A method for assessing the impact of forest disturbance on carbon sinks

Abstract

The present invention belongs to the field of carbon sink assessment, and provides a method for assessing the impact of forest disturbance on carbon sink, including: collecting forest data to be assessed, extracting disturbance data, and calculating using a forest disturbance weight matrix; the construction process of the forest disturbance weight matrix includes: historical forest data collection, data slicing, long-short term memory network training, model prediction, data denoising, original forest disturbance impact matrix construction, and optimal coefficient selection. The present invention improves the reliability and versatility of the model by collecting a variety of forest disturbance data; improves the training efficiency of the model and the prediction accuracy of the model for long-term and short-term data by slicing and analyzing the data; and realizes the carbon sink impact assessment function by constructing an impact matrix and optimal coefficient selection.

Description

Technical Field

[0001] The present invention relates to the field of carbon sink assessment, and in particular to a method for assessing the impact of forest disturbance on carbon sink. Background Art

[0002] In the context of climate change, severe forest disturbances significantly affect carbon sink function. Due to the diversity of disturbance types, time and space, it is challenging to accurately assess the impact of disturbances on carbon storage. Traditional methods mostly rely on ground surveys and remote sensing data at a single time point, which makes it difficult to capture the dynamic changes and long-term impacts of disturbance events.

[0003] Current carbon sink assessment methods are usually based on limited ground sample data, lack long-term dynamic monitoring capabilities, and cannot comprehensively assess the impact of disturbances on carbon storage. In addition, existing methods often ignore the differences in the impact of different types of disturbances on carbon sinks and fail to provide precise guidance for regional forest management. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the purpose of the present invention is to provide a method for evaluating the impact of forest disturbance on carbon sinks, which improves the reliability and versatility of the model by collecting various forest disturbance data; improves the training efficiency of the model and the prediction accuracy of the model for long-term and short-term data by slicing the data; and realizes the carbon sink impact evaluation function by constructing an influence matrix and optimal selection of coefficients.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A method for assessing the impact of forest disturbance on carbon sinks, including:

[0007] Collect data on the forests to be assessed;

[0008] Extracting disturbance data of the forest data to be evaluated to obtain disturbance data to be evaluated;

[0009] Inputting the disturbance data to be evaluated into a preset forest disturbance weight matrix to obtain a target evaluation value;

[0010] The process of constructing the forest disturbance weight matrix includes:

[0011] Collect historical forest data; the historical forest data includes: long-term consistent NPP data, forest disturbance types, forest disturbance data and forest vegetation carbon storage data;

[0012] Slicing the historical forest data by month, season and year respectively to obtain a sliced ​​data set; the sliced ​​data set includes: a monthly forest data subset, a seasonal forest data subset and an annual forest data subset;

[0013] Using the slice data set to train the long short-term memory network to obtain a trained long short-term forest prediction model;

[0014] Using the long-term and short-term forest prediction model to perform a preset number of predictions to obtain a prediction data set;

[0015] Performing noise reduction processing on the predicted data set to obtain a preprocessed data set;

[0016] Constructing the original forest disturbance impact matrix;

[0017] The coefficients of the original forest disturbance type matrix are optimally selected using the k-nearest neighbor technique according to the preprocessed data set to obtain a forest disturbance weight matrix with determined coefficients.

[0018] Preferably, the step of acquiring the long time series consistent NPP data includes:

[0019] Collect CPF_MONTH_NPP dataset and MODIS_MONTH_NPP dataset;

[0020] Reconstructing the CPF_MONTH_NPP dataset and the MODIS_MONTH_NPP dataset respectively by using a mean iteration filtering method;

