A method for predicting the distribution of vegetation in coastal wetlands

Abstract

This invention relates to a method for predicting the distribution of vegetation in coastal wetlands. The purpose of this invention is to address the problem of low accuracy in predicting the distribution of vegetation in coastal wetlands using existing models. The process is as follows: 1. Obtain land cover classification data; 2. Obtain a marine environmental dynamic dataset for the same region; 3. Construct a coastal wetland MIKE 21 hydrodynamic model, with the marine environmental dynamic dataset of the coastal wetland region as input and the spatial distribution of current velocity, current direction, water depth, salinity, and temperature of the coastal wetland as output; 4. Obtain environmental driving factors for the same region; 5. Construct a coastal wetland PLUS model, with land cover classification data and environmental driving factors for the same region as input and the wetland land use type prediction result as output. This invention is applicable to the field of coastal wetland vegetation distribution prediction.

Description

Technical Field

[0001] This invention relates to a method for predicting the distribution of vegetation in coastal wetlands. Background Technology

[0002] Coastal wetlands refer to waters with a depth of less than six meters at low tide and their surrounding wetlands, including permanent waters with a depth not exceeding six meters, intertidal zones (or floodplains), and coastal lowlands. Coastal wetlands possess significant resource, ecological, and environmental benefits. Besides their functions in degrading pollutants, protecting dikes from floods, and providing habitats and food sources for wetland animals, coastal wetlands also have a powerful carbon sequestration function, playing a crucial role in reducing atmospheric carbon dioxide (CO2) concentrations and mitigating global climate change. In recent years, coastal wetlands have faced challenges due to human activities such as land reclamation, port development, and land-based pollution, resulting in a sharp decline in wetland area and degradation of wetland ecosystems. A series of countermeasures are urgently needed to protect coastal wetland resources.

[0003] Ecosystem simulation is an important means of understanding, learning about, and protecting nature. Wetland vegetation provides primary productivity for wetland ecosystems and is also a food source for many birds and benthic organisms. The distribution of wetland vegetation reflects the health status of wetlands to a certain extent. By simulating changes in key wetland vegetation, we can effectively monitor wetland landscape evolution trends and provide strong support for wetland management and planning.

[0004] Coastal wetlands, located at the land-sea interface, are influenced by a multitude of factors, including meteorology, tides, rivers, and land-based pollutant emissions. Similarly, the growth and distribution of wetland vegetation are determined by factors such as meteorology, soil conditions, tides, animal foraging, population competition, and human activities. Therefore, simulating the growth and distribution of wetland vegetation by examining its various mechanistic processes presents significant challenges. Domestic and international scholars have largely utilized cellular automata principles to simulate the distribution patterns of wetland vegetation, achieving promising results. Cellular automata are discrete-time, spatial, and state-based grid dynamics models, with localized spatial interactions and temporal causality. They are renowned for their powerful ability to simulate the spatiotemporal evolution of complex systems. Since the 1960s, cellular automata have been widely applied in geography and ecology. East China Normal University used cellular automata based on the mechanisms of seed formation, dispersal, and growth to simulate the spatial distribution of Spartina alterniflora and Phragmites australis in the Chongming Dongtan Wetland, achieving good results, but lacking predictions for future scenarios.

[0005] Geographic cellular automata models such as GeoSOS, FLUS, and PLUS, developed by teams from Sun Yat-sen University and China University of Geosciences, have been widely applied in urban expansion, land use, and ecological protection, providing relatively accurate land use predictions and offering scientific guidance for ecological management. Currently, geographic cellular automata are widely used in inland areas, but research on estuaries and coastal regions is rarely reported. Coastal wetland environments are complex, and compared to inland ecosystems, variables such as tides and salinity have a more significant impact on vegetation growth. Therefore, when simulating vegetation distribution in coastal wetlands, it is necessary to consider their unique hydrodynamic processes to ensure the accuracy of the simulation results. Summary of the Invention

[0006] The purpose of this invention is to address the problem of low accuracy in predicting the distribution of vegetation in coastal wetlands using existing models, and to propose a method for predicting the distribution of vegetation in coastal wetlands.

[0007] The specific process of a method for predicting the distribution of vegetation in coastal wetlands is as follows:

[0008] Step 1: Collect Landsat 8 satellite images of the coastal wetland area, segment the Landsat 8 satellite images of the coastal wetland area, and obtain land cover classification data;

[0009] Step 2: Obtain the marine environmental dynamics dataset for the same region as in Step 1;

[0010] Step 3: Construct the MIKE 21 hydrodynamic model of the coastal wetland. The input of the MIKE 21 hydrodynamic model of the coastal wetland is the marine environmental dynamic dataset of the coastal wetland area. The output of the MIKE 21 hydrodynamic model of the coastal wetland is the current velocity, current direction, water depth, salinity, and spatial distribution of temperature of the coastal wetland.

