Remote sensing prediction method and system for vegetation net primary productivity

Abstract

The invention provides a remote sensing prediction method and system for vegetation net primary productivity, and relates to the field of remote sensing prediction, and the method comprises the steps: obtaining a remote sensing image and meteorological data of marsh wetland vegetation in a research area, and carrying out the preprocessing; calculating vegetation net primary productivity data and hotspot analysis data through the remote sensing image and the meteorological data; constructing a training set through vegetation net primary productivity data and hotspot analysis data; a vegetation net primary productivity prediction model is constructed through a graph neural network and a Transform model; training the vegetation net primary productivity prediction model through the training set; and acquiring a current remote sensing image and meteorological data, and predicting future vegetation net primary productivity data in combination with the trained vegetation net primary productivity prediction model. According to the method, the GNN-Transform model is adopted, and the prediction precision is remarkably improved through deep mining of the spatial dependency relationship and the time evolution rule.

Description

Technical Field

[0001] This application relates to the field of remote sensing prediction, and particularly to a remote sensing prediction method and system for vegetation net primary productivity. Background Art

[0002] The prediction of vegetation net primary productivity information is an indispensable part of ecological environment monitoring and agricultural production management. For existing methods of predicting vegetation net primary productivity, such as the Chinese patent "A Remote Sensing Prediction Method and Device for Vegetation Net Primary Productivity, Application No. CN202410381278.1", this patent provides a remote sensing prediction method and device for vegetation net primary productivity. The Chinese patent "A Remote Sensing Prediction Method and Device for Vegetation Net Primary Productivity, Application No. CN202410381278.1", this patent provides a remote sensing prediction method and device for vegetation net primary productivity.

[0003] It mainly predicts by using machine learning or deep learning models to combine the vegetation net primary productivity (NPP) inverted from remote sensing data with the meteorological factors in the corresponding region, resulting in large errors in the prediction results when dealing with the vegetation growth process with complex spatio-temporal variations.

[0004] In practical applications, vegetation net primary productivity is affected by multiple factors, such as climate change, seasonal change, human activities, and natural disasters. These factors not only show obvious dynamic changes in time but also exhibit clustering characteristics in space. Therefore, single remote sensing data or meteorological data often cannot fully reflect the complexity and diversity of vegetation growth. In addition, considering the characteristics of vegetation distribution in different regions of our country, the climate conditions and ecological environments in the north (such as North China, Northwest China, and Northeast China) and the south (such as Jiangnan, South China, and Southwest China) are different, and it is difficult to collect climate data in many regions, which brings great challenges to the prediction of vegetation net primary productivity by remote sensing technology. Summary of the Invention

[0005] The purpose of the present invention is to provide a remote sensing prediction method and system for vegetation net primary productivity to solve the problem that the existing technical means cannot accurately predict the future vegetation net primary productivity of marsh wetlands.

[0006] The above object of this application is achieved by the following technical solutions: S1: Obtain remote sensing images and meteorological data of marsh wetland vegetation in the study area and perform preprocessing; S2: Calculate vegetation net primary productivity data and hotspot analysis data through the remote sensing images and meteorological data; S3: Construct a training set through the vegetation net primary productivity data and hotspot analysis data; S4: Construct a vegetation net primary productivity prediction model through a graph neural network and a Transformer model; S5: Train the vegetation net primary productivity prediction model with a training set; S6: Obtain the current remote sensing image and meteorological data, and combine with the trained vegetation net primary productivity prediction model to predict future vegetation net primary productivity data.

[0007] Optionally, step S1 includes: Select the distribution of unchanged swamp wetlands within a preset research time period as the research area; The meteorological data includes: monthly precipitation, monthly temperature, monthly solar radiation, monthly NDVI, and land cover type; Interpolate and resample the meteorological data. Optionally, step S2 includes: Use the CASA model to invert the preprocessed meteorological data and calculate the vegetation net primary productivity data.

[0008] Optionally, step S3 includes: Crop the vegetation net primary productivity data and hotspot analysis data into images with a size of 256×256 pixels and make them into a training set.

