Water and soil loss management and control early warning system and method based on Internet of Things

Abstract

The invention discloses a water and soil loss management and control early warning system and method based on the Internet of Things, and relates to the technical field of water and soil conservation and ecological environment monitoring. Core parameters such as interface double-electric-layer relaxation time sequential change rate, interface tension attenuation amplitude and capacitance distortion coefficient are calculated, microscopic signals of soil colloid binding water stripping and interface instability are captured, and the signals are obviously abnormal 10-15 days before water and soil loss macroerosion occurs; meanwhile, through soil body crack nucleation infrasound-acoustic emission coupling monitoring, soil body microcrack nucleation and expanded infrasound and acoustic emission signals are captured, the signals appear 7-10 days earlier than slope macroscopic deformation, through two-dimensional precursor signal coupling recognition, early-stage pre-judgment of water and soil loss is achieved, and the early-stage early-stage early-stage early-stage early-stage early-stage early-stage early-stage early-stage early-stage early-stage early-stage early-stage early-stage early-stage early-stage early-stage early-stage early-stage early-stage early- the ecological damage degree and the treatment cost caused by water and soil loss are reduced.

Description

Technical Field

[0001] This invention relates to the field of soil and water conservation and ecological environment monitoring technology, and in particular to an Internet of Things-based early warning system and method for soil erosion control. Background Technology

[0003] Currently, after decades of development, my country's soil and water conservation monitoring and early warning technology has formed three major mainstream technical systems: ground-based in-situ monitoring, remote sensing macroscopic monitoring, and numerical model early warning. The general implementation plan involves deploying conventional monitoring equipment or using remote sensing in the target control area to collect macroscopic indicators such as rainfall intensity, soil moisture content, slope macroscopic deformation, and vegetation cover at fixed intervals. The collected data is then transmitted to a backend platform, where soil and water conservation risk is assessed using empirical thresholds or statistical models, and early warning information is pushed out when thresholds are exceeded. However, the following problems exist:

[0004] Firstly, most monitoring indicators focus on macroscopic parameters such as soil moisture content, slope macroscopic deformation, surface runoff, and gully morphology. These parameters only show identifiable abrupt changes when soil erosion has already occurred or even caused significant surface damage. They cannot capture the precursory signals of soil microstructural instability before soil erosion occurs. The early warning lead time of industry-standard solutions is generally less than 24 hours, which can only provide alarm notification after the event occurs and cannot leave sufficient time for pre-disaster prevention and control. At the same time, empirical statistical models that rely heavily on fitting historical data cannot reconstruct the entire internal evolution process of soil erosion from microscopic interface instability and pore structure destruction to macroscopic erosion. The accuracy rate of predicting the time, scope, and intensity of erosion is less than 60%, which cannot provide accurate quantitative support for advanced prevention and control.

[0005] Secondly, many of my country's key areas for soil erosion control are remote mountainous regions without stable power supply or wired network coverage. Existing technologies generally employ a fixed-cycle data collection mode. To ensure monitoring accuracy and capture key erosion signals, a high-frequency data collection mode is required, resulting in high equipment power consumption, short battery life, and frequent on-site battery replacements, leading to extremely high maintenance costs and making long-term continuous and stable operation impossible. On the other hand, extending the data collection cycle to reduce power consumption results in the loss of weak precursor signals before soil erosion occurs, leading to severely insufficient monitoring accuracy. Consequently, the inherent contradiction between high-precision monitoring and low-power long-term operation remains unresolved, making it difficult to achieve large-scale and routine implementation in remote core areas of soil erosion. Therefore, we propose an IoT-based soil erosion control and early warning system and method. Summary of the Invention

[0006] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides an Internet of Things-based early warning system and method for soil and water conservation, which solves the technical problems of delayed early warning, high false alarm rate, incompatibility between monitoring accuracy and low power consumption, low accuracy of risk projection, and disconnect between control and management links.

[0007] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: A method for early warning of soil erosion control based on the Internet of Things, the method includes the following steps: S1. Based on a distributed IoT passive sensing architecture, acquire multi-dimensional monitoring source data covering soil interface electrochemistry, soil microstructure deformation, pore hydrological fractal, surface erosion resistance ecology, and external force-driven data in the target control area. S2. Extract the early warning signals of interface instability and soil fissure nucleation from multi-dimensional monitoring source data, and generate the chaotic phase space adaptive monitoring cycle control command based on the coupling of the two early warning signals. S3. Based on the chaotic phase space adaptive monitoring cycle control command, dynamically reconstruct the acquisition parameters of multi-dimensional monitoring source data to obtain a full-chain continuous monitoring sample set with multiple batches of spatiotemporal registration. S4. Input the full-chain continuous monitoring sample set with multiple batches of spatiotemporal registration into the preset fourth-order multi-scale coupled deduction model of soil and water loss, complete the cross-scale coupled analysis from micro-interface instability to macro-erosion evolution, and obtain the advanced soil and water loss risk assessment value in the whole time domain. S5. Based on the advanced water and soil erosion risk assessment value across the entire time domain, and combined with the control level of the target control area, generate an early warning judgment value for water and soil erosion control. Based on the comparison result between the early warning judgment value and the preset threshold, implement closed-loop control intervention or routine monitoring at the corresponding level.

[0008] Preferably, in step S1, the multi-dimensional monitoring source data includes soil mineral-microorganism interface relaxation electrochemical characteristic data, soil fissure nucleation infrasound-acoustic emission coupling characteristic data, unsaturated soil capillary water hysteresis ring fractal characteristic data, biological crust hyperspectral fluorescence quantum yield characteristic data, and rainfall-wind erosion coupling driving characteristic data. Specifically, step S1 includes: S101. Obtain standardized access points for multiple types of passive IoT sensing nodes within the target control area. The sensing nodes correspond one-to-one with the monitoring dimensions of soil interface electrochemistry, soil microstructure deformation, pore hydrological fractal, surface erosion resistance ecology, and external force-driven monitoring. S102. Simultaneously collect multiple sets of raw passive monitoring data through each standardized access point; S103. Extract the unique features of the physical field corresponding to each group of original passive monitoring data, and classify and calibrate the monitoring data according to the feature dimensions to obtain the multi-dimensional monitoring source data.

[0009] Preferably, step S2 specifically includes: S201. Based on the relaxation electrochemical characteristic data of the soil mineral-microorganism interface, calculate the temporal change rate of the interface double electric layer relaxation time, the amplitude of the interface tension decay and the capacitance distortion coefficient. When any parameter exceeds the corresponding critical threshold, it is determined that a precursor signal of interface instability in the early stage has been captured. The early stage is 10-15 days before the occurrence of macro-erosion of soil and water loss, which can be dynamically adjusted according to the soil type and climate characteristics of the target control area. S202. Based on the infrasound-acoustic emission coupling characteristic data of soil fissure nucleation, calculate the frequency of soil micro-fracture events, energy release rate and infrasound spectrum shift. When any parameter exceeds the corresponding critical threshold, it is determined that a pre-nucleation signal of fissure nucleation has been captured. The pre-nucleation signal of fissure nucleation is 7-10 days earlier than the macroscopic slope deformation of the soil and can be dynamically adjusted according to the geological and geomorphological characteristics of the target control area. S203. Combining the spatial distribution range and intensity level of the dual precursor signals, an adaptive monitoring cycle control command for the corresponding control zone is generated through a chaotic phase space reconstruction algorithm. The control command is used to synchronously adjust the acquisition frequency, sampling bandwidth and spatial resolution of monitoring data in each dimension.

