CN117828992A - Accurate prediction method and system for CCN number concentration with high space-time resolution - Google Patents
Accurate prediction method and system for CCN number concentration with high space-time resolution Download PDFInfo
- Publication number
- CN117828992A CN117828992A CN202410012958.6A CN202410012958A CN117828992A CN 117828992 A CN117828992 A CN 117828992A CN 202410012958 A CN202410012958 A CN 202410012958A CN 117828992 A CN117828992 A CN 117828992A
- Authority
- CN
- China
- Prior art keywords
- ccn
- prediction
- number concentration
- factors
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000012216 screening Methods 0.000 claims abstract description 15
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 7
- 239000000126 substance Substances 0.000 claims description 18
- 239000000203 mixture Substances 0.000 claims description 11
- 238000007637 random forest analysis Methods 0.000 claims description 8
- 239000000356 contaminant Substances 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 5
- 238000012937 correction Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 230000002123 temporal effect Effects 0.000 claims 1
- 238000004088 simulation Methods 0.000 abstract description 14
- 230000003287 optical effect Effects 0.000 abstract description 5
- 230000007547 defect Effects 0.000 abstract description 2
- 239000002245 particle Substances 0.000 description 15
- 239000000443 aerosol Substances 0.000 description 11
- 238000010801 machine learning Methods 0.000 description 9
- RAHZWNYVWXNFOC-UHFFFAOYSA-N Sulphur dioxide Chemical compound O=S=O RAHZWNYVWXNFOC-UHFFFAOYSA-N 0.000 description 8
- 238000002790 cross-validation Methods 0.000 description 8
- 230000008859 change Effects 0.000 description 7
- 239000007789 gas Substances 0.000 description 7
- 230000000694 effects Effects 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 4
- 239000005427 atmospheric aerosol Substances 0.000 description 4
- 239000003344 environmental pollutant Substances 0.000 description 4
- 230000007246 mechanism Effects 0.000 description 4
- JCXJVPUVTGWSNB-UHFFFAOYSA-N nitrogen dioxide Inorganic materials O=[N]=O JCXJVPUVTGWSNB-UHFFFAOYSA-N 0.000 description 4
- 231100000719 pollutant Toxicity 0.000 description 4
- 230000005855 radiation Effects 0.000 description 4
- 229910002651 NO3 Inorganic materials 0.000 description 3
- NHNBFGGVMKEFGY-UHFFFAOYSA-N Nitrate Chemical compound [O-][N+]([O-])=O NHNBFGGVMKEFGY-UHFFFAOYSA-N 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 3
- 238000009833 condensation Methods 0.000 description 3
- 230000005494 condensation Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000001556 precipitation Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- MGWGWNFMUOTEHG-UHFFFAOYSA-N 4-(3,5-dimethylphenyl)-1,3-thiazol-2-amine Chemical compound CC1=CC(C)=CC(C=2N=C(N)SC=2)=C1 MGWGWNFMUOTEHG-UHFFFAOYSA-N 0.000 description 2
- 239000002028 Biomass Substances 0.000 description 2
- UGFAIRIUMAVXCW-UHFFFAOYSA-N Carbon monoxide Chemical compound [O+]#[C-] UGFAIRIUMAVXCW-UHFFFAOYSA-N 0.000 description 2
- RRHGJUQNOFWUDK-UHFFFAOYSA-N Isoprene Chemical compound CC(=C)C=C RRHGJUQNOFWUDK-UHFFFAOYSA-N 0.000 description 2
- CBENFWSGALASAD-UHFFFAOYSA-N Ozone Chemical compound [O-][O+]=O CBENFWSGALASAD-UHFFFAOYSA-N 0.000 description 2
- QAOWNCQODCNURD-UHFFFAOYSA-L Sulfate Chemical compound [O-]S([O-])(=O)=O QAOWNCQODCNURD-UHFFFAOYSA-L 0.000 description 2
- 150000003863 ammonium salts Chemical class 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 239000003738 black carbon Substances 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 229910002091 carbon monoxide Inorganic materials 0.000 description 2
- 238000003066 decision tree Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000006911 nucleation Effects 0.000 description 2
- 238000010899 nucleation Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- VEXZGXHMUGYJMC-UHFFFAOYSA-M Chloride anion Chemical compound [Cl-] VEXZGXHMUGYJMC-UHFFFAOYSA-M 0.000 description 1
- 229910002089 NOx Inorganic materials 0.000 description 1
- 235000017284 Pometia pinnata Nutrition 0.000 description 1
- 239000002253 acid Substances 0.000 description 1
- 230000003213 activating effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 229910002092 carbon dioxide Inorganic materials 0.000 description 1
- 150000001804 chlorine Chemical class 0.000 description 1
- 238000005345 coagulation Methods 0.000 description 1
- 230000015271 coagulation Effects 0.000 description 1
- 238000002485 combustion reaction Methods 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000002156 mixing Methods 0.000 description 1
