Coastal zone ecosystem carbon reserve estimation method, system and equipment based on long-time-sequence remote sensing data and medium

Abstract

The invention provides a coastal zone ecosystem carbon reserve estimation method based on long time sequence remote sensing data, and the method comprises the steps: obtaining satellite remote sensing data, and constructing a high-quality time sequence remote sensing image stack according to the satellite remote sensing data; classifying the remote sensing images based on a random forest method to obtain coastal zone surface land utilization type data; on the basis of the coastal zone surface land utilization type data, coastal zone ecosystem carbon reserve calculation is carried out through an InVEST model, and a coastal zone ecosystem carbon reserve accounting result is obtained; and counting and analyzing the change trend of the coastal zone ecosystem carbon reserve in the aspect of time based on the coastal zone ecosystem carbon reserve accounting result. Technical support can be provided for a coastal zone ecosystem carbon income and expenditure quantitative calculation method.

Description

Technical Field

[0001] The present invention belongs to the technical field of remote sensing images, and in particular relates to a method, system, equipment and medium for estimating carbon reserves of coastal ecosystems based on long-term remote sensing data. Background Art

[0002] As a special zone where land and sea meet, interact and have certain influences, the coastal zone is an important carbon sink area in the world with strong carbon storage capacity. Typical blue carbon ecosystems such as mangroves, salt marshes and seagrass beds can not only absorb a large amount of atmospheric carbon dioxide, but also store organic carbon in sediments for a long time. At present, domestic and foreign scholars have conducted extensive research on the spatiotemporal changes and driving factors of surface carbon storage in coastal zones using field survey methods and ecological model methods, with the help of ranging instruments, global positioning systems (GPS), aerial or satellite images and other tools. The results of the field survey method are relatively reliable and perform well in small-scale applications, but due to the huge manpower and economic costs, it cannot meet the monitoring of long time scales and large spatial ranges. Ecological models are mainly based on the complex relationship between the change process of land use types and environmental factors to estimate the productivity and carbon storage of terrestrial ecosystems. Ecological models often require a large amount of measured data support, and are less applicable to coastal areas that lack field surveys and ground stations.

[0003] There are currently a variety of coastal surface carbon storage assessment methods for different research purposes, and they have been successful in specific fields or under specific conditions. However, due to the limitations of traditional classification methods and the differences in physical mechanisms, data sources and parameter settings of different models, there is still a large degree of uncertainty in the existing surface carbon storage estimation results, resulting in a serious shortage of existing carbon storage data. Therefore, this technology fully considers the advantages of remote sensing technology in long-term series, large spatial range, and high-dynamic surface monitoring. With the support of the GEE cloud platform and the InVEST model, a new type of ecosystem carbon storage remote sensing estimation method based on long-term remote sensing data has been developed for coastal areas. The present invention can provide technical support for the quantitative calculation method of carbon budget of coastal ecosystems. Summary of the invention

[0004] The purpose of the present invention is to address the shortcomings of existing methods and propose a new method for estimating carbon storage in coastal ecosystems based on long-term remote sensing data.

[0005] The technical solution of the present invention is as follows:

[0006] A method for estimating carbon storage in coastal ecosystems based on long-term remote sensing data, the method comprising:

[0007] Acquire satellite remote sensing data, and construct a high-quality time-series remote sensing image stack based on the satellite remote sensing data;

[0008] Remote sensing images were classified based on the random forest method to obtain data on coastal land use types;

[0009] Based on the coastal zone surface land use type data, the carbon storage of the coastal zone ecosystem is calculated using the InVEST model to obtain the carbon storage accounting results of the coastal zone ecosystem;

[0010] Based on the coastal ecosystem carbon stock accounting results, statistics and analysis are conducted on the temporal variation trend of coastal ecosystem carbon stock.

