CN-121859067-B - Method for extracting spring climates of evergreen broad-leaved forest

CN121859067BCN 121859067 BCN121859067 BCN 121859067BCN-121859067-B

Abstract

The application discloses an extraction method of evergreen broadleaf forest spring climates, and relates to the technical field of ecological observation, comprising the steps of carrying out negative value pixel correction and time interpolation on collected sunlight-induced chlorophyll fluorescence data; the method comprises the steps of carrying out fusion on processed sunlight-induced chlorophyll fluorescence data and temperature variable data, constructing a time sequence, carrying out fitting treatment to obtain a target time sequence, respectively extracting evergreen broadleaf forest spring climates based on the target time sequence by adopting a plurality of preset vegetation climates extraction methods, carrying out precision evaluation on the extracted evergreen broadleaf forest spring climates by using ground observation data, and determining the optimal evergreen broadleaf forest spring climates from the extracted evergreen broadleaf forest spring climates based on precision evaluation results. According to the application, through negative value pixel correction and time interpolation processing and fusion of sunlight-induced chlorophyll fluorescence data and temperature variable data, the accuracy of the result is improved by adopting the combination of a multi-fitting method and a multi-extraction method.

Inventors

GE ZHONGXI
TANG FENG
TANG BOHUI
FU ZHITAO

Assignees

昆明理工大学

Dates

Publication Date: 20260512
Application Date: 20260317

Claims (7)

