CN-122016676-A - MAX-DOAS-based aerosol and gas profile inversion method and system

CN122016676ACN 122016676 ACN122016676 ACN 122016676ACN-122016676-A

Abstract

The invention provides an aerosol and gas profile inversion method and system based on MAX-DOAS, wherein the method comprises a profile inversion step and a result monitoring step, wherein the profile inversion step is used for inverting a double-layer MCMC profile, the aerosol profile is inverted by using O 4 diagonal column concentration in the first layer, the probability inversion of gas concentration profile condition is carried out by using fixed aerosol information in the second layer, so as to obtain a profile inversion result, and the result monitoring step is used for providing a quality control index, monitoring inversion progress and anomaly diagnosis in real time so as to ensure the accuracy of the result. According to the invention, the aerosol resolution is improved from 200m to 100m, the trace gas resolution is improved to 50m, the vertical resolution and the error separation precision are improved, the Gaussian assumption limit of the traditional optimal estimation is broken through, the global convergence is ensured through self-adaptive adjustment and step optimization, and the high-precision inversion is realized.

Inventors

XIE PINHUA
YU ZIJUN
TIAN XIN
XU JIN
WANG ZIJIE

Assignees

中国科学院合肥物质科学研究院

Dates

Publication Date: 20260512
Application Date: 20251230

Claims (10)

