Prediction method and device for potential distribution of mangrove forest

Abstract

The invention relates to the field of ecological environment restoration, and discloses a mangrove forest potential distribution prediction method comprising the following steps: 1, obtaining mangrove forest site distribution information and environment variables; 2, Pearson correlation calculation is carried out on the environment variable data, and a target environment variable is screened out; and step 3, inputting the target environment variable and the mangrove forest site distribution information into a maxent model, and outputting prediction data by the maxent model. According to the prediction method, the environmental data is screened, so that the prediction efficiency and the prediction precision of the maxent model are improved. Meanwhile, the invention further provides a mangrove forest potential distribution prediction device.

Description

Technical Field

[0001] The present invention relates to the field of ecological environment restoration, and particularly to a method and device for predicting the potential distribution of mangroves. Background Art

[0002] As a special ecosystem at the land-sea junction, the mangrove ecosystem provides huge ecological service function values and plays an important role in global ecological protection. Predicting the potential distribution of mangroves has important reference significance for subsequent calculation of the carbon storage of mangroves and the restoration of mangroves.

[0003] For the potential distribution of mangroves, generally the maximum entropy model is used for prediction. By inputting the environmental factors affecting the distribution of mangroves into the maximum entropy model, the maximum entropy model can predict the future distribution location of mangroves according to relevant data. There are many environmental factors affecting mangroves. Inputting all environmental factors into the maximum entropy model for prediction will not only affect the operation efficiency of the model, but also due to the high collinearity of some environmental factors, it will cause the maximum entropy model to overfit, thereby reducing the prediction accuracy of the model.

[0004] CN202211212714 discloses a method for estimating the increment of mangrove carbon storage based on the maximum entropy model and the InVEST model. Step S1: Obtain the existing mangrove distribution points through field investigation, and determine the main environmental factors affecting the distribution of mangroves. The main environmental factors include climate factors and marine data. Step S2: Use the maximum entropy model and geographic information system analysis to predict the potential habitat distribution of mangroves, and use the jackknife method to analyze the main environmental factors affecting the distribution of mangroves to obtain the suitable habitat data of mangroves. Use the jackknife method of the maximum entropy model to evaluate the weights of 36 main environmental factors, and obtain the optimal parameter settings through multiple numerical adjustments. The specific model operation parameters are that the number of operations is 1000 times, 25% is used as the training factor, 75% is used as the test factor, the number of repetitions is 10 times, and other parameters are default. This estimation method uses the maximum entropy model to predict the potential habitat distribution of mangroves. However, this method inputs all 36 main environmental factors into the maximum entropy model, which may affect the efficiency and accuracy of the maximum entropy model.

[0005] The technical problem to be solved by the present invention is: how to process environmental data to improve the prediction efficiency and accuracy of the potential distribution of mangroves. Summary of the Invention

[0006] The main object of the present invention is to provide a method for predicting the potential distribution of mangroves. By calculating the Pearson correlation of environmental variables, environmental variables with high correlation are removed to prevent the maxent model from overfitting and reducing the prediction accuracy.

[0007] Meanwhile, a prediction device for the potential distribution of mangroves is also provided.

[0008] To achieve the above object, the technical solution adopted in this application is as follows:

[0009] A method for predicting the potential distribution of mangroves includes the following steps:

[0010] Step 1: Obtain the mangrove site distribution information and environmental variables;

[0011] Step 2: Perform Pearson correlation calculation on the environmental variable data to screen out the target environmental variables;

[0012] Step 3: Input the target environmental variables and the mangrove site distribution information into the maxent model, and the maxent model outputs prediction data.

[0013] Preferably, it further includes Step 4: Input the prediction data into ArcGIS, and use the natural break classification method to draw the distribution map of the potential suitable area for mangroves.

[0014] Preferably, the environmental variables include annual average temperature, average daily temperature range, isothermality, standard deviation of seasonal change in temperature, highest temperature in the warmest month, lowest temperature in the coldest month, annual temperature range, average temperature in the wettest season, average temperature in the driest season, average temperature in the warmest season, average temperature in the coldest season, annual precipitation, precipitation in the wettest month, precipitation in the driest month, seasonal variation coefficient of precipitation, precipitation in the wettest season, precipitation in the driest season, precipitation in the warmest season, precipitation in the coldest season.

