A method for predicting spatial distribution of soil respiration temperature sensitivity

Abstract

The application discloses a method for predicting spatial distribution of soil respiration temperature sensitivity, comprising the following steps: step 1, data acquisition and processing; step 2, calculating parameters of Van't Hoff index chemical reaction-temperature equation; step 3, calculating soil respiration temperature sensitivity Q 10 ; step 4, obtaining influence factor data of soil respiration temperature sensitivity Q 10 and processing; step 5, building a deep neural network model and training; step 6, determining spatial distribution of soil respiration temperature sensitivity Q 10 of a research area. The method greatly enriches the availability of existing data and significantly improves the ability to extract effective information from soil respiration data; at the same time, on the basis of enriching the availability of data, a deep neural network model is used to build a nonlinear relationship between soil respiration temperature sensitivity Q 10 and influence factors of Q 10 , and successfully extends point-scale Q 10 to regional scale, realizing prediction of spatial distribution of soil respiration temperature sensitivity.

Description

Technical Field

[0001] This invention belongs to the field of environmental monitoring technology, and in particular relates to a method for predicting the spatial distribution of soil respiration temperature sensitivity. Background Technology

[0002] The temperature sensitivity of soil respiration indicates the degree to which soil respiration responds to temperature changes, and is usually expressed as the multiple of the change in soil respiration (Q) when the temperature increases by 10°C. 10 The value is used to represent Q. 10 It is a key parameter for determining global terrestrial carbon dynamics under climate change; a higher value indicates that the soil respiration rate is more sensitive to temperature changes. 10 The accuracy of the Q value will determine the accuracy of Earth system models in predicting the intensity of carbon-climate feedback. Therefore, accurate parameterization of Q is crucial. 10 Estimating its spatial distribution is an important prerequisite for the ecosystem modeling process.

[0003] Currently, Q 10 The value is estimated by fitting a function using monitored soil respiration flux and temperature data. Therefore, Q 10 The determination of the value heavily relies on continuously monitored soil respiration flux and temperature data. However, in reality, the available continuous monitoring data is very limited, which leads to a limited pool of Q values ​​available for analysis. 10 Data is severely insufficient. Soil respiration carbon emissions represent the amount of carbon dioxide released by soil respiration over a period of time. While carbon emission data also contains crucial information on soil respiration, it can supplement the limited continuous monitoring data for calculating Q. 10 Value. However, to date, no data has been found to calculate Q from soil respiration carbon emissions. 10 An effective method. Furthermore, improving Q... 10 Data availability will also promote Q 10 Spatial distribution prediction of Q in a region. Current methods can predict the Q distribution of a region. 10 Spatial distribution can be predicted using methods such as terrestrial carbon models, climate-driven models, and random forests, but effective Q-factor prediction is lacking. 10 Value verification introduces significant uncertainty, especially in regions lacking data. Summary of the Invention

[0004] The purpose of this invention is to provide a method for predicting the spatial distribution of soil respiration temperature sensitivity in order to solve the above-mentioned technical problems.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] This invention discloses a method for predicting the spatial distribution of soil respiration temperature sensitivity, the method comprising the following steps:

[0007] Step 1: Data Acquisition and Processing: Acquire soil respiration monitoring data and daily temperature dataset for the study area. The soil respiration monitoring data includes two types of data: the first type is the daily average flux data of soil respiration, and the second type is the carbon emission data of soil respiration. Based on the location and time information of the acquired soil respiration monitoring data, extract the daily average temperature data from the daily temperature dataset.

[0008] Step 2: Calculate the parameters of the Van't Hoff exponential chemical reaction-temperature equation: The Van't Hoff exponential chemical reaction-temperature equation is used to describe the relationship between the daily average flux of soil respiration and the daily average air temperature.

[0009] Rs = ae bT (1)

[0010] In the formula, Rs is the average daily flux of soil respiration, in μmol Cm. -2 s -1 T represents the daily average temperature in °C; a is the first empirical parameter, and b is the second empirical parameter.

[0011] For the first type of data mentioned in step 1, the least squares method is used to fit the Van't Hoff exponential chemical reaction-temperature equation, i.e., formula (1), in order to determine the first empirical parameter a and the second empirical parameter b of the equation.

[0012] The least squares method finds the best-fit function for the data by minimizing the sum of squared errors, and its loss function is as follows:

[0013]

[0014] In the formula, L is the loss function of the least squares method, i.e., the sum of squared errors; (x i y i Let f(i) be the i-th sample; f() is the theoretical function with undetermined parameters;

[0015] For the second type of data mentioned in step 1, the first empirical parameter a and the second empirical parameter b are solved by the following method;

[0016] The daily average flux of soil respiration during the monitoring period is expressed as follows, based on the Van't Hoff index chemical reaction-temperature equation:

[0017]

[0018] In the formula, Rs i T represents the average daily soil respiration flux on day i; i This represents the average daily temperature on day i.