1. An aerosol and gas profile inversion method based on MAX-DOAS is characterized by comprising the following steps: S1, inputting original MAX-DOAS spectrum data, obtaining O 4 inclined column concentration and trace gas inclined column concentration through spectrum fitting, constructing a Bayesian inversion framework, quantizing the coincidence degree of observation and a forward model through a likelihood function, inverting posterior probability distribution of an aerosol extinction profile by using an MCMC sampling algorithm, and extracting posterior mean value from the posterior probability distribution of the aerosol extinction profile to be used as an optimal aerosol extinction profile; S2, fixing the optimal aerosol extinction profile obtained in the step S1, combining the concentration of the trace gas diagonal column, inverting the posterior probability distribution of the gas concentration profile through a conditional likelihood function, gas concentration prior probability distribution and an MCMC sampling algorithm, and extracting a posterior mean value from the posterior probability distribution of the trace gas concentration profile to serve as the optimal trace gas concentration profile; And S3, using a Bayesian error decomposition algorithm, combining the posterior probability distribution of the aerosol extinction profile obtained in the step S1 and the posterior probability distribution of the trace gas concentration profile obtained in the step S2, realizing mathematical strict separation of measurement errors, smoothing errors and total errors, monitoring inversion progress in real time and performing anomaly diagnosis.
2. The method for MAX-DOAS based aerosol and gas profile inversion of claim 1, wherein the main flow of the MCMC sampling algorithm used in steps S1 and S2 is: inputting the concentration of the inclined column and the prior profile into a Bayes inversion framework, and constructing posterior probability distribution; Taking posterior probability as target distribution, randomly walking in a parameter space through an MCMC sampling algorithm, calling a radiation transmission model to calculate a forward simulation value, determining to accept or reject a new state according to the coincidence degree of simulation and observation, and obtaining a posterior sample set after iterative convergence; and extracting posterior mean value, standard deviation and confidence interval through statistical analysis of the posterior sample set to obtain an inversion profile.
3. The method for inverting the aerosol and the gas profile based on MAX-DOAS as recited in claim 2, wherein the specific flow for obtaining the inversion profile by taking posterior probability as target distribution, randomly walking in a parameter space through an MCMC sampling algorithm, calling a radiation transmission model to calculate a forward simulation value, determining to accept or reject a new state according to the coincidence degree of simulation and observation, and obtaining a posterior sample set after iterative convergence, wherein the specific flow for obtaining the inversion profile by statistically analyzing the posterior sample set to extract posterior mean value, standard deviation and confidence interval is as follows: a plurality of walkers are initially set, an initial state is generated by random disturbance near a first-pass average value, and the number of sampling steps and the number of burn-in steps are set; Performing stretching movement on each walker, randomly selecting another walker and proposing a new state, and calling a radiation transmission model to calculate a likelihood function; Calculating posterior probability and acceptance probability through likelihood function and prior probability distribution, generating random numbers between 0 and 1, judging the relation between the random numbers and the acceptance probability, and repeatedly calculating according to the set sampling steps until convergence; And reserving a converged posterior sample, calculating posterior statistics as an inversion profile, taking standard deviation as uncertainty, and extracting a quantile as a confidence interval.
4. The method for inverting aerosol and gas profiles based on MAX-DOAS as set forth in claim 1, wherein the MCMC sampling algorithm of step S1 and step S2 constructs a two-layer MCMC inversion engine, the posterior probability distribution of the aerosol extinction profile is obtained through the MCMC sampling algorithm of the first layer, and the posterior probability distribution of the trace gas concentration profile is obtained through the MCMC sampling algorithm of the second layer.
5. The method for MAX-DOAS based aerosol and gas profile inversion of claim 4, wherein the step of performing the dual layer MCMC inversion engine comprises: Inputting O 4 diagonal column concentration and an aerosol extinction prior profile into a first layer MCMC sampling algorithm of a double-layer MCMC inversion engine, constructing a Bayesian inversion framework, obtaining posterior probability distribution of the aerosol extinction profile through the coincidence degree of likelihood function quantitative observation and a forward model, and extracting posterior mean value from the posterior probability distribution of the aerosol extinction profile to serve as an optimal aerosol extinction profile; And then, taking the optimal aerosol extinction profile obtained in the first layer as a fixed parameter, combining trace gas diagonal concentration data to input the optimal aerosol extinction profile into a second-layer MCMC sampling algorithm, obtaining posterior probability distribution of the gas concentration profile through the MCMC sampling algorithm based on a Bayesian theorem and combining a conditional likelihood function and gas concentration prior probability distribution, performing burn-in processing and convergence diagnosis on an MCMC sampling chain, calculating posterior mean value of the posterior probability distribution of the gas concentration profile as the optimal trace gas concentration profile, and obtaining aerosol and gas vertical profile through vertical integration.
6. The method for MAX-DOAS based aerosol and gas profile inversion of claim 5, wherein the first layer MCMC sampling algorithm comprises: Inputting the O 4 diagonal concentration and the aerosol extinction prior profile into a forward modeling module, and calculating weight functions of each layer of the air quality factor and the aerosol extinction profile through a SCIATRAN radiation transmission model; based on the calculated weight function and the aerosol prior probability distribution, constructing a posterior probability distribution function of the aerosol extinction profile according to the Bayesian theorem, adopting a Metropolis-Hastings algorithm to randomly sample, extracting samples from the posterior probability distribution of the aerosol extinction profile through an acceptance-rejection criterion, and obtaining a posterior sample set of the aerosol extinction profile after burn-in processing and Gelman-Rubin convergence diagnosis; And calculating posterior mean value from a posterior sample set of the aerosol extinction profile as optimal aerosol extinction profile estimation, calculating posterior covariance matrix quantization uncertainty, substituting the optimal aerosol extinction profile into SCIATRAN radiation transmission model to verify consistency of simulated O 4 diagonal concentration and observation, and outputting the optimal aerosol extinction profile and complete uncertainty quantization result thereof.