[0010] Preferably, step S3 specifically includes: S301. According to the adaptive monitoring cycle control instruction, generate a distributed passive acquisition task corresponding to the monitoring data of each dimension, and determine the acquisition configuration parameters of the corresponding sensing node. The acquisition configuration parameters include acquisition time window, acquisition frequency, sampling bandwidth, spatial sampling density, and trigger threshold. S302. Based on the acquisition configuration parameters, complete synchronous and continuous acquisition to obtain monitoring information in various dimensions; S303. All monitoring information of all dimensions is spatiotemporally registered with unified timestamps and spatial geographic coordinates. After data preprocessing such as outlier removal, baseline correction and trend term extraction, it is integrated to form multiple batches of full-chain continuous monitoring sample sets. The sample sets cover the entire process of data from changes in soil microstructure to macro erosion evolution.

[0011] Preferably, step S4 specifically includes: S401. Input the monitoring sample set into the interface instability layer of the fourth-order multi-scale coupled deduction model, calculate the soil particle bound water stripping coefficient and colloidal dispersion degree, and obtain the interface instability potential value of soil erosion. S402. Input the monitoring sample set into the pore failure layer of the model, calculate the rate of change of the fractal dimension of the soil pore structure and the pore connectivity, and combine the interface instability potential value to obtain the pore failure risk value. S403. Input the monitoring sample set into the fracture nucleation layer of the model, calculate the soil micro-fracture propagation rate, connectivity probability and nucleation density, and obtain the fracture nucleation trigger value by combining the pore failure risk value. S404. Input the monitoring sample set into the erosion evolution layer of the model, calculate the slope erosion resistance attenuation coefficient and the external driving force intensity, and combine the output results of the above three layers to predict the occurrence range, erosion modulus, transport path and degree of harm of soil erosion in the next 168 hours, and obtain the full-time risk evolution sequence. The prediction time can be dynamically adjusted according to the control needs. S405. The output results of the above four layers are nonlinearly weighted and fused to obtain the advanced soil and water loss risk assessment value in the whole time domain.

[0012] Preferably, step S5 specifically includes: S501. Obtain the ecological protection red line level and soil and water conservation control level of the target control area, and determine the risk weight correction coefficient and early warning threshold adjustment coefficient of the corresponding level. S502. Multiply the full-time domain advanced soil and water loss risk assessment value by the corresponding correction coefficient in sequence to obtain the soil and water loss control early warning judgment value. S503. Compare the early warning judgment value with the preset five-level graded early warning threshold range. If it exceeds the upper limit of the threshold range, it is judged as an advanced risk event of soil and water loss of the corresponding level, triggering ecological-engineering collaborative closed-loop intervention command and graded early warning response at the matching control level; if it does not exceed the threshold, it is judged as a stable soil state, maintaining a low-power passive normal monitoring mode. The ecological-engineering collaborative closed-loop intervention command and graded early warning response include pushing risk evolution information to the corresponding level of soil and water conservation management departments, initiating slope capillary water control engineering commands, implementing in-situ remediation and treatment plans for underlying biological crusts, linking on-site passive sound and light early warning equipment, and triggering emergency soil stabilization engineering activation plans.

[0013] A water and soil erosion control and early warning system based on the Internet of Things includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the system includes: The passive distributed sensing module is used to acquire multi-dimensional monitoring source data covering the target control area, including soil interface electrochemistry, soil microstructure deformation, pore hydrological fractal, surface erosion resistance ecology, and external force-driven data, based on a distributed IoT passive sensing architecture. The dual precursor coupling identification module is used to extract the precursor signals of interface instability in the early stage of soil erosion and the precursor signals of soil fissure nucleation from multi-dimensional monitoring source data, and generate the chaotic phase space adaptive monitoring cycle control command based on the coupling of the dual precursor signals. The monitoring sequence reconstruction module is used to dynamically reconstruct the acquisition parameters of multi-dimensional monitoring source data according to the adaptive monitoring cycle adjustment command, and obtain a full-chain continuous monitoring sample set with spatiotemporal registration of multiple batches. The cross-scale risk extrapolation module is used to input the monitoring sample set into the preset fourth-order multi-scale coupled extrapolation model of soil and water loss, complete the cross-scale coupled analysis, and obtain the advanced soil and water loss risk assessment value in the whole time domain. The tiered early warning and control module is used to generate early warning judgment values ​​based on the advanced water and soil erosion risk assessment values ​​across the entire time domain and the control level of the target control area. Based on the comparison results, closed-loop control intervention or routine monitoring is implemented.

[0014] Preferably, the passive distributed sensing module includes: The access point management unit is used to acquire standardized access points for multiple types of passive IoT sensing nodes within the target control area, which correspond one-to-one with the monitoring dimensions of soil interface electrochemistry, soil microstructure deformation, pore hydrological fractal, surface erosion resistance ecology, and external force-driven monitoring. The data synchronization acquisition unit is used to synchronously acquire multiple sets of raw passive monitoring data through various standardized access points; The feature classification and calibration unit is used to extract the unique features of the physical field corresponding to each group of raw passive monitoring data, classify and calibrate the monitoring data according to the feature dimensions, and output multi-dimensional monitoring source data.

[0015] Preferably, the dual-precursor coupling identification module includes: The first precursor calculation unit is used to calculate the corresponding characteristic parameters based on the relaxation electrochemical characteristic data of the soil mineral-microorganism interface, and to complete the identification and judgment of the precursor signals of interface instability in the early stage. The second precursor calculation unit is used to calculate the corresponding characteristic parameters based on the infrasound-acoustic emission coupling characteristic data of soil fracture nucleation, and to complete the identification and judgment of the precursor signal of fracture nucleation. The control command generation unit is used to combine the spatial distribution and intensity level of the two precursor signals and generate adaptive monitoring cycle control commands for the corresponding control zone through a chaotic phase space reconstruction algorithm.

[0016] Preferably, the cross-scale risk extrapolation module includes: The interface instability analysis unit is used to perform calculation and analysis of the interface instability layer of the fourth-order multi-scale coupled deduction model and output the instability potential value of the soil erosion interface. The pore failure analysis unit is used to perform calculation and analysis of the pore failure layer of the model, and outputs the pore failure risk value by combining the interface instability potential value. The fracture nucleation analysis unit is used to perform calculation and analysis of the fracture nucleation layer in the model, and outputs the fracture nucleation trigger value by combining the pore failure risk value. The erosion evolution analysis unit is used to perform calculation and analysis of the erosion evolution layer of the model, and to deduce the full-time-domain risk evolution sequence by combining the output results of the aforementioned three layers. The risk fusion assessment unit is used to perform nonlinear weighted fusion of the output results of each layer of the model to output the advanced soil and water loss risk assessment value in the whole time domain.

[0017] (III) Beneficial Effects 1. By continuously collecting relaxation electrochemical characteristic data of soil mineral-microorganism interface, core parameters such as the temporal change rate of interface double layer relaxation time, the amplitude of interface tension attenuation, and the capacitance distortion coefficient are calculated to capture microscopic signals of soil colloidal water stripping and interface instability. These signals show significant anomalies 10-15 days before the occurrence of macroscopic soil erosion. Simultaneously, through infrasound-acoustic emission coupling monitoring of soil fissure nucleation, infrasound and acoustic emission signals of soil microfissure nucleation and expansion are captured. These signals appear 7-10 days earlier than macroscopic slope deformation. By identifying precursor signals in two dimensions, advanced prediction of soil erosion is achieved, reducing the degree of ecological damage and governance costs caused by soil erosion.

[0018] 2. A dual-dimensional precursor coupling and cross-verification system combining chemical interface instability and physical structure damage is constructed. When a single-dimensional precursor signal exceeds the critical threshold, monitoring intensification is only initiated in the corresponding area, without directly determining it as an effective erosion risk. Only when the precursor signals of both soil interface chemical instability and soil physical structure damage exceed the critical threshold simultaneously is it determined as an effective soil and water loss precursor risk event. At the same time, multi-dimensional monitoring data from five physical fields—soil interface electrochemistry, soil microstructure deformation, pore hydrological fractal, surface erosion resistance ecology, and external driving forces—are combined to cross-verify from the entire chain of erosion evolution, effectively eliminating misjudgments caused by non-erosion factors such as soil wet-dry cycles, environmental temperature fluctuations, and surface debris interference.

