A forest water conservation capacity remote sensing inversion method suitable for Nanling cloudy and rainy areas

Abstract

The present application relates to the technical field of ecological protection, and more particularly to a forest water conservation capacity remote sensing inversion method suitable for Nanling cloudy and rainy areas, aiming at the problem that the existing inversion method cannot overcome the cloud cover and terrain influence, and the precision of forest water conservation capacity is low, the present application obtains optical and microwave remote sensing data, pre-processes the optical and microwave remote sensing data, extracts corresponding optical characteristics and microwave characteristics, weights and fuses the optical characteristics and microwave characteristics, calculates the forest water conservation amount by using a water conservation amount physical model, and generates a Nanling forest water conservation capacity spatial distribution map. The method overcomes the errors caused by weather, terrain shadow and distortion, constitutes a full-weather, high-precision, Nanling special environment-adapted forest water conservation capacity remote sensing inversion system, and realizes accurate quantification of the water conservation capacity of the Nanling special forest ecosystem.

Description

Technical Field

[0001] This invention relates to the field of ecological protection technology, and in particular to a remote sensing inversion method for forest water conservation capacity applicable to the cloudy and rainy areas of the Nanling Mountains. Background Technology

[0002] The Nanling Mountains region is an important ecological security barrier in southern China and one of the main sources of the Pearl River. With its high forest coverage and complex vegetation structure, the water conservation function of its forest ecosystems (including canopy interception, water retention in the litter layer, soil infiltration, and water storage) plays an irreplaceable role in regulating the regional hydrological cycle, reducing flood peaks, purifying water quality, and responding to extreme climate change. Therefore, accurate and dynamic monitoring of the water conservation capacity of the forests in the Nanling Mountains region is of great significance for regional water resource management, ecological benefit assessment, and the formulation of ecological compensation mechanisms.

[0003] Currently, the main methods for monitoring forest water conservation capacity include ground-based measurement and remote sensing inversion. Ground-based measurement involves establishing standard plots and conducting fixed-point observations using equipment such as rain gauges, lysimeters, and artificial rainfall simulators. While this method offers high data accuracy, it is time-consuming, labor-intensive, and destructive. Furthermore, it is limited by the high altitude, steep roads, and complex terrain of the Nanling Mountains, making large-scale, continuous spatial monitoring difficult. Remote sensing inversion utilizes satellite remote sensing data, combined with ecological process models or statistical regression models, to extrapolate large-scale vegetation parameters (such as leaf area index (LAI) and vegetation cover (FVC)) and soil parameters, thereby estimating water conservation capacity. Remote sensing inversion offers advantages such as macroscopic, dynamic, and rapid measurement, and has become the mainstream technology for regional ecological monitoring.

[0004] Although remote sensing inversion methods are well-established in plains or sunny areas, their direct application in the Nanling region still faces the following serious technical challenges:

[0005] (1) The Nanling region has a typical subtropical monsoon climate with high temperature and humidity all year round, and is shrouded in clouds and fog. The frequency of precipitation is high. Currently, the mainstream optical remote sensing satellites mainly rely on visible light and infrared band imaging, which cannot penetrate the cloud layer. Under the interference of cloud and rain weather, the region cannot obtain accurate and effective optical remote sensing data. The time resolution is low, which leads to the embarrassing situation that remote sensing monitoring in the region often has no data when monitoring is needed, and data has already passed by when it is available. It is difficult to capture the dynamic changes of forest water conservation capacity during the alternation of rainy and dry seasons.

[0006] (2) The terrain of Nanling is undulating and crisscrossed by gullies. Under this terrain, the signals received by the sensor not only contain the spectral information of the ground objects, but are also severely affected by the terrain slope, aspect and shadow effect (i.e., the phenomenon of "different spectrum for the same object" or "same spectrum for different objects"). The existing inversion model based on the assumption of flat terrain does not consider or does not fully consider the terrain correction factor, resulting in large errors in the vegetation index and biophysical parameters extracted in the Nanling mountain area, which seriously reduces the accuracy of water conservation capacity estimation.

[0007] (3) Existing general inversion systems mostly focus on monitoring single vegetation growth status (such as using only the NDVI index), lacking multi-layer coupling mechanisms for water conservation in specific forest communities in the Nanling Mountains, such as evergreen broad-leaved forests and mixed coniferous and broad-leaved forests. Moreover, existing inversion systems simply equate vegetation indices with water conservation capacity, ignoring the differences in the response of key hydrological parameters such as the thickness of the understory litter layer and soil porosity to remote sensing signals. This results in unclear physical mechanisms in the inversion results, and the accuracy is difficult to meet the needs of refined ecological management.

[0008] In summary, given the cloudy and rainy environment and complex mountainous terrain of the Nanling Mountains, developing a remote sensing inversion method that can effectively overcome cloud and rain obstruction, eliminate the influence of topography, and has high accuracy in retrieving forest water conservation capacity is a pressing technical challenge that needs to be addressed. Summary of the Invention

[0009] To address the aforementioned problems, this invention aims to provide a remote sensing inversion method for forest water conservation capacity in the Nanling Mountains, a region characterized by frequent cloud cover and rain. Addressing the limitations of existing inversion methods in overcoming cloud cover and topographical influences, and the resulting low accuracy in determining forest water conservation capacity, this invention overcomes cloud cover and rain cover by incorporating microwave remote sensing data with multi-source data fusion. It also eliminates topographical influences by introducing a topographic correction factor and constructing a coupled model of multi-source remote sensing and ground-based measurements to achieve precise quantification of the Nanling forest water conservation capacity.

