CN104809572A - Method for inversing population density based on night lamplight data - Google Patents

Abstract

The invention relates to a method for retrieving population density based on night light data, using data including VIIRS DNB night light data and land use data, as well as demographic data, village-level administrative boundary data and DEM data. The processing process is as follows: a. Extract urban functional areas based on land use data; b. Use night light data and population density data to establish the relationship between light value and population density value; c. Use the relationship established in b. Density inversion is performed to obtain the preliminary inversion result of population density; d. Establish the relationship between population density and night light brightness value in different functional areas, and use the relational formula in different functional areas to correct the preliminary inversion results in c.

Description

A Method of Retrieving Population Density Based on Night Light Data

technical field

The invention belongs to the application field of remote sensing information technology in geography, and focuses on solving the problem of using night light data to invert population density on a large scale, and improves the inversion accuracy in combination with urban functional zoning to provide a basis for large-scale applications. The method proposed by the present invention solves the problem of low accuracy in the application of night light data to population density inversion research on a large scale, provides a solution to the difficulty in obtaining data in population density research, and breaks the rule of statistical population data based on administrative boundaries The limitation of statistical units provides the basic conditions for the spatial redistribution of statistical population data.

Background technique

At present, in the application of night light data, DMSP/OLS night light data is mainly used. The resolution of this data is relatively low, it is easy to be saturated, and most of them are applied in a large-scale spatial range, mainly used for the extraction of urban built-up areas, GDP estimation, population density estimation, and population migration.

At present, in the research on inversion of population density based on night light data, DMSP/OLS night light data is mostly used, so the inversion accuracy is relatively low. VIIRS DNB used in the present invention is a new type of night light data. Research on this data relatively few applications.

Contents of the invention

The purpose of this invention is to use the new VIIRS DNB night light data to study the population density inversion of night light data on a smaller scale, and to correct the inversion results from the perspective of urban functional areas to improve the accuracy of the inversion results. The present invention proposes a new method for inverting population density based on night light data on a smaller scale, and provides a basis for large-scale applications. Specifically include the following.

A flow chart of a method for inverting population density based on night light data is shown in Figure 1, which mainly includes the following steps:

A. Establish a geographic grid system based on VIIRS lighting data

Since the population density and land use data need to be combined with the light data for analysis, the night light data is grid data, so the present invention uses the night light data to establish a geographic grid system, and each grid attribute has a corresponding ID and geographic information .

B. Identification of urban functional areas based on land use and geographic grid system

Firstly, based on two land classification standards, an urban functional classification system suitable for studying the application of lighting data is constructed, which provides a reference for the merging of land use types in urban functional classification. Then use the geographic grid system and the weight of urban functions to finally obtain the current data of urban functional areas based on land use.

a. Construct the urban function system based on the land use classification system, the results are shown in Figure 2

b. Determining the weight of urban functions

In the present invention, refer to the Delphi method: organize the investigation into questionnaire form, consult relevant experts respectively, sort the importance of the six functions, the smaller the serial number, the higher the importance, and sort out the first round of expert opinions; Then it is sent to experts for the second round of consultation. The experts set the weight range of each function based on the statistical results of the importance of the seven functions in the first round; the experts in the third round determine the specific weight range of each function based on the existing weight range. Weights. The results are shown in Figure 3.

c. Determine the cell city function

According to the urban function classification system, the urban land use type data will be integrated into the land function classification data. By embedding the land function classification data into the established geographic grid system, the land function type and its area in each grid unit can be obtained. Using formula (2),

A=a*I (2) (where a is the area occupied by each function, and I is the weight value corresponding to each function)

Calculate the relative value of the land function in each grid unit, and the function corresponding to the maximum value of A in the calculation result is the main function of the grid unit, that is, the urban function within the scope of the grid unit.

d. Form urban functional areas

Check whether the grid unit attributes are true. If the grid is too small, the functional area recognition effect is similar to the original land use type data, and the patchy area effect cannot be formed; if the grid is too large, many urban functional types will be lost in the grid. Determine the size of the grid unit according to the research requirements. If the grid properties do not meet the research needs, return to re-establish the geographic grid system and re-determine the shape and size of the grid units, etc., and repeat until the grid unit properties meet the research needs. Since each grid cell represents the surface characteristics of a local area, the sheet-like area formed by adjacent grid cells forms an urban functional area.