[0019] 3. Adopting a distributed IoT passive sensing architecture, all front-end monitoring nodes are powered by passive triggering or micro-energy storage, requiring no external mains power supply. This is naturally suitable for application scenarios in my country's key water and soil erosion areas, which are mostly remote mountainous areas without electricity. At the same time, based on Takens' embedding theorem, chaotic phase space reconstruction is performed. Combined with the spatial distribution range and intensity level of dual precursor signals, the acquisition parameters of each monitoring node are dynamically reconstructed. In risk-free areas, a low-power normal monitoring mode is maintained, and the core unit of the node is in a dormant state. In risk areas triggered by precursor signals, the acquisition frequency is increased, the sampling bandwidth is widened, and the spatial sampling density is increased to achieve high-precision encrypted monitoring of risk areas.

[0020] 4. Construct a full-chain, multi-scale analysis system sequentially from nanoscale soil interface reactions, micrometer-scale pore structure destruction, millimeter-scale microfracture nucleation to meter-scale macroscopic erosion evolution. First, calculate the soil particle bound water stripping coefficient and colloidal dispersion degree through the interface instability layer to obtain the interface instability potential value of soil erosion. Then, calculate the change rate of soil pore structure fractal dimension and pore connectivity through the pore failure layer, and obtain the pore failure risk value by combining the interface instability potential value. Next, calculate the soil microfracture propagation rate, connectivity probability, and nucleation density through the fracture nucleation layer, and obtain the fracture nucleation trigger value by combining the pore failure risk value. Finally, incorporating slope erosion resistance and external force driving data through the erosion evolution layer, predict the future 168-hour... The system fully reconstructs the intrinsic evolution mechanism of soil erosion, including its occurrence range, erosion modulus, sediment transport pathways, and severity, rather than relying on empirical statistical judgments based on historical data fitting by existing technologies. Furthermore, after an early warning is triggered, corresponding control and intervention instructions can be matched according to the warning level, combined with the ecological protection red line level and soil and water conservation management level of the target control area. These instructions include pushing the entire chain of risk evolution information to the corresponding level of soil and water conservation management departments, initiating instructions for slope capillary water control projects, implementing in-situ remediation and treatment plans for underlying biological crusts, linking on-site passive audible and visual early warning equipment, and triggering emergency soil stabilization engineering plans. This achieves seamless integration from risk warning to treatment implementation. Attached Figure Description

[0021] The above description is only an overview. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0022] Figure 1 This is a flowchart of the early warning method for advanced control of soil erosion in Embodiment 1 of the present invention; Figure 2 This is a flowchart of the operation of the early warning system for advanced control of soil erosion in Embodiment 2 of the present invention. Detailed Implementation

[0023] This application provides an IoT-based early warning system and method for soil erosion control, which addresses the technical problems of delayed early warning, high false alarm rate, the incompatibility between monitoring accuracy and low power consumption, low accuracy of risk projection, and disconnect between control and management processes.

[0024] Example 1: Due to the inability to effectively control soil erosion, we conducted research on some typical types of soil erosion and analyzed operational data from existing monitoring systems. We discovered that the severe delay in early warnings stemmed from an overemphasis on macroscopic indicators such as slope deformation and soil moisture content. This resulted in warnings only being issued after erosion had occurred, neglecting the importance of other stages in the soil erosion evolution process. This limited warnings to approximately 24 hours in advance, failing to meet pre-disaster prevention needs. The evolution of soil erosion involves, in sequence, the stripping of bound water at the soil mineral-water interface, colloid dispersion, pore structure destruction, microcrack development, and macroscopic erosion. Furthermore, single-factor monitoring is highly susceptible to interference from complex environmental factors such as wet-dry cycles, temperature fluctuations, trampling by humans and animals, wind vibration, and rain splash, leading to false alarms. This large number of ineffective warnings severely undermines the credibility of the early warning system. To address these issues, we adopted a combined approach of microscopic precursor dual-coupling identification, chaotic adaptive regulation, and cross-scale mechanism deduction for advanced early warning. The specific details are as follows: Taking the key area for soil and water conservation in the Loess Plateau as an example, this area is a typical loess hilly and gully landform, and there is no stable power supply or wired network coverage. The average annual rainfall is about 480 mm, and heavy rainfall during the flood season causes frequent erosion. The original monitoring system only deployed rain gauges and GPS slope deformation monitoring stations.

[0025] Reference Figure 1As shown, we first used UAV aerial topographic mapping, 1m deep soil profile survey, comprehensive watershed survey of erosion channels, and remote sensing interpretation of vegetation cover to classify the regional erosion risk level and determine the spatial distribution of high-risk erosion channels, steep-slope farmland, and sparsely vegetated areas. We then used in-situ field tests to verify the matching between erosion units and monitoring grids, determining differentiated grid layout rules: a 200m×200m grid was used for high-risk erosion areas with a slope >25°; a 500m×500m grid was used for medium-risk areas with a slope of 15°-25°; and a 1000m×1000m grid was used for low-risk areas with a slope <15° and vegetation cover >60%. Each deployed monitoring device has complete multi-dimensional monitoring nodes, ensuring that the precursor signals of each erosion unit can be fully captured without any monitoring blind spots. For monitoring blind spots without sunlight, such as the bottom of erosion gullies and forest slopes, a combination of piezoelectric vibration and thin-film solar energy is used. This allows for self-powered operation by utilizing the piezoelectric effect of raindrops and soil micro-vibrations under no-sunlight conditions. At the same time, a Mesh self-organizing network protocol is used to allow blind spot nodes to complete data relay transmission through adjacent nodes. In the no-sunlight monitoring blind spots of the main gully, this method enables continuous monitoring of erosion precursors at the bottom of the gully, improving the early warning time of gully erosion by 2 days compared to slope monitoring. A passive IoT sensing node system was constructed, but to ensure stable operation in the field, all nodes needed to pass three performance tests: IP68 waterproof and dustproof, high and low temperature cycling, and field corrosion resistance, to adapt to extreme temperature environments ranging from -30℃ to 60℃. To address the weak electrochemical signals at the soil interface, a three-electrode system specifically designed for soil electrochemical monitoring was adopted: a high-purity gold working electrode, a dual liquid-junction Ag / AgCl reference electrode, and a ring platinum counter electrode, resolving signal drift issues caused by soil corrosion and liquid junction potential interference. To address the transient nature of soil micro-fracture signals, a passive triggered infrasound-acoustic emission sensing unit was used, waking up only when the signal amplitude exceeds three times the background noise, balancing transient signal capture with low-power operation. Simultaneously, capillary water characteristic monitoring and slope erosion resistance high-gloss sensors were also included. The system comprises three types of nodes: spectrum monitoring, meteorological driving factor monitoring, and a supporting edge computing and communication platform. This platform utilizes industrial-grade embedded processors and industrial-grade wireless aggregation gateways, coupled with a solar-powered micro-energy storage system adapted for power-free outdoor scenarios. This ensures the gateways can operate stably for over 30 consecutive days even on cloudy or rainy days. The gateways are deployed at high points in the region, with a spacing of ≤5km between the three gateways. A spreading factor of SF12 is used to ensure that the wireless signal coverage strength of all monitoring nodes is ≥-110dBm. Microsecond-level time synchronization of all nodes is then achieved via the IEEE 1588 precise time synchronization protocol to avoid time-series misalignment of multi-dimensional data. A BeiDou satellite communication backup module is included to enable emergency transmission of core early warning data in the event of a public network outage. Furthermore, a built-in SM4 hardware encryption chip ensures secure and compliant data transmission and storage.

