Regional terrestrial ecosystem carbon reserve estimation method

Abstract

The invention discloses a regional terrestrial ecosystem carbon reserve estimation method, and belongs to the technical field of ecosystem carbon reserve estimation. The method is realized through five steps of data acquisition, remote sensing data downloading and preprocessing, data rasterization processing, feature extraction and carbon reserve estimation. By combining the satellite remote sensing data and the ground survey data, the carbon reserve in the region can be estimated more accurately, and the precision and reliability of the estimation result are improved by the multi-source data fusion method. By performing multi-band feature extraction on the remote sensing image, the vegetation coverage and type features can be comprehensively known, so that the carbon reserve can be estimated more accurately. The method not only considers the carbon reserve of the vegetation layer, but also considers the carbon reserve of the litter layer, the withered wood layer and the soil layer. The multi-level estimation method is more comprehensive and can better reflect the carbon reserve state of the ecological system.

Description

Technical Field

[0001] The present invention relates to the technical field of ecosystem carbon storage estimation, and in particular to a method for estimating carbon storage of a regional terrestrial ecosystem. Background Art

[0002] The estimation of carbon storage is of great significance in global climate change research and ecological environmental protection. As an important part of the earth's carbon cycle, terrestrial ecosystems, especially forests, wetlands and other ecosystems, absorb carbon dioxide from the atmosphere through photosynthesis and fix carbon, playing a key role in mitigating the greenhouse effect. Accurately estimating the carbon storage of terrestrial ecosystems is of great guiding significance for the formulation of scientific carbon emission control policies and management strategies.

[0003] Traditional carbon stock estimation methods mainly rely on ground survey data, such as biomass measurement and soil carbon content analysis. Although these methods are highly accurate, they are time-consuming and labor-intensive, and difficult to promote and apply on a large scale. With the development of remote sensing technology, satellite remote sensing data can provide large-scale and continuous ecosystem information, making it possible to achieve large-scale carbon stock estimation.

[0004] Remote sensing data includes vegetation cover types, vegetation index, terrain data and meteorological data, which can comprehensively reflect the state and changes of ecosystems from a spatial perspective. By combining remote sensing technology with ground survey data, multi-source data can be integrated to improve the accuracy and reliability of carbon storage estimation. At the same time, the application of machine learning algorithms in remote sensing data processing and carbon storage estimation makes automation and efficiency possible.

[0005] In the existing technology, many studies have begun to use satellite remote sensing data to estimate carbon storage, but there are generally some shortcomings. For example, the cloud and fog obstruction and noise processing are not sufficient, resulting in data quality damage; the lack of spatiotemporal dynamic monitoring means makes it difficult to capture the spatiotemporal changes of carbon storage; the data processing and estimation methods are relatively simple, and the comprehensive impact of different vegetation types and carbon storage at all levels is not fully considered. Summary of the invention

[0006] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a method for estimating carbon stocks in regional terrestrial ecosystems, which has the following advantages: by combining satellite remote sensing data and ground survey data, the carbon stocks in the region can be estimated more accurately; through spatiotemporal raster data and spatiotemporal convolutional neural networks, the spatiotemporal changes of carbon stocks can be dynamically monitored; convolutional neural networks and other machine learning algorithms are used to automatically process data, thereby improving the efficiency and quality of data processing.

[0007] The above object of the present invention is achieved through the following technical solutions:

[0008] A method for estimating carbon storage of regional terrestrial ecosystems comprises the following steps:

[0009] 1) Data collection: Collect satellite remote sensing data and ground survey data of the area to be collected; the satellite remote sensing data includes vegetation cover type, vegetation index, terrain data and meteorological data, and the ground survey data includes vegetation biomass, soil type and carbon content;

[0010] 2) Download and preprocess remote sensing data: According to the remote sensing data collected in step 1) and the size of the regional scale, download the remote sensing data of the corresponding frequency and preprocess the data;

[0011] 3) Data rasterization: The pre-processed data is rasterized to obtain a first raster map; the raster data includes spatial information and variable values ​​of each raster; spatial coordinates are obtained, a raster image is formed on the coordinates, and then gridding is performed to obtain a gridded raster image, and finally the image is expanded;

[0012] 4) Feature extraction: The remote sensing image data in step 3) is read into a computer, and the computer generates a vector for each data point in the read remote sensing image according to its band to form a matrix as a training sample set. The vegetation coverage sensing feature and the vegetation type sensing feature are extracted from the sensor data of the collection area according to the training sample set;

