Construction method of urban spatial data set for predicting bird distribution

Abstract

The invention relates to the technical field of ecological data, and discloses a method for constructing an urban spatial data set for predicting bird distribution, comprising the following steps: acquiring dominant environmental factors and urban bird point location distribution data, and processing the environmental factors and the urban bird point location distribution data; utilizing Spearman correlation analysis to calculate a correlation relationship of the environmental factors; running the MaxENT model, and screening out key environmental influence factors; carrying out normalization processing on the key environment influence factors; and taking the bird richness distribution data as a dependent variable y and the key environment influence factor data as an independent variable factor xn, constructing a generalized linear model, and finally obtaining an urban space data set formula capable of being used for predicting bird distribution. According to the method, the environmental factors which mainly influence the survival of the birds in the urban area are determined, and the method is suitable for predicting the bird richness spatial distribution in the high-density urban area under the condition of lack of bird distribution data.

Description

Technical Field

[0001] The present invention relates to the technical field of ecological data, and in particular to a method for constructing an urban spatial data set for predicting bird distribution. Background Art

[0002] The disorderly expansion of cities has caused the landscape fragmentation of the original habitat space, restricted the spread and reproduction of animals, and led to the reduction of urban biodiversity. Biodiversity is the result of the continuous evolution of the biosphere. It is not only the natural basis of ecosystem services, but also the guarantee for the survival of human society. Suitable indicator species can reflect a wide range of biodiversity changes and respond sensitively to environmental stress and ecological disturbances. Therefore, identifying the spatial distribution of representative indicator species is of great significance for understanding and evaluating the health of urban ecosystems and habitat changes.

[0003] Birds are considered by scholars to be an important component of urban ecological environment and urban biodiversity, as well as an important indicator of their changes. On the one hand, bird diversity in urban landscapes is relatively easy to monitor and count, and as birds are in the middle and upper layers of the food chain of urban ecosystems, changes in their number and species can reflect the dynamic changes of other species, and are an important indicator for measuring the quality of urban habitats. On the other hand, birds are very sensitive to changes in habitat patterns, especially to dynamic changes in ecosystems and human disturbances. They are also highly mobile and habitat selective, making them the most important indicator species for studying urban habitat evolution and ecosystem changes.

[0004] However, the amount of data on urban bird distribution is usually huge and difficult to obtain. Therefore, the present invention will start from the two core links of identifying key environmental influencing factors and generating urban spatial datasets, and explain the method of constructing an urban spatial dataset that can be used to predict bird distribution.

[0005] SDMs, also known as niche models or environmental response models, use specific algorithms to calculate species distribution information and environmental variable data such as the geographic and climatic environment where the target species live, so as to achieve quantitative analysis of the preference and dependence of species on the surrounding environmental ecological factors, and put the calculation results into the ecological landscape in the form of probability to estimate the spatial distribution of target species. SDMs are based on niche theory, including the maximum entropy model (MaxEnt), the bioclimatic analysis system (BIOCLIM), the niche factor analysis model (ENFA), the classification and regression tree model (CART) and other models. Among them, MaxEnt is based on the niche model and estimates the distribution of species by determining the stable relationship between species and the environment. Because it only requires species occurrence point data and can obtain high-precision results, it can avoid the problem of difficulty in obtaining non-occurrence point data. Therefore, it has become a model that is currently recognized to be less affected by samples and an effective tool for species distribution research. It can accurately identify key environmental factors that have a greater impact on species distribution.

[0006] The generalized linear model (GLM) is a statistical learning method based on generalized probability distribution to predict a continuous or discrete response variable. Unlike the traditional linear regression model, GLM can handle both continuous and discrete data and has greater flexibility to fit various types of response variables. It can model random dependent variables such as normal distribution, binomial distribution, Poisson distribution, and Gamma distribution.

[0007] In view of this, the present invention is proposed. This method constructs a method for constructing an urban spatial data set for predicting bird distribution under the support of the MaxENT model and the generalized linear model. Summary of the invention

[0008] In order to solve the problem that the amount of data on urban bird distribution is usually extremely large and difficult to obtain, the present invention provides a method for constructing an urban spatial data set for predicting bird distribution based on the MaxENT model and the generalized linear model, which determines the key environmental influencing factors affecting the spatial distribution of urban birds, and can predict the spatial distribution of urban bird richness in the absence of bird point data, thereby conducting research on the protection of urban bird diversity and studying the relationship between their habitat conditions and environmental factors.

