Lake wetland migrant bird habitat inversion and optimization regulation and control method based on remote sensing
Through remote sensing interpretation and mathematical statistics, combined with generalized additive model, the water bodies, vegetation and land use types of lake and wetlands were analyzed, and the shortcomings of the analysis of changes in migratory bird populations and community structure in the existing technology were solved, and migratory bird habitat was realized The efficient and accurate inversion and optimization of regulation provide a scientific basis for the rational use of wetland resources and ecological protection.
Patent Information
- Application Number
- CN202411167702.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-23
- Publication Date
- 2024-12-03
AI Technical Summary
The existing technology is difficult to comprehensively analyze the key factors of changes in population and community structure of overwintering migratory birds in lakes and wetlands, and it is easy to ignore the mutual influence and interaction between influencing factors. The accuracy and quality of land use classification are insufficient, making it difficult to provide suitable survival for migratory birds. Optimal threshold condition.
Remote sensing interpretation and mathematical statistics method are used to extract water area by normalized differential water index method, vegetation coverage is extracted, random forest supervision classification method is used to classify land use types, and generalized additive model is constructed based on hydrological and meteorological data. , explore the main drivers of change and analyze their influencing effects.
Through the construction of optimized feedback algorithm and GAM model, the efficiency and accuracy of migratory bird habitat inversion are improved, the changes in wetland water bodies and vegetation are revealed, the impact of the environment on overwintering migratory birds is quantified, and the optimal threshold for suitable habitat for migratory birds is obtained. Promote the rational use of wetland resources and ecosystem protection.
Smart Images

Figure CN119066341A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the technical field of remote sensing inversion, and in particular relates to a remote sensing-based lake wetland migratory bird habitat inversion and optimization control method. Background Art
[0002] Wetlands are one of the most productive ecosystems on Earth and play an important role in conserving water resources, regulating climate, maintaining carbon cycles and protecting biodiversity. They can provide sufficient foraging and habitats for migratory birds. Taking Poyang Lake Wetland, a typical lake wetland in the middle and lower reaches of the Yangtze River, as an example, under the background of declining water levels, accelerated vegetation succession rates, sharp declines in biodiversity, and lack of food for overwintering migratory birds during the dry season, the habitat information of lake wetland water bodies, vegetation, and land use types that have a greater impact on migratory birds is comprehensively and rapidly inverted, the key factors affecting the population size and community structure of overwintering migratory birds are clarified, and the optimal threshold conditions suitable for the survival of overwintering migratory birds are analyzed, which can promote the rational use of wetland resources and the protection and restoration of wetland ecosystems.
[0003] Water bodies are one of the most sensitive indicators to environmental changes. They are closely related to the growth of wetland vegetation and the distribution areas of migratory birds. At present, monitoring water bodies based on remote sensing technology has become a mainstream method. Compared with traditional in-situ monitoring methods, the inversion of remote sensing data has the advantages of a wider range, real-time and high efficiency. The extraction methods mainly include: manual digital extraction, single-band threshold method and inter-spectral relationship method. In the application of the inter-spectral relationship method, McFeeters proposed the normalized difference water index (NDWI). NDWI can effectively enhance water information, but it is very sensitive to built-up land, but it may not be reliable for shallow water areas with boundary changes, which often leads to overestimation of water bodies. Therefore, based on NDWI, the improved differentiated water index method (MDNWI) was proposed and widely used. MNDWI is more suitable for enhancing water information than NDWI, and can extract water bodies more accurately. It is generally calculated using green wave and short-wave infrared (SWIR) bands.
