Mangrove forest chlorophyll estimation method for improving radiation transfer model

Abstract

The invention relates to a mangrove forest chlorophyll estimation method for improving a radiation transfer model. The mangrove forest chlorophyll estimation method comprises the following steps: collecting actually measured data of tree height, crown breadth and leaf area index of a mangrove forest; constructing a mathematical relationship among the leaf area index, the tree height and the crown breadth; using the improved radiation transmission model to generate simulated spectrum data as a training set; screening sensitive characteristics of mangrove forest functional traits; constructing a random forest inversion model based on physical constraints; and estimating the chlorophyll content of the mangrove forest by adopting the physical constraint-based random forest inversion model. The method has the advantages that the chlorophyll inversion precision can be effectively improved, a scientific basis is provided for fine inversion of mangrove forest chlorophyll, and data support and decision reference are provided for protection, restoration and scientific management of mangrove forests.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of vegetation parameter remote sensing inversion, and particularly relates to a mangrove chlorophyll estimation method for improving a radiation transfer model. BACKGROUND

[0002] Mangrove has vital ecological functions in seawater purification, wave protection, maintenance of ecosystem diversity, and carbon storage, and is one of the most productive ecosystems in tropical and subtropical coastal zones. Chlorophyll is a key component for photosynthesis of plants, directly determines the growth, productivity and ecological function of plants, and is an important biophysical parameter for evaluating the function of an ecological system. Accurate estimation of mangrove chlorophyll content is of great significance for understanding the dynamic changes of mangrove ecosystems and evaluating environmental stress.

[0003] Traditional field sampling methods are destructive and have limited sample size, making it difficult to obtain regional-scale mangrove chlorophyll content. Existing methods can be divided into empirical, physical and hybrid methods. The first method is vegetation index, which is the simplest form of empirical method, only requiring a few different bands of calculation methods. Although this method is fast and convenient, it has significant limitations. First, the vegetation index is based on a specific band calculation formula, which is difficult to adapt to the diversity of all plant species, especially in different growth stages or environmental conditions. Taking the special ecological system of mangrove as an example, due to its complex canopy structure, the vegetation index is difficult to fully consider, resulting in low inversion accuracy and large result error. The second method is to apply a physical model to consider the influence of canopy structure on chlorophyll estimation. PROSAIL model is a common radiation transfer model that can simulate the spectral response of different vegetation types. However, the PROSAIL model does not directly include actual growth parameters such as tree height and crown width, so how to combine these parameters with the model is the key to further improve the performance and prediction accuracy of the model. The third method is a hybrid method that combines the advantages of empirical and physical models to more accurately estimate chlorophyll. This method generally uses multivariate statistical analysis or machine learning techniques (such as partial least squares regression, support vector machines, etc.) to process more rich spectral information. However, the hybrid method also has certain challenges, especially in the selection of vegetation structure parameters. For example, although this method can reduce the influence of canopy structure on chlorophyll inversion accuracy to some extent, many models only use leaf area index as a single canopy structure parameter. Given the difficulty in obtaining leaf area index in practice, this may adversely affect model accuracy in some cases. SUMMARY

[0004] The purpose of the present application is to overcome the deficiencies in the prior art and provide a mangrove chlorophyll estimation method for improving a radiation transfer model.

[0005] In a first aspect, a mangrove chlorophyll estimation method for improving a radiation transfer model is provided, comprising:

[0006] Step 1, collecting measured data of tree height, crown width and leaf area index of the mangrove, and removing abnormal data;

[0007] Step 2, constructing a mathematical relationship between the leaf area index and the tree height and the crown width, and substituting the relationship into the radiation transfer model to improve the model;

[0008] Step 3, generating simulated spectral data as training samples using the improved radiation transfer model;

[0009] Step 4, screening sensitive features of mangrove functional traits to determine input features;

[0010] Step 5, constructing a physically constrained random forest inversion model (PIRF) based on the training set;

[0011] Step 6, estimating the mangrove chlorophyll content using the physically constrained random forest inversion model.

[0012] Preferably, in step 2, the mathematical relationship between the leaf area index and the tree height and the crown width is represented as:

[0013]

[0014] where Y is the leaf area index, X1 is the tree height, X2 is the crown width, and a, m, n are undetermined coefficients.

[0015] Preferably, in step 2, the radiation transfer model is the PROSAIL model.