[0021] A regression model is established using the reconstructed CPF_MONTH_NPP dataset and the MODIS_MONTH_NPP dataset; the regression model is: in, is the predicted value of the regression model; C is the data value in the regression model; are model parameters simulated by the least squares method; M i is the maximum value of the i-th month in the MODIS_MONTH_NPP dataset, C i The maximum value of the i-th month in the CPF_MONTH_NPP dataset; is the average value of the monthly maximum value of the MODIS_MONTH_NPP dataset; is the average value of the monthly maximum value of the CPF_MONTH_NPP data set; n is the month number; ε is the residual; M real is the true value of the MODIS_MONTH_NPP dataset, M predicte is the predicted value of the MODIS_MONTH_NPP dataset; m is the month number;

[0022] According to the CPF_MONTH_NPP dataset, the MODIS_MONTH_NPP dataset is extended using the regression model to obtain the long time series consistent NPP data.

[0023] Preferably, the slice data set is used to train the long short-term memory network to obtain a trained long short-term forest prediction model, including:

[0024] Inputting the annual forest data subset into the long short-term memory network for forward propagation to obtain a forward output;

[0025] Calculating a preset loss function based on the forward output to obtain first error data;

[0026] Back-propagating the parameters of the long short-term memory network based on the first error data to obtain a propagated annual forecast model;

[0027] Inputting the seasonal forest data subset and the monthly forest data subset into the annual prediction model for prediction and loss value calculation to obtain second error data;

[0028] The annual prediction model is optimized using the Adam optimizer according to the second error data to obtain the optimized long-term and short-term forest prediction model.

[0029] Preferably, performing noise reduction processing on the predicted data set to obtain a preprocessed data set includes:

[0030] Calculate the monthly average value of the long time series consistent NPP data to obtain the benchmark mean value;

[0031] Set the reference variance range;

[0032] Calculate the average value of each month in the prediction data set to obtain a prediction mean set;

[0033] Calculate the variance of each mean in the predicted mean set and the benchmark mean to obtain a deviation value;

[0034] When the deviation value does not belong to the reference variance range, the prediction data corresponding to the deviation value is eliminated.

[0035] Preferably, the forest disturbance types include: logging, forest fire, farming, construction, pests and diseases, and storms.

[0036] Preferably, the forest vegetation carbon storage data is divided into: low-range carbon storage, medium-low range carbon storage, medium range carbon storage, medium-high range carbon storage and high range carbon storage.

[0037] Preferably, the original forest disturbance impact matrix is:

[0038]

[0039] Wherein, Ann represents the coefficient corresponding to the nth forest disturbance type with index n.

[0040] Preferably, the disturbance data to be evaluated is input into a preset forest disturbance weight matrix to obtain a target evaluation value, including:

[0041] Inputting the disturbance data to be evaluated into the equation group corresponding to the forest disturbance weight matrix to obtain an output vector;

[0042] A sum operation is performed on the output vector to obtain the target evaluation value.

[0043] Preferably, it also includes:

[0044] The historical forest data is updated once a month, and the forest disturbance weight matrix is ​​adjusted using the updated historical forest data.

[0045] The present invention discloses the following technical effects:

[0046] The present invention provides an assessment method for the impact of forest disturbance on carbon sinks. By collecting various forest disturbance data, the defect that conventional technology is difficult to capture long-term impacts is solved, and the long-term impact of forest disturbance data is extracted; by slicing data, the problem of large gap between long-term and short-term prediction effects of existing models is solved, and the training efficiency of the model and the prediction accuracy of long-term and short-term data are improved; by constructing an influence matrix and optimal selection of coefficients, the problem of complex conventional carbon sink assessment process is solved, and the carbon sink impact assessment process is simplified. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0048] Figure 1 A schematic diagram of a process for evaluating the impact of forest disturbance on carbon sinks provided by an embodiment of the present invention;

[0049] Figure 2 A schematic diagram of a construction process of a forest disturbance weight matrix provided in an embodiment of the present invention;

[0050] Figure 3A fitting diagram of the relationship between CPF_MONTH_NPP and MODIS_MONTH_NPP data provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] The purpose of the present invention is to provide a method for evaluating the impact of forest disturbance on carbon sinks, by collecting various forest disturbance data, improving the reliability and versatility of the model; by slicing the data, improving the training efficiency of the model and the prediction accuracy of the model for long-term and short-term data; by constructing an influence matrix and optimal selection of coefficients, realizing the carbon sink impact evaluation function.