[0011] Step 4: Obtain the environmental driving factors for the same region as in Step 1;

[0012] Step 5: Construct the coastal wetland PLUS model. The input of the PLUS model is the land cover classification data and the environmental driving factors of the same area. The output of the PLUS model is the wetland land use type prediction result.

[0013] Preferably, in step one, Landsat 8 satellite imagery of the coastal wetland area (https: / / www.gscloud.cn / search) is acquired, and the Landsat 8 satellite imagery of the coastal wetland area is segmented to obtain land cover classification data; the specific process is as follows:

[0014] Using ENVI 5.3 software, Landsat 8 satellite imagery of the coastal wetland area was segmented into six land use types: water area, mudflats, Suaeda salsa, farmland, reeds, and building land.

[0015] Preferably, in step two, the marine environmental dynamics dataset of the same region as in step one is obtained; the specific process is as follows:

[0016] The marine environmental dynamics dataset comes from the website of the European Copernicus Marine Environmental Monitoring Service.

[0017] Preferably, in step three, a MIKE 21 hydrodynamic model of the coastal wetland is constructed. The input of the MIKE 21 hydrodynamic model is a marine environmental dynamic dataset of the coastal wetland area, and the output of the MIKE 21 hydrodynamic model is the spatial distribution of the coastal wetland's current velocity, current direction, water depth, salinity, and temperature. The specific process is as follows:

[0018] Obtain topographic data of the coastal wetland area;

[0019] The terrain data was provided through the UK Marine Data Centre website and has a spatial resolution of 400m.

[0020] A two-dimensional mesh for coastal wetlands was constructed using the MIKE 21 model, with the mesh spatial resolution A satisfying: 100m≤A≤1km;

[0021] The input for the MIKE 21 hydrodynamic model of coastal wetlands is: a dataset of marine environmental dynamics of the coastal wetland area;

[0022] The output of the MIKE 21 hydrodynamic model for coastal wetlands includes: the flow velocity, flow direction, water depth, salinity, and spatial distribution of temperature in the coastal wetlands.

[0023] The flow velocity, flow direction, water depth, salinity, and temperature of the coastal wetland, output from the MIKE 21 hydrodynamic model, were converted into ".shp" files using the Mike2Shp tool, and then converted into tif images using ArcGIS software.

[0024] Preferably, in step four, the environmental driving factors of the same region as in step one are obtained; the specific process is as follows:

[0025] Environmental driving factors include hydrological factors, topographic factors, and location factors;

[0026] The hydrological factors are: the water depth, salinity, temperature, flow velocity, and flow direction of the coastal wetland as output by the MIKE 21 hydrodynamic model.

[0027] Topographical factors include: elevation, slope, and aspect of the coastal wetland area;

[0028] Location factors are: the distances from each grid within the coastal wetland area to roads, rivers, buildings, and coastlines within the area;

[0029] The hydrological, topographic, and location factor files need to be cropped to the same area and maintain the same TIFF format and image resolution.

[0030] Preferably, in step five, a coastal wetland PLUS model is constructed. The input to the PLUS model is land cover classification data and environmental driving factors of the same area, and the output of the PLUS model is the wetland land use type prediction result; the specific process is as follows:

[0031] Step 51: Extraction of land expansion documents:

[0032] Extract change files for each land use type over N years;

[0033] The land use types are water areas, tidal flats, Suaeda salsa, farmland, reeds, and building land;

[0034] Step 52: Based on land expansion documents and environmental driving factors, generate the development probability of each land use type and the contribution of each environmental driving factor to the change of each land use type; the specific process is as follows:

[0035] In the Land Expansion Analysis Strategy (LEAS) module of the PLUS model, input the land expansion file and environmental drivers to generate the development probability of each land use type and the contribution of each environmental driver to the change of each land use type.

[0036] Step 53: Land use simulation and prediction; the specific process is as follows:

[0037] Based on the number of pixels for each land use type in the first two periods, the Markov Chain in the PLUS model is used to predict the number of pixels for each land use type in future years.

[0038] Set up a conversion matrix where "1" indicates that land use type M can be converted to land use type N, and "0" indicates the opposite.

[0039] The predicted number of pixels for each type of land in future years, the transformation matrix, and the development probability of each land use type generated in step 52 are input into the cellular automata module of the PLUS model. The cellular automata module of the PLUS model outputs the distribution pattern of each type of land in future years.

