Future grid population distribution prediction method

Abstract

The invention discloses a future grid population distribution prediction method, which comprises the following steps of: a) acquiring fine spatial-temporal scale population data in a historical period, and processing the fine spatial-temporal scale population data in the historical period to obtain a population elasticity coefficient; b) acquiring a ratio of crop yield or water resource quantity of each age in the future to a steady-state value of a reference year, and acquiring an environment variable under the influence of climate change; and c) according to the population elasticity coefficient and the environment variable under the climate change influence, generating future urban population and rural population distribution change based on the shared social economy SSPs and the shared social economy and representative forced path SSPs-RCPs composite scene, and finally synthesizing population distribution change. According to the future grid population distribution prediction method provided by the invention, the influence of the natural environment on the population distribution change is combined on the basis that the social and economic change influences the population distribution change, and the future grid population distribution can be predicted more accurately.

Description

Technical Field

[0001] The invention relates to the technical field of demography and geography, and in particular to a future grid population distribution prediction method based on an SSPs-RCPs composite scenario of the IPCC. Background Art

[0002] In the process of migration and settlement, human beings' choice of seeking benefits and avoiding harm in the environment has always been an important survival strategy. In fact, the early discussion of population distribution and migration was more reflected in the "human-food relationship". Starting from the "human-food relationship", researchers further expanded to the "population-industry" relationship and the "population-environment" relationship. This formed the linkage of "population-socio-economics-natural environment" to explore issues such as population distribution and migration. For a long time in the 20th century, due to the criticism of "geographical determinism", environmental factors were rarely involved in the explanation of population temporal and spatial changes, and more discussions were carried out from a socio-economic perspective. In addition, environmental factors and many other socio-economic factors interact with population distribution and migration, making it very difficult to strictly distinguish the degree of these factors. Population distribution simulation is one of the important methods to study how the population is distributed in time and space. In the early days, population distribution information was mainly obtained through regional statistics or population surveys. Although this method can obtain relatively accurate population information on statistical units, it is time-consuming, labor-intensive, and costly. In addition, it ignores the differences in population distribution within statistical units, which may lead to many inaccuracies in population distribution-related analyses. With the development of new technologies and methods such as remote sensing, geographic information systems, artificial intelligence, and optimization computing, population distribution simulation methods have also developed rapidly. Scholars from various countries have successively carried out population distribution simulation research at the global scale, national scale, provincial scale, county scale, and even pixel scale, and have formed some influential global population spatialization products, such as the GPW (Gridded Population of the World) population spatialization product of the International Earth Science Information Network Center of a certain university, and the Lans Scan population spatial distribution data product established by a laboratory in a certain country. These population distribution simulation studies involving multiple spatial scales provide effective basic data for related research in the fields of disasters, urban planning and management.

[0003] The above methods have played an important role in population distribution simulation, but they also have some shortcomings: Many models need to simplify reality when they are established, which may lead to the lack of full consideration of the complexity of the real world. For example, agent models may ignore the complex interactions between individuals, and micro-simulations may not take into account the heterogeneity of all individuals. Many models require a large amount of data for training and verification, including demographic data, geographic data, etc. However, sometimes it may be difficult to obtain these data, or the quality of the data may not be high enough, which will affect the accuracy and reliability of the model. Some models may take a long time to calculate or require a lot of computing resources due to the large amount of calculations. Especially when considering large-scale spatial and temporal ranges, the computational complexity of the simulation may be very high, which limits the practical application of the model. There are a large number of parameters in many models, and determining these parameters usually requires repeated adjustment and optimization of the model, which requires a lot of experiments and computing resources. If the parameters are not selected properly, the prediction results of the model may be inaccurate or unreliable. Since population distribution is affected by many uncertain factors, such as economic development, policy changes, natural disasters, etc., the prediction of future population distribution is often accompanied by great uncertainty. Models often have difficulty accurately capturing and quantifying these uncertainties, which affects the model's predictive ability. Taking cellular automata as an example, it is a bottom-up modeling idea that simulates global complex phenomena through local evolution models, thereby resolving complex problems that are difficult to describe into relatively simple local evolution rules, simplifying the process of solving complex problems. However, in population distribution simulation, the top-down population distribution simulation model has a low degree of automation, making it difficult to analyze the causes of population distribution, and is not suitable for high spatial resolution population distribution simulation because fine-scale population samples are difficult to obtain. At the same time, cellular automata lack exploration of the mechanism of population neighborhood and migration definitions, and also lack the disadvantage of detailed analysis of population migration and distribution between regions. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a method for predicting future grid population distribution in view of the above-mentioned defects of the prior art, and simulate the future spatiotemporal distribution of population based on the SSPs-RCPs composite scenario of the IPCC, aiming to solve the problem of inaccurate prediction and simulation of population distribution.

