Basin ecological health condition evaluation method based on remote sensing and machine learning

Abstract

The invention discloses a drainage basin ecological health condition evaluation method based on remote sensing and machine learning, and the method comprises the steps: selecting a plurality of indexes representing the ecological environment condition of a drainage basin region, and building a remote sensing ecological health condition index system; the method comprises the following steps: constructing an inversion model of each index by referring to the inversion model of each index and combining a random forest algorithm on the basis of multi-source remote sensing data such as a multispectral image, measured data and statistical data, and verifying and optimizing the inversion model of each index by synchronizing the measured data; based on a principal component analysis method, determining a drainage basin remote sensing ecological index for evaluating the health condition of the drainage basin; and evaluating the ecological environment condition of the watershed region based on the watershed remote sensing ecological index. According to the method, the remote sensing technology and the machine learning technology are combined, the drainage basin remote sensing ecological health condition index system is constructed, the ecological environment condition of the drainage basin area is comprehensively and scientifically evaluated, and a scientific basis and technical guidance are provided for protection, restoration and management of the drainage basin area.

Description

Technical Field

[0001] The present invention relates to the technical field of ecological environment monitoring, and in particular to a method for evaluating the ecological health status of a watershed based on remote sensing and machine learning. Background Art

[0002] As the combined impacts of global climate change and human activities intensify, the ecological and environmental quality of river basins, as complex ecosystems coupled with multiple factors—water, soil, air, and life—has become a key focus in the development of ecological civilization. The existing "Technical Specification for Ecological and Environmental Status Assessment" (HJ / T 192-2015) applies to areas above the county level, ecological functional zones, cities / urban clusters, and nature reserves. Furthermore, limitations such as an annual assessment frequency hinder the scientific evaluation of ecological and environmental diversity across river basins. Traditional regional ecological and environmental monitoring mainly relies on ground-based measurement methods, with periodic sampling to obtain basic parameters such as biological abundance index and vegetation cover. Although this method has high accuracy in the evaluation of county-level and above administrative units, its limitations are becoming increasingly prominent: first, it relies on manual sampling, which is costly in terms of manpower and time, and the interannual evaluation frequency makes it difficult to capture the rapid dynamic changes in the basin's ecological environment; second, the monitoring process may cause physical interference to fragile ecosystems, such as soil disturbance and vegetation destruction; third, the evaluation index system is designed based on administrative divisions, which fails to fully reflect the spatial heterogeneity of the basin (such as hydrological connectivity and pollutant migration patterns), resulting in insufficient applicability and scientificity of the evaluation results at the basin scale.

[0003] With the deepening of ecological civilization construction, people's attention to regional ecological environment conditions has gradually increased, and the requirements for the timeliness, continuity, and diversity of ecological environment monitoring and evaluation have also increased accordingly. These requirements provide an opportunity for the application of remote sensing technology in ecological environment monitoring.

[0004] Remote sensing technology has gradually become an important means of watershed ecological monitoring due to its advantages of large range, high timeliness and non-contact. Existing studies mostly construct remote sensing ecological index (RSEI) based on a few spectral indices such as greenness (NDVI), humidity (Wetness), and dryness (NDBSI). However, its indicator dimensions are single and it is difficult to comprehensively characterize the synergistic mechanism of multiple factors in the watershed. For example, the traditional RSEI model does not include key indicators such as water cleanliness, soil stress intensity, and spatial distribution of nutrients, resulting in insufficient sensitivity to typical ecological problems in the watershed, such as water environment degradation and the spread of non-point source pollution. In addition, the existing evaluation system has limited ability to depict dynamic processes such as the expansion of impervious surfaces and changes in biodiversity gradients in the watershed, making it difficult to support the needs of refined ecological protection decision-making. Summary of the Invention

[0005] The purpose of this invention is to provide a method for evaluating the ecological health status of a watershed based on remote sensing and machine learning. By combining remote sensing technology and machine learning technology, a watershed remote sensing ecological health status indicator system is constructed, and multiple indicators are combined to more comprehensively and scientifically evaluate the ecological environment status of the watershed area, providing a scientific basis and technical guidance for the protection, restoration and management of the watershed area.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A method for evaluating the ecological health status of a watershed based on remote sensing and machine learning, comprising the following steps:

[0008] S1. Establish a remote sensing ecological health indicator system for evaluating the ecological status of the watershed, including:

[0009] Select multiple indicators representing the ecological and environmental status of the watershed area, including greenness, vegetation cover, water cleanliness index, humidity, dryness, land stress index, water density index, impervious surface ratio, biological abundance index, chemical oxygen demand, total nitrogen, chlorophyll a, and total phosphorus, and establish a remote sensing ecological health status indicator system based on these multiple indicators;

[0010] S2. Inversion and verification of each indicator in the remote sensing ecological health status indicator system, including:

[0011] Based on multi-source remote sensing data including multispectral images, measured data, and statistical data, referring to the existing inversion models of various indicators and combining with the random forest algorithm, the inversion models of various indicators are constructed. The inversion models of various indicators are verified and optimized by synchronous measured data, and finally the inversion results of various indicators are obtained;

[0012] S3, based on principal component analysis, determines the watershed remote sensing ecological index for assessing watershed health, including:

[0013] Each indicator was normalized and masked, and principal component analysis was performed to obtain the first principal component result. The first principal component result was subtracted from 1 to construct the initial watershed remote sensing ecological index for assessing watershed health status. The index was normalized again to obtain the watershed remote sensing ecological index for assessing watershed health status.

[0014] S4, evaluate the ecological environment status of the watershed based on the watershed remote sensing ecological index, including:

[0015] The larger the value of the basin remote sensing ecological index, the better the ecological environment of the basin area, and vice versa.

[0016] Furthermore, the remote sensing ecological health status index system represents a function of 13 indicators, namely:

[0017] F(WRSEHI)=F(Greenness,Wetness,VFC,WCI,NDBSI,BRI,WDI,LSI,ISP,COD,TN,TP,Chla)

[0018] Among them, WRSEHI (Watershed Remote Sensing Ecological Health Index) stands for watershed remote sensing ecological index, Greenness stands for greenness, Wetness stands for humidity, VFC stands for vegetation cover, WCI stands for water cleanliness index, NDBSI stands for dryness, BRI stands for biological abundance index, WDI stands for water density index, LSI stands for land stress index, ISP stands for impervious surface ratio, COD stands for chemical oxygen demand, TN stands for total nitrogen, TP stands for total phosphorus, and Chla stands for chlorophyll a.

