An automatic observation method for water holding capacity of forest litter

Abstract

The application discloses a kind of forest litter water-holding capacity automatic observation methods, comprising: identifying the topographic factor distribution of different slope direction slope altitude, and combining vegetation index calculation vegetation structure parameter, according to topographic factor distribution and vegetation structure parameter division forest environment heterogeneity area and generation spatial distribution layer, generation has the observation point layout scheme of regional representative, according to observation point layout scheme deployment sensor device, obtain the dynamic change time and change amplitude of litter water-holding capacity;The dynamic change time and change amplitude of litter water-holding capacity are decomposed, and the continuous spatial distribution data of forest litter water-holding capacity is generated, a water-holding capacity prediction model is constructed, and the litter water-holding capacity observation value of different regions of forest is calculated in real time according to the parameter coefficient of the prediction model.The application realizes the accurate observation of the litter water-holding characteristics of forest.

Description

Technical Field

[0001] This invention relates to the field of environmental monitoring technology, and in particular to an automatic method for observing the water holding capacity of forest litter. Background Technology

[0002] As a crucial component of the hydrological cycle, the litter layer in forest ecosystems directly impacts watershed water conservation, soil erosion control, and ecosystem stability. Accurate monitoring of changes in the water-holding capacity of forest litter is of great significance for forest hydrological research, ecological protection, and water resource management.

[0003] Traditional methods for monitoring the water holding capacity of litter mainly rely on regular manual sampling and laboratory analysis. This approach is not only time-consuming and labor-intensive, but also struggles to capture the dynamic changes in water holding capacity. The limited frequency of manual monitoring makes it impossible to reflect real-time changes under the influence of environmental factors such as rainfall and evaporation. Furthermore, the selection of sampling points often lacks systematicity and fails to represent the overall characteristics of complex forest environments.

[0004] The high heterogeneity of forest environments poses a fundamental challenge to water holding capacity observation. Forests with different slope aspects, gradients, and altitudes exhibit drastically different hydrological characteristics, further exacerbated by differences in forest age structure and tree species composition. This complex spatial heterogeneity directly leads to scientific challenges in the placement of observation points; single or random point placements cannot accurately reflect the true distribution pattern of litter water holding capacity within the region. Furthermore, the irrationality of observation point placement further restricts the effective deployment of automated monitoring systems, as the lack of a representative observation network makes it difficult to obtain regionally representative continuous observation data, even with advanced sensor technology.

[0005] Therefore, how to scientifically and rationally deploy observation points in complex and ever-changing forest environments, construct an automated observation network that can comprehensively reflect the water-holding characteristics of litter in different forest types, and achieve continuous and accurate collection of water-holding data has become a key issue in the development of automatic observation technology for forest litter water-holding. Summary of the Invention

[0006] To address the technical problems existing in the prior art, this invention proposes an automatic observation method for the water holding capacity of forest litter, enabling accurate observation of the water holding characteristics of forest litter.

[0007] To achieve the above objectives, the present invention provides an automatic method for monitoring the water holding capacity of forest litter, comprising:

[0008] Aerial scanning of the target forest land is performed to identify the distribution of topographic factors with different slope aspects, slopes, and altitudes. Vegetation structure parameters are calculated in combination with vegetation indices. Based on the distribution of topographic factors and vegetation structure parameters, forest land environmental heterogeneity regions are divided, and a spatial distribution layer is generated.

[0009] By using the area weights and hydrological characteristic differences of each heterogeneous region in the spatial distribution layer, a regionally representative observation point layout scheme is generated. Based on the observation point layout scheme, sensing devices are deployed to collect soil moisture data and soil weight data, respectively, to obtain the dynamic trend and magnitude of change in litter water holding capacity.

[0010] The dynamic changes in water holding capacity of the litter are decomposed into time and magnitude to obtain the water holding characteristics change patterns at different time scales. The water holding characteristics change patterns at different time scales are then extended to the entire forest area to generate continuous spatial distribution data of forest litter water holding capacity.

[0011] Based on the continuous spatial distribution data, a quantitative relationship between litter water holding capacity and meteorological data is obtained, a water holding capacity prediction model is constructed, and the observed values ​​of litter water holding capacity in different areas of the forest are calculated in real time based on the parameter coefficients of the prediction model.