[0040] Preferably, the process of obtaining the elevation, slope, and aspect is as follows:

[0041] Download the DEM data and use ArcGIS software to process the DEM data to generate the elevation, slope, and aspect of the coastal wetland area.

[0042] Preferably, the process of obtaining the distances from each grid point to roads, rivers, buildings, and coastlines within the coastal wetland area is as follows:

[0043] First, using ArcGIS software, the outlines of roads, rivers, and coastlines are drawn in the satellite imagery. Then, the Euclidean distance tool is used to calculate the distance from each pixel in the satellite imagery to the outlines of the roads, rivers, and coastlines.

[0044] The beneficial effects of this invention are as follows:

[0045] This invention relates to the field of marine ecological modeling technology, specifically to a method for studying coastal wetland vegetation distribution based on a coupled "physical-biological" model. The patch-forming land use change simulation software PLUS, developed by Liang Xun's team at China University of Geosciences, has been used to study land use changes in inland areas such as cities, farmland, and forests. However, research on the evolutionary competition patterns of vegetation in estuaries or coastal wetlands is limited. This invention aims to utilize the PLUS model coupled with the two-dimensional hydrodynamic model MIKE 21 from the MIKE water environment software developed by the Danish Hydraulic Institute (DHI) to explore the processes of vegetation expansion, competition, and decline in important estuarine wetlands under the influence of natural causes and human activities. A two-dimensional hydrodynamic model of estuarine wetlands is constructed based on MIKE 21. The model's output variables, such as salinity, temperature, and water depth, are used as input variables for the PLUS model. Combined with meteorological, topographical, and locational factors (distance to the coastline and distance to the river channel), a complete environmental driving factor system for coastal wetlands is established, which contributes to the accurate simulation and prediction of vegetation distribution. By combining vegetation classification results from multiple satellite images with an environmental driving factor system, the input of the PLUS model is constructed. After training by the machine learning module inside the model, the model can output the vegetation distribution for the predicted year, providing scientific guidance for the restoration of habitats for coastal wetland species and the construction of ecological security.

[0046] The coastal wetland vegetation distribution research method proposed in this invention, based on a "physical-biological" element coupling model, has the following advantages compared with existing related technologies:

[0047] (1) Current simulation studies on coastal wetland ecosystems mostly focus on environmental factors, such as hydrodynamics, nutrients, chlorophyll, and planktonic plants and animals. There are few simulation and prediction studies on the distribution of key wetland vegetation. This invention uses the MIKE21 hydrodynamic model coupled with the PLUS land use model to simulate changes in wetland land use types, focuses on the relationship between various wetland elements, and realizes the "environment-biology" driven response in the model, which is conducive to the scientific planning and deployment of coastal wetland ecological restoration work.

[0048] (2) While the PLUS model has been widely used in inland urban expansion, farmland construction, and forestry monitoring, research on coastal wetland protection and wetland vegetation community evolution is scarce. This invention applies the PLUS model to land type change analysis in coastal wetlands, focusing on the relationship between natural wetlands and human activities, as well as the competitive evolution patterns of different wetland vegetation. The model can output the spatiotemporal distribution of various wetland vegetation types, expanding the simulation capabilities of PLUS and providing a scientific reference for wetland ecological restoration in my country.

[0049] (3) Compared with the application cases of the PLUS model in inland ecosystems, this invention has made a specific design for the environmental driving factors of the model input based on the growth characteristics of coastal wetland vegetation. That is, by constructing the MIKE 21 coastal wetland hydrodynamic model, the spatial distribution of hydrological elements in the study area, such as salinity, temperature, and flooding depth, is obtained, which improves the input elements of the PLUS model and enhances the simulation accuracy of the model. Attached Figure Description

[0050] Figure 1 A diagram illustrating the architecture of a "physical-biological" coupled coastal wetland vegetation distribution model.

[0051] Figure 2 This is a simulation area map of the MIKE 21 model. Bathymetry is the water depth measurement, deg is the degree Celsius, Above is above, below is below, and UndefinedValue is an invalid value.