[0004] Since the 20th century, remote sensing technology has been widely used to monitor the biophysical and biochemical characteristics of wetland vegetation (biomass, water content, etc. of wetland vegetation). Related vegetation indices (Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Fractional Vegetation Cover (FVC), etc.) are often used to reflect the growth of wetland vegetation and analyze the causes of changes. Changes in land use types are the biggest driving force for the degradation or disappearance of inland wetlands. Remote sensing technology can be used to quickly and accurately invert land use types and analyze the spatiotemporal distribution characteristics and influencing factors of wetland vegetation. Currently, remote sensing technology is used to classify wetland land use types. The most commonly used remote sensing data are Landsat series, Sentinel series, Gaofen-1, MO DIS data, etc. The main methods are: supervised classification, unsupervised classification, etc. Compared with supervised classification, unsupervised classification is relatively simple and does not require training samples, but the accuracy and quality of the classification results are not as high as supervised classification, and often require manual correction. Therefore, the current land use classification is mainly based on supervised classification. Among them, the random forest supervised classification method based on remote sensing is a powerful ensemble learning algorithm, which has significant advantages in land use classification. Compared with a single decision tree, random forest can process large-scale data sets, including high-dimensional features and a large number of samples, avoiding overfitting problems. This method is widely used in environmental monitoring and other fields, providing reliable technical support for land use research. The use of cloud platforms (Google EarthEngine, PIE-Engine, AI-Earth, etc.) can greatly improve the efficiency of inverting wetland vegetation information and land use types, but the current research on remote sensing inversion using cloud platforms in China is still in its initial development stage.
[0005] The generalized additive model (GAM) is widely used in the analysis of factors affecting the evolution of birds. In ecology, the GAM model is often used to describe the impact of environmental factors on biological communities. It is a relationship model between the mean of the response variable and the smooth function of the predictor variable established through a connection function. Each additive term in the model is estimated using a single smooth function. In each additive term, it can be explained how the dependent variable changes with the change of the independent variable, which enables the model to better reflect the changing characteristics of the data. At present, the analysis of the factors affecting the evolution of wetland wintering migratory birds is mostly based on the correlation between a single factor and the number of migratory birds or the impact on the changing trend of migratory bird habitats. It is rare to comprehensively consider all the influencing factors and give the optimal threshold conditions suitable for the survival of migratory birds, and it is easy to ignore the mutual influence and interaction between the influencing factors.
[0006] Based on the above technical problems existing in the prior art, the present invention proposes a method for inversion and optimization control of lake wetland migratory bird habitat based on remote sensing. Summary of the invention
[0007] The purpose of the present invention is to address the deficiencies in the prior art and to provide a remote sensing-based lake wetland migratory bird habitat inversion and optimization control method, which uses remote sensing interpretation and mathematical statistics to analyze the spatiotemporal succession characteristics of water bodies, vegetation and land use types and the changing patterns of migratory bird communities, and uses an improved optimization feedback algorithm to explore the main driving factors causing the changes and analyze their impact.
[0008] The present invention adopts the following technical solution:
[0009] A remote sensing-based lake wetland migratory bird habitat inversion and optimization control method, comprising:
[0010] Step 1, obtaining wetland water body and vegetation succession characteristic data and performing data preprocessing;
[0011] Step 2, extracting water body area using normalized difference water body index method;
[0012] Step 3, extracting the vegetation coverage of lake wetlands by pixel dichotomy;
[0013] Step 4, classify the land use types using random forest supervised classification;
[0014] Step 5: Analyze the changing patterns of population size and community structure of wintering migratory birds in lake wetlands; use hydrometeorological data, water area inverted by remote sensing, vegetation coverage and land use type data as influencing factors, and biological quantity as response variable. The influencing factors are obtained through two screenings, and the model parameters are adjusted through continuous feedback optimization to construct a GAM model to obtain the driving factors of the evolution of wintering migratory birds and the optimal threshold suitable for the survival of wintering migratory birds.
[0015] Furthermore, in step 1, data preprocessing includes atmospheric correction, mosaic cropping, cloud removal, image filling and stripe repair.
[0016] Further, in step 2, extracting the water area includes:
[0017] Step 21, calculate the MNDWI value for the wetland water area:
[0018]
[0019] In formula (1): GREEN is the green light band; SWIR1 is the short-wave infrared band;
[0020] Step 22, set NDWI threshold Value, extract water area image:
[0021]
[0022] In formula (2), MNDWI thresholdis the empirical threshold, ranging from 0 to 0.3;
[0023] Step 23, use image morphological operations to refine and fill the water area data, and then calculate the water area based on the resolution and number of pixels of the image:
[0024]
[0025] In formula (3): Water area is the water body area; is the number of pixels with Water=1; area pixel is the pixel size.
