A method for monitoring and predicting soil erosion of sloping farmland

Abstract

The application discloses a kind of water and soil loss monitoring and prediction method for slope farmland, it is related to slope farmland water and soil measurement technical field, by a variety of equipment collection including basic parameter set RAW of soil, meteorological and vegetation information, by data processing obtains resolving parameter set RDS, provides comprehensive data basis for accurate prediction.Based on resolving parameter set RDS extraction slope farmland feature vector SVL, reveal the inherent antagonistic mechanism of erosion process.According to slope farmland feature vector SVL, build slope farmland instability mode calculation model, solve the problem that cannot distinguish erosion mode, then compared with erosion index threshold interval vector Pth, generate predicted erosion response scheme AIR, realize the transformation from risk prediction to accurate decision-making.Finally, introduce the measured real-time rainfall kinetic energy flux factor Ek (t) during rainfall Dynamic check, generate the final monitoring erosion response scheme AIL, build closed-loop management and control system from before prediction to in-process correction.

Description

Technical Field

[0001] This invention relates to the field of soil and water conservation technology for sloping farmland, specifically a method for monitoring and predicting soil and water loss on sloping farmland. Background Technology

[0002] In today's world, the fields of environmental monitoring and agricultural technology are developing rapidly, working together to achieve sustainable resource utilization and harmonious coexistence with the ecological environment. Within this vast field, soil and water conservation, as a crucial branch, remains a core issue concerning land productivity, water resource security, and even regional ecological health. Soil and water conservation technologies face diverse challenges, with soil erosion being particularly prominent and complex for sloping farmland, a vital component of agricultural resources. Due to their unique sloping topography, sloping farmland is highly susceptible to soil particle stripping and migration under the hydraulic forces of rainfall or irrigation. This process is not merely simple soil erosion, but a dynamic and variable physical phenomenon. Its specific patterns and intensity profoundly impact agricultural production planning, land resource management, and the prevention of ecological disasters. Therefore, how to scientifically and accurately understand and address the soil erosion process of sloping farmland under specific rainfall events has become a key technical challenge to be solved in the fields of modern precision agriculture and smart soil and water conservation.

[0003] However, existing management techniques for soil erosion on sloping farmland generally have significant shortcomings and deficiencies in addressing the aforementioned challenges. Traditional monitoring methods mostly rely on on-site surveys after heavy rainfall, fixed-point observations of runoff plots, or post-disaster interpretation of remote sensing images. These methods are essentially "after-the-fact" assessments, generally exhibiting severe lag and failing to provide any early warning window for the implementation of preventative measures. Regarding forecasting, while some classic empirical or statistical models exist, such as the general soil erosion equation, these models are mostly based on long-term, macroscopic averages, aiming to assess annual average losses. Therefore, their forecast accuracy is coarse and lacks timeliness, completely unable to provide precise, quantitative risk forecasts for specific rainfall events occurring in the next few hours or days. More importantly, these traditional methods typically only provide a general estimate of total soil loss, failing to differentiate the specific physical patterns of erosion—whether it's large-scale, uniform sheet erosion or highly destructive localized gully erosion. The prevention and control measures for these two patterns are drastically different, often leaving farm managers in a dilemma when making decisions.

[0004] This technological deficiency leads to a series of negative consequences in actual production. For example, when meteorological departments issue rainstorm warnings, managers of sloping farmland, lacking scientific and precise decision-making basis, can only rely on personal experience for "blind guessing" protection. If they choose to invest heavily in comprehensive coverage to prevent sheet erosion, but what actually occurs is concentrated gully erosion, not only are the protection costs essentially ineffective, but the best opportunity to prevent the real disaster is also missed due to incorrect response measures. Conversely, if they choose to dig diversion channels to prevent gully erosion, but sheet erosion occurs instead, it will also result in a waste of resources and damage to the land. This blind decision-making directly leads to a huge waste of protection resources and a continuous decline in land productivity. In the long run, this will not only cause the loss of topsoil and expensive fertilizers, directly affecting crop yields, but the silt washed away will also silt up river channels, pollute downstream water bodies, and in extreme cases, even induce geological disasters such as flash floods or mudslides in small watersheds, seriously threatening the ecological security of the region. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for monitoring and predicting soil erosion on sloping farmland, solving the problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides a method for monitoring and predicting soil erosion on sloping farmland, comprising the following steps:

[0007] S1. Collect the basic parameter set RAW of the sloping farmland area through various types of sensors, API interfaces and image acquisition devices, and process the basic dataset RAW to obtain the solution parameter set RDS.

[0008] S2. Based on the solution parameter set RDS, the core features of the sloping farmland area are extracted to obtain the soil erosion resistance Rs and hydrodynamic stress Sh, and the sloping farmland feature vector SVL is generated.

[0009] S3. Based on the sloping farmland feature vector SVL and the solution parameter set RDS, construct a model for predicting the instability mode of sloping farmland, and generate the predicted erosion index vector PML by generating the sheet erosion prediction index Ms and the gully erosion prediction index Mr respectively.

[0010] S4. Based on the comparison between the predicted erosion index vector PML and the erosion index threshold interval vector Pth, a predicted erosion response plan AIR is generated.

[0011] S5. The real-time rainfall kinetic flux factor Ek(t) is recalculated by analyzing the two-dimensional laser raindrop spectrometer to obtain the real-time cumulative monitoring erosion index vector PNL(t) and compared with the erosion index threshold interval vector Pth. Based on the predicted erosion response plan AIR, adjustments are made to generate the monitoring erosion response plan AIL.

[0012] Preferably, S1 includes S11 and S12.

[0013] S11. Collect basic parameter sets (RAW) of sloping farmland areas through various types of sensors, API interfaces, and image acquisition devices;

[0014] The parameters collected in the basic parameter set RAW by various types of sensors include soil acoustic impedance Zs collected by an acoustic probe, soil micropore connectivity kp collected by a conductivity matrix sensor, soil organic carbon content Cooc collected by a near-infrared spectral probe, and soil moisture content hw collected by a time domain reflectometer probe.

[0015] The parameters collected in the basic parameter set RAW via the API interface include the predicted rainfall intensity time series I(t) for the next 24 hours obtained from the meteorological platform via the meteorological platform API interface, where t represents the time interval sequence number, specifically in 0.5-hour intervals;

[0016] The parameters collected by the image acquisition equipment in the basic parameter set RAW include high-density three-dimensional point cloud data las scanned by UAV LiDAR of sloping farmland area, multi-channel raw images tiff of vegetation in sloping farmland area captured by multispectral camera, and high-definition raw images jpg of root system in sloping farmland area captured by miniature root window camera.

[0017] Preferably, in step S12, the D8 algorithm is used to perform terrain analysis on the high-density three-dimensional point cloud data las to obtain the micro-domain runoff concentration γf.

[0018] Normalized Difference Vegetation Index (NDVI) is calculated for multi-channel raw images using TIFF. Based on the NDVI and the pixel binary model, Leaf Area Index (Lai) is obtained.

[0019] For high-resolution raw root system images (jpg), the UNet model and EnglightenGAN algorithm are used for image segmentation and length statistics to obtain the root density ρ.

[0020] For the predicted rainfall intensity time series I(t) in the next 24 hours, the erosivity power law estimation model based on rainfall intensity is used to estimate the erosivity at each time interval t in the next 24 hours. The predicted rainfall erosivity Rk in the next 24 hours is obtained by summing the values ​​and comparing it with the preset benchmark rainfall erosivity Rth to obtain the predicted rainfall erosion factor R.

[0021] Based on the data processing performed on the basic dataset RAW as described above, the solution parameter set RDS=[Zs, kp, Coc, hw, γf, Lai, ρ, R] is obtained.