[0052] Figure 3 This is a simulation result diagram of land use types in the Liaohe River Estuary wetland based on a coupled model;

[0053] Figure 4 The image shows the simulation results of the PLUS model. Detailed Implementation

[0054] Specific Implementation Method 1: The specific process of this implementation method for predicting the distribution of vegetation in coastal wetlands is as follows:

[0055] Physics: the flow velocity, flow direction, water depth, salinity, and temperature of coastal wetlands;

[0056] Biology: Represents plant types and their distribution areas;

[0057] Step 1: Collect Landsat 8 satellite imagery of the coastal wetland area (https: / / www.gscloud.cn / search), segment the Landsat 8 satellite imagery of the coastal wetland area, and obtain land cover classification data;

[0058] Step 2: Obtain the marine environmental dynamics dataset for the same region as in Step 1;

[0059] Step 3: Construct the MIKE 21 hydrodynamic model of the coastal wetland. The input of the MIKE 21 hydrodynamic model of the coastal wetland is the marine environmental dynamic dataset of the coastal wetland area. The output of the MIKE 21 hydrodynamic model of the coastal wetland is the current velocity, current direction, water depth, salinity, and spatial distribution of temperature of the coastal wetland.

[0060] Step 4: Obtain the environmental driving factors for the same region as in Step 1;

[0061] Step 5: Construct the coastal wetland PLUS model. The input of the PLUS model is the land cover classification data and the environmental driving factors of the same area. The output of the PLUS model is the wetland land use type prediction result.

[0062] Specific Implementation Method Two: This implementation method differs from Specific Implementation Method One in that, in step one, Landsat 8 satellite imagery of the coastal wetland area (https: / / www.gscloud.cn / search) is collected, and the Landsat 8 satellite imagery of the coastal wetland area is segmented to obtain land cover classification data; the specific process is as follows:

[0063] Using ENVI 5.3 software, Landsat 8 satellite images of the coastal wetland area from 2019 to 2021 were segmented into six land use types: water area, mudflat, Suaeda salsa, farmland, reeds, and building land.

[0064] The other steps and parameters are the same as in Specific Implementation Method 1.

[0065] Specific Implementation Method Three: This implementation method differs from Specific Implementation Method One or Two in that step two involves acquiring the marine environmental dynamics dataset of the same region as in step one; the specific process is as follows:

[0066] The marine environmental dynamics dataset comes from the website of the European Copernicus Marine Environmental Monitoring Service (CMEMS, https: / / resources.marine.copernicus.eu / ).

[0067] Other steps and parameters are the same as in specific implementation method one or two.

[0068] Specific Implementation Method Four: This implementation method differs from Specific Implementation Methods One to Three in that, in step three, a coastal wetland MIKE 21 hydrodynamic model is constructed. The input of the coastal wetland MIKE 21 hydrodynamic model is the marine environmental dynamic dataset of the coastal wetland area, and the output of the coastal wetland MIKE 21 hydrodynamic model is the spatial distribution of the coastal wetland's flow velocity, flow direction, water depth, salinity, and temperature.

[0069] The specific process is as follows:

[0070] Obtain topographic data of the coastal wetland area;

[0071] The terrain data is provided through the UK Marine Data Centre website (GEBCO, https: / / www.gebco.net / ) with a spatial resolution of 400m;

[0072] A two-dimensional mesh for coastal wetlands was constructed using the MIKE 21 model, with the mesh spatial resolution A satisfying: 100m≤A≤1km;

[0073] The input for the MIKE 21 hydrodynamic model of coastal wetlands is: a dataset of marine environmental dynamics of the coastal wetland area;

[0074] The output of the MIKE 21 hydrodynamic model for coastal wetlands includes: the flow velocity, flow direction, water depth, salinity, and spatial distribution of temperature in the coastal wetlands.

[0075] The flow velocity, flow direction, water depth, salinity, and temperature of the coastal wetland, output from the MIKE 21 hydrodynamic model, were converted into ".shp" files using the Mike2Shp tool, and then converted into tif images using ArcGIS software.

[0076] The simulation period for the MIKE 21 hydrodynamic model of the coastal wetland was one year, 2020.

[0077] The other steps and parameters are the same as those in one of the specific implementation methods one to three.

[0078] Specific Implementation Method Five: This implementation method differs from Specific Implementation Methods One to Four in that, in step four, the environmental driving factors (12 environmental driving factors) of the same region as in step one are obtained; the specific process is as follows:

[0079] Environmental driving factors include hydrological factors, topographic factors, and location factors;

[0080] The hydrological factors are: the water depth, salinity, temperature, flow velocity, and flow direction of the coastal wetland as output by the MIKE 21 hydrodynamic model.

[0081] Topographical factors include: elevation, slope, and aspect of the coastal wetland area;

[0082] Location factors are the distances from each grid within the coastal wetland area to spatial elements that have a certain impact on vegetation growth, such as roads, rivers, buildings, and coastlines within the area.