[0011] Furthermore, the obtaining of satellite remote sensing data and constructing a high-quality time-series remote sensing image stack according to the satellite remote sensing data is specifically as follows:

[0012] According to the reflectance characteristics of different bands in remote sensing data, images with cloud cover less than 10% are screened, and cloud removal is performed using the quality assessment bands generated by the CFMASK algorithm in the GEE cloud platform;

[0013] A high-quality time-series remote sensing image stack involving multiple features is constructed through cloud-processed remote sensing data, wherein the multiple features include spectral feature parameters, texture feature parameters, thermal feature parameters, elevation feature parameters and principal component feature parameters.

[0014] Furthermore, the remote sensing images are classified based on the random forest method to obtain the coastal zone surface land use type data as follows:

[0015] The “Gini index” is used to rank the importance of feature parameters in high-quality image stacks;

[0016] Eliminate features with contributions less than 10% according to the sorting order, use the remaining features to update the remote sensing feature stack and retrain the random forest classification model, check whether the feature contribution is greater than the threshold, if not, repeat the operation until all feature contributions exceed 10% and stop iteration;

[0017] Recursive feature elimination was used to eliminate the optimal feature parameters, and the remote sensing data were classified through the random forest classification model to obtain the coastal land use type data.

[0018] Furthermore, the calculation formula of the Gini index is:

[0019]

[0020] Among them, K represents K categories, P k Represents the sample weight of category K.

[0021] Then feature X jThe importance of node m, that is, the change in the Gini index before and after the node m branch is:

[0022]

[0023] Among them, GI l and GI r Respectively represent the Gini index of the two new nodes after branching.

[0024] If feature X j The node that appears in decision tree i is in set M, then X j The importance of the i-th tree is:

[0025]

[0026] Assuming there are n trees in the random forest classifier, then feature X j The importance is shown in:

[0027]

[0028] Compared with other importance indicators, the Gini index is interpretable, non-parametric, and can better handle unbalanced data, making it an ideal choice when processing large-scale data sets.

[0029] Furthermore, the InVEST carbon storage module estimates carbon storage based on the pool substitution method, mainly using the LUCC conditions in different phases to calculate the total regional carbon storage and temporal and spatial distribution characteristics through four carbon pools: aboveground carbon storage, underground root carbon storage, soil carbon storage, and dead organic matter carbon storage. The carbon storage calculation formula is:

[0030] C=C above +C below +C soil +C dead

[0031] In the formula, C is the total carbon storage in the region, C above is the aboveground carbon storage, C below is the underground root carbon storage; C soil is soil carbon storage, C dead is the carbon storage of dead organic matter.

[0032] The InVEST carbon storage module runs based on the raster data images of the geographic information system. Each pixel represents a land cover type. For each land cover type, the model requires at least one of the above four carbon pools to be input to run. The more complete the carbon pool types and the more detailed the land use type division, the more accurate the results of the model calculation will be.

[0033] The present invention also provides a coastal ecosystem carbon storage estimation system based on long-term remote sensing data, the system comprising:

[0034] A data processing module, used for acquiring satellite remote sensing data and constructing a high-quality time-series remote sensing image stack based on the satellite remote sensing data;

[0035] The data classification module is used to classify remote sensing images based on the random forest method to obtain the surface land use type data of the coastal zone;

[0036] A carbon storage calculation module calculates the carbon storage of the coastal ecosystem through the InVEST model based on the coastal surface land use type data to obtain the carbon storage accounting results of the coastal ecosystem;

[0037] The carbon storage change module counts and analyzes the temporal change trend of the carbon storage of the coastal ecosystem based on the carbon storage accounting results of the coastal ecosystem.

[0038] The present invention also provides an electronic device, including a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the above-mentioned coastal ecosystem carbon storage estimation method based on long-term remote sensing data.

[0039] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method for estimating carbon reserves of coastal ecosystems based on long-term remote sensing data as described above.