[0019] The daily soil respiration emission is calculated based on the average daily soil respiration flux, using the following formula:

[0020]

[0021] In the formula, This represents the daily soil respiration emissions on day i, expressed in gCm³. -2 m represents the mass of carbon per micromole, which is 12 × 10⁻⁶. -6 gμmol -1 t represents the number of seconds in a day, which is 86400 s; η is a conversion factor with a value of 1.0368, and the unit is g / μmol. -1 s;

[0022] The cumulative value of soil respiration carbon emissions during the monitoring period is calculated based on the daily soil respiration emissions, using the following formula:

[0023]

[0024] In the formula, P represents the number of days in the monitoring period, in days (d). The cumulative value of soil respiration carbon emissions within day P, in gCm. -2 ;

[0025] When two soil respiration carbon emission data exist at the same location but at different monitoring times, the values ​​of parameters a and b are obtained by solving the following set of nonlinear equations:

[0026]

[0027] In the formula, m and n represent different monitoring days, with the unit being days (d).

[0028] Step 3: Calculate the soil respiration temperature sensitivity Q 10 Based on the parameters of the Van't Hoff index chemical reaction-temperature equation obtained in step 2, the soil respiration temperature sensitivity Q was calculated. 10 The calculation formula is:

[0029] Q 10 =e 10b (7)

[0030] In the formula, Q 10 is the multiple of the change in soil respiration when the temperature increases by 10℃, used to represent the temperature sensitivity of soil respiration; b is the second empirical parameter;

[0031] Step 4: Obtain the soil respiration temperature sensitivity Q 10 Impact factor data and processing: Obtain Q within the study area 10 Impact factor data, for obtaining Q 10 The impact factor data is processed, that is, Q is first...10 The impact factor data is resampled to the required resolution, then outliers are removed and missing values ​​are imputed.

[0032] Step 5: Build and train a deep neural network model: This includes the following steps:

[0033] Step 51, Q calculated in step 3 10 As the label, with the calculated Q 10 Q at location 10 The influencing factors are used as features to form the training dataset;

[0034] Step 52: For Q in the training dataset 10 and Q 10 The impact factor data is normalized to generate a normalized training dataset;

[0035] Step 53: Build a deep neural network model and add L1 or L2 regularization to prevent the deep neural network model from overfitting.

[0036] Step 54: Use K-fold cross-validation combined with grid search to determine the optimal hyperparameter combination;

[0037] Step 55: Train a deep neural network model using the normalized training dataset and establish Q... 10 With Q 10 The mapping relationship between the influencing factors is used to obtain the predicted Q. 10 The regression model is the trained deep neural network model;

[0038] Step 6: Determine the soil respiration temperature sensitivity Q in the study area. 10 Spatial distribution: Q of all pixels at the required resolution of the study area 10 The influencing factor data is used as the prediction dataset, and the prediction dataset is normalized to generate a normalized prediction dataset. The generated normalized prediction dataset is then input into the trained deep neural network model to obtain the normalized Q-values ​​of the study area at the corresponding resolution. 10 Value; normalized Q 10 The value was inversely normalized to obtain the soil respiration temperature sensitivity Q over the entire study area. 10 The value, i.e., the soil respiration temperature sensitivity Q in the study area, is determined. 10 Spatial distribution.

[0039] Furthermore, the soil respiration monitoring data mentioned in step 1 is obtained through literature, public databases, or site monitoring; the daily temperature dataset comes from a public raster dataset, which is remote sensing data or reanalysis data.

[0040] Furthermore, Q mentioned in step 4 10 The influencing factor data include annual average temperature, annual precipitation, normalized vegetation index (NDVI), pH, soil organic carbon density (SOCD), elevation (DEM), and soil property data.

[0041] Furthermore, the soil property data includes soil bulk density, clay content, sand content, and silt content.

[0042] Furthermore, the required resolution described in step 4 is 1km × 1km.

[0043] Furthermore, the specific process of building the deep neural network in step 53 is as follows: setting the number of input layer nodes, the number of hidden layers, the number of neurons in the hidden layers, the output layer, the activation function, the loss function, the optimizer, the learning rate, the number of training iterations, and the L1 or L2 regularization parameter; the number of input layer nodes is related to Q... 10 The number of influencing factors is related;

[0044] L1 and L2 regularization are used to limit the complexity of the model, prevent the model from overfitting on the training data, and thus improve the model's generalization ability.

[0045] L1 regularization:

[0046]

[0047] L2 regularization:

[0048]

[0049] In the formula, and These are the loss functions for L1 and L2 regularization, respectively; λ is the data loss of the model; W is the regularization parameter; |||1 and |||2 are the L1 and L2 norms, respectively.