[0015] Preferably, the environmental variables further include ocean surface nitrogen content, ocean surface minimum water temperature, ocean surface average water temperature, ocean surface phosphorus content, ocean surface average salinity, ocean surface pH value, ocean surface highest water temperature.

[0016] Preferably, the specific operation of Step 2 is: perform Pearson correlation calculation on the environmental variables, calculate the Pearson correlation coefficient r between the environmental variables, and when r > |0.8|, according to the empirical conditions, eliminate the corresponding environmental variables, so as to screen out the target environmental variables.

[0017] Preferably, the empirical conditions are as follows:

[0018] When two of the environmental variables are environmental variables of different types, preferably eliminate the environmental variables related to precipitation and retain the environmental variables related to temperature;

[0019] When two of the environmental variables are environmental variables of the same type, preferably eliminate the environmental variable with a larger number of r > |0.8| among the other environmental variables.

[0020] Preferably, the specific operation of step 3 is as follows: input the target environmental variables and the mangrove site distribution information into the maxent model, calculate the target environmental variables by using the jackknife method in the maxent model, obtain the contribution rate of the target environmental variables, and the maxent model outputs the prediction data according to the contribution rate of the target environmental variables.

[0021] Preferably, step 3 further includes: verifying the prediction accuracy of the maxent model for the prediction data by using the receiver operating characteristic curve.

[0022] Among them:

[0023] maxent model: that is, the maximum entropy model, which estimates the probability distribution by maximizing entropy (uncertainty), is used for prediction and decision-making, and aims to make decisions under the most uncertain assumptions;

[0024] ArcGIS: is a comprehensive geographic information system (GIS) platform, and the main functions of ArcGIS include map making, spatial analysis, and data management.

[0025] Meanwhile, a prediction device for the potential distribution of mangroves is also disclosed, including the following units:

[0026] Data acquisition unit: used to acquire mangrove site distribution information and environmental variables;

[0027] Data processing unit: used to perform Pearson correlation calculation on the environmental variable data and screen out the target environmental variable data;

[0028] Distribution prediction unit: used to input the target environmental variables and the mangrove site distribution information into the maxent model, and the maxent model outputs the prediction data.

[0029] Preferably, it further includes a distribution map making unit: used to input the prediction data into ArcGIS and draw the potential suitable area distribution map of mangroves by using the natural break classification method.

[0030] Compared with the prior art, the present solution has the following beneficial effects:

[0031] The prediction method of the present solution calculates the Pearson correlation of the environmental variables affecting the potential distribution of mangroves, eliminates the environmental variables with high correlation, thereby reducing the number of environmental variables input into the maxent model, avoiding model overfitting, and improving the prediction accuracy of the model.

[0032] Secondly, by setting appropriate screening conditions, it can ensure that without reducing the prediction accuracy of the maxent model, the number of environmental variables input into the model is reduced, and the workload is reduced. Description of the Drawings

[0033] Figure 1 Flow chart of the prediction method for the potential distribution of mangroves in Example 1;

[0034] Figure 2 Result graph of the receiver operating characteristic curve for the contemporary potential distribution of mangroves in Example 1;

[0035] Figure 3 Result graph of the receiver operating characteristic curve for the future potential distribution of mangroves in Example 1;

[0036] Figure 4 Structural block diagram of the prediction device for the potential distribution of mangroves in Example 2. Detailed implementation manners

[0037] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. Generally, the components of the embodiments of the present application described and illustrated herein can be arranged and designed in various different configurations.

[0038] Example 1

[0039] Refer to Figures 1-3 , a prediction method for the potential distribution of mangroves, comprising the following steps:

[0040] Step 1: Obtain mangrove site distribution information and environmental variables;

[0041] According to field surveys, literature materials, records in the Flora of China, mangrove specimens in the Chinese Virtual Herbarium (https: / / www.cvh.org.cn / ) and the Global Biodiversity Information Facility (GBIF, http: / / www.gbif.org / (accessed on October 28, 2023)), obtain the coordinate positions of the distribution sites of mangroves in China, delete duplicate, incorrect records and sites with adjacent distances, and a total of 136 effective mangrove site distribution information is obtained.