[0026] In addition, to address the issue of electromagnetic interference, random soil vibration, and rain splashing masking weak microscopic precursor signals in the raw signals collected in the field, an adaptive soft threshold denoising algorithm that dynamically adjusts with the number of wavelet decomposition levels is used. First, the raw monitoring signal is decomposed into five levels of wavelets, and the db4 wavelet basis function, adapted to non-stationary low-frequency signals, is selected to obtain wavelet coefficients at different scales. Then, an adaptive threshold is dynamically calculated based on the number of decomposition levels, decreasing as the number of decomposition levels increases, achieving strong denoising of high-frequency noise at lower levels and weak processing of effective low-frequency signals at higher levels. Subsequently, wavelet coefficients larger than the threshold are subjected to soft threshold shrinkage, while those smaller than the threshold are set to zero. Finally, the denoised effective signal is reconstructed using the processed wavelet coefficients. This method can completely preserve weak low-frequency precursor signals while removing noise.

[0027] The denoised signal is checked for CRC32 cyclic redundancy check to remove erroneous data. Data packets that fail the check are triggered by a 3-retransmission mechanism. If the 3 retransmissions fail, the node is marked as abnormal and an operation and maintenance warning is pushed to the on-site maintenance personnel through the platform. Valid data that passes the check is archived locally according to timestamp, node ID and spatial coordinates, with a local storage period of ≥12 months. At the same time, it is synchronized to the cloud platform for long-term storage with a storage period of ≥5 years, which complies with the water and soil conservation monitoring data management specifications.

[0028] After data cleaning, multi-dimensional feature classification and calibration were completed according to the key links of the chain evolution of soil erosion. Each feature was set with a fixed calculation time window and step size to ensure the consistency and comparability of feature data. That is, the interface double layer relaxation time, the relative change of interface tension, and the degree of double layer capacitance distortion within a 24-hour time window were extracted from the electrochemical data and calibrated as the relaxation electrochemical feature data of soil mineral-microorganism interface, corresponding to the initial link of the erosion chain process. The microfracture infrasound spectrum features, acoustic emission event time sequence, single event energy release features, and b-values ​​within a 1-hour time window are extracted from the acoustic emission data and calibrated as soil fracture nucleation infrasound-acoustic emission coupling characteristic data, corresponding to the fracture development stage in the erosion chain process. The capillary rise-fall hysteresis curve, soil water potential temporal variation characteristics, and multifractal spectrum within a 24-hour time window were extracted from soil water potential data and labeled as capillary hysteresis loop fractal characteristic data of unsaturated soil, corresponding to the pore destruction link in the erosion chain process. Chlorophyll fluorescence quantum yield, photochemical quenching coefficient, and effective photosynthetic efficiency within a single day time window are extracted from hyperspectral data and calibrated as characteristic data of hyperspectral fluorescence quantum yield of biological crust, corresponding to the evaluation of slope erosion resistance. The real-time rainfall intensity, 10-minute peak rainfall intensity, wind erosion kinetic energy, near-surface turbulence intensity, and 168-hour numerical rainfall forecasts from regional meteorological stations were extracted from meteorological data and labeled as rainfall-wind erosion coupled driving characteristic data, corresponding to the analysis of external erosion forces.

[0029] During the calibration process, feature validity verification is performed simultaneously. Invalid features that exceed the physical reasonable range are directly removed, and a remote calibration command is triggered at the node. If three consecutive sets of features are invalid, the device is marked as faulty. The resulting standardized feature dataset is stored in a structured manner according to time series and spatial grid, providing reliable input for subsequent precursor identification.

[0030] Based on this, we preprocessed and enhanced the features of the multi-source heterogeneous data fusion. First, on the basis of the original ground node monitoring data, we fused high-resolution aerial survey data from UAVs and remote sensing inversion data from Sentinel satellites. We completed the accurate registration and fusion preprocessing of the multi-source data by using a unified spatiotemporal benchmark. For the sparse monitoring areas at the edge of the project area, we used vegetation coverage, surface soil moisture, and gully development data retrieved by remote sensing to complete feature enhancement. We filled the monitoring gaps of the ground nodes by using the Kriging spatiotemporal interpolation algorithm to improve the effective monitoring coverage and the feature data completeness rate of low-risk areas. At the same time, we further reduced the error of the feature data by cross-validating the multi-source data.

[0031] The dual microscopic precursor signals of soil interface instability and soil fissure nucleation are used for early warning of soil erosion in the early stage. The problem of early warning lag and high false alarm rate is solved by the dual precursor coupling verification mechanism. 10-15 days before the occurrence of macroscopic soil erosion, rainwater infiltration will cause the stripping of water-bound soil mineral-water interface and dispersion of colloidal particles. The relaxation characteristics of the interface double electric layer will undergo significant regular changes. This is the earliest identifiable signal of the erosion chain process. 7-10 days before the occurrence of erosion, the nucleation of microfissures in the soil will generate stable infrasound-acoustic emission signals, forming the second key precursor window.

[0032] However, traditional ideal medium models have excessive calculation errors in polydisperse soil colloidal systems. Therefore, we introduce a modified soil colloidal relaxation model. By introducing a relaxation time distribution coefficient, we correct the ideal homogeneous medium assumption of the traditional Debye model, adapt it to the polydisperse characteristics of soil colloids, and accurately calculate three core judgment indicators: the relaxation time of the interfacial double layer, the relative change of interfacial tension, and the degree of double layer capacitance distortion. Then, using 10 years of historical erosion event data and indoor experimental data, we use ROC curves and the Youden exponent maximization method to calibrate the critical thresholds of the three indicators. Specifically, we select electrochemical monitoring data from 15 days before the occurrence of historical erosion events as positive samples and continuous monitoring data from the stable soil state as negative samples. We plot ROC curves to calculate the true positive rate and false positive rate under different thresholds. The parameter value corresponding to the maximum value of the Youden exponent is the optimal critical threshold. If any of the three indicators exceeds the critical threshold, it is determined that a precursor signal of interface instability in the early stage has been captured.

[0033] Simultaneously, for the identification of precursors to soil fissure nucleation, the magnitude-frequency relationship from the field of rock mass fracture early warning is introduced into the field of soil erosion monitoring. First, based on the amplitude and cumulative frequency of acoustic emission events, the b-value is obtained through linear fitting using the least squares method. When micro-fissures inside the soil develop rapidly, the proportion of small-scale fracture events increases, and the b-value will show a continuous decrease. This serves as the core criterion for determining fissure nucleation. Simultaneously, three auxiliary judgment indicators are added: the frequency of micro-fracture events within a 1-hour time window, the energy release rate of a single event, and the infrasound spectrum shift. The critical threshold is calibrated using the same ROC curve method. If the core indicator continuously decreases by more than 30% within 24 hours, or if any one of the three auxiliary indicators exceeds the critical threshold, it is determined that a precursor signal of fissure nucleation has been captured. The results can be dynamically adjusted within a range of 7-10 days according to geological and geomorphological characteristics.

[0034] Based on the mechanism that interface instability and fracture nucleation are two consecutive stages in the chain process of soil erosion, the precursor signal of interface instability should appear earlier than the precursor signal of fracture nucleation. The two are strongly correlated in time. Cross-correlation analysis quantifies the temporal coupling degree of the two precursor signals. Only when both precursor signals simultaneously exceed the critical threshold and the temporal coupling coefficient is >0.7 is it determined to be a valid soil erosion precursor risk event, initiating the highest level of monitoring. If a single precursor signal exceeds the limit, it is only marked as a risk concern area, initiating lower-level monitoring without triggering an early warning. This mechanism can reduce the false alarm rate caused by environmental interference. Based on this core dual-precursor identification scheme, the existing electrical... Based on the core dimensions of chemical and acoustic emission precursors, two auxiliary precursor dimensions can be added: soil microbial community carbon source utilization characteristics and soil moisture content vertical gradient. Among them, microbial community characteristics are obtained by inverting in-situ monitoring data of interface electrochemical nodes, and moisture content vertical gradient is obtained by layered capillary water nodes. Then, a lightweight decision tree model is used to complete the intelligent coupling identification of multiple precursors and automatically filter out single index anomalies caused by non-erosive factors. In the adaptation and verification of this extended scheme in the red soil area of ​​southern China, the early warning window of the early warning period is extended and the false alarm rate of the warning is further reduced. In particular, it effectively filters the seasonal dry and wet cycle interference in the red clay area.