[0053] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0054] Figure 1 A schematic diagram of the evaluation process of the impact of forest disturbance on carbon sinks provided by an embodiment of the present invention, Figure 2 A schematic diagram of the construction process of the forest disturbance weight matrix provided in an embodiment of the present invention is shown in FIG. Figure 1 and Figure 2 As shown, the present invention provides a method for evaluating the impact of forest disturbance on carbon sinks, comprising:

[0055] Step 100: Collect data on the forest to be assessed;

[0056] Step 300: extracting disturbance data of the forest data to be evaluated to obtain disturbance data to be evaluated;

[0057] Step 300: inputting the disturbance data to be evaluated into a preset forest disturbance weight matrix to obtain a target evaluation value;

[0058] The process of constructing the forest disturbance weight matrix includes:

[0059] Step 301: Collect historical forest data; the historical forest data includes: long-term consistent NPP data, forest disturbance types, forest disturbance data, and forest vegetation carbon storage data;

[0060] Step 302: Slice the historical forest data by month, season and year to obtain a slice data set; the slice data set includes: a monthly forest data subset, a seasonal forest data subset and an annual forest data subset;

[0061] Step 303: using the slice data set to train the long short-term memory network to obtain a trained long short-term forest prediction model;

[0062] Step 304: using the long-term and short-term forest prediction model to perform a preset number of predictions to obtain a prediction data set;

[0063] Step 305: performing noise reduction processing on the predicted data set to obtain a preprocessed data set;

[0064] Step 306: constructing the original forest disturbance impact matrix;

[0065] Step 307: Optimizing coefficient selection for the original forest disturbance type matrix using the k-nearest neighbor technique according to the preprocessed data set to obtain a forest disturbance weight matrix with determined coefficients.

[0066] Furthermore, the step of acquiring the long time series consistent NPP data includes:

[0067] Collect CPF_MONTH_NPP dataset and MODIS_MONTH_NPP dataset;

[0068] Reconstructing the CPF_MONTH_NPP dataset and the MODIS_MONTH_NPP dataset respectively by using a mean iteration filtering method;

[0069] A regression model is established using the reconstructed CPF_MONTH_NPP dataset and the MODIS_MONTH_NPP dataset; the regression model is: in, is the predicted value of the regression model; C is the data value in the regression model; are model parameters simulated by the least squares method; M i is the maximum value of the i-th month in the MODIS_MONTH_NPP dataset, C i The maximum value of the i-th month in the CPF_MONTH_NPP dataset; is the average value of the monthly maximum value of the MODIS_MONTH_NPP dataset; is the average value of the monthly maximum value of the CPF_MONTH_NPP data set; n is the month number; ε is the residual; M real is the true value of the MODIS_MONTH_NPP dataset, Mpredicte is the predicted value of the MODIS_MONTH_NPP dataset; m is the month number;

[0070] According to the CPF_MONTH_NPP dataset, the MODIS_MONTH_NPP dataset is extended using the regression model to obtain the long time series consistent NPP data.

[0071] Specifically, the slice data set is used to train the long short-term memory network to obtain a trained long short-term forest prediction model, including:

[0072] Inputting the annual forest data subset into the long short-term memory network for forward propagation to obtain a forward output;

[0073] Calculating a preset loss function based on the forward output to obtain first error data;

[0074] Back-propagating the parameters of the long short-term memory network based on the first error data to obtain a propagated annual forecast model;

[0075] Inputting the seasonal forest data subset and the monthly forest data subset into the annual prediction model for prediction and loss value calculation to obtain second error data;

[0076] The annual prediction model is optimized using the Adam optimizer according to the second error data to obtain the optimized long-term and short-term forest prediction model.