[0019] Furthermore, the aforementioned S2, inversion and verification of each indicator in the remote sensing ecological health status indicator system, specifically includes:

[0020] (1) NDVI is selected to represent the greenness index in the remote sensing ecological health status index system, and the formula is as follows:

[0021]

[0022] Where B4 and B8 correspond to the reflectance of the red band and near-red band in Sentinel-2A, respectively;

[0023] (2) The humidity index in the index system is inverted using the Tasselled Cap method, and the formula is as follows:

[0024]

[0025] Where B2, B3, B4, B8, B11, and B12 correspond to the reflectivity of the 2nd, 3rd, 4th, 8th, 11th, and 12th bands in Sentinel-2A;

[0026] (3) The pixel binary model in the mixed pixel method is used to extract the vegetation coverage VFC. The formula is as follows:

[0027]

[0028] Where, NDVI veg and NDVI soil are the NDVI values ​​of pure vegetation pixels and pure soil pixels in remote sensing images respectively;

[0029] (4) Based on the difference in spectral slope, the ratio of the green-blue band to the red-green band is selected as the water cleanliness index (WCI). The formula is as follows:

[0030]

[0031] Where: ρ green , ρ blue , ρ red Represent the reflectivity of the green band, blue band, and red band respectively, B green 、B blue 、B red Represent the central wavelengths of the green band, blue band, and red band respectively; ρ3, ρ2, and ρ4 represent the reflectivity of the green band, blue band, and red band respectively;

[0032] (5) The Normalized Difference Built-up and Bareness Index (NDBSI) is selected to represent the dryness index in the index system. The formula is as follows:

[0033]

[0034] NDBSI=(SI+IBI) / 2

[0035] Where SI represents the bare soil index, IBI represents the building index, and B2, B3, B4, B8, and B11 correspond to the reflectance of the 2nd, 3rd, 4th, 8th, and 11th bands in Sentinel-2A, respectively.

[0036] (6) The impervious surface ratio index is extracted using the assignment analysis method, and the formula is as follows:

[0037]

[0038] Where S 林 、S 草 、S 水 、S 耕地 、S 建筑 、S 未利用地 They represent the area of ​​forest land, grassland, water wetland, cultivated land, construction land and unused land in the region respectively. 总 Indicates the total area of ​​the region;

[0039] (7) The ratio analysis method is used to extract the water density index in the index system. The formula is as follows:

[0040] Water density index = A wat × water area / regional area

[0041] Where A watis the normalization coefficient of water area. Water area includes rivers and lakes in surface water resources and is calculated using the water area in the land use classification results.

[0042] (8) The assignment analysis method was used to extract the biological abundance index, and the formula is as follows:

[0043]

[0044] Where A bio is the normalization coefficient of biological abundance index, S 林 、S 草 、S 水 、S 耕地 、S 建筑 、S 未利用地 They represent the area of ​​forest land, grassland, water wetland, cultivated land, construction land and unused land in the region respectively. 总 Indicates the total area of ​​the region;

[0045] (9) The land stress index was extracted using the assignment analysis method. The formula is as follows:

[0046]

[0047] Where: A ero is the normalized coefficient of land stress index, S 重度侵蚀 、S 中度侵蚀 、S 建设用地 、S 其它土地胁迫 Respectively represent the areas of severe erosion, moderate erosion, construction land, and other land stress in the region, S 总 Indicates the total area of ​​the region;

[0048] (10) The optimal spectral parameter vegetation index was screened by the variable importance criterion, and the random forest model was applied to establish the chlorophyll a remote sensing random forest inversion model, COD remote sensing random forest inversion model, TN remote sensing random forest inversion model, and TP remote sensing random forest inversion model suitable for the watershed.

[0049] Furthermore, the S3 determines the watershed remote sensing ecological index for evaluating the health status of the watershed based on the principal component analysis method, specifically including:

[0050] S301, normalizing 13 indicators including greenness, vegetation cover, water cleanliness index, humidity, dryness, soil stress index, water density index, impervious surface ratio, biological abundance index, chemical oxygen demand, total nitrogen, chlorophyll a, and total phosphorus to obtain 13 normalized indicators;

[0051] S302: Mask the 13 normalized indicators and perform principal component analysis to obtain the principal component results. Subtract the first principal component result from 1 to construct the initial watershed remote sensing ecological index for assessing the health of the watershed, namely:

[0052] WRSEHI0=1-{PCA1[ρ(Greenness', Wetness', VFC', WCI', NDBSI', BRI', WDI', LSI', ISP', COD', TN', TP', Chla')]}

[0053] Among them, WRSEHI0 represents the initial watershed remote sensing ecological index used to assess the health status of the watershed, PCA1 represents the first principal component result, ρ represents mask processing, Greenness', Wetness', VFC', WCI', NDBSI', BRI', WDI', LSI', ISP', COD', TN', TP', and Chla' represent the normalized greenness, wetness, vegetation cover, water cleanliness index, dryness, biological abundance index, water density index, land stress index, impervious surface rate, chemical oxygen demand, total nitrogen, total phosphorus, and chlorophyll a index, respectively;

[0054] S303, normalizing the initial watershed remote sensing ecological index constructed for evaluating the health status of the watershed to obtain the watershed remote sensing ecological index for evaluating the health status of the watershed, namely:

[0055] WRSEHI=(WRSEHI0-WRSEHI min ) / (WRSEHI max -WRSEHI min )

[0056] WRSEHI is a watershed remote sensing ecological index used to assess watershed health. min Indicates the minimum value of the watershed remote sensing ecological index, WRSEHI max Indicates the maximum value of the watershed remote sensing ecological index.

[0057] Furthermore, the step S302 of performing principal component analysis to obtain principal component results specifically includes:

[0058] SPSS was used to construct the correlation matrix between each indicator and the basin remote sensing ecological index, analyze the contribution of each indicator to the basin remote sensing ecological index, and select the principal component with a cumulative contribution rate of more than 85% as the final analysis object to obtain the results of each principal component.