[0005] The technical solution adopted by the present invention to solve the problem is as follows:

[0006] A method for predicting future grid population distribution, the method comprising:

[0007] Step 1: Obtain population data at a fine spatiotemporal scale in the historical period, and process the population data at a fine spatiotemporal scale in the historical period to obtain a population elasticity coefficient, specifically including:

[0008] According to the average migration-residence time scale of the population, the peak age of migration rate and the influence radius of mutual attraction of population potential, the distance damping coefficient β of urban and rural populations is obtained, and the formula is as follows:

[0009]

[0010] Where: T is the average migration-residence time scale of the population, D is the area of ​​the interaction domain; t max is the peak age of migration rate;

[0011] Based on the crop yield and water resources data of the historical period, obtain the environmental variables of the historical period:

[0012]

[0013] Where: A' i represents the environmental variables of the historical period; c' i is the relative change in crop yields during the historical period; w' i is the relative change of water resources in the historical period; the crop yield elasticity coefficient γ and water resource elasticity coefficient δ are obtained by spatial error regression;

[0014] After clarifying the distance damping coefficients of urban and rural populations, the potential values ​​of urban and rural populations are obtained based on the population data of fine spatiotemporal scales in historical periods.

[0015]

[0016] Where: ν' i represents the historical population potential; i represents the grid where the study area is located; represents the population of the surrounding grid; α represents the elasticity coefficient of the population; Represents the distance attenuation level of the potential impact from surrounding areas; d ij represents the distance between the study area grid and the surrounding grids, β represents the distance damping coefficient; l i represents the spatial masking coefficient, j represents the grid set with potential influence around the grid where the study area is located, and m represents the grid set with potential influence around all grids;

[0017] According to the urban population potential value and the rural population potential value, the population potential ratio of the historical period is calculated to simulate the historical population distribution;

[0018] Compare the historical population simulation value with the demographic value to obtain the population elasticity coefficient, the formula is as follows:

[0019]

[0020] Among them, Pi mod Represents the population simulation value of each grid, P i obs represents the demographic value, i represents the grid, and n is the total number of grids;

[0021] Step 2: Obtain the ratio of crop yield and water resources in each future era to the steady-state value of the base year, and obtain environmental variables under the influence of climate change, including:

[0022] According to the spatial error regression method, the crop yield elasticity coefficient and water resource elasticity coefficient are obtained:

[0023]

[0024] ε=λWε+μ

[0025] Based on the crop yield elasticity coefficient and water resource elasticity coefficient, the environmental variables affected by climate change are determined:

[0026]

[0027] represents population change, W is the spatial weight matrix, ε is the regression residual vector, λ is the autoregressive parameter, represents crop yield data, Represents water resources data; A i represents the environmental variables under the influence of climate change, c i is the relative change of crop yield driven by RCPs scenario; γ is c i The elastic modulus w i is the relative change of water resources; δ is w i Elastic coefficient; Residual error μ;

[0028] Step 3: Based on the population elasticity coefficient and environmental variables under the influence of climate change, generate the future urban and rural population distribution changes based on the shared socioeconomic (SSPs) and shared socioeconomic and representative forcing pathways (SSPs-RCPs) composite scenarios, and finally synthesize the population distribution changes, including:

[0029] Based on the shared socioeconomic pathway and representative forcing pathway scenarios, the water resources index and crop yield index driven by the RCPs scenario are introduced to obtain the future population potential model:

[0030]

[0031] According to the future population potential model, the temporal and spatial change trends of urban and rural populations in the future under the SSPs scenario are calculated, that is, only considering socioeconomic changes without considering climate change; then the SSPs and RCPs scenarios are combined to calculate the future urban and rural population distribution changes under the SSPs-RCPs composite scenario, and finally synthesize the population distribution changes:

[0032]

[0033] Where: represents the average of the four simulated populations; z represents four groups of model combinations; P u represents the urban population; P r Represents rural population.