[0042] The environmental variables include bioclimatic variable data. To further improve the prediction accuracy, the environmental variables can also include marine environmental data (marine surface nitrogen content, marine surface minimum water temperature, marine surface average water temperature, marine surface phosphorus content, marine surface average salinity, marine surface pH value, marine surface maximum water temperature). In this embodiment, the bioclimatic variable data is used to specifically illustrate this method.

[0043] Bioclimatic variable data is from the WorldClim database (http: / / www.worldclim.org / ). Data on 19 bioclimatic variables related to temperature and precipitation were downloaded for three periods: the current period (1970 - 2000) and the future periods (2041 - 2060 and 2081 - 2100). Among the future data, four CO2 emission scenarios, RCP 2.6, RCP 4.5, RCP 7.0, and RCP 8.5, are also included, with a resolution of 30s (approximately 1km).

[0044] The environmental variables include 19 bioclimatic variables, namely annual mean temperature (bio1), mean diurnal range of temperature (bio2), isothermality (bio3), standard deviation of temperature seasonality (bio4), maximum temperature of the warmest month (bio5), minimum temperature of the coldest month (bio6), annual temperature range (bio7), mean temperature of the wettest quarter (bio8), mean temperature of the driest quarter (bio9), mean temperature of the warmest quarter (bio10), mean temperature of the coldest quarter (bio11), annual precipitation (bio12), precipitation of the wettest month (bio13), precipitation of the driest month (bio14), coefficient of variation of precipitation seasonality (bio15), precipitation of the wettest quarter (bio16), precipitation of the driest quarter (bio17), precipitation of the warmest quarter (bio18), and precipitation of the coldest quarter (bio19).

[0045] Step 2: Calculate the Pearson correlation between environmental variable data and screen out the target environmental variables;

[0046] The specific operation of Step 2 is as follows: Calculate the Pearson correlation coefficient r between environmental variables. When r > |0.8|, according to empirical conditions, the corresponding environmental variables are excluded to screen out the target environmental variables.

[0047] In this embodiment, after collecting the 19 bioclimatic variables, according to the Pearson correlation calculation formula: where x i and y i are the i-th data of the bioclimatic variable, and [bioclimatic variable] is the bioclimatic variable, to obtain the correlation coefficient r between every two bioclimatic variables. The specific calculation results of the correlation between the 1st to 9th bioclimatic variables and the 19 bioclimatic variables are shown in Table 1; the specific calculation results of the correlation between the 10th to 19th bioclimatic variables and the 19 bioclimatic variables are shown in Table 2.

[0048] Table 1

[0049]

[0050]

[0051] Table 2

[0052] bio10 bio11 bio12 bio13 bio14 bio15 bio16 bio17 bio18 bio19 bio1 0.90 0.98 0.60 0.71 -0.69 0.71 0.73 -0.63 0.66 -0.72 bio2 0.04 -0.16 -0.51 -0.46 0.11 -0.19 -0.44 0.11 -0.50 0.20 bio3 0.78 0.79 0.33 0.42 -0.62 0.56 0.45 -0.60 0.40 -0.56 bio4 -0.65 -0.92 -0.64 -0.70 0.68 -0.62 -0.72 0.63 -0.63 0.66 bio5 0.78 0.57 0.10 0.19 -0.36 0.34 0.20 -0.35 0.12 -0.28 bio6 0.75 0.97 0.64 0.72 -0.71 0.67 0.73 -0.65 0.66 -0.70 bio7 -0.65 -0.91 -0.67 -0.72 0.69 -0.63 -0.73 0.64 -0.67 0.70 bio8 0.82 0.73 0.62 0.71 -0.55 0.70 0.73 -0.50 0.78 -0.65 bio9 0.78 0.90 0.41 0.53 -0.78 0.66 0.55 -0.75 0.46 -0.66 bio10 1.00 0.82 0.47 0.60 -0.61 0.68 0.61 -0.58 0.56 -0.62 bio11 0.82 1.00 0.60 0.70 -0.71 0.69 0.72 -0.65 0.64 -0.71 bio12 0.47 0.60 1.00 0.92 -0.26 0.50 0.92 -0.22 0.91 -0.35 bio13 0.60 0.70 0.92 1.00 -0.43 0.71 0.99 -0.39 0.93 -0.51 bio14 -0.61 -0.71 -0.26 -0.43 1.00 -0.82 -0.47 0.96 -0.46 0.93 bio15 0.68 0.69 0.50 0.71 -0.82 1.00 0.75 -0.81 0.74 -0.82 bio16 0.61 0.72 0.92 0.99 -0.47 0.75 1.00 -0.43 0.95 -0.55 bio17 -0.58 -0.65 -0.22 -0.39 0.96 -0.81 -0.43 1.00 -0.42 0.87 bio18 0.56 0.64 0.91 0.93 -0.46 0.74 0.95 -0.42 1.00 -0.57 bio19 -0.62 -0.71 -0.35 -0.51 0.93 -0.82 -0.55 0.87 -0.57 1.00