[0010] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0011] On the one hand, the present invention provides a remote sensing inversion method for forest water conservation capacity applicable to the cloudy and rainy areas of the Nanling Mountains, the method comprising the following steps:

[0012] Acquire optical and microwave remote sensing data for the Nanling region;

[0013] Preprocessing is performed on optical remote sensing data and microwave remote sensing data to obtain preprocessed optical remote sensing data and microwave remote sensing data;

[0014] Based on the preprocessed optical and microwave remote sensing data, optical and microwave features are extracted and fused to obtain a full-coverage set of forest feature parameters.

[0015] Based on the forest characteristic parameter set, the forest water conservation capacity is calculated using a physical model of water conservation capacity.

[0016] The calculated forest water conservation capacity is rendered in a hierarchical manner to generate a spatial distribution map of forest water conservation capacity in the Nanling region.

[0017] Furthermore, the preprocessing of optical remote sensing data and microwave remote sensing data to obtain preprocessed optical remote sensing data and microwave remote sensing data specifically includes the following steps:

[0018] Radiometric calibration and geometric correction are performed on optical and microwave remote sensing data to eliminate sensor errors and geometric position deviations.

[0019] Cloud detection is performed on radiometrically calibrated and geometrically corrected optical remote sensing data to identify cloud-affected, cloud-shadowed, and cloudless areas in the optical remote sensing data.

[0020] Terrain correction was performed on microwave remote sensing data and optical remote sensing data after cloud detection to eliminate shadows and radiation distortions caused by the complex mountainous terrain of the Nanling Mountains, resulting in preprocessed optical and microwave remote sensing data.

[0021] Furthermore, the cloud detection performed on the radiometrically calibrated and geometrically corrected optical remote sensing data to identify cloud-affected, cloud-shadowed, and cloudless areas specifically includes the following steps:

[0022] Based on multispectral feature analysis and thermal infrared feature analysis, areas affected by clouds and cloud shadows are extracted from optical remote sensing data;

[0023] Based on the extracted areas affected by clouds and cloud shadows, a masking process is performed on these areas, replacing them with invalid values ​​while retaining the original data values ​​to obtain cloudless areas.

[0024] Furthermore, the extraction of optical and microwave features based on preprocessed optical and microwave remote sensing data, and the fusion of these features to obtain a comprehensive set of forest feature parameters, specifically includes the following steps:

[0025] Based on cloudless areas obtained from cloud detection, optical features of cloudless areas are extracted using optical remote sensing data.

[0026] Based on the cloud-affected and cloud-shadow-affected areas obtained from cloud detection, microwave features of the cloud-affected and cloud-shadow-affected areas are extracted using microwave remote sensing data.

[0027] The optical and microwave features are weighted and fused to generate a full-coverage set of forest feature parameters.

[0028] Furthermore, the weighted fusion of optical and microwave features to generate a full-coverage forest feature parameter set specifically includes the following steps:

[0029] Calculate the feature information entropy of optical and microwave features, and determine adaptive weights;

[0030] Based on adaptive weights, optical and microwave features are fused by region-specific weighted feature fusion.

[0031] Based on the feature weighted fusion results, the cloud boundaries are feathered and smoothed to obtain a full-coverage set of forest feature parameters.

[0032] Furthermore, the calculation of forest water conservation capacity based on a set of forest characteristic parameters and using a physical model of water conservation capacity specifically includes the following steps:

[0033] Obtain ground-measured data from typical forest plots in the Nanling Mountains;

[0034] A physical model of water conservation capacity was constructed, and the parameters in the model were calibrated using sample plot measurement data. Then, based on the calibrated physical model of water conservation capacity, the forest water conservation capacity at the pixel level for the entire region was calculated.

[0035] On the other hand, the present invention also provides a remote sensing inversion system for forest water conservation capacity in the Nanling cloudy and rainy region, for implementing the remote sensing inversion method for forest water conservation capacity in the Nanling cloudy and rainy region as described above, the system comprising:

[0036] The data acquisition module acquires optical and microwave remote sensing data for the Nanling region.

[0037] The preprocessing and terrain correction module preprocesses the optical remote sensing data and microwave remote sensing data to obtain preprocessed optical remote sensing data and microwave remote sensing data.

[0038] The multi-source feature fusion module extracts optical and microwave features based on preprocessed optical and microwave remote sensing data, and fuses the optical and microwave features to obtain a full-coverage set of forest feature parameters.

[0039] The inversion calculation module calculates the forest water conservation capacity based on the forest characteristic parameter set and using a physical model of water conservation capacity.

[0040] The results output and visualization module performs hierarchical rendering of the calculated forest water conservation capacity, generating a spatial distribution map of forest water conservation capacity in the Nanling region.

[0041] In another aspect, the present invention also provides an electronic device, the device including at least one processor; and a memory communicatively connected to the processor; wherein the memory stores instructions executed by the processor, the instructions being executed by the processor to enable the processor to perform the remote sensing inversion method for forest water conservation capacity applicable to the Nanling cloudy and rainy region as described above.

[0042] The improvement of the present invention compared with the prior art is that,

[0043] 1. This invention acquires data through a dual-channel system of optical and microwave remote sensing. Utilizing the physical properties of microwave remote sensing data—its ability to penetrate clouds and fog, and its immunity to sunlight and precipitation—it combines optical remote sensing data with a pixel-level fusion mechanism to achieve complementary acquisition: "optical data for cloudless areas, microwave data for cloud-covered areas." This allows the system to operate normally in cloudy and rainy weather, overcoming the dependence of traditional optical remote sensing on clear weather at the data source. Specifically addressing the local microclimate characteristics of the Nanling Mountains shrouded in clouds and fog, microwave remote sensing can effectively cover cloud-covered areas such as valleys and steep slopes that are difficult for optical remote sensing to penetrate. This enables comprehensive monitoring of the entire Nanling region, from mountaintops to valleys, and from sunny to shady slopes, eliminating blind spots in traditional methods. Furthermore, it can accurately capture real-time changes in forest water conservation capacity before and after extreme weather events such as rainy seasons, dry seasons, heavy rains, and typhoons, improving the effective acquisition rate of regional remote sensing data and achieving uninterrupted data collection throughout the year.