[0022] Preferably, S2 includes S21 and S22;

[0023] S21. Based on the solution parameter set RDS, the core features of the sloping farmland area are extracted. According to the soil organic carbon content Cooc as the organic matter basis, the hyperbolic secant function sech is used for the soil acoustic impedance Zs and soil micropore connectivity kp, and the S-shaped saturation curve is used for the root density ρ. The comprehensive soil erosion resistance algorithm is constructed to calculate the soil erosion resistance Rs and quantify the comprehensive erosion resistance of the sloping farmland as an ecosystem.

[0024] The expression for the comprehensive soil erosion resistance algorithm is as follows:

[0025] ;

[0026] In the formula, ln represents the natural logarithm function with the constant e as the base, Zo represents the physical constant used to normalize the soil acoustic impedance Zs, specifically the acoustic impedance value of pure water, α represents the preset soil geological adjustment coefficient, and Km represents the preset half-saturation constant of the root system soil stabilization effect.

[0027] Preferably, in step S22, based on the predicted rainfall erosivity factor R and leaf area index Lai in the next 24 hours as the core drivers of rainfall event erosion on sloping farmland, the ratio of soil moisture content hw to the preset soil saturation moisture content hs is calculated. An exponential saturation model is used for micro-domain runoff concentration γf to construct a hydrodynamic integrated stress algorithm, calculate the hydrodynamic stress Sh, and quantify the erosive capacity of rainfall events in the next 24 hours on the soil.

[0028] The expression for the hydrodynamic integrated stress algorithm is as follows:

[0029] ;

[0030] In the formula, ζ represents the canopy interception coefficient preset according to the vegetation type, β represents the preset surface runoff effect coefficient, and exp represents the natural exponential function;

[0031] Based on soil erosion resistance Rs and hydrodynamic stress Sh, the feature vector SVL=[Rs, Sh] of sloping farmland is generated.

[0032] Preferably, S3 includes S31 and S32;

[0033] S31. Based on the soil erosion resistance Rs and hydrodynamic stress Sh in the feature vector SVL of sloping farmland and the micro-domain runoff concentration γf in the solution parameter set RDS, a model for predicting the instability mode of sloping farmland is constructed. The model for predicting the instability mode of sloping farmland includes a sheet erosion model and a gully erosion model.

[0034] Based on the Weber distribution formula, this paper introduces the micro-domain runoff concentration degree γf in the soil erosion resistance Rs and hydrodynamic stress Sh in the feature vector SVL of sloping farmland and the micro-domain runoff concentration degree γf in the solution parameter set RDS. It uses the natural exponential function exp to represent the inhibitory effect of runoff concentration on sheet erosion, constructs a sheet erosion prediction model, obtains the sheet erosion prediction index Ms, and quantifies the probability of sheet erosion pattern in soil and water loss in sloping farmland.

[0035] The formula for the plaque erosion estimation model is as follows:

[0036] ;

[0037] In the formula, σ represents the preset Weber shape parameter, and λ represents the preset confluence suppression effect coefficient.

[0038] Preferably, in S32, based on the Hill-Langmuir equation, a gully erosion estimation model is constructed for the soil erosion resistance Rs and hydrodynamic stress Sh in the feature vector SVL of sloping farmland and the micro-domain runoff concentration γf in the solution parameter set RDS, to obtain the gully erosion prediction index Mr, and to quantify the probability of gully erosion patterns in soil and water loss in sloping farmland.

[0039] The formula for the estimation model of gully erosion is as follows:

[0040] ;

[0041] In the formula, ω represents the preset bus energy amplification factor, η represents the preset Hill coefficient, and Kr represents the preset half-saturation constant;

[0042] Based on the sheet erosion prediction index Ms and the gully erosion prediction index Mr, a prediction erosion index vector PML=[Ms, Mr] is generated.

[0043] Preferably, S4 includes S41 and S42;

[0044] S41. Based on the comparison between the predicted erosion index vector PML and the erosion index threshold interval vector Pth, a predicted erosion response plan AIR is generated.

[0045] Among them, the erosion index threshold interval vector Pth includes the sheet erosion index threshold interval Ps and the gully erosion index threshold interval Pr. The sheet erosion index threshold interval Ps includes the first-level sheet erosion index threshold Ps1, the second-level sheet erosion index threshold Ps2 and the third-level sheet erosion index threshold Ps3. The gully erosion index threshold interval Pr includes the first-level gully erosion index threshold Pr1, the second-level gully erosion index threshold Pr2 and the third-level gully erosion index threshold Pr3. The predicted erosion response scheme AIR includes the predicted sheet erosion response scheme AIRs and the predicted gully erosion response scheme AIRr.

[0046] For the sheet erosion prediction index Ms, if the sheet erosion prediction index Ms < the secondary sheet erosion index threshold Ps2, the predicted sheet erosion response plan AIRs is no intervention.

[0047] If the threshold of the secondary sheet erosion index Ps2 ≤ the sheet erosion prediction index Ms < the threshold of the tertiary sheet erosion index Ps3, the predicted sheet erosion response plan AIRs is the primary sheet erosion treatment plan: use organic materials such as straw, leaves and hay to spread evenly on the surface of the sloping farmland soil.

[0048] If the sheet erosion prediction index Ms is greater than or equal to the third-level sheet erosion index threshold Ps3, the predicted sheet erosion response plan AIRs is a second-level sheet erosion treatment plan: based on the first-level sheet erosion treatment plan, cover the planting area of ​​the sloping farmland with waterproof cloth and use heavy objects such as stones or sandbags to compact the edges.

[0049] Preferably, in S42, for the gully erosion prediction index Mr, if the gully erosion prediction index Mr < the secondary gully erosion index threshold Pr2, the gully erosion response plan AIRr is to not intervene.

[0050] If the threshold of the secondary gully erosion index Pr2 ≤ the gully erosion prediction index Ms < the threshold of the tertiary gully erosion index Pr3, the predicted gully erosion response plan AIRr is the primary gully erosion treatment plan: On the slope, place multiple rows of obstacles in the horizontal direction. The obstacles are composed of sandbags, stones, logs or tightly bundled hay bales.

[0051] If the gully erosion prediction index Ms ≥ the third-level gully erosion index threshold Pr3, the predicted gully erosion response plan AIRr is a second-level gully erosion treatment plan: based on the first-level treatment plan, dig a temporary irrigation canal parallel to the horizontal line at the highest point of the sloping farmland, and reinforce the temporary irrigation canal and the existing drainage ditch by laying gravel on both sides and at the bottom.

[0052] Preferably, S5 includes S51;

[0053] S51. When rainfall begins, a two-dimensional laser raindrop spectrometer is activated to collect raindrop data pairs RDP=[d_(i,t), v_(i,t)], where d_(i,t) represents the diameter of the i-th raindrop at time interval sequence number t, v_(i,t) represents the velocity of the i-th raindrop at time interval sequence number t, and the total number of raindrops sampled by the two-dimensional laser raindrop spectrometer is Ndr. Based on the kinetic flux calculation formula, the real-time rainfall kinetic flux E(t) is obtained. The real-time rainfall kinetic flux E(t) is compared with the preset baseline rainfall kinetic flux. The real-time rainfall kinetic flux factor Ek(t) is obtained by comparing Eth. The predictive rainfall erosivity factor R is replaced by the real-time rainfall kinetic flux factor Ek(t) and recalculated from step S2. The calculation results are accumulated in real time according to the time interval sequence number t to obtain the real-time cumulative monitoring erosion index vector PNL(t) and compared with the erosion index threshold interval vector Pth. The real-time cumulative monitoring erosion index vector PNL(t) includes the real-time monitoring index Ns(t) for sheet erosion and the real-time monitoring index Nr(t) for gully erosion.