C. Establish the relationship between grid population density and light value

a. Population data gridding

When the data is converted from the administrative unit to the grid unit, the population per unit area (population density) is used as the bridge for data transmission. The specific operation process: create a new field P in the Y layer, and set the field Y-ID in U equal to j Add all the U field values of the Y layer, and put the result into the P field corresponding to the field Y-ID equal to j in the Y layer, such as the formula P _J =∑U _ij , the content of the P field of the Y layer is the result of gridding .

b. Establish a relationship model between population density and light brightness value

D. Using the existing model and lighting data inversion to obtain preliminary results

Because the light brightness value of the industrial land, storage land, and external transportation facilities in the city is relatively high, which is similar to the light value of the actual high population density area in the urban core area, so according to the results of the model simulation The high population density leads to large errors in the results. Therefore, to further improve the accuracy of simulating population density based on night light data, it is necessary to combine land use and urban functional areas for analysis.

E. Establish the relationship between population density and light brightness values in different functional areas

The present invention separately studies the relationship between the population density and the brightness of lights in different functional areas. Since the population density in ecological functional areas and agricultural functional areas is low, the value of light brightness is small. Analyze the population density and light brightness value of five types of functional areas, including management area, commercial service area, industrial area, and traffic area.

F. Preliminary results based on functional zone model correction

Based on the model of population density and light brightness value in each functional area, the preliminary results of population density simulation based on night light data are corrected, and the population density inversion results based on urban functional zoning corrected light data are obtained.

Description of drawings

Figure 1 is a flowchart of a method for retrieving population density based on night light data;

Figure 2 is a diagram of the urban function classification system;

Figure 3 is a map of urban function weights;

Figure 4 is the status map of the urban functional areas in Xiamen;

Figure 5 is the preliminary results of population density retrieval from night light data in Xiamen;

Figure 6 shows the relationship between the population density of different functional areas and the average value of light brightness;

Figure 7 is a population density map corrected based on functional zoning.

Detailed ways

A method of retrieving population density based on night light data in the present invention can be illustrated by the following examples:

A. The present invention selects Xiamen City as the implementation case city, uses the vector boundary of Xiamen City to cut the night light data in 2012, uses the grid attribute as the cell attribute, and establishes a geographic grid system based on the night light data. Each grid There are corresponding ID and geographic information in .

B. The classification of urban functions is based on land functions and is divided into seven types. The result is shown in Figure 2.

C. Use the Delphi method to set the city function weight, the result is shown in Figure 3.

D. Based on the urban functional classification system and functional weights, based on the geographic grid system and the current land use data as the basic data, the current status of the urban functional areas in Xiamen is obtained. The results are shown in Figure 4.

E. Establish a population density and light brightness model

a. The relationship between population density and radiation rate obtained from the analysis of village-level population density and average radiation rate: y=-0.191*x ³ +21.054*x ² -213.613*x+901.182, where y is the average population density, unit is people/square kilometer, and x is the radiance rate of light data at night. In the above regression model, the p value is 0.000, and the R ² is 0.738, indicating that the average population density has a good correlation with the radiance rate of night light data.

F. Obtain preliminary inversion results based on the model

Using the population density and light brightness value model in E(a), the population density inversion was performed using the VIIRS DNB light data in Xiamen in 2012, and the results are shown in Figure 5. The results show that some areas with low population density, such as industrial areas and traffic areas, have high light brightness values, but in fact the population density is relatively small, while the population density in the inversion results is high, as shown in Figure 6. Therefore The results need to be corrected from the perspective of urban functional areas.

G. Establish the model of population density and light brightness value in each functional area

a. Residential functional area

Analysis results of the grid population density and light brightness values in the residential functional area by using SPSS software:

y=-0.002*x ³ +0.255*x ² -3.784*x+51.295 where x is the light brightness value, y is the population density in the cell, and the unit is people/square kilometer. In the above regression model, the P value is 0.000, indicating that the regression model between the average population density and the radiance rate of night light data has statistical significance, and R2 is 0.595, indicating that the population density in residential areas and the radiance rate of night light data have a good correlation .