- 229910000069 nitrogen hydride Inorganic materials 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 239000013618 particulate matter Substances 0.000 description 1
- 239000002243 precursor Substances 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 150000003254 radicals Chemical class 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000004062 sedimentation Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 230000002194 synthesizing effect Effects 0.000 description 1
- 150000003505 terpenes Chemical class 0.000 description 1
- 235000007586 terpenes Nutrition 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Chemical compound O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a precise prediction method and a precise prediction system for high-space-time resolution CCN number concentration, wherein the method comprises the following steps: collecting CCN number concentration samples affecting a region to be predicted by using a WRF-Chem mode; collecting characteristic factors of a region to be predicted by utilizing a multi-source big data network; respectively preprocessing a CCN number concentration sample and a characteristic factor; screening the processed characteristic factors to obtain prediction factors; and constructing a prediction model based on the prediction factor, and completing the prediction of the CCN number concentration by using the prediction model. The CCN prediction system constructed by the invention can optimize the mode simulation result, improves the prediction precision, and has higher concentration precision of the predicted CCN number compared with the traditional optical characteristic inversion algorithm. And meanwhile, the defect of the CCN observation in the space-time scale is also overcome. The method is helpful for accurately predicting the CCN number concentration, parameterizing the time-space characteristics of the CCN number concentration into the mode simulation, and accurately evaluating the influence of the CCN number concentration on weather and climate.
Description
Technical Field
The invention relates to the technical field of atmospheric measurement, in particular to a method and a system for accurately predicting CCN number concentration with high space-time resolution.
Background
The impact of atmospheric aerosols on cloud radiation forcing is the largest source of uncertainty in current climate change predictions. The main reason for this uncertainty is that atmospheric aerosols can act as nuclei for condensation (Cloud condensation nuclei, CCN). CCN refers to atmospheric aerosol particles that can be activated to form cloud droplets under supersaturated conditions. The interaction of CCN with water vapor affects the microscopic and macroscopic physics of the cloud, thereby affecting cloud formation, properties, dynamics and precipitation, further affecting the energy balance of the earth, thereby affecting climate and weather. Therefore, related studies of atmospheric aerosols and CCN have important scientific significance, and are important for improving the parameterization scheme of aerosol-cloud in climate mode to accurately evaluate the indirect climate effect of aerosol.
The CCN number concentration is related to the environmental supersaturation, and the larger the supersaturation, the larger the CCN number concentration, which in turn has more aerosol particles with cloud potential. The aerosol particle spectral distribution, chemical composition and mixing regime at a given supersaturation level play an important role in determining CCN activity. In order to quantify the effect of CCN on the earth's radiation budget, it is critical to obtain the concentration of CCN numbers on a space-time scale. But the characteristics of aerosols may exhibit space-time differences with the physicochemical processes of the particles in the atmosphere. In general, the closer to the region of the emission source, the higher concentration of atmospheric gaseous precursors, oxidized radicals, and primary emitted particles (such as biomass-derived primary organics and BC, etc.), will make various types of atmospheric chemical and physical processes (nucleation, coagulation growth, collision and growth of particles, and homogeneous and heterogeneous reactions, etc.) more intense and rapid. These processes may change the size and composition of aerosol particles on a shorter time scale, thereby changing their hygroscopic and activating properties. The spatial-temporal difference in the characteristic height of aerosol particles is such that CCN number concentration (N CCN ) Also the spatiotemporal distribution of (a) is significantly different, which also highlights the quantization of the different regions N CCN Is of importance.