[0034] Beneficial effects of the present invention: The embodiment of the present invention first processes the population data of the historical period to obtain the population potential, and generates a future grid population distribution prediction method based on the population potential value in combination with the environmental variables under climate change; it can be seen that the future grid population distribution prediction method proposed in the embodiment of the present invention combines the impact of the natural environment on the changes in population distribution due to socio-economic changes, and can more accurately predict the future grid population distribution. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a schematic diagram of simulation of future population spatiotemporal distribution based on the SSPs-RCPs composite scenario according to an embodiment of the present invention;

[0036] Figure 2 This is a schematic diagram of the prediction of China's population distribution in 2050 according to an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The present invention is described in detail below with reference to the accompanying drawings and embodiments.

[0038] See also Figure 1 , the present invention specifically includes:

[0039] A fitting model of future population distribution under the SSPs-RCPs composite scenario was constructed. Future urban and rural population changes follow the population change trend set by the SSPs scenario. On this basis, water resource variables (represented by the relative change of 10-year average river runoff) and crop yield variables (represented by the relative change of 10-year average crop yield) driven by the RCPs scenario were introduced. A future population potential model based on the SSPs-RCPs composite scenario was constructed. The model parameters involved population elasticity coefficient, distance damping coefficient, water resource elasticity coefficient and crop yield elasticity coefficient. Through reasonable parameter setting, using urban and rural population data in 1990 and 2000, crop yield data and water resource data from 1970 to 2010, the population potential model was used to calculate the simulated data of urban population, rural population and total population distribution in 2010, and the data were verified and analyzed with the 2010 population statistics to evaluate the reliability of the model and parameters. The future population distribution in 10-year steps from 2020 to 2050 is fitted under the SSPs scenario (i.e., without considering the impact of climate change) and the SSPs-RCPs composite scenario.

[0040] By using an exponential population potential model, the temporal and spatial trends of urban and rural populations in the future under the SSPs scenario, that is, only considering socioeconomic changes without considering climate change, are calculated. Then, the SSPs and RCPs scenarios are combined to calculate the future distribution changes of urban and rural populations under the SSPs-RCPs composite scenario, and finally synthesize the population distribution changes. In order to reduce the systematic errors in the climate system model, crop model and hydrological model, the crop model and hydrological model driven by four climate models are used to calculate the distribution and trend of urban population, rural population and total population in 2030 and 2050, respectively, and finally synthesize the temporal and spatial distribution of population every 10 years from 2020 to 2100 under the SSPs-RCPs composite scenario.

[0041] Example

[0042] 1. Determination of distance damping coefficient

[0043] The distance damping coefficient is the key coefficient of the population potential model, and the formula is as follows:

[0044]

[0045] Where: T is the average migration-residence time scale of the population. In this example, the average migration-residence time is set to 10 years with a decade interval. D is the domain element area of ​​the interaction domain. The influence radius of the urban population potential model is set to 250 km, and the domain element area is 196250 km. 2 The rural population potential model sets the influence radius as 125km and the domain area as 49062km2 ;t max is the peak age of mobility. Set t max is 25. Then the distance damping coefficient of the urban population potential model is 0.002, and the distance damping coefficient of the rural population potential model is 0.004. That is, compared with the rural population, the distance damping coefficient of the urban population affected by the surrounding areas is smaller.

[0046] 2. Determination of population elasticity coefficient

[0047] After clarifying the distance damping coefficients of urban and rural populations, the population data of 1990 and 2000 are substituted into formula (2) to obtain the potential values ​​of urban and rural populations:

[0048]

[0049] Where: ν' i represents the historical population potential; i represents the grid where the study area is located; represents the population of the surrounding grid; α represents the elasticity coefficient of the population; Represents the distance attenuation level of the potential impact from surrounding areas; d ij represents the distance between the study area grid and the surrounding grids, β represents the distance damping coefficient; l i represents the spatial masking coefficient, j represents the grid set with potential influence around the grid where the study area is located, and m represents the grid set with potential influence around all grids;

[0050] Formula (3) is used to calculate the simulated values ​​of each grid in 2010:

[0051]

[0052] Where: v' i,y represents the population potential of grid i in year y, which is calculated by the population potential model and indicates the population impact potential of the surrounding grids; p i,y represents the population value of grid i in the era; △P T,y+10 represents the change in the total population of the country between the decade y+10 and decade y, and its value is obtained from the population scenario data of SSPs; V T,y Represents the sum of the population potential of all grids across the country.

