Method for evaluating soil hydraulic parameters based on a pedotransfer function model

Abstract

The application discloses a kind of based on ensemble soil conversion function model's soil hydraulic parameter evaluation method, including the following steps: determining the required predictor and multiple pressure water head, from global soil basic attribute database NCSS extraction corresponding pressure water head under measured soil moisture content data and predictor;Pressure water head and corresponding soil moisture content observation value are formed into data pair, and bootstrap resampling method is used to extract data;Select four existing single PTF model, calculate the parameters of each single PTF model respectively;Each parameter of single PTF is endowed with a weight, and the value of each parameter in ensemble PTF model is calculated respectively, so as to obtain weighted ensemble PTF model;The ensemble PTF model is verified, and the calculated soil moisture content is compared with the measured soil moisture content under corresponding pressure water head to verify.

Description

Technical Field

[0001] This invention relates to the field of soil measurement, and specifically to a method for evaluating soil hydraulic parameters based on a set soil transformation function model. Background Technology

[0002] Soil moisture characteristic curves describe the relationship between soil matric potential and soil moisture content, and are an important foundation for studying soil moisture movement. The Van Genuchten equation (hereinafter referred to as the VG equation) is a widely used equation to describe soil moisture characteristic curves.

[0003] Currently, the main methods for obtaining parameters in the VG equation are direct measurement and indirect methods. Direct measurement requires collecting soil samples in the field and determining soil moisture characteristic curves through laboratory experiments. Direct measurement methods mainly include the tensiometer method, pressure membrane method, equilibrium vapor pressure method, and centrifuge method. The advantage of direct measurement is that it can accurately determine the soil moisture characteristic curve, reflecting the true soil hydraulic properties. The disadvantages are that it requires sophisticated experimental equipment, most instruments have long measurement cycles, the measurement process is cumbersome, and it consumes a lot of manpower and resources. Furthermore, due to its limitations, direct measurement is difficult to apply to large-scale soil moisture simulation studies. Indirect methods utilize easily measurable basic soil physicochemical properties to indirectly obtain soil moisture characteristic curves. The most commonly used method is to obtain soil hydraulic parameters through the Pedotransfer function (PTF). The PTF establishes the relationship between information on basic soil properties such as particle composition, bulk density, and organic matter content and soil hydraulic parameters. It uses easily obtainable physicochemical properties of the soil to predict soil hydraulic parameters. This method can compensate for the lack of spatial representativeness in field sampling experiments and is an efficient method for obtaining soil hydraulic parameters.

[0004] Currently, many scholars have developed numerous Predictive Data Forms (PTFs) based on datasets from different regions and using various algorithms. However, most PTFs are constructed based on specific regions or rely on specific databases. Due to the strong spatial heterogeneity of soil, PTFs constructed based on specific regions have regional applicability. Most PTFs currently do not provide their applicable scope, and their prediction accuracy outside their construction region is often unknown. Using only a single PTF to predict parameters in the VG equation cannot fully reflect and explain the spatial heterogeneity of soil. Therefore, how to expand the applicability of PTFs and improve their prediction accuracy based on existing PTFs is a crucial problem that urgently needs to be solved. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a method for evaluating soil hydraulic parameters based on an ensemble soil transformation function model. This method utilizes the NCSS global soil fundamental properties database and employs a genetic algorithm combined with bootstrap resampling to construct an ensemble PTF model coupling multiple individual soil transformation functions (PTFs). This aims to reduce the impact of the regional applicability of the PTFs on the calculation results and expand the scope of application of the PTFs. Subsequently, an independent subset of the NCSS database is used to validate the ensemble PTF model, thereby efficiently and accurately obtaining the parameters in the VG equation.

[0006] A method for evaluating soil hydraulic parameters based on a ensemble soil transformation function model includes the following steps:

[0007] Step 1: Determine the required predictive factors and multiple pressure heads. Extract the measured soil moisture content data and predictive factors under the corresponding pressure heads from the Global Soil Basic Properties Database (NCSS). The predictive factors include the percentage content of sand, silt, and clay, soil bulk density, and soil organic matter content.

[0008] Step 2: Combine the pressure head and corresponding soil moisture content observation values ​​in the soil samples selected in Step 1 into data pairs. Use the bootstrap resampling method to extract data. Divide the above data pairs into two parts: one part is used as a training subset for training, and the other part is used as a validation subset for validation. Both the training subset and the validation subset include the prediction factor and the measured soil moisture content data under the corresponding pressure head.