[0083] The hydrological, topographic, and location factor files need to be cropped to the same area and maintain the same TIFF format and image resolution.

[0084] The other steps and parameters are the same as those in one of the specific implementation methods one to four.

[0085] Specific Implementation Method Six: This implementation method differs from Specific Implementation Methods One to Five in that, in step five, a coastal wetland PLUS model is constructed. The input of the PLUS model is land cover classification data and environmental driving factors in the same area, and the output of the PLUS model is the wetland land use type prediction result.

[0086] The specific process is as follows:

[0087] Step 51: Extraction of land expansion documents:

[0088] Extract change files for each land use type over N years;

[0089] Land use types include water areas, tidal flats, Suaeda salsa, farmland, reeds, and building land, etc.

[0090] Using two years of land use data, we extracted the changes in each land use type over the two years. Data from 2019 and 2020 were selected for extraction.

[0091] Step 52: Based on land expansion documents and environmental driving factors, generate the development probability of each land use type and the contribution of each environmental driving factor to the change of each land use type; the specific process is as follows:

[0092] In the Land Expansion Analysis Strategy (LEAS) module of the PLUS model, input the land expansion files and environmental driving factors to generate the development probability of each land use type (6 files for 6 land types) and the contribution of each environmental driving factor to the change of each land use type (12 files for 12 environmental driving factors).

[0093] Step 53: Land use simulation and prediction; the specific process is as follows:

[0094] Based on the number of pixels for each land use type in the first two periods, the Markov Chain in the PLUS model is used to predict the number of pixels for each land use type in future years.

[0095] Set up a conversion matrix where "1" indicates that land use type M can be converted to land use type N, and "0" indicates the opposite.

[0096] The predicted number of pixels for each type of land in future years, the transformation matrix, and the development probability of each land use type generated in step 52 are input into the cellular automata (CARS) module of the PLUS model. The cellular automata (CARS) module of the PLUS model outputs the distribution pattern of each type of land in future years.

[0097] The other steps and parameters are the same as those in one of the specific implementation methods one to five.

[0098] Specific Implementation Method Seven: This implementation method differs from Specific Implementation Methods One through Six in that the process of obtaining the elevation, slope, and aspect is as follows:

[0099] Download the DEM data (https: / / www.gscloud.cn / search) and use ArcGIS software to process the DEM data to generate the elevation, slope, and aspect of the coastal wetland area.

[0100] The other steps and parameters are the same as those in one of the specific implementation methods one to six.

[0101] Specific Implementation Method Eight: This implementation method differs from Specific Implementation Methods One to Seven in that the process of obtaining the distances from each grid within the coastal wetland area to spatial elements that have a certain impact on vegetation growth, such as roads, rivers, buildings, and coastlines, is as follows:

[0102] First, using ArcGIS software, the outlines of roads, rivers, and coastlines are drawn in the satellite imagery. Then, the Euclidean distance tool is used to calculate the distance from each pixel in the satellite imagery to the outlines of the roads, rivers, and coastlines.

[0103] The other steps and parameters are the same as those in any of the specific implementation methods one to seven.

[0104] The beneficial effects of the present invention are verified using the following embodiments:

[0105] Example 1:

[0106] This invention takes the Liaohe River estuary wetland as an example to introduce the implementation process of a coastal wetland vegetation distribution model based on the coupling of "physical-biological" elements, such as... Figure 1 As shown.

[0107] (1) Construction of the MIKE 21 hydrodynamic model of the Liaohe River Estuary Wetland; the specific process is as follows:

[0108] A two-dimensional mesh for the Liaohe River estuary wetland was constructed using the MIKE 21 model, with a spatial resolution of 100m-1km. The simulation range was 121.8316-122.114 degrees east longitude and 40.7999-40.8996 degrees north latitude. Figure 2 The topographic data was provided by the UK Marine Data Centre website (GEBCO, https: / / www.gebco.net / ), with a spatial resolution of 400m. To accurately simulate temperature and salinity variations, modules for evaporation, precipitation, and atmospheric heat exchange were added to the model. Water level, current velocity, and temperature and salinity boundary condition data were obtained from the European Copernicus Marine Environment Monitoring Service website (CMEMS, https: / / resources.marine.copernicus.eu / ). The model was run for the entire year of 2020.

