An improved method of forest structure parameter fusion runoff model

Abstract

The application belongs to the technical field of forest ecological hydrological modeling, and discloses an improved runoff generation model fusing forest structure parameters. The method comprises the following steps: dividing the forest in the research area into six types based on a two-dimensional classification system of forest origin and forest age stage; establishing a nonlinear coupling relationship between the crown water storage capacity and the leaf area index and the crown height, and improving the calculation module of the forest canopy interception; introducing a forest age adjustment coefficient based on the Logistic function to correct the remote sensing evapotranspiration data product; realizing the fine calculation of the runoff on the grid scale; and adopting a multi-source data cross-validation strategy to evaluate the model precision. Through the rigid embedding of the fine forest structure parameters into the core modules of the forest canopy interception and evapotranspiration of the hydrological model, the application overcomes the systematic estimation deviation caused by the parameter homogenization of the traditional model, and realizes the spatial fine expression of the runoff generation process in the complex forest habitat.

Description

Technical Field

[0001] This invention belongs to the field of forest eco-hydrological modeling technology, and relates to an improved method for runoff generation model that integrates forest structural parameters. Background Technology

[0002] The quantitative simulation of traditional forest eco-hydrological processes has long relied on a dual-track approach of empirical statistical models and mechanistic physical models. However, both face significant bottlenecks in expressing mechanisms under complex environments. Empirical statistical models are primarily based on extensive observational data, establishing empirical relationships between macroscopic forest characteristic parameters (such as leaf area index (LAI) and vegetation cover) and hydrological flux indicators through statistical regression. For example, the classic Gash interception model uses simplified empirical formulas to calculate canopy interception based on parameters such as canopy water storage capacity and rainfall intensity; the Budyko hydrothermal coupling framework establishes empirical relationships between regional evapotranspiration and precipitation, as well as potential evapotranspiration, through aridity indices. These models are computationally simple and suitable for rapid assessment in data-scarce regions, but their parameters often exhibit regional specificity, making it difficult to directly generalize to different forest types and climatic conditions.

[0003] Mechanistic physical models are based on the physical laws of hydrological cycle and energy balance, simulating forest hydrological processes by solving a system of partial differential equations. Representative models include distributed hydrological models such as SWAT and VIC, and land surface process models such as CLM. These models consider the coupling of multiple physical processes such as rainfall interception, soil infiltration, evapotranspiration, and slope runoff, and can reflect the hydrological cycle mechanism well. However, physical models have high requirements for input data, with numerous parameters, and require detailed spatial distribution information such as forest structure, soil texture, and topography.

[0004] Traditional models often assume relatively homogeneous forest stand properties, using a few macroscopic parameters to characterize the forest's impact on precipitation and evapotranspiration, which fails to reflect the high heterogeneity of forest internal structure and components. On the one hand, even if different forests have similar macroscopic parameters such as forest stand area index (LAI), their canopy structures can vary greatly, resulting in significantly different characteristics in precipitation interception and evapotranspiration. On the other hand, the spatial heterogeneity of forest vertical structure has a profound impact on hydrological processes; the canopy interception rate of different structures can range from 10% to 50% of rainfall, showing a huge difference.

[0005] Regarding evapotranspiration estimation, while existing remote sensing evapotranspiration products (such as MOD16) have considered the influence of vegetation parameters such as leaf area index, their canopy conductance parameters are set based on a globally unified vegetation classification scheme, failing to distinguish the differences in transpiration capacity at different age stages within the same vegetation type. Studies have shown that tree transpiration capacity exhibits a non-monotonic variation with forest age: young forests have shallow root systems and small canopies, resulting in weaker transpiration capacity; middle-aged forests are in their vigorous growth period, reaching peak transpiration; and mature forests show a slight decrease in transpiration capacity but still maintain a high level. This age effect has not been fully reflected in traditional evapotranspiration products.