[0026] Furthermore, in step 3, extracting the vegetation coverage of lake wetlands includes:
[0027] Step 31, calculate the normalized difference vegetation index NDVI:
[0028]
[0029] In formula (4), NIR is the near infrared band; RED is the visible red band;
[0030] Step 32: Use NDVI to estimate vegetation coverage FVC by pixel dichotomy:
[0031]
[0032] In formula (5), NDVI min With NDVI max is the minimum and maximum value of NDVI in the area. To eliminate the influence of noise, NDVI min With NDVI max Take the 5% and 95% values of confidence for the cumulative frequency of NDVI values.
[0033] Furthermore, in step 4, classifying the land use type includes:
[0034] Step 41, based on the scientific investigation results and the main vegetation and land use types of the wetlands recorded in the literature, and comparing the remote sensing image features, a land use type classification system for the wetland study area is established;
[0035] Step 42, according to the land use type classification system of the wetland study area, in accordance with the principles of representativeness, uniformity and accuracy of sample point selection, combined with field investigation and literature research, and referring to existing achievements, a certain number of classified sample point data are directly selected on the remote sensing image through visual interpretation, where representativeness means that the sample points should fully represent the characteristics of the category to which they belong, and the sample points of different categories should have significant differences, and uniformity means that the sample points should be distributed as evenly as possible to prevent the classification algorithm from overfitting or neglecting specific areas;
[0036] Step 43, randomly divide the sample point data into a training data set and a validation data set, wherein 70% of the sample point data are used for training; 30% are used for validation, and the basic bands of the remote sensing image and the enhanced vegetation index EVI, NDVI and MNDWI after band operation are used as classification features of different land use types. Finally, the training data set is input into the random forest model for training:
[0037]
[0038] In formula (6), NIR is the near infrared band; RED is the visible red band; BLUE is the visible blue band; C1 is a constant of 6.0; C2 is a constant of 7.5;
[0039] Step 44, morphological processing is performed on the classification results to eliminate the influence of noise and smooth the image edge, wherein the morphological processing is to perform opening and closing operations, the opening operation is to first perform an erosion operation and then perform an expansion operation; the closing operation is to first perform an expansion operation and then perform an erosion operation to reduce or eliminate the boundary of the object in the image, by sliding the structural element on the image, the central pixel is 1 only when all pixels in the area covered by the structural element are 1, otherwise it is 0; and the expansion is the opposite, in order to expand or connect the object in the image, as long as the structural element intersects with any pixel in the image, the central pixel is set to 1;
[0040] Step 45, checking the classification accuracy, when the ACC (Accuracy) coefficient or Kappa coefficient of the classification result reaches more than 85%, the result is saved and plotted.
[0041] Further, step 5 includes:
[0042] Step 51: The collected hydrological, meteorological and remote sensing inversion data are regarded as influencing factors, and the wintering migratory bird data are regarded as response variables, and a single-factor GAM model of a single influencing factor and the number of wintering migratory birds is established, wherein the GAM model formula is:
[0043] g(μ)=α+f(X1+…+X i )……(7),
[0044] In formula (7), g() is the link function, α is the intercept, and μ is the mathematical expectation of the response variable, that is, μ=E(Y / X1,...,X i ), f(X i ) is a single explanatory variable X i A single variable function of ;
[0045] Step 52, considering the living habits of different wintering migratory birds, the correlation between the influencing factors and the migratory birds, and the parameters of the single factor model, preliminarily screen the key influencing factors for different wintering migratory birds, wherein the absolute value of the Pearson correlation coefficient is greater than 0.2, and the Pearson correlation analysis is shown in formula (8);
[0046]
[0047] In formula (8), R xy is the correlation coefficient between the number of migratory birds and the influencing factors, x i and i are the number of migratory birds and the impact factor value in the i-th year, and are the average number of migratory birds and the average impact factor over n years, n is the number of samples, R xy If it is greater than 0, it means that the two variables are positively correlated, and if it is less than 0, it means that the variables are negatively correlated; the larger the absolute value, the higher the correlation;
[0048] Step 53: In order to more comprehensively analyze the impact of influencing factors on migratory birds, the GAM model was used to explore the interaction between different influencing factors, and the interaction terms of influencing factors were added to the model. In order to avoid overfitting of the model, Pearson correlation analysis was used to exclude the repeated effects between influencing factors, and six main influencing factors were determined for each wintering migratory bird.