The existing direct observation of the CCN is mostly based on short-term outfield observation of a plurality of single sites, but the current satellite remote sensing technology cannot realize direct detection of the CCN. Although aerosol optical thickness characteristics in the global range can be obtained by using ground-based and space-based remote sensing observation to invert CCN number concentration, the inversion accuracy of the satellite optical method has larger error due to the fact that the satellite optical method is more sensitive to particles in a large-particle-size section. In addition, the current model has great uncertainty in predicting CCN number concentration in contaminated areas, and obtaining accurate values of CCN number concentration in a large spatial scale range remains a challenge. Closed experimental studies of CCN help to clarify the mechanism of aerosol particle conversion to CCN, and the relationship of other atmospheric variables and parameters to CCN, but a broad range of ubiquitous CCN parameterization schemes cannot be obtained due to the limited observations and the nonlinear relationship of CCN number concentrations to these factors.
Disclosure of Invention
In order to solve the technical problems in the background, the invention combines WRF-Chem numerical mode simulation and machine learning technology, establishes a high space-time resolution CCN number concentration accurate prediction system based on observation-mode-intelligent measurement and calculation, and realizes correction of the current climate mode result, thereby being beneficial to elucidating the characteristic rule of the refined CCN number concentration space-time variation.
In order to achieve the above object, the present invention provides a method for accurately predicting CCN number concentration with high space-time resolution, comprising the steps of:
collecting CCN number concentration samples affecting a region to be predicted by using a WRF-Chem mode; collecting characteristic factors of a region to be predicted by utilizing a multi-source big data network;
preprocessing the CCN number concentration sample and the characteristic factor respectively;
screening the processed characteristic factors to obtain prediction factors;
and constructing a prediction model based on the prediction factor, and completing the prediction of the CCN number concentration by using the prediction model.
Preferably, the feature factors include: chemical composition data, conventional gas contaminant data, and meteorological element data sets.
Preferably, the method for performing the pretreatment comprises: normalizing the characteristic factors; and performing memory correction and generalization processing on the CCN data to obtain the processed data.
Preferably, the processed characteristic factors are screened, and characteristic factors with small influence on the output result of the model are removed to obtain the prediction factors.
Preferably, the method for constructing the prediction model comprises the following steps: establishing a nonlinear relation between the characteristic factors and the CCN number concentration data; wherein the CCN number concentration data is a target variable.
Preferably, a random forest algorithm is adopted to construct the prediction model; meanwhile, cartesian coordinates are added in the construction process.
The invention also provides a high space-time resolution CCN number concentration accurate prediction system, which is used for realizing the method and comprises the following steps: the device comprises an acquisition module, a preprocessing module, a screening module and a prediction module;
the acquisition module is used for acquiring CCN number concentration samples affecting the region to be predicted by using a WRF-Chem mode; collecting characteristic factors of a region to be predicted by utilizing a multi-source big data network;
the pretreatment module is used for respectively carrying out pretreatment on the CCN number concentration sample and the characteristic factors;
the screening module is used for screening the processed characteristic factors to obtain prediction factors;
the prediction module is used for constructing a prediction model based on the prediction factors and completing the prediction of the CCN number concentration by using the prediction model.
Preferably, the feature factors include: chemical composition data, conventional gas contaminant data, and meteorological element data sets.
Compared with the prior art, the invention has the following beneficial effects:
the CCN prediction system constructed by the invention can optimize the mode simulation result, improves the prediction precision, and has higher concentration precision of the predicted CCN number compared with the traditional optical characteristic inversion algorithm. And meanwhile, the defect of the CCN observation in the space-time scale is also overcome. Meanwhile, the prediction method disclosed by the invention has universality, and is favorable for parameterizing the time-space characteristics of the CCN to the mode simulation for accurately predicting the CCN concentration, so that the influence of the CCN on weather and climate is accurately estimated.
Drawings
In order to more clearly illustrate the technical solutions of the present invention, the drawings that are needed in the embodiments are briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of a method according to an embodiment of the invention;
FIG. 2 is a schematic diagram of a model architecture according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of ranking importance of feature factors of a CCN number concentration prediction model according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of determining a super parameter of a CCN number concentration prediction model according to an embodiment of the present invention; wherein, (a) the change trend of the cross-validation score along with the number of the trees, (b) the change trend of the cross-validation score along with the maximum feature number, (c) the change trend of the cross-validation score along with the minimum sample size of the leaf nodes, (d) the change trend of the cross-validation score along with the minimum segmentation sample size of the nodes, and (e) the change trend of the cross-validation score along with the depth of the decision tree.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.