7. The method for MAX-DOAS based aerosol and gas profile inversion of claim 5, wherein the second layer MCMC sampling algorithm comprises: the optimal aerosol extinction profile obtained in the first layer is used as a fixed parameter, trace gas diagonal column concentration, the fixed aerosol extinction profile and trace gas priori profile are input into a forward modeling module, and a sensitivity weight function of the trace gas diagonal column concentration to the gas concentration of each layer is calculated through a SCIATRAN radiation transmission model; Based on a sensitivity weight function and a trace gas concentration priori profile construction condition posterior probability distribution, sampling is carried out by adopting a Stretch Move set sampling algorithm and combining with Metropolis-Hastings criteria, and a posterior sample set of the trace gas concentration profile is obtained through collaborative optimization of a plurality of sampling chains; And extracting posterior mean value from posterior samples of the trace gas concentration profile as an optimal trace gas concentration profile, calculating posterior covariance quantization uncertainty, vertically integrating the optimal trace gas concentration profile with the atmospheric density profile, calculating trace gas vertical column concentration, and outputting the trace gas concentration profile and an uncertainty quantization result thereof.
8. The MAX-DOAS claimed in claim 1, wherein the posterior probability distribution of the aerosol extinction profile and the optimal aerosol extinction profile inversion process comprises: Constructing a Bayesian inversion framework: Wherein y is an observation vector, and is the O 4 diagonal concentration obtained by MAX-DOAS observation, and x is a state vector, and is the inversion profile which is finally needed to be obtained; The posterior probability distribution of the extinction profile of the aerosol is represented by the probability distribution of x obtained after observation of the observation vector y; representing the probability of observing the observation vector y given the true state vector x as a likelihood function; The prior probability distribution is a state vector x obtained by speculating according to the prior experience; Is a normalization constant; according to the above formula, the likelihood function The Gaussian distribution form is adopted: Wherein, the Mapping the state vector x to the observation space, using SCIATRAN radiation transmission models in the above formula; In order to observe the data error covariance, The error covariance inverse matrix is a weight matrix of residual errors, and the weight is smaller when the error is larger; Constructing likelihood functions according to the above formula Log form of (c): Wherein, the For the aerosol observations at the i-th elevation angle, For the model observations corresponding to the elevation angle, Is an observation error; According to the Bayesian inversion framework, the prior probability distribution constructed by adopting the multi-element Gaussian distribution: Wherein the prior mean value Indicating that a typical aerosol extinction profile is present, Representing uncertainty and vertical correlation of the profile for a priori covariance matrix; according to the algorithm described above, the logarithmic form of the posterior probability distribution of the aerosol extinction profile is calculated: Wherein the method comprises the steps of Is a weighted average sum of aerosol observation residuals. According to the algorithm, the posterior mean value of the posterior probability distribution of the aerosol extinction profile is calculated: Wherein, the The posterior mean value of the aerosol extinction profile is used for representing the optimal aerosol extinction profile, and N is the total number of steps of sampling the first layer of MCMC.
9. The method for MAX-DOAS based aerosol and gas profile inversion of claim 8, wherein said posterior probability distribution of gas concentration profile and optimal trace gas concentration profile inversion comprises: calculating posterior probability distribution of the gas concentration profile: Wherein, the To obtain the gas profile distribution that eventually needs to be inverted, For the posterior mean value of the aerosol extinction profile obtained in the inversion step, In order to invert the resulting gas profile to be needed, To observe the resulting concentration of trace gas in the diagonal column, Is a function of the likelihood probability of the gas, Is the prior probability distribution of the gas; according to the algorithm, the posterior mean value of the posterior probability distribution of the trace gas concentration profile is calculated: Wherein, the The posterior mean value of the trace gas concentration profile is expressed as the optimal trace gas concentration profile, and M is the total number of steps of sampling the second layer MCMC.
10. An aerosol and gas profile inversion system based on MAX-DOAS, which is characterized by comprising: The posterior probability distribution inversion module is used for inputting original MAX-DOAS spectrum data, obtaining O 4 inclined column concentration and trace gas inclined column concentration through spectrum fitting, constructing a Bayes inversion framework, quantizing the coincidence degree of observation and a forward model through a likelihood function, inverting the posterior probability distribution of the aerosol extinction profile by using an MCMC sampling algorithm, and extracting posterior mean value from the posterior probability distribution of the aerosol extinction profile as an optimal aerosol extinction profile; The posterior probability distribution inversion module is used for fixing the optimal aerosol extinction profile obtained by the module, combining trace gas diagonal column concentration, inverting the posterior probability distribution of the gas concentration profile through a conditional likelihood function, gas concentration prior probability distribution and an MCMC sampling algorithm, and extracting posterior mean value from the posterior probability distribution of the trace gas concentration profile to serve as the optimal trace gas concentration profile; And the monitoring module is used for realizing mathematical strict separation of measurement errors, smooth errors and total errors by using a Bayesian error decomposition algorithm and combining the obtained posterior probability distribution of the aerosol extinction profile and the obtained posterior probability distribution of the trace gas concentration profile, and monitoring inversion progress in real time and carrying out abnormality diagnosis.