[0049] Step 54, selecting the optimal parameters of the model through the feedback optimization algorithm, establishing the optimal multi-factor mixed GAM model, and performing an F test on the established model;
[0050] Step 55: Use the model to analyze the changes in the number of wintering migratory bird populations under the combined effects of multiple factors, and obtain the optimal threshold values of the influencing factors suitable for migratory bird survival.
[0051] Furthermore, in step 5, the hydrological and meteorological data include average water level, average temperature, accumulated precipitation, and average sunshine time.
[0052] The beneficial effects of the present invention are:
[0053] 1. The remote sensing-based lake wetland migratory bird habitat inversion and optimization control method of the present invention, by exploring the changing laws of wetland water bodies, reveals the growth and succession of wetland vegetation, and is then used to quantify and analyze the impact of the current rapidly changing environment on wintering migratory birds and other wetland organisms, and obtain the optimal threshold value of suitable habitat for migratory birds;
[0054] 2. The remote sensing-based lake wetland migratory bird habitat inversion and optimization control method described in the present invention greatly improves the construction efficiency of the GAM model by optimizing the feedback algorithm. Combined with the comprehensive collection and analysis of influencing factors, the optimal threshold analysis of wetland biological habitat is faster and more accurate. BRIEF DESCRIPTION OF THE DRAWINGS
[0055] Figure 1 Schematic diagram of the process of the inversion and optimization control method of the habitat of migratory birds in lake wetlands based on remote sensing in an embodiment of the present invention;
[0056] Figure 2 A schematic diagram of the process of constructing a GAM model in an embodiment of the present invention;
[0057] Figure 3-1 This is a schematic diagram of the water area in region A during the dry season in 1990 in an embodiment of the present invention;
[0058] Figure 3-2 Schematic diagram of the water area of region A during the dry season in 2001 in an embodiment of the present invention;
[0059] Figure 3-3 Schematic diagram of the water area in region A during the dry season in 2010 in an embodiment of the present invention;
[0060] Figure 3-4 Schematic diagram of the water area in region A during the dry season in 2020 in an embodiment of the present invention;
[0061] Figure 4-1 This is a schematic diagram of the vegetation coverage in the dry season of Region A in 1990 in an embodiment of the present invention;
[0062] Figure 4-2 Schematic diagram of vegetation coverage in region A during the dry season in 2000 according to an embodiment of the present invention;
[0063] Figure 4-3 Schematic diagram of vegetation coverage in region A during the dry season in 2010 in an embodiment of the present invention;
[0064] Figure 4-4 Schematic diagram of vegetation coverage in region A during the dry season in 2020 in an embodiment of the present invention;
[0065] Figure 5-1 This is a schematic diagram of the dry season land use type in Region A in 1990 in an embodiment of the present invention;
[0066] Figure 5-2 It is a schematic diagram of the land use types in the dry season of region A in 2000 according to an embodiment of the present invention;
[0067] Figure 5-3 It is a schematic diagram of the land use type in the dry season of region A in 2010 in an embodiment of the present invention;
[0068] Figure 5-4 This is a schematic diagram of the land use type in the dry season in Region A in 2020 in an embodiment of the present invention;
[0069] Figure 6a Schematic diagram of the impact of BAR on certain migratory birds in an embodiment of the present invention;
[0070] Figure 6b Schematic diagram of the impact of MUD on certain migratory birds in an embodiment of the present invention;
[0071] Figure 6c Schematic diagram of the impact of TAI on certain migratory birds in an embodiment of the present invention;
[0072] Figure 6d Schematic diagram of the impact of LU on certain migratory birds in an embodiment of the present invention;
[0073] Figure 6e Schematic diagram of the impact of MAR on certain migratory birds in an embodiment of the present invention;
[0074] Figure 6f Schematic diagram of the impact of QI on certain migratory birds in an embodiment of the present invention. DETAILED DESCRIPTION
[0075] In order to more clearly understand the above-mentioned purposes, features and advantages of the present invention, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.