[0012] Compared with the prior art, the present invention has the following advantages and technical effects:

[0013] This invention discloses an automatic observation method for forest litter water holding capacity. It utilizes multispectral scanning by unmanned aerial vehicles (UAVs) to acquire topographic and vegetation parameters, delineates heterogeneous forest regions, and deploys observation points. A network of soil moisture and weighing sensors is then deployed to achieve automated monitoring. Kalman filtering and time series analysis are employed to process the monitoring data, extracting water holding capacity variation characteristics, and spatial interpolation is used to generate continuous distribution data. A water holding capacity prediction model is established by combining meteorological data, enabling dynamic prediction and model optimization of litter water holding capacity. This invention achieves accurate observation and prediction of forest litter water holding characteristics, providing an effective tool for forest hydrological process research and water resource management, improving forest water regulation capabilities, and promoting the sustainable development of forest ecosystems. Attached Figure Description

[0014] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0015] Figure 1 This is a flowchart of an automatic monitoring method for water holding capacity of forest litter according to an embodiment of the present invention. Detailed Implementation

[0016] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0017] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0018] This embodiment proposes an automatic method for monitoring the water holding capacity of forest litter, such as... Figure 1 ,include:

[0019] Aerial scanning of the target forest land is performed to identify the distribution of topographic factors with different slope aspects, slopes, and altitudes. Vegetation structure parameters are calculated in combination with vegetation indices. Based on the distribution of topographic factors and vegetation structure parameters, forest land environmental heterogeneity regions are divided, and a spatial distribution layer is generated.

[0020] By using the area weights and hydrological characteristic differences of each heterogeneous region in the spatial distribution layer, a regionally representative observation point layout scheme is generated. Based on the observation point layout scheme, sensing devices are deployed to collect soil moisture data and soil weight data, respectively, to obtain the dynamic trend and magnitude of change in litter water holding capacity.

[0021] The dynamic changes in water holding capacity of the litter are decomposed into time and magnitude to obtain the water holding characteristics change patterns at different time scales. The water holding characteristics change patterns at different time scales are then extended to the entire forest area to generate continuous spatial distribution data of forest litter water holding capacity.

[0022] Based on the continuous spatial distribution data, a quantitative relationship between litter water holding capacity and meteorological data is obtained, a water holding capacity prediction model is constructed, and the observed values ​​of litter water holding capacity in different areas of the forest are calculated in real time based on the parameter coefficients of the prediction model.

[0023] Furthermore, aerial scanning of the target forest area was conducted to identify the distribution of topographic factors at different slope aspects, gradients, and elevations, and vegetation structure parameters were calculated in conjunction with vegetation indices, including:

[0024] A drone equipped with a multispectral sensor was used to take aerial photos of the target forest to acquire multispectral image data and generate a spectral reflectance dataset.

[0025] The vegetation index was calculated using the normalized difference vegetation index formula based on the spectral reflectance dataset, and the vegetation index distribution was obtained.

[0026] Based on the vegetation index distribution, the random forest algorithm is used to classify the age structure and tree species composition to obtain the vegetation structure parameters.

[0027] Specifically, drones equipped with multispectral sensors are used to conduct aerial photography of the forest, acquiring multispectral image data and generating a spectral reflectance dataset. Using this dataset, the Normalized Difference Vegetation Index (NDVI) formula NDVI = (NIR - RED) / (NIR + RED) is applied, where NIR is the near-infrared reflectance and RED is the red band reflectance, to calculate the vegetation index and obtain its distribution. If the NDVI value of a certain region in the vegetation index distribution is higher than a preset normalization threshold, a support vector machine (SVM) algorithm is used to classify the spectral reflectance data, determining the aspect, slope, and elevation distributions to obtain the topographic factor distribution. Based on the topographic factor distribution, a K-means clustering algorithm is applied to cluster the aspect, slope, and elevation data, dividing the environment into heterogeneous regions and obtaining the heterogeneous region distribution. Using the heterogeneous region distribution, combined with the vegetation index distribution, a random forest algorithm is used to classify the age structure and tree species composition, obtaining the vegetation structure parameters.