[0044] 2. This invention uses a high-precision DEM terrain correction subunit and introduces terrain correction factors such as slope and aspect to perform terrain normalization processing on optical remote sensing data and microwave remote sensing data. It corrects the radiation values ​​after terrain distortion to "equivalent radiation values ​​for flat terrain", fundamentally eliminating the phenomenon of "different spectra for the same object and different objects for the same spectrum" caused by mountainous terrain, eliminating errors caused by complex mountainous terrain distortion, greatly improving the accuracy of parameter extraction and inversion, and avoiding the problem of remote sensing distortion in mountainous areas.

[0045] 3. This invention obtains ground-measured data from typical forest plots in the Nanling Mountains and constructs a physical model of water conservation. The parameters in the model are calibrated using the measured data from the plots. Then, based on the calibrated physical model of water conservation, the forest water conservation at the pixel level for the entire region is calculated. This invention overcomes the limitations of existing general inversion models that rely on a single vegetation index, achieving precise adaptation of the model to the Nanling forest ecosystem. Furthermore, it quantifies the water conservation and contribution ratio of the canopy, litter, and soil layers, accurately reflecting the multi-level and multi-stage actual hydrological regulation process of the Nanling forest, making the physical mechanism of the inversion results clearer and the ecological significance more explicit. Attached Figure Description

[0046] Figure 1This is a flowchart illustrating the steps of the remote sensing inversion method for forest water conservation capacity applicable to the cloudy and rainy areas of the Nanling Mountains, as described in this invention.

[0047] Figure 2 This is a spatial distribution map of the Nanling forest water conservation capacity in 2024 generated by the present invention. Detailed Implementation

[0049] To enable those skilled in the art to better understand the technical solutions of the present invention, the technical solutions of the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0050] Example 1: Refer to Appendix Figure 1 As shown in the figure, this embodiment provides a remote sensing inversion method for forest water conservation capacity applicable to the cloudy and rainy areas of the Nanling Mountains. The method includes the following steps:

[0051] Step 1: Acquire optical and microwave remote sensing data for the Nanling region;

[0052] Specifically, the optical remote sensing data in this invention comes from medium- and high-resolution optical satellite imagery and is used to obtain vegetation spectral information in cloudless areas;

[0053] More specifically, medium to high resolution optical satellite imagery can be obtained using Sentinel-2 or Landsat 8;

[0054] Microwave remote sensing data comes from synthetic aperture radar (SAR) data and is used to obtain surface scattering information in cloud-covered areas.

[0055] The synthetic aperture radar can be Sentinel-1, which utilizes the characteristic of radar signals to penetrate clouds and fog to obtain surface scattering information in cloud-covered areas.

[0056] Step 2: Preprocess the optical remote sensing data and microwave remote sensing data to obtain preprocessed optical remote sensing data and microwave remote sensing data;

[0057] Specifically, the preprocessing steps in this process include radiometric calibration, geometric correction, cloud detection, and terrain correction.

[0058] Specifically, step 2 includes the following sub-steps:

[0059] Sub-step 201: Perform radiometric calibration and geometric correction on the optical remote sensing data and microwave remote sensing data to eliminate sensor errors and geometric position deviations;

[0060] More specifically, radiometric calibration includes radiometric calibration of optical remote sensing data and radiometric calibration of microwave remote sensing data (SAR);

[0061] Radiometric calibration of optical remote sensing data involves converting the raw digital quantization values ​​(DN) recorded by the sensor into physically meaningful quantities that reflect the true reflectivity or radiance of the Earth's surface. This process is achieved by applying radiometric calibration coefficients (including gain and offset) provided or verified by the sensor manufacturer. Specifically, it utilizes the following formula:

[0062]

[0063] in, For band The reflectivity of the top of the atmosphere, For band The original digital quantization value, and bands The corresponding radiation calibration gain coefficient and offset. This step aims to eliminate radiation distortion caused by factors such as sensor response differences and decay over time, ensuring the consistency and comparability of radiation values ​​from different time phases and different sensors.

[0064] Radiometric calibration of microwave remote sensing data (SAR) involves converting the raw complex signals (amplitude and phase) recorded by SAR sensors into a standard physical quantity representing the backscattering characteristics of the Earth's surface—the backscattering coefficient. The process consists of the following three parts:

[0065] 1) Apply the calibration constant and antenna pattern correction factor provided by the SAR sensor.

[0066] 2) Consider the radiation variations in range and azimuth (such as range attenuation and antenna gain mode).

[0067] 3) Normalize the signal power and calculate the backscattering coefficient. (Unit: dB), usually using the formula:

[0068]

[0069] Where DN is the calibrated amplitude (or intensity) value, and K is a constant containing all calibration parameters (provided by the SAR sensor metadata). For Sentinel-1 data, precise radiometric calibration is performed using its detailed metadata information to ensure the acquired... This value accurately reflects the microwave scattering characteristics of the Earth's surface, eliminating the influence of gain variations and noise within the SAR sensor system. This step is crucial for the subsequent accurate extraction of microwave features.

[0070] In some embodiments, geometric correction includes geometric correction of optical remote sensing data and geometric correction of microwave remote sensing data (SAR);

[0071] Geometric correction of optical remote sensing data involves establishing a precise mapping between image coordinates (rows and columns) and geographic coordinates based on satellite orbital parameters, sensor attitude (pitch, roll, yaw) data, and ground control points (GCPs). This mapping is achieved by constructing a physical sensor model or rational polynomial coefficients (RPCs). By applying this mapping, the image is resampled and corrected to a specified map projection coordinate system (such as UTM projection), resulting in geometrically corrected optical remote sensing data. This step aims to eliminate image geometric distortions (such as positional shifts, stretching, and warping) caused by factors such as Earth curvature, terrain undulations, sensor attitude instability, and platform motion (changes in speed and altitude).