[0076] Example
[0077] like Figure 1 As shown, the remote sensing-based lake wetland migratory bird habitat inversion and optimization control method includes:
[0078] Step 1: Data preprocessing. The basic settings and data processing using PIE-engine cloud platform, ArcGIS 10.8 and ENVI 5.3 are as follows: setting the boundaries of the study area, selecting the study time period and screening the basic image data, atmospheric correction, mosaicking, cropping, and cloud removal of the basic image ("QA_PIXEL" band), using nearby time images to fill the cloud removal area of the basic image, and constructing the minimum cloud cover image set;
[0079] For example, the remote sensing data used for the remote sensing inversion of lakes and wetlands comes from the Landsat series secondary data on the PIE-engine cloud platform (https: / / engine.piesat.cn / );
[0080] Among them, for remote sensing data with poor data quality, such as data with stripes in Landsat-7, the Landsat-gapfill plug-in in ENVI5.3 is used to repair it, which improves the accuracy and quality of remote sensing inversion;
[0081] Step 2: Use the improved normalized difference water index method to extract the water area. For example, use the PIE-engine cloud platform and ArcGIS software to extract the water area based on the secondary data products of Landsat5, 7 and 8 surface reflectance data, and use the improved normalized difference water index (MNDWI) as follows:
[0082]
[0083] In the formula, GREEN is the green light band; SWIR1 is the short-wave infrared band. After calculating the MNDWI value, set the NDWI threshold The water area image is extracted by using image morphological operations, and the water area data is refined and filled. The water area is calculated by combining the image resolution and pixel number as follows:
[0084]
[0085]
[0086] Among them, MNDWI threshold is the empirical threshold, ranging from 0 to 0.3. area is the water body area; is the number of pixels with Water=1; area pixel is the pixel size;
[0087] Step 3: Extract vegetation coverage using pixel dichotomy. Fractional vegetation coverage (FVC) is a comprehensive quantitative indicator of the surface coverage of plant communities and an important parameter for describing vegetation communities and ecosystems. The normalized difference vegetation index (NDVI) is used to estimate FVC using pixel dichotomy. The calculation formula is as follows:
[0088]
[0089] Among them, NIR is the near infrared band, RED is the visible red band, NDVI1 is the NDVI value of the pixel completely covered by vegetation, NDVI0 is the NDVI value of the pixel with bare soil or no vegetation cover, and FVC is calculated by the following formula:
[0090]
[0091] Among them, NDVI min With NDVI max is the minimum and maximum value of NDVI in the area. To eliminate the influence of noise, NDVI min With NDVI max The 5% and 95% values of confidence level of cumulative frequency of NDVI values were taken;
[0092] Step 4: Random forest supervised classification method is used to classify land use types;
[0093] Based on the PIE-engine cloud platform and ArcGIS software, the land cover types of lake wetlands during the dry season from 1990 to 2022 were classified using the random forest supervised classification method. The random forest supervised classification constructs multiple decision trees for ensemble learning to process large-scale data sets and avoid overfitting problems.
[0094] Firstly, according to the scientific investigation results and the main vegetation and land use types of wetlands recorded in the literature, a land use type classification system for the wetland study area was established by comparing the remote sensing image features;
[0095] Taking Poyang Lake Wetland as an example (but the technical solution described in the embodiment of the present application is not limited to Poyang Lake Wetland), based on the results of the first Poyang Lake comprehensive scientific survey and the second Poyang Lake comprehensive scientific survey and the main vegetation and land use types of Poyang Lake Wetland in the relevant materials of the Jiangxi Provincial Water Resources Department, with reference to the distribution of various land use types in the research of relevant scholars, and comparing the remote sensing image characteristics, a land use type classification system for the Poyang Lake Wetland study area was established, as shown in Table 1, wherein bare land refers to exposed ground without plant growth; mudflat refers to flat land formed by silt accumulation in the water or by the water; swamp refers to muddy areas that have been soaked in stagnant water for a long time and are densely covered with aquatic plants; moss community, phalaenopsis and Polygonum community, and reed community refer to vegetation areas mainly composed of moss, phalaenopsis and Polygonum, and reed as community species; other vegetation types refer to vegetation areas such as other grasses, sedges and weeds;
