Deep Learning-Based Satellite Remote Sensing Method and System for Monitoring Urban Green Space Carbon Sequestration

Abstract

This invention relates to the field of carbon sink monitoring technology, specifically to a satellite remote sensing method and system for monitoring urban green space carbon sinks based on deep learning. The method includes identifying key points characterizing the trend trajectory of carbon sink changes based on historical carbon sink information and satellite remote sensing images; classifying the carbon sink impact items corresponding to the key points into positive and negative impact items; and generating future carbon sink prediction results based on the positive and negative impact items. This invention outputs positive contribution trends and negative consumption trends respectively based on the differentiated positive and negative impact items; by combining these two to construct future carbon sink change trends, it completes the carbon sink prediction, thereby decomposing the carbon sink change process into two independent dimensions: promotion and consumption. This helps managers clearly identify key drivers and constraints, providing support for formulating precise control strategies.

Description

Technical Field

[0001] This invention belongs to the field of carbon sink monitoring technology, specifically relating to a satellite remote sensing method and system for monitoring urban green space carbon sinks based on deep learning. Background Technology

[0002] Urban green spaces play an irreplaceable role in absorbing greenhouse gases, regulating microclimates, and purifying air quality. As an important component of the urban ecosystem and a core carbon sink unit, the carbon sink capacity of urban green spaces needs to be dynamically and quantitatively assessed to measure the level of urban ecological health and to assist in the formulation of scientific urban planning, environmental governance solutions, and carbon reduction policies.

[0003] In practical applications, existing carbon sequestration monitoring and assessment methods often rely on static parameters such as fixed vegetation types and areas, or on estimations based on simplified linear models. This makes it difficult to effectively capture and respond to dynamic environmental changes. For short-term disturbances such as sudden temperature changes, extreme precipitation, air pollution events, or surrounding construction, it is impossible to accurately quantify the instantaneous impact of these complex factors on the carbon sequestration capacity of green spaces in real time. This leads to a certain deviation between the assessment results and the actual situation, resulting in reduced timeliness and reliability of the data.

[0004] To address the aforementioned issues, this invention provides a satellite remote sensing method and system for monitoring urban green space carbon sequestration based on deep learning. Summary of the Invention

[0005] The purpose of this invention is to provide a satellite remote sensing method and system for monitoring urban green space carbon sequestration based on deep learning, so as to solve the technical problem that the amount of urban green space carbon sequestration cannot be accurately assessed because it is affected by a variety of growth environment parameters.

[0006] To achieve the above-mentioned objectives, the technical solution adopted by the present invention includes:

[0007] Deep learning-based satellite remote sensing methods for monitoring carbon sequestration in urban green spaces include:

[0008] Based on historical carbon sink information and satellite remote sensing images, key points characterizing the trend trajectory of carbon sink changes are identified; based on these key points, the carbon sink impact items corresponding to the key points are divided into positive impact items and negative impact items; based on the positive and negative impact items, future carbon sink prediction results are generated.

[0009] Among them, based on historical carbon sink information and satellite remote sensing images, key points characterizing the trend trajectory of carbon sink change are identified, including: forming a trend trajectory based on historical carbon sink information and satellite remote sensing images; performing difference analysis on the trend trajectory to determine the peak position in the trend trajectory as the key point.

[0010] Preferably, forming a trend trajectory based on historical carbon sink information and satellite remote sensing images includes: inputting a historical data sequence composed of parameters related to carbon sinks into a prediction model to obtain the model output as a trend trajectory.

[0011] Preferably, the difference analysis of the trend trajectory includes:

[0012] Calculate the numerical difference between adjacent time points in the trend trajectory to form a difference sequence; when the numerical difference in the difference sequence is greater than a first set threshold or less than a second set threshold, the corresponding position is determined as the peak position.

[0013] Preferably, based on key points, the carbon sink impact items corresponding to the key points are divided into positive impact items and negative impact items, including:

[0014] If the carbon sequestration of urban green space at the key point's time location is higher than the baseline carbon sequestration, the carbon sequestration impact item corresponding to the key point will be determined as a positive impact item; if the carbon sequestration of urban green space at the key point's time location is not higher than the baseline carbon sequestration, the carbon sequestration impact item corresponding to the key point will be determined as a negative impact item.

