments of the application. Obviously, the described embodiments are part of the embodiments of the application, not all. The components of the embodiments of the application described and shown in the drawings here can be arranged and designed in various different configurations.

[0045] The division of modules appearing in the application is a logical division, and in actual application, there can be another division mode, for example, multiple modules can be combined or integrated in another system, or some features can be ignored or not executed.

[0046] The independently described modules or sub-modules can be physically separated or not physically separated: can be software implemented, or hardware implemented, and part of the modules or sub-modules can be implemented by software, the function of the part of the modules or sub-modules is called by the processor, and the other part of the modules or sub-modules is implemented by hardware, for example, by hardware circuit. In addition, part or all of the modules can be selected to achieve the purpose of the application scheme according to actual needs.

[0047] Please refer to Figs. 1-2 The application provides a global temperature change comprehensive evaluation method based on multi-dimensional time series analysis and pattern fusion, comprising:

[0048] S101: Obtain the ground surface temperature data of the current pixel at different times, obtain the target indicators based on the ground surface temperature data, the target indicators include average temperature, maximum temperature, minimum temperature and temperature difference, and extract the multi-dimensional features of the target indicators;

[0049] Obtain the ERA5 reanalysis ground surface 2-meter air temperature dataset from 2003 to 2024 globally, which is provided by the European Centre for Medium-Range Weather Forecasts (ECMWF). Extract the monthly time series of four key indicators: average temperature (Tmean), maximum temperature (Tmax), and minimum temperature (Tmin). The data is global grid data with uniform spatial resolution of 0.25°x0.25°. Based on the extracted Tmax and Tmin, further calculate the derived indicator temperature difference Trg (Trg = Tmax - Tmin).

[0050] Area weighting: Considering that the actual geographical area represented by the earth grid is different at different latitudes, when performing any global or regional average calculation, the contribution of each pixel must be cosine latitude weighted to ensure the accuracy of the results. The specific steps are as follows:

[0051] For any pixel i in the study area, obtain the latitude of its center point (unit: degree).

[0052] Calculate the weight of this pixel To ensure that the weight is positive, the latitude needs to be converted to radian system for calculation:

[0053]

[0054] Suppose we want to calculate the regional average temperature at a certain time , the region contains N pixels, and the temperature value of each pixel is . The calculation formula is:

[0055]

[0056] S102: Establish several temperature change patterns, set the judgment range of the temperature change patterns on several multi-dimensional features, and establish a feature comparison table based on the temperature change patterns and the judgment range of the corresponding multi-dimensional features;

[0057] S103: Obtain several target multi-dimensional features of the target indicators of the current pixel element, index the target multi-dimensional features in the feature comparison table, and obtain several temperature change patterns of different target indicators of the current pixel;

[0058] S104: Based on the several temperature change patterns of the target indicators, output the final evaluation result.

[0059] The scheme provided by the application mainly includes obtaining surface temperature data of a current pixel at different times, obtaining target indicators based on the surface temperature data, the target indicators including average temperature, maximum temperature, minimum temperature and temperature difference, and extracting multi-dimensional features of the target indicators, which innovatively integrates three key dimensions of direction and rate of trend (linear analysis), dominance (trend signal-to-noise ratio analysis) and kinetic characteristics (nonlinear mutation analysis). All classification criteria are based on data-driven statistics, not subjective thresholds, ensuring the objectivity and repeatability of classification.

[0060] In an example embodiment of the application, the extraction of multi-dimensional features of the target indicators includes:

[0061] S201: performing linear trend analysis, volatility analysis and nonlinear change detection analysis on each target indicator, respectively obtaining overall slope, significance p value, trend signal-to-noise ratio, breakpoint number, breakpoint probability, slope before and after breakpoint, and mean jump amplitude.

[0062] Specifically, the overall trend slope of the time series is estimated by the non-parametric Theil-Sen method (unit: K / decade). This method determines the overall trend by calculating the median of all slopes between pairs of data points, so it is very robust to individual outliers in the time series. The specific steps are as follows: the linear trend analysis includes:

[0063] S202: obtaining a time series composed of data at different times for a certain target indicator, calculating the slope between any two different time points in the time series;

[0064] Given a time series data of length n , , where j=1,2,…,n. Calculate the slope between any two different time points , and , (where j>k): :

[0065] In the formula, and are the jth time point and the corresponding temperature, respectively, and are the jth time point and the corresponding temperature, respectively.

[0066] S203: form a slope set by all slopes, sort the slope set in ascending order, obtain the median of the sorted slope set as the overall slope ;

[0067] S204: all calculated slope values are grouped into a slope set, and the slope set is sorted in ascending order. The Theil-Sen slope estimate is the median of the sorted set.