[0054] For the real-time monitoring index Ns(t) of sheet erosion, when comparing it with the erosion index threshold interval vector Pth, the trigger sensitivity is improved. The first-level sheet erosion index threshold Ps1 replaces the second-level sheet erosion index threshold Ps2, and the second-level sheet erosion index threshold Ps2 replaces the third-level sheet erosion index threshold Ps3.

[0055] For the real-time monitoring index Nr(t) of gully erosion, when comparing it with the erosion index threshold interval vector Pth, the trigger sensitivity is increased, and the first-level gully erosion index threshold Pr1 replaces the second-level gully erosion index threshold Pr2, and the second-level gully erosion index threshold Pr2 replaces the third-level gully erosion index threshold Pr3.

[0056] Based on the comparison results after increasing the trigger sensitivity, adjustments were made to the predictive erosion response plan AIR.

[0057] If the predicted sheet erosion response plan AIRs and the predicted gully erosion response plan AIRr are upgraded from no intervention to Level 1 sheet erosion treatment plan and Level 1 gully erosion treatment plan or above, or upgraded from Level 1 sheet erosion treatment plan and Level 1 gully erosion treatment plan to Level 2 sheet erosion treatment plan and Level 2 gully erosion treatment plan, then the monitoring erosion response plan AIL is generated based on the upgraded treatment plan.

[0058] Conversely, the generated monitoring erosion response plan AIL is the current predicted sheet erosion response plan AIRs and the current predicted gully erosion response plan AIRr, without any adjustments.

[0059] This invention provides a method for monitoring and predicting soil erosion on sloping farmland, which has the following beneficial effects:

[0060] (1) By quantitatively modeling soil erosion resistance Rs and hydrodynamic stress Sh, the dynamic antagonistic relationship between the intrinsic resistance and external stress faced by sloping farmland is profoundly revealed. On this basis, the differential prediction of sheet erosion prediction index Ms and gully erosion prediction index Mr is creatively realized, solving the fundamental problem that traditional technology cannot distinguish erosion patterns. Furthermore, the predicted erosion index vector PML containing Ms and Mr is intelligently compared with the multi-level erosion index threshold interval vector Pth to automatically generate a scientific and accurate predicted erosion response plan AIR, transforming the predicted information into actual productivity with high economic benefits. Finally, by introducing the measured real-time rainfall kinetic flux factor Ek(t) during rainfall for monitoring, the purpose of dynamic correction is achieved, generating the monitored erosion response plan AIL, which constitutes a complete technical closed loop from prediction to action to adaptive adjustment, providing a solid and reliable technical guarantee for achieving cost reduction and efficiency improvement in agricultural production and long-term sustainable use of land resources.

[0061] (2) By using various types of data acquisition equipment to conduct a multi-source data acquisition system, the basic parameter set RAW was comprehensively acquired. After rigorous data processing, a solution parameter set RDS was generated, which includes key information such as soil acoustic impedance Zs, micro-domain runoff concentration γf, leaf area index Lai, and predictive rainfall erosivity factor R based on API prediction. Based on this, these discrete parameters were condensed into two core features, soil erosion resistance Rs representing "defense capability" and hydrodynamic stress Sh representing "attack capability", through our constructed soil comprehensive erosion resistance algorithm and hydrodynamic comprehensive stress algorithm, and the sloping farmland feature vector SVL was generated. On this basis, a differentiated probability model based on the Weiber distribution and the Hill-Langmuir equation was further used to realize the quantitative prediction of sheet erosion prediction index Ms and gully erosion prediction index Mr. This series of processing elevated the prediction from a simple total estimation to a patterned prediction with tactical guidance significance reflected in the final predicted erosion index vector PML.

[0062] (3) By comparing the sheet erosion prediction index Ms and gully erosion prediction index Mr in the predicted erosion index vector PML with an erosion index threshold interval vector Pth containing multiple threshold levels, a clear and explicit predicted erosion response plan AIR is generated. This completely changes the previous predicament of blindly making decisions based on experience, enabling managers to accurately allocate the most appropriate protection resources based on the predicted main erosion patterns and risk levels, thereby maximizing protection benefits and minimizing resource costs. Furthermore, by using a two-dimensional laser raindrop spectrometer to collect raindrop data for RDP during rainfall, the real real-time rainfall kinetic flux factor Ek(t) is accurately calculated and used to replace the predicted rainfall erosion force factor R for real-time verification, resulting in a dynamic real-time cumulative monitoring erosion index vector PNL(t). Based on preset rules, such as replacing the secondary sheet erosion index threshold Ps2 with the primary sheet erosion index threshold Ps1 and replacing the tertiary sheet erosion index threshold Ps3 with the secondary sheet erosion index threshold Ps2 during monitoring, the system intelligently improves the trigger sensitivity of the early warning and adjusts and generates the final monitoring erosion response plan AIL based on this, ensuring that the most timely and reliable risk management can be carried out in all situations. Attached Figure Description

[0063] Figure 1 This is a schematic diagram of the steps of a method for monitoring and predicting soil and water loss on sloping farmland according to the present invention;

[0064] Figure 2 This is a schematic diagram of the data flow of a method for monitoring and predicting soil erosion on sloping farmland according to the present invention.

[0065] Figure 3 This is a schematic diagram of the correlation curve between the sheet / groove erosion estimation model of the present invention and the hydrodynamic stress Sh. Detailed Implementation

[0066] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0067] Example 1

[0068] This invention provides a method for monitoring and predicting soil erosion on sloping farmland. Please refer to [link / reference]. Figure 1 This includes the following steps:

[0069] S1. Collect the basic parameter set RAW of the sloping farmland area through various types of sensors, API interfaces and image acquisition devices, and process the basic dataset RAW to obtain the solution parameter set RDS.

[0070] S2. Based on the solution parameter set RDS, the core features of the sloping farmland area are extracted to obtain the soil erosion resistance Rs and hydrodynamic stress Sh, and the sloping farmland feature vector SVL is generated.

[0071] S3. Based on the sloping farmland feature vector SVL and the solution parameter set RDS, construct a model for predicting the instability mode of sloping farmland, and generate the predicted erosion index vector PML by generating the sheet erosion prediction index Ms and the gully erosion prediction index Mr respectively.

[0072] S4. Based on the comparison between the predicted erosion index vector PML and the erosion index threshold interval vector Pth, a predicted erosion response plan AIR is generated.

[0073] S5. The real-time rainfall kinetic flux factor Ek(t) is recalculated by analyzing the two-dimensional laser raindrop spectrometer to obtain the real-time cumulative monitoring erosion index vector PNL(t) and compared with the erosion index threshold interval vector Pth. Based on the predicted erosion response plan AIR, adjustments are made to generate the monitoring erosion response plan AIL.

[0074] In this embodiment, a comprehensive set of basic parameters, RAW, is processed to obtain a set of computational parameters, RDS. This process can be completed before rainfall events occur, overcoming the significant lag in traditional monitoring methods. Based on this data, soil erosion resistance (Rs) and hydrodynamic stress (Sh), which characterize the intrinsic antagonistic relationship of the land surface, are extracted. Based on these, the sheet erosion prediction index Ms and gully erosion prediction index Mr are creatively calculated. The acquisition of this pair of differentiated indices solves the problem of being unable to distinguish erosion patterns, ensuring that the subsequently generated predictive erosion response plan AIR is no longer a blind, general protection measure, but a highly targeted tactical deployment, thus avoiding resource waste and prevention failure. Furthermore, a crucial closed-loop correction mechanism is included: when rainfall actually occurs, the system introduces the measured rainfall kinetic flux Ek(t) to dynamically verify the risk and adjust the original response plan, ultimately forming the monitoring erosion response plan AIL. This complete process from pre-event prediction to in-event correction ensures that the entire risk management is both forward-looking and highly reliable.

