Search

CN-121996893-A - Self-adaptive localization method based on integrated ancient climate data assimilation frame

CN121996893ACN 121996893 ACN121996893 ACN 121996893ACN-121996893-A

Abstract

The invention discloses a self-adaptive localization method based on an aggregate ancient climate data assimilation frame, which comprises the steps of constructing an observation density field, updating the localization radius of a mode lattice point, calculating the weight of a substitute record according to a Gaspari-Cohn fifth-order polynomial if the substitute record is positioned in the localization radius, calculating the correlation information of the climate mode lattice point and the substitute record if the substitute record is positioned outside the localization radius, calculating the weight of the substitute record by using the correlation information if the correlation information meets a correlation threshold, otherwise, calculating the weight of the substitute record positioned outside the localization radius by using a Gaspari-Cohn fifth-order polynomial, combining the statistical correlation information in a reconstruction time period to obtain a final mixed weight matrix, and adjusting a covariance matrix by using the matrix. On the premise of ensuring that each mode lattice point is kept with a certain amount of observation influence, the false remote correlation is reduced by utilizing the statistical correlation, and the reconstruction effect is improved.

Inventors

  • LUO GE
  • ZENG YUEFEI

Assignees

  • 南京信息工程大学

Dates

Publication Date
20260508
Application Date
20260401

Claims (8)

  1. 1. The self-adaptive localization method based on the aggregate ancient climate data assimilation framework is characterized by comprising the following steps of: Step 1, for a research area, obtaining surrogate record data of a time period to be assimilated by data, dividing the time period to be assimilated by data into a plurality of assimilation windows, obtaining the observation density of each climate pattern lattice point under the assimilation window by a Gaussian kernel probability density function for each assimilation window, and carrying out normalization treatment to obtain the relative observation density of each climate pattern lattice point; Step2, setting upper and lower limits of the localization radius, and converting the relative observation density at each climate mode lattice point into the localization radius of each climate mode lattice point; Step 3, for each climate mode grid point, if the substitute record is located in the localization radius of the climate mode grid point, calculating the weight of the substitute record located in the localization radius according to a Gaspari-Cohn fifth-order polynomial; Step 4, if the substitute record is located outside the localized radius, calculating the correlation information of the weather pattern lattice points and the substitute record, judging whether the correlation information meets a correlation threshold, if so, calculating the weight of the substitute record located outside the localized radius by using the correlation information, otherwise, calculating the weight of the substitute record located outside the localized radius by using a Gaspari-Cohn fifth-order polynomial; And 5, integrating the surrogate record weights inside and outside the localized radius of the climatic mode pattern points to obtain a mixed weight matrix, and adjusting a covariance matrix in the set Kalman filtering method by using the mixed weight matrix to realize final data assimilation.
  2. 2. The adaptive localization method based on the aggregate archaic climate data assimilation framework of claim 1, wherein in step 1, based on the obtained surrogate record data, the observed density at each climate pattern lattice point under the assimilation window is estimated by a gaussian kernel probability density function: , Wherein, the Represent the first Individual climate pattern grid The observed density at which the position is observed, Represents the total number of surrogate recordings within the assimilation window, Represent the first Record of substitution pair The contribution of the individual climate pattern grid points, The bandwidth parameter is represented by a parameter of the bandwidth, Representing a gaussian kernel function, the expression is: , Wherein, the ; Using Scott rules, the expression is determined as: , Wherein, the Standard deviation representing surrogate recordings within an assimilation window; Thereby obtaining the observation density field of each assimilation window, normalizing the observation density field to obtain the relative observation density of each climate pattern lattice point under the assimilation window 。
  3. 3. The adaptive localization method based on the aggregate archaic climate data assimilation framework of claim 1, wherein in the step 2, the relative observed density at the climate pattern lattice point is converted into the localization radius of the climate pattern lattice point expressed as: , Wherein, the Represent the first Individual climate pattern grid Is defined by a localization radius of (c) in the drawing, Represent the first The relative observed densities at the individual climate pattern grid points, 、 Respectively represent a preset maximum value and a preset minimum value of the localization radius, Representing a curvature parameter controlling the density to radius mapping.
  4. 4. The adaptive localization method based on the aggregate archaic climate data assimilation framework of claim 1, wherein in the step 3, the surrogate recording weights within the localization radius for the pattern lattice are calculated by Gaspari-Cohn fifth-order polynomials: , Wherein, the A Gaspari-Cohn fifth order polynomial is represented, , Represent the first Individual climate pattern grid And the first The distance between the individual surrogate recordings is chosen, Represent the first Individual climate pattern grid Is included.
  5. 5. The adaptive localization method based on the aggregate archaic climate data assimilation framework of claim 4, wherein in the step 4, the correlation information of the climate pattern lattice point and the surrogate record is obtained by calculating Pearson correlation coefficient in the period of time to be assimilated with data, and the expression is: , Wherein, the Represent the first Under the assimilation window Individual climate pattern grid points and the first The correlation of the individual surrogate recordings, Representing the length of the period of time to be data assimilated, Represent the first Under the assimilation window The values of the individual climate pattern grid points, Represent the first Under the assimilation window The value of the individual surrogate record is recorded, And Respectively representing average values of all weather pattern grid points and substitute records in the period of time to be assimilated; , Wherein, the The representation is located at the first Individual climate pattern grid Is out of the localized radius of (2) Weights of the surrogate records.
  6. 6. The adaptive localization method based on the aggregate archaic climate data assimilation framework of claim 1, wherein in the step 5, the covariance matrix in the aggregate kalman filtering method is adjusted by using the mixed weight matrix, and the expression is: , Wherein, the A matrix of mixing weights is represented and, Representing the multiplication of the matrix, The covariance function is represented by a function of the covariance, Representing the background field of the investigation region, Representing the pattern priors obtained by the surrogate records after passing through the surrogate system model.
  7. 7. A computer device comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that the processor, when executing the computer program, implements the steps of the adaptive localization method based on an aggregate archaic data assimilation framework as claimed in any of claims 1 to 6.
  8. 8. A computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the steps of the adaptive localization method based on an aggregate archaic data assimilation framework of any of claims 1 to 6.

