Method and device for evaluating the adaptability of a gulf salinity to a mangrove forest

Abstract

The application discloses a kind of gulf salinity and mangrove adaptability evaluation method and device, it is related to remote sensing monitoring technology and mangrove ecological protection technical field.The method comprises: obtaining Landsat8 OLI satellite image and mangrove visual image;Extract band data, construct vegetation index, water index, double-band feature vector and three-band feature vector;The salinity prediction ability of feature vector is evaluated using random forest regressor, and after screening, redundant features are removed by Pearson correlation coefficient to obtain target feature vector;Based on target feature vector, SHAP interpretability analysis is carried out on XGBoost model, and the optimal feature vector is determined;Salinity inversion is carried out using XGBoost model, salinity level area is divided, and adaptability is evaluated in combination with mangrove visual image.The application can accurately evaluate the adaptability of gulf salinity and mangrove, improve evaluation efficiency and accuracy.

Description

Technical Field

[0001] This application relates to the fields of remote sensing monitoring technology and mangrove ecological protection technology, and in particular to a method for assessing the compatibility between bay salinity and mangroves. Background Technology

[0002] Sea surface salinity (SSS) is a key environmental factor regulating the distribution, growth, and resistance to disturbance of mangrove species. Especially in semi-enclosed bays and other areas, fluctuations in freshwater input can easily lead to changes in salinity, which in turn can cause ecological problems such as mangrove mortality and vine invasion. Accurately obtaining the spatiotemporal distribution of salinity and establishing its correlation with mangrove growth is crucial for coastal ecological protection.

[0003] In existing technologies, the accuracy of salinity retrieval is effectively improved by introducing machine learning models to retrieve marine salinity.

[0004] However, although machine learning models have improved the inversion results, they have not made full use of model-driven predictive ability assessment to initially screen high-response features. Furthermore, after screening, they have not effectively addressed the multicollinearity problem among features. Highly correlated features are redundantly input into the model, which can easily lead to model overfitting, decreased generalization ability, and confusion in the explanation of physical mechanisms. Summary of the Invention

[0005] Therefore, it is necessary to provide a method for assessing the compatibility between bay salinity and mangroves to address the aforementioned technical issues.

[0006] The present invention adopts the following technical solution: This invention provides a method for assessing the compatibility between bay salinity and mangrove forests, comprising: Acquire Landsat-8 OLI satellite imagery and mangrove visual imagery of the Gulf region; and extract the first few band data from the Landsat-8 OLI imagery. Based on the band data, several vegetation indices, several water body indices, and several band combination feature vectors are determined as the total feature vectors of the bay area. The salinity prediction capability of all feature vectors in the bay area is evaluated by using a random forest regressor. Several feature vectors with the highest salinity prediction capability are selected, and the Pearson correlation coefficient between each pair of feature vectors is determined. Feature vectors with a Pearson correlation coefficient greater than a preset threshold are removed to obtain the target feature vector of the bay area. Based on the target feature vector of the bay area, SHAP interpretability analysis is performed on the trained XGBoost model, and the target feature vector most related to salinity is determined as the optimal feature vector of the bay area. Based on the optimal feature vector, the trained XGBoost model is used to perform salinity inversion in the bay area, dividing the bay area into regions with different salinity levels; based on the regions with different salinity levels and visual images of mangroves, the compatibility between the salinity of the bay area and the mangroves is evaluated.

[0007] Preferably, acquiring Landsat-8 OLI imagery of the Gulf region specifically includes: Acquire raw Landsat-8 OLI satellite imagery of the target bay area with cloud cover below 10%; Radiometric calibration and atmospheric correction were performed sequentially on the original Landsat-8 OLI satellite imagery to obtain Landsat-8 OLI imagery of the Gulf region.

[0008] Preferably, the acquisition of visual images of mangroves specifically includes: Aerial photography of the mangroves was obtained by using multi-rotor drones to take aerial photos of the estuaries of rivers flowing into the sea in the bay area and several synchronous stations set up in the core distribution area of ​​mangroves. Visual interpretation of the aerial images of the mangrove forest yields visual images of the mangrove forest.