[0028] Furthermore, based on the distribution of topographic factors and vegetation structure parameters, heterogeneous regions of the forest environment are delineated, generating a spatial distribution layer, including:

[0029] If the age structure in the vegetation structure parameters matches the preset age threshold, then the slope, aspect and elevation are extracted from the topographic factor data to obtain the topographic factor distribution.

[0030] Based on the distribution of topographic factors, geographic information system technology is used to perform spatial overlay analysis of the distribution of topographic factors and vegetation structure parameters to generate a preliminary spatial distribution layer.

[0031] The initial spatial distribution layer is converted to a vector format using rasterization technology.

[0032] If there are regions in the vector format distribution layer with spatial resolution lower than a preset resolution threshold, then an interpolation algorithm is used to perform spatial interpolation on the low-resolution regions to obtain an optimized spatial distribution layer.

[0033] Based on the optimized spatial distribution layer, vector analysis techniques are used to extract the boundaries of heterogeneous environmental regions, generating the final spatial distribution layer.

[0034] Specifically, the initial spatial distribution layer is converted into a vector format using rasterization technology. Rasterization first divides the layer data into regular grid units, such as a 10-meter × 10-meter grid, and then converts it into a vector format based on the attribute values ​​of each grid to generate the final spatial distribution layer.

[0035] For example, the converted vector layer can clearly show the distribution of mature coniferous forests in a high-altitude area, covering 15% of the total forest land, providing spatial basis for precise forest resource management. Such vector layers facilitate subsequent analysis and visualization, improving the efficiency of data utilization.

[0036] The application of the above methods can significantly improve the accuracy and efficiency of forest monitoring. UAV multispectral imagery data provides a high-resolution data source for vegetation index calculation; support vector machines and K-means clustering effectively delineate heterogeneous regions based on topography and environment; random forest further refines vegetation structure classification; and geographic information systems and rasterization enable spatial visualization of the data. These technologies support each other, forming a complete scheme for forest ecological assessment and providing a scientific basis for forest protection, resource management, and ecological planning.

[0037] Furthermore, a regionally representative observation point layout scheme is generated, including:

[0038] By using the spatial distribution layer, the area weight and hydrological characteristic data of each heterogeneous region are obtained, and the area proportion and hydrological characteristic variation coefficient of each region are calculated to obtain the regional characteristic dataset.

[0039] Based on the aforementioned regional feature dataset, a stratified sampling algorithm is used to calculate the number of observation points in each heterogeneous region, taking into account area weight and coefficient of variation, to obtain the distribution of the number of observation points. Then, a grid partitioning technique is applied to divide each heterogeneous region into a uniform grid. Based on the point density, the spatial coordinates of the observation points are generated within the grid to obtain the point coordinate dataset.

[0040] Based on the data set of point coordinates and the coefficient of variation of the hydrological features, the point coordinates are classified, the representative weight of each point is determined, and the representative distribution of the points is obtained.

[0041] By overlaying the point coordinates with the spatial distribution of heterogeneous regions using the representative distribution of the points, a spatial distribution layer of the observation points is generated, thus generating the observation point layout scheme.

[0042] Specifically, by using a spatial distribution layer, the area weights and hydrological characteristic data of each heterogeneous region are obtained. The area proportion and coefficient of variation of the hydrological characteristics of each region are calculated to obtain a regional characteristic dataset. Based on the regional characteristic dataset, a stratified sampling algorithm is used to calculate the number of observation points in each heterogeneous region for the area weights and coefficients of variation, resulting in a point quantity distribution. If the area weight of a certain region in the point quantity distribution is greater than a preset threshold and the coefficient of variation is higher than a preset value, the density adjustment formula D = k(AV), where D is the point density, A is the area weight, V is the coefficient of variation, and k is a constant coefficient, is used to calculate the increased point density, resulting in an adjusted point density distribution. Based on the adjusted point density distribution, a grid partitioning technique is applied to divide each heterogeneous region into a uniform grid. Spatial coordinates of observation points are generated within the grid based on the point density, resulting in a point coordinate dataset. Based on the point coordinate dataset and the coefficient of variation of the hydrological characteristics, a random forest algorithm is used to classify the point coordinates, determine the representative weight of each point, and obtain a representative point distribution. By using representative point distribution and applying geographic information system technology, the point coordinates are overlaid with the spatial distribution of heterogeneous areas to generate a spatial distribution layer of observation points, thus obtaining the final observation point layout scheme.