[0050] Furthermore, the specific process of determining the optimal hyperparameter combination using K-fold cross-validation combined with grid search in step 54 is as follows:

[0051] (1) Identify the hyperparameters that need to be adjusted and their possible values, and create a hyperparameter grid, which is the possible combinations of hyperparameter values;

[0052] (2) Perform grid search and K-fold cross-validation on all possible hyperparameter combinations;

[0053] The specific steps of the K-fold cross-validation are as follows:

[0054] 1) Split the entire dataset into K subsets;

[0055] 2) Select one subset for validation without repetition, and use the remaining K-1 subsets for model training;

[0056] 3) Summarize the results of K verifications and calculate the average value as the final verification result;

[0057] (3) Compare the verification results of all hyperparameter combinations to determine the optimal combination of hyperparameters.

[0058] Furthermore, the training of the deep neural network described in step 55 continuously adjusts and optimizes the weights and biases {W,b} of the deep neural network through the backpropagation algorithm in order to minimize the loss function.

[0059]

[0060] In the formula, It is the loss function, which is usually the root mean square error between the predicted and the measured data.

[0061] The beneficial effects of this invention are: This invention provides a spatial distribution prediction method for soil respiration temperature sensitivity, constructing a system for calculating soil respiration temperature sensitivity Q from different types of soil respiration data. 10 Methods, particularly for determining the soil respiration temperature sensitivity Q from soil respiration carbon emissions. 10 This greatly enriches the availability of existing data and significantly improves the ability to extract useful information from soil respiration data. Simultaneously, this invention enriches the soil respiration temperature sensitivity Q... 10 Based on data availability, a deep neural network model was used to construct a soil respiration temperature sensitivity Q-value. 10 With Q 10 The nonlinear relationship between influencing factors was investigated, and the point-scale Q was successfully implemented. 10 Extending to the regional scale, it enables spatial distribution prediction of soil respiration temperature sensitivity.

[0062] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0063] Figure 1 This is a flowchart of the method described in this invention;

[0064] Figure 2 This is a schematic diagram of deep neural network model training and prediction in Example 1;

[0065] Figure 3 The soil respiration temperature sensitivity Q of the Tibetan Plateau predicted in Example 1. 10 Spatial distribution map. Detailed Implementation

[0066] This invention discloses a method for predicting the spatial distribution of soil respiration temperature sensitivity, such as... Figure 1 As shown, the method includes the following steps:

[0067] Step 1: Data Acquisition and Processing

[0068] Soil respiration monitoring data and daily temperature datasets were acquired for the study area. The soil respiration monitoring data included two types of data: the first type was the daily average soil respiration flux data, and the second type was the soil respiration carbon emission data. These were generally obtained through literature, publicly available databases, or on-site monitoring. The unit for the daily average soil respiration flux data was μmol Cm. -2 s -1 Soil respiration carbon emissions data refers to the cumulative value of soil respiration carbon emissions during the monitoring period, expressed in gCm. -2 The daily temperature dataset comes from publicly available raster datasets, which can be remote sensing data or reanalysis data, such as the China Regional Surface Meteorological Element Driven Dataset (CMFD).

[0069] Then, based on the location and time information of the acquired soil respiration monitoring data, daily average temperature data is extracted from the daily temperature dataset. For the first type of data, the daily average temperature corresponding to the time of the daily average flux is extracted; for the second type of data, the daily average temperature of each day within the monitoring period is extracted. This is because the acquired soil respiration monitoring data lacks temperature information, therefore, it is necessary to extract the corresponding temperature information from the daily temperature dataset based on the location and time information of the soil respiration monitoring data.

[0070] Step 2: Calculate the parameters of the Van't Hoff exponential chemical reaction-temperature equation:

[0071] The relationship between daily average soil respiration flux and daily average air temperature is described using the Van't Hoff index chemical reaction-temperature equation:

[0072] Rs = ae bT (1)

[0073] In the formula, Rs is the average daily flux of soil respiration, in μmol Cm. -2 s -1 T represents the daily average temperature in °C; a is the first empirical parameter, and b is the second empirical parameter.

[0074] For the first type of data mentioned in step 1, the least squares method is used to fit the Van't Hoff exponential chemical reaction-temperature equation, i.e., formula (1), in order to determine the first empirical parameter a and the second empirical parameter b of the equation.

[0075] Least squares is a commonly used method for solving regression problems. It finds the best-fit function for the data by minimizing the sum of squared errors. Its loss function is as follows:

[0076]

[0077] In the formula, L is the loss function of the least squares method, i.e., the sum of squared errors; (x i y i ) represents the i-th sample; f() is the theoretical function with undetermined parameters.

[0078] For the second type of data mentioned in step 1, the first empirical parameter a and the second empirical parameter b are solved by the following method.