[0077] Specifically, the Long Short-Term Memory (LSTM) network is a special type of Recurrent Neural Network (RNN) used to process sequence data. This network design is used to solve the gradient vanishing and gradient exploding problems encountered by traditional RNN in long sequence learning, making LSTM perform well in many application areas, such as natural language processing, time series prediction, and speech recognition. The advantages of LSTM are: solving the long dependency problem: LSTM can effectively capture the dependency relationship in long time series and is suitable for tasks that require long-term memory; resistance to gradient vanishing: by designing the gate mechanism, LSTM can keep information and flow in long sequences, thereby preventing gradient vanishing; can handle variable length sequences: can flexibly handle input sequences of different lengths. The LSTM network contains special structures and units, mainly composed of the following parts: Forget Gate: determines which information is discarded from the cell state. Input Gate: determines how much new information is stored in the cell state. Output Gate: determines which part of the cell state will be used for output. In each time step, the LSTM unit performs the following steps: the forget gate reads the hidden state of the previous time step and the input of the current time step, and then decides which information will be forgotten; the input gate reads the hidden state of the previous time step and the input of the current time step at the same time to decide which new information will be stored; the cell state is updated according to the output of the forget gate and the input gate; the output gate reads the updated cell state and the input of the current time step to determine the final hidden state. Long short-term memory network (LSTM) is mainly used to solve the following problems: Long sequence dependency problem: In traditional recurrent neural networks (RNN), as the length of the sequence increases, the gradient may become very small (gradient vanishing) or very large (gradient explosion) during the back propagation process, making it difficult for the model to learn long-distance dependencies. LSTM effectively solves this problem through a gating mechanism, allowing the model to learn longer-term dependencies in the sequence. Sequence prediction problem: LSTM performs well in sequence prediction tasks, such as stock price prediction, weather prediction, etc. These tasks require the model to capture long-term trends in sequence data. Natural Language Processing (NLP) tasks: LSTM is widely used in the field of NLP, including language model construction, machine translation, text summarization, sentiment analysis, question-answering systems, etc. LSTM can handle long-distance dependencies in text sequences and understand contextual information. Speech recognition: In automatic speech recognition systems, LSTM can process audio sequences and identify patterns and structures in speech, thereby improving recognition accuracy.Time Series Analysis: LSTM can be applied to the analysis of time series data such as financial market analysis and medical health monitoring to capture long-term trends and cyclical changes in time series. Generative Models: LSTM can also be used as part of a generative model to generate text, music, artwork, etc. Reinforcement Learning: In the field of reinforcement learning, LSTM can be used as part of an intelligent agent to help the agent learn how to make decisions based on the state of the environment. Image Processing: Although LSTM is mainly used for sequence data, it can also be used in combination with other types of neural networks, such as convolutional neural networks (CNNs), to process image data. These applications of LSTM demonstrate its powerful ability to process data with time series characteristics, especially in scenarios where long-term dependent information needs to be captured and utilized.

[0078] Furthermore, the prediction data set is subjected to noise reduction processing to obtain a preprocessed data set, including:

[0079] Calculate the monthly average value of the long time series consistent NPP data to obtain the benchmark mean value;

[0080] Set the reference variance range;

[0081] Calculate the average value of each month in the prediction data set to obtain a prediction mean set;

[0082] Calculate the variance of each mean in the predicted mean set and the benchmark mean to obtain a deviation value;

[0083] When the deviation value does not belong to the reference variance range, the prediction data corresponding to the deviation value is eliminated.

[0084] Preferably, the forest disturbance types include: logging, forest fire, farming, construction, pests and diseases, and storms.

[0085] Optionally, the forest vegetation carbon storage data is divided into: low-range carbon storage, medium-low range carbon storage, medium range carbon storage, medium-high range carbon storage and high range carbon storage.