[0009] Preferably, the vegetation indices include: Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), Soil-Adjusted Vegetation Index (SAVI), Transformed Vegetation Index (TVI), and Simple Ratio Vegetation Index (SR); the water indices include: Normalized Difference Water Index (NDWI), Modified Normalized Difference Water Index (MNDWI), Multiband Difference Water Index (MBWI), Enhanced Normalized Difference Water Index (ANDWI), and Difference Water Index 2015WI2015.

[0010] Preferably, the multiple numerical calculation methods for two non-overlapping band data include: subtraction, multiplication, division, and the numerical calculation method of dividing the difference of the two bands by the sum of the two bands; the numerical calculation method for the three non-overlapping band data is the numerical calculation method of subtracting the reciprocal of the second band from the reciprocal of the first band and then multiplying by the numerical calculation method of the third band.

[0011] Preferably, the salinity prediction capability of the random forest regressor is evaluated by using all feature vectors of the bay area, specifically including: All feature vectors are input into a random forest regressor, and salinity prediction is performed for each feature vector by constructing multiple decision trees; The ability of a feature to predict salinity is evaluated by averaging the salinity predictions of all decision trees for each feature vector. The higher the average value of the salinity prediction results, the stronger the predictive ability of the feature vector; conversely, the lower the average value, the weaker the predictive ability of the feature vector.

[0012] Preferably, when performing SHAP interpretability analysis on the trained XGBoost model, the average absolute value of SHAP is used as the feature importance metric for the target feature vector of the bay area, so as to determine the optimal water feature vector with the highest response to salinity prediction.

[0013] Preferably, a trained XGBoost model is used to perform salinity inversion in the bay area, dividing the bay area into regions with different salinity levels; based on the regions with different salinity levels and visual images of mangroves, a compatibility assessment of the salinity of the bay area and the mangroves is performed, specifically including: The spatial distribution of salinity across the entire target bay was inverted using the optimal XGBoost model, and a salinity spatial distribution map was drawn, dividing the bay area into high-salinity and low-salinity zones according to salinity levels. Based on the regional distribution of high-salinity and low-salinity areas, and combined with visual images of mangroves, an assessment of mangrove salinity adaptability was conducted, specifically including: Based on the distribution areas of mangroves and the regional distribution of high-salinity and low-salinity areas in visual images of mangroves, the critical salinity threshold for mangrove survival was determined. Based on the degradation of mangroves in visual images and the regional distribution of high-salinity and low-salinity areas, the optimal salinity range for the growth of *Derris tectorum* was determined. Based on the critical salinity threshold for mangrove survival and the optimal salinity range for drupe vine growth, a correlation model between salinity gradient and mangrove distribution and growth status is established for the protection of mangroves in the target area and the special management of drupe vine.

[0014] This invention provides a device for assessing the compatibility between bay salinity and mangrove forests, comprising: The data acquisition module is used to acquire Landsat-8 OLI satellite images and mangrove visual images of the Gulf region; and to extract the first few band data from the Landsat-8 OLI images. The feature construction module is used to determine several vegetation indices, several water body indices, and several band combination feature vectors based on the band data, as the total feature vectors of the bay area. The feature selection module is used to evaluate the salinity prediction capability of all feature vectors in the bay area using a random forest regressor, select a number of feature vectors with the highest salinity prediction capability, determine the Pearson correlation coefficient between every two feature vectors, and remove feature vectors with a Pearson correlation coefficient greater than a preset threshold to obtain the target feature vector of the bay area; it is also used to perform SHAP interpretability analysis on the trained XGBoost model based on the target feature vector of the bay area, and determine the target feature vector most correlated with salinity as the optimal feature vector of the bay area; The adaptation evaluation module is used to perform salinity inversion on the bay area using the trained XGBoost model based on the optimal feature vector, and divide the bay area into regions with different salinity levels; based on the regions with different salinity levels and visual images of mangroves, the module evaluates the compatibility between the salinity of the bay area and the mangroves.

[0015] The present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned method for assessing the compatibility between bay salinity and mangroves.

[0016] The present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-mentioned method for assessing the compatibility between bay salinity and mangroves.