[0013] 5) Carbon stock estimation: A machine learning algorithm is used to construct a wetland carbon stock estimation model. Remote sensing data and meteorological data are used to realize the net primary productivity of vegetation. The vegetation coverage characteristics are input into the classification model to obtain the vegetation type of each vegetation coverage sub-area. The vegetation characteristics are input into the vegetation quantity prediction model to obtain the vegetation quantity of each vegetation coverage sub-area. The carbon storage of the vegetation type and vegetation quantity per unit area of ​​litter layer and per unit area of ​​dead wood layer of each vegetation coverage sub-area is calculated to calculate the carbon storage per unit area of ​​soil layer.

[0014] Furthermore, in the step 1), the data collection of carbon dioxide emissions is added, obtained through an API and integrated into the main data set.

[0015] Furthermore, in step 2), data quality assessment standards and processes are defined and implemented, and convolutional neural networks are used to perform cloud and fog occlusion processing and noise processing to improve data quality.

[0016] Furthermore, in the step 3), time information is added to the raster data to form spatiotemporal raster data, the spatiotemporal raster data is stored in GeoTIFF format, and the spatiotemporal convolutional neural network and spatiotemporal prediction model are used to capture the dynamic changes of carbon storage.

[0017] Preferably, the total amount of vegetation carbon storage is calculated as follows:C =sum{i=1} {n} (B i ×R i ×S i ), where V C represents the total carbon storage of vegetation; B i represents the biomass of the i-th vegetation type; R i represents the carbon content of the i-th vegetation type; S i represents the area of ​​the i-th vegetation type; n represents the total number of vegetation types.

[0018] Preferably, the formula B{total}=(a·V+b)·A{total} is also included for estimating the total biomass of a certain tree species, wherein B{total} is the total biomass of a certain tree species; B is the biomass per unit area of ​​a certain tree species; V is the stock per unit area of ​​a certain tree species; A{total} is the total area of ​​a certain tree species; and a and b are model parameters.

[0019] In summary, the present invention includes at least one of the following beneficial technical effects:

[0020] 1) The present invention can more accurately estimate the carbon storage in a region by combining satellite remote sensing data and ground survey data. This multi-source data fusion method improves the accuracy and reliability of the estimation results.

[0021] 2) Through spatiotemporal raster data and spatiotemporal convolutional neural networks, the spatiotemporal changes of carbon storage can be dynamically monitored, which is very important for long-term ecological monitoring and management, and helps to timely detect and respond to changes in ecosystems.

[0022] 3) Automatically process data using convolutional neural networks and other machine learning algorithms, improving the efficiency and quality of data processing. Obtain and integrate meteorological data and carbon dioxide emission data through APIs, and flexibly expand and update the main data set, which improves the richness and real-time nature of the data.

[0023] 4) By extracting multi-band features from remote sensing images, we can fully understand the vegetation coverage and type characteristics, and thus estimate carbon storage more accurately.

[0024] 5) The method of the present invention not only considers the carbon storage of the vegetation layer, but also considers the carbon storage of the litter layer, dead wood layer and soil layer. This multi-level estimation method is more comprehensive and can better reflect the carbon storage status of the ecosystem. DETAILED DESCRIPTION

[0025] Example 1

[0026] A method for estimating carbon storage of regional terrestrial ecosystems comprises the following steps:

[0027] 1) Data collection: Collect satellite remote sensing data and ground survey data of the area to be collected; the satellite remote sensing data includes vegetation cover type, vegetation index, terrain data and meteorological data, and the ground survey data includes vegetation biomass, soil type and carbon content; by adding data collection on carbon dioxide emissions, obtain and integrate it into the main data set through API.

[0028] 2) Download and preprocess remote sensing data: According to the remote sensing data collected in step 1) and the size of the regional scale, download the remote sensing data of the corresponding frequency and preprocess the data; define and implement data quality assessment standards and processes, use convolutional neural networks to perform cloud occlusion processing and noise processing, and improve data quality.

[0029] 3) Data rasterization: The preprocessed data is rasterized to obtain the first raster map; the raster data includes the spatial information and variable values ​​of each grid; the spatial coordinates are obtained, a raster image is formed on the coordinates, and then gridding is performed to obtain a gridded raster image, and finally the image is expanded; time information is added to the raster data to form spatiotemporal raster data, the GeoTIFF format is used to store the spatiotemporal raster data, and the spatiotemporal convolutional neural network and spatiotemporal prediction model are used to capture the dynamic changes of carbon storage.