[0068]

[0069] The present application adopts a non-parametric Mann-Kendall (MK) rank test method to evaluate the statistical significance of the monotonic trend of the time series data without assuming that the time series data conforms to a specific distribution. The core of the method is to calculate an S statistic, which represents the difference between the number of positive and negative differences of all data points in the time series. The calculation formula is:

[0070]

[0071] where n is the length of the time series, and are the values of the time series at times j and k, and sgn is the sign function. Then, the S statistic is standardized to the MK test statistic that conforms to the standard normal distribution, and the p value is calculated accordingly. The calculated p value is compared with the preset significance level α (set to 0.05 in the present application). If p < 0.05, it is determined that the trend of the time series is statistically significant.

[0072] In an example embodiment of the present application, the volatility analysis includes:

[0073] The STL decomposition method is used to decompose the time series of each temperature index of the current pixel into three independent components: a trend term, a seasonal term, and a residual term;

[0074] The present application uses the STL decomposition method to robustly decompose the temperature index time series (Yt) of each pixel into three independent components: a long-term trend term (Tt), a periodic seasonal term (St), and an irregular residual term (Rt). STL is a non-parametric method based on local weighted regression (Loess), which can effectively separate different components in the time series through iterative calculation.

[0075]

[0076] where represents the temperature index time series, the component represents the smooth, long-term change direction of the time series over the entire study period; the component represents the periodic fluctuation pattern within a year;​​ The component represents the random or irregular fluctuations remaining after removing the trend and seasonal effects.

[0077] The standard deviation of the residual term and the cumulative change of the trend term are obtained, and the trend signal-to-noise ratio is calculated based on the standard deviation and the cumulative change.

[0078] The trend term Tt and the residual term Rt obtained by the STL decomposition are mainly used in the present application. Among them, Tt is used to quantify the cumulative change of the trend, and the standard deviation of Rt is used to quantify the random fluctuation level between years, which are jointly used as inputs for subsequent calculation of the trend signal-to-noise ratio.

[0079] The cumulative change of the trend term throughout the study period is calculated as |ΔT|=|Tend - Tstart|, where Tend represents the value of the trend term time series Tt at the end of the entire study period, and Tstart represents the value of the trend term time series Tt at the beginning of the entire study period.

[0080] The standard deviation of the residual term Rt is calculated as , which represents the random fluctuation or noise level between years.

[0081] ④ Define and calculate the trend signal-to-noise ratio r = |ΔT| / . The value of r quantifies the strength of the long-term trend signal relative to the short-term fluctuations.

[0082] ⑤ Based on the global statistical distribution of all pixel r values, determine their 33% and 67% quantiles (r33% and r67%) as data-driven thresholds to divide the pixels into three categories: "trend dominant" (r>r67%), "trend and fluctuations comparable" (r33% ≤ r ≤r67%), and "fluctuation dominant" (r

[0083] In an example embodiment of the present application, a Bayesian ensemble algorithm Rbeast is used to detect structural change points for each time series. Rbeast is based on Bayesian theory and reversible jump Markov chain Monte Carlo (RJMCMC) method, which explores and averages among a large number of models containing different numbers and locations of change points. It models the time series as a piecewise function, where the connection points between segments are the change points, and the nonlinear change detection analysis includes:

[0084] A Bayesian ensemble algorithm Rbeast is used to detect structural change points for each time series, and the time series is modeled as a piecewise function, where the connection points between each piecewise function are change points, and the number of breakpoints, breakpoint probability and slope before and after the breakpoint are obtained.

[0085] A probability model Yt=Tt+St+ is constructed inside the algorithm where both the trend term Tt and the seasonality term St are allowed to be piecewise. The unknown parameters in the model include: the number of breakpoints bp, the location of each breakpoint , , and the trend slope and intercept within each piece.

[0086] The algorithm initiates a MCMC simulation process that can jump around the model space in different dimensions.

[0087] After sufficient sampling, the algorithm obtains the posterior distribution of all parameters. This is not a single answer, but a series of possible results with their corresponding probabilities.

[0088] The final detection results are a statistical summary of the posterior probability distribution:

[0089] a. The probability of the number of breakpoints (output bp): The algorithm gives the posterior probability of the number of breakpoints being 0, 1, 2, … The invention adopts the result with the highest probability, i.e., the number of breakpoints is 1, as the detection bp.

[0090] b. The probability curve of the breakpoint (output pb): The algorithm generates a curve showing the likelihood of a breakpoint occurring at each time point. The invention takes the peak probability of this curve as the posterior probability pb of the existence of a breakpoint.