[0016] Preferably, in step 4, further comprising: using filtering technology to denoise the simulated spectral data.

[0017] Preferably, further comprising: step 7, using R 2 and RMSE to evaluate the accuracy of the physically constrained random forest inversion model.

[0018] Preferably, step 8, selecting at least one vegetation index to construct a vegetation index inversion model; comparing the accuracy of the physically constrained random forest inversion model and the vegetation index inversion model.

[0019] In a second aspect, a mangrove chlorophyll estimation device for improving a radiation transfer model is provided, which is used to execute the method of any one of the first aspect, comprising:

[0020] A collection module for collecting measured data of tree height, crown width and leaf area index of the mangrove, and removing abnormal data;

[0021] a constructing module configured to construct a mathematical relationship between leaf area index and tree height and crown width, and to substitute the relationship into a radiation transfer model to improve the model;

[0022] a generating module configured to generate simulated spectral data as a training set by using the improved radiation transfer model;

[0023] a screening module configured to screen sensitive features of mangrove functional traits and determine input features;

[0024] a constructing module configured to construct a physically-constrained random forest inversion model based on the training set;

[0025] an estimating module configured to estimate mangrove chlorophyll content by using the physically-constrained random forest inversion model.

[0026] In a third aspect, a computer storage medium is provided, and the computer storage medium stores a computer program; when the computer program runs on a computer, the computer program causes the computer to execute the method according to any one of the first aspect.

[0027] In a fourth aspect, an electronic device is provided, and the electronic device comprises:

[0028] a memory configured to save a computer program;

[0029] a processor configured to execute the computer program to implement the method according to any one of the first aspect.

[0030] The present application has the following beneficial effects: the present application firstly establishes the relationship between mangrove leaf area index (LAI) and tree height and crown width based on measured data, and introduces the relationship into the PROSAIL model, so that the mangrove leaf area index can be estimated simply and efficiently. Then, based on the improved PROSAIL model, a mangrove chlorophyll estimation method is proposed, which is verified by combining UAV data and measured data. The method can effectively improve the accuracy of chlorophyll inversion, and provide scientific basis for fine inversion of mangrove chlorophyll, and provide data support and decision reference for protection and scientific management of mangrove. Therefore, the method proposed in the present application has important practical application significance. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1 is a flowchart of a mangrove chlorophyll estimation method for improving a radiation transfer model provided by the present application;

[0032] Figure 2 is a schematic diagram of global sensitivity analysis of PROSAIL model parameters provided by the present application;

[0033] Figure 3It is a structural schematic diagram of a mangrove chlorophyll estimation device for improving a radiation transmission model provided by the application. DETAILED DESCRIPTION

[0034] The application will be further described below in conjunction with examples. The following examples are only used to help understand the application. It should be pointed out that for ordinary people in the technical field, some modifications can be made without departing from the principles of the application, and these improvements and modifications also fall within the protection scope of the claims of the application.

[0035] Example 1

[0036] Leaf area index, crown width and tree height are important parameters for describing the structure of mangrove canopy. Leaf area index (LAI) is an important parameter for measuring the number and distribution of plant leaves, and is widely used in vegetation monitoring, crop growth research and ecological environment assessment. However, direct measurement of leaf area index is usually complex, costly and affected by factors such as light conditions, leaf shape and distribution, resulting in inaccurate chlorophyll estimation. Therefore, how to accurately and efficiently estimate the leaf area index has become a key problem. Tree height and crown width are basic indicators of canopy structure and are easy to obtain through remote sensing images or ground measurements. In practical applications, tree height and crown width are usually related to leaf area index. Using these easily accessible indicators to estimate leaf area index can not only reduce the measurement cost, but also improve the timeliness and accuracy of the data. In the existing leaf area index estimation methods, most of them ignore the influence of tree height and crown width, and do not establish the relationship between leaf area index and these parameters, so as to improve the accuracy of chlorophyll estimation.

[0037] Therefore, in order to solve the problems of difficult acquisition of existing mangrove leaf area index and low precision of physical model, the application embodiment 1 provides a mangrove chlorophyll estimation method for improving a radiation transmission model. First, the relationship between leaf area index (LAI) and tree height and crown width is established based on measured data, and the relationship is introduced into the PROSAIL model. Then, a random forest inversion method based on physical constraints is constructed, and compared with vegetation index. Finally, the method is verified by combining unmanned aerial vehicle data and measured data.