[0059] Furthermore, the above S4 evaluates the ecological environment of the watershed region based on the watershed remote sensing ecological index, specifically including:

[0060] The WRSEHI value range is 0-1, and the WRSEHI is divided into 5 levels with an interval of 0.2. The values ​​are statistically divided into excellent, good, fair, poor, and poor from large to small;

[0061] The [0-0.2) interval is the “poor” level, indicating that the ecological health status is seriously damaged;

[0062] The [0.2-0.4) interval is the "poor" level, indicating that there are obvious problems with the ecological health status;

[0063] The range [0.4-0.6) is “general”, indicating that ecological health is at a medium level;

[0064] The range [0.6-0.8) is “good”, indicating that the ecological health status is relatively good;

[0065] The [0.8-1) range is the "excellent" level, indicating that the ecological health status is excellent.

[0066] On the other hand, the present invention also provides a watershed ecological health assessment system based on remote sensing and machine learning, which is used to implement the above-mentioned watershed ecological health assessment method based on remote sensing and machine learning, including:

[0067] The indicator system construction module is used to establish a remote sensing ecological health indicator system for evaluating the ecological status of the watershed, including: selecting multiple indicators representing the ecological environment status of the watershed area, including greenness, vegetation cover, water cleanliness index, humidity, dryness, land stress index, water density index, impervious surface ratio, biological abundance index, chemical oxygen demand, total nitrogen, chlorophyll a, and total phosphorus, and establishing a remote sensing ecological health indicator system based on multiple indicators;

[0068] The indicator inversion module is used to invert and verify each indicator in the remote sensing ecological health status indicator system. This module includes: building an inversion model for each indicator based on multi-source remote sensing data including multispectral images, measured data, and statistical data, referring to existing inversion models for each indicator, and combining it with the random forest algorithm. The inversion model for each indicator is then verified and optimized using synchronized measured data to ultimately obtain the inversion results for each indicator.

[0069] The WRSEHI calculation module is used to determine the watershed remote sensing ecological index for assessing watershed health based on the principal component analysis method, including: normalizing and masking each indicator, performing principal component analysis transformation to obtain the first principal component result, subtracting the first principal component result from 1 to construct the initial watershed remote sensing ecological index for assessing watershed health, and normalizing again to obtain the watershed remote sensing ecological index for assessing watershed health;

[0070] The evaluation module is used to evaluate the ecological environment status of the watershed area based on the watershed remote sensing ecological index, including: the larger the value of the watershed remote sensing ecological index, the better the ecological environment status of the watershed area, and vice versa.

[0071] In addition, the present invention provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, the method for evaluating the ecological health status of a watershed based on remote sensing and machine learning as described above is implemented.

[0072] The present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed, implements the above-mentioned method for evaluating the ecological health status of a watershed based on remote sensing and machine learning.

[0073] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects: the watershed ecological health status evaluation method based on remote sensing and machine learning provided by the present invention selects 13 indicators including greenness, vegetation coverage, water cleanliness index, humidity, dryness, land stress index, water density index, impervious surface rate, biological abundance index, chemical oxygen demand, total nitrogen, chlorophyll a, and total phosphorus to establish a remote sensing ecological health status index system; among them, 8 indicators including greenness, vegetation coverage, water cleanliness index, humidity, dryness, land stress index, water density index, and impervious surface rate are closely related to human activities and are important indicators for humans to intuitively feel the quality of the watershed ecological environment. The biological abundance index, chemical oxygen demand, total nitrogen, chlorophyll a, and total phosphorus are closely related to human activities and are important indicators for humans to intuitively feel the quality of the watershed ecological environment. Five indicators, including oxygen demand, total nitrogen, chlorophyll a, and total phosphorus, can effectively reflect the ecological and environmental status of the ecosystem in the basin area. Combining diversified indicators to achieve multivariate analysis, thereby improving the scientific nature of the remote sensing ecological health status index system; after completing the construction of the index system, referring to the direct calculation methods or related inversion models of existing indicators, an inversion model suitable for evaluating the basin area is established, and the dimensional inconsistency of each inversion result is eliminated through the normalization method. Finally, the WRSEHI of the evaluated basin area is obtained through the principal component analysis (PCA) method, and the ecological and environmental status of the basin is evaluated and analyzed. The ecological and environmental status of the basin area can be obtained simply and intuitively, which has a certain universality and is easy to promote and use.

[0074] In summary, the present invention constructs a basin remote sensing ecological health status index system to evaluate the ecological environment status of the basin area, which not only provides a scientific basis and technical guidance for the protection, restoration and management of basin areas of various scales and types, but also has a certain driving effect on the development of a universal model of basin-scale remote sensing ecological index. It solves the limitation factors of the "Technical Specifications for Ecological Environment Status Evaluation" (HJ / T 192-2015) that it is applicable to areas above the county level, ecological functional zones, cities / urban agglomerations and nature reserves, and the evaluation frequency is 1 year / time. It can more scientifically and comprehensively evaluate the ecological environment difference characteristics of the basin. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0076] Figure 1 This is a flowchart of a method for evaluating the ecological health status of a watershed based on remote sensing and machine learning according to an embodiment of the present invention;

[0077] Figure 2 This is a schematic diagram of the dynamic changes of WRSEHI in the Yingwen River Basin in an embodiment of the present invention, wherein (a) represents the dynamic monitoring schematic diagram from 2019 to 2020, (b) represents the dynamic monitoring schematic diagram from 2020 to 2021, (c) represents the dynamic monitoring schematic diagram from 2021 to 2022, and (d) represents the dynamic monitoring schematic diagram from 2022 to 2023. DETAILED DESCRIPTION

[0078] The following are detailed examples of the present invention. These examples are intended to explain the present invention and are not to be construed as limiting the present invention. Where specific techniques or conditions are not specified in the examples, the techniques or conditions described in the literature in the art or in the product specifications were used. Materials or instruments used, where the manufacturer is not specified, are commercially available conventional products.

[0079] like Figure 1 As shown, the method for evaluating the ecological health status of a watershed based on remote sensing and machine learning provided by an embodiment of the present invention includes the following steps:

[0080] S1. Establish a remote sensing ecological health indicator system for evaluating the ecological status of the watershed, including:

[0081] We selected several indicators representing the ecological and environmental status of the watershed, including greenness, wetness, vegetation fractional cover (VFC), water cleanliness index (WCI), normalized difference built-up and soil index (NDBSI), biological richness index (BRI), water density index (WDI), land stress index (LSI), impervious surface percentage (ISP), chemical oxygen demand (COD), total nitrogen (TN), total phosphorus (TP), and chlorophyll a (Chla). A remote sensing ecological health index system was established based on these 13 indicators. The function is expressed as follows:

[0082] F(WRSEHI)=F(Greenness, Wetness, VFC, WCI, NDBSI, BRI, WDI, LSI, ISP, COD, TN, TP, Chla).