[0030] 4) Feature extraction: The remote sensing image data in step 3) is read into a computer, and the computer generates a vector for each data point in the read remote sensing image according to its band to form a matrix as a training sample set. The vegetation coverage sensing feature and the vegetation type sensing feature are extracted from the sensor data of the collection area according to the training sample set;

[0031] 5) Carbon stock estimation: A machine learning algorithm is used to construct a wetland carbon stock estimation model. Remote sensing data and meteorological data are used to realize the net primary productivity of vegetation. The vegetation coverage characteristics are input into the classification model to obtain the vegetation type of each vegetation coverage sub-area. The vegetation characteristics are input into the vegetation quantity prediction model to obtain the vegetation quantity of each vegetation coverage sub-area. The carbon storage of the vegetation type and vegetation quantity per unit area of ​​litter layer and per unit area of ​​dead wood layer of each vegetation coverage sub-area is calculated to calculate the carbon storage per unit area of ​​soil layer.

[0032] The total amount of vegetation carbon storage is calculated as: V C =sum{i=1} {n} (B i ×R i ×S i ), where V C represents the total carbon storage of vegetation; B i represents the biomass of the i-th vegetation type; R irepresents the carbon content of the i-th vegetation type; S i represents the area of ​​the i-th vegetation type; n represents the total number of vegetation types. The formula B{total}=(a·V+b)·A{total} is used to estimate the total biomass of a tree species, where B{total} is the total biomass of a tree species; B is the biomass per unit area of ​​a tree species; V is the stock per unit area of ​​a tree species; A{total} is the total area of ​​a tree species; a and b are model parameters.

[0033] 1. Obtain and integrate data through API

[0034] ① Acquiring and integrating data involves the following steps:

[0035] ② Select API service: Determine the API that provides the required data. For example, meteorological data can be obtained through the OpenWeatherMap API, and carbon dioxide emissions data can be obtained through the Carbon Monitor API.

[0036] ③Send request: Send HTTP request according to API documentation to obtain data. Usually API requires API key and query parameters.

[0037] ④Parse data: The data returned by the API is usually in JSON or XML format and needs to be parsed.

[0038] ⑤Integrate into the main dataset: Integrate the parsed data into the existing dataset. Here, you can use Python’s Pandas library to store the data neatly in a DataFrame.

[0039] Sample code (take weather data as an example):

[0040] import requests

[0041] import pandas as pd

[0042] #Define API URL and parameters

[0043] api_url="http: / / api.openweathermap.org / data / 2.5 / weather"

[0044] params={

[0045] 'q':'CityName',

[0046] 'appid':'YourAPIKey'

[0047] }

[0048] #Send request and get data

[0049] response=requests.get(api_url,params=params)

[0050] data = response.json()

[0051] #Parse the data and integrate it into the main dataset

[0052] main_dataset = pd.DataFrame() #Assuming that there is a main data set

[0053] weather_data = {

[0054] 'temperature':data['main']['temp'],

[0055] 'humidity':data['main']['humidity'],

[0056] 'pressure':data['main']['pressure']

[0057] }

[0058] #Create DataFrame and merge

[0059] weather_df=pd.DataFrame([weather_data])

[0060] main_dataset=pd.concat([main_dataset,weather_df],axis=1)

[0061] 2. Image expansion processing

[0062] Image dilation is a morphological operation in image processing, the main purpose of which is to expand the target area in the image. Commonly used libraries are OpenCV and scikit-image.

[0063] Sample code:

[0064] import cv2

[0065] import numpy as np

[0066] #Read in the image

[0067] image=cv2.imread('image.tif',0)

[0068] #Define expansion structure

[0069] kernel=np.ones((5,5),np.uint8)

[0070] #Expansion processing

[0071] dilated_image=cv2.dilate(image,kernel,iterations=1)

[0072] #Save the results

[0073] cv2.imwrite('dilated_image.tif',dilated_image)

[0074] 3. Generate vectors and form matrices based on bands

[0075] For remote sensing images, each pixel has multiple bands of reflectance values. By reading the image data, the band of each pixel forms a vector, and the band vectors of all pixels form a matrix (or a data cube), where each row represents multiple band values ​​of a pixel.