[0053] After calculating the correlation coefficients among the 19 bioclimatic variables, the bioclimatic variables with correlation coefficient r > |0.8| are removed according to the following empirical conditions.

[0054] The said empirical conditions are as follows:

[0055] When two of the environmental variables are of different types, the environmental variable related to precipitation is preferentially removed, and the environmental variable related to temperature is retained; the specific operation is: when r > |0.8| for two bioclimatic variables, if one is an environmental variable related to precipitation and the other is an environmental variable related to temperature, then the environmental variable related to precipitation is removed.

[0056] When two of the environmental variables are of the same type, the environmental variable with a larger number of r > |0.8| among the other environmental variables is preferentially removed. For this empirical condition, two environmental variables A and B are used for illustration. Suppose r > |0.8| between A and B, A also has r > |0.8| with 6 other environmental variables, while B has r > |0.8| with 3 other environmental variables, then B is removed.

[0057] By removing the 19 bioclimatic variables through the empirical conditions, the target environmental variables obtained are: diurnal range of temperature (bio2), isothermality (bio3), maximum temperature of warmest month (bio5), annual temperature range (bio7), mean temperature of wettest quarter (bio8), mean temperature of driest quarter (bio9), annual precipitation (bio12), precipitation of driest month (bio14), coefficient of variation of precipitation seasonality (bio15).

[0058] It should be noted that since the correlation coefficient is related to two bioclimatic variables, when r < |0.8|, one of them needs to be removed. By setting empirical conditions, the object to be removed is determined, so as to reduce the number of environmental variables and improve the processing efficiency of the maxent model while ensuring the prediction accuracy of the maxent model.

[0059] Step 3: Input the target environmental variables and the mangrove site distribution information into the maxent model, and the maxent model outputs the prediction data.

[0060] Preferably, the specific operation of step 3 is as follows: input the target environmental variables and the mangrove site distribution information into the MaxEnt model, use the jackknife method in the MaxEnt model to calculate the target environmental variables, obtain the contribution rate of the target environmental variables, and the MaxEnt model outputs the prediction data according to the contribution rate of the target environmental variables.

[0061] In this embodiment, the contemporary data in the obtained target environmental variables and the mangrove site distribution information are used as the test training data and input into the MaxEnt model, and the jackknife method is used to analyze the contribution rate of each environmental variable in the model. Among them, the random test percentage is set to 25%, repeated 10 times, and the repeated operation category is cross-validation. The contribution degree of the target environmental variables affecting the contemporary potential distribution of mangroves is shown in Table 3.

[0062] Table 3

[0063]

[0064]

[0065] Then, the prediction results of the MaxEnt model are verified using the receiver operating characteristic curve. The average AUC value of 10 verifications is 0.996, and the mangrove distribution range predicted by the MaxEnt model covers all actual distributions, indicating that the data obtained by the MaxEnt model is accurate and reliable. The results of the receiver operating characteristic curve are as Figure 2 shown.