[0083] S2. Inversion and verification of each indicator in the remote sensing ecological health status indicator system, including:

[0084] Based on multi-source remote sensing data including multispectral images, measured data, and statistical data, referring to the existing inversion models of various indicators and combining with the random forest algorithm, the inversion models of various indicators are constructed. The inversion models of various indicators are verified and optimized by synchronous measured data, and finally the inversion results of various indicators are obtained;

[0085] The inversion process of each indicator is as follows:

[0086] (1) NDVI is selected to represent the greenness index in the remote sensing ecological health status index system, and the formula is as follows:

[0087]

[0088] Where B4 and B8 correspond to the reflectance of the red band and near-red band in Sentinel-2A, respectively;

[0089] (2) The humidity index in the index system is inverted using the Tasselled Cap method, and the formula is as follows:

[0090]

[0091] Where B2, B3, B4, B8, B11, and B12 correspond to the reflectivity of the 2nd, 3rd, 4th, 8th, 11th, and 12th bands in Sentinel-2A;

[0092] (3) The pixel binary model in the mixed pixel method is used to extract the vegetation coverage VFC. The formula is as follows:

[0093]

[0094] Where, NDVI veg and NDVI soil are the NDVI values ​​of pure vegetation pixels and pure soil pixels in remote sensing images, respectively. In view of the existence of outliers and noise in image data, the NDVI value when the cumulative histogram reaches 5% is taken as the NDVIsoil value, and the NDVI value when the cumulative histogram reaches 95% is taken as the NDVIveg value.

[0095] (4) Based on the difference in spectral slope, the ratio of the green-blue band to the red-green band is selected as the water cleanliness index (WCI). The formula is as follows:

[0096]

[0097] Where: ρ green , ρ blue , ρ red Represent the reflectivity of the green band, blue band, and red band respectively, B green 、B blue 、B red Represent the central wavelengths of the green band, blue band, and red band respectively; ρ3, ρ2, and ρ4 represent the reflectivity of the green band, blue band, and red band respectively;

[0098] (5) The normalized difference building-bare soil index is selected to represent the dryness index in the index system. The formula is as follows:

[0099]

[0100] NDBSI=(SI+IBI) / 2

[0101] Where SI represents the bare soil index, IBI represents the building index, and B2, B3, B4, B8, and B11 correspond to the reflectance of the 2nd, 3rd, 4th, 8th, and 11th bands in Sentinel-2A, respectively.

[0102] (6) The impervious surface ratio index is extracted using the assignment analysis method, and the formula is as follows:

[0103]

[0104] Where S 林 、S 草 、S 水 、S 耕地 、S 建筑 、S 未利用地 They represent the area of ​​forest land, grassland, water wetland, cultivated land, construction land and unused land in the region respectively. 总 Indicates the total area of ​​the region;

[0105] (7) The ratio analysis method is used to extract the water density index in the index system. The formula is as follows:

[0106] Water density index = A wat × water area / regional area

[0107] Where A wat is the normalization coefficient of water area. Water area includes rivers and lakes in surface water resources and is calculated using the water area in the land use classification results.

[0108] (8) The assignment analysis method was used to extract the biological abundance index, and the formula is as follows:

[0109]

[0110] Where A bio is the normalization coefficient of biological abundance index, S 林 、S 草 、S 水 、S 耕地 、S 建筑 、S 未利用地 They represent the area of ​​forest land, grassland, water wetland, cultivated land, construction land and unused land in the region respectively. 总 Indicates the total area of ​​the region;

[0111] (9) The land stress index was extracted using the assignment analysis method. The formula is as follows:

[0112]

[0113] Where: A ero is the normalized coefficient of land stress index, S 重度侵蚀 、S 中度侵蚀 、S 建设用地 、S 其它土地胁迫 Respectively represent the areas of severe erosion, moderate erosion, construction land, and other land stress in the region, S 总 Indicates the total area of ​​the region;

[0114] (10) The optimal spectral parameter vegetation index was screened by the variable importance criterion, and the random forest model was applied to establish the random forest inversion models for the chlorophyll a, COD, TN, and TP contents based on remote sensing in the watershed, respectively.

[0115] Traditional monitoring methods for water indicators such as chlorophyll a, COD, TN, and TP levels rely on spot sampling and interpolation analysis to determine the spatial distribution of these indicators. However, inversion models, through mathematical modeling, map these indicators to remote sensing data, yielding the spatial distribution of the entire water body. This approach not only addresses the nonlinear correlations of complex systems but also yields more accurate spatial distribution results with greater efficiency.

[0116] S3, based on principal component analysis, determines the watershed remote sensing ecological index for assessing watershed health, including:

[0117] S301, normalizing 13 indicators including greenness, vegetation cover, water cleanliness index, humidity, dryness, soil stress index, water density index, impervious surface ratio, biological abundance index, chemical oxygen demand, total nitrogen, chlorophyll a, and total phosphorus to obtain 13 normalized indicators;

[0118] S302: Mask the 13 normalized indicators and perform principal component analysis to obtain the principal component results. Subtract the first principal component result from 1 to construct the initial watershed remote sensing ecological index for assessing the health of the watershed, namely:

[0119] WRSEHI0=1-{PCA1[ρ(Greenness', Wetness', VFC', WCI', NDBSI', BRI', WDI', LSI', ISP', COD', TN', TP', Chla')]}

[0120] Among them, WRSEHI0 represents the initial watershed remote sensing ecological index used to assess the health status of the watershed, PC1 represents the first principal component result, ρ represents masking (by unifying the spatial range and eliminating interference areas to provide reliable and clean input data for principal component analysis), Greenness', Wetness', VFC', WCI', NDBSI', BRI', WDI', LSI', ISP', COD', TN', TP', and Chla' represent the normalized greenness, humidity, vegetation cover, water cleanliness index, dryness, biological abundance index, water density index, land stress index, impervious surface rate, chemical oxygen demand, total nitrogen, total phosphorus, and chlorophyll a index, respectively;

[0121] S303, normalizing the initial watershed remote sensing ecological index constructed for evaluating the health status of the watershed to obtain the watershed remote sensing ecological index for evaluating the health status of the watershed, namely:

[0122] WRSEHI=(WRSEHI0-WRSEHI min ) / (WRSEHI max -WRSEHI min )

[0123] WRSEHI is a watershed remote sensing ecological index used to assess watershed health. min Indicates the minimum value of the watershed remote sensing ecological index, WRSEHI max Indicates the maximum value of the watershed remote sensing ecological index.