[0079] The daily average flux of soil respiration during the monitoring period is expressed as follows, based on the Van't Hoff index chemical reaction-temperature equation:

[0080]

[0081] In the formula, Rs i T represents the average daily soil respiration flux on day i; i Let represent the average daily temperature on day i.

[0082] The daily soil respiration emission is calculated based on the average daily soil respiration flux, using the following formula:

[0083]

[0084] In the formula, This represents the daily soil respiration emissions on day i, expressed in gCm³. -2 m represents the mass of carbon per micromole, which is 12 × 10⁻⁶. -6 gμmol -1 t represents the number of seconds in a day, which is 86400 s; η is a conversion factor with a value of 1.0368, and the unit is g / μmol. -1 s.

[0085] The cumulative value of soil respiration carbon emissions during the monitoring period is calculated based on the daily soil respiration emissions, using the following formula:

[0086]

[0087] In the formula, P represents the number of days in the monitoring period, in days (d). The cumulative value of soil respiration carbon emissions within day P, in gCm. -2 .

[0088] When two soil respiration carbon emission data exist at the same location but at different monitoring times, the values ​​of parameters a and b are obtained by solving the following set of nonlinear equations:

[0089]

[0090] In the formula, m and n represent different monitoring days, with the unit being days (d).

[0091] Step 3: Calculate the soil respiration temperature sensitivity Q 10 :

[0092] Based on the parameters of the Van't Hoff index chemical reaction-temperature equation obtained in step 2, the soil respiration temperature sensitivity Q was calculated. 10 The calculation formula is:

[0093] Q 10 =e 10b (7)

[0094] In the formula, Q 10 is the multiple of the change in soil respiration when the temperature increases by 10℃, used to represent the temperature sensitivity of soil respiration; b is the second empirical parameter.

[0095] Step 4: Obtain the soil respiration temperature sensitivity Q 10 Impact factor data and processing:

[0096] Obtain Q within the study area 10 Influencing factor data, including annual mean temperature, annual precipitation, Normalized Vegetation Index (NDVI), pH (potential for hydrogen), Soil Organic Carbon Density (SOCD), Digital Elevation Model (DEM), soil properties, etc., Q 10 The impact factor data also comes from publicly available raster datasets and is of various types.

[0097] For the obtained Q 10 The impact factor data is processed, that is, Q is first... 10 The impact factor data is resampled to the required resolution, and then outliers are removed and missing values ​​are imputed.

[0098] Step 5: Build and train a deep neural network model:

[0099] Specifically, the following steps are included:

[0100] Step 51, Q calculated in step 3 10As the label, with the calculated Q 10 Q at location 10 The influencing factors are used as features to form the training dataset.

[0101] Step 52: For Q in the training dataset 10 and Q 10 The impact factor data is normalized to generate a normalized training dataset.

[0102] Step 53: Build a deep neural network model and add L1 or L2 regularization to prevent the deep neural network model from overfitting.

[0103] The specific process of building a deep neural network is as follows: Set the number of nodes in the input layer of the deep neural network (and Q). 10 The factors influencing the number of factors include: the number of hidden layers, the number of neurons in the hidden layers, the output layer, the activation function, the loss function, the optimizer, the learning rate, the number of training iterations, and the L1 or L2 regularization parameters.

[0104] L1 and L2 regularization are used to limit the complexity of the model, prevent the model from overfitting to the training data, and thus improve the model's generalization ability.

[0105] L1 regularization:

[0106]

[0107] L2 regularization:

[0108]

[0109] In the formula, and These are the loss functions for L1 and L2 regularization, respectively; λ is the data loss of the model; W is the regularization parameter; |||1 and |||2 are the L1 and L2 norms, respectively.

[0110] Step 54: Determine the optimal hyperparameter combination using K-fold cross-validation combined with grid search. The specific process is as follows:

[0111] (1) Identify the hyperparameters that need to be adjusted and their possible values, and create a hyperparameter grid (i.e., possible combinations of hyperparameter values);

[0112] (2) Perform grid search and K-fold cross-validation on all possible hyperparameter combinations;

[0113] The specific steps of the K-fold cross-validation are as follows:

[0114] 1) Split the entire dataset into K subsets;

[0115] 2) Select one subset for validation without repetition, and use the remaining K-1 subsets for model training;

[0116] 3) Summarize the results of K verifications and calculate the average value as the final verification result;

[0117] (3) Compare the verification results of all hyperparameter combinations to determine the optimal combination of hyperparameters.

[0118] Step 55: Train a deep neural network model using the normalized training dataset and establish Q... 10 With Q 10 The mapping relationship between the influencing factors is used to obtain the predicted Q. 10 The regression model is the trained deep neural network model.

[0119] In this process, the training of the deep neural network involves continuously adjusting and optimizing the weights and biases {W,b} of the deep neural network through the backpropagation algorithm in order to minimize the loss function.