[0006] In summary, the standardized parameter settings of traditional runoff generation models inevitably lead to systematic drift in runoff estimation when dealing with forest ecosystems with extremely heterogeneous structures. Therefore, there is an urgent need for an improved method that can integrate the three-dimensional structural characteristics of forests into the core calculation module of runoff generation models. Summary of the Invention

[0007] The purpose of this invention is to provide an improved method for runoff generation model that integrates forest structure parameters. By establishing a nonlinear coupling relationship between runoff generation calculation parameters and forest structure characteristics (canopy height, leaf area index, and stand age), the method achieves a spatially refined expression of the runoff generation process in complex forest habitats, overcoming the systematic estimation bias caused by parameter homogenization in traditional models.

[0008] The technical solution of the present invention: An improved method for runoff generation model incorporating forest structure parameters, comprising the following steps: A two-dimensional classification system was established based on the forest origin attribute and age stage, dividing the forests in the study area into natural forests and plantations. Natural forests were further divided into natural young forests, natural middle-aged forests, and natural mature forests, while plantations were divided into artificial young forests, artificial middle-aged forests, and artificial mature forests. For each type of forest, a nonlinear coupling relationship between canopy water storage capacity and leaf area index and canopy height is established, and the canopy interception calculation module is improved. An evapotranspiration age correction module is constructed by introducing a forest age adjustment coefficient to correct remote sensing evapotranspiration data products. At the grid scale, the flow rate of each grid cell is calculated based on the water balance equation; A multi-source data cross-validation strategy was adopted to evaluate the reliability of the improved model at both the watershed macro-scale and the regional scale.

[0009] Furthermore, the nonlinear coupling relationship between canopy water storage capacity and leaf area index and canopy height is established as follows: in, The water storage capacity of the canopy is expressed in mm. Leaf area index, in m²2 / m 2 ; The height of the canopy is in meters (m). The average height of the study area was used as a reference. The basic water storage coefficient reflects the basic water storage capacity per unit leaf area; The leaf area power exponent reflects the nonlinear saturation effect of water storage capacity as leaf area increases. <1; The canopy height adjustment coefficient quantifies the additional contribution of the canopy's three-dimensional structure to water storage capacity.

[0010] Furthermore, differentiated parameter combinations are adopted for different forest types, specifically: Natural young forests: =0.16, =0.75, =0.10; Natural middle-aged forests: =0.18, =0.80, =0.15; Natural mature forest: =0.20, =0.82, =0.18; Artificial young forests: =0.13, =0.65, =0.06; Artificial Middle-Aged Forests: =0.15, =0.70, =0.10; Mature plantations: =0.17, =0.73, =0.12.

[0011] Furthermore, the method of introducing a forest age adjustment coefficient to correct remote sensing evapotranspiration data products includes: The evapotranspiration rate is corrected using the following formula: in, The corrected evaporation rate is expressed in mm. Estimated values ​​for MOD16 remote sensing evapotranspiration data products, in mm; ) The age adjustment function is used. The tree age adjustment function adopts the form of a Logistic function: in, Forest age, in years; and These are the minimum adjustment coefficient and the maximum adjustment coefficient, respectively. This represents a forest age where transpiration capacity has reached a moderate level. The rate of change parameter controls the steepness of the curve.

[0012] Furthermore, different age adjustment function parameters are used for natural forests and plantations: Natural forests: =0.78, =1.08, =45 years =0.06; Planted forests: =0.72, =1.05, =15 years, =0.12.

[0013] Furthermore, the calculation of the flow rate of each grid cell based on the water balance equation at the grid scale includes: For each grid cell The production flow rate is calculated as follows: in, For precipitation, This is the corrected evaporation rate. Canopy interception; The calculation of canopy interception is combined with the precipitation intensity adjustment function: in, Canopy coverage; The water storage capacity is calculated using the canopy water storage capacity formula; This is a precipitation intensity adjustment function; To intercept saturated precipitation.