[0053] When searching for the minimum cumulative sum of the differences between the population simulation values ​​and the population statistics values ​​of each grid in 2010 in the interval (0-1), the parameter value S is obtained:

[0054]

[0055] The population elasticity coefficient α in the rural population potential model is calculated by formula (2) to be 0.275, and the population elasticity coefficient α in the urban potential model is 0.3. That is, compared with the rural population, the urban population in the surrounding areas has a greater potential to influence the local urban population. i mod Represents the population simulation value of each grid, P i obs Represents a demographic value.

[0056] 3. Determination of crop yield elasticity coefficient and water resource elasticity coefficient

[0057] The crop yield elasticity coefficient γ and water resource elasticity coefficient δ in formula (5) are obtained by spatial error regression. The natural logarithm of formula (4) is obtained, and lnc and lnw are approximately equal to the change rate shown in formula (3) according to Taylor's first-order expansion, which is expressed as and

[0058]

[0059] ε=λWε+μ (6)

[0060] Where: represents population change; W is the spatial weight matrix; represents crop yield data, represents water resources data; ε is the regression residual vector; λ is the autoregression parameter, estimated using the maximum likelihood method. The principle of using spatial error regression for parameter estimation is that population changes are not only affected by the relative changes in local water resources and crop yields, but also by various other factors in the surrounding areas, represented by the spatial residual vector. The impact of other local potential factors on the population is represented by the residual residual μ.

[0061] 4. Determination of environmental variables under the influence of climate change

[0062] Compared with the rural population, the urban population is more affected by the surrounding areas. The influence radius is set to 20 grids around grid i, about 250km; the influence radius of the rural population is set to 10 grids around grid i, about 125km. Based on the future population potential model of Jones et al., the impact of climate change is considered. The relative change is the ratio of crop yield or water resources in each future era (expressed as a 10-year average) to the steady-state value in the base year (expressed as a 40-year average from 1970 to 2009):

[0063]

[0064] Where: A i represents environmental variables under the influence of climate change; ci is the relative change of crop yield driven by RCPs scenario; γ is c i The elastic modulus w i is the relative change of water resources; δ is w i The elastic coefficient.

[0065] 5. Determination of population potential ratio

[0066] The population potential ratio of the future population of grid i is simulated by the population potential model:

[0067]

[0068] Where: v i,y represents the population potential of grid i in year y, which is calculated by the future population potential model and indicates the population impact potential of the surrounding grids; p i,y represents the population value of grid i in the era; △P T,y+10 represents the change in the total population of the country between the decade y+10 and decade y, and its value is obtained from the population scenario data of SSPs; V T,y represents the sum of the population potential of all grids in the country. T,y+10 When it is a positive value, it means that the total population of the country is growing positively within a certain 10 years, that is, the larger the population potential value of the grid, the higher the possibility of future population growth; △P T,y+10 When it is a negative value, it means that the total population of the country has negative growth within a certain 10 years, that is, the larger the population potential value of the grid, the lower the possibility of population decline in the future.

[0069] 6. Determination of future population potential

[0070] The population potential model is a basic model for exploring population distribution and migration. Jones et al. combined the SSPs scenario with the population potential model to develop a future population potential model driven by the Shared Socioeconomic Pathways (SSPs). Based on this model, the present invention introduces the water resource index and crop yield index driven by the RCPs scenario to construct a future population potential model based on the SSPs-RCPs composite scenario:

[0071]

[0072] Where: ν i represents the future population potential; i represents the grid where the study area is located; represents the population of the surrounding grid; α represents the elasticity coefficient of the population; Represents the distance attenuation level of the potential impact from surrounding areas; d ij represents the distance between the study area grid and the surrounding grids, β represents the distance damping coefficient; l irepresents the spatial masking coefficient, which aims to control the population migration into areas that are not suitable for human habitation, such as water surfaces, extremely high altitude areas, and deserts in the model; j represents a set of grids with influence potential around the grid where the study area is located.

