Early detection and evaluation method and system for non-biological stress tolerance of new forest germplasm

Abstract

This invention discloses an early detection and evaluation method and system for the abiotic stress tolerance of new forest tree germplasm. The method includes: collecting chlorophyll fluorescence indicators, phenotypic indicators, and physiological and biochemical indicators of forest trees under abiotic stress conditions, and performing data cleaning and standardization; performing principal component analysis on the multi-source standardized dataset to screen candidate indicators; performing grey relational analysis using the comprehensive score of forest tree stress tolerance as a reference sequence and the candidate indicators as a comparison sequence; determining the core indicators and their weights based on the variance contribution rate of the principal component analysis and the correlation degree of the grey relational analysis; obtaining the abiotic stress assessment index from the core indicators using a membership function; and classifying the tolerance level of forest tree germplasm using a clustering algorithm based on this index. This invention achieves early, rapid, and accurate evaluation of the abiotic stress tolerance of forest tree germplasm, and can effectively guide forest tree stress resistance breeding work.

Description

Technical Field

[0001] This invention relates to the field of plant physiological and ecological detection technology, specifically to an early detection and evaluation method and system for the abiotic stress tolerance of new forest tree germplasm. Background Technology

[0002] Abiotic stress refers to the negative impacts on plant growth and development caused by various adverse conditions in the environment other than biotic factors. These mainly include drought, salinity, extreme temperatures, high light levels, and heavy metal pollution. Statistics show that global crop yield losses due to abiotic stress reach 65%-87% annually, seriously threatening food security and sustainable agricultural development. Traditional methods for detecting plant stress largely rely on manual observation and destructive sampling analysis, which suffer from problems such as detection lag, low efficiency, and inability to perform high-throughput screening.

[0003] With the development of science and technology, the application of spectral imaging technology, artificial intelligence, and plant phenomics has brought new opportunities for the detection of abiotic stress. These technologies enable non-destructive, rapid, and high-throughput detection of plants, thereby greatly improving detection efficiency and accuracy. Among them, chlorophyll fluorescence technology, as a non-destructive detection method, is particularly suitable for detecting abiotic stress. It can sensitively reflect the functional state of photosynthetic photosystem II (PSII). By measuring fluorescence indices such as maximum photochemical efficiency (Fv / Fm), actual photochemical efficiency (Y(II)), and non-photochemical quenching (NPQ), chlorophyll fluorescence technology can detect early damage to plant photosynthetic structures before stress symptoms appear.

[0004] Despite some advancements in existing technologies, stress assessment typically relies on single or a few fluorescence indicators, lacking a comprehensive multi-dimensional evaluation system integrating phenotypic, physiological, and biochemical indicators. Plant stress resistance is influenced by multiple factors, including genetic characteristics, environmental conditions, and physiological mechanisms, making it difficult to accurately assess stress resistance characteristics based on a single indicator. Furthermore, different abiotic stresses can lead to similar changes in fluorescence indicators, making stress type identification challenging. For example, both high-temperature and drought stress can cause a decrease in Fv / Fm and an increase in NPQ, making it difficult to distinguish stress types based solely on fluorescence indicators. In addition, fluorescence indicator responses to stress exhibit significant species specificity. Studies have shown that different plant species may display drastically different fluorescence response patterns to the same stress. For instance, under drought stress, drought-resistant varieties show a greater increase in NPQ, while drought-intolerant varieties show a greater decrease in Fv / Fm and Fv / Fo. Summary of the Invention

[0005] To overcome the aforementioned problems in the prior art, this application provides an early detection and evaluation method and system for the abiotic stress tolerance of new forest tree germplasm, which can identify new forest tree germplasm with strong abiotic stress tolerance at an early and non-destructive stage.