[0091] c. The piecewise trend parameters (output , , ls): For the detected breakpoint, the algorithm gives the posterior distribution of the trend slope (i.e., , ) and intercept within the two pieces before and after the breakpoint. The mean jump amplitude ls is calculated by the difference between the intercepts of the two pieces at the breakpoint time.

[0092] Set strict screening criteria to determine significant breakpoints: (a) the number of detected breakpoints bp is 1; (b) the posterior probability pb of this breakpoint is greater than 0.5; (c) the change slope before the breakpoint is not equal to the change slope after the breakpoint.

[0093] For the detected significant breakpoint, further determine whether it contains a significant mean jump: if the mean jump amplitude |ls| exceeds 1.96 times the residual standard deviation (i.e., |ls|>1.96 × ), it is considered that a statistically significant mean jump or drop occurs at the breakpoint.

[0094] Extracted multi-dimensional features (bp, p-value, r, bp, pb, , , , ls), a change mode is independently assigned to each temperature index (Tmean, Tmax, Tmin, Trg) of each pixel. The present application presets 14 fine single-index change modes, and the determination rule is based on the logical decision tree of the above-mentioned feature combination, as shown in Fig. 2 .

[0095] In an example embodiment of the present application, this step is the core technical solution of the present application, which aims to fuse the 14 single-index change modes independently evaluated for the four temperature indexes (Tmean, Tmax, Tmin, Trg) of each pixel in the previous step into a final evaluation result that can comprehensively reflect the regional temperature change direction, process dynamics and diurnal asymmetry through a set of systematic decision rules. The fusion process specifically includes the following three key links.

[0096] Specifically, the output of the final evaluation result based on the same target index includes:

[0097] S301: Change direction determination (robust evaluation based on the majority voting principle)

[0098] This step aims to determine the overall direction of the pixel temperature change (warming, cooling or unclear trend) to obtain a robust macroscopic judgment. The present application adopts the majority voting principle. Specifically, the single-index classification results of the three core temperature indexes, i.e., the average temperature (Tmean), the maximum temperature (Tmax) and the minimum temperature (Tmin), are reviewed, and the specific steps are as follows:

[0099] Based on the number of temperature change modes belonging to the cooling class and the warming class, respectively, the temperature change trend of the current pixel is obtained;

[0100] For example, the number of single-index classification results belonging to the "warming class" (i.e., the categories 1, 4, 6, 10 and 12 preset in Fig. 2 ) is counted among the above-mentioned three indexes. If the number of "warming class" is greater than or equal to 2, it is determined that the overall change direction of the pixel is warming (Warming).

[0101] ②If the above condition is not met, further count the number of "cooling class" (i.e., categories 2, 5, 7, 11 and 13). If the number is greater than or equal to 2, it is determined that the overall change direction is cooling (Cooling).

[0102] ③All other cases (such as one warming and one cooling, or no obvious trend), are determined as mixed / without clear trend (Mixed).

[0103] S302: set several different priorities of the influence degree of temperature change mode on climate system, obtain the priority of the current pixel based on the temperature change mode of the target index of the current pixel;

[0104] This step aims to identify and determine the most significant dynamic characteristics in the temperature change process of the pixel, capturing those nonlinear, non-stationary change events that are of important indicative significance to the climate system. In the technical field to which the present invention belongs, the dynamics type refers to the process characteristics of temperature evolution over time, rather than just the net result of its change. It is used to distinguish between smooth, linear changes and complex changes containing features such as acceleration, abrupt change or turning point. Determining the dynamics type is of great significance because different dynamics types often correspond to different physical driving mechanisms or climate system feedback processes, for example, an abrupt change may be related to a sudden event or a system critical point, and acceleration may indicate the enhancement of positive feedback, which is crucial for risk assessment and attribution analysis of climate change. The present invention adopts the priority principle of any trigger. The "complex dynamics mode" here is specifically defined as all modes other than stable and fluctuation dominant among the 14 single-index modes, and is divided into four levels of dominant dynamics types according to the degree of influence on the climate system:

[0105] High priority - reversal class: categories 8 (temperature rise to fall) and 9 (temperature fall to rise).

[0106] Medium priority - abrupt class: categories 12 (jump type temperature rise) and 13 (jump type temperature fall).

[0107] Low priority - acceleration / deceleration class: categories 6 (accelerated temperature rise), 7 (accelerated temperature fall), 10 (decelerated temperature rise), and 11 (decelerated temperature fall).

[0108] Basic class - stable / fluctuation class: categories 1-5.