[0014] Furthermore, an upper limit constraint of no more than 35% of precipitation is set for the canopy interception, and a non-negative constraint is set for the runoff of each grid unit.

[0015] Furthermore, the two-dimensional classification system established based on forest origin attributes and forest age stages includes: The origin of forests is determined based on spatial distribution data of natural and planted forests; The age classification system is categorized into three levels: young forests, middle-aged forests, and mature forests. The age classification thresholds for natural forests are: 0-20 years old as young forests, 21-60 years old as middle-aged forests, and 61 years and older as mature forests. The age classification thresholds for planted forests are: 0-10 years old as young forests, 11-25 years old as middle-aged forests, and 26 years and older as mature forests. Spatial distribution maps of six types of forests were generated using a raster overlay method.

[0016] Furthermore, the improved runoff generation model method integrating forest structure parameters also includes determining the runoff direction based on the D8 algorithm and calculating the runoff accumulation in each grid cell: in, For grid cells The cumulative flow of the confluence; For grid cells The set of all upstream meshes; To obtain from grid cells To grid cell The transmission coefficient.

[0017] 10. The method for improving a runoff generation model by integrating forest structure parameters according to claim 1, characterized in that the multi-source data cross-validation strategy includes: At the macro scale of the watershed, the monthly runoff output by the model is compared with the measured monthly runoff data from the hydrological station. Parameter calibration is performed using calibration period data, and accuracy is evaluated using independent validation period data. At the regional scale, the improved evapotranspiration estimates are compared with global land surface evapotranspiration data products for spatial consistency. The Monte Carlo method was used to conduct sensitivity analysis on key parameters in the interception module and evapotranspiration module to assess the impact of parameter uncertainty on the production flow calculation results.

[0018] The beneficial effects of this invention are: This invention embeds forest structure parameters (canopy height, leaf area index, and stand age) into the core modules of the runoff generation model, specifically the canopy interception and evapotranspiration modules. It establishes nonlinear response equations between canopy interception parameters and the three-dimensional canopy structure, and a logistic regulation relationship between evapotranspiration parameters and stand age development stages. This overcomes the systematic estimation bias caused by the parameter homogenization assumption in traditional models. During the validation period at three hydrological stations, the improved model showed an average improvement of 0.06 in the Nash-Sutcliffe efficiency coefficient (NSE), an average reduction of approximately 13% in the root mean square error, and an average improvement of 0.05 in the Kling-Gupta efficiency coefficient (KGE). The accuracy improvement remained stable during both the calibration and validation periods. This method possesses a clear physical mechanism, good time extrapolation robustness, and compatibility with remote sensing data, making it applicable to runoff generation simulation in other complex forest ecosystems.

[0019] This invention rigidly embeds refined forest structure parameters (canopy height, leaf area index, and stand age) into the core modules of the runoff generation model, specifically the canopy interception and evapotranspiration modules. It establishes nonlinear response equations between interception parameters and the three-dimensional canopy structure, and a Logistic regulation relationship between evapotranspiration parameters and stand age development stages. This overcomes the systematic estimation bias caused by the parameter homogenization assumption in traditional models. During the validation period at three hydrological stations, the improved model showed an average improvement of 0.06 in the Nash-Sutcliffe efficiency coefficient, an average reduction of approximately 13% in the root mean square error, and an average improvement of 0.05 in the Kling-Gupta efficiency coefficient. The accuracy improvement remained stable during both the calibration and validation periods. This method possesses a clear physical mechanism, good time extrapolation robustness, and compatibility with remote sensing data, making it applicable to runoff simulation in other complex forest ecosystems. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating the improved runoff generation model method that integrates forest structure parameters as described in this invention.

[0021] Figure 2 The diagram shows the comparison of natural forest yield before and after the improvement. (a) represents the natural-young forest yield before the improvement, (b) represents the natural-young forest yield after the improvement, (c) represents the natural-middle-aged forest yield before the improvement, (d) represents the natural-middle-aged forest yield after the improvement, (e) represents the natural-mature forest yield before the improvement, and (f) represents the natural-mature forest yield after the improvement.