[0109] The MIKE 21 model can output the spatial distribution of parameters such as water level, flow velocity, flow direction, water depth, salinity, and temperature in the Liaohe River Estuary wetland on an hourly basis. Considering that soil salinity and flooding depth are key factors affecting the growth of vegetation in salt marshes, the water depth, salinity, and temperature output by the MIKE 21 model were selected as environmental driving factors input into the PLUS model. To explore the differences in the impact of environmental factors on vegetation growth at different times, the average water depth, salinity, and temperature data during the vegetation germination period from February to April 2020, and the data corresponding to the rapid vegetation growth period from May to July, were selected and input into the PLUS model. Since the environmental factor data input to the PLUS model is in ".tif" format, the simulation results of MIKE 21 need to be converted to ".shp" files using the Mike2Shp tool, and then converted into tif images using ArcGIS software.

[0110] (2) Construction of the Liaohe River Estuary Wetland PLUS Model; the specific process is as follows:

[0111] Before running the PLUS model, data preparation is required. The input data for the PLUS model consists of two parts: multi-period land use classification data and the distribution of environmental driving factors in the same region. The land use classification data is derived from Landsat 8 satellite imagery of the Liaohe River Estuary region (https: / / www.gscloud.cn / search) using ENVI 5.3 software image segmentation. Specifically, it is divided into six land use types: water area, mudflat, Suaeda salsa, farmland, reeds, and built-up land. Three periods of satellite imagery from 2019 to 2021 are selected as the input for the PLUS model.

[0112] In addition to hydrological factors such as water depth, salinity, and temperature output by the MIKE 21 model, environmental driving factors also include some topographic and locational factors.

[0113] Download DEM data (https: / / www.gscloud.cn / search) and use ArcGIS software to generate topographic factors such as elevation, slope, and aspect of the simulated area.

[0114] Location factors refer to the distances from each grid within the simulated area to spatial elements that influence vegetation growth, such as roads, rivers, buildings, and coastlines. The simulated area has relatively few buildings, but includes one river (Liaohe River) and a coastal highway. Therefore, ArcGIS software was first used to depict the outlines of roads, rivers, and coastlines in satellite imagery, and the Euclidean distance tool was used to calculate the distances from each pixel in the image to these outlines.

[0115] The average water depth, salinity, and temperature within the study area from February to April; the average water depth, salinity, and temperature from May to July; elevation, slope, and aspect; and distances to roads, rivers, and coastlines constitute the environmental driving system of the PLUS model. The spatial distribution files of the 12 environmental driving factors need to be cropped to the same region and kept in a consistent format (tif) and image resolution.

[0116] (3) Prediction of wetland vegetation distribution; the specific process is as follows:

[0117] The simulation prediction of the PLUS model consists of three steps:

[0118] a. Land expansion extraction:

[0119] Using two years of land use data, we extracted the changes in each land use type over the two years. Data from 2019 and 2020 were selected for extraction.

[0120] b. Land Expansion Analysis:

[0121] In the Land Expansion Analysis Strategy (LEAS) module of the PLUS model, input the land expansion files and environmental drivers calculated in the previous step to generate the development probability of each land use type (6 files for 6 land types) and the contribution of each environmental driver to the change of each land use type (12 files for 12 environmental drivers).

[0122] c. Land use simulation and prediction:

[0123] Before land use simulation, it is necessary to obtain the number of pixels for each type of land and the land type transformation matrix for the predicted year. The Markov Chain in the PLUS model can be used to predict the number of pixels for each type of land in the future.

[0124] The transformation matrix represents the transformation relationship between different land use types. A "1" in the matrix indicates that land use type M can be transformed into land use type N, and a "0" indicates the opposite. Different environmental policies will produce different transformation matrices. For example, this invention adopts a wetland protection policy, so wetland vegetation and tidal flats cannot be transformed into farmland and building land. The transformation matrix used in this example is shown in Table 1.

[0125] Table 1 Transformation Matrix

[0126] waters mudflats Suaeda salsa farmland reed Construction land waters 1 1 1 0 1 0 mudflats 1 1 1 0 1 0 Suaeda salsa 1 1 1 0 1 0 farmland 0 0 0 1 0 0 reed 1 1 1 0 1 0 Construction land 0 0 0 0 0 1

[0127] After determining the number of pixels for each type and the transformation matrix for the predicted year, the development probability of each land use type obtained in step b is input into the cellular automata (CARS) module of the PLUS model, and the simulated land use results are output. Figure 3 ).

[0128] (4) PLUS simulation accuracy verification; the specific process is as follows:

[0129] The kappa coefficient and FoM coefficient were calculated using the accuracy verification module in the PLUS model to evaluate the consistency between the simulation results and the actual land use distribution.