[0073] 7. Population distribution calculation

[0074] Through the discussion of the above models and parameters, this paper adopts an exponential population potential model to calculate the temporal and spatial variation trends of China's future urban and rural population under the SSPs scenario, that is, only considering socioeconomic changes without considering climate change. i Set to 1. Then, the SSPs and RCPs scenarios are combined to calculate the distribution changes of China's future urban and rural populations under the SSPs-RCPs composite scenario, and finally synthesize the population distribution changes. In order to reduce the systematic errors in the climate system model, crop model and hydrological model, the LPJmL, GEPIC crop model and HadGEM2-ES, IPSL-CM5A-LR hydrological model driven by the SSP126, SSP245, SSP360 and SSP585 climate models are used to calculate the distribution and change trends of China's urban population, rural population and total population in 2030 and 2050, respectively. Finally, the temporal and spatial distribution of China's population every 10 years from 2020 to 2050 under the SSPs-RCPs composite scenario is synthesized.

[0075]

[0076] Where: represents the average of the four simulated populations; z represents the four model combinations; P u represents the urban population; P r Represents rural population.

[0077] See also Figure 2 , based on the schematic diagram of China's population distribution forecast in 2050 given by the present invention.

Claims

1. A method for predicting future grid population distribution, characterized in that: The method comprises: Step 1: Obtain population data at a fine spatiotemporal scale in the historical period, and process the population data at a fine spatiotemporal scale in the historical period to obtain a population elasticity coefficient, specifically including: According to the average migration-residence time scale of the population, the peak age of migration rate and the influence radius of mutual attraction of population potential, the distance damping coefficient β of urban and rural populations is obtained, and the formula is as follows: Where: T is the average migration-residence time scale of the population, D is the area of ​​the interaction domain; t max is the peak age of migration rate; Based on the crop yield and water resources data of the historical period, obtain the environmental variables of the historical period: Where: A' i represents the environmental variables of the historical period; c' i is the relative change in crop yields during the historical period; w' i is the relative change of water resources in the historical period; the crop yield elasticity coefficient γ and water resource elasticity coefficient δ are obtained by spatial error regression; After clarifying the distance damping coefficients of urban and rural populations, the potential values ​​of urban and rural populations are obtained based on the population data of fine spatiotemporal scales in historical periods. Where: ν' i represents the historical population potential; i represents the grid where the study area is located; represents the population of the surrounding grid; α represents the elasticity coefficient of the population; Represents the distance attenuation level of the potential impact from surrounding areas; d ij represents the distance between the study area grid and the surrounding grids, β represents the distance damping coefficient; l i represents the spatial masking coefficient, j represents the grid set with potential influence around the grid where the study area is located, and m represents the grid set with potential influence around all grids; According to the urban population potential value and the rural population potential value, the population potential ratio of the historical period is calculated to simulate the historical population distribution; Compare the historical population simulation value with the demographic value to obtain the population elasticity coefficient, the formula is as follows: in, Represents the simulated population value of each grid, represents the demographic value, i represents the grid, and n is the total number of grids; Step 2: Obtain the ratio of crop yield and water resources in each future era to the steady-state value of the base year, and obtain environmental variables under the influence of climate change, including: According to the spatial error regression method, the crop yield elasticity coefficient and water resource elasticity coefficient are obtained: ε=λWε+μ Based on the crop yield elasticity coefficient and water resource elasticity coefficient, the environmental variables affected by climate change are determined: represents population change, W is the spatial weight matrix, ε is the regression residual vector, λ is the autoregressive parameter, represents crop yield data, Represents water resources data; A i represents the environmental variables under the influence of climate change, c i is the relative change of crop yield driven by RCPs scenario; γ is c i The elastic modulus w i is the relative change of water resources; δ is w i Elastic coefficient; Residual error μ; Step 3: Based on the population elasticity coefficient and environmental variables under the influence of climate change, generate the future urban and rural population distribution changes based on the shared socio-economic SSPs and the shared socio-economic and representative forcing pathway SSPs-RCPs composite scenarios, and finally synthesize the population distribution changes, including: Based on the shared socioeconomic pathway and representative forcing pathway scenarios, the water resources index and crop yield index driven by the RCPs scenario are introduced to obtain the future population potential model: According to the future population potential model, the temporal and spatial change trends of urban and rural populations in the future under the SSPs scenario are calculated, that is, only considering socioeconomic changes without considering climate change; then the SSPs and RCPs scenarios are combined to calculate the future urban and rural population distribution changes under the SSPs-RCPs composite scenario, and finally synthesize the population distribution changes: Where: represents the average of the four simulated populations; z represents four groups of model combinations; P u represents the urban population; P r Represents rural population.

Images