[0043] In one embodiment, based on the point coordinate dataset and the coefficient of variation of hydrological features, a random forest algorithm is used to classify the points and determine their representative weights. For example, points in high-altitude steep slope areas have a higher weight value (0.7) due to the large variation in hydrological features; points in low-altitude gentle slope areas have a weight value of 0.5. This classification method highlights the monitoring priority of key areas. The random forest analyzes the relationship between point coordinates and hydrological features through multiple decision trees to ensure the reliability of the classification results.

[0044] Geographic Information System (GIS) technology is used to overlay point coordinates onto heterogeneous areas, generating a spatial distribution layer of observation points. It is assumed that 12 points in the high-altitude steep slope area are concentrated in areas with slopes greater than 30 degrees, while 20 points in the low-altitude gentle slope area are distributed across flat terrain. The final layer visually demonstrates the relationship between the points and topographic and hydrological features, facilitating subsequent monitoring planning. This layout scheme optimizes the spatial representativeness of observation points and improves monitoring efficiency.

[0045] Furthermore, sensing devices are deployed according to the observation point layout plan to collect soil moisture data and soil weight data, including:

[0046] Topographic factors and vegetation structure parameters are extracted from the observation point layout scheme, and the points are classified using the K-means clustering algorithm to obtain the classified observation point set.

[0047] Based on the classified set of observation points, combined with topographic factors and vegetation structure parameters, the weighted average method is used to calculate the data collection priority of each point and determine the differentiated collection frequency scheme.

[0048] The sensor network is configured using the LoRa wireless communication protocol. The sampling frequency parameters of each point are obtained from the differentiated sampling frequency scheme. The sampling frequencies of the soil moisture sensor and the weighing sensor are set to obtain the configured sensor network.

[0049] Soil moisture and soil weight data are collected from the configured sensor network. If the data integrity is lower than the preset integrity threshold, the missing data is supplemented by linear interpolation to obtain a complete dataset.

[0050] Specifically, topographic factor data is acquired through remote sensing imagery and digital elevation models, and vegetation structure parameters are extracted using the vegetation cover index to generate an observation point distribution map. Topographic factors and vegetation structure parameters are extracted from the distribution map, and the points are classified using the K-means clustering algorithm to obtain a set of classified observation points. Based on the classified observation point set, and combining the topographic factors and vegetation structure parameters, a weighted average method is used to calculate the data acquisition priority for each point, determining a differentiated acquisition frequency scheme. A sensor network is configured using the LoRa wireless communication protocol, and the acquisition frequency parameters for each point are obtained from the differentiated acquisition frequency scheme. The acquisition frequencies of the soil moisture sensor and the weighing sensor are set to obtain the configured sensor network. Soil moisture and soil weight data are collected from the configured sensor network. If the data integrity is lower than a preset integrity threshold, missing data is supplemented using linear interpolation to obtain a complete dataset.

[0051] Furthermore, the dynamic changes in the water holding capacity of litter were obtained, including the timing and magnitude of these changes:

[0052] The soil moisture data and soil weight data are smoothed in real time by using the Kalman filter algorithm, and the multi-sensor data are fused by the state estimation method to obtain smoothed time series data.

[0053] The smoothed time series data is processed using the sliding window method, and the data gradient within each time window is calculated to obtain the dynamic change sequence of the data gradient.

[0054] If a gradient value in the dynamic change sequence of the data gradient exceeds a preset gradient threshold, it is determined to be a rainfall event or an evaporation event, and the time and type of the triggering event are determined.

[0055] Based on the time and type of the triggering event, soil moisture and weight data for the corresponding time period are extracted from the time series data, and the change range of litter water holding capacity is calculated.

[0056] The dynamic trend of water holding capacity of the litter was obtained by extracting the trend of the change range of the litter water holding capacity using time series analysis.