[0009] Optionally, step S6 includes: Based on the current remote sensing image and meteorological data, calculate the current vegetation net primary productivity data and current hotspot analysis data; input the current vegetation net primary productivity data and current hotspot analysis data into the trained vegetation net primary productivity prediction model to predict future vegetation net primary productivity data.

[0010] An electronic device includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions. The user interface and the network interface are used to communicate with other devices. The processor is used to execute the instructions stored in the memory so that the electronic device executes a remote sensing prediction method for vegetation net primary productivity.

[0011] A computer-readable storage medium stores instructions, and when the instructions are executed, a remote sensing prediction method for vegetation net primary productivity is executed.

[0012] The beneficial effects brought by the technical solution provided by this application are: Compared with the prior art, in the technical solution proposed by the present invention, by aggregating hotspot analysis data, the prediction model can better learn the spatial aggregation characteristics in the input data. By combining the time-series vegetation net primary productivity (NPP) data with the hotspot analysis data, the model can capture the spatio-temporal dynamics of vegetation changes simultaneously. In particular, the ability to identify stressed or abnormally changing areas in space has been significantly improved. Traditional deep learning methods fail to fully utilize this information, while the GNN-Transformer model adopted in the present invention significantly improves the prediction accuracy by deeply mining the spatial dependence relationship and the law of temporal evolution. Especially in the face of complex climate change, seasonal fluctuations, and human activity impacts, the model can accurately capture potential anomalies in vegetation growth, providing more reliable and accurate support for ecological environment monitoring, disaster warning, and agricultural production decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The following will further illustrate the present application in conjunction with the drawings and embodiments. In the drawings: Figure 1 is the flowchart in the embodiment of the present application; Figure 2 is the structure diagram of vegetation net primary productivity prediction in the embodiment of the present application; Figure 3 is the structure diagram of the deep learning automated prediction method in the embodiment of the present application; Figure 4 is the schematic diagram of the electronic device structure in the embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0014] In order to have a clearer understanding of the technical features, objectives, and effects of the present application, the specific embodiments of the present application will now be described in detail with reference to the drawings.

[0015] The embodiment of the present application provides a remote sensing prediction method for vegetation net primary productivity.

[0016] Please refer to Figure 1 , Figure 1 which is the flowchart of a remote sensing prediction method for vegetation net primary productivity in the embodiment of the present application, including: S1: Obtain the remote sensing images and meteorological data of the marsh wetland vegetation in the study area and perform preprocessing; S2: Calculate the vegetation net primary productivity data and hotspot analysis data through the remote sensing images and meteorological data; S3: Construct a training set through the vegetation net primary productivity data and hotspot analysis data; S4: Construct a vegetation net primary productivity prediction model through a graph neural network and a Transformer model; S5: Train the vegetation net primary productivity prediction model using the training set; S6: Obtain the current remote sensing image and meteorological data, and combine with the trained vegetation net primary productivity prediction model to predict the future vegetation net primary productivity data.

[0017] As an embodiment, select the distribution of marsh wetlands that has not changed during a certain research time period as the research area; obtain and preprocess the remote sensing data and meteorological data of marsh wetland vegetation in the research area; interpolate the meteorological data and perform resampling; extract the net primary productivity and the values of each meteorological element corresponding to all marsh vegetation pixels in the research area; construct a prediction model; combine with the meteorological data under the future climate change scenario, and use the prediction model to predict the future net primary productivity of marsh wetland vegetation. The present invention utilizes the advantages of large spatial scale and easy acquisition of remote sensing data, and combines the existing meteorological data to predict the net primary productivity of marsh vegetation under the influence of future climate change.

[0018] As an embodiment, Figure 2 is the structure diagram of the vegetation net primary productivity prediction method of the technical solution of the present invention. The CASA model is used to perform inversion calculation on the original data to obtain the vegetation net primary productivity data. The hotspot analysis is performed using the vegetation net primary productivity data to obtain its spatial aggregation information, and the two are used as the input data of the model, thereby greatly improving the learning ability of the deep learning model for spatial aggregation features and significantly improving the prediction effect.

[0019] As an embodiment, Figure 3 is the structure diagram of the deep learning automatic prediction method of the technical solution of the present invention. The GNN-Transformer is used as the backbone prediction model. This model can better learn the spatial distribution characteristics in the data through the graph neural network and better learn the changing trend of the data over time through the Transformer module.