[0006] According to one aspect of the present invention, an early detection and evaluation method for the abiotic stress tolerance of new forest tree germplasm is provided, the method comprising: Chlorophyll fluorescence, phenotypic, and physiological and biochemical indicators of forest trees under abiotic stress conditions were collected. The collected raw data were cleaned and standardized to obtain a multi-source standardized dataset. Principal component analysis was performed on the multi-source standardized dataset to screen out candidate indicators with high explanatory power for data variation; Using the comprehensive score of forest stress tolerance as a reference sequence and the candidate indicators as a comparison sequence, a grey relational analysis was performed to calculate the correlation degree. Based on the variance contribution rate of the principal component analysis and the correlation degree of the grey relational analysis, the core indicators and their weights are determined from the candidate indicators. The standardized values ​​of each core indicator are converted into membership values ​​using a membership function, and the membership values ​​are weighted and summed according to the weights to obtain the abiotic stress assessment index. Based on the aforementioned abiotic stress assessment index, a clustering algorithm was used to classify the tolerance levels of forest tree germplasm.

[0007] According to another aspect of the present invention, an early detection and evaluation system for the abiotic stress tolerance of new forest tree germplasm is provided, comprising: The multi-dimensional indicator acquisition module is used to collect chlorophyll fluorescence indicators, phenotypic indicators and physiological and biochemical indicators of forest trees under abiotic stress conditions, and to clean and standardize the collected raw data to obtain a multi-source standardized dataset. The principal component analysis module is used to perform principal component analysis on the multi-source standardized dataset to screen out candidate indicators with high explanatory power for data variation. The grey relational analysis module is used to perform grey relational analysis and calculate the correlation degree using the comprehensive score of forest stress tolerance as a reference sequence and the candidate indicators as a comparison sequence. The core indicator screening module is used to determine the core indicators and their weights from the candidate indicators based on the variance contribution rate of the principal component analysis and the correlation between the grey relational analysis. The abiotic stress assessment index construction module is used to convert the standardized values ​​of each core indicator into membership values ​​using a membership function, and to perform a weighted summation of the membership values ​​according to the weights to obtain the abiotic stress assessment index. The tolerance level classification module is used to classify the tolerance level of forest tree germplasm based on the abiotic stress assessment index and using a clustering algorithm.

[0008] The beneficial effects of implementing the above technical solutions are as follows: 1. By integrating multiple dimensions of indicators such as chlorophyll fluorescence, morphological phenotype, and stress resistance physiology and biochemistry, an early detection and evaluation system is constructed, overcoming the one-sidedness and instability of single-indicator evaluation; 2. By combining principal component analysis and grey relational analysis to optimize the weights of core indicators, the interpretability of the model and its ability to distinguish different stress responses are enhanced; 3. By automatically classifying tolerance levels through clustering algorithms, the objectivity and efficiency of the evaluation are improved; This invention can balance the early detection and minimal / non-destructive nature, and is particularly suitable for large-scale, high-throughput early screening of new forest tree germplasm, significantly improving the accuracy and efficiency of stress resistance breeding work. Attached Figure Description

[0009] Figure 1 A schematic diagram of a method for early detection and evaluation of the abiotic stress tolerance of new forest tree germplasm provided in an embodiment of the present invention;

[0010] Figure 2 A schematic diagram of the variance contribution rate of principal component analysis provided in an embodiment of the present invention;

[0011] Figure 3 A schematic diagram illustrating the results of grey relational analysis of candidate indicators provided in an embodiment of the present invention;

[0012] Figure 4 This is a schematic diagram illustrating the fitting regression between the measured classification results and the algorithm classification results provided in an embodiment of the present invention. Detailed Implementation

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0014] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., in this application are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0015] Traditional methods for detecting abiotic stress in forest trees primarily rely on manual observation and destructive sampling, which have limitations such as detection lag, low efficiency, and inability to perform high-throughput screening. Existing technologies often rely on single or a few chlorophyll fluorescence indicators for stress diagnosis, lacking a comprehensive assessment of multi-dimensional indicators. Furthermore, different stresses can cause similar parameter changes, making identification difficult. In addition, these methods are highly species-specific, lacking universality, and making it difficult to establish a standardized, comprehensive evaluation system.

[0016] In response, this application proposes an early detection and evaluation method for the abiotic stress tolerance of new forest tree germplasm. This method achieves comprehensive, objective, and efficient early detection of the abiotic stress tolerance of new forest tree germplasm by collecting multi-dimensional indicators, screening candidate indicators, optimizing weights, and classifying tolerance levels.