[0040] Compared with the prior art, the present invention has the following advantages:

[0041] In view of the poor effect of traditional land use information extraction methods in complex coastal environments and the unclear spatiotemporal distribution of ecosystem carbon reserves, the present invention constructs a random forest land use classification model that integrates multiple characteristic parameters such as spectrum, texture, temperature, elevation and principal components, and optimizes characteristic parameters based on the Gini index threshold to achieve accurate classification of land use in coastal areas; then based on the InVEST carbon reserve model, the carbon reserve of coastal ecosystems is calculated on the premise of accurately determining the carbon density of different land use types, and spatiotemporal evolution analysis is carried out. The process of the present invention is easy to implement, the meaning is clear, and it can effectively supplement the existing carbon reserve accounting methods in space, which is conducive to the spatial and temporal expression of carbon reserves in coastal ecosystems, and can provide basic theory and technical support for scientific management of coastal resources and regional sustainable development, so it has important practical application significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The accompanying drawings generally illustrate various embodiments by way of example and not limitation, and together with the description and claims, serve to illustrate the embodiments of the invention. Where appropriate, the same reference numerals are used throughout the drawings to refer to the same or similar parts. Such embodiments are illustrative and are not intended to be exhaustive or exclusive embodiments of the present apparatus or method.

[0043] Figure 1 A schematic diagram of the method flow of the present invention is shown;

[0044] Figure 2 The land use classification result of the present invention is shown;

[0045] Figure 3 The schematic diagram of the spatial expression results of carbon storage of Hangzhou Bay ecosystem in 9 phases in the past 40 years (1985-2023) is shown;

[0046] Figure 4 The schematic diagram of carbon storage changes of four types of land features, namely, forest land, cultivated land, bare land and tidal flat, is shown in the present invention. DETAILED DESCRIPTION

[0047] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0048] Hangzhou Bay is located in the northeast of Zhejiang Province, my country. It is the outer coastal part of the Qiantang River estuary area. The bay mouth is connected to the Ningshao Plain in the south, Shanghai in the north, and Zhoushan Islands in the east. It is an important ecological monitoring area and a world-famous trumpet-shaped tidal estuary. The area around Hangzhou Bay has many rivers, lakes, and rich wetland resources, with rich topography and landforms. The silt carried by the rich water systems such as the Yangtze River and Qiantang River gradually accumulated and extended on the south bank of Hangzhou Bay under the action of hydrodynamics, forming a typical coastal salt marsh wetland composed of sub-clay, sub-sand and silt, which has important ecological significance.

[0049] GEE is a planetary-scale geospatial information processing and analysis platform. Due to its massive data storage and powerful computing power, it has become the world's most popular remote sensing big data processing platform.

[0050] According to the needs of long-term research, the present invention selects Landsat5, Landsat 7 and Landsat 9 images covering the study area on the GEE platform. Considering the influence of satellite revisit period, in order to obtain higher quality remote sensing images, the present invention selects the Landsat surface reflectance (SR) data set in the GEE cloud platform, sets a 1-year time window, and uses the median synthesis algorithm to synthesize inter-annual images to obtain remote sensing images that can clearly and completely display the surface coverage information of the study area. The images of 1985, 1990, 1995, 2000, 2005, and 2010 are from the Landsat 5 Thematic Mapper (TM), and the images of 2015 and 2020 are from the Landsat 8 Operational LandImager (OLI). The images of 2023 are from Landsat 9OLI-2. The specific information is shown in Table 1. The data products of the TM sensor are atmospherically corrected using the Land Resources Satellite Ecosystem Perturbation Adaptive Processing System (LEDAPS). Data products from the OLI / OLI-2 sensors are atmospherically corrected using the Land Surface Reflectance Code (LaSRC) integrated with the U.S. Geological Survey's in-house algorithms.

[0051] In order to solve the problems that traditional LUCC information extraction methods are ineffective in complex bay area environments and the spatiotemporal distribution of carbon stocks is still unclear, a LUCC dataset is obtained through the RF land use classification model with optimized feature parameters, and a method for estimating surface carbon stocks in coastal zones based on long-term remote sensing images is developed in combination with the InVEST model.