[0120]

[0121] In the formula, It is the loss function, which is usually the root mean square error between the predicted and the measured data.

[0122] Step 6: Determine the soil respiration temperature sensitivity Q in the study area. 10 Spatial distribution:

[0123] Q of all pixels at the required resolution for the study area 10 The influencing factor data is used as the prediction dataset, and the prediction dataset is normalized to generate a normalized prediction dataset. The generated normalized prediction dataset is then input into the trained deep neural network model to obtain the normalized Q-values ​​of the study area at the corresponding resolution. 10 Value; normalized Q 10 The value was inversely normalized to obtain the soil respiration temperature sensitivity Q over the entire study area. 10 The value, i.e., the soil respiration temperature sensitivity Q in the study area, is determined. 10 Spatial distribution.

[0124] Example 1

[0125] This embodiment is an application example of the above method.

[0126] This embodiment uses the Qinghai-Tibet Plateau as the application area and 2000-2020 as the baseline period. The above method is used to spatially map the sensitivity of soil respiration on the Qinghai-Tibet Plateau, with a spatial resolution of 1km × 1km. Specific steps include:

[0127] Step 1: Data Acquisition and Processing

[0128] Soil respiration monitoring data and daily temperature datasets were acquired for the study area. The soil respiration monitoring data included two types of data: the first type was the daily average soil respiration flux data, and the second type was the soil respiration carbon emission data. The unit for the daily average soil respiration flux data is μmol Cm. -2 s -1 Soil respiration carbon emissions data refers to the cumulative value of soil respiration carbon emissions during the monitoring period, expressed in gCm. -2 The daily temperature dataset is derived from publicly available raster datasets.

[0129] Then, based on the location and time information of the acquired soil respiration monitoring data, daily average temperature data is extracted from the daily temperature dataset. For the first type of data, the daily average temperature corresponding to the time of the daily average flux is extracted; for the second type of data, the daily average temperature of each day within the monitoring period is extracted.

[0130] Step 2: Calculate the parameters of the Van't Hoff exponential chemical reaction-temperature equation:

[0131] The relationship between daily average soil respiration flux and daily average air temperature is described using the Van't Hoff index chemical reaction-temperature equation:

[0132] Rs = ae bT (1)

[0133] In the formula, Rs is the average daily flux of soil respiration, in μmol Cm. -2 s -1 T represents the daily average temperature in °C; a is the first empirical parameter, and b is the second empirical parameter.

[0134] For the first type of data mentioned in step 1, the least squares method is used to fit the Van't Hoff exponential chemical reaction-temperature equation, i.e., formula (1), in order to determine the first empirical parameter a and the second empirical parameter b of the equation.

[0135] Least squares is a commonly used method for solving regression problems. It finds the best-fit function for the data by minimizing the sum of squared errors. Its loss function is as follows:

[0136]

[0137] In the formula, L is the loss function of the least squares method, i.e., the sum of squared errors; (x i y i ) represents the i-th sample; f() is the theoretical function with undetermined parameters.

[0138] For the second type of data mentioned in step 1, the first empirical parameter a and the second empirical parameter b are solved by the following method.

[0139] The daily average flux of soil respiration during the monitoring period is expressed as follows, based on the Van't Hoff index chemical reaction-temperature equation:

[0140]

[0141] In the formula, Rs i T represents the average daily soil respiration flux on day i; i Let represent the average daily temperature on day i.

[0142] The daily soil respiration emission is calculated based on the average daily soil respiration flux, using the following formula:

[0143]

[0144] In the formula, This represents the daily soil respiration emissions on day i, expressed in gCm³. -2 m represents the mass of carbon per micromole, which is 12 × 10⁻⁶. -6 gμmol -1 t represents the number of seconds in a day, which is 86400 s; η is a conversion factor with a value of 1.0368, and the unit is g / μmol. -1 s.

[0145] The cumulative value of soil respiration carbon emissions during the monitoring period is calculated based on the daily soil respiration emissions, using the following formula:

[0146]

[0147] In the formula, P represents the number of days of monitoring, in days; The cumulative value of soil respiration carbon emissions within day P, in gCm. -2 .

[0148] When two soil respiration carbon emission data exist at the same location but at different monitoring times, the values ​​of parameters a and b are obtained by solving the following set of nonlinear equations:

[0149]

[0150] In the formula, m and n represent different monitoring days, with the unit being days (d).

[0151] Step 3: Calculate the soil respiration temperature sensitivity Q 10 :

[0152] Based on the parameters of the Van't Hoff index chemical reaction-temperature equation obtained in step 2, the soil respiration temperature sensitivity Q was calculated.10 The calculation formula is:

[0153] Q 10 =e 10b (7)

[0154] In the formula, Q 10 is the multiple of the change in soil respiration when the temperature increases by 10℃, used to represent the temperature sensitivity of soil respiration; b is the second empirical parameter.