[0086] Specifically, the original forest disturbance impact matrix is:

[0087]

[0088] Wherein, Ann represents the coefficient corresponding to the nth forest disturbance type with index n.

[0089] Furthermore, the disturbance data to be evaluated is input into a preset forest disturbance weight matrix to obtain a target evaluation value, including:

[0090] Inputting the disturbance data to be evaluated into the equation group corresponding to the forest disturbance weight matrix to obtain an output vector;

[0091] A sum operation is performed on the output vector to obtain the target evaluation value.

[0092] Preferably, it also includes:

[0093] The historical forest data is updated once a month, and the forest disturbance weight matrix is ​​adjusted using the updated historical forest data.

[0094] Alternatively, forests play an important role in the global response to climate change. The net primary productivity of forest vegetation reflects the ability of forests to fix carbon through photosynthesis, while carbon storage represents the total amount of carbon accumulated and stored in forest ecosystems. Together, they reveal the contribution of forests to the global carbon cycle. Changes in forest carbon sinks caused by a continuous stream of disturbance events directly affect the global carbon cycle. After a forest is disturbed (such as logging, fire, storms, or pests and diseases), it may damage plant structure, reduce the efficiency of photosynthesis, and lead to a decrease in NPP; it may also cause tree death and carbon emissions, thereby affecting carbon storage; it may also have medium- and long-term effects such as secondary forest formation, changes in vegetation structure, and ecosystem recovery and stability, thereby affecting carbon fixation capacity. Quantitative assessment of changes in productivity and carbon storage before and after forest disturbance helps to understand the impact of forest disturbance on carbon sinks.

[0095] Specifically, the method for constructing the long-term consistent forest vegetation NPP dataset is as follows:

[0096] This embodiment specifically uses CPF_MONTH_NPP data and MODIS_MONTH_NPP data. MOD17A2H and MODIS17A3HGF data (2001-2020, 500m) are widely used around the world. Compared with CPF_NPP (1985-2015, 1km), the two have obvious differences in time coverage, time resolution, spatial resolution, production methods, etc., and are limited by the time scale when conducting long-term time series research. Therefore, when combining them to conduct long-term NPP change trend research, it is necessary to perform necessary preprocessing on the existing data. Based on the CPF_MONTH_NPP and MODIS_MONTH_NPP data sets, a long-term consistent NPP data set is constructed. The steps are as follows:

[0097] S1: Dataset filtering and reconstruction: The two datasets mentioned above are reconstructed using the mean iteration filtering method with the help of Python tools. That is, the mean of the adjacent time series before and after a certain year is calculated first, and then the absolute value of the difference between the annual value and the mean is determined to be greater than the threshold. If so, the mean is used to replace the original value, otherwise the original value is used.

[0098] S2: Modeling of two datasets in overlapping years: If you want to extend the CPF_MONTH_NPP data to the MODIS_MONTH_NPP data, you need to check whether the trends of the two data in the overlapping time period are consistent. Only when the consistency is passed can CPF_MONTH_NPP be used to extend MODIS_MONTH_NPP. Based on step S1, 1500 sample points were randomly selected, and the corresponding pixel values ​​(multi-year monthly averages) of the two types of raster data in the overlapping time period were extracted and fitted. It was found that the linear relationship between the two was relatively significant (reference Figure 3 ). You can use MODIS_MONTH_NPP data and CPF_MONTH_NPP data to build a regression model.