[0015] Preferably, the carbon sink impact items are compiled from carbon sink-related parameters extracted from satellite remote sensing images.

[0016] Preferably, the parameters related to carbon sinks include temperature, humidity, wind speed, rainfall intensity, vegetation health index, and light intensity.

[0017] Preferably, the method further includes: comparing the predicted future carbon sink with historical carbon sink information to obtain a deviation value; and when the deviation value exceeds a deviation threshold, correcting the prediction model based on the deviation value.

[0018] This invention also discloses a deep learning-based satellite remote sensing urban green space carbon sequestration monitoring system, used to execute the deep learning-based satellite remote sensing urban green space carbon sequestration monitoring method, comprising:

[0019] The key point identification module is used to identify key points that characterize the trend trajectory of carbon sink changes based on historical carbon sink information and satellite remote sensing images;

[0020] The carbon sink prediction module is used to classify the carbon sink impact items corresponding to key points into positive impact items and negative impact items based on key points, and generate future carbon sink prediction results based on the positive impact items and negative impact items.

[0021] In addition, there is a model correction module, which is used to compare the future carbon sink prediction results with historical carbon sink information to obtain the deviation value, and correct the prediction model based on the deviation value when the deviation value exceeds the deviation threshold.

[0022] Preferably, the key point identification module is configured to: form a trend trajectory based on historical carbon sink information and satellite remote sensing images; perform difference analysis on the trend trajectory to determine the peak position in the trend trajectory as a key point.

[0023] Preferably, when the carbon sink prediction module distinguishes carbon sink impact items corresponding to key points into positive impact items and negative impact items, it is configured as follows:

[0024] If the carbon sequestration of urban green space at the key point's time location is higher than the baseline carbon sequestration, the carbon sequestration impact item corresponding to the key point will be determined as a positive impact item; if the carbon sequestration of urban green space at the key point's time location is not higher than the baseline carbon sequestration, the carbon sequestration impact item corresponding to the key point will be determined as a negative impact item.

[0025] Beneficial effects

[0026] 1. The urban green space carbon sink monitoring method provided by this invention forms a carbon sink change trend trajectory for each carbon sink influencing item, identifies fluctuations in the trajectory through difference analysis and determines the peak position as the key point, and finally distinguishes the corresponding items as positive and negative influencing items based on the key point. This enables dynamic identification of the influencing factors that play a decisive role in carbon sink changes at a specific time and their direction of action, overcoming the shortcomings of traditional static models that are difficult to capture the nonlinear relationship between various influencing factors and urban green space carbon sink, and improving the accuracy of future carbon sink prediction results.

[0027] 2. This invention outputs positive contribution trends and negative consumption trends respectively based on the distinguished positive and negative impact items; by combining the two, it constructs future carbon sink change trends and completes carbon sink quantity prediction, thereby decomposing the carbon sink change process into two independent dimensions of promotion and consumption, helping managers to clearly identify key drivers and constraints, and providing support for formulating precise control strategies. Attached Figure Description

[0028] Figure 1 This is a flowchart of the method of the present invention;

[0029] Figure 2 This is a system module diagram of the present invention. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are merely for explaining the invention and are not intended to limit the scope of protection of the invention.

[0031] Example 1

[0032] like Figure 1As shown, this embodiment provides a satellite remote sensing method for monitoring urban green space carbon sequestration based on deep learning, specifically including the following steps:

[0033] S1. Obtain the basic data for analysis;

[0034] The basic data specifically includes two parts: the first part is historical satellite remote sensing images covering the green areas of the target city, arranged in a time series; the second part is historical carbon sink information corresponding to the satellite remote sensing images in time.

[0035] The acquisition process includes:

[0036] Based on satellite remote sensing images, carbon sink areas, which are the main bodies of carbon absorption, and carbon source areas, which are the sources of carbon emissions, are accurately defined through land feature identification and classification rules. Among them, the preferred land feature identification and classification rules are image segmentation and classification processing based on spectral features, texture features, and spatial location relationships.