[0130]

[0131] In the formula, k represents the kappa coefficient, p o It is the sum of the number of correctly classified samples in each class divided by the total number of samples, which is the overall classification accuracy. Let the number of true samples in each class be a1, a2, ..., a... c The predicted number of samples for each class are b1, b2, ..., b c If the total number of samples is n, then:

[0132]

[0133] The FoM coefficient is an indicator of the degree of conformity between the overall simulation results and the actual spatial pattern.

[0134]

[0135] In the formula, A represents the number of incorrectly predicted cells where the actual land use type changes but the land use type remains unchanged in the simulation results; B represents the number of correctly predicted cells where the land use type changes in both the actual and simulation results; C represents the number of incorrectly predicted cells where the actual land use type changes but the land use type is incorrectly predicted in the simulation results; and D represents the number of incorrectly predicted cells where the actual land use type remains unchanged but the land use type is incorrectly predicted in the simulation results.

[0136] (1) This invention couples the water environment model MIKE 21 and the land use model PLUS, and can be extended to various ecosystems in inland and coastal areas, such as typical vegetation growth areas like mangroves, salt marsh vegetation, and seagrass beds. This invention can also be applied to the expansion analysis of invasive vegetation, such as Spartina alterniflora and Spartina simonii.

[0137] (2) The environmental driving factors provided by the MIKE 21 model in this invention are only some basic hydrological parameters (temperature, salinity, and depth). Corresponding ecological modules can be added later to simulate the impact of nutrient levels in water bodies on vegetation growth. This invention does not involve soil parameters. The MIKE SHE model in the MIKE software can be used to simulate the soil water and salt content in the Liaohe River Basin, thereby improving the environmental driving system of the PLUS model.

[0138] The PLUS model is coupled with a Markov chain, which can output the number of pixels for each land use type in the predicted year based on the number of pixels for each land use type in the previous two land use results. PLUS uses a Random Forest classification algorithm to explore the contribution of each driving factor to the change of different land use types and the development probability distribution of each land use type (Equation (1)). The contribution of each driving factor for each land use type is determined by the Variable Importance Measures (VIMs) method in the OOB error estimation of the Random Forest algorithm.

[0139] This algorithm can solve the problem of multicollinearity among multiple variables by drawing random samples from the original dataset and finally determining the probability of k land use types (e.g., 6 types) appearing on grid i.

[0140]

[0141] In the formula: the value of d is 0 or 1;

[0142] If d = 1, it means that other land use types have been transformed into the kth land use type;

[0143] When d = 0, it means that the land use type has changed to a land use type other than the kth type.

[0144] x is a vector composed of several environmental driving factors (12 in total);

[0145] Function I is the indicator function of the decision tree set;

[0146] h n (x) is the prediction type of the nth decision tree for vector x;

[0147] M represents the total number of decision trees.

[0148] The calculated land development probability, combined with the previous Markov chain results, is input into the CA cellular automata module of the PLUS model to obtain the land use pattern for the predicted year.

[0149] This invention may have other embodiments. Without departing from the spirit and essence of this invention, those skilled in the art can make various corresponding changes and modifications according to this invention, but these corresponding changes and modifications should all fall within the protection scope of the appended claims.

Claims

1. A method for predicting the distribution of vegetation in coastal wetlands, characterized in that: The specific process of the method is as follows: Step 1: Collect Landsat 8 satellite images of the coastal wetland area, segment the Landsat 8 satellite images of the coastal wetland area, and obtain land cover classification data; Step 2: Obtain the marine environmental dynamics dataset for the same region as in Step 1; Step 3: Construct the MIKE 21 hydrodynamic model of the coastal wetland. The input of the MIKE 21 hydrodynamic model of the coastal wetland is the marine environmental dynamic dataset of the coastal wetland area. The output of the MIKE 21 hydrodynamic model of the coastal wetland is the current velocity, current direction, water depth, salinity, and spatial distribution of temperature of the coastal wetland. Step 4: Obtain the environmental driving factors for the same region as in Step 1; Step 5: Construct the coastal wetland PLUS model. The input to the PLUS model is land cover classification data and environmental driving factors of the same area. The output of the PLUS model is the wetland land use type prediction result. The specific process is as follows: Step 51: Extraction of land expansion documents: Extract change files for each land use type over N years; The land use types are water areas, tidal flats, Suaeda salsa, farmland, reeds, and building land; Step 52: Based on land expansion documents and environmental driving factors, generate the development probability of each land use type and the contribution of each environmental driving factor to the change of each land use type; the specific process is as follows: In the land expansion analysis strategy module of the PLUS model, input the land expansion file and environmental driving factors to generate the development probability of each land use type and the contribution of each environmental driving factor to the change of each land use type. Step 53: Land use simulation and prediction; the specific process is as follows: Based on the number of pixels for each land use type in the first two periods, the Markov chain in the PLUS model is used to predict the number of pixels for each land use type in future years. Set up a conversion matrix, where 1 in the matrix indicates that land use type M can be converted to land use type N, and 0 indicates the opposite. The pixel count of each type of land in the predicted future year, the transformation matrix, and the development probability of each land use type generated in step 52 are input into the cellular automata module of the PLUS model. The cellular automata module of the PLUS model outputs the distribution pattern of each type of land in the future year. In the land use expansion analysis strategy module of the PLUS model, the land expansion file and environmental driving factors are input to generate the development probability of each land use type, expressed as follows: in, For the first Land use types in grid The probability of development on; The value can be 0 or 1; like This indicates that other land use types have been converted to the first type. Land use type; when This indicates that the land use type has changed to, except for the first one. Other land use types besides Class A; It is a vector composed of environmental driving factors; function It is the indicator function of the decision tree set; It is a vector The The prediction type of a decision tree; This represents the total number of decision trees.