[0009] Four existing single PTF models based on VG equations were selected: the Rawls PTF model, the Wosten PTF model, the Weynants PTF model, and the Rosetta3 PTF model.

[0010] The predictive factors extracted in step one are input into each individual PTF model to calculate the parameters in the VG equation; the VG equation is as follows:

[0011]

[0012] The parameters include soil residual moisture content θ r Soil saturated water content θ s The parameter α, which is related to the reciprocal of the soil air intake value, and the shape parameter n, which is related to the soil pore distribution;

[0013] Step 3: Assign a weight to each parameter of a single PTF and calculate the value of each parameter involved in the ensemble PTF model using the following formula:

[0014]

[0015] In the formula, N represents the value of a parameter in the VG ensemble model. m It is the number of PTFs, β i δ is the weight of this parameter. i The value of this parameter corresponds to the value of a single PTF; the weighted set PTF model can be obtained by calculating using formula (2).

[0016] Step 4: Validate the ensemble PTF model

[0017] The ensemble PTF model obtained in step three is validated using the validation subset obtained in step two. The predictors in the validation subset are substituted into the ensemble PTF model to obtain the hydraulic parameters of each soil sample in the validation subset, thereby obtaining the corresponding soil moisture characteristic curve. Then, the pressure head in the corresponding sample is substituted into the VG equation to calculate the corresponding soil moisture content. Finally, the calculated soil moisture content is compared with the measured soil moisture content under the corresponding pressure head for validation.

[0018] Furthermore, the weights of each parameter of the single PTF are determined by minimizing the objective function using a genetic algorithm with least squares. The objective function is as follows:

[0019]

[0020] Where, N d θ represents the number of measured soil moisture contents in soil samples from the NCSS database; i (h i ) represents the measured soil moisture content, θ i ′(h i ) represents the predicted soil moisture content, and P is the weight of the soil hydraulic parameters obtained for each PTF.

[0021] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0022] The soil hydraulic parameter evaluation method based on the ensemble soil transformation function model described in this invention is based on the NCSS global soil basic attribute database. It uses a genetic algorithm combined with the bootstrap resampling method to construct an ensemble PTF model that couples multiple individual PTFs. This aims to reduce the impact of the regional applicability of PTFs on the calculation results, overcome the problems of systematic bias, underestimation of uncertainty, and overestimation of predictive ability in the predicted soil hydraulic parameters caused by using a single PTF model, and expand the scope of application of PTFs. Then, the ensemble PTF model is validated using an independent subset of the NCSS database, thereby efficiently and accurately obtaining the parameters in the VG equation. Attached Figure Description

[0023] Figure 1 This is a flowchart of the soil hydraulic parameter evaluation method based on the ensemble soil transformation function model described in this invention. Detailed Implementation

[0024] To make the objectives, technical solutions, beneficial effects, and significant advancements of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings provided in the examples of the present invention. Obviously, all the described embodiments are only some embodiments of the present invention, and 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.

[0025] This invention, based on the NCSS global soil fundamental properties database, constructs an ensemble PTF model using a genetic algorithm combined with a bootstrap resampling method. The ensemble PTF model is then validated using an independent subset of the NCSS database. Multiple evaluation criteria are employed to assess the performance of the ensemble PTF model and commonly used single PTF models applied to the NCSS database. Specifically, the soil hydraulic parameter evaluation method based on the ensemble soil transformation function model includes, for example: Figure 1 The following steps are shown:

[0026] Step 1: Determine the required predictive factors and pressure head based on actual needs, and extract the measured soil moisture content data and predictive factors under the corresponding pressure head from the Global Soil Basic Properties Database (NCSS).

[0027] The NCSS encompasses measured data on basic soil properties from around the world. In this embodiment, available data includes 63,565 soil samples with multiple soil profiles, totaling 397,212 soil layers; however, data meeting the requirements for assessing PTF (Potentially Tolerant Float) is limited. After screening, this invention selected soil samples with at least one measured soil moisture content data point at pressure heads of -0.06, -0.1, -0.33, -1, -2, and -15 bar, and containing information on soil texture, bulk density, and organic carbon content. Furthermore, outliers potentially caused by data entry errors were removed. Based on the above data screening criteria, a total of 49,855 soil samples (containing 118,599 soil water holding data points) are available for further analysis. The selected data includes 43,613 soil samples from the United States and 6,242 soil samples from other regions.