Description

Self-adaptive localization method based on integrated ancient climate data assimilation frame Technical Field The invention relates to a self-adaptive localization method based on an aggregate ancient climate data assimilation frame, and belongs to the technical field of data assimilation. Background Reconstructing the ancient climate state is a key link for understanding the evolution mechanism and influence of the earth system. At present, the field mainly depends on two methods, namely ancient climate substitute recording and earth system model simulation. However, both of these methods have significant limitations in reconstructing past climate conditions. On the one hand, surrogate recordings (such as tree wheels, ice cores, stalagmites, corals, etc.) play an indispensable role in reconstructing the climate for the past thousands of years due to their long time span, but they often face problems of uneven spatial-temporal distribution, discontinuities, sparsity, etc. Furthermore, climate signals are susceptible to noise contamination during storage, making it extremely challenging to establish quantitative relationships between surrogate recordings and true climate states, and reconstruction of the same climate state from different surrogate sources often shows inconsistencies. On the other hand, a climate model based on physical processes can provide a physically consistent, globally complete, climate field with full-time-space coverage, effectively capturing large-scale features of the climate system. However, model simulations may suffer from systematic bias due to inaccuracy in the parameterization of the physical process and uncertainty in the externally forced response, making it difficult to fully reproduce the observed climate change rate, resulting in limited predictive power. Therefore, to overcome the limitations of the single approach, ancient climate data assimilation techniques (PDAs) have been developed as powerful tools to optimally fuse information of surrogate recordings (sparsity, noise, indirection indicators of past climate) with the kinetic constraints of climate models. In this way, the PDA produces a spatially complete and physically consistent estimate of the climate field, similar to the re-analysis product of the organ age, while also quantifying the uncertainty of the reconstruction. Among the various approaches, ensemble kalman filtering (EnKF) is often used at PDAs because of its ability to easily implement and handle high-dimensional nonlinear systems. Despite its great potential, PDA faces unique challenges that are distinct from modern meteorological data assimilation, mainly because surrogate recordings are quite sparse in time and space and typically represent time-averaged climate signals (e.g., annual average) rather than instantaneous observations. In addition, surrogate data is subject to complex errors such as measurement errors, age uncertainty, imperfect relationships between surrogate signals and target climate variables, and the like. In PDAs, the observed error variance is typically specified empirically based on the residual variance of the proxy system model calibrated with measured data. However, this approach typically underestimates true proxy uncertainty because it fails to account for calibration errors (e.g., non-stationarity between past time periods and time periods), structural errors (arising from missing physical, biological, or chemical processes in the PSM), and representative errors (due to mismatch between point-scale proxy and grid-noded state variables). One key factor in the success of EnKF is the use of covariance localization. Early implementations typically employed a fixed localized radius. Under this framework, one common approach is to use a smooth, distance-dependent function to process the sample covariance. One standard choice for this function is the gaussian-like Gaspari-Cohn function. However, as several studies have shown, gaussian-like decay functions are not necessarily optimal. To address the limitations of fixed localization, a series of adaptive localization methods have been developed. These methods typically adjust localization based on state-dependent correlation patterns or aggregate estimation errors. Such as Anderson, bishop and Hodyss, and the techniques developed by mencarrier et al. However, these methods have not been applied to the surrogate recording assimilation background. In PDAs, large covariance localization radii are typically used due to sparse surrogate recordings distribution. Although necessary, an excessively large localization radius may introduce false long-distance correlations, ultimately degrading the quality of the analysis. This limitation suggests that fixed radius localization is not optimal for PDA applications. In contrast, adaptive localization is advantageous because it can optimize the radius of influence of surrogate recordings according to potential spatially dependent structures, a