[0072] Microwave remote sensing data (SAR) geometric correction includes initial positioning and orthorectification;

[0073] Initial positioning uses precise satellite ephemeris data and orbital parameters, combined with a SAR imaging geometric model (range-Doppler model), to initially locate each pixel in the original slant range image into a geocentric coordinate system (such as WGS84).

[0074] Orthorectification aims to completely eliminate positional shifts (overlay, perspective shrinkage, top-bottom displacement) and geometric distortions caused by terrain undulations. It introduces a high-precision Digital Elevation Model (DEM), using a DEM with a resolution matching or higher than that of the SAR data. Using the SAR imaging geometry model and the DEM, the precise location (or vice versa) of each ground point in the SAR image is calculated in a geocentric coordinate system (such as WGS84). The slant-range image is then projected onto the ground plane (or a specified projection coordinate system) and resampled to generate an orthorectified SAR image with accurate geocoding.

[0075] Sub-step 202: Perform cloud detection on the radiometrically calibrated and geometrically corrected optical remote sensing data to identify cloud-affected, cloud-shadowed, and cloudless areas in the optical remote sensing data;

[0076] Specifically, the cloud detection process includes the following sub-steps:

[0077] Sub-step 2021: Based on multispectral feature analysis and thermal infrared feature analysis, extract the areas affected by clouds and cloud shadows from the optical remote sensing data;

[0078] In some embodiments, cloud detection of optical remote sensing data can be performed using multispectral feature analysis and thermal infrared feature analysis. The specific extraction process is as follows:

[0079] 1. Multispectral Feature Analysis

[0080] Cirrus / Thin Cloud Detection: Utilizing the cirrus band (Landsat-8 / 9 B9: 1.36-1.38 μm), which is located in the center of a strong water vapor absorption zone. Water vapor in the lower troposphere almost completely absorbs surface radiation in this band, causing the surface signal to approach zero; however, cirrus clouds have extremely low water vapor content above them, allowing sunlight scattering and reflection from cirrus ice crystals to penetrate the atmosphere and reach the sensor, resulting in significantly high reflectivity (>0.01–0.02).

[0081] Thick cloud detection: Thick clouds exhibit a "white light" characteristic in the visible to near-infrared range—meaning that the reflectance across all bands is similar and relatively high. The green band (B3: 0.53-0.59μm) or the red band (B4: 0.64-0.67μm) is preferentially used, with a reflectance >0.3. The NIR band (B5) is then used to establish an inter-band ratio for verification, setting B4 / B5≈0.8-1.2 to exclude highly reflective vegetation.

[0082] Cloud and snow differentiation: The standard NDSI (Normalized Differential Snow Cover Index) is used, as shown in the following formula:

[0083]

[0084] In the formula, Green represents the green band, and SWIR1 is the first shortwave infrared band. If NDSI > 0.4, it is judged as snow / ice; if NDSI ≤ 0.4, it is judged as cloud.

[0085] 2. Thermal Infrared Feature Analysis

[0086] Brightness temperature threshold method: The higher the cloud top, the lower the temperature: the discrimination threshold is set at 280-290 K in summer and 270 K in winter. Clouds with a brightness temperature less than the threshold are considered clouds.

[0087] The day-night dual-difference method uses a day-night temperature difference of <10 K and high visible light reflectance during the day as a comprehensive indicator of cloud areas to improve the robustness of the judgment.

[0088] In some embodiments, the specific method for cloud shadow detection is as follows:

[0089] Geometric positioning (key step): based on solar altitude angle Cloud base height H cloud and DEM high H terrain Establish a cloud-cloud shadow displacement model:

[0090]

[0091] In the formula, This represents the horizontal displacement of the cloud shadow in the direction away from the sun.

[0092] Based on the results of multispectral feature analysis and thermal infrared feature analysis, the parts that are identified as clouds and cloud-affected areas by both methods are considered as areas affected by clouds and cloud shadows.

[0093] Sub-step 2022: Based on the extracted cloud-affected and cloud-shadowed areas, perform masking on the cloud-affected and cloud-shadowed areas, replace the cloud-affected and cloud-shadowed areas with invalid values ​​(such as NaN), and retain the original data values ​​for the rest to obtain cloudless areas.

[0094] Sub-step 203: Perform terrain correction on the microwave remote sensing data and the optical remote sensing data after cloud detection to eliminate shadows and radiation distortion caused by the complex mountainous terrain of the Nanling Mountains, and obtain the preprocessed optical remote sensing data and microwave remote sensing data.

[0095] Specifically, based on the high-precision digital elevation model (DEM) and C-model of the Nanling region, and combined with slope and aspect data, terrain normalization processing was performed on microwave remote sensing data and optical remote sensing data after cloud detection to eliminate terrain shadows and radiation distortion. The core element of terrain correction is the high-precision DEM. Using a 12.5-meter spatial resolution digital elevation model of the Nanling region, terrain data such as slope angle α and aspect angle β are generated based on the DEM. Terrain normalization is then performed on the optical remote sensing data to eliminate complex mountain shadows and radiation distortion.

[0096] When using the C-model to perform terrain correction on optical remote sensing data, the C-model correction principle is to perform radiometric compensation by establishing a mathematical relationship between terrain factors and illumination intensity, as shown in the following formula:

[0097]

[0098]

[0099] in: Represents the equivalent reflectance of a horizontal surface;

[0100] This represents the solar zenith angle, derived from the solar altitude angle in the metadata. =90° - Solar altitude angle;

[0101] This indicates the solar azimuth angle, obtained directly from metadata.