[0075] Example 2

[0076] This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 and Figure 2 Specifically: S1 includes S11 and S12;

[0077] S11. Collect basic parameter sets (RAW) of sloping farmland areas through various types of sensors, API interfaces, and image acquisition devices;

[0078] The parameters collected in the basic parameter set RAW by various types of sensors include soil acoustic impedance Zs collected by an acoustic probe, soil micropore connectivity kp collected by a conductivity matrix sensor, soil organic carbon content Cooc collected by a near-infrared spectral probe, and soil moisture content hw collected by a time domain reflectometer probe.

[0079] The parameters collected in the basic parameter set RAW via the API interface include the predicted rainfall intensity time series I(t) for the next 24 hours obtained from the meteorological platform via the meteorological platform API interface, where t represents the time interval number, specifically in 0.5-hour intervals;

[0080] The parameters collected by the image acquisition equipment in the basic parameter set RAW include high-density three-dimensional point cloud data las scanned by UAV LiDAR in sloping farmland area, multi-channel raw images tiff of vegetation in sloping farmland area captured by multispectral camera, and high-definition raw images jpg of root system in sloping farmland area captured by miniature root window camera.

[0081] S12. Use the D8 algorithm to perform terrain analysis on the high-density three-dimensional point cloud data las to obtain the micro-domain runoff concentration γf.

[0082] Normalized Difference Vegetation Index (NDVI) is calculated for multi-channel raw images using TIFF. Based on the NDVI and the pixel binary model, Leaf Area Index (Lai) is obtained.

[0083] For high-resolution raw root system images (jpg), the UNet model and EnglightenGAN algorithm are used for image segmentation and length statistics to obtain the root density ρ.

[0084] For the predicted rainfall intensity time series I(t) in the next 24 hours, the erosivity power law estimation model based on rainfall intensity is used to estimate the erosivity at each time interval t in the next 24 hours. The predicted rainfall erosivity Rk in the next 24 hours is obtained by summing the values ​​and comparing it with the preset benchmark rainfall erosivity Rth to obtain the predicted rainfall erosion factor R.

[0085] The power-law estimation model for erosion force is expressed as follows:

[0086] ;

[0087] ;

[0088] In the formula, N represents the total number of time steps in 24 hours, a and b represent the regional rainfall erosivity model parameters obtained by relevant personnel through statistical analysis of long-term rainfall observation data, and ∑ represents cumulative calculation;

[0089] The purpose of this formula is to obtain the predicted rainfall intensity time series I(t), and then apply a power-law estimation model of rainfall intensity based on long-term observation data. This model can convert the predicted rainfall intensity time series I(t) for the next 24 hours into the magnitude of erosivity for that period. Subsequently, by summing the erosivity for all time steps N within the next 24 hours, a single index that comprehensively reflects the total erosivity potential of the entire predicted rainfall event is finally obtained, namely the predictive rainfall erosivity factor R.

[0090] Based on the data processing performed on the basic dataset RAW as described above, the solution parameter set RDS=[Zs, kp, Coc, hw, γf, Lai, ρ, R] is obtained.

[0091] In this embodiment, a multi-source heterogeneous acquisition system using various types of sensors, API interfaces, and image acquisition devices constructs an unprecedented basic parameter set, RAW. This design transcends the limitations of traditional methods that rely on a single or a few parameters, enabling a comprehensive and three-dimensional "digital portrait" of sloping farmland from multiple dimensions, including soil structure, vegetation physiology, micro-topography, and future weather patterns. This provides a highly faithful and complete data foundation for all subsequent analyses. Furthermore, the unique aspect of this method lies in its approach: it does not simply list raw data. Instead, it introduces a series of advanced computational models, such as the D8 algorithm for processing 3D point cloud data, the UNet and EnlightenGAN algorithms for processing high-resolution root system images, and a power-law estimation model for meteorological data. These models accurately transform abstract raw signals into a computational parameter set, RDS, with clear physical meaning. This intelligent computational process from raw signals to physical parameters significantly improves the quality and usability of the input data, ensuring the scientific rigor and reliability of subsequent core feature extraction.

[0092] Example 3

[0093] This embodiment is an explanation based on Embodiment 2. Please refer to it. Figure 1 and Figure 2 Specifically: S2 includes S21 and S22;

[0094] S21. Based on the solution parameter set RDS, the core features of the sloping farmland area are extracted. According to the soil organic carbon content Cooc as the organic matter basis, the hyperbolic secant function sech is used for the soil acoustic impedance Zs and soil micropore connectivity kp, and the S-shaped saturation curve is used for the root density ρ. The comprehensive soil erosion resistance algorithm is constructed to calculate the soil erosion resistance Rs and quantify the comprehensive erosion resistance of the sloping farmland as an ecosystem.

[0095] The expression for the comprehensive soil erosion resistance algorithm is as follows:

[0096] ;

[0097] In the formula, ln represents the natural logarithm function with base e, Zo represents the physical constant used to normalize the soil acoustic impedance Zs, specifically the acoustic impedance value of pure water, and α represents the preset soil geological adjustment coefficient. Specifically, undisturbed soil samples are collected from sloping farmland areas by researchers in this field and returned to the laboratory to prepare various test samples with different soil acoustic impedances Zs and soil micropore connectivity kp. The shear strength of the undisturbed soil samples is measured using a soil direct shear tester. The soil acoustic impedance Zs and soil micropore connectivity kp of the various test samples are then substituted into the hyperbolic secant function sech in the soil comprehensive erosion resistance algorithm for nonlinear least-multiplication fitting, so that the result of the hyperbolic secant function sech is transformed into multiple soil acoustic impedances Zs and soil micropore connectivity kp. The soil shear strength of the undisturbed soil sample was measured by the corresponding soil direct shear apparatus and converged. The soil geological adjustment coefficient α was calculated in reverse. Km represents the preset half-saturation constant of the root soil stabilization effect. Specifically, the staff of this field conducted a root pull-out experiment on the "root-soil complex" with different root densities ρ. The maximum pulling force required to pull the plant and its roots out of the soil per unit area was measured to obtain the "root density ρ-maximum pulling force" data pair. The root density ρ in "root density ρ-maximum pulling force" was used as the independent variable and the maximum pulling force was used as the dependent variable to draw a scatter plot and obtain an experimental S-shaped growth curve. The root density ρ in the soil comprehensive erosion resistance algorithm was fitted with the experimental S-shaped growth curve using the S-shaped saturation curve to obtain the half-saturation constant Km of the root soil stabilization effect.

[0098] The purpose of this formula is to regard soil organic carbon content (Coc) as the foundation of soil health and structural stability, and to construct an organic matter-based enhancing factor. Using the organic matter enhancement factor as the fundamental multiplier in the entire formula, it directly reflects the linear enhancement effect of soil fertility and aggregate quality on erosion resistance. The physical structure stability of soil is a bounded optimal range problem—excessive looseness (low soil acoustic impedance Zs, high soil micropore connectivity kp) or excessive compaction (high soil acoustic impedance Zs, low soil micropore connectivity kp) are both detrimental to erosion resistance. Therefore, the hyperbolic secant function sech is chosen to simulate this characteristic. This function has a maximum value of 1 and decreases towards both sides, effectively characterizing the optimal nonlinear relationship in this moderate state. Its internal independent variables... The relative relationship between soil acoustic impedance Zs (representing compaction) and soil micropore connectivity kp (representing permeability) was established, and a root network anchoring factor was constructed based on the Hill equation describing the synergistic effect. The nonlinear reinforcement effect of root density ρ on soil was simulated. When the root system is sparse, its soil-fixing effect increases rapidly with the increase of density; however, when the root density ρ reaches a certain level, its marginal reinforcement effect will gradually weaken and tend to saturate. This saturation curve reflects the soil-fixing and slope protection mechanism of the root network.