1. The method for extracting the spring climate of the evergreen broad-leaved forest is characterized by comprising the following steps of: Performing negative value pixel correction and time interpolation on the collected sunlight-induced chlorophyll fluorescence data; Fusing the processed sunlight-induced chlorophyll fluorescence data with temperature variable data, constructing a time sequence, and performing fitting treatment to obtain a target time sequence, wherein the temperature variable data comprises daytime temperature, night temperature and average temperature, and the fusion is performed according to the following formula: sunlight-induced chlorophyll fluorescence data fused with daytime temperature: sunlight-induced chlorophyll fluorescence data fusion at night temperature: sunlight-induced chlorophyll fluorescence data fused with average temperature: Wherein, SIF is sunlight-induced chlorophyll fluorescence data, T max is daytime temperature, T min is nighttime temperature, T mean is average temperature, ST max is time sequence of SIF fused with daytime temperature, ST min is time sequence of SIF fused with nighttime temperature, ST mean is time sequence of SIF fused with average temperature; Respectively extracting evergreen broadleaf forest spring weather based on the target time sequences by adopting a plurality of preset vegetation weather extraction methods, wherein for each group of the target time sequences, performing weather parameter inversion by respectively applying a dynamic threshold method and a double-logic function fitting method to extract a plurality of evergreen broadleaf forest spring weather results; The method comprises the steps of carrying out precision evaluation on the extracted evergreen broad-leaved forest spring weather by using ground observation data, wherein the ground observation data comprises a weather period extracted from total primary productivity data of a flux station and a weather period recorded by ground manual observation; matching the extracted multiple evergreen broad-leaved forest spring climatic period results with the ground observation data to form a verification sample pair; Adopting at least one index of a decision coefficient, a root mean square error and a significance test to carry out statistical analysis on the verification sample pair, and quantitatively extracting consistency and deviation between a evergreen broadleaf forest spring season result and a ground truth value; based on the accuracy evaluation result, an optimal evergreen broadleaf forest spring climate is determined from the extracted evergreen broadleaf forest spring climate.
2. The method for extracting the spring climate from the evergreen broad-leaved forest according to claim 1, wherein the method for performing negative value pixel correction and time interpolation on the collected sunlight-induced chlorophyll fluorescence data further comprises: Calculating a mean value of sunlight-induced chlorophyll fluorescence data at a position corresponding to each negative value pixel in the sunlight-induced chlorophyll fluorescence data; extracting seasonal time periods close to the time points of the negative image elements, and calculating a sunlight-induced chlorophyll fluorescence data average value in the seasonal time periods; Performing weighted average calculation according to the historical sunlight-induced chlorophyll fluorescence data and the current sunlight-induced chlorophyll fluorescence data to obtain a correction value for correction; And performing time interpolation processing on the corrected sunlight-induced chlorophyll fluorescence data by using a preset neural network model.
3. The method for extracting the verdant-green broad-leaved forest spring climate according to claim 2, wherein the time interpolation processing is performed on the corrected sunlight-induced chlorophyll fluorescence data by using a preset neural network model, and the method further comprises: The preset neural network model is a long-term and short-term memory neural network model which introduces an attention mechanism; the long-term and short-term memory neural network model is trained through a self-defined loss function, and the self-defined loss function at least comprises: a mean square error term for constraining a numerical difference between the predicted value and the true value; a first derivative smoothing term for constraining smoothness of the output sequence; the shape constraint item is used for punishing the prediction result of the transformation trend of the reverse U-shaped transformation; And performing time interpolation processing on the missing time periods in the input data through the neural network model.
4. The method for extracting a spring climate from an evergreen broadleaf forest according to claim 1, wherein the determining an optimal evergreen broadleaf forest spring climate from the extracted evergreen broadleaf forest spring climate based on the accuracy evaluation result further comprises: comparing the precision evaluation indexes of all the spring climatic period results obtained by combining different temperature variables, different fitting methods and different climatic extraction methods; And selecting an optimal result from the climatic extraction results as a final evergreen broadleaf forest spring climatic period according to a preset screening criterion.
5. An extraction element of evergreen broadleaf woods spring weather, characterized by comprising: The data processing module is configured to perform negative value pixel correction processing and time interpolation processing on the collected sunlight-induced chlorophyll fluorescence data; The time sequence module is configured to fuse the processed sunlight-induced chlorophyll fluorescence data with temperature variable data, construct a time sequence and perform fitting processing to obtain a target time sequence; The temperature variable data comprises daytime temperature, night temperature and average temperature, and the fusion is carried out according to the following formula: sunlight-induced chlorophyll fluorescence data fused with daytime temperature: sunlight-induced chlorophyll fluorescence data fusion at night temperature: sunlight-induced chlorophyll fluorescence data fused with average temperature: Wherein, SIF is sunlight-induced chlorophyll fluorescence data, T max is daytime temperature, T min is nighttime temperature, T mean is average temperature, ST max is time sequence of SIF fused with daytime temperature, ST min is time sequence of SIF fused with nighttime temperature, ST mean is time sequence of SIF fused with average temperature; The system comprises a target time sequence, a climatic extraction module, a dynamic threshold method and a double-logic function fitting method, wherein the climatic extraction module is configured to respectively extract evergreen broadleaf forest spring climates by adopting a plurality of preset vegetation climatic extraction methods based on the target time sequence, and each group of target time sequences is respectively subjected to climatic parameter inversion by adopting the dynamic threshold method and the double-logic function fitting method so as to extract a plurality of evergreen broadleaf forest spring climatic period results; the precision evaluation module is configured to evaluate the precision of the extracted evergreen broad-leaved forest spring weather by using ground observation data, wherein the ground observation data comprises a weather period extracted by the total primary productivity data of the flux station and a weather period recorded by ground manual observation; matching the extracted multiple evergreen broad-leaved forest spring climatic period results with the ground observation data to form a verification sample pair; Adopting at least one index of a decision coefficient, a root mean square error and a significance test to carry out statistical analysis on the verification sample pair, and quantitatively extracting consistency and deviation between a evergreen broadleaf forest spring season result and a ground truth value; a result confirmation module configured to determine an optimal evergreen broadleaf spring climate from the extracted evergreen broadleaf spring climate based on the accuracy evaluation result.
6. An extraction device for evergreen broadleaf forest spring climates, characterized by comprising at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program, which when executed by the processing unit, causes the processing unit to perform the steps of the method for extracting evergreen broadleaf forest spring climates according to any one of claims 1 to 4.
7. A storage medium storing a computer program for execution by an extraction device of evergreen broadleaf forest spring climates, the computer program, when run on the extraction device of evergreen broadleaf forest spring climates, causing the extraction device of evergreen broadleaf forest spring climates to perform the steps of the method of evergreen broadleaf forest spring climates of any one of claims 1 to 4.