[0028] Among them, the predictor factors required to extract the parameters in the PTF prediction VG equation from the NCSS database include the percentage content of sand, silt, and clay, soil bulk density, and soil organic matter content.

[0029] Step 2: Combine the pressure head and corresponding soil moisture content observation values ​​in the soil samples selected in Step 1 into data pairs. Use the bootstrap resampling method to extract data, setting 100 subsets to be taken from the total dataset. Divide the above data pairs into two parts: one part is used as a training subset for training, and the other part is used as a validation subset for validation. Both the training subset and the validation subset include the prediction factor and the measured soil moisture content data under the corresponding pressure head.

[0030] Four commonly used existing single PTF models based on VG equations were selected: the Rawls PTF model, the Wosten PTF model, the Weynants PTF model, and the Rosetta3 PTF model.

[0031] The Rosetta3PTF model is a PTF model constructed using a neural network.

[0032] The Rawls PTF, Wosten PTF, and Weynants PTF models are constructed based on regression analysis, and their formulas are as follows:

[0033] Rawls model:

[0034] θ s =1 - BD / 2.65

[0035] θ r =-0.0182482+0.00087269(sand)+0.00513488(clay)+0.02939286(θ s -0.00015395 (clay) 2 - 0.0010827(sand)(θ s -0.00018233 (clay) 2 (θ s ) 2 +0.00030703 (clay) 2 (θ s -0.0.0023584(clay)(θ) s ) 2

[0036] α=1 / exp[5.3396738+0.1845038(clay)-2.48394546(θ s -0.00213853 (clay) 2 -0.04356349(sand)(θ s )- 0.61745089(clay)(θ s) + 0.00143598 (sand) 2 (θ s ) 2 - 0.00855375 (clay) 2 (θ s ) 2 - 0.00001282 (sand) 2 (clay) + 0.00895359 (clay) 2 (θ s ) - 0.00072472 (sand) 2 (θ s ) + 0.0000054 (clay) 2 (sand) + 0.5002806 (clay)(θ s ) 2

[0037] n = 1 + exp[-0.7842831 + 0.0177544 (sand) - 1.062498 (θ s ) - 0.00005304 (sand)2 - 0.00273493 (clay) 2 + 1.11134946 (θ s ) 2 - 0.03088295 (sand)(θ<� s ) + 0.00026587 (sand) 2 (θ s ) 2 - 0.00610522 (clay) 2 (θ s ) 2 - 0.00000235 (sand) 2 (clay) + 0.00798746 (clay) 2 (θ s ) - 0.00610522 (clay) 2 (θ s ) 2

[0038] Wosten model:

[0039] θ s = 0.7919 + 0.001691 (clay) - 0.29619 (BD) - 0.000001491 (silt) 2 + 0.0000821 (SOM) 2 ​+0.02427 / (clay)+0.01113 / (silt)+0.01472ln(silt)- 0.0000733(SOM)(clay)-0.000619(BD)(clay)-0.001183(BD)(SOM)-0.0001664(topsoil)(silt)

[0040] θ r = 0

[0041] α = exp[-14.96+0.03135(clay)+0.0351(silt)+0.646(SOM)+15.29(BD)-0.192(topsoil)-4.671(BD) [[ID=...]] 2 - 0.000781(clay) 2 -0.00687(SOM) 2 +0.0449 / (SOM)+0.0663ln(silt)+0.1482ln(SOM)- 0.4546(BD)(silt)-0.4582(BD)(SOM)+0.00673(topsoil)(clay)]

[0042] n = 1.0+exp[-25.23-0.02195(clay)+0.0074(silt)-0.1940(SOM)+45.5(BD)-7.24(BD) 2 +0.0003658(clay) 2 +0.002885(SOM) 2 -12.81 / (BD)-0.1524 / (silt)-0.01958 / (SOM)- 0.2876ln(silt)-0.0709ln(SOM)-44.6ln(BD)- 0.02264(BD)(clay)+0.0896(BD)(SOM)+0.00718(topsoil)(clay)

[0043] Weynants model:

[0044] θ s = 0.06355+0.0013(clay)-0.1631(BD)

[0045] θ r = 0

[0046] α=exp[-4.3003-0.0097(clay)+0.0138(sand)-0.0097(clay)+0.0138(sand)-0.00992(SOC)]

[0047] n=1.0+exp[-1.0846-0.0236(clay)-0.0085(sand)+0.0001(sand) 2 ]

[0048] The predictive factors extracted in step one are input into each of the single PTF models to calculate the four parameters in the VG equation of Equation 1.