[0102] α and β are the slope angle and aspect angle derived from the DEM;

[0103] c represents the empirical adjustment parameter (range 0.5-1.5 in the Nanling area).

[0104] The specific execution flow for terrain correction of optical remote sensing data using the C model is as follows:

[0105] ① Calculate the slope angle α and aspect angle β for each pixel based on the DEM;

[0106] ② Select model C, substitute the slope angle α and aspect angle β, and calculate. ;

[0107] ③ Input the surface reflectance product after radiometric calibration Pixel-by-pixel generation of horizontal surface equivalent reflectance .

[0108] When performing terrain correction on microwave remote sensing data (SAR), radiometric terrain correction (RTC) is used: the backscattering coefficient is adjusted. Normalization to a reference terrain (usually a horizontal surface) eliminates local variations in incident angle and geometric distortions caused by terrain. The formula is as follows:

[0109]

[0110] in: This represents the backscattering coefficient after terrain normalization;

[0111] This represents the local incident angle calculated from the DEM and SAR geometric models;

[0112] n represents the experience index.

[0113] The specific workflow for terrain correction of microwave remote sensing data (SAR) is as follows:

[0114] ①Based on the DEM and SAR geometric models, the local incident angle of each pixel is calculated using elevation data. The calculation process is existing technology and will not be described in detail in this invention;

[0115] ② Based on contemporaneous optical and radar imagery, the land surface is classified. For different land surface types, pure pixel regions of each category are delineated on the classification map, and their values ​​are extracted. and Perform its logarithmic field (usually taking log( ) and log(cos( In the linear regression analysis of ), the absolute value of the slope of the regression line is the n value for this type of land surface.

[0116] ③ Generate a spatial raster map of n values ​​based on the land surface classification map and the n values ​​of each type;

[0117] ④ Substitute each pixel into the above formula for calculation, and output the result after eliminating terrain effects. Microwave data.

[0118] Step 3: Based on the preprocessed optical and microwave remote sensing data, extract optical and microwave features, and fuse the optical and microwave features to obtain a full-coverage set of forest feature parameters;

[0119] Specifically, step 3 includes the following sub-steps:

[0120] Sub-step 301: Based on the cloudless areas obtained from cloud detection, extract the optical features of the cloudless areas using optical remote sensing data;

[0121] More specifically, the optical features in this invention include the Normalized Difference Vegetation Index (NDVI) and the Leaf Area Index (LAI).

[0122] The Normalized Difference Vegetation Index (NDVI) reflects vegetation growth and cover, and its calculation formula is as follows:

[0123]

[0124] In the formula, Indicates near-infrared reflectivity. This indicates the reflectivity in the red light band.

[0125] Enhanced Vegetation Index (EVI): By introducing the blue light band and soil / atmosphere moderating coefficients, it overcomes the problems of NDVI saturation and aerosol influence in the dense vegetation area of ​​the Nanling Mountains. The calculation formula is as follows:

[0126]

[0127] in, This is the gain factor, typically taken as 2.5. This is the atmospheric adjustment coefficient for red light, typically taken as 6.0. is the blue light atmospheric adjustment factor, typically taken as 7.5; L is the canopy background adjustment term, typically taken as 1. This refers to the reflectivity in the blue light band.

[0128] Leaf area index (LAI) extraction: The LAI was extracted using a lookup table method based on the canopy radiative transfer model. The specific steps are as follows:

[0129] ① Forward model construction: The PROSAIL canopy radiative transfer model was selected. The biochemical parameters of leaves (such as chlorophyll content and water content) and canopy structure parameters (such as leaf tilt angle distribution) of typical forests in the Nanling area were input. By changing the leaf area index (LAI) and observation geometry, the surface reflectance spectrum under different LAI was simulated in a forward manner.

[0130] ② Lookup table generation: Construct a multidimensional lookup table (LUT) by combining multiple sets of parameters generated by simulation (such as LAI, chlorophyll content, leaf tilt angle distribution, etc.) with their corresponding spectral reflectance.

[0131] ③ Cost function inversion: Substitute the actual observed remote sensing reflectance (red light, near-infrared, etc.) of the cloudless area with the simulated reflectance from the lookup table into the cost function:

[0132]

[0133] in To observe reflectivity, To simulate reflectance, i represents the band index (e.g., i=1 may represent the red light band, i=2 may represent the near-infrared band, and so on), and x represents the index of the parameter combination. x identifies a specific set of parameter combinations in the lookup table (e.g., different values ​​of parameters such as LAI, chlorophyll content, leaf tilt angle distribution, etc.).

[0134] ④ LAI determination: Traverse all parameter combinations in the lookup table and search for the top k smallest cost functions. For the minimum k corresponding to i and x, look up the multidimensional lookup table (LUT), select the LAI corresponding to i and x in the table, and take the median of these LAIs as the inversion result of the pixel to enhance the inversion stability under the complex terrain of Nanling.

[0135] Sub-step 302: Based on the cloud-affected and cloud-shadow-affected areas obtained from cloud detection, microwave features of the cloud-affected and cloud-shadow-affected areas are extracted using microwave remote sensing data;

[0136] More specifically, the microwave features in this invention include processing the backscattering coefficients through VV and VH polarization, and retrieving vegetation structure information and soil moisture information from the polarized backscattering coefficients. The specific extraction method is as follows:

[0137] (1) Extraction of basic polarization features:

[0138] Extracting terrain-corrected As fundamental characteristics, VV polarization is sensitive to canopy volume scattering and subsoil moisture response, while VH polarization is more significantly sensitive to trunk-ground secondary scattering and forest biomass structure response.