[0037] Furthermore, the carbon sink areas that are the main carbon absorbers are parks, woodlands, and lawns, while the carbon source areas that are the sources of carbon emissions are industrial areas, transportation hubs, and high-density building areas.

[0038] Carbon budget calculations were performed for the two types of regions separately, specifically:

[0039] For carbon source regions, their historical carbon dioxide emissions are estimated by linking their spatial location with urban energy consumption statistics and external databases including traffic flow records, and by applying a pre-defined emission factor calculation system.

[0040] For carbon sink areas, vegetation health indicators, including normalized vegetation index or enhanced vegetation index, extracted from remote sensing images are used, combined with standard carbon density coefficients for different vegetation types, and a conversion process is used to estimate their historical carbon dioxide absorption.

[0041] The difference between the historical carbon dioxide absorption and historical carbon dioxide emissions measured within the same time period is calculated to determine the urban green space carbon sink for that period, which is the net carbon dioxide absorption of the target urban green space area within a specific time period; this is then compiled into a time-seriesd historical carbon sink information.

[0042] The emission factor calculation system is specifically a computational framework used to estimate carbon dioxide emissions from carbon source regions. This framework links regional spatial locations with external databases and includes a series of preset coefficients that convert specific activity levels into carbon dioxide emissions.

[0043] S2. Based on satellite remote sensing images and associated meteorological and environmental data records, extract a series of parameters related to carbon sequestration;

[0044] These parameters are selected factors that significantly affect vegetation photosynthesis and respiration, specifically including:

[0045] Temperature, which characterizes thermal conditions; humidity, which affects plant physiological activities and soil moisture; wind speed, which affects gas exchange and water evaporation; rainfall intensity, which directly replenishes soil moisture; vegetation health index, which directly characterizes vegetation growth and photosynthetic potential; and light intensity, which is the core energy source for photosynthesis.

[0046] After acquiring the multi-source data, the data points are aligned, grouped, and merged according to their time labels. The time series data of each parameter is then encapsulated into an independent structured data unit, which is defined as a carbon sink impact item for subsequent time series analysis.

[0047] S3. Establish processing rules to reveal the nonlinear, long-term time-series dependence between various influencing factors and carbon sink changes;

[0048] This processing rule can analyze the complex, time-varying correspondence between historical data sequences of multiple carbon sink impact items and historical carbon sink information.

[0049] Through iterative calculations and analysis of a large amount of historical data, the internal parameters of this processing rule are continuously adjusted and determined until it can generate an output sequence that highly matches the actual historical trend of carbon sink changes based on a set of input historical data on carbon sink impact items. This output sequence is determined as the carbon sink change trend trajectory, which reflects a smoothed and denoised representation of the contribution of each influencing factor to carbon sink within the historical period.

[0050] Furthermore, the above processing rules can be specifically regarded as a computational model for representing the nonlinear long-term time-series dependency relationship between historical data and historical carbon sink information of multiple carbon sink impact items, and its specific definition is as follows:

[0051]

[0052]

[0053] In the formula, The value represents the trend trajectory, which means the denoised and smoothed carbon sink estimate for the corresponding time point after model processing.

[0054] This represents the hidden state vector, which means the state of the model after processing time step 1. After input, the encoded representation of all historical information up to the present;

[0055] This represents the input vector, which indicates the factors affecting carbon sequestration at a specific point in time. A quantitative snapshot of an environmental factor;

[0056] This represents the input dimension, which signifies the number of carbon sink-related parameters, i.e., the input vector. The dimension;

[0057] The weight matrix represents the internal parameters that the model determines by learning from historical data, used to capture the relationship between input and output.

[0058] This represents the bias vector, which is an internally adjustable parameter of the model used to adjust the baseline of the activation function;

[0059] This represents a nonlinear activation function, which means introducing a nonlinear transformation to enable the model to learn the complex relationship between the input and the output;

[0060] Subscript Indicates the current time step, subscript Indicates the previous time step.