[0035] Since the evolution of precursor signals of soil erosion is a typical chaotic process, it is sensitive to initial conditions and predictable in the short term. By reconstructing the phase space, the one-dimensional precursor time series signal can be converted into a high-dimensional chaotic system, and the evolution characteristics of the system can be fully restored. Then, the evolution trend of the system can be judged by calculating the maximum Lyapunov exponent. When the exponent is greater than 0, it indicates that the chaotic system is in a divergent state, the precursor signal continues to strengthen, and the risk of soil erosion continues to rise. When the exponent is less than or equal to 0, it indicates that the system is in a stable state and the risk does not show an upward trend.

[0036] For example, the CC method is used to simultaneously determine the optimal delay time and optimal embedding dimension of a time series. Based on Takens' embedding theorem, chaotic phase space reconstruction is performed, converting the one-dimensional precursor time series signal into a high-dimensional phase space matrix. Then, the Wolf algorithm is used to calculate the maximum Lyapunov exponent. Based on this calculation result and the intensity and spatial distribution of the precursor signal, three risk levels—low, medium, and high—are defined. Corresponding adaptive monitoring cycle control instructions are generated for each control zone. In the high-risk zone, the acquisition frequency and sampling bandwidth are significantly increased; in the medium-risk zone, the acquisition frequency of core indicators is specifically increased; and in the low-risk zone, a normal low-power acquisition strategy is maintained. That is, in the high-risk zone... The control instructions for the risk zone are as follows: the acquisition cycle of the interface electrochemical node is shortened from 12 hours / time to 48 minutes / time; the sampling bandwidth of the infrasound-acoustic emission node is widened from 1kHz to 8kHz; the wake-up threshold is reduced from 3 times the background noise to 2 times; the monitoring cycle of the capillary water morphology node is shortened from 6 hours / time to 45 minutes / time; the sampling frequency of the hyperspectral node is increased from 1 time / 24 hours to 4 times / 24 hours; the data acquisition frequency of the meteorological node is increased from 1 time / 1 hour to 1 time / 10 minutes; in the medium-risk zone, only the acquisition frequency of the electrochemical and acoustic emission nodes is increased, while the other nodes maintain normal parameters; and in the low-risk zone, normal low-power acquisition parameters are maintained.

[0037] The command is sent to the corresponding node via the LoRa network. After receiving the command, the node immediately updates the collection parameters and sends back the execution confirmation information. If the node does not send back the command, the gateway will send the command three times. If there is still no response, the node will be marked as a device abnormality, triggering an operation and maintenance warning. At the same time, the encrypted collection of adjacent nodes will be enabled to fill the monitoring blind spot.

[0038] According to the control instructions, the gateway generates a step-by-step data acquisition task for each node, specifying the acquisition time window, acquisition frequency, sampling bandwidth, spatial sampling density, and trigger threshold. This ensures that the acquisition time windows of all nodes in the risk area remain synchronized at the microsecond level. After the data acquired by the nodes is transmitted back, a secondary verification is performed to remove invalid data caused by node failure, transmission interruption, or signals exceeding the physically reasonable range. Missing critical data is supplemented using Kriging spatiotemporal interpolation of adjacent nodes in the same grid to ensure the continuity of the data sequence. Finally, the data is processed using UTC timestamps as the time base and CGCS2000 national geodetic coordinates. The standard system completes the spatiotemporal registration of all-dimensional data as a spatial benchmark, ensuring that multi-dimensional data at the same time and spatial location correspond one-to-one. After registration, the 3σ criterion is used to remove outliers exceeding the mean ± 3 times the standard deviation. First-order polynomial fitting is used to complete signal baseline correction to eliminate the influence of sensor zero-point drift. The 5-point moving average method is used to complete data smoothing, denoising and trend term extraction. Finally, the data are integrated to form a full-chain continuous monitoring sample set with multiple batches of spatiotemporal registration, which is divided into a model training set and a validation set in a 7:3 ratio to provide standardized input for subsequent erosion risk projection.

[0039] In multiple project areas across the Loess Plateau, the Southern Red Soil Region, the Karst Desertification Region, and the Northeast Black Soil Region, an edge-side federated learning framework was established. The edge gateways in each region do not share raw monitoring data, but only upload the gradient parameters of the locally trained model. The globally optimal adaptive control strategy model is then collaboratively trained in the cloud, and the optimized model is then distributed to the edge gateways in each region. In the freeze-thaw period application in the Northeast Black Soil Region, the control strategy was optimized through cross-regional collaborative training to address the special erosion patterns of the black soil freeze-thaw cycle, further improving the winter endurance of nodes and increasing the capture rate of precursory freeze-thaw erosion signals.

[0040] A four-level multi-scale coupled erosion risk simulation is constructed based on the chain evolution mechanism of soil erosion, namely interface instability, porosity failure, fissure nucleation, and macroscopic erosion. This model enables accurate dynamic simulation of the entire chain of soil erosion from micro to macro levels. The four-level model divides the erosion evolution process into four progressive levels: interface instability layer, porosity failure layer, fissure nucleation layer, and erosion progression layer. The output of the previous level serves as the input of the next, forming a progressive cross-scale coupling. Details are as follows: By using nonlinear cross-scale coupled control equations, the evolution processes at four levels are coupled and correlated. At least 10 experts from the fields of soil and water conservation, soil science, and geotechnical mechanics are invited to complete the scoring of the importance of indicators through the analytic hierarchy process, and the coupling weight coefficients of each level are calibrated to achieve the coupling of the entire chain mechanism from microscopic interface changes to macroscopic erosion.

[0041] First, in the unstable layer of the interface, the potential value of the interface instability of soil and water loss is calculated by combining the interfacial tension and capacitance distortion coefficient. The initial attenuation of the erosion resistance caused by the stripping of water and the dispersion of colloids at the soil interface is quantified. The value ranges from 0 to 1, and the larger the value, the higher the degree of interface instability. Subsequently, the study proceeds to the pore failure layer. Addressing the multifractal characteristics of soil pore structure failure, traditional single fractal dimension methods cannot accurately quantify the non-uniform failure process of pores. Therefore, multifractal detrended fluctuation analysis is employed to calculate the multifractal spectrum of the capillary hysteresis loop. Specifically, the steps involve constructing a cumulative deviation sequence from the capillary potential time series, dividing the cumulative deviation sequence into fixed-length non-overlapping segments, calculating the detrended fluctuation function for each segment, and finally calculating the q-order fluctuation function. A generalized Hurst exponent is obtained through linear fitting, and the multifractal spectrum is then calculated. The rate of change of fractal dimension is extracted, and combined with the soil pore connectivity calculated using the pore network model, along with the interface instability potential value from the previous layer, the pore failure risk value is calculated. This quantifies the secondary attenuation of corrosion resistance caused by pore structure failure, with a value range of 0-1. Next, we enter the fracture nucleation layer. Combining the microfracture propagation rate, connectivity probability, and nucleation density obtained from acoustic emission spatiotemporal localization, and superimposing the pore failure risk value, we calculate the fracture nucleation trigger value to quantify the structural instability caused by soil fracture development. The value range is 0-1. Finally, we enter the erosion evolution layer. Based on the traditional soil loss equation, we introduce two microscopic precursor correction terms for the first time: interface instability potential and fracture nucleation triggering. This achieves deep coupling between microscopic precursor signals and macroscopic erosion models. The corrected equation follows relevant industry standards for soil erosion classification and grading. Combined with a 1m resolution digital elevation model and water system distribution data, we use the D8 algorithm to extrapolate the occurrence range, erosion modulus, sediment transport path, and severity of soil erosion in the next 168 hours, generating an hourly full-time-domain risk evolution sequence. The extrapolation duration can be dynamically adjusted within the range of 24-168 hours according to control requirements. Finally, through nonlinear weighted fusion, we obtain the full-time-domain advanced soil erosion risk assessment value, ranging from 0 to 1. The larger the value, the higher the soil erosion risk. The risk assessment value before a heavy rainfall event in this project area is 0.78.