Description

Method for extracting spring climates of evergreen broad-leaved forest Technical Field The application relates to the technical field of ecological observation, in particular to a method for extracting the spring climate of evergreen broadleaf forests. Background Vegetation weather is a key index for researching periodic events (such as germination, flowering, fruiting and defoliation) and interactions thereof with environmental factors in the vegetation life cycle, and has important significance for understanding the carbon circulation of an ecosystem, climate change response and biodiversity protection. The evergreen broad-leaved forest is taken as one of the important forest types worldwide, and the accurate extraction of the spring climate is not only beneficial to monitoring vegetation growth dynamics, but also provides data support for climate models and ecological predictions. In recent years, with the development of remote sensing technology, sunlight-induced chlorophyll fluorescence (SIF) data can directly reflect the photosynthesis activity of vegetation to become an emerging means of monitoring the climate, and particularly, the sunlight-induced chlorophyll fluorescence (SIF) data has potential in complex ecosystems such as evergreen broadleaf forests. However, how to efficiently and accurately extract spring climates using SIF data remains a hotspot and difficulty in current research. In the prior art, the extraction method of broadleaf forest weathers mainly depends on remote sensing vegetation index and surface reflectivity data, and weathered event points are identified through time sequence analysis. For example, some methods use Savitzky-Golay (savitz-Golay) filtering to smooth the NDVI (normalized vegetation index) time sequence and then determine the weather period based on threshold or change point detection, and other methods use environmental data such as temperature or precipitation as auxiliary variables to improve extraction accuracy through data fusion or machine learning models. In recent years, as the availability of SIF data increases, some studies have begun to attempt to combine SIF with traditional indices or to perform weatheridentification directly based on SIF time series using an analog weathered extraction algorithm to exploit the sensitivity of SIF to the physiological state of vegetation. However, the prior art has several drawbacks, especially the problem of insufficient extraction accuracy. Firstly, the traditional vegetation index is easily influenced by background noise, seasonal saturation effect and cloud pollution in evergreen broad-leaved forest, so that the identification deviation of climatic events is larger, and the subtle change of the climates in spring cannot be accurately captured. Secondly, the existing SIF data processing often ignores the problems of negative image elements and time sequence discontinuity, and the direct application can lead to the reduction of data quality, thereby affecting the reliability of the climatic extraction. In addition, most methods do not fully consider the synergistic effect of key environmental factors such as temperature, and the like, and only rely on a single data source, so that the adaptability of the model in a complex environment is limited. These drawbacks make it difficult to meet the need for high-precision climate monitoring in the prior art, and a method that can integrate multi-source data, optimize process flows, and improve accuracy based on evaluation feedback is needed. Disclosure of Invention Aiming at least one defect or improvement requirement of the prior art, the invention provides an extraction method of the evergreen broadleaf forest spring climate, which is used for solving the problems that the extraction of the evergreen broadleaf forest spring climate in the prior art does not fully consider the synergic effect of key environmental factors such as temperature and the like, only depends on a single data source, is easily influenced by background noise, seasonal saturation effect and cloud pollution, causes larger climate identification deviation and has low extraction precision of the spring climate. In order to achieve the above object, according to a first aspect of the present invention, there is provided a method for extracting a spring climate of a evergreen broadleaf forest, comprising: Performing negative value pixel correction and time interpolation on the collected sunlight-induced chlorophyll fluorescence data; fusing the processed sunlight-induced chlorophyll fluorescence data with temperature variable data, constructing a time sequence, and performing fitting treatment to obtain a target time sequence; respectively extracting the spring weather of the evergreen broad-leaved forest by adopting a plurality of preset vegetation weather extraction methods based on a target time sequence; performing precision evaluation on the extracted spring weather of the evergreen broad-leaved fore