[0124] S4, evaluate the ecological environment status of the watershed based on the watershed remote sensing ecological index, including:

[0125] The larger the value of the basin remote sensing ecological index, the better the ecological environment of the basin area, and vice versa.

[0126] Taking the Yingwen River Basin as an example, the embodiment of the present invention provides a method for evaluating the ecological health status of the basin based on remote sensing and machine learning. The evaluation process is as follows:

[0127] ① In order to reflect the overall ecological environment status of the current evaluation basin area, the values ​​of each indicator in the remote sensing ecological health status index system are first calculated and statistically analyzed.

[0128] The Water Cleanliness Index (WCI) is an important indicator for evaluating water quality, particularly for identifying black and odorous water. When 0 ≤ WCI ≤ 1, the water quality is poor; when WCI > 1, the water quality is good.

[0129] NDVI values ​​range from -1 to 1, where -1 indicates high reflectance of visible light; 0 indicates rocky or bare soil; and NIR and R are approximately equal. Positive values ​​indicate vegetation cover, and the value increases with vegetation cover. Vegetation cover directly reflects the degree of greenness in an area and is an important indicator of vegetation growth.

[0130] The value range of VFC is [0-100], in %, where [0-10) represents low coverage; [10-30) represents relatively low coverage; [30-50) represents medium coverage; [50-70) represents relatively high coverage; and [70-100) represents high coverage.

[0131] The range of the dryness index is [-1, 1], and the larger the value, the drier it is.

[0132] Humidity can effectively reflect the humidity conditions of water bodies, soil and vegetation. The humidity index range is [-1, 1]. The larger the value, the higher the air humidity.

[0133] The biological abundance index is an index to evaluate the abundance of biological species in the study area. The biological abundance index ranges from [0-100]. The higher the biological abundance index value, the richer the biodiversity.

[0134] The land stress index can measure the degree to which natural land has been damaged by human activities. The land stress index ranges from [0-1]. The higher the land stress index, the greater the degree to which natural land has been damaged by human activities.

[0135] The water density index can measure the abundance level of water resources in the study area. The water density index range is [0-100]. The higher the water density index value, the better the water resources in the area.

[0136] The impervious surface ratio can be used to assess the urbanization level and environmental quality of a region. The impervious surface ratio ranges from 0 to 100, expressed in %. Higher values ​​indicate a higher urbanization rate.

[0137] Chemical oxygen demand (COD), total nitrogen (TN), and total phosphorus (TP) are important indicators for determining water pollution and measuring water quality. The Surface Water Environmental Quality Standard (GB3838-2002) categorizes surface water into five functional categories based on their environmental functions and protection objectives.

[0138] Chlorophyll a (Chl-a) in water is an important indicator for assessing plant growth and water quality. Studies have shown that a chlorophyll a concentration below 0.5 μg / L indicates high water transparency, clear water without noticeable turbidity, and no odor. A chlorophyll a concentration between 0.5 and 3 μg / L indicates moderate water transparency, slight turbidity, and normal water quality with a slight odor. A chlorophyll a concentration between 3 and 10 μg / L indicates poor water transparency, noticeable turbidity, and poor water quality, with odor and signs of corruption. A chlorophyll a concentration above 10 μg / L indicates very poor water transparency, turbidity, and extremely poor water quality, with the presence of algal blooms and eutrophication.

[0139] With reference to relevant standards such as the "Surface Water Environmental Quality Standard" (GB 3838-2002) and the national ecological environmental benchmark "Lake Nutrient Benchmark - Central and Eastern Lake Area (Total Phosphorus, Total Nitrogen, Chlorophyll a)" (2020 edition), as well as existing relevant research, this method defines the corresponding range of the quality level of each indicator in the remote sensing ecological health status index system. The specific division is detailed in Table 1 below.

[0140] Table 1 Corresponding range of each indicator

[0141]

[0142]

[0143] Note: NDVI, values ​​below 0 are generally considered to be water bodies, so [Difference] starts at 2.

[0144] To analyze the contribution of each component to PC1 (WRSEHI), this method used a correlation matrix to analyze the correlations between each component and with the WRSEHI, examining whether the WRSEHI effectively integrates the information from each component. Using SPSS, we constructed a correlation matrix to analyze the contribution of each component to PC1, and selected principal components with a cumulative contribution exceeding 85% as the final analysis targets.

[0145] ③ To more clearly reflect the spatial distribution of the current ecological and environmental status of the watershed, this method divides the WRSEHI (range 0-1) into five levels with an interval of 0.2, and the values ​​are statistically divided from large to small as excellent, good, fair, poor, and poor. Specifically, the range [0-0.2) is classified as "poor", indicating severe ecological health; the range [0.2-0.4) is classified as "poor", indicating obvious ecological health problems; the range [0.4-0.6) is classified as "fair", indicating moderate ecological health; the range [0.6-0.8) is classified as "good", indicating good ecological health; and the range [0.8-1) is classified as "excellent", indicating excellent ecological health.

[0146] ④ In order to accurately reflect the spatiotemporal evolution of the ecological environment in the basin in recent decades, this method uses the same period of a typical year as the time node to calculate the PC1 value and WRSEHI mean of each component in the basin remote sensing ecological health status index system, and analyze the overall change pattern of the basin's WRSEHI.

[0147] In this example, 13 indicators of the Yingwen River Basin's ecological environment were calculated by acquiring remote sensing imagery and measured hydrological data from September 2019 to September 2023. Principal component analysis (PCA) was used to synthesize these 13 indicators. The PC1 values ​​and mean WRSEHI values ​​for each component in the Yingwen River Basin's remote sensing ecological health indicator system were calculated. The mean WRSEHI for the Yingwen River Basin was 0.85, and the PC1 values ​​for each component are shown in the table below. The contribution of each component to PC1 reveals that NDVI, humidity, biomass abundance index, and VFC, which are positively correlated with ecological health, have positive values, while TN, TP, dryness, and impervious surface ratio, which are negatively correlated with ecological health, have negative values. As shown in Table 2, in the statistical results for the study area based on the PCA method, the contribution rates of PC1, PC2, PC3, and PC4 were 53.65%, 13.74%, 9.04%, and 8.68%, respectively. The contribution rate is the ratio of the variance of a single principal component to the total variance of all principal components. It directly reflects the principal component's ability to explain the variation in the original data. For example, PC1 explains 53.65% of the total variance in the data, making it the principal component with the most information. PC2 explains 13.74% of the variance, making it the second most important principal component. PC3 and PC4 explain 9.04% and 8.68% of the variance, respectively, with their importance gradually decreasing.