[0017] The above-mentioned at least one technical solution adopted in this invention can achieve the following beneficial effects: In the adaptability assessment method for bay salinity and mangroves provided by this invention, a random forest regressor is used to assess salinity prediction capability, and Pearson correlation coefficient is combined to remove redundant features. This retains key spectral information and eliminates the interference of multicollinearity on model stability, ensuring the efficiency and reliability of input features. Furthermore, based on the selected target feature vector with high salinity responsiveness, the physical logic of the salinity inversion results is traceable and the credibility is greatly improved. The high-precision salinity inversion results are used to divide regions with different salinity levels, and a precise correlation between salinity gradient and mangrove growth and vine invasion is established by combining mangrove visual images. This breaks through the limitation of the disconnect between salinity inversion and ecological assessment in traditional monitoring, and provides technical support with high precision, interpretability and application value for the ecological protection and precise management of coastal mangroves. Attached Figure Description

[0018] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0019] Figure 1 A flowchart illustrating the compatibility assessment method between bay salinity and mangroves provided by this invention; Figure 2 A schematic diagram of a device for assessing the compatibility of bay salinity with mangroves provided by the present invention; Figure 3 This is a schematic diagram of an apparatus for implementing a method for assessing the compatibility between bay salinity and mangroves, as provided by the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments in the specification without creative effort are within the scope of protection of this application.

[0021] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0022] Figure 1 This is a schematic diagram of a method for assessing the compatibility between bay salinity and mangroves in this invention, specifically including the following steps: S101: Acquire Landsat-8 OLI satellite imagery and mangrove visual imagery of the bay area; and extract the first few band data from the Landsat-8 OLI imagery.

[0023] Optionally, acquiring Landsat-8 OLI imagery of the Gulf region specifically includes: acquiring raw Landsat-8 OLI satellite imagery of the target Gulf region with cloud cover below 10%; performing radiometric calibration and atmospheric correction on the raw Landsat-8 OLI satellite imagery sequentially to obtain Landsat-8 OLI imagery of the Gulf region.

[0024] Optionally, the acquisition of visual images of mangroves specifically includes: using a multi-rotor UAV to take aerial photos of the estuaries of rivers flowing into the bay area and several synchronous stations deployed in the core distribution area of ​​mangroves to obtain aerial images of mangroves; and performing visual interpretation on the aerial images of mangroves to obtain visual images of mangroves.

[0025] Specifically, the core of this step is to acquire high-quality, spatiotemporally matched measured salinity data, remote sensing image data, and UAV ecological data to lay the foundation for subsequent inversion and evaluation.

[0026] Water Sampling and Salinity Measurement: Sampling was conducted according to monitoring requirements. Several sampling points were set up in the target bay and surrounding river estuaries, covering different hydrological areas such as estuaries, inner bays, and bay necks, to ensure a full characterization of the 12‰-30‰ salinity range. GPS was used for precise positioning, and water samples were collected at a depth of 0.5m below the water surface. Salinity was measured on-site using a YSI Professional Plus multi-functional water quality analyzer. During the sampling period and satellite transit window, there was no precipitation and the water surface was stable to eliminate runoff interference and meteorological noise, ensuring high-fidelity matching between measured data and remote sensing imagery.

[0027] Landsat-8 OLI image preprocessing: Landsat-8 OLI satellite images with cloud cover below 10% and matching the sampling period within ±14 days were selected and acquired through the Google Earth Engine platform. Image preprocessing included: first, radiometric calibration to convert the raw DN values ​​into top-atmosphere radiance; then, atmospheric correction using the FLAASH module to eliminate atmospheric scattering and absorption effects and obtain the true surface reflectance; finally, data from the first seven bands (B1-B7) were extracted, all with a spatial resolution of 30m, meeting the requirements for refined inversion of various small-scale bay areas. The parameters for the seven bands are: B1 (coastal band, 0.435-0.451 μm), B2 (blue band, 0.452-0.512 μm), B3 (green band, 0.533-0.590 μm), B4 (red band, 0.636-0.673 μm), B5 (near-infrared band, 0.851-0.879 μm), B6 ​​(shortwave infrared 1, 1.566-1.651 μm), and B7 (shortwave infrared 2, 2.107-2.294 μm).

[0028] Unmanned Aerial Vehicle (UAV) Image Acquisition and Interpretation: Using multi-rotor UAVs, aerial photography was conducted at several synchronous stations deployed at the estuaries of various rivers flowing into the sea and in the core distribution area of ​​mangroves within three days after sampling. Clear, windless weather was selected to ensure image clarity. Visual interpretation of the aerial images was performed to identify the distribution areas of typical mangrove species such as *Avicennia marina*, *Kandelia candel*, and *Avicennia marina*, as well as *Derris tectorum*. Dead and degraded mangrove areas were marked, and a one-to-one correspondence was established between the image interpretation results and the salinity data of the sampling points.