[0076] Sample code:

[0077] import numpy as np

[0078] from osgeo import gdal

[0079] #Read remote sensing image data

[0080] dataset=gdal.Open('image.tif')

[0081] bands=[dataset.GetRasterBand(i+1).ReadAsArray()for iin range(dataset.RasterCount)]

[0082] #Stack the band data into a three-dimensional array (cube)

[0083] image_data = np.dstack(bands)

[0084] num_pixels=image_data.shape[0]*image_data.shape[1]

[0085] num_bands=image_data.shape[2]

[0086] #Flatten the 3D array into a matrix

[0087] matrix=image_data.reshape((num_pixels,num_bands))

[0088] print(matrix.shape)#(num_pixels,num_bands)

[0089] Each row of this matrix represents a pixel, and each column represents the reflectance value of a band. For example, if the image is 100×100 pixels and has 4 bands, the shape of the flattened matrix will be (10000,4).

[0090] The embodiments of the specific implementation methods of the present invention are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention.

Claims

1. A method for estimating carbon storage in regional terrestrial ecosystems, characterized in that The steps include: 1) Data collection: Collect satellite remote sensing data and ground survey data of the area to be collected; the satellite remote sensing data includes vegetation cover type, vegetation index, terrain data and meteorological data, and the ground survey data includes vegetation biomass, soil type and carbon content; 2) Download and preprocess remote sensing data: According to the remote sensing data collected in step 1) and the size of the regional scale, download the remote sensing data of the corresponding frequency and preprocess the data; 3) Data rasterization: The pre-processed data is rasterized to obtain a first raster map; the raster data includes spatial information and variable values ​​of each raster; spatial coordinates are obtained, a raster image is formed on the coordinates, and then gridding is performed to obtain a gridded raster image, and finally the image is expanded; 4) Feature extraction: The remote sensing image data in step 3) is read into a computer, and the computer generates a vector for each data point in the read remote sensing image according to its band to form a matrix as a training sample set. The vegetation coverage sensing feature and the vegetation type sensing feature are extracted from the sensor data of the collection area according to the training sample set; 5) Carbon stock estimation: A machine learning algorithm is used to construct a wetland carbon stock estimation model. Remote sensing data and meteorological data are used to realize the net primary productivity of vegetation. The vegetation coverage characteristics are input into the classification model to obtain the vegetation type of each vegetation coverage sub-area. The vegetation characteristics are input into the vegetation quantity prediction model to obtain the vegetation quantity of each vegetation coverage sub-area. The carbon storage of the vegetation type and vegetation quantity per unit area of ​​litter layer and per unit area of ​​dead wood layer of each vegetation coverage sub-area is calculated to calculate the carbon storage per unit area of ​​soil layer.

2. A method for estimating carbon storage in regional terrestrial ecosystems according to claim 1, characterized in that In the step 1), data collection of carbon dioxide emissions is added, obtained through API and integrated into the main data set.

3. A method for estimating carbon storage in regional terrestrial ecosystems according to claim 1, characterized in that In the step 2), data quality assessment standards and processes are defined and implemented, and convolutional neural networks are used to perform cloud and fog occlusion processing and noise processing to improve data quality.

4. A method for estimating carbon storage in regional terrestrial ecosystems according to claim 1, characterized in that In the step 3), time information is added to the raster data to form spatiotemporal raster data, the spatiotemporal raster data is stored in GeoTIFF format, and the spatiotemporal convolutional neural network and spatiotemporal prediction model are used to capture the dynamic changes of carbon storage.

5. A method for estimating carbon storage in regional terrestrial ecosystems according to claim 1, characterized in that The total amount of vegetation carbon storage is calculated as follows: V C =sum{i=1} {n} (B i ×R i ×S i ), where V C represents the total carbon storage of vegetation; B i represents the biomass of the i-th vegetation type; R i represents the carbon content of the i-th vegetation type; S i represents the area of ​​the i-th vegetation type; n represents the total number of vegetation types.

6. A method for estimating carbon storage in regional terrestrial ecosystems according to claim 1, characterized in that It also includes the formula B{total}=(a·V+b)·A{total} for estimating the total biomass of a certain tree species, where B{total} is the total biomass of a certain tree species; B is the biomass per unit area of ​​a certain tree species; V is the stock per unit area of ​​a certain tree species; A{total} is the total area of ​​a certain tree species; and a and b are model parameters.