[0042] Based on the original fourth-order model, future rainfall scenario data from the CMIP6 global climate model and human activity impact factors from regional land use planning and ecological restoration projects were incorporated. The model's extrapolation scale was extended from a short-term 168-hour period to a long-term 10-30-year period, extrapolating the evolution trend of soil erosion under different climate scenarios and governance measures. In the application of the Yanhe project area, this extended scheme successfully identified three new high-risk erosion channels to be added in the next 15 years, and predicted in advance the increase in flood season erosion risk caused by climate change. This provided accurate data support for the layout of long-term soil and water conservation planning and ecological restoration projects in the region. At the same time, the benefits simulation of different ecological restoration measures was completed, providing a quantitative basis for the selection of governance schemes.

[0043] Based on statistical analysis of soil erosion disasters, and combined with the national emergency response plan for sudden environmental incidents and the governance needs of areas at different control levels, a five-level tiered early warning system was established. An adaptive correction mechanism linked to the control level was also designed. Details are as follows: Values ​​are assigned based on the ecological protection red line level and the soil and water conservation control level. The risk weight correction coefficient and the early warning threshold adjustment coefficient are linearly adjusted. The higher the control level, the larger the risk weight correction coefficient and the smaller the early warning threshold adjustment coefficient, and the higher the early warning sensitivity, ensuring that the early warning sensitivity is fully matched with the regional control requirements.

[0044] The calculated soil erosion risk assessment value is multiplied by two correction coefficients corresponding to the control level to obtain the final warning judgment value. This value is then compared with the preset five-level warning threshold range to determine the corresponding warning level. Different warning levels specify the response time limit, responsible party, handling measures, and effectiveness evaluation standards: Level 1 warning (extremely high risk) corresponds to a warning judgment value ≥ 0.8, with a response time limit of 2 hours. The responsible parties are the provincial soil and water conservation department and the regional management unit. Handling measures include pushing the entire risk chain information to the provincial department, issuing emergency response plans, activating on-site warning equipment, transferring personnel at risk, and initiating emergency soil stabilization projects, with on-site handling completed within 24 hours; Level 2 warning (high risk)... The corresponding early warning threshold is 0.6-0.8, with a response time limit of 12 hours. The responsible entity is the municipal-level soil and water conservation department, and the response measures include pushing risk information, increasing monitoring, making advance preparations for response, and conducting on-site inspections. The corresponding warning threshold is 0.4-0.6, with a response time limit of 24 hours. The responsible entity is the regional management unit, and the response measures include continuously tracking risk changes, increasing monitoring of key nodes, and ensuring daily protection. The corresponding warning threshold is 0.2-0.4, with routine monitoring maintained and risk briefings pushed out weekly. The corresponding warning threshold is <0.2, with routine monitoring maintained at low power consumption and monitoring briefings pushed out monthly.

[0045] During implementation, a full-scenario anomaly handling mechanism is established simultaneously, covering all common anomaly scenarios in field engineering. For example, when node communication is interrupted, data is abnormal, or power supply fails, an equipment fault warning is immediately triggered. At the same time, adjacent nodes are activated to encrypt data collection to fill monitoring blind spots. The faulty node is repaired or replaced within 72 hours, and recalibrated after replacement. When more than three sets of feature data are invalid, remote node calibration is immediately triggered. If calibration is invalid, it is marked as an equipment fault, and on-site investigation is arranged. When the model simulation results exceed the reasonable range for 24 consecutive hours, automatic simulation is suspended, and the model parameters are reset using historical benchmark data. After recalibration based on on-site inspection data, operation is resumed. When the gateway's public network communication is interrupted, it automatically switches to local offline operation mode, fully storing all monitoring data and early warning information. After communication is restored, all offline data is automatically retransmitted. Under extreme weather conditions such as heavy rain, heavy snow, and high temperatures, the equipment protection mode is automatically activated, and the data collection strategy is adjusted to ensure that core data is not lost. After the weather recovers, the missing data is automatically collected to ensure the stable operation of the system in various complex field scenarios. Based on the existing closed-loop early warning and response system, a new module for quantifying the benefits of soil and water conservation measures has been added. This module quantifies the reduction rate of erosion modulus, sediment interception benefits, and improvement in slope erosion resistance by using monitoring data before and after the early warning for projects such as in-situ remediation of biological crusts, horizontal terrace land preparation on slopes, and soil stabilization in gullies and valleys. The benefit assessment results are then fed back into the model to iteratively optimize the model's coupling weights, early warning thresholds, and response scheme library. After a level-one early warning for heavy rainfall in the project area, the benefit assessment of biological crust remediation measures was completed through this extended scheme. Based on the assessment results, the model weight coefficients for the red clay area were optimized, improving the accuracy of erosion risk projection for subsequent similar rainfall events.

[0046] Example 2: Reference Figure 2 As shown, an Internet of Things-based soil erosion control and early warning system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the system further includes: A passive distributed sensing module is used to acquire multi-dimensional monitoring source data covering soil interface electrochemistry, soil microstructure deformation, pore hydrological fractals, surface erosion resistance ecology, and external force-driven factors within a target control area, based on a distributed IoT passive sensing architecture. The passive distributed sensing module includes an access point management unit, used to acquire standardized access points for multiple types of passive IoT sensing nodes within the target control area, each corresponding to one of the monitoring dimensions: soil interface electrochemistry, soil microstructure deformation, pore hydrological fractals, surface erosion resistance ecology, and external force-driven factors; a data synchronization acquisition unit, used to synchronously acquire multiple sets of raw passive monitoring data through each standardized access point; and a feature classification and calibration unit, used to extract the unique features of the physical field corresponding to each set of raw passive monitoring data, classify and calibrate the monitoring data according to the feature dimensions, and output multi-dimensional monitoring source data. A dual-precursor coupling identification module is used to extract precursor signals of interface instability and soil fissure nucleation from multi-dimensional monitoring source data. Based on the coupling of the dual precursor signals, an adaptive monitoring cycle control command in chaotic phase space is generated. The dual-precursor coupling identification module includes a first precursor calculation unit, used to calculate corresponding characteristic parameters based on the relaxation electrochemical characteristic data of the soil mineral-microorganism interface, to complete the identification and judgment of the precursor signals of interface instability in the early stage; a second precursor calculation unit, used to calculate corresponding characteristic parameters based on the infrasound-acoustic emission coupling characteristic data of soil fissure nucleation, to complete the identification and judgment of the precursor signals of fissure nucleation; and a control command generation unit, used to combine the spatial distribution and intensity level of the dual precursor signals and generate adaptive monitoring cycle control commands for the corresponding control zones through a chaotic phase space reconstruction algorithm. The monitoring sequence reconstruction module is used to dynamically reconstruct the acquisition parameters of multi-dimensional monitoring source data according to the adaptive monitoring cycle adjustment command, and obtain a full-chain continuous monitoring sample set with spatiotemporal registration of multiple batches. The cross-scale risk extrapolation module is used to input the monitoring sample set into a preset fourth-order multi-scale coupled extrapolation model of soil and water loss, complete cross-scale coupled analysis, and obtain the full-time domain advanced soil and water loss risk assessment value. The cross-scale risk extrapolation module includes: an interface instability analysis unit, used to perform calculation and analysis of the interface instability layer of the fourth-order multi-scale coupled extrapolation model and output the interface instability potential value of soil and water loss; a pore failure analysis unit, used to perform calculation and analysis of the pore failure layer of the model and output the pore failure risk value in combination with the interface instability potential value; a fracture nucleation analysis unit, used to perform calculation and analysis of the fracture nucleation layer of the model and output the fracture nucleation trigger value in combination with the pore failure risk value; an erosion evolution analysis unit, used to perform calculation and analysis of the erosion evolution layer of the model and extrapolate the full-time domain risk evolution sequence in combination with the output results of the above three layers; and a risk fusion assessment unit, used to perform nonlinear weighted fusion of the output results of each layer of the model and output the full-time domain advanced soil and water loss risk assessment value. The tiered early warning and control module is used to generate early warning judgment values ​​based on the advanced water and soil erosion risk assessment values ​​across the entire time domain and the control level of the target control area. Based on the comparison results, closed-loop control intervention or routine monitoring is implemented.