[0096] Table 1 Classification system of land use types
[0097]
[0098]
[0099] Secondly, to ensure the accuracy and reliability of the classification results, according to the land use type classification system of the wetland study area, in accordance with the principles of representativeness, uniformity and accuracy of sample point selection, combined with field investigation and literature research, and referring to existing results, a certain number of classified sample point data were directly selected on the remote sensing image through visual interpretation;
[0100] Again, the sample point data set was input into the random forest model for training, and 70% of the data set was randomly used for training; 30% was used for verification, among which the classification features selected the basic bands of the image and the enhanced vegetation index EVI, NDVI and improved MNDWI after band operation; the calculation formula of the enhanced vegetation index EVI is as follows:
[0101]
[0102] Among them, NIR is the near infrared band, RED is the visible red band, BLUE is the visible blue band, C1 is a constant of 6.0, and C2 is a constant of 7.5;
[0103] Then, the classification results are subjected to morphological processing (opening and closing operations) to eliminate the influence of noise and smooth the image edges;
[0104] Finally, check the classification accuracy (ACC (Accuracy) coefficient or Kappa coefficient is above 85%), save and plot;
[0105] Step 5: Optimal threshold optimization feedback algorithm for migratory bird habitat;
[0106] Use the "mgcv" package in RStudio software to build a GAM model:
[0107] g(μ)=α+f(X1+…+X i ),
[0108] Where g() is the link function, α is the intercept, and μ is the mathematical expectation of the response variable, that is, μ=E(Y / X1,...,X i ), f(X i ) is a single explanatory variable X i The single variable function of the response variable is established by the connection function in the formula. The relationship between the mathematical expectation of the response variable and the smooth function of the explanatory variable is established. The response variable is regarded as the sum of smooth functions of some explanatory variables. The model form is more flexible and there is no requirement for the form of the explanatory variables. The only assumption is that each function term is smooth and additive. The GAM model needs to set the basic dimension (parameter) of each smooth term. For the parameter setting, the embodiment of the present application uses R language to create a feedback optimization algorithm, which improves the calculation efficiency and can more accurately and quickly find the optimal parameters of the influencing factors in the model. The detailed technical route is as follows. Figure 2As shown in the figure, remote sensing inversion data is used as the influencing factor, and the number of different wintering migratory bird populations is used as the response variable. First, a GAM model of a single influencing factor and wintering migratory birds (response variable) is established. Then, the model explanation rate and significance of the influencing factor on the number of different wintering migratory bird populations are used to analyze the correlation between wintering migratory birds and the influencing factor. The Pearson correlation analysis formula is as follows:
[0109]
[0110] In the formula, R xy is the correlation coefficient between the number of migratory birds and the influencing factors, x i and i are the number of migratory birds and the impact factor value in the i-th year, and are the average number of migratory birds and the average value of the impact factor over n years, n is the number of samples;
[0111] After comprehensively considering the living habits of wintering migratory birds, the correlation between influencing factors and migratory birds, and the parameters of the single factor model (for example, p value less than 0.1, deviation explanation rate greater than 5%), the key influencing factors were preliminarily screened for different migratory bird populations, and then the interactions between the influencing factors were considered and the repeated effects between the preliminarily selected influencing factors were eliminated. Finally, six main influencing factors were determined for each migratory bird population. The optimal parameters of the model were selected through the feedback optimization algorithm, and the optimal multi-factor mixed GAM model was established. The established model was subjected to an F test. After the test, the model was applied to analyze the changes in the number of wintering migratory bird populations under the joint action of multiple factors, and to explore the optimal threshold values of the above-mentioned influencing factors suitable for the survival of migratory birds.