[0148] The cumulative contribution rate of the eigenvalues ​​of PC1, PC2, PC3, and PC4 reached 85.11% (>85%), indicating that the first four principal components have covered more than 85% of the original information, indicating the rationality of using the principal components of WRSEHI to evaluate the ecological health status of the watershed.

[0149] Table 2 Principal component analysis of WRSEHI in the Yingwen River Basin

[0150]

[0151] ⑤ To further illustrate the temporal and spatial variations in the basin's ecological environment over recent decades, the WRSEHI values ​​at each time point were similarly divided into five levels, with intervals of 0.2. The range [0-0.2) is designated "poor"; the range [0.2-0.4) is designated "poor"; the range [0.4-0.6) is designated "fair"; the range [0.6-0.8) is designated "good"; and the range [0.8-1) is designated "excellent." Furthermore, by statistically analyzing the area of ​​each ecological level in the basin and the corresponding area of ​​WRSEHI dynamics, we analyzed the rate of good and good ecological levels in the basin and their changes, and analyzed the effectiveness of ecological restoration at each stage. In this example, as shown in Table 3, the WRSEHI values ​​in the Yingwen River basin, from highest to lowest, correspond to "good," "excellent," "fair," and "poor," accounting for 25%, 46.43%, 10.17%, and 14.29%, respectively. The Yingwen River Basin's ecological rating of "Excellent" (Good) has remained above 60% over the past five years, reaching 66%, 60%, 67%, and 68% respectively. The "Poor" rating has remained below 10% over the past five years. Meanwhile, the area and proportion of "Poor" and "Relatively Poor" ecological ratings have declined significantly, with the combined area of ​​these two categories shrinking to near-minimum levels in 2023, representing only 0.18% of the total basin area.

[0152] Table 3 Ecological grade area and proportion in the Yingwen River Basin

[0153]

[0154] ⑥ In order to more intuitively express the spatiotemporal variation of the ecological environment in the basin in terms of spatial distribution, the WRSEHI results at each time node in the basin were processed for change detection, and different colors such as red, yellow, and blue were used to represent the different trends of the ecological environment in the basin, such as deterioration, improvement, and stability (as shown in Table 4 and Table 5). Figure 2). In this example, from 2019 to 2023, the annual fluctuation range in the Yingwen River Basin was small, and all fluctuated within the adjacent grade range. From 2019 to 2020, the ecological health of the basin remained basically stable, with most areas (87.81%) not experiencing significant changes. At the same time, 12.19% of the areas showed a positive trend, indicating that the water environment was relatively stable during this period. However, from 2020 to 2021, the ecological health of the basin declined to a certain extent, with 26.35% of the areas experiencing a decrease in ecological health, and only 4.69% of the areas experiencing improvement. This may be related to changes in the external environment or increased pollution loads. From 2021 to 2022, the ecological health of the basin gradually recovered, with 28.56% of the areas experiencing an improvement in ecological health, but 12.07% of the areas still deteriorated, indicating that water health still fluctuates during the recovery process. From 2022 to 2023, the health of the river basin tended to stabilize, with no significant changes in 93.25% of the areas, indicating that the river basin ecosystem gradually entered a relatively balanced state under governance and natural regulation. However, the ecological health level of 3.53% of the areas still declined, indicating that the local water environment has not yet fully recovered.

[0155] Table 4 Dynamic changes and proportions of WRSEHI in the Yingwen River Basin

[0156]

[0157] This method involves measuring chlorophyll a in water samples from the watershed in the laboratory using spectrophotometry. Pearson correlation analysis is then performed between the spectral band ratios and the measured chlorophyll a content. Variable importance criteria are used to select the optimal spectral parameters, and a random forest model is used to establish an inversion model for chlorophyll a content. COD is measured in water samples from the watershed in the laboratory using the dichromate method. Pearson correlation analysis is then performed between the spectral band ratios and the measured COD content, and the optimal spectral parameters are selected using the variable importance criteria. A random forest model is used to establish an inversion model for COD content. TN is measured in water samples from the watershed in the laboratory using alkaline potassium persulfate digestion UV spectrophotometry. Pearson correlation analysis is then performed between the spectral band ratios and the measured TN content, and the optimal spectral parameters are selected using the variable importance criteria. A random forest model is used to establish an inversion model for TN content. TP is measured in water samples from the watershed in the laboratory using ammonium molybdate spectrophotometry. Then, Pearson correlation analysis was performed on the spectral band ratios and the measured TP content, and the optimal spectral parameters were screened using the variable importance criterion. Finally, an inversion model for TP content was established using the random forest model.

[0158] Among the many methods for calculating comprehensive indices, principal component analysis (PCA) can transform relevant multivariate spatial data into a small number of low-correlation comprehensive indices by calculating the standardized matrix eigenvalues ​​and eigenvectors, and then calculating the principal component contribution rate and cumulative contribution rate. This method maximizes the information reflected by the larger number of variables using fewer comprehensive indices. Therefore, this method uses the PCA variation method to obtain the WRSEHI for the evaluated watershed area.

[0159] In summary, to accurately reflect the spatiotemporal evolution of the ecological and environmental status of a watershed over recent decades, this method uses the same time period in a typical year as a time node to calculate the PC1 values ​​and mean WRSEHI values ​​of each component of the watershed remote sensing ecological health index system and analyze the overall variation of the watershed WRSEHI. To further illustrate the spatiotemporal variation of the watershed's ecological and environmental status over recent decades, the WRSEHI values ​​at each time node are similarly divided into five levels with an interval of 0.2. Furthermore, by calculating the area within each ecological level and the corresponding area of ​​dynamic WRSEHI changes, the rate of good and excellent ecological levels in the watershed and its changes are analyzed, as well as the effectiveness of ecological restoration at each stage. To more intuitively and three-dimensionally represent the spatiotemporal variation of the watershed's ecological and environmental status, change detection and difference processing are performed on the WRSEHI results at each time node in the watershed, using red, yellow, and blue to represent different trends in the watershed's ecological and environmental status, such as deterioration, improvement, and stability.