[0052] Table 1 Landsat satellite image information used in the present invention

[0053]

[0054] As an embodiment, the present invention uses the Landsat surface reflectance dataset as a data source and is implemented through the GEE cloud platform. The process is as follows: Figure 1 As shown:

[0055] Step 1: Acquire satellite remote sensing data, and construct a high-quality time-series remote sensing image stack based on the satellite remote sensing data;

[0056] Step 1 includes:

[0057] Step 1.1: According to the reflectance characteristics of different bands in remote sensing images, images with cloud cover less than 10% are selected, and the quality assessment band (QA) generated by the CFMASK algorithm in the GEE cloud platform is used for cloud removal;

[0058] Step 1.2: Build a high-quality image stack involving multiple features.

[0059] The multi-features include spectrum, texture, temperature, elevation, and principal component. The specific feature parameters are shown in Table 2.

[0060] Table 2 Feature parameters included in remote sensing image stack

[0061]

[0062] Step 2: Classify the remote sensing images based on the random forest method to obtain the coastal zone surface land use type data;

[0063] Considering the spatial resolution of Landsat data, the present invention refers to the National Remote Sensing Monitoring Land Use / Cover Classification System and combines the distribution characteristics of Hangzhou Bay land features to classify the land features in the study area into 6 categories, and the specific information is shown in Table 3. Among the selected sample points, 70% of the samples are used for training and 30% of the samples are used for verification.

[0064] Table 3 Classification scheme and image features

[0065]

[0066]

[0067] The present invention extracts 30 feature parameters to generate an image feature stack, and carries out feature parameter contribution ranking and feature recursive elimination based on the "Gini index importance". The present invention sets five groups of decision trees with an interval of 100. Through normalization and averaging, it accepts the influence of fluctuations to a certain extent while ensuring the accuracy of feature sorting.

[0068] The optimal feature combination after threshold optimization is input into the random forest classifier to achieve high-precision LUCC information extraction. Figure 2 This is the land use classification result.

[0069] In order to verify the accuracy of the results of the present invention, the accuracy of the land use classification data of each research year was evaluated. The overall accuracy and Kappa coefficient of each research year are shown in Table 4. The overall accuracy of each year exceeded 94.00%, and the Kappa coefficient exceeded 0.9300. The classification results are accurate and can provide relatively accurate data support for subsequent carbon storage analysis.

[0070] Table 4 Overall classification accuracy and Kappa coefficient for each research year

[0071]

[0072] Step 3: Combine the land use type data obtained in step 2 and use the InVEST model to calculate the carbon storage of coastal ecosystems;

[0073] This invention refers to the global above-ground and underground carbon density dataset released by NASA in 2020, the 2010 China terrestrial ecosystem carbon density dataset produced by the National Ecological Science Data Center based on literature review and field sampling, and combines the carbon density data of the adjacent Greater Bay Area. The reference values ​​of carbon density of various types of land features in Hangzhou Bay are determined as shown in Table 5.

[0074] Table 5 Reference values ​​of carbon density for each surface type of coastal ecosystem (t / km 2 )

[0075]

[0076] The present invention combines land use classification data and uses the Carbon Storage and Sequestration module in the InVEST software to calculate carbon reserves. Figure 2 This is the spatial expression result of carbon storage in the Hangzhou Bay ecosystem for 9 phases in the past 40 years (1985-2023).

[0077] Step 4: Analysis of spatiotemporal evolution of carbon storage in coastal ecosystems;

[0078] The statistics of ecosystem carbon storage, carbon storage changes and change rates of each land use type and research phase in Hangzhou Bay are shown in Table 6.

[0079] Table 6 Carbon storage, carbon storage change and change rate of Hangzhou Bay ecosystem

[0080]

[0081]

[0082] according to Figure 3 As can be seen from Table 6, the overall carbon storage in Hangzhou Bay has been declining, with an average annual decrease of 0.68 Mt. Compared with the previous research phase, each research phase decreased by 0.24, 1.86, 2.22, 3.25, 4.58, 7.08, 4.41, and 2.06 Mt, respectively, and the corresponding change rates were -0.22%, -1.72%, -2.09%, -3.13%, -4.55%, -7.37%, -4.96%, and -2.44%, respectively. The carbon storage in the Hangzhou Bay region has been declining in the past 40 years. From 1985 to 2015, the decline rate of carbon storage in Hangzhou Bay continued to increase, and from 2015 to 2023, the decline rate gradually slowed down.