[0112] As a specific implementation method, the method is applied to the analysis of the optimal threshold of wetland water bodies, vegetation succession and a certain wintering migratory bird habitat in area A in a certain year, wherein the remote sensing inversion process of water body and vegetation related information is as follows: Figure 1 As shown in Figure 2, the analysis process of the optimal threshold of factors affecting the habitat of overwintering migratory birds is as follows: Figure 2 As shown in Figure 2, the remote sensing inversion of water and vegetation related information lays the foundation for the subsequent optimal threshold analysis of migratory bird habitat influencing factors. Considering the characteristics of wintering migratory birds, the dry season is selected for analysis. The results of the extraction of water area in the dry season are shown in Figure 2. Figures 3-1 to 3-4 The results of vegetation coverage in the dry season are shown in Figures 4-1 to 4-4 As shown in Figure 2, according to the established classification system, the results of land use classification in the dry season are as follows: Figures 5-1 to 5-4 As shown;
[0113] The collected hydrological and meteorological data and the remote sensing inversion results of the above water bodies, vegetation and land use types were used as influencing factors, and the population of a certain migratory bird in area A was used as the response variable. After two-step screening (Table 2), in Table 2-3, DW is the deep water area, BAR is the bare land area, MUD is the mudflat area, TAI is the moss community area, LU is the reed community area, QI is the other vegetation area, KW is the dry season water area, FVCM is the average vegetation coverage, and TK is the average temperature in the dry season. A GAM model reflecting the intrinsic mapping relationship between the two was constructed. The model construction results are shown in Tables 3 and Figure 6a-6f As shown;
[0114] Table 2 Model influencing factor screening
[0115]
[0116] As shown in Table 3, the deviation explanation rate of the model is above 90%, which is statistically significant (p<0.001), indicating that the constructed GAM model can be further used to analyze the habitat suitability of migratory birds;
[0117] Table 3 Model information statistics
[0118]
[0119] According to Figure 6, the number of migratory birds is mainly affected by bare land, mudflats, sedge communities, reed clusters, swamps and other vegetation. The bare land area shows a positive linear correlation with the number of migratory birds, while the other influencing factors show a nonlinear relationship. The number of migratory birds increases first, then decreases, and finally increases with the increase of the reed cluster area; increases first, then decreases with the increase of other vegetation areas, and the opposite is true with the increase of mudflat area; decreases with the increase of sedge community area; and decreases gently with the change of swamp area. From the perspective of the number of migratory birds, the larger the bare land area, the more beneficial it is, and the optimal area of reed clusters is 150km. 2 The optimal area of other vegetation is 800km 2 The optimal area of the moss community is about 800km 2 Below, the optimal mudflat area is 50km 2 Below, remote sensing interpretation and mathematical statistics are used to analyze the temporal and spatial distribution characteristics of wetland water bodies and wetland vegetation and the corresponding biological habitat change patterns, and to identify the main driving factors affecting the structure of biological communities and population sizes, which can promote the rational use of wetland land resources and provide a scientific basis for protecting the health of wetland ecosystems.
[0120] The present invention is not limited by the above embodiments. The above embodiments and descriptions are only for explaining the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention may be subject to various changes and improvements, which are within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the appended claims.
Claims
1. A method for inversion and optimization of lake wetland migratory bird habitat based on remote sensing, characterized in that: include: Step 1, obtaining wetland water body and vegetation succession characteristic data and performing data preprocessing; Step 2, extracting water body area using normalized difference water body index method; Step 3, extracting the vegetation coverage of lake wetlands by pixel dichotomy; Step 4, classify the land use types using random forest supervised classification; Step 5: Analyze the changing patterns of population size and community structure of wintering migratory birds in lake wetlands; use hydrometeorological data, water area inverted by remote sensing, vegetation coverage and land use type data as influencing factors, and biological quantity as response variable. The influencing factors are obtained through two screenings, and the model parameters are adjusted through continuous feedback optimization to construct a GAM model to obtain the driving factors of the evolution of wintering migratory birds and the optimal threshold suitable for the survival of wintering migratory birds.
2. The method for inversion and optimization of habitat for migratory birds in lake wetlands based on remote sensing according to claim 1, characterized in that: In step 1, data preprocessing includes atmospheric correction, mosaic cropping, cloud removal, image filling and stripe repair.
3. The method for inversion and optimization of habitat for migratory birds in lake wetlands based on remote sensing according to claim 1, characterized in that: In step 2, extracting the water area includes: Step 21, calculate the MNDWI value for the wetland water area: In formula (1): GREEN is the green light band; SWIR1 is the short-wave infrared band; Step 22, set NDWI threshold Value, extract water area image: In formula (2), MNDWI threshold is the empirical threshold, ranging from 0 to 0.3; Step 23, use image morphological operations to refine and fill the water area data, and then calculate the water area based on the resolution and number of pixels of the image: In formula (3): Water area is the water body area; is the number of pixels with Water=1; aera pixel is the pixel size.