Example 1
As shown in fig. 1, a flow chart of a method of the present embodiment includes the steps of:
s1, acquiring CCN number concentration samples affecting a region to be predicted by using a WRF-Chem mode; collecting characteristic factors of a region to be predicted by utilizing a multi-source big data network;
wherein the feature factors affecting the target variable include: chemical composition data (using the TAP dataset from Qinghai university), conventional gas contaminant data (using national control site data), meteorological element data (using ERA-5 analysis data).
Whereas the simulation of the concentration data of the target variable CCN employs the regional on-line coupled meteorological Chemistry model (The Weather Research and Forecasting Model/Chemistry, WRF-Chem). The meteorological data adopted by the numerical mode is FNL analysis data issued by a numerical weather forecast center (NWP) of the United states atmospheric ocean (NOAA) in real time; the emissions source data list mainly includes an artificial source list, a natural source (biological source), and biomass combustion emissions data. The artificial source list is derived from a Multi-scale emission list model (Multi-resolution Emission Inventory for China, MEIC) in china, the meshed list data set is developed by the university of bloom, the spatial-temporal resolution is respectively month and 0.25 ° by0.25 °, the data range scale can cover the whole china area, the pollution industry relates to industry, electric power, residents, transportation and agriculture, more than 700 artificial emission sources are covered, and pollutants comprise BC, OC, PM2.5, PMcoarse, CO, CO2, NH3, NOx, SO2 and VOC. The biological source list was derived from the biological emissions source model results of MEGAN (Model ofemissions ofgases and aerosols fromNature). Biological sources mainly comprise organic matters such as isoprene, terpene and the like discharged by plants. And selecting different micro-physical schemes and chemical mechanisms, comprehensively simulating and predicting the time-space variation condition of CCN number concentration in the long time scale of the research area, comprehensively analyzing errors, predicting the duration and the mode setting complexity, and selecting an optimal mode to output a result.
S2, respectively preprocessing the CCN number concentration sample and the characteristic factors;
matching the space-time scale of the sample data to ensure the unification of the time-space resolution of the sample data; selecting near-surface CCN number concentration data of mode output (taking the typical supersaturation in cloud s=0.2% as an example); normalizing each variable of the characteristic factors, and converting all data of each characteristic factor list item into between [0,1], so as to cancel errors caused by the order-of-magnitude difference between the data of each dimension on model prediction; in the embodiment, whether the characteristic factors contain chemical component information or not is considered respectively, and the prediction performance of the model is estimated.
S3, screening the processed characteristic factors to obtain the predictive factors.
In the embodiment, feature factors which are input are screened by combining with feature parameter importance evaluation, and feature factors which have small influence on an output result, such as chlorine salt ChL in chemical components, sea level air pressure MSL in meteorological factors, total cloud quantity TCC and the like, are removed. Feature parameter importance is assessed by a method of rearranging feature importance. The main principle is that all values of a certain feature are randomly arranged on a data set to recalculate a prediction result, if the difference between the two prediction results is large, the influence of the feature factor on the model prediction result is obvious, and if the difference between the two prediction results is small, the importance of the feature parameter is low, so that the prediction factor is obtained.
S4, constructing a prediction model based on the prediction factor, and completing the prediction of the CCN number concentration by using the prediction model.
The specific model construction thought is as follows:
the machine learning model is built, namely, a nonlinear relation between a target variable and a characteristic parameter is built, the target variable of the embodiment is obtained by selecting different micro-physical schemes and chemical mechanisms by using a WRF-Chem numerical mode, carrying out comprehensive simulation and prediction on the time-space variation condition of CCN number concentration of a long-time scale of a research area, carrying out comprehensive error analysis, predicting duration and mode setting complexity, and selecting an optimal mode to output a result, wherein the overall model architecture is shown in figure 2.
Based on the WRF-Chem numerical mode and combined with more common atmospheric state variables, the invention selects PM2.5 mass concentration and chemical composition data (organic matters OM, ammonium salt NH4, sulfate SO4, nitrate NO3 and black carbon BC), conventional gas pollutant data (ozone O3, sulfur dioxide SO2, nitrogen dioxide NO2 and carbon monoxide CO), and conventional atmospheric data such as weather data (temperature TEM, relative humidity RH, wind speed WS, wind direction WD, boundary layer height BLH, precipitation total amount PRE, surface air pressure SP and solar radiation SSR) from European centers to be re-analyzed as input features. And constructing a CCN number concentration prediction method by using a random forest model of a machine learning integration algorithm, performing a comparison experiment with an established numerical mode and an actual observation result, synthesizing prediction duration, and searching an optimal set prediction scheme by model complexity and error analysis.