[0083] The carbon storage changes of four types of landforms, namely forest land, cultivated land, bare land and tidal flats, are as follows: Figure 4shown.

[0084] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for estimating carbon storage in coastal ecosystems based on long-term remote sensing data, characterized in that: The method comprises: Acquire satellite remote sensing data, and construct a high-quality time-series remote sensing image stack based on the satellite remote sensing data; Remote sensing images were classified based on the random forest method to obtain data on coastal land use types; Based on the coastal zone surface land use type data, the carbon storage of the coastal zone ecosystem is calculated using the InVEST model to obtain the carbon storage accounting results of the coastal zone ecosystem; Based on the coastal ecosystem carbon stock accounting results, statistics and analysis are conducted on the temporal variation trend of coastal ecosystem carbon stock.

2. The method for estimating carbon storage in coastal ecosystems based on long-term remote sensing data according to claim 1 is characterized in that: The obtaining of satellite remote sensing data and constructing a high-quality time-series remote sensing image stack according to the satellite remote sensing data specifically includes: According to the reflectance characteristics of different bands in remote sensing data, images with cloud cover less than 10% are screened, and cloud removal is performed using the quality assessment bands generated by the CFMASK algorithm in the GEE cloud platform; Build a high-quality time-series remote sensing image stack with multiple features through cloud-processed remote sensing data.

3. The method for estimating carbon storage in coastal ecosystems based on long-term remote sensing data according to claim 1 is characterized in that: The multiple features include spectral feature parameters, texture feature parameters, thermal feature parameters, elevation feature parameters and principal component feature parameters.

4. The method for estimating carbon storage of coastal ecosystems based on long-term remote sensing data according to claim 1, characterized in that: The random forest method is used to classify remote sensing images and obtain coastal surface land use type data as follows: Use the "Gini index" to rank the importance of feature parameters in high-quality image stacks; Eliminate features with contributions less than 10% according to the sorting order, use the remaining features to update the remote sensing feature stack and retrain the random forest classification model, check whether the feature contribution is greater than the threshold, if not, repeat the operation until all feature contributions exceed 10% and stop iteration; Recursive feature elimination was used to eliminate the optimal feature parameters, and the remote sensing data were classified through the random forest classification model to obtain the coastal land use type data.

5. The method for estimating carbon storage of coastal ecosystems based on long-term remote sensing data according to claim 1, characterized in that: The calculation formula of the Gini index is: Among them, K represents K categories, P k Represents the sample weight of category K.

6. The method for estimating carbon storage in coastal ecosystems based on long-term remote sensing data according to claim 1, characterized in that: The carbon storage calculation formula is: C=C above +C below +C soil +C dead In the formula, C is the total carbon storage in the region, C above is the aboveground carbon storage, C below is the underground root carbon storage; C soil is soil carbon storage, C dead is the carbon storage of dead organic matter.

7. A coastal ecosystem carbon storage estimation system based on long-term remote sensing data, characterized in that: The system comprises: A data processing module, used for acquiring satellite remote sensing data and constructing a high-quality time-series remote sensing image stack based on the satellite remote sensing data; The data classification module is used to classify remote sensing images based on the random forest method to obtain the surface land use type data of the coastal zone; A carbon storage calculation module calculates the carbon storage of the coastal ecosystem through the InVEST model based on the coastal surface land use type data to obtain the carbon storage accounting results of the coastal ecosystem; The carbon storage change module counts and analyzes the temporal change trend of the carbon storage of the coastal ecosystem based on the carbon storage accounting results of the coastal ecosystem.

8. An electronic device, characterized in that: It comprises a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the coastal ecosystem carbon storage estimation method based on long-term remote sensing data according to claims 1-6.

9. A computer-readable storage medium, characterized in that: It stores a computer program, which, when executed by a processor, implements the coastal ecosystem carbon storage estimation method based on long-term remote sensing data as described in claims 1-6.

Images