[0109] Specific implementation process: check whether any of the temperature change trends of the four indicators belongs to the reversal class. If so, the dominant dynamics type is determined to be reversal, and the subsequent check is terminated.

[0110] If there is no temperature change trend of the reversal class, then check whether any of the temperature change trends belongs to the medium priority abrupt class. If so, the dominant dynamics type of the pixel is immediately determined to be abrupt, and the subsequent check is terminated.

[0111] If neither of the first two is, then check whether any of them belongs to the acceleration / deceleration class. If so, the dominant dynamics type is determined to be accelerated / decelerated.

[0112] If none of the above complex dynamics patterns occur, the dominant dynamics type of the pixel is determined to be the lowest priority Stable / Fluctuation.

[0113] S303: Output the final judgment result based on the temperature change trend and priority of the current pixel.

[0114] This step uses the method of decision rule combination to combine the change direction and dominant dynamics type determined in the previous two steps, perform logical matching and conditional judgment, and finally output 9 types of refined comprehensive temperature change patterns. The specific steps are as follows:

[0115] Highest priority determination: If the dominant dynamics type determined in step S302 is trend reversal, the pixel is finally classified as trend reversal pattern, which is the highest priority rule and does not need to consider the change direction.

[0116] High priority combination: If the dominant dynamics type is mutation, then according to the change direction determined in step S301, it is subdivided as follows:

[0117] If the direction is warming, it is classified as jump warming;

[0118] If the direction is cooling, it is classified as jump cooling.

[0119] Second high priority combination: If the dominant dynamics type is acceleration / deceleration, then according to the change direction determined in step S301, it is subdivided as follows:

[0120] If the direction is warming, it is classified as acceleration / deceleration warming;

[0121] If the direction is cooling, it is classified as acceleration / deceleration cooling.

[0122] Basic type combination: If the dominant dynamics type is stable / fluctuation, then according to the change direction determined in step S301, it is subdivided as follows:

[0123] If the direction is warming, it is classified as stable warming;

[0124] If the direction is cooling, it is classified as stable cooling.

[0125] Specific condition determination (handle mixed / no clear trend cases):

[0126] If the direction is mixed / no clear trend and the dominant dynamics type is stable / fluctuation, a specific condition check is performed: whether the single index pattern of Tmax and Tmin presents a significant opposite trend (for example, one is stable warming and the other is stable cooling). If this condition is met, it is classified as day-night opposite trend.

[0127] All the remaining cases that do not satisfy any of the above rules (mainly the cases that the direction is mixed / no clear trend and no significant diurnal reverse trend) are finally classified as fluctuation dominant.

[0128] Output the comprehensive classification results: Based on the above steps, the pixels are finally classified into 9 categories of comprehensive temperature change patterns, which are: fluctuation dominant, stable warming, stable cooling, accelerated / decelerated warming, accelerated / decelerated cooling, jump warming, jump cooling, trend reversal, and diurnal trend opposite.

[0129] In a second aspect, the present application also provides a global temperature change comprehensive evaluation system based on multi-dimensional time series analysis and pattern fusion, which is used to execute the above-mentioned global temperature change comprehensive evaluation method based on multi-dimensional time series analysis and pattern fusion, and includes:

[0130] The data acquisition module is configured to acquire the ground temperature data of different times of the current pixel, acquire target indicators based on the ground temperature data, the target indicators including average temperature, maximum temperature, minimum temperature and temperature difference, and extract multi-dimensional features of the target indicators;

[0131] The comparison module is configured to establish a plurality of temperature change patterns, set the judgment range of the temperature change patterns on a plurality of multi-dimensional features, establish a feature comparison table based on the temperature change patterns and the judgment range of the corresponding multi-dimensional features, acquire a plurality of target multi-dimensional features of the target indicators of the current pixel, index the target multi-dimensional features in the feature comparison table, and obtain a plurality of temperature change patterns of different target indicators of the current pixel;

[0132] The output module is configured to output the final evaluation results based on the plurality of temperature change patterns of the target indicators.