[0139] (2) Construction of polarization-derived features:

[0140] To address the issue of saturation of single polarization characteristics in the dense forests of the Nanling Mountains, polarization ratio characteristics were calculated. , This represents the backscattering coefficient after VH polarization normalization. This represents the backscattering coefficient after VV polarization normalization, characterized by the polarization ratio. Enhance the ability to characterize forest structure.

[0141] (3) Extracting spatial texture features based on the polarized backscattering coefficient:

[0142] To address the fragmented terrain and complex forest features of the Nanling Mountains, a gray-level co-occurrence matrix (GLCM) is used. A fixed-size window slides across the entire image, moving one pixel at a time. Based on the pixel values ​​within the window, the gray-level co-occurrence matrix, i.e., the counting matrix C, is calculated. , This matrix records specific spatial relationships (such as distance d and direction). Under grayscale and The number of times they co-occur; then, normalize this counting matrix to obtain the probability matrix P( , The normalization method involves dividing each element of the counting matrix by the sum of all elements in the counting matrix (i.e., the total number of valid pixel pairs within the window). In the normalized probability matrix P( , On the surface, various texture features (mean, contrast, homogeneity, entropy, etc.) are calculated.

[0143] Mean: Smoothed speckle noise, reflecting the average level of forest backscattering, as shown in the following formula:

[0144]

[0145] Contrast ratio: Reflects the local scattering heterogeneity caused by the micro-topography of the forest canopy and the gaps between canopies. The formula is as follows:

[0146]

[0147] Homogeneity: Characterized by the relatively uniform scattering characteristics of the canopy of typical evergreen broad-leaved forests in the Nanling Mountains, the formula is as follows:

[0148]

[0149] Entropy: Reflects the complexity and randomness of forest structure. The more complex the structure of a natural forest, the higher its entropy value. The formula is as follows:

[0150]

[0151] Sub-step 303: Perform weighted fusion of optical and microwave features to generate a full-coverage forest feature parameter set.

[0152] More specifically, the weighted fusion process includes the following sub-steps:

[0153] Sub-step 3031: Calculate the feature information entropy of optical and microwave features, and determine the adaptive weights:

[0154] To objectively reflect the information richness of different features within a local region and avoid biases caused by subjective weighting, the information entropy method is used to dynamically assign weights to each pixel. Let the optical feature set of the i-th pixel be... The microwave feature set is (All have been normalized to [0, 1]), calculate their local information entropy respectively. and The calculation formula is:

[0155] In the formula, Represents optical or microwave characteristics; M is the feature dimension; It represents the proportion probability of the j-th feature value in the i-th pixel.

[0156] Adaptive weights of optical features for the i-th pixel are calculated based on information entropy. Adaptive weights for microwave features The calculation formula is as follows:

[0157]

[0158]

[0159] Sub-step 3032: Based on adaptive weights, perform feature weighted fusion of optical and microwave features by region;

[0160] A cloud mask matrix (Mask) is introduced, where Mask=1 for cloudless areas and Mask=0 for areas affected by clouds or cloud shadows. Differentiated fusion strategies are employed for different areas.

[0161] In cloudless regions (Mask=1): both optical and microwave features are effective, and adaptive weighted fusion is used, as shown in the following formula:

[0162]

[0163] In areas affected by clouds or cloud shadows (Mask=0): optical features are missing and are directly replaced by microwave features.

[0164]

[0165] Sub-step 3033: Based on the feature weighted fusion result, perform feathering smoothing on the cloud boundary to obtain a full-coverage forest feature parameter set;

[0166] Due to the terrain of the Nanling Mountains, cloud shadows have fragmented edges. If the mask is directly and rigidly stitched together, the fused feature map will exhibit spectral / texture jumps (step effect) at the cloud boundaries. Therefore, distance feathering is applied to the cloud mask edges: a buffer distance d (2-3 pixels) is defined, and the normalized distance weights from the cloud boundary pixels to the cloudless areas are calculated. (Inside the cloudless area) =1, Deep cloud region =0, transition zone (linearly varying with distance). The final fusion formula is modified as follows:

[0167]

[0168] Step 4: Based on the forest characteristic parameter set, calculate the forest water conservation capacity using the physical model of water conservation capacity;

[0169] Specifically, step 4 includes the following sub-steps:

[0170] Sub-step 401: Obtain ground-measured data of typical forest plots in the Nanling Mountains;

[0171] More specifically, the ground-measured data of typical forest plots in the Nanling Mountains include ground-measured data such as canopy interception, litter water holding capacity, and soil porosity. The specific process for obtaining the ground-measured data is as follows:

[0172] ① Sample plot selection: Thirteen typical forest sample plots were selected in the Nanling area (located at altitudes of <300m, 300-600m, 600-900m, 900-1200m, and >1200m respectively); Plot size: 30m×30m (matched with satellite pixels); Cover type: Evergreen broad-leaved forest, mixed coniferous and broad-leaved forest, and other major forest types.

[0173] ② Field investigation:

[0174] 1) Canopy interception I: Calculated by installing rain gauges outside the forest, throughfall rain gauges inside the forest, and trunk runoff collection devices. The formula is as follows:

[0175] Canopy interception = Rainfall outside the forest - (Throughfall + Trunk runoff)

[0176] 2) Water holding capacity of litter Five 1m × 1m quadrats were set up in each plot; all litter in each quadrat was collected and weighed by fresh weight. ); Soak all the fallen leaves for 24 hours and weigh them at saturation ( ); Final calculation:

[0177]

[0178] 3) Soil porosity: First, undisturbed soil samples are collected using a standard volumetric sampler. After drying, the dry weight per unit volume of soil is calculated. Then, the soil bulk density (BD) is calculated using the following formula:

[0179]

[0180] In the formula, Indicates the dry soil mass (unit: g). Indicates the volume of the ring cutter (unit: cm). 3 );

[0181] Based on soil bulk density and soil density (taken as 2.65 g / cm³), calculate the total soil porosity = (1- (2.65) × 100%.

[0182] The above ground measurement data can be combined with GPS precise positioning (WGS84 coordinate system) and time series (multi-period observations during rainy / dry seasons) and stored in a database format for easy subsequent retrieval. The database format can specifically adopt the PostGIS spatial database.

[0183] Sub-step 402: Construct a physical model of water conservation capacity (WCC), calibrate the parameters in the model using ground-measured data, and then calculate the forest water conservation capacity at the pixel level for the entire region based on the calibrated physical model of water conservation capacity.

[0184] Specifically, the canopy interception I of the sample plots obtained from ground-measured data... , Substitute into the following calculation formula:

[0185]

[0186] In the formula, I represents the canopy interception amount (unit: mm); LAI represents the leaf area index (dimensionless); R represents the rainfall (unit: mm); a represents the canopy interception parameter (dimensionless); and k represents the canopy extinction coefficient (dimensionless). a and k can be calculated using the formula.

[0187] The litter water holding capacity L of the sample plot was obtained from ground-measured data. Substitute into the following calculation formula:

[0188]

[0189] In the formula, L is the litter water holding capacity (unit: mm); NDVI is the normalized vegetation index; DEM is the elevation (m); and b and c are parameters. b and c can be calculated using the formula.

[0190] The soil bulk density of the sample plot was obtained from the ground measurement data. Substitute into the following calculation formula:

[0191]

[0192] In the formula, S represents soil water storage (unit: mm); Clay% represents clay content (%); BD represents soil bulk density (g / cm³); Slope represents slope (°); and d represents soil structure parameter (mm·g / cm³). d can be calculated using the formula.

[0193] In summary, using ground-measured data, the parameters a, k, b, c, and d in the formulas for parameters I, L, and S were obtained through nonlinear regression fitting. Subsequently, parameters a, k, b, c, and d were substituted into their respective formulas to calculate the canopy interception I, litter water holding capacity L, and soil water storage S for pixels outside the sample plots. Finally, based on the calibrated water conservation capacity (WCC) physical model, the forest water conservation capacity at the pixel level for the entire region was calculated.

[0194] The physical model for water conservation capacity (WCC) is as follows:

[0195]

[0196] Step 5: Perform hierarchical rendering on the calculated forest water conservation capacity to generate a spatial distribution map of forest water conservation capacity in the Nanling area.

[0197] Specifically, the inverted forest water conservation capacity data is stored in GeoTIFF format using a mapping unit. Taking Python as an example, GDAL is used to read the GeoTIFF, and then matplotlib is used to draw a hierarchical rendering map, add map elements such as legends, and generate an image, which is then transformed into an intuitive spatial distribution map of forest water conservation capacity in the Nanling area. The transformed spatial distribution map of forest water conservation capacity in the Nanling area is then displayed using a display terminal.

[0198] In this invention, the mapping unit is a pixel, specifically the water conservation value of each 30m × 30m pixel obtained in step 4. Hierarchical rendering divides continuous values ​​into several levels according to a specific grading method, assigning a specific color to each level to form an intuitive gradation distribution map. The specific steps of hierarchical rendering are as follows:

[0199] (1) Determining the grading method and number of grades: Considering the uneven spatial distribution of water conservation capacity, the natural breakpoint method is adopted to find the optimal grading point based on the distribution characteristics of the data itself, so as to minimize the intra-class differences and maximize the inter-class differences. In this invention, the number of grades is 5: low, lower, medium, higher, and high.

[0200] (2) Set colors for each level: Use a gradient from cool to warm colors, with warm colors for low values ​​and cool colors for high values.

[0201] (3) Spatial distribution map generation: Rendering is performed using programming libraries (Python's matplotlib, GDAL).

[0202] Example 2: This example provides a remote sensing inversion system for forest water conservation capacity in the Nanling cloudy and rainy area, which is used to implement the remote sensing inversion method for forest water conservation capacity in the Nanling cloudy and rainy area described in Example 1.

[0203] Specifically, the system includes:

[0204] The data acquisition module acquires optical and microwave remote sensing data for the Nanling region.

[0205] In this invention, Sentinel-1 (SAR) and Sentinel-2 (optical) data of the Nanling region are automatically retrieved from the European Space Agency (ESA) Sentinel satellite data hub via an Internet interface.

[0206] The preprocessing and terrain correction module preprocesses the optical remote sensing data and microwave remote sensing data to obtain preprocessed optical remote sensing data and microwave remote sensing data.

[0207] In this invention, the built-in SRTM 30-meter resolution DEM data interface is used to run the GDAL library for geometric correction and terrain correction based on the Minnaert model.

[0208] The multi-source feature fusion module extracts optical and microwave features based on preprocessed optical and microwave remote sensing data, and fuses the optical and microwave features to obtain a full-coverage set of forest feature parameters.

[0209] This invention is developed using Python and integrates the OpenCV library for image processing, achieving pixel-level fusion of optical and radar data.

[0210] The inversion calculation module calculates the forest water conservation capacity based on the forest characteristic parameter set and using a physical model of water conservation capacity.

[0211] The results output and visualization module performs hierarchical rendering of the calculated forest water conservation capacity, generating a spatial distribution map of forest water conservation capacity in the Nanling region.

[0212] Example 3: This example provides an electronic device, including at least one processor; and a memory communicatively connected to the processor; wherein the memory stores instructions to be executed by the processor, the instructions being executed by the processor to enable the processor to execute the remote sensing inversion method for forest water conservation capacity applicable to the cloudy and rainy areas of the Nanling Mountains as described in Example 1.

[0213] A high-performance computing server (configured with dual Xeon Gold processors, 128GB of memory, and 4 NVIDIA A100 GPUs) is connected to a large-capacity disk array for storing massive amounts of remote sensing images and historical data; the system described in Example 2 is deployed on this server.

[0214] Real-world application case: Regional monitoring system based on high-performance servers

[0215] Based on the forest water conservation capacity remote sensing inversion system and electronic equipment of this invention, a comprehensive satellite remote sensing application platform for Shaoguan City, deployed at the Shaoguan Meteorological Bureau, was constructed. The platform serves as the carrier for the electronic equipment, and the corresponding user interface is shown in the attached diagram. Figure 2 As shown.

[0216] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A remote sensing inversion method for forest water conservation capacity applicable to the cloudy and rainy areas of the Nanling Mountains, characterized by: The method includes the following steps: Acquire optical and microwave remote sensing data for the Nanling region; Preprocessing is performed on optical remote sensing data and microwave remote sensing data to obtain preprocessed optical remote sensing data and microwave remote sensing data; Based on the preprocessed optical and microwave remote sensing data, optical and microwave features are extracted and fused to obtain a full-coverage set of forest feature parameters. Based on the forest characteristic parameter set, the forest water conservation capacity is calculated using a physical model of water conservation capacity. The calculated forest water conservation capacity is rendered in a hierarchical manner to generate a spatial distribution map of forest water conservation capacity in the Nanling region.

2. The remote sensing inversion method for forest water conservation capacity applicable to the cloudy and rainy areas of the Nanling Mountains according to claim 1, characterized in that: The preprocessing of optical and microwave remote sensing data to obtain preprocessed optical and microwave remote sensing data specifically includes the following steps: Radiometric calibration and geometric correction are performed on optical and microwave remote sensing data to eliminate sensor errors and geometric position deviations. Cloud detection is performed on radiometrically calibrated and geometrically corrected optical remote sensing data to identify cloud-affected, cloud-shadowed, and cloudless areas in the optical remote sensing data. Terrain correction was performed on microwave remote sensing data and optical remote sensing data after cloud detection to eliminate shadows and radiation distortions caused by the complex mountainous terrain of the Nanling Mountains, resulting in preprocessed optical and microwave remote sensing data.

3. The remote sensing inversion method for forest water conservation capacity applicable to the cloudy and rainy areas of the Nanling Mountains according to claim 2, characterized in that: The aforementioned cloud detection of radiometrically calibrated and geometrically corrected optical remote sensing data, identifying cloud-affected, cloud-shadowed, and cloudless areas in the optical remote sensing data, specifically includes the following steps: Based on multispectral feature analysis and thermal infrared feature analysis, areas affected by clouds and cloud shadows are extracted from optical remote sensing data; Based on the extracted areas affected by clouds and cloud shadows, a masking process is performed on these areas, replacing them with invalid values ​​while retaining the original data values ​​to obtain cloudless areas.

4. The remote sensing inversion method for forest water conservation capacity applicable to the cloudy and rainy areas of the Nanling Mountains according to claim 3, characterized in that: The aforementioned method, which extracts optical and microwave features from preprocessed optical and microwave remote sensing data and fuses these features to obtain a full-coverage set of forest feature parameters, specifically includes the following steps: Based on cloudless areas obtained from cloud detection, optical features of cloudless areas are extracted using optical remote sensing data. Based on the cloud-affected and cloud-shadow-affected areas obtained from cloud detection, microwave features of the cloud-affected and cloud-shadow-affected areas are extracted using microwave remote sensing data. The optical and microwave features are weighted and fused to generate a full-coverage set of forest feature parameters.

5. The remote sensing inversion method for forest water conservation capacity applicable to the cloudy and rainy areas of the Nanling Mountains according to claim 4, characterized in that: The weighted fusion of optical and microwave features to generate a full-coverage forest feature parameter set specifically includes the following steps: Calculate the feature information entropy of optical and microwave features, and determine adaptive weights; Based on adaptive weights, optical and microwave features are fused by region-specific weighted feature fusion. Based on the feature weighted fusion results, the cloud boundaries are feathered and smoothed to obtain a full-coverage set of forest feature parameters.

6. The remote sensing inversion method for forest water conservation capacity applicable to the cloudy and rainy areas of the Nanling Mountains according to claim 1, characterized in that: The calculation of forest water conservation capacity based on a set of forest characteristic parameters and a physical model of water conservation capacity includes the following steps: Obtain ground-measured data from typical forest plots in the Nanling Mountains; A physical model of water conservation capacity was constructed, and the parameters in the model were calibrated using sample plot measurement data. Then, based on the calibrated physical model of water conservation capacity, the forest water conservation capacity at the pixel level for the entire region was calculated.

7. A remote sensing inversion system for forest water conservation capacity in the Nanling cloudy and rainy region, used to implement the remote sensing inversion method for forest water conservation capacity in the Nanling cloudy and rainy region as described in any one of claims 1-6, characterized in that: The system includes: The data acquisition module acquires optical and microwave remote sensing data for the Nanling region. The preprocessing and terrain correction module preprocesses the optical remote sensing data and microwave remote sensing data to obtain preprocessed optical remote sensing data and microwave remote sensing data. The multi-source feature fusion module extracts optical and microwave features based on preprocessed optical and microwave remote sensing data, and fuses the optical and microwave features to obtain a full-coverage set of forest feature parameters. The inversion calculation module calculates the forest water conservation capacity based on the forest characteristic parameter set and using a physical model of water conservation capacity. The results output and visualization module performs hierarchical rendering of the calculated forest water conservation capacity, generating a spatial distribution map of forest water conservation capacity in the Nanling region.

8. An electronic device, characterized in that, The device includes at least one processor; and a memory communicatively connected to the processor; wherein the memory stores instructions that are executed by the processor to enable the processor to perform the remote sensing inversion method for forest water conservation capacity applicable to the cloudy and rainy areas of the Nanling Mountains as described in any one of claims 1-6.

Images