[0155] Step 4: Obtain the soil respiration temperature sensitivity Q 10 Impact factor data and processing:

[0156] Obtain Q within the study area 10 Impact factor data, Q obtained in this embodiment 10 The influencing factors include annual average temperature, annual precipitation, NDVI, pH, SOCD, DEM, soil bulk density, clay content, sand content, and silt content.

[0157] Get Q 10 The impact factor data were resampled to a spatial resolution of 1km×1km, and then outliers were removed and missing values ​​were imputed.

[0158] Step 5: Build and train a deep neural network model:

[0159] Specifically, the following steps are included:

[0160] Step 51, Q calculated in step 3 10 As the label, with the calculated Q 10 Q at location 10 The influencing factors are used as features to form the training dataset.

[0161] Step 52: For Q in the training dataset 10 and Q 10 The impact factor data is normalized to generate a normalized training dataset.

[0162] Step 53: Build a deep neural network model and add L1 or L2 regularization to prevent the deep neural network model from overfitting.

[0163] The specific process of building a deep neural network is as follows: Set the number of nodes in the input layer of the deep neural network (and Q). 10 The factors influencing the number of factors include: the number of hidden layers, the number of neurons in the hidden layers, the output layer, the activation function, the loss function, the optimizer, the learning rate, the number of training iterations, and the L1 or L2 regularization parameters.

[0164] L1 and L2 regularization are used to limit the complexity of the model, prevent the model from overfitting to the training data, and thus improve the model's generalization ability.

[0165] L1 regularization:

[0166]

[0167] L2 regularization:

[0168]

[0169] In the formula, and These are the loss functions for L1 and L2 regularization, respectively; λ is the data loss of the model; W is the regularization parameter; |||1 and |||2 are the L1 and L2 norms, respectively.

[0170] In this embodiment, the number of neurons in the input layer is set to 10(Q). 10 The number of influence factors is 10, the number of neurons in the output layer is 1, the activation function is ReLU, the optimizer is Adam, and the loss function is MSE. To obtain the deep neural network model with the optimal structure, the number of hidden layers, the number of neurons in the hidden layers, the learning rate, and the regularization parameter are determined using K-fold cross-validation in step 54.

[0171] Step 54: Determine the optimal hyperparameter combination using K-fold cross-validation combined with grid search. The specific process is as follows:

[0172] (1) Identify the hyperparameters that need to be adjusted and their possible values, and create a hyperparameter grid (i.e., possible combinations of hyperparameter values); In this embodiment, the optimal hidden layer structure (number of layers and number of neurons) and learning rate search grid are set as follows:

[0173] Hidden layer structure: [10, 10], [20, 20], [10, 10, 10], [20, 20, 20], [10, 20, 10]. Where [10, 10] indicates that there are 2 hidden layers and 10 neurons in each layer.

[0174] Learning rates: 0.01, 0.001, 0.0001, 0.00001, 0.000001.

[0175] (2) Perform grid search and K-fold cross-validation on all possible hyperparameter combinations;

[0176] The specific steps of the K-fold cross-validation are as follows:

[0177] 1) Split the entire dataset into K subsets; in this embodiment, the number of folds in the K-fold cross-validation is set to 10 (i.e., K = 10).

[0178] 2) Select one subset for validation without repetition, and use the remaining K-1 subsets for model training; In this embodiment, select one subset for validation without repetition, and use the remaining 9 subsets for model training, for a total of 10 validations.

[0179] 3) Summarize the results of K verifications and calculate the average value as the final verification result; in this embodiment, the results of 10 verifications are summarized and the average value is calculated as the final verification result.

[0180] (3) Compare the validation results of all hyperparameter combinations to determine the optimal combination of hyperparameters. In this embodiment, the optimal parameters are: {Hidden layer structure: [10,10,10]; Learning rate: 0.0001}. This embodiment does not limit the type of optimal parameters selected for cross-validation.

[0181] Step 55: Train a deep neural network model using the normalized training dataset to establish the temperature sensitivity Q. 10 With Q 10 The mapping relationship between the influencing factors is used to obtain the predicted Q. 10 The regression model is the trained deep neural network model, such as... Figure 2 As shown.

[0182] In this process, the training of the deep neural network involves continuously adjusting and optimizing the weights and biases {W,b} of the deep neural network through the backpropagation algorithm in order to minimize the loss function.

[0183]

[0184] In the formula, It is the loss function, which is usually the root mean square error between the predicted and the measured data.

[0185] Step 6: Determine the soil respiration temperature sensitivity Q in the study area. 10 Spatial distribution:

[0186] Q of all pixels at the required resolution for the study area 10 The influencing factor data is used as the prediction dataset. This embodiment uses Q data with a 1km × 1km resolution from the Qinghai-Tibet Plateau. 10 Impact factor data (2000-2020) predicts soil respiration temperature sensitivity Q in the Tibetan Plateau during the period 2000-2020. 10 The spatial distribution of the prediction dataset is Q within all 1km×1km pixels. 10 Impact factor data.