[0009] The present application embodiment provides a method for constructing an urban spatial dataset for predicting bird distribution, comprising the following steps:

[0010] S110: Obtaining dominant environmental factors and urban bird point distribution data, and processing the environmental factors and urban bird point distribution data;

[0011] S120: Spearman correlation analysis was used to calculate the correlation between environmental factors;

[0012] S130: Run the MaxENT model to test the prediction accuracy of the model and test the contribution of different environmental factors to the distribution gain by the jackknife method to screen out key environmental influencing factors;

[0013] S140: Normalize key environmental impact factors;

[0014] S150: Use bird abundance distribution data as dependent variable y and key environmental impact factor data as independent variable factor x n , a generalized linear model was constructed, and finally a formula for urban spatial data set that can be used to predict bird distribution was obtained.

[0015] Furthermore, in S110, the dominant environmental factors include: phenological data, basic geographic data, vegetation quality data, land property data and human activity data.

[0016] Furthermore, in S110, the urban bird point distribution data is obtained through the China Bird Watching Network and the GBIF database; the phenological data comes from the World Climate Database; the basic geographic data comes from the National Earth System Science Data Center; the vegetation quality data comes from Google Earth Engine and the National Snow and Ice Data Center of the United States; the land property data is calculated from the land use data; and the human activity data comes from the National Oceanic and Atmospheric Administration of the United States; the above data are processed and unified with coordinates using ArcGIS, with a spatial resolution of no more than 10m, and output as an .asc format file for Maxent model analysis.

[0017] Furthermore, in S120, Spearman correlation analysis is used to calculate the correlation between 19 phenological data and NDVI, DR, DW, NL, CHM, AP and TP; environmental factors with a correlation coefficient lower than 0.5, i.e., no significant correlation with other variables, are screened out for MaxENT model construction, and the screened environmental factors are bio_1, bio_2, bio_4, bio_13, NDVI, DR, DW, NL, CHM and AP.

[0018] Furthermore, the S130 includes the following specific steps:

[0019] S131: Input the selected environmental variables and species distribution data into MaxEnt software, randomly select 25% of the distribution point data as the test set, and the remaining 75% of the data as the training set, and keep the other parameters as default;

[0020] S132: Run the model 10 times and take the average of the 10 validations as the final prediction result;

[0021] S133: Select the knife cutting method in the environment parameter setting and output the raster layer in ASCII format.

[0022] Furthermore, in S130, the key environmental factors are AP, CHM, and NDVI.

[0023] Furthermore, in S140, the key environmental impact factor data are normalized using the calculation variable module in the SPSS software.

[0024] Furthermore, in S150, the bird abundance is subjected to a generalized linear model with normal error distribution using SPSS software, wherein the bird abundance is the dependent variable y, and the key environmental influencing factors AP, CHM, and NDVI are the independent variable factors x, respectively. 1 、x 2 and x 3 , run the model, and obtain the data set formula y = -11.604x 1 +1.796x 2 +0.635x 3 .

[0025] Compared with the prior art, the technical solution of this application has the following beneficial effects:

[0026] The present invention analyzes the selected environmental variables by the Spearman correlation analysis method, screens out environmental variables with correlation coefficients lower than 0.5 and no significant correlation with other variables, inputs the environmental variables together with species distribution data into MaxEnt V3.4.1 software, randomly selects 25% of the distribution point data as a test set, and the remaining 75% of the data as a training set, and keeps the other parameters as default, runs the model 10 times, takes the average value of the 10 verifications as the final prediction result, selects the jackknife method in the environmental parameter setting, quickly determines the key environmental factors affecting the spatial distribution of birds in the city, and then obtains the urban spatial data set for predicting the distribution of birds by constructing a generalized linear model, which provides important guidance for the prediction of the spatial distribution of urban bird diversity and the protection of bird diversity. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0028] Figure 1 It is a schematic flow chart of a method for constructing an urban spatial data set for predicting bird distribution according to the present invention;

[0029] Figure 2 A technical roadmap for the method of constructing an urban spatial data set for predicting bird distribution in the present invention;

[0030] Figure 3 A heat map of environmental factor correlations for the method of constructing an urban spatial data set for predicting bird distribution according to the present invention;

[0031] Figure 4 This is a diagram showing the test results of the model jackknife method of the method for constructing an urban spatial data set for predicting bird distribution according to the present invention;

[0032] Figure 5 This is a graph showing the accuracy of the model and the impact of 10 environmental factors on the distribution of urban birds. DETAILED DESCRIPTION

[0033] The present invention is further described below in conjunction with the accompanying drawings and embodiments, but they are not intended to limit the present invention.