[0066] After verifying and training the MaxEnt model using the contemporary data in the target environmental variables to accurately predict the distribution of mangroves, then input the future data in the target environmental variables into the MaxEnt model, use the jackknife method to analyze the contribution rate of each environmental variable in the model, predict the potential distribution of mangroves, and use the receiver operating characteristic curve to determine whether the MaxEnt model is accurately predicted.

[0067] Taking the data of the target environmental variables from 2041 to 2060 under the CO2 emission scenario of RCP 2.6 as an example for the future data, the contribution degree of the target environmental variables affecting the future potential distribution of mangroves is shown in Table 4.

[0068] Table 4

[0069] Variable code Environmental variable Contribution rate / % bio14 Precipitation in the driest month (mm) 44.4 bio2 Mean diurnal range of temperature (°C) 18.1 bio7 Annual range of temperature (°C) 13.6 bio8 Mean temperature in the wettest season (°C) 10.1 bio12 Annual precipitation (mm) 7 bio9 Mean temperature in the driest season (°C) 3.4 bio3 Isothermality (×100) 1.5 bio15 Coefficient of variation of precipitation seasonality (mm) 1.3 bio5 Highest temperature in the warmest month (°C) 0.7

[0070] The receiver operating characteristic curve of the potential distribution of mangroves under the RCP 2.6 carbon emission concentration from 2041 to 2060 is as Figure 3 shown.

[0071] Other specific data of the target environmental variables can be used to predict the potential distribution of mangroves according to the above steps, which will not be described in detail here.

[0072] It should be noted that the contemporary data is for the target environmental variables from 1970 to 2000, and the future data is for the target environmental variables from 2041 to 2060 or from 2081 to 2100. Among them, the future data includes four CO2 emission scenarios: RCP 2.6, RCP 4.5, RCP 7.0, and RCP 8.5. Since the future CO2 emission scenarios are uncertain, by introducing the CO2 emission scenarios and predicting the potential distribution of mangroves under different CO2 emission scenarios, the potential distribution of mangroves can be predicted more accurately.

[0073] Preferably, it further includes step 4: inputting the prediction data into ArcGIS, and using the natural break classification method to draw the distribution map of the potential suitable areas of mangroves.

[0074] In this embodiment, the prediction data of the maxent model is input into the ArcGIS software. The ArcGIS software divides the distribution area according to the prediction data, and then uses the natural break classification method to divide the distribution area of mangroves into three levels, namely: high suitability (FI > 0.7), medium suitability (0.7 ≥ FI > 0.35), and low suitability (0.35 ≥ FI > 0), so as to draw the distribution map of the suitable areas of mangroves, which is convenient for the staff to observe the potential distribution of mangroves.

[0075] The core of the present invention lies in:

[0076] 1. Calculate the correlation coefficients between all environmental variables using the Pearson correlation formula, and eliminate the environmental variables with a correlation coefficient > |0.8|, so as to avoid overfitting caused by multiple environmental variables with a correlation coefficient > |0.8| being input into the maxent model and improve the prediction accuracy;

[0077] 2. Conditionally eliminate the environmental variables with a correlation coefficient > |0.8| using empirical conditions, and improve the efficiency of the model while ensuring the prediction accuracy of the maxent model.

[0078] Embodiment 2

[0079] Reference Figure 4 , a prediction device for the potential distribution of mangroves, includes the following units:

[0080] Data acquisition unit: used to acquire the mangrove site distribution information and environmental variables;

[0081] Data processing unit: used to perform Pearson correlation calculation between environmental variable data and screen out the target environmental variables;

[0082] Distribution prediction unit: It is used to input the target environmental variables and the mangrove site distribution information into the MaxEnt model, and the MaxEnt model outputs prediction data.

[0083] Preferably, it further includes a distribution map making unit: It is used to input the prediction data into ArcGIS, and use the natural break classification method to draw the mangrove potential suitable area distribution map.