[0029] S102: Based on the band data, determine several vegetation indices, several water body indices, and several band combination feature vectors as the total feature vectors of the bay area.

[0030] Optionally, vegetation indices include: Normalized Difference Vegetation Index (NDVI), Enhanced Difference Vegetation Index (EVI), Soil-Adjusted Difference Vegetation Index (SAVI), Transformed Difference Vegetation Index (TVI), and Simple Ratio Difference Vegetation Index (SR); water indices include: Normalized Difference Water Index (NDWI), Modified Normalized Difference Water Index (MNDWI), Multiband Difference Water Index (MBWI), Enhanced Normalized Difference Water Index (ANDWI), and Difference Water Index 2015WI2015.

[0031] Specifically, based on the preprocessed Landsat-8 OLI imagery, a total of 290 feature variables were constructed, covering four categories: ① 7 original bands (B1-B7); ② 168 band combination variables, with two bands using multiple calculation forms such as Ri-Rj, Ri*Rj, Ri / Rj, and (Ri-Rj) / (Ri+Rj), and three bands using the calculation form (1 / Ri-1 / Rj)*Rk; ③ 5 vegetation indices (NDVI, EVI, SAVI, TVI, SR); ④ 5 water indices (NDWI, MNDWI, MBWI, ANDWI, WI2015).

[0032] The formulas for calculating each index are as follows: EVI=2.5×(NIR-Red) / (NIR+6×Red-7.5×Blue+1); SAVI=(NIR-Red) / (NIR+Red+L)×(1+L)(L=0.5); TVI=0.5×[120×(NIR-Green)-200×(Red-Green)]; SR=NIR / Red, NDWI=(Green-NIR) / (Green+NIR); MNDWI=(Green-SWIR1) / (Green+SWIR1); MBWI=2×Green−Red−NIR−SWIR1−SWIR2; ANDWI=(Blue×Green+Red-NIR-SWIR1-SWIR2) / (Blue+Green+Red+NIR+SWIR1+SWIR2); WI2015=0.1446×Blue+0.1761×Green+0.3322×Red+0.3396×NIR+0.6210×SWIR1-0.4186×SWIR2-3.3828; Optionally, the various numerical operation methods for two non-overlapping band data include: subtraction, multiplication, division, and the numerical operation method of dividing the difference of the two bands by the sum of the two bands; the numerical operation method for the three non-overlapping band data is the numerical operation method of subtracting the reciprocal of the second band from the reciprocal of the first band and then multiplying by the numerical operation method of the third band.

[0033] Optionally, the numerical calculation methods for two non-overlapping band data include: subtraction, multiplication, division, and the numerical calculation method of dividing the difference between the two bands by the sum of the two bands; the numerical calculation method for three non-overlapping band data is the numerical calculation method of subtracting the reciprocal of the second band from the reciprocal of the first band and then multiplying by the numerical calculation method of the third band.

[0034] S103: Use a random forest regressor to evaluate the salinity prediction capability of all feature vectors in the bay area, select a number of feature vectors with the highest salinity prediction capability, determine the Pearson correlation coefficient between each pair of feature vectors, and remove feature vectors with a Pearson correlation coefficient greater than a preset threshold to obtain the target feature vector of the bay area.

[0035] Optionally, a random forest regressor is used to evaluate the salinity prediction capability of all feature vectors in the bay area. Specifically, this includes: inputting all feature vectors into the random forest regressor, constructing multiple decision trees to predict the salinity of each feature vector; and evaluating the feature's salinity prediction capability based on the average of the salinity prediction results of all decision trees for each feature vector. The larger the average salinity prediction result, the stronger the prediction capability of the feature vector; conversely, the weaker the prediction capability of the feature vector.

[0036] Specifically, 290 feature variables were input into a random forest regressor. The predictive ability of the features on salinity was evaluated by constructing multiple decision trees and averaging the results, and the top 30 features by importance were selected. To eliminate the effects of multicollinearity, the Pearson correlation coefficients among the selected features were calculated. Redundant features with an absolute correlation coefficient exceeding 0.95 were removed, and unique and representative core features were retained, ultimately obtaining a set of key features that can accurately characterize salinity changes.