[0061] S4. Perform difference analysis on the generated carbon sink change trend trajectory to automatically capture abrupt events that cause significant changes in carbon sink capacity;

[0062] The difference analysis process specifically includes:

[0063] Continuous time points are extracted from the trend trajectory of carbon sink changes, and the numerical difference between each pair of adjacent time points is calculated to form a difference sequence, which reflects the rate of trend change.

[0064] A threshold judgment is performed on each numerical difference in the difference sequence. Here, a positive first threshold and a negative second threshold are set to filter out normal, small-range random fluctuations. If a numerical difference is greater than the first threshold or less than the second threshold, it is determined that there is a significant fluctuation at that time point.

[0065] Furthermore, the location of the point with the fastest rate of change in this fluctuation, that is, the point with the largest absolute value of the difference, is determined as the key point, representing the core moment when carbon sequestration capacity undergoes a sudden change.

[0066] The first threshold is a preset positive number, used as a critical value to determine whether the trend trajectory shows a significant positive increase. The second threshold is a preset negative number, used as a critical value to determine whether the trend trajectory shows a significant negative decrease.

[0067] S5. After identifying the key points, their properties need to be categorized, that is, it needs to be determined whether the carbon sequestration impact items corresponding to the key points promote or inhibit carbon sequestration. This determination is performed based on the preset differentiation logic:

[0068] Obtain the actual urban green space carbon sink at the key point in time; calculate the baseline carbon sink for comparison, which can be obtained by calculating the moving average of the carbon sink over a stable time window prior to the key point, representing the expected level under normal conditions.

[0069] If the actual carbon sequestration at a critical point is higher than the baseline carbon sequestration, the carbon sequestration impact item at that critical point will be judged as a positive impact item.

[0070] If it is not higher than the benchmark carbon sink, it is judged as a negative impact item.

[0071] The stable time window is a preset time period used to extract historical data when calculating the baseline carbon sink. Data within this time period is considered to represent the average performance of the system under normal, non-mutational conditions.

[0072] S6. Based on the classification results of each influencing factor, generate the future carbon sink change trend for a specified time period, specifically including:

[0073] For each item classified as having a positive impact, based on its historical data, growth rate, periodicity, and other historical patterns, an extrapolation calculation process is used to generate a positive contribution trend representing the combined effect of future positive factors.

[0074] For all negative impact items, a negative consumption trend representing the combined effect of future negative factors is generated; the extrapolation calculation process preferably uses a time series model or regression analysis to extend the trend;

[0075] The positive contribution trend and the negative consumption trend are combined and calculated at corresponding time points within a specified future time period. The preferred method of this combined calculation is to use weighted superposition to construct a comprehensive future carbon sink change trend, i.e., a comprehensive future carbon sink dynamic change prediction curve.

[0076] Based on the comprehensive future carbon sink change trend, the predicted future carbon sink of urban green space is obtained and output through cumulative calculation or integral operation.

[0077] The future carbon sink forecast is the final predicted value of the urban green space carbon sink within a specified future time period.

[0078] S7. After obtaining the actual historical carbon sink information for a specified future time period, compare it with the generated future carbon sink prediction result, calculate the deviation value between the two, and compare the deviation value with the preset deviation threshold.

[0079] If the deviation value does not exceed the threshold, it indicates that the current processing rule is still effective; if the deviation value exceeds the threshold, it indicates that the actual situation has changed unexpectedly. At this time, the correction mechanism will be triggered, that is, based on the magnitude and direction of the deviation value, the internal parameters of the established processing rule will be adjusted, such as adjusting the weight of those positive or negative impact items that are overestimated or underestimated.

[0080] Then, steps S3 to S6 are re-executed using the revised processing rules to generate prediction results that are closer to the actual situation, thereby forming a closed-loop feedback and continuous optimization monitoring system.

[0081] Example 2

[0082] See Figure 2 As shown, this embodiment provides a satellite remote sensing urban green space carbon sequestration monitoring system based on deep learning, including the following modules:

[0083] The key point identification module is configured to identify key points that characterize the trajectory of carbon sink change trends based on historical carbon sink information and satellite remote sensing images.