[0034] like Figure 1-Figure 5 As shown, the present invention provides a method for constructing an urban spatial data set for predicting bird distribution based on a maximum entropy model (MaxENT) and a generalized linear model, comprising the following steps:

[0035] Step 1: Acquisition and processing of dominant environmental factors and urban bird distribution data, where the dominant environmental factors include: phenological data, basic geographic data, vegetation quality data, land property data and human activity data, as shown in Table 1:

[0036] Table 1 Environmental variables for predicting bird distribution

[0037]

[0038]

[0039]

[0040] The distribution points of bird observation records were obtained through the China Bird Watching Network and the GBIF database; there are 19 phenological data (bio_1~bio_19) from the World Climate Database; basic geographic data including distance to water system DW and distance to road DR from the National Earth System Science Data Center; vegetation quality data including normalized vegetation index NDVI and vegetation canopy height CHM from Google Earth Engine and the National Snow and Ice Data Center; land property data including vegetation coverage TP and building proportion AP, calculated from land use data; human activity data including night light index NTL from the National Oceanic and Atmospheric Administration of the United States. The above data were processed and unified with ArcGIS, with a spatial resolution of 10m, and output as .asc format files for Maxent model analysis;

[0041] Step 2: The specific operation is: use Spearman correlation analysis to calculate the correlation between 19 phenological data and NDVI, DR, DW, NL, CHM, AP and TP, such as Figure 2 As shown. Generally speaking, when the absolute value of the correlation coefficient is between 0.1 and 0.3, 0.3 and 0.5, and greater than 0.5, it is considered that there is a weak correlation, moderate correlation, and strong correlation between the variables. This method screened out environmental factors with a correlation coefficient lower than 0.5, that is, no significant correlation with other variables, to construct the MaxENT model. The screened environmental factors are bio_1, bio_2, bio_4, bio_13, NDVI, DR, DW, NL, CHM, and AP.

[0042] The specific operation of step three is as follows: input the selected environmental variables into MaxEnt software together with the species distribution data, randomly select 25% of the distribution point data as the test set, and the remaining 75% of the data as the training set. The other parameters remain default, and the model is run 10 times. The average value of 10 verifications is taken as the final prediction result. Select the jackknife method in the environmental parameter setting, and output the ASCII format raster layer. This method uses the receiver operating characteristic curve (AUC) to evaluate the model accuracy. The average value of the 10 AUC values ​​of all environmental factors and the dominant environmental factors is 0.82, indicating that the model has good prediction accuracy and is suitable for simulating the range of suitable habitats for birds.

[0043] Summarizing the model results, it can be observed that there are significant differences in the contribution of environmental factors to the model. Among them, AP has the highest contribution, followed by CHM and NDVI, which indicates that these factors are key environmental factors that determine the suitable habitat for birds. The contribution of different environmental factors to distribution gain was tested by the jackknife method. The results showed that the scores of AP, CHM, and NDVI were particularly significant during the model training process and were determined to be key environmental influencing factors. The details are shown in Table 2.

[0044] Table 2 Main influencing factors of urban bird distribution

[0045] variable Contribution AP 58.2 CHM 13.6 NDVI 12.8 bio_2 5.2 bio_4 4.8 DR 4.5 bio_1 2.4 bio_13 2.2 DW 1.9 NTL 0.86

[0046] Step 4: The specific operation is: use the calculation variable module in SPSS software to normalize the data of the three key environmental factors.