[0160] Example 2

[0161] The present invention also provides a watershed ecological health assessment system based on remote sensing and machine learning, which is used to implement the above-mentioned watershed ecological health assessment method based on remote sensing and machine learning, including:

[0162] The indicator system construction module is used to establish a remote sensing ecological health indicator system for evaluating the ecological status of the watershed, including: selecting multiple indicators representing the ecological environment status of the watershed area, including greenness, vegetation cover, water cleanliness index, humidity, dryness, land stress index, water density index, impervious surface ratio, biological abundance index, chemical oxygen demand, total nitrogen, chlorophyll a, and total phosphorus, and establishing a remote sensing ecological health indicator system based on multiple indicators;

[0163] The indicator inversion module is used to invert and verify each indicator in the remote sensing ecological health status indicator system. This module includes: building an inversion model for each indicator based on multi-source remote sensing data including multispectral images, measured data, and statistical data, referring to existing inversion models for each indicator, and combining it with the random forest algorithm. The inversion model for each indicator is then verified and optimized using synchronized measured data to ultimately obtain the inversion results for each indicator.

[0164] The WRSEHI calculation module is used to determine the watershed remote sensing ecological index for assessing watershed health based on principal component analysis, including: normalizing and masking each indicator, performing principal component analysis transformation to obtain each principal component result, subtracting the first principal component result from 1 to construct the initial watershed remote sensing ecological index for assessing watershed health, and performing normalization again to obtain the watershed remote sensing ecological index for assessing watershed health;

[0165] The evaluation module is used to evaluate the ecological environment status of the watershed area based on the watershed remote sensing ecological index, including: the larger the value of the watershed remote sensing ecological index, the better the ecological environment status of the watershed area, and vice versa.

[0166] Example 3

[0167] The present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method for evaluating the ecological health status of a watershed based on remote sensing and machine learning is implemented.

[0168] The present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed, implements the above-mentioned method for evaluating the ecological health status of a watershed based on remote sensing and machine learning.

[0169] Matters not covered by the present invention are known technologies.

[0170] While various embodiments of the present invention have been described above, the above descriptions are intended to be illustrative, non-exhaustive, and not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is selected to best explain the principles of the embodiments, their practical applications, or improvements to existing technologies, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for evaluating the ecological health status of a watershed based on remote sensing and machine learning, characterized in that: The following steps are involved: S1. Establish a remote sensing ecological health indicator system for evaluating the ecological status of the watershed, including: Select multiple indicators representing the ecological and environmental status of the watershed area, including greenness, vegetation cover, water cleanliness index, humidity, dryness, land stress index, water density index, impervious surface ratio, biological abundance index, chemical oxygen demand, total nitrogen, chlorophyll a, and total phosphorus, and establish a remote sensing ecological health status indicator system based on these multiple indicators; S2. Inversion and verification of each indicator in the remote sensing ecological health status indicator system, including: Based on multi-source remote sensing data including multispectral images, measured data, and statistical data, referring to the existing inversion models of various indicators and combining with the random forest algorithm, the inversion models of various indicators are constructed. The inversion models of various indicators are verified and optimized by synchronous measured data, and finally the inversion results of various indicators are obtained; S3, based on principal component analysis, determines the watershed remote sensing ecological index for assessing watershed health, including: Each indicator was normalized and masked, and principal component analysis was performed to obtain the first principal component result. The first principal component result was subtracted from 1 to construct the initial watershed remote sensing ecological index for assessing watershed health status. The index was normalized again to obtain the watershed remote sensing ecological index for assessing watershed health status. S4, evaluate the ecological environment status of the watershed based on the watershed remote sensing ecological index, including: The larger the value of the basin remote sensing ecological index, the better the ecological environment of the basin area, and vice versa.

2. The method for evaluating the ecological health status of a watershed based on remote sensing and machine learning according to claim 1, characterized in that: The remote sensing ecological health status index system represents a function of 13 indicators, namely: F(WRSEHI)=F(Greenness,Wetness,VFC,WCI,NDBSI,BRI,WDI,LSI,ISP,COD,TN,TP,Chla) Among them, WRSEHI stands for Watershed Remote Sensing Ecological Index, Greenness stands for greenness, Wetness stands for humidity, VFC stands for vegetation cover, WCI stands for Water Cleanliness Index, NDBSI stands for Dryness, BRI stands for Biological Richness Index, WDI stands for Water Density Index, LSI stands for Land Stress Index, ISP stands for Impervious Surface Ratio, COD stands for Chemical Oxygen Demand, TN stands for Total Nitrogen, TP stands for Total Phosphorus, and Chla stands for Chlorophyll a.

3. The method for evaluating the ecological health status of a watershed based on remote sensing and machine learning according to claim 2 is characterized in that: S2, the inversion and verification of each indicator in the remote sensing ecological health status indicator system, specifically includes: (1) NDVI is selected to represent the greenness index in the remote sensing ecological health status index system, and the formula is as follows: Where B4 and B8 correspond to the reflectance of the red band and near-red band in Sentinel-2A, respectively; (2) The humidity index in the index system is inverted using the tasseled cap transformation method. The formula is as follows: Where B2, B3, B4, B8, B11, and B12 correspond to the reflectivity of the 2nd, 3rd, 4th, 8th, 11th, and 12th bands in Sentinel-2A; (3) The pixel binary model in the mixed pixel method is used to extract the vegetation coverage VFC. The formula is as follows: Where, NDVI veg and NDVI soil are the NDVI values ​​of pure vegetation pixels and pure soil pixels in remote sensing images respectively; (4) Based on the difference in spectral slope, the ratio of the green-blue band to the red-green band is selected as the water cleanliness index (WCI). The formula is as follows: Where: ρ green , ρ blue , ρ red Represent the reflectivity of the green band, blue band, and red band respectively, B green 、B blue 、B red Represent the center wavelengths of the green band, blue band, and red band respectively; ρ3, ρ2, and ρ4 represent the reflectivity of the green band, blue band, and red band respectively; (5) The normalized difference building-bare soil index is selected to represent the dryness index in the index system. The formula is as follows: NDBSI=(SI+IBI) / 2 Where SI represents the bare soil index, IBI represents the building index, and B2, B3, B4, B8, and B11 correspond to the reflectance of the 2nd, 3rd, 4th, 8th, and 11th bands in Sentinel-2A, respectively. (6) The impervious surface ratio index is extracted using the assignment analysis method, and the formula is as follows: Where S 林 、S 草 、S 水 、S 耕地 、S 建筑 、S 未利用地 They represent the area of ​​forest land, grassland, water wetland, cultivated land, construction land and unused land in the region respectively. 总 Indicates the total area of ​​the region; (7) The ratio analysis method is used to extract the water density index in the index system. The formula is as follows: Water density index = A wat × water area / regional area Where A wat is the normalization coefficient of water area. Water area includes rivers and lakes in surface water resources and is calculated using the water area in the land use classification results. (8) The assignment analysis method was used to extract the biological abundance index, and the formula is as follows: Where A bio is the normalization coefficient of biological abundance index, S 林 、S 草 、S 水 、S 耕地 、S 建筑 、S 未利用地 They represent the area of ​​forest land, grassland, water wetland, cultivated land, construction land and unused land in the region respectively. 总 Indicates the total area of ​​the region; (9) The land stress index was extracted using the assignment analysis method. The formula is as follows: Where: A ero is the normalized coefficient of land stress index, S 重度侵蚀 、S 中度侵蚀 、S 建设用地 、S 其它土地胁迫 Represent the areas of severe erosion, moderate erosion, construction land, and other land stress in the region, S 总 Indicates the total area of ​​the region; (10) The optimal spectral parameter vegetation index was screened by the variable importance criterion, and the random forest model was applied to establish the chlorophyll a remote sensing random forest inversion model, COD remote sensing random forest inversion model, TN remote sensing random forest inversion model, and TP remote sensing random forest inversion model suitable for the watershed.