4. The method for inversion and optimization of habitat for migratory birds in lake wetlands based on remote sensing according to claim 1, characterized in that: In step 3, extracting the vegetation coverage of lake wetlands includes: Step 31, calculate the normalized difference vegetation index NDVI: In formula (4), NIR is the near infrared band; RED is the visible red band; Step 32: Use NDVI to estimate vegetation coverage FVC by pixel dichotomy: In formula (5), NDVI min With NDVI max is the minimum and maximum value of NDVI in the area. To eliminate the influence of noise, NDVI min With NDVI max Take the 5% and 95% values of confidence for the cumulative frequency of NDVI values.
5. The method for inversion and optimization of habitat for migratory birds in lake wetlands based on remote sensing according to claim 1, characterized in that: In step 4, land use type classification includes: Step 41, based on the scientific investigation results and the main vegetation and land use types of the wetlands recorded in the literature, and comparing the remote sensing image features, a land use type classification system for the wetland study area is established; Step 42, according to the land use type classification system of the wetland study area, with reference to literature records and visual interpretation, select the classification feature sample point data set; Step 43, randomly divide the sample point data into a training data set and a validation data set, use the basic band of the remote sensing image and the enhanced vegetation index EVI, NDVI and MNDWI after band operation as classification features of different land use types, and finally input the training data set into the random forest model for training, wherein: In formula (6), NIR is the near infrared band; RED is the visible red band; BLUE is the visible blue band; C1 is a constant of 6.0; C2 is a constant of 7.5; Step 44, performing morphological processing on the classification results to eliminate noise influence and smooth the image edge; Step 45, checking the classification accuracy, when the ACC coefficient or Kappa coefficient of the classification result reaches more than 85%, the result is saved and plotted.
6. The method for inversion and optimization of habitat for migratory birds in lake wetlands based on remote sensing according to claim 1, characterized in that: Step 5 includes: Step 51: The collected hydrological, meteorological and remote sensing inversion data are regarded as influencing factors, and the wintering migratory bird data are regarded as response variables, and a single-factor GAM model of a single influencing factor and the number of wintering migratory birds is established, wherein the GAM model formula is: g(μ)=α+f(X1+...+X i )……(7), In formula (7), g() is the link function, α is the intercept, and μ is the mathematical expectation of the response variable, that is, μ = E(Y / X1,...,X i ), f(X i ) is a single explanatory variable X i A single variable function of ; Step 52, considering the living habits of different wintering migratory birds, the correlation between the influencing factors and the migratory birds, and the parameters of the single factor model, the key influencing factors for different wintering migratory birds are preliminarily screened, wherein the absolute value of the Pearson correlation coefficient is greater than 0.2, and the Pearson correlation analysis is as shown in formula (8); In formula (8), R xy is the correlation coefficient between the number of migratory birds and the influencing factors, x i and i are the number of migratory birds and the impact factor value in the i-th year, and are the average number of migratory birds and the average value of the impact factor over n years, n is the number of samples; Step 53: Use the GAM model to explore the interaction between different influencing factors, and select the interaction terms of the influencing factors to be added to the model. Use Pearson correlation analysis to eliminate the repeated effects between the influencing factors, and determine six main influencing factors for each wintering migratory bird. Step 54, selecting the optimal parameters of the model through the feedback optimization algorithm, establishing the optimal multi-factor mixed GAM model, and performing an F test on the established model; Step 55: Use the model to analyze the changes in the number of wintering migratory bird populations under the combined effects of multiple factors, and obtain the optimal threshold values of the influencing factors suitable for migratory bird survival.
7. The method for inversion and optimization of habitat for migratory birds in lake wetlands based on remote sensing according to claim 1, characterized in that: In step 5, the hydrological and meteorological data include average water level, average temperature, accumulated precipitation, and average sunshine time.
Citation Information
Cited By
Urban wetland fine classification method considering combination of different seasonal facies
CN119559513A
Water body extraction method based on integration of multiple artificial intelligence algorithms
CN119741618A
Navigation water level determination method and system based on ship noise interference evaluation
CN121488872A