[0047] Step 5: The specific operation is as follows: Use SPSS software to conduct a generalized linear model with normal error distribution on bird abundance, where bird abundance is the dependent variable y, and key environmental influencing factors AP, CHM, and NDVI are independent variable factors x. 1 、x 2 and x 3, run the model, and obtain the data set formula y = -11.604x 1 +1.796x 2 +0.635x 3 The details are shown in Table 3.

[0048] Table 3 Final GLM results for bird richness: bird richness, key environmental factors AP, CHM, and NDVI.

[0049]

[0050] Probability distribution: Normal

[0051] Correlation Function: Identity

[0052] a. Target: y: Bird abundance

[0053] The present invention determines the main environmental factors affecting the survival of birds in urban areas, and is suitable for predicting the spatial distribution of bird richness in high-density urban areas in the absence of bird distribution data.

[0054] The invention concept is described in detail using specific examples herein, and the description of the above embodiments is only used to help understand the core idea of ​​the invention. It should be pointed out that any obvious modification, equivalent substitution or other improvement made by a person of ordinary skill in the art without departing from the invention concept should be included in the protection scope of the present invention.

Claims

1. A method for constructing an urban spatial dataset for predicting bird distribution, characterized in that: The following steps are involved: S110: Obtaining dominant environmental factors and urban bird point distribution data, and processing the environmental factors and urban bird point distribution data; S120: Spearman correlation analysis was used to calculate the correlation between environmental factors; S130: Run the MaxENT model to test the prediction accuracy of the model and test the contribution of different environmental factors to the distribution gain by the jackknife method to screen out key environmental influencing factors; S140: Normalize key environmental impact factors; S150: Use bird abundance distribution data as dependent variable y and key environmental impact factor data as independent variable factor x n , a generalized linear model was constructed, and finally a formula for urban spatial data set that can be used to predict bird distribution was obtained.

2. The method for constructing an urban spatial data set for predicting bird distribution according to claim 1, characterized in that: In S110, the dominant environmental factors include: phenological data, basic geographic data, vegetation quality data, land property data and human activity data.

3. The method for constructing an urban spatial data set for predicting bird distribution according to claim 2, characterized in that: In S110, the urban bird point distribution data is obtained through the China Bird Watching Network and the GBIF database; the phenological data comes from the World Climate Database; the basic geographic data comes from the National Earth System Science Data Center; the vegetation quality data comes from Google Earth Engine and the U.S. National Snow and Ice Data Center; the land property data is calculated from the land use data; Human activity data comes from the National Oceanic and Atmospheric Administration of the United States. The above data are processed and unified with ArcGIS, with a spatial resolution of no more than 10m, and output as .asc format files for Maxent model analysis.

4. The method for constructing an urban spatial data set for predicting bird distribution according to claim 3, characterized in that: In S120, Spearman correlation analysis is used to calculate the correlation between 19 phenological data and NDVI, DR, DW, NL, CHM, AP and TP; environmental factors with a correlation coefficient lower than 0.5, i.e., no significant correlation with other variables, are screened out for MaxENT model construction, and the screened environmental factors are bio_1, bio_2, bio_4, bio_13, NDVI, DR, DW, NL, CHM and AP.

5. The method for constructing an urban spatial data set for predicting bird distribution according to claim 4, characterized in that: The S130 includes the following specific steps: S131: Input the selected environmental variables and species distribution data into MaxEnt software, randomly select 25% of the distribution point data as the test set, and the remaining 75% of the data as the training set, and keep the other parameters as default; S132: Run the model 10 times and take the average of the 10 validations as the final prediction result; S133: Select the knife cutting method in the environment parameter setting and output the raster layer in ASCII format.

6. The method for constructing an urban spatial data set for predicting bird distribution according to claim 5, characterized in that: In S130, the key environmental factors are AP, CHM, and NDVI.

7. The method for constructing an urban spatial data set for predicting bird distribution according to claim 6, characterized in that: In S140, the key environmental impact factor data are normalized using the calculation variable module in the SPSS software.

8. The method for constructing an urban spatial data set for predicting bird distribution according to claim 7, characterized in that: In S150, a generalized linear model with normal error distribution for bird richness is established using SPSS software, wherein bird richness is the dependent variable y, and key environmental influencing factors AP, CHM, and NDVI are independent variable factors x1, x2, and x3, respectively. The model is run to obtain the data set formula y=-11.604x1+1.796x2+0.635x3.

Images