[0022] Figure 3The diagram shows the comparison of plantation yield before and after improvement. (a) is the yield of plantation-young forest before improvement, (b) is the yield of plantation-young forest after improvement, (c) is the yield of plantation-middle-aged forest before improvement, (d) is the yield of plantation-middle-aged forest after improvement, (e) is the yield of plantation-mature forest before improvement, and (f) is the yield of plantation-mature forest after improvement. Detailed Implementation

[0023] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings and technical solutions.

[0024] Example 1 Using Hainan Tropical Rainforest National Park as the study area, this paper provides a detailed explanation of the improved runoff generation model that integrates forest structure parameters as described in this invention.

[0025] I. Data Preparation; Forest structure parameters include: canopy height data, which is generated by fusing data from GEDI and ICESat-2 ATLAS dual-platform satellite lidar using a neural network-guided interpolation method, with a spatial resolution of 30m; forest age data, which is obtained by inversion using the XGBoost algorithm with canopy height data as the core driving feature, with a spatial resolution of 30m; and leaf area index (LAI) data, which is generated by spatial continuous raster mapping using GEDI L2B data products after correction with the help of a digital elevation model (SRTM DEM), with a resolution of 500m.

[0026] Meteorological and hydrological data include: CHIRPS satellite inverted precipitation data products (0.05° resolution, daily scale), MOD16A2 evapotranspiration data products (1km resolution, 8-day composite), and ERA5-Land reanalysis meteorological data (0.1° resolution).

[0027] The auxiliary data include: SRTM DEM data (30m resolution), land cover classification data, SoilGrids soil type data, and measured runoff data from hydrological stations.

[0028] II. Classification of Forest Types; A two-dimensional classification system combining forest origin and age stage was adopted. The age classification thresholds for natural forests were: 0-20 years old as young forests (NF-Y), 21-60 years old as middle-aged forests (NF-M), and over 60 years old as mature forests (NF-O). For plantations, the age classification thresholds were: 0-10 years old as young forests (PF-Y), 11-25 years old as middle-aged forests (PF-M), and over 25 years old as mature forests (PF-O). Spatial classification employed a raster overlay method: the binary maps of natural and plantations were resampled to a 30m resolution, the age raster was reclassified according to the thresholds, and finally, spatial overlay operations were used to generate six forest distribution maps.

[0029] III. Improved retention calculation; The improved formula for canopy water storage capacity is: The physical meaning of this formula is that the power function form of the leaf area index (β<1) reflects the diminishing marginal return of canopy water storage capacity with increasing leaf area; the logarithmic adjustment term of canopy height reflects that tall canopies have richer foliage layers and greater surface roughness, providing additional water storage space. The differentiated parameter settings for the six forest types are as follows: natural young forests α=0.16, β=0.75, γ=0.10, canopy coverage c=0.55, S c Range 0.6-1.2 mm; natural middle-aged forests α=0.18, β=0.80, γ=0.15, c=0.75, S c Range 1.2-2.0 mm; Natural mature forest α=0.20, β=0.82, γ=0.18, c=0.85, S c Range 1.8-2.8 mm; For young artificial forests, α=0.13, β=0.65, γ=0.06, c=0.50, S c Range 0.4-0.8 mm; For planted middle-aged forests, α=0.15, β=0.70, γ=0.10, c=0.70, S c Range 0.9-1.5 mm; In mature plantations, α=0.17, β=0.73, γ=0.12, c=0.78, S c Range 1.2-1.8mm.