[0084] In the specific execution process, this module obtains historical carbon sink information of a designated urban green area over a period of time, as well as multiple satellite remote sensing images covering the area during the same period; and forms a carbon sink change trend trajectory based on the historical carbon sink information and satellite remote sensing images.

[0085] The historical data sequence, consisting of parameters related to carbon sinks, extracted from historical data, is input into a pre-trained prediction model to obtain the model output as a trajectory of carbon sink change trends. This trajectory is a continuous curve or data sequence depicting the change of carbon sink volume over time.

[0086] A difference analysis is performed on the carbon sink change trend trajectory to identify the peak position in the carbon sink change trend trajectory as the key point. This process includes:

[0087] Calculate the numerical differences between adjacent time points in the carbon sink change trend trajectory to form a difference sequence, which reflects the rate of change in carbon sink volume;

[0088] Traverse the difference sequence. When a certain numerical difference in the sequence is greater than a preset first threshold or less than a preset second threshold, the position corresponding to the numerical difference is determined as the peak position, that is, the identified key point. These key points represent the time nodes when the carbon sequestration capacity changes significantly.

[0089] The carbon sink prediction module is configured to classify the corresponding carbon sink impact items into positive impact items and negative impact items based on key points, and generate future carbon sink prediction results based on this classification results.

[0090] In the specific execution process, this module receives key point information output by the key point recognition module, and processes satellite remote sensing images corresponding to the time and location of each key point to extract parameters related to carbon sequestration.

[0091] Parameters related to carbon sequestration include: temperature, humidity, wind speed, rainfall intensity, vegetation health index, and light intensity.

[0092] The carbon sink impact items corresponding to each key point are distinguished. Specifically, the urban green space carbon sink at the time location of the key point is compared with the preset baseline carbon sink, which is obtained by calculating the moving average of the carbon sink within a stable time window before the key point.

[0093] If the carbon sequestration of urban green space at the key point in time is higher than the baseline carbon sequestration, it indicates that the combination of factors associated with the key point has a positive effect on carbon sequestration. The module then determines the carbon sequestration impact item corresponding to the key point as a positive impact item. Conversely, if the carbon sequestration of urban green space at the key point in time is not higher than the baseline carbon sequestration, the corresponding carbon sequestration impact item is determined as a negative impact item.

[0094] Based on all the distinguished positive and negative impact items, the future carbon sink forecast is generated.

[0095] For example, based on all the distinguished positive and negative impact items, this module generates, through an extrapolation calculation process, a positive contribution trend representing the combined effect of future positive factors and a negative consumption trend representing the combined effect of future negative factors.

[0096] The positive contribution trend and the negative consumption trend are then combined to construct the future carbon sink change trend; based on this future carbon sink change trend, the future carbon sink volume prediction results for one or more time points are calculated and output.

[0097] The model correction module is configured to compare future carbon sink predictions with historical carbon sink information to obtain a deviation value, and correct the prediction model based on the deviation value when the deviation value exceeds a deviation threshold. This module is activated after a working cycle ends when new historical carbon sink information is available.

[0098] The carbon sink prediction module compares the previously generated future carbon sink prediction results with newly acquired historical carbon sink information for the same period point by point or as a whole, thereby calculating the deviation between the two.

[0099] This module compares the calculated deviation value with a preset deviation threshold. If the deviation value does not exceed the threshold, the current prediction model is considered to be stable and no adjustment is needed. If the deviation value exceeds the deviation threshold, it indicates that there is a decline in model performance or a systematic deviation. In this case, the module will trigger a correction mechanism.

[0100] Specifically, the prediction model used in the key point identification module is corrected based on this deviation value.

[0101] Correction methods may include retraining the model using a training set containing new data, adjusting the model's hyperparameters, or updating the model weights to better fit the latest data distribution, thereby improving the accuracy of subsequent predictions.