[0099] S3: Use CPF_MONTH_NPP data to model MODIS_MONTH_NPP data. Since the overlap period of the two data sources is from 2001 to 2015, the required modeling data and verification data are selected from the overlapping year data along the time series in an 8:2 ratio. Use CPF_MONTH_NPP data to model MODIS_MONTH_NPP data pixel by pixel. The basic form of the regression model is:

[0100]

[0101] S4: Dataset extension: MODIS_MONTH_NPP data are extended based on the modeling results. According to the regression model obtained, the MODIS_MONTH_NPP data and CPF_MONTH_NPP data are input into the model during the non-overlapping period (1985-2000), and the MODIS_MONTH_NPP data extension results are obtained. Using band operations (accumulated within the year), the long-term consistent annual YEAR_NPP dataset from 1986 to 2020 is finally constructed, with a temporal resolution of yearly, a spatial resolution of 1km, a unit of gC·m-2·a-1, and a .TIF format.

[0102] Specifically, when using the k-nearest neighbor technique to determine the coefficients of the original forest disturbance impact matrix, it is first necessary to construct a corresponding impact prediction function; the impact prediction function is a combination of the coefficient value of the original forest disturbance impact matrix and the product of the corresponding variable in advance, and its form is as follows:

[0103]

[0104] Wherein, Z is the impact prediction function; The variable corresponding to the nth forest disturbance type with index n.

[0105] The forest disturbance type and the forest disturbance data are taken as input, the impact assessment value is taken as the target value, and the coefficient is optimized using the k-nearest neighbor technique.

[0106] Specifically, the K-NearestNeighbors (KNN) algorithm is a simple and commonly used non-parametric classification and regression algorithm. It makes predictions based on the labels of the input data points and the nearest K neighbors in the training data set. The KNN algorithm is used in many fields, especially in classification and regression problems. KNN is classified by measuring the distance between different feature values. Its idea is: if most of the k most similar samples (i.e., the nearest ones in the feature space) of a sample belong to a certain category, then the sample also belongs to this category, where K is usually an integer not greater than 20. In the KNN algorithm, the selected neighbors are all objects that have been correctly classified. This method only determines the category of the sample to be classified based on the category of the nearest one or several samples in the classification decision. The k-nearestneighbor (kNN) method is a basic classification and regression method; a model based on labeled training data; and a supervised learning algorithm. The three key points of the basic approach are: first, determine the distance metric; second, select the k value (find the k instance points in the training set that are closest to the estimated point); and third, the classification decision rule. In classification tasks, the "voting method" can be used, that is, the most frequently appearing label category in these k instances is selected as the prediction result; in regression tasks, the "average method" can be used, that is, the average value of the real-valued output labels of these k instances is used as the prediction result; and weighted averaging or weighted voting can also be performed based on the distance, and the closer the distance, the greater the weight of the instance.

[0107] The beneficial effects of the present invention are as follows:

[0108] The present invention improves the reliability and versatility of the model by collecting diverse forest disturbance data; improves the training efficiency of the model and the prediction accuracy of the model for long-term and short-term data by performing slice analysis on the data; and realizes the carbon sink impact assessment function by constructing an impact matrix and optimizing coefficient selection.

[0109] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0110] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for assessing the impact of forest disturbance on carbon sinks, characterized in that: include: Collect data on the forests to be assessed; Extracting disturbance data of the forest data to be evaluated to obtain disturbance data to be evaluated; Inputting the disturbance data to be evaluated into a preset forest disturbance weight matrix to obtain a target evaluation value; The process of constructing the forest disturbance weight matrix includes: Collect historical forest data; the historical forest data includes: long-term consistent NPP data, forest disturbance types, forest disturbance data and forest vegetation carbon storage data; Slicing the historical forest data by month, season and year respectively to obtain a sliced ​​data set; the sliced ​​data set includes: a monthly forest data subset, a seasonal forest data subset and an annual forest data subset; Using the slice data set to train the long short-term memory network to obtain a trained long short-term forest prediction model; Using the long-term and short-term forest prediction model to perform a preset number of predictions to obtain a prediction data set; Performing noise reduction processing on the predicted data set to obtain a preprocessed data set; Constructing the original forest disturbance impact matrix; The coefficients of the original forest disturbance type matrix are optimally selected using the k-nearest neighbor technique according to the preprocessed data set to obtain a forest disturbance weight matrix with determined coefficients.