[0020] Step S1 includes: Select the distribution of marsh wetlands that has not changed during the preset research time period as the research area; The meteorological data includes: monthly precipitation, monthly temperature, monthly solar radiation, monthly NDVI, and land cover type; Interpolate and resample the meteorological data.

[0021] As an embodiment, the monthly NDVI (Normalized Difference Vegetation Index) is an index used to evaluate the surface vegetation coverage and growth vitality.

[0022] Step S2 includes: The preprocessed meteorological data is inverted using the CASA model to calculate the vegetation net primary productivity data.

[0023] As an example, the CASA model (Carnegie-Ames-Stanford Approach) is a process-based remote sensing model mainly used to estimate the net primary productivity (NPP) of terrestrial ecosystems.

[0024] Step S3 includes: The vegetation net primary productivity data and the hotspot analysis data are cropped into images with a size of 256×256 pixels and made into a training set.

[0025] Step S6 includes: Based on the current remote sensing image and meteorological data, calculate the current vegetation net primary productivity data and the current hotspot analysis data; input the current vegetation net primary productivity data and the current hotspot analysis data into the trained vegetation net primary productivity prediction model to predict the future vegetation net primary productivity data.

[0026] A remote sensing prediction system for vegetation net primary productivity, the system includes: a training set construction module, a model training module, and a prediction module; The training set construction module, the model training module, and the prediction module are sequentially connected in series; The training set construction module is used to obtain the remote sensing image and meteorological data of the marsh wetland vegetation in the study area and perform preprocessing; The training set construction module is also used to calculate the vegetation net primary productivity data and the hotspot analysis data through the remote sensing image and the meteorological data; The training set construction module is also used to construct a training set through the vegetation net primary productivity data and the hotspot analysis data; The model training module is used to construct a vegetation net primary productivity prediction model through a graph neural network and a Transformer model; The model training module is also used to train the vegetation net primary productivity prediction model through the training set; The prediction module is used to obtain the current remote sensing image and meteorological data, and combine the trained vegetation net primary productivity prediction model to predict the future vegetation net primary productivity data.

[0027] As an embodiment, the present invention is based on the idea of using vegetation net primary productivity (NPP) data and hotspot analysis data as multi-modal input for a prediction model. By combining the rich temporal trend information in the temporal vegetation net primary productivity data with the rich spatial aggregation characteristics in the hotspot analysis data, the learning of the spatial aggregation information of the vegetation net primary productivity by the prediction model is improved. Through this multi-modal input method, the model can better capture the spatio-temporal dynamics of vegetation changes and identify potential stresses or abnormal changes during the vegetation growth process. The temporal vegetation net primary productivity data provides the changes in the productivity levels of vegetation at different time points, reflecting the impacts of factors such as climate change, seasonal changes, and human activities on vegetation growth. The hotspot analysis data, on the other hand, reveals the spatial aggregation patterns of vegetation growth in specific regions and can effectively identify high-risk or stressed regions, such as areas with abnormal vegetation growth caused by soil erosion, drought, pests, etc.

[0028] As an embodiment, in the actual environment, due to the large differences in vegetation net primary productivity in different time periods and spatial regions, traditional deep learning methods have not been able to effectively learn the temporal change trends and spatial aggregation characteristics of vegetation net primary productivity. By using the GNN-Transformer as the basic model and combining the graph neural network (GNN) and the Transformer model in deep learning methods, the present invention can effectively mine the spatial dependence relationships and temporal evolution laws in the data, thereby accurately predicting the vegetation net primary productivity at different time scales and spatial scales, improving the prediction accuracy and robustness, and providing more reliable decision-making support for ecological environment monitoring and agricultural production.

[0029] In another embodiment, the hotspot analysis of vegetation net primary productivity (NPP) data refers to analyzing the distribution of NPP data in the geographical space to identify regions with higher NPP values, which are called "hotspots". The core of hotspot analysis lies in identifying the high-value regions where NPP values are spatially aggregated, and these regions usually indicate higher vegetation productivity and better ecosystem health conditions.