[0049] The following is the expression for the VG equation:

[0050]

[0051] In the formula, θ represents the soil moisture content (cm). 3 / cm 3 ), involving 4 parameters: θ r θ s These are the soil residual moisture contents (cm) 3 / cm 3 ) and soil saturated water content, α is a parameter that is related to the reciprocal of soil air intake value, and n is a shape parameter that is related to soil pore distribution.

[0052] Step 3: Calculate the weights of the parameters obtained for each individual PTF model.

[0053] Each parameter of each individual PTF is assigned a weight, and the value of each parameter in the ensemble PTF model is calculated using the following formula:

[0054]

[0055] In the formula, N represents the value of a parameter in the VG ensemble model. m It is the number of PTFs, β i δ is the weight of this parameter. i This corresponds to the value of this parameter in a single PTF. The weighted ensemble PTF model can be obtained by calculating using formula (2).

[0056] Where β i The weights are determined by minimizing the objective function using a genetic algorithm. The objective function is as follows:

[0057]

[0058] Where, N dN represents the number of soil samples with measured soil moisture content in the NCSS database, and N represents the total number of samples in the entire dataset. d =118599. θ i (h i ) represents the measured soil moisture content, θ i ′(h i ) represents the predicted soil moisture content, and P is the weight of the soil hydraulic parameters calculated for each PTF, corresponding to β in Equation 2. The VG ensemble model consists of 4 individual PTFs, and each PTF calculates 4 parameters (θ). r θ s There are 16 in total (α, n).

[0059] The expression for the PTF model of the set after determining the weights is as follows:

[0060] θ r =0.262124012Rawls_θ r +0.223882394Wosten_θ r +0.213324577Weynants_θ r +0.300669017Rose tta3_θ r

[0061] θ s =0.086589213Rawls_θ s +0.340702917Wosten_θ s +0.53371747Weynants_θ s +0.0389904Rosetta3 _θ s

[0062] α=0.055785648Rawls_α+0.236500256Wosten_α+0.029478989Weynants_α+0.678235106Rosetta 3_α

[0063] n=0.230863636Rawls_n+0.325687661Wosten_n+0.320355963Weynants_n+0.12309274Rosetta3_ n

[0064] Step 4: Validate the ensemble PTF model

[0065] The ensemble PTF model obtained in step three was validated using the validation subset extracted by the bootstrap resampling method in step two. Specifically, the predictors (including soil texture, soil bulk density, soil organic matter, etc.) in the validation subset were substituted into the newly constructed ensemble PTF model to obtain the hydraulic parameters of each soil sample in the validation subset, thereby obtaining the corresponding soil moisture characteristic curve. Then, the pressure head in the corresponding sample was substituted into the VG equation to calculate the corresponding soil moisture content. Finally, the calculated soil moisture content was compared with the measured soil moisture content under the corresponding pressure head for validation.

[0066] Experimental verification

[0067] The results of individual PTF and ensemble PTF models were compared using root mean square error (RMSE), mean error (ME), and coefficient of determination (R²). 2 The three evaluation indicators were used to evaluate the ensemble PTF model, and the results are shown in Table 1.

[0068] Table 1: Comparison of Single PTF and Ensemble PTF Models

[0069]

[0070] Among them, the closer the root mean square error is to 0, the smaller the deviation between the simulated and observed values, and the better the simulation effect. ME represents the overall deviation between the simulated and observed values; the closer the ME is to 0, the closer the simulated values ​​are to the measured values. Furthermore, the coefficient of determination R... 2 To determine the model's fit, R0 2 The closer the value is to 1, the better the model fits the measured data.