[0037] S104: Based on the target feature vector of the bay area, perform SHAP interpretability analysis on the trained XGBoost model, and determine the target feature vector most related to salinity as the optimal feature vector of the bay area.

[0038] Optionally, when performing SHAP interpretability analysis on the trained XGBoost model, the average absolute value of SHAP is used as a measure of the feature importance of the target feature vector in the bay area to determine the optimal water feature vector with the highest response to salinity prediction.

[0039] Specifically, based on the AutoML-FLAML framework, six mainstream machine learning algorithms—XGBoost, Catboost, Extra_tree, LGBM, XGB_Limitdepth, and RF—were selected to construct and optimize salinity inversion models. With a time budget of 120 seconds, the coefficient of determination (R²) was used as the core optimization metric. The Cost-Frugal Optimization (CFO) efficient search algorithm adaptively searched the hyperparameter space, automatically completing the entire process of data preprocessing, model selection, and hyperparameter optimization without manual intervention. The model performance was comprehensively evaluated using three metrics: R², RMSE (root mean square error), and MAE (mean absolute error). XGBoost was determined to be the optimal salinity inversion model, with test set accuracy metrics of R²=0.9062‰, RMSE=0.9062‰, and MAE=1.1097‰, demonstrating significantly better adaptability than other algorithms.

[0040] Specifically, the SHAP (Shapley Additive exPlanations) method was used to interpret the optimal XGBoost model. Based on the Shapley value theory in cooperative game theory, each feature was assigned a value reflecting its marginal impact on the prediction result, and the average absolute value of SHAP was used to measure the importance of the feature. The analysis results show that the B3 (green band)-B4 (red band) combination is the core sensitive feature, accounting for 53.3% of the importance, and has a high responsiveness to both high and low salinity values. Multiple combination features constructed with the blue band (B2) also showed significant contributions, consistent with the strong absorption characteristics of CDOM in the blue light band, verifying the reliability of indirect salinity inversion. The SHAP analysis clarified the response relationship between each feature and salinity, completely solving the "black box" problem of traditional machine learning models and improving the credibility of the inversion results.

[0041] S105: Based on the optimal feature vector, the trained XGBoost model is used to perform salinity inversion on the bay area, dividing the bay area into regions with different salinity levels; based on the regions with different salinity levels and visual images of mangroves, the compatibility between the salinity of the bay area and the mangroves is evaluated.

[0042] Optionally, a trained XGBoost model is used to retrieve salinity in the bay area, dividing it into regions with different salinity levels. Based on these regions and visual images of mangroves, a compatibility assessment of the bay's salinity and mangrove ecosystem is conducted. This includes: retrieving the spatial distribution of salinity across the entire target bay using the optimal XGBoost model, creating a salinity spatial distribution map, and dividing the bay area into high-salinity and low-salinity zones according to salinity levels; and, based on the regional distribution of high-salinity and low-salinity zones and combined with visual mangrove images, conducting [further analysis / assessment]. The assessment of mangrove salinity suitability specifically includes: determining the critical salinity threshold for mangrove survival based on the distribution area of ​​mangroves and the regional distribution of high-salinity and low-salinity areas in visual images of mangroves; determining the optimal salinity range for the growth of *Derris tectorum* based on the degradation status of mangroves and the regional distribution of high-salinity and low-salinity areas in visual images of mangroves; and establishing a correlation model between salinity gradient and mangrove distribution and growth status based on the critical salinity threshold for mangrove survival and the optimal salinity range for the growth of *Derris tectorum*, for the protection of mangroves in the target area and the special management of *Derris tectorum*.

[0043] Specifically, the optimal XGBoost model was used to invert the spatial distribution of salinity across the entire target bay, and a salinity spatial distribution map was drawn to clarify the general distribution pattern of salinity in the order of "bay neck > inner bay > estuary". Low salinity areas (12‰-20‰) are concentrated in the estuary and coastal areas, while high salinity areas (25‰-30‰) are distributed in the bay neck and some river mouth areas. Based on the visual interpretation results of UAV imagery, a salinity suitability assessment of mangroves was conducted: ① A region-specific critical salinity threshold was defined, and 25‰ was determined to be the critical salinity for the growth of most common mangrove species (such as *Symplocos edulis*, *Kandelia candel*, and *Avicennia marina*). When the salinity is >25‰, the mangrove mortality phenomenon is most significant in some river estuaries. ② A suitable habitat for *Derris tectorum* was delineated, and the optimal salinity range for *Derris tectorum* was identified as 10‰-20‰. Within this salinity range, the degree of mangrove degradation due to *Derris tectorum* invasion is the most severe. ③ A correlation model between salinity gradient and mangrove distribution and growth status was established to provide precise basis for mangrove protection and *Derris tectorum*-specific management in the target area.