[0030] This application also discloses an electronic device. Referring to Figure 4 , Figure 4 is a schematic structural diagram of an electronic device disclosed in an embodiment of this application. The electronic device 500 may include: at least one processor 501, at least one network interface 504, a user interface 503, a memory 505, and at least one communication bus 502.

[0031] Among them, the communication bus 502 is used to realize the connection and communication between these components.

[0032] Among them, the user interface 503 may include a display screen. Optionally, the user interface 503 may further include a standard wired interface and a wireless interface.

[0033] Among them, the network interface 504 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface).

[0034] This application also discloses a computer-readable storage medium, which stores multiple instructions. The instructions are suitable for being loaded by a processor to execute the above-mentioned remote sensing prediction method for vegetation net primary productivity.

[0035] The above are only exemplary embodiments of the present disclosure, and the scope of the present disclosure cannot be limited thereby. That is, any equivalent changes and modifications made according to the teachings of the present disclosure still fall within the scope covered by the present disclosure.

[0036] This application aims to cover any variations, uses or adaptations of the present disclosure. These variations, uses or adaptations follow the general principles of the present disclosure and include well-known common knowledge or conventional technical means in the technical field not recorded in the present disclosure. The description and embodiments are only regarded as exemplary, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A remote sensing prediction method for vegetation net primary productivity, characterized in that: The method comprises the following steps: S1: Obtain remote sensing images and meteorological data of marsh wetland vegetation in the study area and perform preprocessing; S2: Calculate vegetation net primary productivity data and hotspot analysis data through remote sensing images and meteorological data; S3: Construct a training set using vegetation net primary productivity data and hotspot analysis data; S4: Construct a vegetation net primary productivity prediction model through graph neural network and Transformer model; S5: Train the vegetation net primary productivity prediction model using the training set; S6: Obtain current remote sensing images and meteorological data, and combine them with the trained vegetation net primary productivity prediction model to predict future vegetation net primary productivity data.

2. A remote sensing prediction method for vegetation net primary productivity according to claim 1, characterized in that: Step S1 includes: The distribution of marsh wetlands that did not change during the preset study period was selected as the study area; Meteorological data include: monthly precipitation, monthly temperature, monthly solar radiation, monthly NDVI and land cover type; Interpolate and resample meteorological data.

3. The remote sensing prediction method of vegetation net primary productivity according to claim 1, characterized in that: Step S2 includes: The CASA model was used to invert the preprocessed meteorological data and calculate the vegetation net primary productivity data.

4. The remote sensing prediction method of vegetation net primary productivity according to claim 1, characterized in that: Step S3 includes: The vegetation net primary productivity data and hotspot analysis data were cropped into images of size 256 × 256 pixels and made into a training set.

5. The remote sensing prediction method of vegetation net primary productivity according to claim 1, characterized in that: Step S6 includes: According to the current remote sensing images and meteorological data, the current vegetation net primary productivity data and the current hot spot analysis data are calculated; the current vegetation net primary productivity data and the current hot spot analysis data are input into the trained vegetation net primary productivity prediction model to predict the future vegetation net primary productivity data.

6. A remote sensing prediction system for net primary productivity of vegetation, used to implement a remote sensing prediction method for net primary productivity of vegetation as claimed in any one of claims 1 to 5, characterized in that: The system comprises: a training set construction module, a model training module and a prediction module; The training set construction module, the model training module and the prediction module are connected in sequence; The training set construction module is used to obtain remote sensing images and meteorological data of marsh wetland vegetation in the study area and perform preprocessing; The training set building module is also used to calculate vegetation net primary productivity data and hot spot analysis data through remote sensing images and meteorological data; The training set construction module is also used to construct a training set through vegetation net primary productivity data and hot spot analysis data; The model training module is used to construct a vegetation net primary productivity prediction model through a graph neural network and a Transformer model; The model training module is also used to train the vegetation net primary productivity prediction model through a training set; The prediction module is used to obtain current remote sensing images and meteorological data, and combine them with the trained vegetation net primary productivity prediction model to predict future vegetation net primary productivity data.

7. An electronic device, characterized in that: It includes a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes the method as described in any one of claims 1-5.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed by a computer, the method according to any one of claims 1 to 5 is executed.

Images