[0038] Specifically, the application provides a mangrove chlorophyll estimation method for improving a radiation transmission model. Taking the data of mangroves in the Aojiang Estuary as an example, the method is realized based on the PROSAIL model, as shown in Figure 1 , comprising:

[0039] Step 1: Collect the measured data of tree height, crown width and leaf area index of mangroves, and eliminate abnormal data.

[0040] Specifically, the measured data of mangrove parameters, including tree height, east-west crown width, north-south crown width, leaf area index, etc., are sorted out, the mean values of the east-west crown width and the north-south crown width are calculated as the crown width, and the abnormal parameter groups are removed by using the three-sigma principle.

[0041] Step 2, a mathematical relationship between the leaf area index and the tree height and the crown width is constructed, and the relationship is substituted into the radiation transfer model to improve the model.

[0042] In step 2, the radiation transfer model is a PROSAIL model.

[0043] In step 2, the mathematical relationship between the leaf area index and the tree height and the crown width is represented as:

[0044]

[0045] wherein Y is the leaf area index, X1 is the tree height, X2 is the crown width, and a, m, and n are undetermined coefficients.

[0046] Step 3, the improved radiation transfer model is used to generate simulated spectral data as training samples.

[0047] For example, the SIMLAB software and the improved PROSAIL model are used to obtain the simulated spectral data as training samples.

[0048] In step 3, it also includes reducing random noise in the spectral data by using filtering technology.

[0049] For example, the obtained simulated spectral data is shown in Table 1.

[0050] Table 1

[0051]

[0052] Step 4, sensitive feature screening of mangrove functional traits is performed to determine the input features.

[0053] For example, as shown in Table 2, the SIMLAB software is used to perform sensitive feature screening of mangrove functional traits, which is applied to the training set to determine the input features of the next step model. Figure 2

[0054] Step 5, based on the training set, a random forest inversion model based on physical constraints (PIRF) is constructed.

[0055] Specifically, on the basis of the traditional random forest (RF), a series of strategies and steps of systematically integrating physical constraints into the construction, training and prediction process of the random forest inversion model are designed, the physical mechanism of the PROSAIL model is deeply integrated, an architecture with physical interpretability is designed, and the simple application of the existing RF / SVM / PLS is replaced. ​

[0056] Step 6, estimating the mangrove chlorophyll content by using the random forest inversion model based on physical constraints.

[0057] Embodiment 2

[0058] Based on embodiment 1, the application embodiment 2 provides a more specific method for improving the mangrove chlorophyll estimation model, including:

[0059] Step 1, collecting the measured data of tree height, crown width and leaf area index of mangrove, and eliminating abnormal data.

[0060] Step 2, constructing the mathematical relationship between leaf area index and tree height, crown width, and substituting the relationship into the radiation transfer model to improve the model.

[0061] Step 3, generating simulated spectral data as a training set by using the improved radiation transfer model.

[0062] Step 4, screening the sensitive features of mangrove functional traits to determine the input features.

[0063] Step 5, constructing a random forest inversion model based on physical constraints based on the training set.

[0064] Step 6, estimating the mangrove chlorophyll content by using the random forest inversion model based on physical constraints.

[0065] Step 7, using R 2 and RMSE to evaluate the accuracy of the random forest inversion model based on physical constraints.

[0066] For example, as shown in Table 2, the model accuracy is evaluated by using R 2 and RMSE.

[0067] Table 2

[0068] Conventional RF Random forest based on physical constraints CIgreen VOG2 [R 2 ]] 0.71 0.82 0.75 0.91 RMSE (pg / cm 2 )]]> 15.2 10.2 28.42 30.17

[0069] Step 8, selecting at least one vegetation index to construct a vegetation index inversion model; comparing the accuracy of the random forest inversion model based on physical constraints and the vegetation index inversion model.

[0070] For example, using vegetation indexes CIgreen and VOG2, as shown in Table 3.

[0071] Table 3

[0072] Vegetation index Calculation formula Inversion model CIgreen [R 780 / R 550 –1]]> y = 26.705x + 26.421 VOG2 (R 734 –R 747 ) / (R 715 +R 726 )]]> y = -512.21x + 35.888

[0073] Wherein, R n represents the reflectivity of the wavelength n band.

[0074] It should be noted that the same or similar parts in this embodiment as in embodiment 1 can be mutually referred to and will not be repeated in this application.