[0017] Example 1 See Figure 1 This embodiment provides an early detection and evaluation method for the abiotic stress tolerance of new forest tree germplasm, specifically including the following steps: S1. Collect chlorophyll fluorescence index, phenotypic index and physiological and biochemical index of forest trees under abiotic stress conditions, and clean and standardize the collected raw data to obtain a multi-source standardized dataset.

[0018] Specifically, under abiotic stress conditions, micro-invasive multi-point sampling technology combined with a portable photosynthesis measurement system can be used to simultaneously collect data on chlorophyll fluorescence, phenotypic indicators, and physiological and biochemical parameters. The sampling time for these three parameters is locked at Δt ≤ 30 min, eliminating interference from inter-individual differences and temporal drift on data correlation. The collected raw data are cleaned and standardized using Python's Pandas library to obtain a multi-source standardized dataset.

[0019] Among them, chlorophyll fluorescence indicators are used to reflect the state of the photosynthetic system, including electron transport rate (ETR), maximum photochemical efficiency of photosystem II (Fv / Fm), actual photochemical efficiency (Y(II)), photochemical quenching coefficient (qP), regulated energy dissipation quantum efficiency (Y(NPQ)), non-regulated energy dissipation quantum efficiency (Y(NO)) and non-photochemical quenching NPQ; phenotypic indicators are morphological parameters used to visually characterize the plant's growth status, including plant height, leaf width ratio, stem-to-root ratio, and SPAD value; physiological indicators are biochemical parameters used to reflect the plant's stress resistance physiological response, including malondialdehyde (MDA) content, peroxidase content, and catalase content.

[0020] S2. Perform principal component analysis on the multi-source standardized dataset to screen out candidate indicators with high explanatory power for data variation.

[0021] Specifically, based on the multi-source standardized dataset, candidate indicators are screened and weighted using Python's Scikit-learn library and a custom algorithm; principal component analysis (PCA) is performed using sklearn.decomposition to calculate the variance contribution rate of each indicator in the multi-source standardized dataset, and candidate indicators with high explanatory power for data variation are selected.

[0022] S3. Using the comprehensive score of forest stress tolerance as the reference sequence and the candidate indicators as the comparison sequence, perform grey relational analysis to calculate the correlation degree.

[0023] Specifically, using the comprehensive score of forest stress tolerance as the reference sequence and candidate indicators as the comparison sequence, the mean value method was used to perform dimensionless processing, and the correlation coefficient and correlation degree between each comparison sequence and the reference sequence were calculated.

[0024] S4. Based on the variance contribution rate of principal component analysis and the correlation with grey relational analysis, determine the core indicators and their weights from the candidate indicators.

[0025] Specifically, candidate indicators are ranked according to their correlation degree. In this embodiment, indicators with a correlation degree greater than 0.7 are selected for inclusion in the core indicator set. By combining the variance contribution rate of principal component analysis and the correlation degree of grey relational analysis, 3-5 of the most representative core indicators are selected, and the weight value W of each core indicator is calculated. i , where ΣW i =1.

[0026] S5. Use the membership function to convert the standardized values ​​of each core indicator into membership values, and then perform a weighted summation of the membership values ​​according to the weights to obtain the abiotic stress assessment index.

[0027] Specifically, the membership degree value x of each core indicator is calculated based on the membership function. i x i ∈[0,1], construct the plant abiotic stress assessment index. The formula for calculating the plant abiotic stress assessment index is: ; Where Y represents the abiotic stress assessment index, x i W represents the membership value of the i-th core indicator. i This represents the weight value of the i-th core indicator.

[0028] S6. Based on the abiotic stress assessment index, a clustering algorithm is used to classify the tolerance level of forest tree germplasm.

[0029] Specifically, based on the abiotic stress assessment index, a clustering algorithm is used to cluster forest tree germplasm, automatically identifying three levels: highly tolerant, intermediate, and sensitive.