[0099] S22. Based on the predicted rainfall erosivity factor R and leaf area index Lai in the next 24 hours as the core drivers of rainfall event erosion on sloping farmland, the ratio of soil moisture content hw to the preset soil saturation moisture content hs is calculated. An exponential saturation model is used for micro-domain runoff concentration γf to construct a hydrodynamic integrated stress algorithm, calculate the hydrodynamic stress Sh, and quantify the erosive capacity of rainfall events in the next 24 hours on the soil.

[0100] The expression for the hydrodynamic integrated stress algorithm is as follows:

[0101] ;

[0102] In the formula, ζ represents the canopy interception coefficient preset according to the vegetation type. Specifically, researchers in this field construct an artificial rainfall simulation device in the laboratory, and evenly distribute a large number of rain collectors below the crop canopy to collect the through-rainfall that passes through the canopy and finally reaches the ground surface. Through multiple sets of experiments, the "leaf area index Lai-canopy reduction rate" data pair is obtained, where the canopy reduction rate is the difference between the total rainfall and the through-rainfall divided by the total rainfall. The "leaf area index Lai-canopy reduction rate" data pair is then compared with the hydrodynamic integrated stress algorithm. Nonlinear curve fitting was performed to calculate the canopy interception coefficient ζ, which is preset according to the vegetation type. β represents the preset surface runoff effect coefficient, and exp represents the natural exponential function. Specifically, researchers in the field constructed multiple standard-sized runoff plots in the laboratory and collected micro-domain runoff convergence γf at different locations. Artificial rainfall simulation devices were used to induce rainfall in the runoff plots, and the peak flow and velocity of the runoff were accurately measured at the outlet using a flow meter and a flow meter. The runoff amplification factor was obtained by calculating the ratio of the peak flow at different locations to the average peak flow on flat surfaces. Multiple experiments were conducted to obtain multiple sets of "micro-domain runoff convergence γf - runoff amplification factor" data pairs, which were then incorporated into the hydrodynamic integrated stress algorithm. Nonlinear curve fitting was performed to calculate the surface runoff effect coefficient β;

[0103] The purpose of this formula is to construct a net rainfall erosivity factor based on the predicted rainfall erosivity factor R and leaf area index Lai for the next 24 hours. This study simulates the reduction of raindrop energy during penetration of the vegetation canopy due to interception and buffering by the leaf area index Lai, representing the net rainfall erosion force that ultimately reaches the land surface. When this net rainfall erosion force acts on the surface, its destructive effect is influenced by micro-topography. A topographic amplification factor is constructed using an exponential saturation model. Simulating this process, we find that the greater the micro-domain runoff concentration γf, the stronger the amplification effect on erosive forces. However, this amplification effect has an upper limit and tends to saturate. Ultimately, the soil's antecedent moisture level determines its sensitivity to erosive forces. Therefore, the ratio of the current soil moisture content hw to the soil saturation moisture content hs is used as an antecedent moisture content sensitivity factor. The closer the soil moisture content is to saturation, the faster the runoff occurs, the more sensitive the response to erosion, and the stronger the overall stress.

[0104] Based on soil erosion resistance Rs and hydrodynamic stress Sh, the feature vector SVL=[Rs, Sh] of sloping farmland is generated.

[0105] In this embodiment, multidimensional physical parameters, including soil acoustic impedance Zs, root density ρ, predicted rainfall erosivity factor R, and leaf area index Lai, are received from the solution parameter set RDS, and their core features are extracted. The unique aspect of this process is that it is not a simple combination of parameters, but rather a creative integration and dimensionality enhancement of these parameters through two deeply coupled nonlinear algorithms based on physical mechanisms. Specifically, the comprehensive soil erosion resistance algorithm uses soil organic carbon content Coc as a foundation and innovatively employs a hyperbolic secant function sech to couple soil compaction and porosity. Simultaneously, an S-shaped saturation curve is used to simulate the nonlinear anchoring effect of the root system, resulting in a soil erosion resistance Rs that can accurately quantify the surface's defense capabilities. Simultaneously, the comprehensive hydrodynamic stress algorithm couples the predicted rainfall erosivity factor R with the leaf area index Lai of the vegetation canopy through an exponential decay model, and incorporates the amplification effects of topography and previous water content, ultimately obtaining a hydrodynamic stress Sh that comprehensively reflects the external attack capability. This approach, which elevates multidimensional discrete parameters to two core counter-effects—soil erosion resistance Rs and hydrodynamic stress Sh—results in the generation of the sloping farmland feature vector SVL. This not only profoundly reveals the intrinsic physical mechanism of the erosion process but also provides an unprecedented and highly interpretable data foundation for subsequent accurate and differentiated probability extrapolation of erosion patterns.

[0106] Example 4

[0107] This embodiment is an explanation based on Embodiment 3. Please refer to it. Figure 1 , Figure 2 and Figure 3 Specifically: S3 includes S31 and S32;

[0108] S31. Based on the soil erosion resistance Rs and hydrodynamic stress Sh in the feature vector SVL of sloping farmland and the micro-domain runoff concentration γf in the solution parameter set RDS, a model for predicting the instability mode of sloping farmland is constructed. The model for predicting the instability mode of sloping farmland includes a sheet erosion model and a gully erosion model.

[0109] Based on the Weber distribution formula, this paper introduces the micro-domain runoff concentration degree γf in the soil erosion resistance Rs and hydrodynamic stress Sh in the feature vector SVL of sloping farmland and the micro-domain runoff concentration degree γf in the solution parameter set RDS. It uses the natural exponential function exp to represent the inhibitory effect of runoff concentration on sheet erosion, constructs a sheet erosion prediction model, obtains the sheet erosion prediction index Ms, and quantifies the probability of sheet erosion pattern in soil and water loss in sloping farmland.

[0110] The formula for the plaque erosion estimation model is as follows:

[0111] ;

[0112] In the formula, σ represents the preset Weber shape parameter, and λ represents the preset confluence suppression effect coefficient;

[0113] The purpose of this formula is to construct a core failure probability factor based on the Weber distribution failure model in the field of reliability engineering. This study treats the surface of sloping farmland as a system subject to erosion due to continuous hydrodynamic impact, and constructs a stress-erosion resistance ratio... As the core variable of the model, Sh represents the direct interaction between external hydrodynamic stress and surface soil erosion resistance Rs, and is the fundamental cause of erosion. Meanwhile, sheet erosion is characterized by uniformity, while an increase in micro-area runoff concentration γf indicates a tendency for water flow to concentrate. These two states are mutually competitive and inhibitory. Therefore, combining micro-area runoff concentration γf, an exponential decay model is adopted. To quantify this inhibitory effect, when the micro-domain runoff concentration γf is very small, it indicates that the water flow is dispersed, and the factor approaches 1, producing almost no inhibition; when the micro-domain runoff concentration γf is very large, it indicates that the water flow is highly concentrated, and the factor approaches 0, which will greatly reduce the probability of sheet erosion, because at this time energy is more likely to trigger gully erosion.

[0114] S32. Based on the Hill-Langmuir equation, a gully erosion estimation model is constructed for the soil erosion resistance Rs and hydrodynamic stress Sh in the feature vector SVL of sloping farmland and the micro-domain runoff concentration γf in the solution parameter set RDS. The gully erosion prediction index Mr is obtained to quantify the probability of gully erosion patterns in soil and water loss in sloping farmland.