[0047] Example 3: Based on regional erosion characteristics, soil types, and control requirements, parameter calibration and model correction were completed to ensure the universality and replicability of the technical solution. Targeted optimizations were performed for the erosion characteristics of different regions, resulting in standardized scenario adaptation specifications. In the southern red soil region, the soil type is Quaternary red clay, characterized by high clay content, small porosity, and poor permeability. The average annual rainfall is 1580 mm, concentrated between April and June. Biological crusting is dominated by bryophytes, and water erosion is the primary erosion type. The erosion characteristics show that pore structure destruction precedes interfacial chemical instability, significantly differing from the erosion patterns of loess soils in the Loess Plateau. The optimal optimization scheme involves recalibrating the weighting coefficients of a fourth-order multi-scale coupled model through indoor red clay trench erosion tests, and adjusting the pore-damaged layer... The weight was increased from 0.25 to 0.3, and the weight of the interface instability layer was adjusted from 0.3 to 0.25. To match the erosion pattern of the red clay pore structure being destroyed first, the critical threshold of the precursor signal was recalibrated using regional historical erosion event data. The critical threshold of the relaxation time time series change rate was adjusted to ±18% / 24h, the critical threshold of the micro-fracture event frequency was adjusted to 9 times / hour, the early warning window of the precursor signal of the interface instability in the early stage was dynamically adjusted to 10 days, and the early warning window of the precursor signal of fracture nucleation was dynamically adjusted to 7 days. In the karst rocky desertification area of ​​Southwest China, the soil type is limestone soil with an average soil layer thickness of less than 30 cm and a bedrock exposure rate of over 60%. Soil erosion is mainly underground leakage, which is completely different from the slope hydraulic erosion pattern of the Loess Plateau. The adaptation and optimization scheme is to adjust the node layout and burial depth, adjust the burial depth of the interface electrochemical node and the acoustic emission node to 10 cm and 5 cm respectively to meet the monitoring needs of shallow soil layers, and densify the grid to 200 m × 200 m to meet the characteristics of strong spatial heterogeneity of erosion. The bedrock fissure leakage correction term is added to the modified soil loss equation, and the weight coefficient of the fourth-order model is adjusted to increase the weight of the fissure nucleation layer to 0.35. The core control factor of underground leakage is the development law of bedrock fissures. Through indoor limestone soil test and regional historical leakage event data, the critical threshold of precursor signals is recalibrated, the critical threshold of relaxation time time series change rate is adjusted to ±15% / 24h, and the critical threshold of micro-fracture event frequency is adjusted to 8 times / hour. In the Northeast Black Soil Region, the soil type is black soil with high organic matter content. The erosion types are mainly slope hydraulic erosion and freeze-thaw erosion. The erosion risk is highest during the spring freeze-thaw period. The adaptation and optimization scheme is to add a freeze-thaw cycle correction term to the fourth-order model, recalibrate the critical threshold of the precursor signal during the freeze-thaw period, adjust the low-power operation strategy of the node in winter, and optimize the adaptive control mechanism during the freeze-thaw period. It has been verified after one freeze-thaw cycle.

[0048] Finally, it should be noted that the above embodiments are merely examples for clearly illustrating the present invention and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for early warning and control of soil erosion based on the Internet of Things, characterized in that, The method includes the following steps: S1. Based on a distributed IoT passive sensing architecture, acquire multi-dimensional monitoring source data covering soil interface electrochemistry, soil microstructure deformation, pore hydrological fractal, surface erosion resistance ecology, and external force-driven data in the target control area. S2. Extract the early warning signals of interface instability and soil fissure nucleation from multi-dimensional monitoring source data, and generate the chaotic phase space adaptive monitoring cycle control command based on the coupling of the two early warning signals. S3. Based on the chaotic phase space adaptive monitoring cycle control command, dynamically reconstruct the acquisition parameters of multi-dimensional monitoring source data to obtain a full-chain continuous monitoring sample set with multiple batches of spatiotemporal registration. S4. Input the full-chain continuous monitoring sample set with multiple batches of spatiotemporal registration into the preset fourth-order multi-scale coupled deduction model of soil and water loss, complete the cross-scale coupled analysis from micro-interface instability to macro-erosion evolution, and obtain the advanced soil and water loss risk assessment value in the whole time domain. S5. Based on the advanced water and soil erosion risk assessment value across the entire time domain, and combined with the control level of the target control area, generate an early warning judgment value for water and soil erosion control. Based on the comparison result between the early warning judgment value and the preset threshold, implement closed-loop control intervention or routine monitoring at the corresponding level.

2. The method for early warning and control of soil erosion based on the Internet of Things according to claim 1, characterized in that, In step S1, the multi-dimensional monitoring source data includes soil mineral-microorganism interface relaxation electrochemical characteristic data, soil fissure nucleation infrasound-acoustic emission coupling characteristic data, unsaturated soil capillary water hysteresis ring fractal characteristic data, biological crust hyperspectral fluorescence quantum yield characteristic data, and rainfall-wind erosion coupling driving characteristic data. Specifically, step S1 includes: S101. Obtain standardized access points for multiple types of passive IoT sensing nodes within the target control area. The sensing nodes correspond one-to-one with the monitoring dimensions of soil interface electrochemistry, soil microstructure deformation, pore hydrological fractal, surface erosion resistance ecology, and external force-driven monitoring. S102. Simultaneously collect multiple sets of raw passive monitoring data through each standardized access point; S103. Extract the unique features of the physical field corresponding to each group of original passive monitoring data, and classify and calibrate the monitoring data according to the feature dimensions to obtain the multi-dimensional monitoring source data.

3. The method for early warning and control of soil erosion based on the Internet of Things according to claim 2, characterized in that, Step S2 specifically includes: S201. Based on the relaxation electrochemical characteristic data of the soil mineral-microorganism interface, calculate the temporal change rate of the interface double electric layer relaxation time, the amplitude of the interface tension decay and the capacitance distortion coefficient. When any parameter exceeds the corresponding critical threshold, it is determined that a precursor signal of interface instability in the early stage has been captured. The early stage is 10-15 days before the occurrence of macro-erosion of soil and water loss, which can be dynamically adjusted according to the soil type and climate characteristics of the target control area. S202. Based on the infrasound-acoustic emission coupling characteristic data of soil fissure nucleation, calculate the frequency of soil micro-fracture events, energy release rate and infrasound spectrum shift. When any parameter exceeds the corresponding critical threshold, it is determined that a pre-nucleation signal of fissure nucleation has been captured. The pre-nucleation signal of fissure nucleation is 7-10 days earlier than the macroscopic slope deformation of the soil and can be dynamically adjusted according to the geological and geomorphological characteristics of the target control area. S203. Combining the spatial distribution range and intensity level of the dual precursor signals, an adaptive monitoring cycle control command for the corresponding control zone is generated through a chaotic phase space reconstruction algorithm. The control command is used to synchronously adjust the acquisition frequency, sampling bandwidth and spatial resolution of monitoring data in each dimension.