4. The method for evaluating the ecological health status of a watershed based on remote sensing and machine learning according to claim 1, wherein: S3, based on the principal component analysis method, determines the watershed remote sensing ecological index for assessing the health status of the watershed, specifically including: S301, normalizing 13 indicators including greenness, vegetation cover, water cleanliness index, humidity, dryness, soil stress index, water density index, impervious surface ratio, biological abundance index, chemical oxygen demand, total nitrogen, chlorophyll a, and total phosphorus to obtain 13 normalized indicators; S302: Mask the 13 normalized indicators and perform principal component analysis to obtain the principal component results. Subtract the first principal component result from 1 to construct the initial watershed remote sensing ecological index for assessing the health of the watershed, namely: WRSEHI0=1-{PCA1[ρ(Greenness', Wetness', VFC', WCI', NDBSI', BRI', WDI', LSI', ISP', COD', TN', TP', Chla')]} Among them, WRSEHI0 represents the initial watershed remote sensing ecological index used to assess the health status of the watershed, PCA1 represents the first principal component result, ρ represents mask processing, Greenness', Wetness', VFC', WCI', NDBSI', BRI', WDI', LSI', ISP', COD', TN', TP', and Chla' represent the normalized greenness, wetness, vegetation cover, water cleanliness index, dryness, biological abundance index, water density index, land stress index, impervious surface rate, chemical oxygen demand, total nitrogen, total phosphorus, and chlorophyll a index, respectively; S303, normalizing the initial watershed remote sensing ecological index constructed for evaluating the health status of the watershed to obtain the watershed remote sensing ecological index for evaluating the health status of the watershed, namely: WRSEHI=(WRSEHI0-WRSEHI min ) / (WRSEHI max -WRSEHI min ) WRSEHI is a watershed remote sensing ecological index used to assess watershed health. min Indicates the minimum value of the watershed remote sensing ecological index, WRSEHI max Indicates the maximum value of the watershed remote sensing ecological index.

5. The method for evaluating the ecological health status of a watershed based on remote sensing and machine learning according to claim 4 is characterized in that: The step S302 of performing principal component analysis to obtain principal component results specifically includes: SPSS was used to construct the correlation matrix between each indicator and the basin remote sensing ecological index, analyze the contribution of each indicator to the basin remote sensing ecological index, and select the principal component with a cumulative contribution rate of more than 85% as the final analysis object to obtain the results of each principal component.

6. The method for evaluating the ecological health status of a watershed based on remote sensing and machine learning according to claim 4, wherein: S4, evaluating the ecological environment status of the watershed area based on the watershed remote sensing ecological index, specifically includes: The WRSEHI value range is 0-1, and the WRSEHI is divided into 5 levels with an interval of 0.

2. The values ​​are statistically divided into excellent, good, fair, poor, and poor from large to small; The [0-0.2) interval is the "poor" level, indicating that the ecological health status is seriously damaged; The [0.2-0.4) range is "poor", indicating that there are obvious problems with the ecological health status; The range [0.4-0.6) is "general", indicating that the ecological health is at a medium level; The range [0.6-0.8) is "good", indicating that the ecological health is relatively good; The [0.8-1) range is "excellent" grade, indicating that the ecological health status is excellent.

7. A watershed ecological health assessment system based on remote sensing and machine learning, applied to implement the watershed ecological health assessment method based on remote sensing and machine learning as described in any one of claims 1 to 6, comprising: The indicator system construction module is used to establish a remote sensing ecological health indicator system for evaluating the ecological status of the watershed, including: selecting multiple indicators representing the ecological environment status of the watershed area, including greenness, vegetation cover, water cleanliness index, humidity, dryness, land stress index, water density index, impervious surface ratio, biological abundance index, chemical oxygen demand, total nitrogen, chlorophyll a, and total phosphorus, and establishing a remote sensing ecological health indicator system based on multiple indicators; The indicator inversion module is used to invert and verify each indicator in the remote sensing ecological health status indicator system. This module includes: building an inversion model for each indicator based on multi-source remote sensing data including multispectral images, measured data, and statistical data, referring to existing inversion models for each indicator, and combining it with the random forest algorithm. The inversion model for each indicator is then verified and optimized using synchronized measured data to ultimately obtain the inversion results for each indicator. The WRSEHI calculation module is used to determine the watershed remote sensing ecological index for assessing watershed health based on the principal component analysis method, including: normalizing and masking each indicator, performing principal component analysis transformation to obtain the first principal component result, subtracting the first principal component result from 1 to construct the initial watershed remote sensing ecological index for assessing watershed health, and normalizing again to obtain the watershed remote sensing ecological index for assessing watershed health; The evaluation module is used to evaluate the ecological environment status of the watershed area based on the watershed remote sensing ecological index, including: the larger the value of the watershed remote sensing ecological index, the better the ecological environment status of the watershed area, and vice versa.

8. An electronic device, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method for evaluating the ecological health status of a watershed based on remote sensing and machine learning is implemented as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that A computer program is stored thereon, which, when executed, implements the watershed ecological health status assessment method based on remote sensing and machine learning as described in any one of claims 1 to 6.

Images