[0044] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this invention.

[0045] The above describes a method for assessing the compatibility between bay salinity and mangrove forests, provided by one or more embodiments of the present invention. Based on the same idea, the present invention also provides a corresponding device for assessing the compatibility between bay salinity and mangrove forests, such as... Figure 2 As shown.

[0046] Figure 2A schematic diagram of a compatibility assessment device for bay salinity and mangroves provided by the present invention, the device comprising: Data acquisition module 201 is used to acquire Landsat-8 OLI satellite imagery and mangrove visual imagery of the Gulf region; and extract several band data in the order of the first band from the Landsat-8 OLI imagery; The feature construction module 202 is used to determine several vegetation indices, several water body indices and several band combination feature vectors based on the band data, as the total feature vectors of the bay area. The feature selection module 203 is used to evaluate the salinity prediction capability of all feature vectors in the bay area using a random forest regressor, select a number of feature vectors with the highest salinity prediction capability, determine the Pearson correlation coefficient between every two feature vectors, and remove feature vectors with a Pearson correlation coefficient greater than a preset threshold to obtain the target feature vector of the bay area. Based on the target feature vector of the bay area, the trained XGBoost model is subjected to SHAP interpretability analysis, and the target feature vector most correlated with salinity is determined as the optimal feature vector of the bay area. The adaptation evaluation module 204 is used to perform salinity inversion on the bay area using the trained XGBoost model based on the optimal feature vector, and divide the bay area into areas with different salinity levels; and to evaluate the compatibility between the salinity of the bay area and the mangroves based on the areas with different salinity levels and visual images of the mangroves.

[0047] Specific limitations regarding the compatibility assessment device for bay salinity and mangroves can be found in the limitations of the compatibility assessment method for bay salinity and mangroves described above, and will not be repeated here. Each module in the aforementioned compatibility assessment device for bay salinity and mangroves can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0048] The present invention also provides a computer-readable storage medium storing a computer program that can be used to execute the above-described... Figure 1 This paper presents a method for assessing the compatibility between bay salinity and mangroves.

[0049] The present invention also provides Figure 3 The schematic diagram of the computer device shown is as follows: Figure 3As shown, at the hardware level, this computer device includes a processor, internal bus, network interface, memory, and non-volatile memory, and may also include other hardware required for business operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then executes it to achieve the above. Figure 1 This paper presents a method for assessing the compatibility between bay salinity and mangroves.

[0050] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

Claims

1. A method for assessing the compatibility between bay salinity and mangrove forests, characterized in that, include: Acquire Landsat-8 OLI satellite imagery and mangrove visual imagery of the Gulf region; and extract the first few band data from the Landsat-8 OLI imagery. Based on the band data, several vegetation indices, several water body indices, and several band combination feature vectors are determined as the total feature vectors of the bay area. The salinity prediction capability of all feature vectors in the bay area was evaluated using a random forest regressor. Several feature vectors with the highest salinity prediction capability were selected, and the Pearson correlation coefficient between each pair of feature vectors was determined. Feature vectors with Pearson correlation coefficients greater than a preset threshold were removed to obtain the target feature vectors of the bay area. Based on the target feature vectors of the bay area, the trained XGBoost model was subjected to SHAP interpretability analysis, and the target feature vector most correlated with salinity was determined as the optimal feature vector of the bay area. Based on the optimal feature vector, the trained XGBoost model is used to perform salinity inversion in the bay area, dividing the bay area into regions with different salinity levels; based on the regions with different salinity levels and visual images of mangroves, the compatibility between the salinity of the bay area and the mangroves is evaluated.

2. The method for assessing the compatibility between bay salinity and mangroves as described in claim 1, characterized in that, The acquisition of Landsat-8 OLI imagery of the bay area specifically includes: Acquire raw Landsat-8 OLI satellite imagery of the target bay area with cloud cover below 10%; Radiometric calibration and atmospheric correction were performed sequentially on the original Landsat-8 OLI satellite imagery to obtain Landsat-8 OLI imagery of the Gulf region.