Description

MAX-DOAS-based aerosol and gas profile inversion method and system Technical Field The invention relates to the technical field of environmental monitoring, in particular to an aerosol and gas profile inversion method and system based on MAX-DOAS. Background At the present time of rapid economic development, industrialized and urban human activities have a bad influence on the environment. Important components of atmospheric pollutants are aerosols and trace gases, which play an important role in the physical process of the atmosphere. MAX-DOAS (Multiaxis differential absorption Spectrometry) is a foundation remote sensing technology, which obtains the concentration information of the inclined column of aerosol and trace gas in the atmosphere by measuring the solar scattered light with different angles, and then matches with a vertical distribution inversion algorithm to realize the profile inversion of the gas, so as to detect the atmospheric pollutants. The existing MAX-DOAS-based inversion algorithm is mainly divided into a table look-up method and an optimal estimation method, and has obvious defects: 1. The vertical resolution of the existing MAX-DOAS inversion algorithm is generally lower and is 200m, so that the near-ground atmosphere layering is insufficient, the urban scale pollution monitoring and fine research requirements cannot be met, and particularly, complex structures inside a boundary layer cannot be resolved. Meanwhile, due to lower vertical resolution, the smoothing error accounts for more than 70% of inversion errors, and the accuracy of an inversion algorithm is greatly limited. 2. The existing MAX-DOAS aerosol and gas profile inversion algorithm based on the optimal estimation method basically uses the method of carrying out iterative solution on the value function based on the Gauss Newton method, but has the following serious defects: 1) The local optimal solution is calculated by using a gradient descent method when the optimal solution is solved by the optimal estimation method, singular values are generated when the atmosphere layering number is large, the local optimal solution is trapped, the global optimal solution cannot be calculated continuously, the convergence result of the gradient method is strongly dependent on an initial guess value, different initial values can lead to completely different inversion results, and objective initial value selection criteria are lacked. 2) The optimal estimation method is developed by using a Taylor formula, the forward model is assumed to be linearized near a priori point, the atmospheric radiation transmission process has strong nonlinear characteristics, when the linearization error is large, the Gaussian Newton method iteration process is easy to diverge or false convergence, the traditional convergence criterion (the gradient modulus is smaller than a threshold value) is caused to fail in a nonlinear system, and the algorithm can stop at the saddle point or oscillate among a plurality of local extremums. 3) Incomplete uncertainty estimation is that the prior gas inversion error analysis assumes Gaussian distribution of posterior distribution, but actual posterior distribution often has non-Gaussian characteristics such as multimodal, deflection, thick tail and the like, which causes serious distortion of confidence intervals, actual coverage rate is far lower than a calculation result, a main source of errors cannot be accurately identified, and error prediction results under different observation conditions are unreliable. Patent document with publication number CN113834792 proposes a 50 m resolution trace gas profile inversion method based on MAX-DOAS, which obtains an aerosol profile through first inversion calculation of a Monte Carlo sampling algorithm, obtains a trace gas profile through second inversion calculation of a Monte Carlo sampling algorithm of an aerosol profile combined gas diagonal concentration and a gas prior profile, and adopts the Monte Carlo method to replace Gaussian Newton iterative solution. However, a Monte Carlo random sampling method is adopted, the optimal solution is found by randomly generating a weight factor alpha in a range from zero to one, calculating a cost function for each alpha value and arranging the values, the search space is a one-dimensional scalar parameter, and based on the preset sampling times, no explicit convergence diagnosis standard exists, so that a certain inversion error still exists. Disclosure of Invention The invention aims to solve the technical problem of how to reduce inversion errors of inversion of aerosol and trace gas profiles based on an MAX-DOAS inversion algorithm. In order to solve the technical problems, the invention adopts the following technical scheme that the method for inverting the aerosol and the gas profile based on MAX-DOAS comprises the following steps: S1, inputting original MAX-DOAS spectrum data, obtaining O 4 inclined column concentration and tra