Because of the information relating to the spatio-temporal scale, cartesian coordinates are added to the feature factors at the beginning of the model construction. Because of the time information DOY (expressed as what day of the year), the seasonality of time cannot be reflected because the first and last days of the year are both in winter, but DOY values are very different. To solve this problem, day of the year DOY is converted into a cartesian coordinate system:
wherein t is x ，t y Is a cartesian coordinate system; t is the total number of days of the year; DOY can be normalized to [02 pi ] based on equation (1)]Within the interval and further converted to polar coordinates. The spatial information is generally given in terms of polar coordinates of longitude and latitude, and is not suitable for representing distance due to uneven variation of the polar coordinates of longitude and latitude. The Cartesian coordinate system can realize uniform numerical variation, so that longitude and latitude information of the national control site is converted into the Cartesian coordinate system as a characteristic parameter input model, and a specific conversion formula is as follows:
wherein,latitude, θ is longitude, and R is earth radius.
The adjustment of the model super-parameters mainly limits the model parameters, so that the complexity of the model is reduced on the premise of better model output results. The hyper-parameters adjustment referred to herein are mainly the number of trees, the maximum growth depth of the trees, the minimum sample size of the leaves, the minimum sample size of the branch nodes, the maximum number of eigenvalues, and criteria, typically the default genigini index. Either too simple or too complex a model results in poor generalization ability of the model, and thus it is desirable to find the best model complexity to minimize generalization errors. Based on the idea of cross-validation, namely setting different random forest parameters, firstly dividing a training set into n parts, wherein n is 10 parts, using n-1 parts to train a model, using the rest 1 parts to predict the model, and obtaining the optimal parameters of the model based on the average value of 10-fold cross-validation results. The larger the number of trees, the more complex the model, and the better the prediction capability of the model; the maximum growth depth of the tree is not limited by default, but is limited by considering that the sample size is large, and the parameter is limited by simplifying the model; the minimum sample number of the leaves and the minimum sample number of the branch nodes are both the branch reduction operation, and the two parameters need to be considered and adjusted when the sample number is large; the maximum selection feature number, i.e. how many features are selected from all input features to establish a decision tree, can be a default value or can be a direct designated number, and if the number is set as None, the maximum feature number is selected.
Example two
In the following, the present embodiment will be described in detail to solve the technical problems in actual life.
Taking North China plain as an example, the WRF-Chem mode is utilized to perform aerosol simulation by using two mechanisms, namely a carbon bond mechanism-a particle size separation method (CBMZ-MOSAIC) and a regional acid sedimentation mechanism-a mode method (RADM 2-SORGAM). And (3) integrating the numerical mode calculation efficiency, comparing with six external field observation results of the early North China plain, and selecting an optimal mode simulation result to output. The characteristic variables are selected by referring to discussion of the influence of early external field observation on the CCN number concentration, such as the mass concentration of the particles for representing the pollution degree, meteorological factors for influencing the secondary process of the particles, the concentration of the gas pollutants for influencing the nucleation rate of the particles, the chemical composition information of the particles and the like. Therefore, PM2.5 mass concentration and chemical component data (organic matters, ammonium salts, sulfate, nitrate and black carbon) are selected in the initial stage of model construction, conventional gas pollutant data come from observations of national control sites (ozone, sulfur dioxide, nitrogen dioxide and carbon monoxide), and weather data come from European centers to re-analyze conventional atmospheric data such as weather data (temperature, relative humidity, wind speed, wind direction, boundary layer height, precipitation total amount, surface air pressure and solar radiation) and the like as input features. The model result verification is based on the outside field observation of North China plain, namely, the observation of Beijing city sites in autumn and winter in 2014, autumn and winter in 2015, autumn and winter in 2016 and Beijing city sites in summer in 2017, the observation of Hebei in spring and summer in 2016 and the observation of Hebei in Qing Ji in winter in 2018.
Then correcting and assimilating all the data to ensure the unification of time resolution; normalizing the characteristic factors; in the embodiment, whether the characteristic factors consider chemical component information or not is considered respectively, and the prediction performance of the model is estimated.