[0187] The predicted dataset is normalized to generate a normalized predicted dataset; the normalized predicted dataset of the study area is then input into the trained deep neural network model, such as... Figure 2 As shown, the normalized Q value of the corresponding resolution in the study area is obtained. 10 Value; normalized Q 10 The value was inversely normalized to obtain the soil respiration temperature sensitivity Q over the entire study area. 10 The value, i.e., the soil respiration temperature sensitivity Q in the study area, is determined. 10 Spatial distribution.

[0188] This embodiment predicts the soil respiration temperature sensitivity Q of the Qinghai-Tibet Plateau during the period 2000-2020. 10 Spatial distribution such as Figure 3 As shown. By Figure 3 It can be seen that the soil respiration temperature sensitivity Q on the Qinghai-Tibet Plateau is... 10 The values ​​range from 1 to 10, showing a distribution pattern of smaller values ​​in the northwest and larger values ​​in the southeast. The deep neural network effectively represents the soil respiration sensitivity Q of the Qinghai-Tibet Plateau. 10 Spatial distribution.

[0189] This invention fully utilizes limited soil respiration monitoring data (including flux data and carbon emission data) and extracts valuable information, significantly improving the usability of data in the study area. Simultaneously, this invention incorporates deep learning technology to transform point-scale Q... 10 Extending the value to a regional scale improves Q. 10 Accuracy of spatial distribution prediction. This invention is applicable to soil respiration temperature sensitivity Q in any region. 10 Spatial distribution prediction is more advantageous for regions lacking continuous soil respiration monitoring (such as the Qinghai-Tibet Plateau).

[0190] Finally, it should be noted that the above description is only used to illustrate the technical solution of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred arrangement, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for predicting the spatial distribution of soil respiration temperature sensitivity, characterized in that, The method includes the following steps: Step 1: Data Acquisition and Processing: Acquire soil respiration monitoring data and daily temperature dataset for the study area. The soil respiration monitoring data includes two types of data: the first type is the daily average flux data of soil respiration, and the second type is the carbon emission data of soil respiration. Based on the location and time information of the acquired soil respiration monitoring data, extract the daily average temperature data from the daily temperature dataset. Step 2: Calculate the parameters of the Van't Hoff exponential chemical reaction-temperature equation: The Van't Hoff exponential chemical reaction-temperature equation is used to describe the relationship between the daily average flux of soil respiration and the daily average air temperature. （1） In the formula, Rs is the average daily flux of soil respiration, in μmol·C·m³. -2 ·s -1 T represents the daily average temperature in °C; a is the first empirical parameter, and b is the second empirical parameter. For the first type of data mentioned in step 1, the least squares method is used to fit the Van't Hoff exponential chemical reaction-temperature equation, i.e., formula (1), in order to determine the first empirical parameter a and the second empirical parameter b of the equation. The least squares method finds the best-fit function for the data by minimizing the sum of squared errors, and its loss function is as follows: （2） In the formula, L is the loss function of the least squares method, i.e., the sum of squared errors; (x i y i Let f be the i-th sample; f() is the theoretical function with undetermined parameters; For the second type of data mentioned in step 1, the first empirical parameter a and the second empirical parameter b are solved by the following method; The daily average flux of soil respiration during the monitoring period is expressed as follows, based on the Van't Hoff index chemical reaction-temperature equation: （3） In the formula, Rs i T represents the average daily soil respiration flux on day i; i This represents the daily average temperature of day i; the daily soil respiration emission is calculated based on the daily average soil respiration flux, using the following formula: （4） In the formula, This represents the daily soil respiration emissions on day i, expressed in g·C·m³. -2 m represents the mass of carbon per micromole, which is 12 × 10⁻⁶. -6 g·μmol -1 t represents the number of seconds in a day, which is 86400 s; η is a conversion factor with a value of 1.0368, and the unit is g·s·μmol. -1 ; The cumulative value of soil respiration carbon emissions during the monitoring period is calculated based on the daily soil respiration emissions, using the following formula: （5） In the formula, P represents the number of days in the monitoring period, in days (d). P represents the cumulative carbon emissions from soil respiration within one day, expressed in g·C·m³. -2 ; When two soil respiration carbon emission data exist at the same location but at different monitoring times, the values ​​of parameters a and b are obtained by solving the following set of nonlinear equations: （6） In the formula, m and n represent different monitoring days, with the unit being days (d). Step 3: Calculate the soil respiration temperature sensitivity Q 10 Based on the parameters of the Van't Hoff index chemical reaction-temperature equation obtained in step 2, the soil respiration temperature sensitivity Q was calculated. 10 The calculation formula is: （7） In the formula, Q 10 is the multiple of the change in soil respiration when the temperature increases by 10℃, used to represent the temperature sensitivity of soil respiration; b is the second empirical parameter; Step 4: Obtain the soil respiration temperature sensitivity Q 10 Impact factor data and processing: Obtaining Q within the study area 10 Impact factor data, for obtaining Q 10 The impact factor data is processed, that is, Q is first... 10 The impact factor data is resampled to the required resolution, then outliers are removed and missing values ​​are imputed. Step 5: Build and train a deep neural network model: This includes the following steps: Step 51, Q calculated in step 3 10 As the label, with the calculated Q 10 Q at location 10 The influencing factors are used as features to form the training dataset; Step 52: For Q in the training dataset 10 and Q 10 The impact factor data is normalized to generate a normalized training dataset; Step 53: Build a deep neural network model and add L1 or L2 regularization to prevent the deep neural network model from overfitting. Step 54: Use K-fold cross-validation combined with grid search to determine the optimal hyperparameter combination; Step 55: Train a deep neural network model using the normalized training dataset and establish Q... 10 With Q 10 The mapping relationship between the influencing factors is used to obtain the predicted Q. 10 The regression model is the trained deep neural network model; Step 6: Determine the soil respiration temperature sensitivity Q in the study area. 10 Spatial distribution: Q of all pixels at the required resolution of the study area 10 The influencing factor data is used as the prediction dataset, and the prediction dataset is normalized to generate a normalized prediction dataset. The generated normalized prediction dataset is then input into the trained deep neural network model to obtain the normalized Q-values ​​of the study area at the corresponding resolution. 10 Value; normalized Q 10 The value was inversely normalized to obtain the soil respiration temperature sensitivity Q over the entire study area. 10 The value, i.e., the soil respiration temperature sensitivity Q in the study area, is determined. 10 Spatial distribution.