[0084] The specific working process of this prediction device is as follows: First, the data acquisition unit acquires the contemporary mangrove site distribution information, the contemporary data and future data of environmental variables, and sends the relevant data to the data processing unit. The data processing unit first calculates the Pearson correlation of the environmental variables, screens out the target environmental variables with a correlation < |0.8|, and inputs the contemporary data of the target environmental variables and the contemporary mangrove site distribution information into the MaxEnt model for verification training, and uses the receiver operating characteristic curve for verification. After verification, the future data in the target environmental variables and the CO2 emission situation are input into the MaxEnt model to predict the future potential distribution of mangroves, obtain the prediction data and send the prediction data to the distribution map making unit. The distribution map making unit divides the distribution area according to the prediction data, and uses the natural break classification method to divide the distribution map into different levels and draw the mangrove potential suitable area distribution map.

[0085] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the claims and their equivalents.

Claims

1. A method for predicting the potential distribution of mangroves, characterized in that: The following steps are involved: Step 1: Obtain mangrove site distribution information and environmental variables; Step 2: Calculate the Pearson correlation between environmental variable data to screen out target environmental variables; Step 3: Input the target environmental variables and mangrove site distribution information into the Maxent model, and the Maxent model outputs the predicted data.

2. The method for predicting the potential distribution of mangroves according to claim 1, characterized in that: It also includes step 4: input the predicted data into ArcGIS, use the natural breakpoint classification method to draw a distribution map of potential suitable areas for mangroves.

3. The method for predicting the potential distribution of mangroves according to claim 1, characterized in that: The environmental variables include annual average temperature, daily range of average temperature, isothermality, standard deviation of seasonal temperature variation, maximum temperature in the warmest month, minimum temperature in the coldest month, annual temperature range, average temperature in the wettest quarter, average temperature in the driest quarter, average temperature in the warmest quarter, average temperature in the coldest quarter, annual precipitation, precipitation in the wettest month, precipitation in the driest month, seasonal coefficient of variation of precipitation, precipitation in the wettest quarter, precipitation in the driest quarter, precipitation in the warmest quarter, and precipitation in the coldest quarter.

4. The method for predicting the potential distribution of mangroves according to claim 3, characterized in that: The environmental variables also include the nitrogen content of the ocean surface, the lowest water temperature of the ocean surface, the average water temperature of the ocean surface, the phosphorus content of the ocean surface, the average salinity of the ocean surface, the pH value of the ocean surface, and the highest water temperature of the ocean surface.

5. The method for predicting the potential distribution of mangroves according to claim 1, characterized in that: The specific operation of step 2 is: perform Pearson correlation calculation between environmental variables, calculate the Pearson correlation coefficient r between environmental variables, and when r>|0.8|, eliminate the corresponding environmental variables according to empirical conditions, so as to screen out the target environmental variables.

6. The method for predicting the potential distribution of mangroves according to claim 5, characterized in that: The experience conditions are: When two of the environmental variables are of different types, the environmental variable related to precipitation is eliminated first, and the environmental variable related to temperature is retained; When two of the environment variables are of the same type, the environment variable with the largest number of r>|0.8| among the other environment variables is preferentially eliminated.

7. The method for predicting the potential distribution of mangroves according to claim 1, characterized in that: The specific operation of step 3 is: input the target environmental variables and mangrove site distribution information into the maxent model, use the jackknife method in the maxent model to calculate the target environmental variables, and obtain the contribution rate of the target environmental variables. According to the contribution rate of the target environmental variables, the maxent model outputs the predicted data.

8. The method for predicting the potential distribution of mangroves according to claim 7, characterized in that: The step 3 also includes: using a receiver operating characteristic curve to verify the prediction accuracy of the maxent model on the prediction data.

9. A device for predicting the potential distribution of mangroves, characterized in that: The following units are included: Data acquisition unit: used to obtain mangrove site distribution information and environmental variables; Data processing unit: used to calculate the Pearson correlation between environmental variable data and filter out target environmental variable data; Distribution prediction unit: used to input target environmental variables and mangrove site distribution information into the Maxent model, which outputs predicted data.

10. The device for predicting the potential distribution of mangroves according to claim 9, characterized in that: It also includes a distribution map production unit: it is used to input the predicted data into ArcGIS, and use the natural breakpoint classification method to draw a distribution map of potential suitable areas for mangroves.

Images