After the data processing is completed, the prediction model factors are interpreted. Screening model input prediction factors according to the model arrangement importance, screening out characteristic factors with smaller influence on a mode output result, and reducing model complexity; such as chloride ChL in chemical composition, sea level air pressure MSL in meteorological factors, total cloud amount TCC, etc., and FIG. 3 shows a ranking chart of feature factor importance selected for the model.
Finally, constructing a model, wherein Cartesian coordinates are added when the model is constructed due to the information related to the space-time scale; determining a random forest algorithm of machine learning, constructing a model, adjusting model super-parameters, combining model feature factors and target variables, and establishing a high-space-time resolution CCN number concentration accurate prediction method based on 'observation-mode-machine learning'. FIG. 4 is an optimization result of the random forest model hyper-parameters constructed in the North China region. The CV cross-validation results showed that when the number of trees (n_evastiators) was less than 200, the prediction accuracy increased rapidly with increasing number of trees, followed by a gradual plateau. Earlier studies indicated that the number of trees for the random forest algorithm defaults to 500, and this example combines CV score values with data sample size selection parameters set to 500. The effect of the maximum depth (max_depth) of the tree on the accuracy is manifested in that as the depth increases, the accuracy of the model moves to the right, i.e. the complexity of the model increases, here the value 28. The larger the minimum sample number of the branch operation leaves and the minimum sample number value of the branch nodes, the larger the model generalization error, which indicates that the model is at a level close to the complexity of the optimal model. In view of the large sample size, this is slightly increased based on the default value. The effect of the selection of the maximum selection feature number on the CV score appears as a trend of increasing and then decreasing, so the highest value of the selected CV curve corresponds to 16. The comparison of the model test set and the WRF-Chem simulation value in this embodiment is: the correlation is 0.89-0.91, the slope is 0.83-0.86, and the root mean square error in the winter half year is slightly higher than that in the summer half year. Further comparing the observed value of CCN number concentration, the WRF-Chem mode simulation value and the random forest model prediction value, taking supersaturation degree of 0.20% as an example, the high space-time resolution CCN number concentration accurate prediction method based on observation-mode-machine learning can realize better prediction of the CCN number concentration in the North China region, compared with the WRF-Chem mode simulation result, the method is improved by about 30%, and the error of six observation comparison verification is within about 20%. The method shows that the constructed machine learning model can further realize the prediction of CCN number concentration in a larger space-time range of a composite pollution area.
In this embodiment, the model prediction accuracy of the feature factor considering the component information is higher than that of the model including only PM2.5 particulate matter concentration and other atmospheric state variables, compared with the prediction performance of the model considering the chemical component information or not.
On the whole, on the basis of WRF-Chem mode simulation, prediction of the concentration of the condensation nuclei of the cloud with larger space-time scale can be achieved by combining a machine learning technology and a conventional atmospheric state variable, and compared with an actual observation result, the WRF-Chem numerical mode simulation result is optimized by a machine learning model. The CCN number concentration prediction method proposed by this embodiment is critical to improve the aerosol-cloud parameterization scheme in climate mode to accurately evaluate the aerosol indirect climate effect.
Example III
The embodiment also provides a precise prediction system for the concentration of the CCN number with high space-time resolution, which comprises the following steps: the device comprises an acquisition module, a preprocessing module, a screening module and a prediction module; the acquisition module is used for acquiring CCN number concentration samples affecting the region to be predicted by using a WRF-Chem mode; collecting characteristic factors of a region to be predicted by utilizing a multi-source big data network; the characteristic factors are characteristic factor data acquired by utilizing a multi-source big data network; the pretreatment module is used for respectively carrying out pretreatment on the CCN number concentration samples and the characteristic factors; the screening module is used for screening the processed characteristic factors to obtain predicted factors; the prediction module is used for constructing a prediction model based on the prediction factors and completing the prediction of the CCN number concentration by using the prediction model. Wherein the characteristic factors include: chemical composition data, conventional gas contaminant data, and meteorological element data sets.
The above embodiments are merely illustrative of the preferred embodiments of the present invention, and the scope of the present invention is not limited thereto, but various modifications and improvements made by those skilled in the art to which the present invention pertains are made without departing from the spirit of the present invention, and all modifications and improvements fall within the scope of the present invention as defined in the appended claims.
Claims (8)
1. A precise prediction method for the concentration of CCN numbers with high space-time resolution is characterized by comprising the following steps:
collecting CCN number concentration samples affecting a region to be predicted by using a WRF-Chem mode; collecting characteristic factors of a region to be predicted by utilizing a multi-source big data network;
preprocessing the CCN number concentration sample and the characteristic factor respectively;
screening the processed characteristic factors to obtain prediction factors;
and constructing a prediction model based on the prediction factor, and completing the prediction of the CCN number concentration by using the prediction model.