[0030] IV. Evapotranspiration age correction; The corrected formula is: ,in Natural forest parameters: f min =0.78、 f max =1.08、 A mid =45 years, k=0.06, meaning that the evapotranspiration of natural young forests is approximately 78% of the MOD16 remote sensing evapotranspiration data product value, gradually increasing with forest age, reaching a moderate level around 45 years, and reaching 108% in mature forests. Plantation parameters: f min =0.72、 f max =1.05、 A mid =15 years, k=0.12, reflecting that the plantation forest reaches the mature state of transpiration capacity in a relatively short period of time.

[0031] V. Grid-based flow generation calculation; For each 30m×30m grid cell i, the flow rate is calculated as follows: Among them, the amount of canopy interception Precipitation intensity adjustment function , P sat The value is approximately 50 mm. An upper limit constraint of no more than 35% of precipitation is set for canopy interception, and a non-negative constraint is set for runoff. The runoff direction is determined based on the D8 algorithm, and the cumulative runoff volume is calculated. .

[0032] VI. Model Validation Results; Calibration (2013-2018) and validation (2019-2022) were conducted at the Songtao Station on the Nandu River, the Baoqiao Station on the Changhua River, and the Jiaji Station on the Wanquan River. The improved model achieved NSEs of 0.82, 0.85, and 0.78 at the three stations during the calibration period, with R... 2 The values ​​were 0.85, 0.88, and 0.82, respectively; the validation period NSEs were 0.76, 0.79, and 0.71, respectively, and R... 2 The values ​​were 0.80, 0.83, and 0.76, respectively. Compared to the baseline model (traditional homogenized parameters), the validation period NSE improved by 0.06, 0.07, and 0.06, respectively, and the RMSE decreased from 19.2, 14.8, and 22.4 mm / month to 16.8, 12.6, and 19.5 mm / month, respectively.

[0033] In terms of evapotranspiration validation, the spatial correlation coefficient with GLEAM v3.6 global evapotranspiration data products reached 0.81, with a root mean square deviation of 62.3 mm / year. Sensitivity analysis showed that runoff was most sensitive to precipitation input (±20% disturbance resulted in changes of ±22.3%–±25.1%), followed by the evapotranspiration regulation coefficient. f max (±6.8%-±7.5%) and leaf area power index β (±4.2%-±5.1%).

[0034] This document uses specific examples to illustrate the accounting framework and method of the present invention. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

[0035] Table 1. Differentiated interception and evapotranspiration parameter settings for six types of forests.

[0036] Note: α Basic water storage coefficient; βThe leaf area is the power exponent. γ This is the canopy height adjustment coefficient; c This represents a typical value for canopy coverage.

Claims

1. An improved method for runoff generation model incorporating forest structure parameters, characterized in that, The steps are as follows: A two-dimensional classification system was established based on the forest origin attribute and age stage, dividing the forests in the study area into natural forests and plantations. Natural forests were further divided into natural young forests, natural middle-aged forests, and natural mature forests, while plantations were divided into artificial young forests, artificial middle-aged forests, and artificial mature forests. For each type of forest, a nonlinear coupling relationship between canopy water storage capacity and leaf area index and canopy height is established, and the canopy interception calculation module is improved. An evapotranspiration age correction module is constructed by introducing a forest age adjustment coefficient to correct remote sensing evapotranspiration data products. At the grid scale, the flow rate of each grid cell is calculated based on the water balance equation; A multi-source data cross-validation strategy was adopted to evaluate the reliability of the improved model at both the watershed macro-scale and the regional scale.

2. The improved runoff generation model method incorporating forest structure parameters according to claim 1, characterized in that, The nonlinear coupling relationship between canopy water storage capacity and leaf area index and canopy height is established as follows: in, The water storage capacity of the canopy is expressed in mm. Leaf area index, in m² 2 / m 2 ; The height of the canopy is in meters (m). The average height of the study area was used as a reference. The basic water storage coefficient reflects the basic water storage capacity per unit leaf area; The leaf area power exponent reflects the nonlinear saturation effect of water storage capacity as leaf area increases. <1; The canopy height adjustment coefficient quantifies the additional contribution of the canopy's three-dimensional structure to water storage capacity.