[0115] The formula for the estimation model of gully erosion is as follows:

[0116] ;

[0117] In the formula, ω represents the preset bus energy amplification factor, η represents the preset Hill coefficient, and Kr represents the preset half-saturation constant;

[0118] The objective of this formula is to combine the sloping farmland feature vector (SVL) and the micro-domain runoff concentration γf, and construct the core attack term using hydrodynamic stress Sh and micro-domain runoff concentration γf. Among them, the runoff energy amplification index ω was used to extremely amplify the influence of micro-domain runoff concentration γf, accurately simulating the high concentration of energy after water flow convergence. This is the key to forming the dagger effect to pierce the surface. Dividing the core attack term by the soil erosion resistance Rs, a relatively concentrated erosion intensity is constructed. We obtained a driving variable that can fully reflect the conditions that trigger gully erosion. For the relatively concentrated erosion intensity, we used the S-shaped curve structure of the Hill-Langmuir equation to describe the transition from low probability to high probability. Among them, the Hill coefficient η is the essence of the whole model. When η>1, it can make the probability curve extremely steep, perfectly simulating the positive feedback and cooperative mutation characteristics of gully erosion that deteriorates rapidly once it starts.

[0119] Based on the sheet erosion prediction index Ms and the gully erosion prediction index Mr, generate the prediction erosion index vector PML=[Ms, Mr];

[0120] The following is a specific example of predicting the erosion index vector PML:

[0121] Solving parameter set RDS=[Coc: 1.5%, Zs: 1.8E+06 (kg*m -2 s -1 ), kp: 0.45, ρ: 1.3 (kg / m 3 ), R: 5, Lai: 2.2, γf: 0.7, hw: 30%, hs: 45%];

[0122] The preset model intrinsic coefficients are as follows:

[0123] Reference acoustic impedance value Zo = 1.5E + 0.6 (kg*m) -2 s -1 Soil geological adjustment coefficient α = 0.8;

[0124] The half-saturation constant of the root system's soil-stabilizing effect is Km = 1.0 (kg / m²). 3 The canopy retention coefficient ζ = 0.5;

[0125] The surface runoff effect coefficient β = 1.2, the Weber shape parameter σ = 2.5, and the runoff inhibition effect coefficient λ = 1.5;

[0126] The energy amplification index ω = 1.8, the Hill coefficient η = 3.0, and the half-saturation constant Kr = 1.0;

[0127] Soil erosion resistance Rs is calculated as follows:

[0128] ;

[0129] The hydrodynamic stress Sh is calculated as follows:

[0130] ;

[0131] The flaky erosion prediction index Ms is calculated as follows:

[0132] ;

[0133] The gully erosion prediction index Mr is calculated as follows:

[0134] ;

[0135] The predicted erosion index vector is PML = [Ms: 0.031, Mr: 0.513].

[0136] In this embodiment, by receiving soil erosion resistance Rs and hydrodynamic stress Sh from the feature vector SVL of sloping farmland, and combining them with the micro-domain runoff concentration γf from the solution parameter set RDS, the instability mode of sloping farmland is probabilistically inferred. Its unique advantage lies in the fact that it does not employ a single, universal model, but creatively matches specific, differentiated inference models to two erosion modes with drastically different physical mechanisms. Specifically, it uses a Weber distribution-based inference model to quantify the wide-area fatigue failure-type sheet erosion prediction index Ms, and introduces the natural exponential function exp to accurately describe the inhibitory effect of runoff concentration on this mode. Simultaneously, it uses a Hill-Langmuir equation-based inference model to quantify the local cooperative abrupt change-type gully erosion prediction index Mr, and profoundly characterizes the suddenness and positive feedback characteristics of this mode through parameters such as the Hill coefficient η. This design, which tailors algorithms to different physical processes, makes the final generated Predictive Erosion Index Vector (PML) no longer a general risk value, but a high-resolution risk profile containing rich tactical information. It fundamentally solves the problem that traditional methods cannot distinguish erosion patterns, and provides core decision-making basis for subsequent implementation of accurate and efficient differentiated response strategies.

[0137] Example 5

[0138] This embodiment is an explanation based on Embodiment 4. Please refer to it. Figure 1 and Figure 2 Specifically:

[0139] S4 includes S41 and S42;

[0140] S41. Based on the comparison between the predicted erosion index vector PML and the erosion index threshold interval vector Pth, a predicted erosion response plan AIR is generated.

[0141] Among them, the erosion index threshold interval vector Pth includes the sheet erosion index threshold interval Ps and the gully erosion index threshold interval Pr. The sheet erosion index threshold interval Ps includes the first-level sheet erosion index threshold Ps1, the second-level sheet erosion index threshold Ps2 and the third-level sheet erosion index threshold Ps3. The gully erosion index threshold interval Pr includes the first-level gully erosion index threshold Pr1, the second-level gully erosion index threshold Pr2 and the third-level gully erosion index threshold Pr3. The predicted erosion response scheme AIR includes the predicted sheet erosion response scheme AIRs and the predicted gully erosion response scheme AIRr.

[0142] For the sheet erosion prediction index Ms, if the sheet erosion prediction index Ms < the secondary sheet erosion index threshold Ps2, the predicted sheet erosion response plan AIRs is no intervention.

[0143] If the threshold of the secondary sheet erosion index Ps2 ≤ the sheet erosion prediction index Ms < the threshold of the tertiary sheet erosion index Ps3, the predicted sheet erosion response plan AIRs is the primary sheet erosion treatment plan: use organic materials such as straw, leaves and hay to spread evenly on the surface of the sloping farmland soil.

[0144] If the sheet erosion prediction index Ms ≥ the third-level sheet erosion index threshold Ps3, the predicted sheet erosion response plan AIRs is the second-level sheet erosion treatment plan: based on the first-level sheet erosion treatment plan, cover the planting area of ​​the sloping farmland with waterproof cloth and use heavy objects such as stones or sandbags to compact the edges.

[0145] S42. For the gully erosion prediction index Mr, if the gully erosion prediction index Mr < the secondary gully erosion index threshold Pr2, the predicted gully erosion response plan AIRr is no intervention.

[0146] If the threshold of the secondary gully erosion index Pr2 ≤ the gully erosion prediction index Ms < the threshold of the tertiary gully erosion index Pr3, the predicted gully erosion response plan AIRr is the primary gully erosion treatment plan: On the slope, place multiple rows of obstacles in the horizontal direction. The obstacles are composed of sandbags, stones, logs or tightly bundled hay bales.

[0147] If the gully erosion prediction index Ms ≥ the third-level gully erosion index threshold Pr3, the predicted gully erosion response plan AIRr is a second-level gully erosion treatment plan: based on the first-level treatment plan, dig a temporary irrigation canal parallel to the horizontal line at the highest point of the sloping farmland, and reinforce the temporary irrigation canal and the existing drainage ditch by laying gravel on both sides and the bottom.

[0148] S5 includes S51;

[0149] S51. When rainfall begins, a two-dimensional laser raindrop spectrometer is activated to collect raindrop data pairs RDP=[d_(i,t), v_(i,t)], where d_(i,t) represents the diameter of the i-th raindrop at time interval sequence number t, v_(i,t) represents the velocity of the i-th raindrop at time interval sequence number t, and the total number of raindrops sampled by the two-dimensional laser raindrop spectrometer is Ndr. Based on the kinetic flux calculation formula, the real-time rainfall kinetic flux E(t) is obtained. The real-time rainfall kinetic flux E(t) is compared with the preset baseline rainfall kinetic flux. The real-time rainfall kinetic flux factor Ek(t) is obtained by comparing Eth. The predictive rainfall erosivity factor R is replaced by the real-time rainfall kinetic flux factor Ek(t) and recalculated from step S2. The calculation results are accumulated in real time according to the time interval sequence number t to obtain the real-time cumulative monitoring erosion index vector PNL(t) and compared with the erosion index threshold interval vector Pth. The real-time cumulative monitoring erosion index vector PNL(t) includes the real-time monitoring index Ns(t) for sheet erosion and the real-time monitoring index Nr(t) for gully erosion.

[0150] The formula for calculating the real-time rainfall kinetic flux factor Ek(t) is as follows:

[0151] ;

[0152] ;

[0153] In the formula, π represents the preset constant of pi, specifically 3.141592653, ρ_w represents the preset density of water, and A represents the preset sampling area of ​​the two-dimensional laser raindrop spectrometer.

[0154] For the real-time monitoring index Ns(t) of sheet erosion, when comparing it with the erosion index threshold interval vector Pth, the trigger sensitivity is improved. The first-level sheet erosion index threshold Ps1 replaces the second-level sheet erosion index threshold Ps2, and the second-level sheet erosion index threshold Ps2 replaces the third-level sheet erosion index threshold Ps3.

[0155] For the real-time monitoring index Nr(t) of gully erosion, when comparing it with the erosion index threshold interval vector Pth, the trigger sensitivity is increased, and the first-level gully erosion index threshold Pr1 replaces the second-level gully erosion index threshold Pr2, and the second-level gully erosion index threshold Pr2 replaces the third-level gully erosion index threshold Pr3.

[0156] Based on the comparison results after increasing the trigger sensitivity, adjustments were made to the predictive erosion response plan AIR.

[0157] If the predicted sheet erosion response plan AIRs and the predicted gully erosion response plan AIRr are upgraded from no intervention to Level 1 sheet erosion treatment plan and Level 1 gully erosion treatment plan or above, or upgraded from Level 1 sheet erosion treatment plan and Level 1 gully erosion treatment plan to Level 2 sheet erosion treatment plan and Level 2 gully erosion treatment plan, then the monitoring erosion response plan AIL is generated based on the upgraded treatment plan.

[0158] Conversely, the generated monitoring erosion response plan AIL is the current predicted sheet erosion response plan AIRs and the current predicted gully erosion response plan AIRr, without any adjustments.

[0159] In this embodiment, by intelligently comparing the predicted erosion index vector PML generated in the preceding steps with an erosion index threshold interval vector Pth containing multiple threshold levels, this method can accurately transform abstract probability predictions into specific, tiered predicted erosion response plans (AIR). The unique aspect of this process is that it matches specific response strategies to different erosion patterns. For example, it matches a coverage plan to the sheet erosion prediction index Ms, and a blocking or diversion plan to the gully erosion prediction index Mr, thereby achieving targeted responses and solving the problems of blind decision-making and resource waste caused by the inability to distinguish patterns in traditional methods. Furthermore, this embodiment also achieves a closed-loop risk management system by introducing a dynamic monitoring mechanism based on the real-time rainfall kinetic flux factor Ek(t). Specifically, in the process of generating the monitored erosion response plan AIL, the system not only performs real-time verification using the real-time monitored erosion index vector PNL, but also innovatively adopts a strategy to improve the sensitivity of early warning triggers, stipulating that the response plan can only be upgraded and not downgraded. This design ensures that the system not only has the ability to plan ahead, but also has high reliability and risk avoidance capabilities in real rainfall events, achieving a high degree of unity between foresight and robustness.

[0160] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring and predicting soil erosion of sloping land, characterized by: The method comprises the following steps: S1, collecting a basic parameter set RAW of a slope farmland region through various types of sensors, API interfaces and image acquisition devices, performing data processing on the basic data set RAW, and obtaining a calculation parameter set RDS; S2, comprising S21 and S22; S21, performing core feature extraction of the slope farmland region based on the calculation parameter set RDS, using soil organic carbon content Coc as the basis of organic matter, using a hyperbolic secant function sech for soil acoustic impedance Zs and soil micropore connectivity kp, and using an S-shaped saturation curve for root density p, constructing a soil comprehensive erosion resistance algorithm, calculating soil erosion resistance Rs, and quantifying the comprehensive resistance of the slope farmland as an ecological system to erosion; Wherein, the soil comprehensive erosion resistance algorithm expression is as follows: ； In the formula, ln represents a natural logarithm function with constant e as the base, Zo represents a physical constant for normalizing soil acoustic impedance Zs, and the specific value is the acoustic impedance value of pure water, a represents a preset soil geology adjustment coefficient, and Km represents a preset half-saturation constant of root soil fixation effect; S22, according to the predictive rainfall erosion factor R and the leaf area index Lai within the next 24 hours as the core driving of rainfall events on slope farmland, calculating the ratio of soil water content hw to preset soil saturated water content hs, using an exponential saturation model for micro-domain runoff convergence degree γf, constructing a hydrodynamic comprehensive stress algorithm, calculating hydrodynamic stress Sh, and quantifying the erosion ability of the rainfall event within the next 24 hours on the soil; Wherein, the hydrodynamic comprehensive stress algorithm expression is as follows: ； In the formula, ζ represents a preset canopy interception coefficient according to the vegetation type, β represents a preset surface runoff effect coefficient, and exp represents a natural exponential function; According to the soil erosion resistance Rs and the hydrodynamic stress Sh, a slope farmland feature vector SVL = [Rs, Sh] is generated; S3, based on the slope farmland feature vector SVL and the calculation parameter set RDS, a slope farmland instability mode calculation model is constructed, a sheet erosion prediction index Ms and a gully erosion prediction index Mr are obtained respectively, and a prediction erosion index vector PML is generated; S4, comparing the prediction erosion index vector PML with an erosion index threshold interval vector Pth, a prediction erosion response scheme AIR is generated; S5, analyzing real-time rainfall kinetic energy flux factor Ek(t) through a two-dimensional laser raindrop spectrometer, recalculating from step S2 by replacing the predictive rainfall erosion factor R with the real-time rainfall kinetic energy flux factor Ek(t), t is a time interval serial number, obtaining a real-time cumulative monitoring erosion index vector PNL(t) and comparing it with the erosion index threshold interval vector Pth, adjusting based on the prediction erosion response scheme AIR, and generating a monitoring erosion response scheme AIL.

2. The method for monitoring and predicting soil erosion of slope land according to claim 1, wherein: S1 comprises S11 and S12; S11, collecting a basic parameter set RAW of a slope farmland region through various types of sensors, API interfaces and image acquisition devices, The parameters collected by various types of sensors in the basic parameter set RAW include soil acoustic impedance Zs collected by a sound wave probe, soil micro-pore connectivity kp collected by an electrical conductivity matrix sensor, soil organic carbon content Coc collected by a near-infrared spectrum probe, and soil water content hw collected by a time domain reflectometer probe; The parameters collected through the API interface in the basic parameter set RAW include a predicted rainfall intensity time series I(t) within the next 24 hours obtained from a weather platform by using the weather platform API interface, where t represents a time interval sequence number, and the time interval is specifically 0.5 hours; The parameters collected by image acquisition devices in the basic parameter set RAW include high-density three-dimensional point cloud data las of the slope farmland region scanned by a UAV LiDAR, multi-channel raw images tiff of the vegetation in the slope farmland region taken by a multispectral camera, and high-definition root system raw images jpg of the slope farmland region taken by a miniature root window camera; S12, terrain analysis is performed on the high-density three-dimensional point cloud data las using a D8 algorithm to obtain micro-domain runoff convergence degree γf; The normalized difference vegetation index NDVI is calculated for the multi-channel raw images tiff, and the leaf area index Lai is obtained according to the normalized difference vegetation index NDVI and the pixel bisection model; The UNet model and the EnglightenGAN algorithm are used for image segmentation and length statistics for the high-definition root system raw images jpg to obtain root density ρ; For the predicted rainfall intensity time series I(t) within the next 24 hours, the erosive power of rainfall intensity is estimated based on the power-law estimation model of erosive power, and the predicted rainfall erosive power Rk within the next 24 hours is obtained by accumulation. The predicted rainfall erosive power factor R is obtained by comparing the predicted rainfall erosive power Rk with the preset reference rainfall erosive power Rth. According to the above data processing on the basic data set RAW, the calculation parameter set RDS = [Zs, kp, Coc, hw, γf, Lai, ρ, R] is obtained.

3. The method for monitoring and predicting water and soil loss of slope farmland according to claim 1, characterized in that: S3 includes S31 and S32; S31, based on the soil erosion resistance Rs and the hydrodynamic stress Sh in the slope farmland feature vector SVL and the micro-domain runoff convergence degree γf in the calculation parameter set RDS, a slope farmland instability mode calculation model is constructed, which includes a sheet erosion calculation model and a gully erosion calculation model; Based on the Weber distribution formula, the micro-domain runoff convergence degree γf is introduced and represented by a natural exponential function exp to represent the inhibitory effect of runoff convergence on sheet erosion, and a sheet erosion calculation model is constructed to obtain a sheet erosion prediction index Ms, which quantifies the probability of the sheet erosion mode in the water and soil loss of the slope farmland; wherein σ represents a preset Weber shape parameter, and λ represents a preset convergence inhibition effect coefficient; ； ​ S32, based on the Hill-Romiller equation, the soil resistance Rs and the water dynamic stress Sh in the slope farmland feature vector SVL and the micro-domain runoff convergence degree γf in the calculation parameter set RDS are used to construct a gully erosion prediction model to obtain a gully erosion prediction index Mr, which quantifies the probability of the gully erosion mode in the soil and water loss of the slope farmland; The gully erosion prediction model is expressed as follows: ； In the formula, ω represents a preset convergence energy amplification coefficient, η represents a preset Hill coefficient, and Kr represents a preset half-saturation constant. According to the sheet erosion prediction index Ms and the gully erosion prediction index Mr, a prediction erosion index vector PML=[Ms, Mr] is generated.

4. The soil and water loss monitoring and prediction method for slope farmland according to claim 3, characterized in that: S4 includes S41 and S42; S41, the prediction erosion index vector PML is compared with the erosion index threshold interval vector Pth to generate a predicted erosion response scheme AIR; The erosion index threshold interval vector Pth includes a sheet erosion index threshold interval Ps and a gully erosion index threshold interval Pr, the sheet erosion index threshold interval Ps includes a first sheet erosion index threshold Ps1, a second sheet erosion index threshold Ps2 and a third sheet erosion index threshold Ps3, the gully erosion index threshold interval Pr includes a first gully erosion index threshold Pr1, a second gully erosion index threshold Pr2 and a third gully erosion index threshold Pr3, and the predicted erosion response scheme AIR includes a predicted sheet erosion response scheme AIRs and a predicted gully erosion response scheme AIRr; For the sheet erosion prediction index Ms, if the sheet erosion prediction index Ms is less than the second sheet erosion index threshold Ps2, the predicted sheet erosion response scheme AIRs is no intervention; If the second sheet erosion index threshold Ps2 is less than or equal to the sheet erosion prediction index Ms and less than the third sheet erosion index threshold Ps3, the predicted sheet erosion response scheme AIRs is a sheet erosion first-level treatment scheme: organic materials such as rice straw, leaves and hay are uniformly laid on the surface of the slope farmland soil; If the sheet erosion prediction index Ms is greater than or equal to the third sheet erosion index threshold Ps3, the predicted sheet erosion response scheme AIRs is a sheet erosion second-level treatment scheme: on the basis of the sheet erosion first-level treatment scheme, waterproof cloth is covered on the planting area of the slope farmland and stones or sandbags are used to compact the edges; S42, for the gully erosion prediction index Mr, if the gully erosion prediction index Mr is less than the second gully erosion index threshold Pr2, the predicted gully erosion response scheme AIRr is no intervention; If the second gully erosion index threshold Pr2 is less than or equal to the gully erosion prediction index Ms and less than the third gully erosion index threshold Pr3, the predicted gully erosion response scheme AIRr is a first-level gully erosion treatment scheme: a plurality of rows of obstacles are placed in the horizontal direction on the slope surface, and the obstacles are composed of sandbags, stones, logs or tightly bundled straw bales. If the gully erosion prediction index Ms is greater than or equal to the third gully erosion index threshold Pr3, the predicted gully erosion response scheme AIRr is the second gully erosion treatment scheme: on the basis of the first treatment scheme, a temporary diversion ditch parallel to the horizontal line is dug at the highest point of the slope farmland, and the temporary diversion ditch and the existing drainage ditch are reinforced by laying stones on both sides and the bottom.

5. The method according to claim 4, wherein: S5 comprises S51; S51, when starting to rain, enabling a two-dimensional laser raindrop spectrometer to collect raindrop data RDP=[d_(i,t), v_(i,t)], wherein d_(i,t) represents the diameter of the i-th raindrop at time interval sequence number t, v_(i,t) represents the speed of the i-th raindrop at time interval sequence number t, and the total number of raindrops sampled by the two-dimensional laser raindrop spectrometer is Ndr, the real-time rainfall kinetic energy flux E(t) is obtained based on the kinetic energy flux calculation formula, the real-time rainfall kinetic energy flux E(t) is compared with the preset reference rainfall kinetic energy flux Eth to obtain the real-time rainfall kinetic energy flux factor Ek(t), the predictive rainfall erosivity factor R is replaced with the real-time rainfall kinetic energy flux factor Ek(t) to recalculate from step S2, and the calculation result is accumulated in real time according to time interval sequence number t to obtain a real-time cumulative monitoring erosion index vector PNL(t) and compare it with the erosion index threshold interval vector Pth, wherein the real-time cumulative monitoring erosion index vector PNL(t) comprises a real-time monitoring index Ns(t) of sheet erosion and a real-time monitoring index Nr(t) of gully erosion. For the real-time monitoring index Ns(t) of sheet erosion, the triggering sensitivity is improved when comparing with the erosion index threshold interval vector Pth, the first sheet erosion index threshold Ps1 is replaced with the second sheet erosion index threshold Ps2, and the second sheet erosion index threshold Ps2 is replaced with the third sheet erosion index threshold Ps3. For the real-time monitoring index Nr(t) of gully erosion, the triggering sensitivity is improved when comparing with the erosion index threshold interval vector Pth, the first gully erosion index threshold Pr1 is replaced with the second gully erosion index threshold Pr2, and the second gully erosion index threshold Pr2 is replaced with the third gully erosion index threshold Pr3. According to the comparison result after improving the triggering sensitivity, the monitoring erosion response scheme AIL is generated based on the adjustment of the predicted erosion response scheme AIR. If the predicted sheet erosion response scheme AIRs and the predicted gully erosion response scheme AIRr are improved from no intervention to the first sheet erosion treatment scheme and the first gully erosion treatment scheme or above, or from the first sheet erosion treatment scheme and the first gully erosion treatment scheme to the second sheet erosion treatment scheme and the second gully erosion treatment scheme, the monitoring erosion response scheme AIL is generated according to the treatment scheme after the level is improved. Otherwise, the monitoring erosion response scheme AIL is the current predicted sheet erosion response scheme AIRs and the current predicted gully erosion response scheme AIRr, and no adjustment is made.

Images