2. The method for accurate prediction of high spatial-temporal resolution CCN number concentration according to claim 1, wherein said characteristic factors include: chemical composition data, conventional gas contaminant data, and meteorological element data sets.
3. The method for accurate prediction of high spatial-temporal resolution CCN number concentration according to claim 1, wherein said method for performing said preprocessing comprises: normalizing the characteristic factors; and performing memory correction and generalization processing on the CCN data to obtain the processed data.
4. The accurate prediction method of the high spatial-temporal resolution CCN number concentration according to claim 3, wherein the processed characteristic factors are screened, and characteristic factors with small influence on a model output result are removed to obtain the prediction factors.
5. The method for accurately predicting the CCN number concentration in high spatial and temporal resolution of claim 1, wherein the method for constructing the prediction model comprises: establishing a nonlinear relation between the characteristic factors and the CCN number concentration data; wherein the CCN number concentration data is a target variable.
6. The precise prediction method of the concentration of the high spatial-temporal resolution CCN numbers according to claim 5, wherein a random forest algorithm is adopted to construct the prediction model; meanwhile, cartesian coordinates are added in the construction process.
7. A high spatial-temporal resolution CCN number concentration accurate prediction system for implementing the method of any one of claims 1-6, comprising: the device comprises an acquisition module, a preprocessing module, a screening module and a prediction module;
the acquisition module is used for acquiring CCN number concentration samples affecting the region to be predicted by using a WRF-Chem mode; collecting characteristic factors of a region to be predicted by utilizing a multi-source big data network;
the pretreatment module is used for respectively carrying out pretreatment on the CCN number concentration sample and the characteristic factors;
the screening module is used for screening the processed characteristic factors to obtain prediction factors;
the prediction module is used for constructing a prediction model based on the prediction factors and completing the prediction of the CCN number concentration by using the prediction model.
8. The high spatial-temporal resolution CCN number concentration accurate prediction system of claim 7, wherein said feature factors comprise: chemical composition data, conventional gas contaminant data, and meteorological element data sets.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410012958.6A CN117828992A (en) | 2024-01-04 | 2024-01-04 | Accurate prediction method and system for CCN number concentration with high space-time resolution |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410012958.6A CN117828992A (en) | 2024-01-04 | 2024-01-04 | Accurate prediction method and system for CCN number concentration with high space-time resolution |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117828992A true CN117828992A (en) | 2024-04-05 |
Family
ID=90507483
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410012958.6A Pending CN117828992A (en) | 2024-01-04 | 2024-01-04 | Accurate prediction method and system for CCN number concentration with high space-time resolution |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117828992A (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20110082946A (en) * | 2010-01-12 | 2011-07-20 | 연세대학교 산학협력단 | Apparatus and method for estimating precipitation with cloud microphysics scheme, apparatus and method for weather forecasting, and recording medium thereof |
| CN112016696A (en) * | 2020-08-14 | 2020-12-01 | 武汉大学 | PM integrating satellite observation and ground observation1Concentration inversion method and system |
| CN113537515A (en) * | 2021-07-27 | 2021-10-22 | 江苏蓝创智能科技股份有限公司 | PM2.5 prediction method, system, device and storage medium |
| KR20220040081A (en) * | 2020-09-23 | 2022-03-30 | 경북대학교 산학협력단 | Meteorological prediction apparatus and method capable of adjusting the parameters of cloud microphysics, and recorded media recording program to perform the same |
| CN114974453A (en) * | 2022-05-30 | 2022-08-30 | 中南大学 | High-coverage reconstruction method for concentration of atmospheric carbon dioxide column |
| US20220373984A1 (en) * | 2021-05-19 | 2022-11-24 | Shandong University | Hybrid photovoltaic power prediction method and system based on multi-source data fusion |