b. Public service and public management area

The following are the analysis results of the grid population density and light brightness values in the residential functional area using SPSS software in the public service and public management area:

y=-0.0003*x ³ +0.032*x ² +2.367*x+26.774 Among them, x is the light brightness value, y is the population density in the cell, and the unit is people/square kilometer. In the above regression model, the P value is 0.000, indicating that the regression model between the average population density and the radiance rate of night light data is statistically significant, and the R2 is 0.439.

c. Commercial service area

The following is the analysis result of the grid population density and light brightness value in the residential function area using SPSS software in the commercial service area: y=0.0003*x ³ -0.1003*x ² +10.223*x-42.432 where x is the light brightness value , y is the population density in the cell, and the unit is person/km2. In the above regression model, the P value is 0.000, indicating that the regression model between the average population density and the radiance rate of night light data is statistically significant, and the R2 is 0.529.

d. Industrial functional area

The following is the analysis result of grid population density and light brightness value using SPSS software in the industrial area: y=-0.001*x ³ +0.14*x ² -1.518*x+42.101 where, x is the light brightness value, y is the grid The population density in yuan, the unit is person/square kilometer. In the above regression model, the P value is 0.000, indicating that the regression model between the average population density and the radiance rate of night light data is statistically significant, and the R2 is 0.527.

e. Traffic function area

The following is the analysis result of the grid population density and light brightness value using SPSS software in the traffic function area: y=-0.012*x ² +1.844*x+22.51 where x is the light brightness value, and y is the population density in the cell , the unit is person/square kilometer. In the above regression model, the P value is 0.000, indicating that the regression model between the average population density and the radiance rate of night light data is statistically significant, and the R2 is 0.254.

H. Calibration of simulation results based on functional zone model

Based on the model of population density and light brightness value in each functional area obtained from the above analysis, the simulation results of the population density map based on night light data simulation are corrected. Figure 7 is the population density map corrected based on functional zoning, and the correction results The population density of some industrial areas, ports, and residential areas in the middle is lower than the original simulation results and is closer to the actual population distribution.

I． Verification of accuracy

Taking demographic data as reference data, the accuracy of the corrected population simulation results is verified. First, the total simulated population at the village level is calculated using the village-level administrative boundary as a unit; then the relative error is calculated based on the statistical population and the simulated population. The average error within the city is 0.36, as shown in Table 1 for the relative error values of some villages.

Table 1 Accuracy of simulation calibration results.

Using the method of retrieving population density based on night light data has obvious advantages:

(1) The simulation results of population density and light brightness value based on the geographic grid system show that there is a certain correlation between the light brightness value and the population density on the urban scale, and the population density can be calculated using night light data on a large regional scale. simulation;

(2) In the present invention, the application is based on the identification results of urban functional areas. Due to the overflow characteristics of lights, the radiation of lights makes the difference between the brightness values of lights in units of regions. At the same time, the extraction based on land use in this study For the urban functional area, use the function weight value to determine the dominant function of the cell, so that the urban functional area forms a contiguous area. This method makes the two data have homogeneity in space and improves the research accuracy;

(3) The present invention simulates the population data from the urban scale, and finds that the population density in some functional areas is abnormal, and the simulated population density in the traffic and industrial areas is obviously high. By analyzing the relationship between population density and light brightness values at the district scale, it is found that there are differences in the scatter distributions of population density and light brightness values in different functional areas, and the coefficient of determination of the regression model for traffic function areas and public service areas is relatively low. Therefore, in order to improve the simulation accuracy, the relationship between population and light brightness values can be analyzed from the perspective of urban functional areas, so as to correct the error of large-scale population and light brightness values.

Claims (3)

1. A method for retrieving population density based on night light data, characterized in that the following steps: A. Retrieve the population density on a smaller scale based on the new night light data VIIRS DNB light data; B. Extract urban functional areas based on land use, establish the correlation between population density and night light brightness values in each urban functional area, and correct the inversion results. 2. A method for retrieving population density based on night light data according to claim 1, characterized in that: taking advantage of the spatial resolution of the new night light data and the advantages of the improved saturated light brightness value, on a small scale Perform population density inversion to improve the accuracy of night light data in the application of population density inversion. 3. A method for retrieving population density based on night light data according to claim 1, characterized in that: identify urban functional areas based on land use and geographic grid system, and establish light brightness values and The relationship between population density, using this relationship to correct the results obtained by directly inverting population density from light data, in order to improve the accuracy of large-scale population density inversion in existing research.