[0057] Specifically, in one possible implementation, the Kalman filter algorithm, when used for real-time smoothing of soil moisture and soil weight data, can be viewed as a state estimation tool. Through prediction and updating steps, it combines sensor observations with a system model to reduce noise interference. For example, suppose a forest soil moisture sensor collects data every 10 minutes. The raw data may fluctuate significantly due to environmental noise, such as humidity values ​​jumping irregularly between 50% and 55%. The Kalman filter predicts the current state using historical data and then corrects it with new observations, generating a smoothed humidity value sequence, such as stabilizing at around 52%. This smoothing process more accurately reflects the true changes in soil moisture, facilitating subsequent analysis.

[0058] Specifically, when fusing multi-sensor data, Kalman filtering can integrate the outputs of soil moisture sensors and weighing sensors. For example, a soil moisture sensor might measure a humidity of 53%, while a weighing sensor might record weight changes suggesting potential rainfall. Through state estimation, the algorithm fuses these two types of data to generate consistent time-series data, such as hourly sequences of soil moisture content and litter weight changes. This fusion comprehensively reflects forestland moisture dynamics and improves data reliability.

[0059] Furthermore, the dynamic changes in water holding capacity of litter were decomposed in terms of timing and magnitude to obtain the changing patterns of water holding characteristics at different time scales, including:

[0060] The water holding capacity data is processed by a time series decomposition algorithm to extract trend components, seasonal components and random components, and obtain the time series characteristics of each component.

[0061] The trend components are analyzed using the autocorrelation function to calculate the periodic parameters and determine the long-term variation law of the trend components.

[0062] For the seasonal component, Fourier transform is applied to extract periodic frequency features to obtain the periodic pattern of the seasonal component.

[0063] If the autocorrelation function value of the random component is lower than the preset autocorrelation function threshold, the random component is smoothed by the moving average method to obtain the smoothed random component sequence.

[0064] Based on the smoothed random component sequence, combined with the periodic patterns of the trend component and the seasonal component, the water-holding characteristic change patterns at different time scales are obtained.

[0065] Specifically, a seasonal decomposition algorithm is used to process the water holding capacity time series data to extract trend, seasonal, and random components, resulting in sequence sequences for each component. For the extracted trend component, an autoregressive model is applied to analyze its time series characteristics and determine its long-term variation pattern. Based on the seasonal component, the periodic frequency is calculated using a fast Fourier transform to obtain its periodicity. If the autocorrelation function value of the random component is below a preset threshold, exponential smoothing is used to process the random component, resulting in a smoothed random component sequence; if it is above the preset threshold, the original random component sequence is retained. By combining the long-term variation pattern of the trend component, the periodicity of the seasonal component, and the smoothed random component sequence, the water holding capacity variation patterns at different time scales are obtained. Principal component analysis is used to reduce the dimensionality of the water holding capacity variation patterns, extracting the main feature vectors and identifying the key driving factors for water holding capacity changes.

[0066] Furthermore, continuous spatial distribution data of forest litter water holding capacity are generated, including:

[0067] The spatial interpolation algorithm is used to extend the water-holding characteristic change pattern of each observation point to the entire forest area. The water-holding capacity of unobserved locations is estimated by using the inverse distance weighting method. The interpolation weight coefficient is adjusted according to the similarity of topographic factors and vegetation structure to generate continuous spatial distribution data of forest litter water-holding capacity.

[0068] Specifically, when acquiring data on litter water holding capacity, topographic factors, and vegetation structure at various observation points within a forest area, an initial dataset can be constructed using a combination of field sampling and remote sensing techniques. Assuming 100 observation points are selected within a forest area, each point records data on litter water holding capacity, topographic slope, aspect, vegetation cover, and tree species distribution. Water holding capacity data is determined using a weighing method, topographic factors are extracted using a digital elevation model, and vegetation structure data is based on analysis of UAV aerial imagery. The initial dataset must ensure a uniform distribution of observation points, covering different terrain and vegetation types within the forest area to guarantee data representativeness.

[0069] In one possible implementation, when estimating the water holding capacity at unobserved locations using an inverse distance weighting algorithm, spatial interpolation can be performed based on the water holding capacity data of the observed points. For example, an unobserved point is 50 meters, 100 meters, and 150 meters away from three observed points, with water holding capacities of 200 mm, 180 mm, and 160 mm, respectively. Using a distance-inverse weighted calculation, the initial estimate of the water holding capacity at this point is approximately 185 mm. This method uses the inverse of spatial distance as a weight; points that are closer contribute more to the estimate, making it suitable for scenarios with relatively uniform spatial distribution.

[0070] Specifically, when calculating the terrain similarity coefficient, factors such as slope, aspect, and altitude can be used to quantify the terrain differences between unobserved and observed points using a weighted Euclidean distance formula. For example, if an unobserved point has a slope of 5 degrees and an aspect of north, while an observed point has a slope of 6 degrees and an aspect of northeast, their terrain similarity coefficient can be calculated using standardized factor differences, yielding a value such as 0.85. Similarly, the vegetation similarity coefficient can be calculated based on the degree of matching between coverage and tree species. For example, if an unobserved point has a coverage of 70% and its main tree species is pine, while an observed point has a coverage of 65% and the tree species are the same, the similarity coefficient can be set at 0.90. These coefficients are used to adjust the weights of the inverse distance weighting algorithm, improving the estimation accuracy.

[0071] Furthermore, a water holding capacity prediction model is constructed, including:

[0072] Continuous spatial distribution data of water holding capacity of litter within the forest area and meteorological data are obtained to construct an initial dataset, wherein the meteorological data includes rainfall, evaporation, temperature and humidity meteorological data;

[0073] Using the initial dataset, the correlation between meteorological data and water holding capacity was analyzed by stepwise regression, and significant influencing factors were screened out to obtain a set of significant influencing factors.

[0074] Multiple regression analysis was used to establish a quantitative relationship between water holding capacity and meteorological factors for a set of significant influencing factors, and to determine the parameter coefficients. Using the parameter coefficients and continuous spatial distribution data, the predicted water holding capacity values ​​at each location were calculated to obtain the predicted value distribution. If the deviation between the predicted value distribution and the actual observed value exceeded the prediction threshold, the parameter coefficients were adjusted using the cross-validation method to obtain the optimized parameter coefficients.

[0075] Using the optimized parameter coefficients, the multiple regression analysis is run again to update the distribution of water holding capacity prediction values, resulting in an adjusted prediction distribution. Based on the adjusted prediction distribution and combined with the spatiotemporal variation characteristics of meteorological data, the water holding capacity prediction model is generated.

[0076] The construction of water holding capacity prediction models also includes:

[0077] We obtained measured data on the water holding capacity of litter in various areas of the forest and collected environmental variables using sensors to obtain an initial dataset.

[0078] Using the initial dataset and the parameters of the water holding capacity prediction model, the random forest algorithm is applied to calculate the predicted water holding capacity of litter in each region.

[0079] For the deviation between the predicted value and the measured data, the prediction error is calculated, and the root mean square error formula is used to obtain the error quantification result. If the error quantification result exceeds the preset quantification threshold, the error distribution characteristics are extracted, and the gradient descent method is used to optimize the parameters of the water holding capacity prediction model to obtain the updated model parameters.

[0080] Based on the updated model parameters, the predicted values ​​of water holding capacity of litter in each region are recalculated to obtain a new prediction dataset.

[0081] The convergence of the model is determined by comparing the new prediction dataset with the measured data. If the error still exceeds the preset threshold, the optimization steps are repeated until the final optimized model parameters are obtained.

[0082] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An automatic method for observing the water holding capacity of forest litter, characterized in that, include: Aerial scanning of the target forest land is performed to identify the distribution of topographic factors with different slope aspects, slope gradients, and altitudes. Vegetation structure parameters are calculated in combination with vegetation indices. Based on the distribution of topographic factors and vegetation structure parameters, forest land environmental heterogeneity regions are divided, and a spatial distribution layer is generated. By using the area weights of each heterogeneous region in the spatial distribution layer and the coefficient of variation of hydrological characteristics, a regionally representative observation point layout scheme is generated. Based on the observation point layout scheme, sensing devices are deployed to collect soil moisture data and soil weight data respectively, thereby obtaining the dynamic trend and magnitude of change in litter water holding capacity. The dynamic trend and magnitude of the water holding capacity of the litter are decomposed to obtain the water holding characteristic change pattern at different time scales, and the water holding characteristic change pattern at different time scales is extended to the entire forest area to generate continuous spatial distribution data of forest litter water holding capacity. Based on the continuous spatial distribution data, a quantitative relationship between litter water holding capacity and meteorological data is obtained, a water holding capacity prediction model is constructed, and the observed values ​​of litter water holding capacity in different areas of forest land are calculated in real time based on the parameter coefficients of the prediction model. Aerial scanning of the target forest area was conducted to identify the distribution of topographic factors at different slope aspects, gradients, and elevations. Vegetation structure parameters were then calculated using vegetation indices, including: A drone equipped with a multispectral sensor was used to take aerial photos of the target forest to acquire multispectral image data and generate a spectral reflectance dataset. The vegetation index was calculated using the normalized difference vegetation index formula based on the spectral reflectance dataset, and the vegetation index distribution was obtained. Based on the vegetation index distribution, the random forest algorithm is used to classify the age structure and tree species composition to obtain the vegetation structure parameters; The dynamic trend and magnitude of the water-holding capacity of the litter are decomposed to obtain the water-holding characteristic change patterns at different time scales, including: The water holding capacity data is processed by a time series decomposition algorithm to extract trend components, seasonal components and random components, and obtain the time series characteristics of each component. The autocorrelation function is used to analyze the trend components, calculate the periodic parameters, and determine the long-term variation law of the trend components. For the seasonal component, Fourier transform is applied to extract periodic frequency features to obtain the periodic pattern of the seasonal component. If the autocorrelation function value of the random component is lower than the preset autocorrelation function threshold, the random component is smoothed by the moving average method to obtain the smoothed random component sequence. Based on the smoothed random component sequence, combined with the periodic patterns of the trend component and the seasonal component, the water-holding characteristic change patterns at different time scales are obtained.

2. The automatic monitoring method for water holding capacity of forest litter as described in claim 1, characterized in that, Based on the distribution of topographic factors and vegetation structure parameters, forest environment heterogeneity regions are delineated, and a spatial distribution layer is generated, including: If the age structure in the vegetation structure parameters matches the preset age threshold, then the slope, aspect and elevation are extracted from the topographic factor data to obtain the topographic factor distribution. Based on the distribution of topographic factors, geographic information system technology is used to perform spatial overlay analysis of the distribution of topographic factors and vegetation structure parameters to generate a preliminary spatial distribution layer. The initial spatial distribution layer is converted to a vector format using rasterization technology. If there are regions in the vector format distribution layer with spatial resolution lower than a preset resolution threshold, then an interpolation algorithm is used to perform spatial interpolation on the low-resolution regions to obtain an optimized spatial distribution layer. Based on the optimized spatial distribution layer, vector analysis techniques are used to extract the boundaries of heterogeneous environmental regions, generating the final spatial distribution layer.

3. The automatic monitoring method for water holding capacity of forest litter as described in claim 1, characterized in that, Generate a regionally representative observation point layout scheme, including: By using the spatial distribution layer, the area weight and hydrological characteristic data of each heterogeneous region are obtained, and the area proportion and hydrological characteristic variation coefficient of each heterogeneous region are calculated to obtain the regional characteristic dataset. Based on the aforementioned regional feature dataset, a stratified sampling algorithm is used to calculate the number of observation points in each heterogeneous region, taking into account area weight and coefficient of variation, to obtain the distribution of the number of observation points. Then, a grid partitioning technique is applied to divide each heterogeneous region into a uniform grid. Based on the point density, the spatial coordinates of the observation points are generated within the grid to obtain the point coordinate dataset. Based on the data set of point coordinates and the coefficient of variation of the hydrological features, the point coordinates are classified, the representative weight of each point is determined, and the representative distribution of the points is obtained. By overlaying the point coordinates with the spatial distribution of heterogeneous regions using the representative distribution of the points, a spatial distribution layer of the observation points is generated, thus generating the observation point layout scheme.

4. The automatic monitoring method for water holding capacity of forest litter as described in claim 3, characterized in that, According to the observation point layout plan, sensor devices are deployed to collect soil moisture data and soil weight data, including: Topographic factors and vegetation structure parameters are extracted from the observation point layout scheme, and the points are classified using the K-means clustering algorithm to obtain a set of classified observation points. Based on the classified set of observation points, combined with topographic factors and vegetation structure parameters, the weighted average method is used to calculate the data collection priority of each point and determine the differentiated collection frequency scheme. The sensor network is configured using the LoRa wireless communication protocol. The sampling frequency parameters of each point are obtained from the differentiated sampling frequency scheme. The sampling frequencies of the soil moisture sensor and the weighing sensor are set to obtain the configured sensor network. Soil moisture and soil weight data are collected from the configured sensor network. If the data integrity is lower than the preset integrity threshold, the missing data is supplemented by linear interpolation to obtain a complete dataset.

5. The automatic monitoring method for water holding capacity of forest litter according to claim 4, characterized in that, Obtaining the dynamic trend and magnitude of the water holding capacity of the litter includes: The soil moisture data and soil weight data are smoothed in real time by using the Kalman filter algorithm, and the multi-sensor data are fused by the state estimation method to obtain smoothed time series data. The smoothed time series data is processed using the sliding window method, and the data gradient within each time window is calculated to obtain the dynamic change sequence of the data gradient. If a gradient value in the dynamic change sequence of the data gradient exceeds a preset gradient threshold, it is determined to be a rainfall event or an evaporation event, and the time and type of the triggering event are determined. Based on the time and type of the triggering event, soil moisture and weight data for the corresponding time period are extracted from the time series data, and the change range of litter water holding capacity is calculated. The dynamic trend of water holding capacity of litter was obtained by extracting the trend of the change range of the litter water holding capacity using time series analysis.

6. The automatic monitoring method for water holding capacity of forest litter according to claim 1, characterized in that, Generate continuous spatial distribution data of forest litter water holding capacity, including: Spatial interpolation algorithms are used to extend the water-holding characteristics change patterns of each observation point to the entire forest area. The water-holding capacity of unobserved locations is estimated by using the inverse distance weighting method. The interpolation weight coefficients are adjusted according to topographic factors and vegetation structure similarity to generate continuous spatial distribution data of forest litter water-holding capacity.

7. The automatic monitoring method for water holding capacity of forest litter according to claim 1, characterized in that, Constructing the water holding capacity prediction model includes: Continuous spatial distribution data of water holding capacity of litter within the forest area and meteorological data are obtained to construct an initial dataset, wherein the meteorological data includes rainfall, evaporation, temperature and humidity meteorological data; Using the initial dataset, the correlation between meteorological data and water holding capacity was analyzed by stepwise regression, and significant influencing factors were screened out to obtain a set of significant influencing factors. Multiple regression analysis was used to establish a quantitative relationship between water holding capacity and meteorological factors for a set of significant influencing factors, and to determine the parameter coefficients. Using the parameter coefficients and continuous spatial distribution data, the predicted water holding capacity values ​​at each location were calculated to obtain the predicted value distribution. If the deviation between the predicted value distribution and the actual observed value exceeded the prediction threshold, the parameter coefficients were adjusted using the cross-validation method to obtain the optimized parameter coefficients. Using the optimized parameter coefficients, the multiple regression analysis is run again to update the distribution of water holding capacity prediction values, resulting in an adjusted prediction distribution. Based on the adjusted prediction distribution and combined with the spatiotemporal variation characteristics of meteorological data, the water holding capacity prediction model is generated.

8. The automatic monitoring method for water holding capacity of forest litter according to claim 7, characterized in that, The construction of the water holding capacity prediction model also includes: We obtained measured data on the water holding capacity of litter in various areas of the forest and collected environmental variables using sensors to obtain an initial dataset. Using the initial dataset and the parameters of the water holding capacity prediction model, the random forest algorithm is applied to calculate the predicted water holding capacity of litter in each region. For the deviation between the predicted value and the measured data, the prediction error is calculated, and the root mean square error formula is used to obtain the error quantification result. If the error quantification result exceeds the preset quantification threshold, the error distribution characteristics are extracted, and the gradient descent method is used to optimize the parameters of the water holding capacity prediction model to obtain the updated model parameters. Based on the updated model parameters, the predicted values ​​of water holding capacity of litter in each region are recalculated to obtain a new prediction dataset. The convergence of the model is determined by comparing the new prediction dataset with the measured data. If the error still exceeds the preset threshold, the optimization steps are repeated until the final optimized model parameters are obtained.

Images