| US20230186173A1 (en) * | 2021-12-14 | 2023-06-15 | Wuhan University | Method of analyzing influence factor for predicting carbon dioxide concentration of any spatiotemporal position |
-
2024
- 2024-01-04 CN CN202410012958.6A patent/CN117828992A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20110082946A (en) * | 2010-01-12 | 2011-07-20 | 연세대학교 산학협력단 | Apparatus and method for estimating precipitation with cloud microphysics scheme, apparatus and method for weather forecasting, and recording medium thereof |
| CN112016696A (en) * | 2020-08-14 | 2020-12-01 | 武汉大学 | PM integrating satellite observation and ground observation1Concentration inversion method and system |
| KR20220040081A (en) * | 2020-09-23 | 2022-03-30 | 경북대학교 산학협력단 | Meteorological prediction apparatus and method capable of adjusting the parameters of cloud microphysics, and recorded media recording program to perform the same |
| US20220373984A1 (en) * | 2021-05-19 | 2022-11-24 | Shandong University | Hybrid photovoltaic power prediction method and system based on multi-source data fusion |
| CN113537515A (en) * | 2021-07-27 | 2021-10-22 | 江苏蓝创智能科技股份有限公司 | PM2.5 prediction method, system, device and storage medium |
| US20230186173A1 (en) * | 2021-12-14 | 2023-06-15 | Wuhan University | Method of analyzing influence factor for predicting carbon dioxide concentration of any spatiotemporal position |
| CN114974453A (en) * | 2022-05-30 | 2022-08-30 | 中南大学 | High-coverage reconstruction method for concentration of atmospheric carbon dioxide column |
Non-Patent Citations (2)
| Title |
|---|
| 刘冲: ""长三角地区气溶胶-云参数化及其对降水的影响研究"", 《中国博士学位论文全文数据库•基础科学辑》, vol. 2021, no. 12, 15 December 2021 (2021-12-15), pages 1 - 160 * |
| 郭世钰等: ""国内典型人工影响天气云降水观测试验进展"", 《气象科技进展》, vol. 13, no. 04, 20 August 2023 (2023-08-20), pages 41 - 49 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Stirnberg et al. | Meteorology-driven variability of air pollution (PM 1) revealed with explainable machine learning | |
| Jumin et al. | Solar radiation prediction using boosted decision tree regression model: A case study in Malaysia | |
| CN106920007B (en) | PM based on second-order self-organizing fuzzy neural network2.5Intelligent prediction method | |
| CN112884079A (en) | Method for estimating near-surface nitrogen dioxide concentration based on Stacking integrated model | |
| CN109060617B (en) | Simulation and analysis method for influences of different agricultural ammonia emission situations on PM2.5 concentration | |
| KR102100739B1 (en) | Method for predicting fine dust occurrence of target area | |
| Tian et al. | Estimating PM2. 5 from multisource data: A comparison of different machine learning models in the Pearl River Delta of China | |
| CN109858686A (en) | A kind of ground emission inventories inverting optimization method based on EnKF | |
| Zu et al. | Investigation of relationships between meteorological conditions and high PM 10 pollution in a megacity in the western Yangtze River Delta, China | |
| Moxim et al. | The meteorological nature of variable soluble iron transport and deposition within the North Atlantic Ocean basin | |
| CN113987912A (en) | Pollutant on-line monitoring system based on geographic information | |
| CN117171546A (en) | Urban air quality prediction method combined with pollution diffusion index | |
| Lu et al. | Multistation collaborative prediction of air pollutants based on the CNN-BiLSTM model | |
| CN116665802B (en) | Assessment method and system for regional ozone concentration | |
| Rumaling et al. | Forecasting particulate matter concentration using nonlinear autoregression with exogenous input model | |
| CN117828992A (en) | Accurate prediction method and system for CCN number concentration with high space-time resolution | |
| Ettouney et al. | Forecasting of ozone pollution using artificial neural networks | |
| Li et al. | Characteristics of secondary inorganic aerosols and contributions to PM2. 5 pollution based on machine learning approach in Shandong Province | |
| CN115545316A (en) | Atmospheric ozone prediction method based on deep learning hybrid model | |
| Song et al. | Spatial prediction of PM2. 5 concentration using hyper-parameter optimization XGBoost model in China | |
| Ju et al. | Analysis and research of absorbing aerosols in Beijing-Tianjin-Hebei region | |
| CN113095499A (en) | Insulator equivalent salt deposit density prediction method | |
| Zhang et al. | A multi-step prediction method of urban air quality index based on meteorological factors analysis | |
| Kaminska et al. | Lag variables in nitrogen oxide concentration modelling: A case study in wroc law, poland | |
| CN117589948B (en) | Cloud condensation nucleus source analysis method and system based on particle number spectrum |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |