A method and device for ecological division of river beach based on hyper-frequency flood and multi-source remote sensing

Abstract

The application discloses a kind of river beach ecological zoning method and device based on super frequency flood and multi-source remote sensing, the method is first calculated by Weibull formula under the bank beach flooding time distribution of super frequency flood scene;Secondly, based on multi-source remote sensing image extraction vegetation and topographic information, combined with flooding time division plant suitable zoning;Then construct the multidimensional habitat evaluation system including vegetation, hydrology, topography and soil, the quality rating of each partition is carried out;Finally, coupling plant zoning and habitat rating results, machine learning model is used to predict and optimize the generation of beach ecological zoning scheme.The application realizes the accurate coupling of flood hydrological process and beach ecosystem, solves the problem that traditional ecological restoration measures lack "hydrology-ecology" collaborative basis, improves the scientificity and feasibility of zoning scheme, and is suitable for arid, semi-arid area northern river beach, lake wetland and other water surrounding ecological zoning, shoreline management, etc..

Description

Technical Field

[0001] This invention relates to the field of ecological restoration technology, specifically a method and apparatus for ecological zoning of riverbanks based on high-frequency floods and multi-source remote sensing. Background Technology

[0002] Riverbanks, as transitional zones connecting land and water, play vital ecological roles in water conservation, water purification, and maintaining biodiversity. However, influenced by climate change and human activities, frequent and excessive flooding has led to abnormally long periods of flooding on riverbanks, resulting in prominent issues such as vegetation degradation and habitat fragmentation. Accurate ecological zoning is a crucial foundation for effective ecological restoration.

[0003] Traditional riverbank ecological zoning methods often rely on field surveys to determine vegetation configuration schemes, lacking consideration for disturbances caused by hydrological events. This results in low vegetation survival rates and unsustainable ecological functions. Furthermore, traditional methods depend on manual surveys to obtain vegetation and habitat information, which is inefficient, costly, and makes it difficult to achieve precise restoration of large-scale riverbanks.

[0004] In existing technologies, flood frequency calculations mostly employ empirical frequency methods such as the Weibull formula, but fail to integrate them with riparian flooding duration and habitat conditions. While multi-source remote sensing technology has been applied to vegetation information extraction, it lacks coupling analysis with hydrological processes, thus failing to provide a scientific basis for plant zoning. Therefore, there is an urgent need to construct a riparian ecological zoning method that integrates ultra-high frequency flood calculation with multi-source remote sensing technology to achieve accurate zoning of riparian ecosystems. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and device for riverbank zoning based on ultra-frequency flood and multi-source remote sensing. By coupling ultra-frequency flood inundation time with multi-source remote sensing vegetation information, suitable plant zoning is obtained, and based on ecological conditions and habitat quality assessment, scientific riverbank ecological zoning is determined.

[0006] A method for ecological zoning of riverbanks based on ultra-frequent floods and multi-source remote sensing includes the following steps:

[0007] Step S1: Calculate the temporal and spatial distribution of flooding

[0008] Hydrological, topographic, and remote sensing data of the target riverbanks were collected. After preprocessing and frequency analysis, the peak flow was calculated using the Weibull formula under different super-frequency flood return periods. A model relating peak flow to inundation depth was established to calculate the inundation time of each spatial unit on the riverbank. Based on field control point data, the spatial distribution map of the inundation time was generated through spatial interpolation.

[0009] Step S2: Divide the suitable areas for planting

[0010] Acquire multi-source remote sensing images and extract vegetation type, coverage and topographic information of the beach; based on the flooding time and space distribution map generated in step S1, and combined with the flood tolerance characteristics of vegetation, divide the beach space into multiple suitable vegetation zones and generate a suitable vegetation zone map.

[0011] Step S3: Evaluate and rate habitat quality

[0012] Construct a habitat quality evaluation index system that includes vegetation status, hydrological conditions, topographic features, and soil quality; for each suitable plant zone divided in step S2, collect or calculate data of various indicators, calculate the habitat quality index of each zone using the comprehensive index method, and rate the quality level based on the habitat quality index.

[0013] Step S4: Generate a shoreline ecological zoning scheme

[0014] The plant suitability zoning map output in step S2 is coupled with the habitat quality rating result output in step S3. The relevant factors are used as features to input into the machine learning model for training and prediction to obtain preliminary ecological zoning prediction results. The final shoreline ecological zoning scheme map is generated by combining the digital elevation model for terrain constraints and spatial continuity optimization.

[0015] Furthermore, the flooding time-space distribution map generated in step S1 is specifically a raster map obtained by extending the flooding time data of at least 50 field control points to the entire shoreline area using the Kriging interpolation method, and is verified using the flooding range extracted from the concurrent remote sensing image, with an overlap of ≥85%.

[0016] Furthermore, when dividing suitable plant zones in step S2, the flooding time thresholds are as follows: Long-term flood-tolerant plant zone: flooding time > 60-90 days / year; Short-term flood-tolerant plant zone: flooding time 30-60 days / year; Flood-avoiding plant zone: flooding time 5-30 days / year; Unsuitable pioneer plant improvement zone: flooding time < 5 days or > 120 days / year.

[0017] Furthermore, in step S3, the secondary indicators used to evaluate habitat quality include vegetation cover, vegetation diversity, and vegetation drought stress index for vegetation status; secondary indicators of hydrological conditions include groundwater level suitability, flooding time suitability, and surface water-groundwater connectivity for hydrological conditions; secondary indicators of topographic features include topographic slope and aspect for topographic features; and secondary indicators of soil quality include soil organic matter content, soil moisture content, and soil salinization degree for soil quality. The weights of each secondary indicator are determined by combining the analytic hierarchy process (AHP) with the entropy weight method.

[0018] Furthermore, the machine learning model used in the coupling analysis in step S4 is a random forest model; the features input to the model include at least: flooding time standardized score, groundwater level standardized score, vegetation diversity score, soil salinization score, slope score, aspect score, and comprehensive score obtained by multi-factor weighted summation; the model output is a category code corresponding to different suitable plant zones.

[0019] Furthermore, in step S4, when generating the final shoreline ecological zoning scheme map, the following steps are further included: aligning and fusing the vector data of the preliminary ecological zoning results with the DEM raster data; performing median filtering on the fused raster map to eliminate isolated pixels, and merging small patches based on patch area; the final output zoning map includes zoning codes, suitable vegetation types, and habitat quality level attribute information.

[0020] A riverbank ecological zoning device for implementing the method described above, comprising:

[0021] The data processing unit is used to collect hydrological, topographic, and remote sensing data of the target riverbank. After preprocessing and frequency analysis, the Weibull formula is used to calculate the peak flow under different super-frequency flood return periods. A relationship model between peak flow and inundation depth is established to calculate the flooding time of each spatial unit of the riverbank. Based on the field control point data, the spatial distribution map of the flooding time is generated by spatial interpolation.

[0022] The remote sensing image processing unit is used to acquire multi-source remote sensing images and extract vegetation type, coverage and topographic information of the beach; based on the flooding time and space distribution map generated by the data processing unit, and combined with the flood tolerance characteristics of vegetation, the beach space is divided into multiple suitable vegetation zones, and a suitable vegetation zone map is generated.

[0023] The habitat evaluation unit is used to construct a habitat quality evaluation index system that includes vegetation status, hydrological conditions, topographic features, and soil quality. For each suitable plant zone divided in step S2, data of various indicators are collected or calculated, the habitat quality index of each zone is calculated using the comprehensive index method, and the quality level is rated based on the habitat quality index.

[0024] An ecological zoning generation unit is used to couple the plant suitability zoning map output by the remote sensing image processing unit with the habitat quality rating result output by the habitat evaluation unit. Relevant factors are used as features to input into a machine learning model for training and prediction to obtain preliminary ecological zoning prediction results. Combined with a digital elevation model, terrain constraints and spatial continuity optimization are performed to generate the final shoreline ecological zoning scheme map.

[0025] Furthermore, it also includes:

[0026] The data input interface is used to receive user-input hydrological station data, remote sensing image data, field survey data, and topographic data.

[0027] The results display and output module is used to visualize the flooding time distribution map, plant suitability zone map, habitat quality rating map and final ecological zoning scheme map, and supports map genus query and export functions.

[0028] The beneficial effects of this invention are:

[0029] 1. This invention achieves precise coupling of hydrological and ecological processes, enhancing the scientific rigor of zoning: Traditional riparian ecological restoration often neglects the disturbance effects of extreme hydrological events such as floods. This invention innovatively links the duration of high-frequency flooding calculated based on the Weibull formula directly to plant ecological suitability, achieving for the first time a quantitative coupling between extreme hydrological processes and riparian ecosystem responses. This allows the zoning scheme to consider not only static habitat conditions but also dynamic hydrological stresses, significantly improving the alignment between ecological restoration measures and natural hydrological patterns, and enhancing the scientific rigor and ecological sustainability of the zoning scheme.

[0030] 2. Integrating multi-source remote sensing technology significantly improves survey efficiency and scope: Traditional methods relying on manual field surveys are costly, inefficient, and unable to cover large areas. This invention comprehensively utilizes multi-source remote sensing data, including 10-meter resolution Sentinel-2 and 30-meter resolution Landsat-9 data, to automatically extract vegetation type, coverage, and topographic information, achieving efficient and accurate monitoring of large-scale riverbank ecological environments. Compared to purely manual surveys, this method can reduce costs by more than 60% and has the capability for regular updates and dynamic monitoring, providing reliable data support for long-term ecological management.

[0031] 3. A multi-dimensional and comprehensive habitat quality evaluation system was constructed, providing a more comprehensive basis for decision-making: Targeting the characteristics of arid and semi-arid regions, this invention constructed a comprehensive evaluation index system covering four dimensions: vegetation status, hydrological conditions, topographic features, and soil quality. The weights were determined using a combination of the analytic hierarchy process (AHP) and the entropy weight method, highlighting key limiting factors such as drought stress and groundwater. This system provides more comprehensive and objective evaluation results, accurately reflecting the current habitat quality of riverbanks and recommending appropriate vegetation configurations and restoration measures for different zones, thus providing a basis for management decisions.

[0032] 4. A predictable and adjustable ecological zoning method is proposed, with strong applicability and scalability: This invention's method is not a simple current-state mapping, but rather, by coupling flooding time, habitat quality, and machine learning models such as random forests, it can predict the potential vegetation distribution, forming an ecological zoning scheme that combines current-state assessment with future-oriented guidance. The method has a clear logic, and the parameters can be flexibly adjusted according to the hydrological, soil, and vegetation characteristics of different regions. Therefore, it is not only applicable to semi-arid riverbanks, but can also be extended to the restoration and management of various aquatic ecosystems such as lake wetlands and reservoir drawdown zones, demonstrating broad applicability and significant promotional value. Attached Figure Description

[0033] Figure 1 This is a flowchart of the riverbank ecological zoning method based on ultra-frequent floods and multi-source remote sensing of the present invention. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] Please see Figure 1 This invention provides a method for ecological zoning of riverbanks based on ultra-frequent floods and multi-source remote sensing, comprising the following steps:

[0036] Step 1. Calculation of flooding time

[0037] (1) Data preprocessing: Collect the annual peak flow series of control hydrological stations near the target riverbank (sample size ≥ 30 years), use the water conservancy project influence restoration method for consistency correction, supplement missing data by neighboring year interpolation, and use Grubbs test to remove outliers. If the hydrological station is far from the target riverbank, perform water level and flow interpolation of upstream and downstream hydrological stations or water level stations.

[0038] (2) Major flood handling: Identify major floods (such as flow exceeding twice the series mean), determine the historical survey period N (such as a complete record period including historical floods), adjust the sorting sequence, and calculate the empirical frequency using the Weibull formula:

[0039] P=m / (N+1) (major flood) and P=m / (n+1) (ordinary flood) are converted to obtain the return period T=1 / P.

[0040] (3) Construction of flooding time model: Based on measured flood inundation data, a model relating peak flow Q to inundation depth h is established:

[0041] h = aQ + b (a and b are fitting parameters)

[0042] The flooding time is calculated based on the receding water coefficient: t = h / k, where k is the receding water rate. A response model between flooding time and peak flow is constructed.

[0043] (4) Spatial distribution of flooding time: Using 10m and 30m resolution DEM data, combined with the topographic features of the beach, the calculated flooding time of each control point is extended to the entire beach area through spatial analysis methods such as Kriging interpolation, and a spatial distribution map of flooding time is generated.

[0044] The specific steps are as follows:

[0045] Step 1.1: Data Preparation:

[0046] Collect hydrological, topographical, and remote sensing data to ensure the accuracy of the flood time-elevation model parameter fitting.

[0047] Hydrological data include: a sample size of ≥15 years, containing measured data from at least two high-frequency floods (e.g., a 50-year return period). Annual maximum peak flow series (Q, unit: m³ / s) for each frequency control station, and corresponding flood hydrographs (water level / flow rate changes over time).

[0048] Topographic data: digital elevation model of the shoreline (DEM, resolution ≥30m), field inundation survey data (such as the inundation range of historical floods, inundation depth of typical points), with ≥50 field survey points covering different areas of nearshore, middle and farshore.

[0049] Remote sensing data: Multi-source remote sensing images during the flood season (such as Sentinel-2 and Landsat-9) are used to verify the inundation range. Cloud cover should be less than 10%, and the data should be synchronized with the actual flood time.

[0050] Step 1.2: Construct the "peak flow - inundation depth" response relationship

[0051] 1.2.1 The inundation depth (h, unit: m) of a spatial unit (grid) on the shore refers to the difference between the water level of that unit and the ground elevation during the flood season, and the formula is:

[0052]

[0053] in: Beach The flooding depth (m) of the grid, if This indicates that the unit was not submerged;

[0054] : Flood level (m, Wusong elevation / local benchmark elevation) corresponding to the peak flow rate Q;

[0055] : Ground elevation (m) of the (x,y) grid. (The baseline elevation is consistent), extracted from the DEM.

[0056] 1.2.2 Peak flow rate - flood level relationship (QZ) w Model)

[0057] The response relationship between peak flow Q and flood level is established using a power function fitting method, and the formula is as follows:

[0058]

[0059] Selecting "peak flow Q - corresponding water level Z" from historical flood measurement data w "Measured data pairs (at least 10 sets, such as..." );

[0060] Alternatively, the least squares method can be used to fit the parameters a, b, and c, ensuring that the fitting error is <0.1m (measured Z). w With calculation Z w The root mean square error (RMSE) is used to measure the average error of the riverbed. In arid and semi-arid regions, since the riverbed is mainly sandy and the changes in riverbed erosion and deposition are small, parameters a, b, and c can be fitted with measured data from the past 10 years without frequent corrections (e.g., in the Ningxia section of the upper reaches of the Yellow River, a=0.002, b=0.5, c=1015.2, RMSE=0.08m).

[0061] 1.2.3 Determination of Inundation Range

[0062] Combining DEM and Z w (Q) The flooding range is determined by the "elevation threshold method": Flooding range = {(x,y)|Z} g (x,y)≤Z w (Q)}.

[0063] Remote sensing image verification: Compare the calculated inundation extent with the inundation extent extracted from the same period's Sentinel-2 image (using the NDWI water index). If the spatial overlap is <85%, QZ needs to be adjusted. w Adjust model parameters (such as the c-value) until the overlap is ≥85%.

[0064] Step 1.3: Constructing the flood receding curve (modeling the receding pattern)

[0065] 1.3.1 Definition of Drainage Curve

[0066] The receding water curve describes the pattern of the inundation depth decreasing over time after a flood. In arid and semi-arid regions, due to strong evaporation and rapid infiltration, the receding water process conforms to the "exponential decay model".

[0067] 1.3.2 Formula for Drainage Curve

[0068] The maximum inundation depth corresponding to a certain peak flow rate Q

[0069] hmax(x,y) represents the initial value and the submergence depth at any time t during the receding process.

[0070] h t The formula for (x,y) is:

[0071]

[0072] : The flooding depth (m) of the (x,y) raster at time t;

[0073] The maximum inundation depth (m) of the (x,y) grid during the flood season, i.e., h(x,y) (calculated in step 1.2.1);

[0074] k: Regression coefficient (d⁻¹), reflecting the rate of regression. The k value is greater in arid areas than in humid areas (e.g., k=0.05-0.08d⁻¹ in Ningxia section, k=0.02-0.03d⁻¹ in Dongting Lake).

[0075] t: Receding time (d), starting from the moment the flood peak passes.

[0076] 1.3.3 Fitting of the drainage coefficient k

[0077] Select measured receding water data from typical flood events (such as a 50-year flood): record the inundation depth at different times t after the flood peak. (At least 5 sets of data points, obtained through water level monitoring or on-site measurement);

[0078] For the formula: Taking the logarithm of both sides transforms the relationship into a linear one.

[0079]

[0080] Linear regression is used to fit the k value, requiring a goodness of fit R² ≥ 0.9;

[0081] The k-value is determined according to the soil type (sandy, loam): sandy soil has fast infiltration, so the k-value is 0.07-0.08 d⁻¹; loam soil has a k-value of 0.05-0.06 d⁻¹ (arid areas are dominated by sandy soil, so soil sampling points need to be increased to ensure the accuracy of zoning).

[0082] Step 1.4: Calculate the flooding time for each spatial unit of the beach.

[0083] Flooding time (T, unit: d) refers to the total time from when a spatial unit is flooded (h>0) to when the water recedes completely (h≤0). It needs to be calculated separately for the "flooding period" and the "receding period". In arid areas, the flooding period is short (usually 1-3 days), which can be simplified to "receding period duration + flooding period constant".

[0084] 1.4.1 Calculation of the duration of the receding water period

[0085] When the water recedes to When t corresponds to the duration of the receding water period (T_receding), substituting it into the receding water curve formula:

[0086]

[0087] because In actual calculations, (The trace amount of water accumulation in sandy soils in arid areas can be ignored) is the endpoint of water receding, and the formula is adjusted as follows:

[0088]

[0089] Solve for T by taking the logarithm of both sides. 退

[0090]

[0091] 1.4.2 Determining the duration of the flood season

[0092] Duration of the flood season (T) 涨 This refers to the time from when the flood begins to inundate the unit until it reaches its maximum inundation depth. In arid areas, floodwaters rise rapidly. Statistics are obtained through measured flood hydrographs.

[0093] Nearshore area (h > 1.0m): T rise = 2-3 days;

[0094] Central region (0.5m < h ≤ 1.0m): T rise = 1-2 days;

[0095] Offshore areas (0 < h ≤ 0.5 m): T rise = 1 day;

[0096] 1.4.3 Formula for Total Flooding Time

[0097] Total flooding time T 总 (x, y) represents the sum of the durations of the flood season and the receding season:

[0098]

[0099] If h max (x,y)≤0.05m (trace flooding), judged as "ineffective flooding", T 总(x,y)=0

[0100] Step 1.5: Calculation of flood duration for floods with different return periods

[0101] Based on the Weibull formula in step 1, the peak flow Q for different return periods (e.g., 50-year, 20-year, and 10-year return periods) is calculated. T Substituting into the above model, we obtain the temporal and spatial distribution of flooding along the entire riverbank:

[0102] 1. Calculate the corresponding peak flow rate Q using the Weibull formula for the target return period T. T ;

[0103] 2. Q T Substitute Q Z w model (step 1.2.1), calculate flood level Z w (Q T );

[0104] 3. Based on Z w (Q T ) and DEM, calculate h for each grid cell max( (x, y) (Step 1.2.1);

[0105] 4. Substitute into the recession curve formula to calculate T. 退 (x,y) (Step 1.4.1);

[0106] 5. Combine with T 涨( x,y), to obtain T 总 (x,y) generates a flood time raster map for this recurrence period.

[0107] Step 1.6: Generation of Temporal and Spatial Distribution of Floodwater

[0108] 1.6.1 Control Point Selection

[0109] Fifty field survey points were evenly distributed along the shoreline, covering different areas near the shore, in the middle, and far shore. The DEM elevation Zg of each point was extracted, and the corresponding frequency flood was calculated (e.g., 50-year return period, Q=1600m). 3 The total flooding time is ( / s).

[0110] 1.6.2 Spatial Interpolation

[0111] Kriging interpolation was used to extend the flooding time of the 50 control points to the entire shoreline area (grid resolution 10m):

[0112] Interpolation model: spherical variogram, range = 500m, sill value = 450, nugget value = 50;

[0113] Generate a temporal and spatial distribution map of flood inundation (selectable return periods of 10 years, 20 years, 30 years, 50 years, etc.):

[0114] The flooding time is 60-90 days / year in the nearshore area (Z≤1015.0m), 30-60 days / year in the central area (1015.0m<Z≤1016.5m), and 5 days / year in the offshore area (Z>1016.5m).

[0115] 1.6.3 Remote Sensing Verification

[0116] Sentinel-2 imagery (cloud cover < 8%) from the 2018 flood season (synchronized with the measured flood time) was selected, and the inundation extent was extracted using the NDWI water index. Comparison with the calculated inundation extent: spatial overlap ≥ 85%, no adjustment of the Q-Zw model parameters is needed, and the interpolation results are reliable. Otherwise, the Q-Zw model parameters should be readjusted.

[0117] Step 2. Plant Zoning

[0118] (1) Multi-source remote sensing data preprocessing: acquire 10m resolution Sentinel-2 image and 30m resolution Landsat-9 image, perform geometric correction, radiometric correction and atmospheric correction to ensure that the image accuracy meets the requirements of GB / T17941-2008.

[0119] (2) Vegetation information extraction: The object-oriented classification method was adopted, combined with the normalized vegetation index (NDVI), enhanced vegetation index (EVI), and vegetation water supply index (VSWI) to extract the vegetation type and coverage information of the beach. The classification accuracy was ≥88% (10m resolution) and ≥85% (30m resolution).

[0120] (3) Construction of a database of plant flood and drought tolerance characteristics: Through data survey, collect the flood tolerance thresholds of common beach plants (such as reeds, bermudagrass, and Kentucky bluegrass) in the target beach area, establish a database of plant flood tolerance characteristics, and clarify the suitable flood time range for different plants.

[0121] (4) Conduct a field survey of the riparian vegetation in different zones. The survey content includes: vegetation species composition, diversity, aboveground and belowground biomass, and cover. Search the literature for the stress tolerance (flooding, drought) characteristics of the vegetation in the relevant zones.

[0122] (5) Suitable vegetation zones: By overlaying the spatial distribution map of flooding time with the vegetation information extraction results, the shoreline was divided into 4 types of suitable vegetation zones according to the flooding time:

[0123] Class I: Plants tolerant of prolonged flooding: Flooding time 60-90 days / year, suitable for planting tamarisk, as well as aquatic plants such as reeds, calamus, and cattails;

[0124] Class II: Areas tolerant of short-term flooding and drought: Flooding time 30-60 days / year, suitable for planting water plants such as Bermuda grass, barnyard grass, and water onion;

[0125] Category III: Flood-Averse and Drought-Resistant Plant Zone: Flooding period is 5-30 days. Suitable for planting mesophytes such as Kentucky bluegrass, sheepgrass, and ryegrass.

[0126] Category IV: Unsuitable area → Pioneer improved plant area: T < 5 days / year, suitable for planting pioneer improved plants.

[0127] The main types are as follows:

[0128] Class I area (areas resistant to prolonged flooding and drought)

[0129] Key characteristics of the zone: flooding for 60-90 days, VSWI < 0.6, EC < 8.0 dS / m, groundwater level 0.5-2.0m, resistant to flooding, dampness, and mild salinity.

[0130] Typical plant types and characteristics are shown in Table 1:

[0131] Table 1

[0132]

[0133] Class II area (area resistant to short-term flooding and drought)

[0134] Key characteristics: flooding duration 30-60 days, VSWI < 0.55, EC < 6.0 dS / m, groundwater depth 2.0-4.0 m, resistant to short-term flooding, drought, and shade (see Table 2).

[0135] Table 2

[0136] Class III Area (Flood-Averse and Drought-Resistant Area)

[0137] Key characteristics of the zone: flooding for 5-30 days, VSWI < 0.45, EC < 5.0 dS / m, groundwater level 4.0-8.0 m, drought-resistant, shade-resistant, and wind-erosion resistant (see Table 3).

[0138] Table 3

[0139]

[0140] Category IV area (unsuitable area → pioneer improved plants)

[0141] Key characteristics of the zoning: T < 5 or > 120 days, VSWI > 0.7, EC > 8.0 dS / m. Poorly ventilated areas are mostly severely saline-alkali / extremely arid environments, requiring pioneer plants to improve the soil before replanting target species, see Table 4.

[0142] Table 4

[0143]

[0144] Step 3. Shoreline Habitat Zoning

[0145] 3.1 Habitat quality assessment

[0146] (1) Construction of evaluation index system for riparian habitat: 11 indicators were selected from four dimensions: vegetation status, hydrological conditions, topographic features, and soil quality, as detailed in Table 5:

[0147] Table 5

[0148]

[0149] (2) Determining the weight of indicators:

[0150] The comprehensive weights were determined using a combination of the analytic hierarchy process (AHP) (subjective) and the entropy weight method (objective), taking into account the characteristics of arid and semi-arid regions and highlighting the impact of drought stress. Specific weights are shown in Table 6.

[0151] Table 6

[0152]

[0153] (3) Quality rating:

[0154] The habitat quality index (0-100 points) is calculated using the comprehensive index method. The scoring formula is as follows:

[0155]

[0156] The standardized scores of individual indicators are converted using the "extreme value method".

[0157] Positive indicator: Score = (Actual value - Minimum value) / (Maximum value - Minimum value);

[0158] Negative indicator: Score = (maximum value - actual value) / (maximum value - minimum value).

[0159] The scores are divided into four levels, with adjusted thresholds for arid regions (emphasizing groundwater and drought stress compliance requirements):

[0160] (1) Excellent (85-100 points): The groundwater level is suitable (0.5-2.0m), the vegetation is not under drought stress, the soil is not salinized, and the water supply and demand are balanced;

[0161] (2) Good (70-84 points): The groundwater level is relatively suitable (2.0-4.0m), the vegetation is under mild drought stress, the soil is slightly salinized (EC<4dS / m), and the water supply and demand are basically balanced;

[0162] (3) Moderate (55-69 points): Groundwater level deviation (4.0-6.0m), vegetation under moderate drought stress, soil moderate salinization (EC4-8dS / m), and water supply and demand imbalance;

[0163] (4) Poor (<55 points): The groundwater level is seriously unsuitable (>6.0m), the vegetation is severely drought stressed, the soil is severely salinized (EC>8dS / m), and the water is seriously insufficient.

[0164] Step 3.2 Zoning Status Assessment

[0165] Step 3.2.1 uses a standardization method to convert factors of different dimensions into scores within the [0,1] interval. The specific formula and calculation example are as follows:

[0166] 3.2.1.1. Flooding duration (positive range type: 60-90 days is optimal)

[0167]

[0168] Among them, S 淹水 The standardized score is the flooding time; T is the actual flooding time (days) of the grid.

[0169] 3.2.1.2. Groundwater level (for positive range type: 0.5-8.0m is suitable)

[0170]

[0171] Among them, S 地下水 The score represents the standardized score for groundwater level; H represents the actual groundwater depth (m) of the grid.

[0172] 3.2.1.3. Vegetation diversity

[0173] A zoned vegetation quadrat survey was conducted on the beach, including the species, quantity, and biomass of the vegetation.

[0174]

[0175] Among them, S 多样性 H' is the standardized score for diversity; H' is the Shannon-Wiener index of the raster; H min =0 (minimum value in arid regions), H max =2.5 (theoretical maximum value for arid regions).

[0176] 3.2.14. Soil quality, including soil nutrients such as nitrogen and phosphorus, and the degree of soil salinization. Among them, the degree of soil salinization (negative type: the lower the EC, the higher the score)

[0177]

[0178] S 盐碱化 : represents the salt-alkali normalization score; EC represents the actual grid conductivity (dS / m); EC min =0, EC max =10.0 (threshold for severe salinization in arid areas).

[0179] 3.2.1.5 Slope (for positive range type: 5-10° is optimal)

[0180]

[0181] 3.2.1.6 Slope aspect (classification and assignment), see Table 7.

[0182] Table 7

[0183]

[0184] 3.3 Vegetation Habitat Rating

[0185] Multi-factor weighted coupling evaluation based on vegetation habitat

[0186] The factors in the aforementioned evaluation system are evaluated according to their weights, and the calculation formula is as follows (weighted sum of factor scores, weights are shown in step 3.1):

[0187]

[0188] in, The standardized score of the i-th indicator (e.g., S1 = vegetation cover score, S2 = vegetation diversity score, ... S... 11 =Soil EC value score);

[0189] The comprehensive weight of the i-th indicator (determined by the analytic hierarchy process and entropy weight method, such as groundwater level weight of 0.18 and vegetation drought stress index weight of 0.12, etc.).

[0190] F∈[0,1], the closer it is to 1, the better the habitat quality and the stronger the ecological suitability. This parameter F is one of the input feature values ​​of the random forest vegetation zoning prediction model.

[0191] For example:

[0192] Where F is the multi-factor coupled comprehensive score (0-1 interval).

[0193] Step 4: Shoreline Ecological Zoning

[0194] By comprehensively analyzing the vegetation and habitat data from steps 2 and 3 (inputting feature values ​​to the prediction model), the ecological zoning of the target beach is predicted.

[0195] 4.1 Vegetation distribution prediction based on random forest

[0196] 4.1.1 Random forest was used for training and computation using PythonScikit-learn (random forest algorithm package).

[0197] The steps are as follows:

[0198] (1) Sample set construction calculation

[0199] Sample feature vector: Each sample contains: S 淹水 S 地下水 S 多样性 S 盐碱化 (or other soil characteristics)

[0200] S 坡度 S 坡向 11 features including F;

[0201] Sample coding: Long-term flood-tolerant and drought-tolerant area = 1, Short-term flood-tolerant and drought-tolerant area = 2, Drought-intolerant area = 3, Unsuitable area (pioneer plant area) = 0;

[0202] Sample augmentation calculation: For the original 500 quadrats, each sample generates 3 new samples through "spectral perturbation" (e.g., S_flooding ± 0.05), and the total number of samples = 500 + 500 × 3 = 2000.

[0203] (2) Model parameter calculation and optimization

[0204] Number of decision trees (n_estimators): 100 (determined through 5-fold cross-validation; the validation set OA is highest when n=100).

[0205] Maximum depth (max_depth): 12 (to avoid overfitting, calculate the minimum depth when the validation set error rate is <5%).

[0206] Subsample ratio: 0.8 (randomly select 80% of the training set samples to construct a single decision tree);

[0207] Minimum number of samples for node split (min_samples_split): 5 (to ensure statistical significance of node split).

[0208] (3) Model prediction calculation logic

[0209] Single decision tree prediction: For the 7 features of the input raster, traverse the decision tree nodes (e.g., "F > 0.7 → determine S: flooding > 0.8 → output category 1").

[0210] Ensemble prediction: 100 decision trees vote, and the category with the most votes is the final distribution type of the raster;

[0211] Prediction accuracy calculation:

[0212]

[0213]

[0214] Where: N is the total number of validation samples; k is the number of partition categories (4 categories); To obfuscate the diagonal elements of the matrix; Summing the i-th row; Summate the i-th column.

[0215] 4.1.2.3 Distribution Prediction Output

[0216] Raster classification rules: Model output category code → corresponding partition type (1=Long-term flood-resistant and drought-resistant area, 2=Short-term flood-resistant and drought-resistant area, 3=Flood-avoiding and drought-resistant area, 0=Unsuitable area);

[0217] Spatial continuity calculation: 3×3 window mid-range filtering, the formula is:

[0218]

[0219] Among them, C 最终 (x, y) represents the raster class after filtering; C(x±1, y±1) represents the original class of the surrounding 8 rasters.

[0220] Based on the vegetation distribution prediction results from step 4.1, and combined with existing DEM data, a DEM-based shoreline ecological zoning map is generated. The specific steps are as follows:

[0221] 4.2 DEM Preprocessing and Terrain Constraint Extraction

[0222] Topographic filling and smoothing: The original DEM is filled with depressions (to eliminate local low-lying water artifacts), and a 3×3 window smoothing filter is used to preserve the macroscopic features of the terrain;

[0223] Topographic factor extraction: Based on DEM, slope raster (unit: °) and aspect raster (azimuth: 0°-360°) are generated, with the raster resolution consistent with the random forest prediction results (≥30m).

[0224] Elevation zoning constraints: Based on the DEM elevation, the area is divided into "nearshore low elevation zone (Z≤Zw-1.0m), central medium elevation zone (Zw-1.0m<Z≤Zw+0.5m), and farshore high elevation zone (Z>Zw+0.5m)" (Zw is the multi-year average flood level) to ensure that the zoning is compatible with the topographic elevation (e.g., the nearshore zone is given priority to be a zone resistant to long-term flooding).

[0225] 4.3. Raster Map Generation and Optimization

[0226] (1) Initial raster image generation

[0227] Rasterization: Align the partitioning results (vector format) predicted by the random forest with the DEM raster, assign each DEM raster a corresponding partitioning type code (0 / 1 / 2 / 3), and generate an initial partitioning raster map;

[0228] Resolution matching: If the DEM resolution (e.g., 10m) is higher than the prediction result resolution (e.g., 30m), use the "nearest neighbor method" to resample to ensure that the raster image resolution is uniform (10-30m is recommended to balance accuracy and efficiency).

[0229] (2) Spatial continuity optimization (3×3 window mid-range filtering)

[0230] For any "isolated rasters" that may exist in the initial raster image (such as a single type 3 raster interspersed within type 1), noise is eliminated using median filtering, as shown in the following formula:

[0231]

[0232] in:

[0233] : The partition type of the filtered (x,y) raster;

[0234] (x,y) The original partition type of the 8 adjacent rasters surrounding the current raster;

[0235] Filtering rule: Retain continuous patches with an area ≥ 0.01 km², and merge isolated patches smaller than this area into the adjacent dominant partition (the adjacent partition with the larger area).

[0236] (3) Raster output

[0237] Output zoning maps, which can be in GeoTIFF (a mainstream GIS-compatible format) or ENVI raster format;

[0238] The raster attribute table contains key information such as "zone code, zone name, suitable vegetation type, and habitat quality level", which can be easily queried when selecting subsequent restoration measures.

[0239] This invention has the following features and effects:

[0240] 1. It achieves quantitative coupling of hydrological and ecological processes, significantly improving the scientific nature and ecological adaptability of regional zoning.

[0241] Traditional shoreline restoration often overlooks the disturbance effects of extreme hydrological events such as floods. This invention innovatively links the inundation time of high-frequency floods calculated based on the Weibull formula directly with the flood tolerance characteristics of plants, achieving precise coupling between dynamic hydrological processes and static habitat conditions. This makes the zoning scheme more in line with natural hydrological rhythms and improves the success rate and sustainability of ecological restoration measures.

[0242] 2. Integrate multi-source remote sensing technologies to achieve efficient and large-scale extraction of ecological and environmental information.

[0243] This invention comprehensively utilizes multi-source remote sensing data such as Sentinel-2 and Landsat-9 to automatically extract vegetation type, coverage, and topographic information, overcoming the shortcomings of traditional manual surveys, such as high cost, low efficiency, and limited coverage. Compared with pure field surveys, this method can reduce survey costs by more than 60% and has the capability for regular updates and dynamic monitoring.

[0244] 3. A multi-dimensional comprehensive evaluation system has been constructed, providing more comprehensive and objective decision support.

[0245] In response to the ecological characteristics of arid and semi-arid regions, a habitat quality evaluation index system was constructed, covering four dimensions: vegetation status, hydrological conditions, topographic features, and soil quality. The weights were determined by combining the analytic hierarchy process (AHP) and the entropy weight method, highlighting key limiting factors such as drought stress and groundwater, making zoning decisions more scientific and targeted.

[0246] 4. Introduce machine learning prediction models to achieve intelligent optimization and wide applicability of partitioning schemes.

[0247] By employing machine learning models such as random forests and coupling multi-source features including flooding time and habitat quality, this method enables the prediction and zoning optimization of potential vegetation distribution. It is not only based on current status assessment but also possesses greater predictive and guiding capabilities. Parameters can be flexibly adjusted according to different regions, making it applicable to various ecological scenarios such as riverbanks, lake wetlands, and reservoir drawdown zones, and demonstrating strong scalability.

[0248] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for ecological zoning of riverbanks based on ultra-frequent floods and multi-source remote sensing, characterized in that, Includes the following steps: Step S1: Calculate the temporal and spatial distribution of flooding Hydrological, topographic, and remote sensing data of the target riverbanks were collected. After preprocessing and frequency analysis, the peak flow was calculated using the Weibull formula under different super-frequency flood return periods. A model relating peak flow to inundation depth was established to calculate the inundation time of each spatial unit of the riverbank. Based on the on-site control point data, the spatial distribution map of flooding time is generated by spatial interpolation. Step S2: Divide the suitable areas for planting Acquire multi-source remote sensing images and extract vegetation type, coverage and topographic information of the beach; based on the flooding time and space distribution map generated in step S1, and combined with the flood tolerance characteristics of vegetation, divide the beach space into multiple suitable vegetation zones and generate a suitable vegetation zone map. Step S3: Evaluate and rate habitat quality Construct a habitat quality evaluation index system that includes vegetation status, hydrological conditions, topographic features, and soil quality; for each suitable plant zone divided in step S2, collect or calculate data of various indicators, calculate the habitat quality index of each zone using the comprehensive index method, and rate the quality level based on the habitat quality index. Step S4: Generate a shoreline ecological zoning scheme The plant suitability zoning map output in step S2 is coupled with the habitat quality rating result output in step S3. The relevant factors are used as features to input into the machine learning model for training and prediction, and preliminary ecological zoning prediction results are obtained. By combining digital elevation models to optimize terrain constraints and spatial continuity, the final shoreline ecological zoning scheme map is generated.

2. The method according to claim 1, characterized in that, The flooding time-space distribution map generated in step S1 is specifically a raster map obtained by extending the flooding time data of at least 50 field control points to the entire beach area using the Kriging interpolation method, and is verified using the flooding range extracted from the concurrent remote sensing image, with an overlap of ≥85%.

3. The method according to claim 1, characterized in that, In step S2, when dividing the suitable plant zones, the flooding time thresholds are as follows: Long-term flood-tolerant plant zone: flooding time > 60-90 days / year; Short-term flood-tolerant plant zone: flooding time 30-60 days / year; Flood-avoiding plant zone: flooding time 5-30 days / year; Unsuitable pioneer plant improvement zone: flooding time < 5 days or > 120 days / year.

4. The method according to claim 1, characterized in that, In step S3, the indicators used to evaluate habitat quality include secondary indicators for vegetation status such as vegetation cover, vegetation diversity, and vegetation drought stress index; secondary indicators for hydrological conditions such as groundwater level suitability, flooding time suitability, and surface water-groundwater connectivity; secondary indicators for topographic features such as topographic slope and aspect; and secondary indicators for soil quality such as soil organic matter content, soil moisture content, and soil salinization degree. The weights of each secondary indicator are determined by combining the analytic hierarchy process (AHP) with the entropy weight method.

5. The method according to claim 1, characterized in that, The machine learning model used in the coupling analysis in step S4 is a random forest model; the features input to the model include at least: flooding time standardized score, groundwater level standardized score, vegetation diversity score, soil salinity score, slope score, aspect score, and comprehensive score obtained by multi-factor weighted summation; the model output is the category code corresponding to different suitable plant zones.

6. The method according to claim 1, characterized in that, When generating the final shoreline ecological zoning scheme map in step S4, the following steps are further included: aligning and fusing the vector data of the preliminary ecological zoning results with the DEM raster data; performing median filtering on the fused raster map to eliminate isolated pixels, and merging small patches according to patch area; the final output zoning map includes zoning code, suitable vegetation type and habitat quality level attribute information.

7. A riverbank ecological zoning device for implementing the method as described in any one of claims 1-6, characterized in that, include: The data processing unit is used to collect hydrological, topographic, and remote sensing data of the target riverbank. After preprocessing and frequency analysis, the Weibull formula is used to calculate the peak flow under different super-frequency flood return periods. A model of the relationship between peak flow and inundation depth is established to calculate the inundation time of each spatial unit of the riverbank. Based on the on-site control point data, the spatial distribution map of flooding time is generated by spatial interpolation. The remote sensing image processing unit is used to acquire multi-source remote sensing images and extract vegetation type, coverage and topographic information of the beach; based on the flooding time and space distribution map generated by the data processing unit, and combined with the flood tolerance characteristics of vegetation, the beach space is divided into multiple suitable vegetation zones, and a suitable vegetation zone map is generated. The habitat evaluation unit is used to construct a habitat quality evaluation index system that includes vegetation status, hydrological conditions, topographic features, and soil quality. For each suitable plant zone divided in step S2, data of various indicators are collected or calculated, the habitat quality index of each zone is calculated using the comprehensive index method, and the quality level is rated based on the habitat quality index. An ecological zoning generation unit is used to couple the plant suitability zoning map output by the remote sensing image processing unit with the habitat quality rating result output by the habitat evaluation unit. Relevant factors are used as features to input into the machine learning model for training and prediction to obtain preliminary ecological zoning prediction results. By combining digital elevation models to optimize terrain constraints and spatial continuity, the final shoreline ecological zoning scheme map is generated.

8. The apparatus as claimed in claim 7, characterized in that, Also includes: The data input interface is used to receive user-input hydrological station data, remote sensing image data, field survey data, and topographic data. The results display and output module is used to visualize the flooding time distribution map, plant suitability zone map, habitat quality rating map and final ecological zoning scheme map, and supports map genus query and export functions.

Images