[0133] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0134] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. The computer software product stored in a storage medium includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the embodiments of the method of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0135] The above merely describes the preferred embodiments of the present application and is not intended to limit the present application. The present application can have various modifications and changes for those skilled in the art. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A comprehensive assessment method for global temperature change based on multidimensional time series analysis and model fusion, characterized in that, include: The system acquires surface temperature data for the current pixel at different times, and obtains target indicators based on the surface temperature data. These target indicators include average temperature, maximum temperature, minimum temperature, and the difference between the maximum and minimum temperatures. Multidimensional features are then extracted from these target indicators. The extracted multidimensional features include: The overall slope and trend significance p-value are obtained through linear trend analysis, wherein the overall slope is calculated using the Theil-Sen estimator, which is the median of the slopes of all point pairs in the time series; The trend signal-to-noise ratio is obtained through volatility analysis, wherein the trend signal-to-noise ratio is obtained by decomposing the time series into a trend term and a residual term using the STL decomposition method, and calculating the ratio of the cumulative change of the trend term to the standard deviation of the residual term, and the threshold used to classify the signal-to-noise ratio level is dynamically determined based on the statistical quantile of this ratio in global pixels. The number of breakpoints, breakpoint probability, slope before and after the breakpoint, and mean jump amplitude are obtained through nonlinear change detection and analysis. The Bayesian ensemble algorithm Rbeast is used to detect mutation points, and a preset probability threshold is applied to the detected mutation points for screening. Establish several temperature change patterns, set the judgment range of the temperature change patterns with respect to several multidimensional features, and establish a feature comparison table based on the temperature change patterns and the corresponding judgment range of multidimensional features. Obtain several multidimensional features of the target indicators of the current image element, and index them in the feature lookup table based on the multidimensional features of the target elements to obtain several temperature change patterns of different target indicators of the current image element. Based on several temperature change patterns with the same target index, the final evaluation results are output, including: Determining the direction of change: Based on the pattern of three indicators, namely average temperature, maximum temperature and minimum temperature, the majority voting principle is used to determine the overall direction of change of the pixels as warming, cooling or a mixture. Dynamics type determination: Preset pattern categories with different priorities, scan in order of priority from high to low, when the pattern of any index belongs to a certain priority category, that priority is determined as the dominant dynamics type of the pixel; Comprehensive judgment: Combining the overall direction of change and the dominant dynamic type, the final judgment result is output. The judgment result includes fluctuation dominance, stable heating, stable cooling, accelerated / decelerated heating, accelerated / decelerated cooling, abrupt heating, abrupt cooling, trend reversal, and opposite day and night trends.

2. The comprehensive assessment method for global temperature change based on multidimensional time series analysis and model fusion as described in claim 1, characterized in that, The linear trend analysis includes: Obtain a time series of data for a target indicator, composed of data from different times, and calculate the slope between any two different time points in the time series; All slopes are combined into a slope set, and the slope set is sorted in ascending order. The median of the sorted slope set is used as the overall slope. The nonparametric Mann-Kendall rank test method is used to evaluate the statistical significance of the monotonicity of time series data without assuming that the time series data follows a specific distribution, and to obtain the significance p-value.

3. The comprehensive assessment method for global temperature change based on multidimensional time series analysis and model fusion as described in claim 2, characterized in that, The volatility analysis includes: The STL decomposition method is used to decompose the time series of each temperature index of the current pixel into three independent components: trend, seasonal and residual. Obtain the standard deviation of the residual term and the cumulative change of the trend term, and calculate the trend signal-to-noise ratio based on the standard deviation and the cumulative change.

4. The comprehensive assessment method for global temperature change based on multidimensional time series analysis and model fusion as described in claim 2, characterized in that, The nonlinear change detection and analysis includes: The Bayesian ensemble algorithm Rbeast is used to detect structural abrupt changes in each time series. The time series is modeled as a piecewise function, and the connection points between each piecewise function are abrupt changes. The number of abrupt changes, the probability of abrupt changes, and the slope before and after the abrupt changes are obtained.

5. A comprehensive global temperature change assessment system based on multidimensional time series analysis and model fusion, used to execute the comprehensive global temperature change assessment method based on multidimensional time series analysis and model fusion as described in any one of claims 1-4, characterized in that, include: The data acquisition module is configured to acquire surface temperature data of the current pixel at different times, acquire target indicators based on the surface temperature data, the target indicators include average temperature, maximum temperature, minimum temperature and the difference between the maximum and minimum temperatures, and extract multi-dimensional features from the target indicators. The comparison module is configured to establish several temperature change patterns, set the judgment range of the temperature change patterns with respect to several multidimensional features, establish a feature comparison table based on the temperature change patterns and the corresponding multidimensional feature judgment ranges, obtain several target multidimensional features of the target indicators of the current image element, index the target multidimensional features in the feature comparison table, and obtain several temperature change patterns of different target indicators of the current pixel. The output module is configured to output the final evaluation results based on several temperature change patterns based on the target index.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the comprehensive assessment method for global temperature change based on multidimensional time series analysis and model fusion as described in any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements a comprehensive assessment method for global temperature change based on multidimensional time-series analysis and model fusion as described in any one of claims 1-4.

Images