2. A method for assessing the impact of forest disturbance on carbon sinks according to claim 1, characterized in that: The step of acquiring the long time series consistency NPP data includes: Collect CPF_MONTH_NPP dataset and MODIS_MONTH_NPP dataset; Reconstructing the CPF_MONTH_NPP dataset and the MODIS_MONTH_NPP dataset respectively by using a mean iteration filtering method; A regression model is established using the reconstructed CPF_MONTH_NPP dataset and the MODIS_MONTH_NPP dataset; the regression model is: in, is the predicted value of the regression model; C is the data value in the regression model; are model parameters simulated by the least squares method; M i is the maximum value of the i-th month in the MODIS_MONTH_NPP dataset, C i The maximum value of the i-th month in the CPF_MONTH_NPP dataset; is the average value of the monthly maximum value of the MODIS_MONTH_NPP dataset; is the average value of the monthly maximum value of the CPF_MONTH_NPP data set; n is the month number; ε is the residual; M real is the true value of the MODIS_MONTH_NPP dataset, M predicte is the predicted value of the MODIS_MONTH_NPP dataset; m is the month number; According to the CPF_MONTH_NPP dataset, the MODIS_MONTH_NPP dataset is extended using the regression model to obtain the long time series consistent NPP data.

3. The method for evaluating the impact of forest disturbance on carbon sink according to claim 1, characterized in that: The slice data set is used to train the long short-term memory network to obtain a trained long short-term forest prediction model, including: Inputting the annual forest data subset into the long short-term memory network for forward propagation to obtain a forward output; Calculating a preset loss function based on the forward output to obtain first error data; Back-propagating the parameters of the long short-term memory network based on the first error data to obtain a propagated annual forecast model; Inputting the seasonal forest data subset and the monthly forest data subset into the annual prediction model for prediction and loss value calculation to obtain second error data; The annual prediction model is optimized using the Adam optimizer according to the second error data to obtain the optimized long-term and short-term forest prediction model.

4. The method for evaluating the impact of forest disturbance on carbon sink according to claim 1, characterized in that: The prediction data set is subjected to noise reduction processing to obtain a preprocessed data set, including: Calculate the monthly average value of the long time series consistent NPP data to obtain the benchmark mean value; Set the reference variance range; Calculate the average value of each month in the prediction data set to obtain a prediction mean set; Calculate the variance of each mean in the predicted mean set and the benchmark mean to obtain a deviation value; When the deviation value does not belong to the reference variance range, the prediction data corresponding to the deviation value is eliminated.

5. The method for evaluating the impact of forest disturbance on carbon sink according to claim 1, characterized in that: The types of forest disturbance include: logging, forest fires, farming, construction, pests and diseases, and storms.

6. The method for evaluating the impact of forest disturbance on carbon sink according to claim 1, characterized in that: The forest vegetation carbon storage data are divided into: low-range carbon storage, medium-low range carbon storage, medium range carbon storage, medium-high range carbon storage and high range carbon storage.

7. The method for evaluating the impact of forest disturbance on carbon sink according to claim 1, characterized in that: The original forest disturbance impact matrix is: Wherein, Ann represents the coefficient corresponding to the nth forest disturbance type with index n.

8. The method for evaluating the impact of forest disturbance on carbon sink according to claim 1, characterized in that: The disturbance data to be evaluated is input into a preset forest disturbance weight matrix to obtain a target evaluation value, including: Inputting the disturbance data to be evaluated into the equation group corresponding to the forest disturbance weight matrix to obtain an output vector; A sum operation is performed on the output vector to obtain the target evaluation value.

9. The method for evaluating the impact of forest disturbance on carbon sink according to claim 1, characterized in that: Also includes: The historical forest data is updated once a month, and the forest disturbance weight matrix is ​​adjusted using the updated historical forest data.

Images