3. The method for assessing the compatibility between bay salinity and mangroves as described in claim 1, characterized in that, The acquisition of the visual images of the mangrove forest specifically includes: Aerial photography of the mangroves was obtained by using multi-rotor drones to take aerial photos of the estuaries of rivers flowing into the sea in the bay area and several synchronous stations set up in the core distribution area of ​​mangroves. Visual interpretation of the aerial images of the mangrove forest yields visual images of the mangrove forest.

4. The method for assessing the compatibility between bay salinity and mangroves as described in claim 1, characterized in that, The band combination feature vector includes a dual-band combination feature vector and a tri-band combination feature vector; The dual-band combined feature vector is determined through multiple numerical operations on two non-overlapping band data; the three-band combined feature vector is determined through numerical operations on three non-overlapping band data.

5. The method for assessing the compatibility between bay salinity and mangroves as described in claim 1, characterized in that, The evaluation of the salinity prediction capability of using a random forest regressor on all feature vectors of the bay area specifically includes: All feature vectors are input into a random forest regressor, and salinity prediction is performed for each feature vector by constructing multiple decision trees; The ability of a feature to predict salinity is evaluated by averaging the salinity predictions of all decision trees for each feature vector. The higher the average value of the salinity prediction results, the stronger the predictive ability of the feature vector; conversely, the lower the average value, the weaker the predictive ability of the feature vector.

6. The method for assessing the compatibility between bay salinity and mangroves as described in claim 1, characterized in that, When performing SHAP interpretability analysis on the trained XGBoost model, the average absolute value of SHAP is used as a measure of the feature importance of the target feature vector in the bay area to determine the optimal water feature vector with the highest response to salinity prediction.

7. The method for assessing the compatibility between bay salinity and mangroves as described in claim 1, characterized in that, The process involves using a trained XGBoost model to retrieve salinity in the bay area, dividing it into regions with different salinity levels. Based on these regions and visual images of the mangroves, a compatibility assessment of the bay area's salinity and mangrove ecosystem is performed, specifically including: The spatial distribution of salinity across the entire target bay was inverted using the optimal XGBoost model, and a salinity spatial distribution map was drawn, dividing the bay area into high-salinity and low-salinity zones according to salinity levels. Based on the regional distribution of high-salinity and low-salinity areas, and combined with visual images of mangroves, an assessment of mangrove salinity adaptability was conducted, specifically including: Based on the distribution areas of mangroves and the regional distribution of high-salinity and low-salinity areas in visual images of mangroves, the critical salinity threshold for mangrove survival was determined. Based on the degradation of mangroves in visual images and the regional distribution of high-salinity and low-salinity areas, the optimal salinity range for the growth of *Derris tectorum* was determined. Based on the critical salinity threshold for mangrove survival and the optimal salinity range for drupe vine growth, a correlation model between salinity gradient and mangrove distribution and growth status is established for the protection of mangroves in the target area and the special management of drupe vine.

8. A device for assessing the compatibility of bay salinity with mangroves, characterized in that, include: The data acquisition module is used to acquire Landsat-8 OLI satellite images and mangrove visual images of the Gulf region; and to extract the first few band data from the Landsat-8 OLI images. The feature construction module is used to determine several vegetation indices, several water body indices, and several band combination feature vectors based on the band data, as the total feature vectors of the bay area. The feature selection module is used to evaluate the salinity prediction capability of all feature vectors in the bay area using a random forest regressor, select a number of feature vectors with the highest salinity prediction capability, determine the Pearson correlation coefficient between each pair of feature vectors, and remove feature vectors with a Pearson correlation coefficient greater than a preset threshold to obtain the target feature vectors of the bay area. The target feature vector based on the bay area is used to perform SHAP interpretability analysis on the trained XGBoost model, and the target feature vector most related to salinity is determined as the optimal feature vector for the bay area. The adaptation evaluation module is used to perform salinity inversion on the bay area using the trained XGBoost model based on the optimal feature vector, and divide the bay area into regions with different salinity levels; based on the regions with different salinity levels and visual images of mangroves, the module evaluates the compatibility between the salinity of the bay area and the mangroves.

Images