2. The method for predicting the spatial distribution of soil respiration temperature sensitivity according to claim 1, characterized in that, The soil respiration monitoring data mentioned in step 1 is obtained through literature, public databases, or site monitoring; the daily temperature dataset comes from a public raster dataset, which is remote sensing data or reanalysis data.

3. The method for predicting the spatial distribution of soil respiration temperature sensitivity according to claim 1, characterized in that, Q mentioned in step 4 10 The influencing factor data include annual average temperature, annual precipitation, normalized vegetation index (NDVI), pH, soil organic carbon density (SOCD), elevation (DEM), and soil property data.

4. The method for predicting the spatial distribution of soil respiration temperature sensitivity according to claim 3, characterized in that, The soil property data includes soil bulk density, clay content, sand content, and silt content.

5. The method for predicting the spatial distribution of soil respiration temperature sensitivity according to claim 1, characterized in that, The required resolution described in step 4 is 1 km × 1 km.

6. The method for predicting the spatial distribution of soil respiration temperature sensitivity according to claim 1, characterized in that, The specific process of building a deep neural network in step 53 is as follows: setting the number of input layer nodes, the number of hidden layers, the number of neurons in the hidden layers, the output layer, the activation function, the loss function, the optimizer, the learning rate, the number of training iterations, and the L1 or L2 regularization parameter; the number of input layer nodes is related to Q... 10 The number of influencing factors is related; L1 and L2 regularization are used to limit the complexity of the model, prevent the model from overfitting on the training data, and thus improve the model's generalization ability. L1 regularization: （8） L2 regularization: （9） In the formula, and These are the loss functions for L1 and L2 regularization, respectively; λ is the data loss of the model; W is the regularization parameter; W is the weight matrix. and These are the L1 and L2 norms, respectively.

7. The method for predicting the spatial distribution of soil respiration temperature sensitivity according to claim 1, characterized in that, The specific process of determining the optimal hyperparameter combination using K-fold cross-validation combined with grid search in step 54 is as follows: (1) Identify the hyperparameters that need to be adjusted and their possible values, and create a hyperparameter grid, i.e., the possible combinations of hyperparameter values; (2) Perform grid search and K-fold cross-validation on all possible hyperparameter combinations; The specific steps of the K-fold cross-validation are as follows: 1) Split the entire dataset into K subsets; 2) Select one subset for validation without repetition, and use the remaining K-1 subsets for model training; 3) Summarize the results of K verifications and calculate the average value as the final verification result; (3) Compare the verification results of all hyperparameter combinations to determine the optimal combination of hyperparameters.

8. The method for predicting the spatial distribution of soil respiration temperature sensitivity according to claim 1, characterized in that, The training of the deep neural network described in step 55 involves continuously adjusting and optimizing the weights and biases of the deep neural network through the backpropagation algorithm. In order to minimize the loss function: （10） In the formula, It is the loss function, which is the root mean square error between the predicted and measured data.

Images