2. The method for predicting the distribution of coastal wetland vegetation according to claim 1, characterized in that: In step one, Landsat 8 satellite imagery of the coastal wetland area is acquired, and the Landsat 8 satellite imagery of the coastal wetland area is segmented to obtain land cover classification data; the specific process is as follows: Using ENVI 5.3 software, Landsat 8 satellite imagery of the coastal wetland area was segmented into six land use types: water area, mudflats, Suaeda salsa, farmland, reeds, and building land.

3. The method for predicting the distribution of coastal wetland vegetation according to claim 2, characterized in that: In step two, the marine environmental dynamics dataset for the same region as in step one is obtained; the specific process is as follows: The marine environmental dynamics dataset comes from the website of the European Copernicus Marine Environmental Monitoring Service.

4. The method for predicting the distribution of coastal wetland vegetation according to claim 3, characterized in that: In step three, a MIKE 21 hydrodynamic model of the coastal wetland is constructed. The input to the MIKE 21 hydrodynamic model is the marine environmental dynamic dataset of the coastal wetland area, and the output of the MIKE 21 hydrodynamic model is the spatial distribution of the coastal wetland's current velocity, current direction, water depth, salinity, and temperature. The specific process is as follows: Obtain topographic data of coastal wetland areas; The terrain data was provided through the UK Marine Data Centre website and has a spatial resolution of 400m. A two-dimensional mesh for coastal wetlands was constructed using the MIKE 21 model, with the mesh spatial resolution A satisfying: 100m≤A≤1km; The input for the MIKE 21 hydrodynamic model of coastal wetlands is: a dataset of marine environmental dynamics of the coastal wetland area; The output of the MIKE 21 hydrodynamic model for coastal wetlands includes: the flow velocity, flow direction, water depth, salinity, and spatial distribution of temperature in the coastal wetlands. The flow velocity, flow direction, water depth, salinity, and temperature of the coastal wetland, output from the MIKE 21 hydrodynamic model, were converted into ".shp" files using the Mike2Shp tool, and then converted into tif images using ArcGIS software.

5. The method for predicting the distribution of coastal wetland vegetation according to claim 4, characterized in that: In step four, the environmental driving factors of the same region as in step one are obtained; the specific process is as follows: Environmental driving factors include hydrological factors, topographic factors, and location factors; The hydrological factors are: the water depth, salinity, temperature, flow velocity, and flow direction of the coastal wetland as output by the MIKE 21 hydrodynamic model. Topographical factors include: elevation, slope, and aspect of the coastal wetland area; Location factors are: the distances from each grid within the coastal wetland area to roads, rivers, buildings, and coastlines within the area; The hydrological, topographic, and location factor files need to be cropped to the same area and maintain the same TIFF format and image resolution.

6. The method for predicting the distribution of coastal wetland vegetation according to claim 5, characterized in that: The process for obtaining the elevation, slope, and aspect is as follows: Download the DEM data and use ArcGIS software to process the DEM data to generate the elevation, slope, and aspect of the coastal wetland area.

7. The method for predicting the distribution of coastal wetland vegetation according to claim 6, characterized in that: The process of obtaining the distances from each grid point to roads, rivers, buildings, and coastlines within the coastal wetland area is as follows: First, using ArcGIS software, the outlines of roads, rivers, and coastlines are drawn in the satellite imagery. Then, the Euclidean distance tool is used to calculate the distance from each pixel in the satellite imagery to the outlines of the roads, rivers, and coastlines.

Images