[0075] Embodiment 3:

[0076] On the basis of embodiment 1, the application embodiment 3 provides a mangrove chlorophyll estimation device for improving a radiation transfer model, comprising:

[0077] The collection module is configured to collect measured data of tree height, crown width and leaf area index of the mangrove, and to remove abnormal data.

[0078] The construction module is configured to construct a mathematical relationship between the leaf area index and the tree height and the crown width, and to substitute the relationship into the radiation transfer model to improve the model.

[0079] The generation module is configured to generate simulated spectral data as a training set by using the improved radiation transfer model.

[0080] The screening module is configured to screen sensitive features of functional traits of the mangrove, and to determine input features.

[0081] The construction module is configured to construct a random forest inversion model based on physical constraints based on the training sample.

[0082] The estimation module is configured to estimate the mangrove chlorophyll content by using the random forest inversion model based on physical constraints.

[0083] It should be noted that the system provided in this embodiment is a device corresponding to the method provided in embodiment 1, and therefore, in this embodiment, the same or similar parts as in embodiment 1 can be mutually referred to and will not be repeated in this application.

[0084] In summary, the application provides a mangrove chlorophyll estimation method for improving a radiation transfer model, which combines a radiation transfer model and machine learning. The method can effectively improve the accuracy of chlorophyll inversion and provide a scientific basis for fine chlorophyll inversion. It provides data support and decision-making reference for the protection and scientific management of mangroves. Therefore, the method provided by the application has important practical application significance.

Claims

1. A method for estimating chlorophyll in mangroves based on an improved radiation transfer model, characterized in that: include: Step 1: Collect measured data on tree height, crown width, and leaf area index of mangroves and remove abnormal data; Step 2: Construct a mathematical relationship between leaf area index and tree height and crown width, and substitute this relationship into the radiation transfer model to improve the model; Step 3: Generate simulated spectral data as a training set using the improved radiation transfer model; Step 4: Screen the sensitive features of mangrove functional traits and determine the input features; Step 5: Based on the training set, construct a physical constraint-based random forest inversion model; Step 6: Use the physical constraint-based random forest inversion model to estimate the chlorophyll content of mangroves.

2. The method for estimating chlorophyll in mangroves based on the improved radiation transfer model according to claim 1, characterized in that: In step 2, the mathematical relationship between the leaf area index, tree height, and crown width is expressed as: Among them, Y is the leaf area index, X1 is the tree height, X2 is the crown width, and a, m, and n are unknown coefficients.

3. The method for estimating chlorophyll in mangroves based on the improved radiation transfer model according to claim 2, characterized in that: In step 2, the radiation transfer model is the PROSAIL model.

4. The method for estimating chlorophyll in mangroves based on the improved radiation transfer model according to claim 3, characterized in that: Step 3 also includes: using filtering technology to perform denoising on the simulated spectral data.

5. The method for estimating chlorophyll in mangroves based on the improved radiation transfer model according to claim 4, characterized in that: Also included: Step 7, using R 2 The accuracy of the physical constraint-based random forest inversion model is evaluated by using the RMSE.

6. The method for estimating chlorophyll in mangroves based on the improved radiation transfer model according to claim 5, characterized in that: The method further includes: step 8, selecting at least one vegetation index to construct a vegetation index inversion model; and comparing the accuracy of the physical constraint-based random forest inversion model with the vegetation index inversion model.

7. A mangrove chlorophyll estimation device based on an improved radiation transfer model, characterized in that: Used to perform the method according to any one of claims 1 to 6, comprising: The collection module is used to collect the measured data of mangrove tree height, crown width and leaf area index, and eliminate abnormal data; A construction module is used to construct a mathematical relationship between leaf area index and tree height and crown width, and substitute this relationship into the radiation transfer model to improve the model; A generation module, used to generate simulated spectral data as a training set using the improved radiation transfer model; Screening module, used to screen sensitive features of mangrove functional traits and determine input features; A construction module, configured to construct a physical constraint-based random forest inversion model based on the training set; An estimation module is used to estimate the chlorophyll content of mangroves using the physical constraint-based random forest inversion model.

8. A computer storage medium, characterized in that The computer storage medium stores a computer program; when the computer program is run on a computer, the computer executes the method according to any one of claims 1 to 6.

9. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor, configured to execute the computer program to implement the method according to any one of claims 1 to 6.

Images