[0102] The above description is merely a preferred embodiment of this application and is not intended to limit this application. For those skilled in the art, this application can have various modifications and variations. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A satellite remote sensing method for monitoring urban green space carbon sequestration based on deep learning, characterized in that, include: Based on historical carbon sequestration information and satellite remote sensing images, key points characterizing the trajectory of carbon sequestration change trends are identified. Based on key points, the carbon sink impact items corresponding to the key points are divided into positive impact items and negative impact items; based on the positive impact items and negative impact items, the future carbon sink volume prediction results are generated; Among them, based on historical carbon sink information and satellite remote sensing images, key points characterizing the trend trajectory of carbon sink change are identified, including: forming a trend trajectory based on historical carbon sink information and satellite remote sensing images; performing difference analysis on the trend trajectory to determine the peak position in the trend trajectory as the key point; Difference analysis of trend trajectories includes: Calculate the numerical difference between adjacent time points in the trend trajectory to form a difference sequence; when the numerical difference in the difference sequence is greater than a first set threshold or less than a second set threshold, the corresponding position is determined as the peak position; Based on key points, the carbon sequestration impact items corresponding to the key points are divided into positive impact items and negative impact items, including: If the carbon sequestration of urban green space at the key point's time location is higher than the baseline carbon sequestration, the carbon sequestration impact item corresponding to the key point will be determined as a positive impact item; if the carbon sequestration of urban green space at the key point's time location is not higher than the baseline carbon sequestration, the carbon sequestration impact item corresponding to the key point will be determined as a negative impact item.

2. The satellite remote sensing method for monitoring urban green space carbon sequestration based on deep learning according to claim 1, characterized in that, The process of forming a trend trajectory based on historical carbon sink information and satellite remote sensing images involves inputting a historical data sequence composed of parameters related to carbon sinks into a prediction model to obtain the model output as a trend trajectory.

3. The satellite remote sensing method for monitoring urban green space carbon sequestration based on deep learning according to claim 1, characterized in that, The carbon sink impact items are compiled based on carbon sink-related parameters extracted from satellite remote sensing images.

4. The satellite remote sensing method for monitoring urban green space carbon sequestration based on deep learning according to claim 3, characterized in that, Parameters related to carbon sequestration include temperature, humidity, wind speed, rainfall intensity, vegetation health index, and light intensity.

5. The satellite remote sensing method for monitoring urban green space carbon sequestration based on deep learning according to claim 1, characterized in that, Also includes: The future carbon sink forecast results are compared with historical carbon sink information to obtain the deviation value; When the deviation value exceeds the deviation threshold, the prediction model is corrected based on the deviation value.

6. A satellite remote sensing urban green space carbon sequestration monitoring system based on deep learning, used to execute the satellite remote sensing urban green space carbon sequestration monitoring method based on deep learning as described in any one of claims 1-5, characterized in that, include: The key point identification module is used to identify key points that characterize the trend trajectory of carbon sink changes based on historical carbon sink information and satellite remote sensing images; The carbon sink prediction module is used to classify the carbon sink impact items corresponding to key points into positive impact items and negative impact items based on key points, and generate future carbon sink prediction results based on the positive impact items and negative impact items. In addition, there is a model correction module, which is used to compare the future carbon sink prediction results with historical carbon sink information to obtain the deviation value, and correct the prediction model based on the deviation value when the deviation value exceeds the deviation threshold.

7. The satellite remote sensing urban green space carbon sequestration monitoring system based on deep learning according to claim 6, characterized in that, The key point identification module is configured to: generate a trend trajectory based on historical carbon sink information and satellite remote sensing images; perform difference analysis on the trend trajectory to determine the peak position in the trend trajectory as the key point.

8. The satellite remote sensing urban green space carbon sequestration monitoring system based on deep learning according to claim 7, characterized in that, When the carbon sink prediction module distinguishes between positive and negative impact items corresponding to key points, it is configured as follows: If the carbon sequestration of urban green space at the key point in time is higher than the baseline carbon sequestration, the carbon sequestration impact item corresponding to the key point will be judged as a positive impact item. If the carbon sequestration of urban green space at the key point's time location is not higher than the baseline carbon sequestration, then the carbon sequestration impact item corresponding to the key point will be judged as a negative impact item.

Images