[0071] The results showed that the RMSE of a single PTF ranged from 0.056 to 0.093 cm⁻¹. 3 / cm 3 Between, R 2 The values ​​are between 0.730 and 0.790. The RMSE and ME values ​​of the ensemble PTF model are 0.051 and -0.001 cm, respectively. 3 / cm 3 It is less than the RMSE and ME values ​​of all individual PTFs, while R 2 The value is 0.819, which is greater than the R of all individual PTFs. 2 These values ​​reflect that the simulation results of the ensemble PTF model are more accurate. Furthermore, the RMSE, ME, and R values ​​of the training and validation sets are also more accurate. 2The values ​​showed little difference, indicating that the constructed ensemble PTF model is relatively stable. This ensemble PTF model was trained and validated based on a global soil database and constructed using a combination of genetic algorithms and bootstrap resampling, resulting in higher accuracy and wider applicability compared to a single PTF model.

[0072] The soil hydraulic parameter evaluation method of the present invention is relatively simple. Table 2 shows the prediction factors required for the selected PTF. After simple processing of the target soil sample in the laboratory, the basic properties of the target soil sample can be obtained through basic experiments and simple calculations. By inputting these basic properties into the ensemble PTF model, the parameters in the VG equation can be obtained quickly and accurately.

[0073] Table 2: The four PTFs and their predictive factors used in this invention

[0074]

[0075] The comparison between the soil moisture content estimated by the ensemble PTF model used in this invention and the observed soil moisture content shows good results. Compared with the single PTF widely used in current Earth system models, the ensemble PTF model can provide more accurate soil moisture content estimates. This is beneficial for providing more accurate prediction results and reducing prediction uncertainty for global climate models, hydrological and ecological models, crop growth models, land surface models, weather forecasting models, and air quality models. Furthermore, the effective ensemble PTF model can provide soil hydraulic parameters for areas lacking basic soil property data, enabling soil water-related research in these regions.

[0076] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention. Non-essential improvements, adjustments or substitutions made by those skilled in the art based on the content of this specification are all within the scope of protection claimed by the present invention.

Claims

1. A method for evaluating soil hydraulic parameters based on a ensemble soil transformation function model, characterized in that, Includes the following steps: Step 1: Determine the required predictive factors and multiple pressure heads. Extract the measured soil moisture content data and predictive factors under the corresponding pressure heads from the Global Soil Basic Properties Database (NCSS). The predictive factors include the percentage content of sand, silt, and clay, soil bulk density, and soil organic matter content. Step 2: Combine the pressure head and corresponding soil moisture content observation values ​​in the soil samples selected in Step 1 into data pairs. Use the bootstrap resampling method to extract data. Divide the above data pairs into two parts: one part is used as a training subset for training, and the other part is used as a validation subset for validation. Both the training subset and the validation subset include the prediction factor and the measured soil moisture content data under the corresponding pressure head. Four existing single PTF models based on VG equations were selected: the Rawls PTF model, the Wosten PTF model, the Weynants PTF model, and the Rosetta3 PTF model. The predictive factors extracted in step one are input into each individual PTF model to calculate the parameters in the VG equation; the VG equation is as follows: (Equation 1) The parameters include residual soil moisture content. Soil saturated water content Parameters related to the reciprocal of soil air intake value Shape parameters associated with soil pore distribution ; Step 3: Assign a weight to each parameter of a single PTF and calculate the value of each parameter involved in the ensemble PTF model using the following formula: (Equation 2) In the formula, For the value of a parameter in the VG ensemble model, It is the number of PTFs. The weight of this parameter, The value of this parameter corresponds to the value of a single PTF; the weighted set PTF model is obtained by calculating using formula (2); Step 4: Validate the ensemble PTF model The PTF model set obtained in step three is validated using the validation subset obtained in step two. The predictors in the validation subset are substituted into the ensemble PTF model to obtain the hydraulic parameters of each soil sample in the validation subset, thereby obtaining the corresponding soil moisture characteristic curve. Then, the pressure head in the corresponding sample is substituted into the VG equation to calculate the corresponding soil moisture content. Finally, the calculated soil moisture content is compared and verified with the measured soil moisture content under the corresponding pressure head.

2. The method for evaluating soil hydraulic parameters based on a ensemble soil transformation function model according to claim 1, characterized in that, The weights of each parameter of the single PTF are determined by minimizing the objective function using a genetic algorithm with least squares. The objective function is as follows: (Equation 3) in, This represents the number of measured soil moisture contents contained in soil samples from the NCSS database. To measure soil moisture content, P represents the predicted soil moisture content, and P is the weight of the soil hydraulic parameters calculated for each PTF.

Images