4. The method for early warning and control of soil erosion based on the Internet of Things according to claim 3, characterized in that, Step S3 specifically includes: S301. According to the adaptive monitoring cycle control instruction, generate a distributed passive acquisition task corresponding to the monitoring data of each dimension, and determine the acquisition configuration parameters of the corresponding sensing node. The acquisition configuration parameters include acquisition time window, acquisition frequency, sampling bandwidth, spatial sampling density, and trigger threshold. S302. Based on the acquisition configuration parameters, complete synchronous and continuous acquisition to obtain monitoring information in various dimensions; S303. All monitoring information of all dimensions is spatiotemporally registered with unified timestamps and spatial geographic coordinates. After data preprocessing such as outlier removal, baseline correction and trend term extraction, it is integrated to form multiple batches of full-chain continuous monitoring sample sets. The sample sets cover the entire process of data from changes in soil microstructure to macro erosion evolution.

5. The method for early warning and control of soil erosion based on the Internet of Things according to claim 4, characterized in that, Step S4 specifically includes: S401. Input the monitoring sample set into the interface instability layer of the fourth-order multi-scale coupled deduction model, calculate the soil particle bound water stripping coefficient and colloidal dispersion degree, and obtain the interface instability potential value of soil erosion. S402. Input the monitoring sample set into the pore failure layer of the model, calculate the rate of change of the fractal dimension of the soil pore structure and the pore connectivity, and combine the interface instability potential value to obtain the pore failure risk value. S403. Input the monitoring sample set into the fracture nucleation layer of the model, calculate the soil micro-fracture propagation rate, connectivity probability and nucleation density, and obtain the fracture nucleation trigger value by combining the pore failure risk value. S404. Input the monitoring sample set into the erosion evolution layer of the model, calculate the slope erosion resistance attenuation coefficient and the external driving force intensity, and combine the output results of the above three layers to predict the occurrence range, erosion modulus, transport path and degree of harm of soil erosion in the next 168 hours, and obtain the full-time risk evolution sequence. The prediction time can be dynamically adjusted according to the control needs. S405. The output results of the above four layers are nonlinearly weighted and fused to obtain the advanced soil and water loss risk assessment value in the whole time domain.

6. The method for early warning and control of soil erosion based on the Internet of Things according to claim 5, characterized in that, Step S5 specifically includes: S501. Obtain the ecological protection red line level and soil and water conservation control level of the target control area, and determine the risk weight correction coefficient and early warning threshold adjustment coefficient of the corresponding level. S502. Multiply the full-time domain advanced soil and water loss risk assessment value by the corresponding correction coefficient in sequence to obtain the soil and water loss control early warning judgment value. S503. Compare the early warning judgment value with the preset five-level graded early warning threshold range. If it exceeds the upper limit of the threshold range, it is judged as an advanced risk event of soil and water loss of the corresponding level, triggering ecological-engineering collaborative closed-loop intervention command and graded early warning response at the matching control level; if it does not exceed the threshold, it is judged as a stable soil state, maintaining a low-power passive normal monitoring mode. The ecological-engineering collaborative closed-loop intervention command and graded early warning response include pushing risk evolution information to the corresponding level of soil and water conservation management departments, initiating slope capillary water control engineering commands, implementing in-situ remediation and treatment plans for underlying biological crusts, linking on-site passive sound and light early warning equipment, and triggering emergency soil stabilization engineering activation plans.

7. A water and soil erosion control and early warning system based on the Internet of Things, characterized in that, The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the Internet of Things-based soil erosion control and early warning method according to any one of claims 1 to 6. The system comprises: The passive distributed sensing module is used to acquire multi-dimensional monitoring source data covering the target control area, including soil interface electrochemistry, soil microstructure deformation, pore hydrological fractal, surface erosion resistance ecology, and external force-driven data, based on a distributed IoT passive sensing architecture. The dual precursor coupling identification module is used to extract the precursor signals of interface instability in the early stage of soil erosion and the precursor signals of soil fissure nucleation from multi-dimensional monitoring source data, and generate the chaotic phase space adaptive monitoring cycle control command based on the coupling of the dual precursor signals. The monitoring sequence reconstruction module is used to dynamically reconstruct the acquisition parameters of multi-dimensional monitoring source data according to the adaptive monitoring cycle adjustment command, and obtain a full-chain continuous monitoring sample set with spatiotemporal registration of multiple batches. The cross-scale risk extrapolation module is used to input the monitoring sample set into the preset fourth-order multi-scale coupled extrapolation model of soil and water loss, complete the cross-scale coupled analysis, and obtain the advanced soil and water loss risk assessment value in the whole time domain. The tiered early warning and control module is used to generate early warning judgment values ​​based on the advanced water and soil erosion risk assessment values ​​across the entire time domain and the control level of the target control area. Based on the comparison results, closed-loop control intervention or routine monitoring is implemented.

8. The Internet of Things-based soil and water conservation early warning system according to claim 7, characterized in that, The passive distributed sensing module includes: The access point management unit is used to acquire standardized access points for multiple types of passive IoT sensing nodes within the target control area, which correspond one-to-one with the monitoring dimensions of soil interface electrochemistry, soil microstructure deformation, pore hydrological fractal, surface erosion resistance ecology, and external force-driven monitoring. The data synchronization acquisition unit is used to synchronously acquire multiple sets of raw passive monitoring data through various standardized access points; The feature classification and calibration unit is used to extract the unique features of the physical field corresponding to each group of raw passive monitoring data, classify and calibrate the monitoring data according to the feature dimensions, and output multi-dimensional monitoring source data.

9. The Internet of Things-based soil and water conservation and early warning system according to claim 7, characterized in that, The dual-precursor coupling identification module includes: The first precursor calculation unit is used to calculate the corresponding characteristic parameters based on the relaxation electrochemical characteristic data of the soil mineral-microorganism interface, and to complete the identification and judgment of the precursor signals of interface instability in the early stage. The second precursor calculation unit is used to calculate the corresponding characteristic parameters based on the infrasound-acoustic emission coupling characteristic data of soil fracture nucleation, and to complete the identification and judgment of the precursor signal of fracture nucleation. The control command generation unit is used to combine the spatial distribution and intensity level of the two precursor signals and generate adaptive monitoring cycle control commands for the corresponding control zone through a chaotic phase space reconstruction algorithm.

10. The Internet of Things-based soil erosion control and early warning system according to claim 7, characterized in that, The cross-scale risk simulation module includes: The interface instability analysis unit is used to perform calculation and analysis of the interface instability layer of the fourth-order multi-scale coupled deduction model and output the instability potential value of the soil erosion interface. The pore failure analysis unit is used to perform calculation and analysis of the pore failure layer of the model, and outputs the pore failure risk value by combining the interface instability potential value. The fracture nucleation analysis unit is used to perform calculation and analysis of the fracture nucleation layer in the model, and outputs the fracture nucleation trigger value by combining the pore failure risk value. The erosion evolution analysis unit is used to perform calculation and analysis of the erosion evolution layer of the model, and to deduce the full-time-domain risk evolution sequence by combining the output results of the aforementioned three layers. The risk fusion assessment unit is used to perform nonlinear weighted fusion of the output results of each layer of the model to output the advanced soil and water loss risk assessment value in the whole time domain.

Images