3. The improved runoff generation model method incorporating forest structure parameters according to claim 2, characterized in that, Different parameter combinations are used for different forest types, specifically: Natural young forests: =0.16, =0.75, =0.10; Natural middle-aged forests: =0.18, =0.80, =0.15; Natural mature forest: =0.20, =0.82, =0.18; Artificial young forests: =0.13, =0.65, =0.06; Artificial Middle-Aged Forests: =0.15, =0.70, =0.10; Mature plantations: =0.17, =0.73, =0.

12.

4. The improved runoff generation model method incorporating forest structure parameters according to claim 1, characterized in that, The method of introducing a forest age adjustment coefficient to correct remote sensing evapotranspiration data products includes: The evapotranspiration rate is corrected using the following formula: in, The corrected evaporation rate is expressed in mm. Estimated values ​​for MOD16 remote sensing evapotranspiration data products, in mm; ) The age adjustment function is used. The tree age adjustment function adopts the form of a Logistic function: in, Forest age, in years; and These are the minimum adjustment coefficient and the maximum adjustment coefficient, respectively. This represents a forest age where transpiration capacity has reached a moderate level. The rate of change parameter controls the steepness of the curve.

5. The improved runoff generation model method incorporating forest structure parameters according to claim 4, characterized in that, Different age adjustment function parameters are used for natural forests and plantations: Natural forests: =0.78, =1.08, =45 years =0.06; Planted forests: =0.72, =1.05, =15 years, =0.

12.

6. The improved runoff generation model method incorporating forest structure parameters according to claim 1, characterized in that, The calculation of the flow rate of each grid cell based on the water balance equation at the grid scale includes: For each grid cell The production flow rate is calculated as follows: in, For precipitation, This is the corrected evaporation rate. Canopy interception; The calculation of canopy interception is combined with the precipitation intensity adjustment function: in, Canopy coverage; The water storage capacity is calculated using the canopy water storage capacity formula; This is a precipitation intensity adjustment function; To intercept saturated precipitation.

7. The improved runoff generation model method incorporating forest structure parameters according to claim 6, characterized in that, An upper limit constraint of no more than 35% of precipitation is set for canopy interception, and a non-negative constraint is set for the runoff of each grid cell.

8. The improved runoff generation model method incorporating forest structure parameters according to claim 1, characterized in that, The two-dimensional classification system based on forest origin attributes and forest age stages includes: The origin of forests is determined based on spatial distribution data of natural and planted forests; The age classification system is categorized into three levels: young forests, middle-aged forests, and mature forests. The age classification thresholds for natural forests are: 0-20 years old as young forests, 21-60 years old as middle-aged forests, and 61 years and older as mature forests. The age classification thresholds for planted forests are: 0-10 years old as young forests, 11-25 years old as middle-aged forests, and 26 years and older as mature forests. Spatial distribution maps of six types of forests were generated using a raster overlay method.

9. The improved runoff generation model method incorporating forest structure parameters according to claim 1, characterized in that, The improved runoff generation model method that integrates forest structure parameters also includes determining the runoff direction based on the D8 algorithm and calculating the runoff accumulation in each grid cell: in, For grid cells The cumulative flow of the confluence; For grid cells The set of all upstream meshes; To obtain from grid cells To grid cell The transmission coefficient.

10. The improved runoff generation model method incorporating forest structure parameters according to claim 1, characterized in that, The multi-source data cross-validation strategy includes: At the macro scale of the watershed, the monthly runoff output by the model is compared with the measured monthly runoff data from the hydrological station. Parameter calibration is performed using calibration period data, and accuracy is evaluated using independent validation period data. At the regional scale, the improved evapotranspiration estimates are compared with global land surface evapotranspiration data products for spatial consistency. The Monte Carlo method was used to conduct sensitivity analysis on key parameters in the interception module and evapotranspiration module to assess the impact of parameter uncertainty on the production flow calculation results.

Images