[0030] The method provided in this embodiment enables comprehensive, objective, and efficient early detection of the abiotic stress tolerance of new forest tree germplasm. This solves the problems of traditional methods, such as detection lag, low efficiency, and inability to perform high-throughput screening. It also overcomes the limitations of single-indicator assessment (incompleteness), difficulty in identifying stress types, and insufficient universality, providing a precise, rapid, and non-destructive evaluation tool for forest tree stress resistance breeding.

[0031] In some embodiments described above in this application, the aforementioned chlorophyll fluorescence indices are further proposed to reflect the state of the photosynthetic system, including at least electron transport rate, maximum photochemical efficiency of photosystem II, actual photochemical efficiency, photochemical quenching coefficient, regulated energy dissipation quantum efficiency, non-regulated energy dissipation quantum efficiency, and non-photochemical quenching. Through the above technical solutions, the dynamic response and functional state of the forest photosynthetic system under abiotic stress can be characterized from multiple dimensions, including energy absorption, transfer, conversion, dissipation, and photosynthesis. This comprehensive consideration of multiple indices overcomes the limitations of evaluation using only one or a few indices.

[0032] In some of the embodiments described above in this application, the above-mentioned evaluation method is further proposed, wherein the phenotypic indicators are morphological parameters used to characterize the plant growth status, including at least plant height, leaf width ratio, stem-to-root ratio, and SPAD.

[0033] These indicators cover key dimensions of plant growth. Plant height directly indicates overall development, leaf width ratio describes leaf morphology and is related to photosynthetic efficiency, stem-to-root ratio reflects resource allocation strategies and stress adaptation mechanisms, while SPAD value indirectly characterizes photosynthetic potential.

[0034] In some embodiments described above in this application, the aforementioned evaluation method is further proposed, wherein physiological and biochemical indicators are used to reflect the biochemical parameters of the plant's physiological response to stress, including at least malondialdehyde (MDA) content, peroxidase content, and catalase content. The introduction of these specific biochemical parameters allows the evaluation of the abiotic stress tolerance of new forest tree germplasm to move beyond a macroscopic level and delve into the cellular and molecular levels, providing a more scientific and operable quantitative basis, thereby enhancing the accuracy and reliability of the evaluation results.

[0035] In some embodiments described above in this application, a further step for determining core indicators is proposed, including: calculating the variance contribution rate of each indicator in the multi-source standardized dataset based on principal component analysis, selecting indicators whose variance contribution rate exceeds a preset threshold as candidate indicators, and using the candidate indicators as a comparison sequence; using the comprehensive score of forest stress tolerance as a reference sequence, calculating the correlation degree between the comparison sequence and the reference sequence based on a grey relational analysis algorithm; calculating the comprehensive score of each candidate indicator based on the variance contribution rate and the grey relational degree, and sorting them according to the comprehensive score to select core indicators that meet preset conditions.

[0036] This embodiment calculates the comprehensive score of each candidate indicator and sorts and screens them accordingly. This ensures that the determined core indicators not only have high statistical explanatory power for data variation, but also have a strong biological correlation with the abiotic stress tolerance of forest trees. This improves the accuracy and reliability of core indicator screening, avoids indicator redundancy or omission, and enables the subsequent construction of the abiotic stress assessment index to more accurately reflect the true tolerance level of forest trees.

[0037] In some embodiments described above in this application, the aforementioned evaluation method is further proposed. The membership function can be either an S-shaped membership function or a linear membership function. Specifically, for indicators that have a non-linear relationship with forest tree tolerance or exhibit gradual changes around a specific threshold, an S-shaped membership function can be used for smooth transformation, avoiding abrupt changes in membership or information distortion caused by inappropriate function selection. Meanwhile, for indicators that have an approximately linear relationship with forest tree tolerance, a linear membership function can be used for efficient transformation, simplifying the calculation process and ensuring the intuitiveness and accuracy of the transformation. This flexible selection mechanism allows the evaluation model to be optimized for adaptability based on the inherent characteristics of each core indicator, thereby significantly enhancing the construction accuracy of the abiotic stress assessment index and the model's scenario adaptability, ensuring the reliability and universality of early detection and evaluation results of the abiotic stress tolerance of new forest tree germplasm.

[0038] In some of the schemes described above in this application, a formula for calculating the abiotic stress assessment index is further proposed as follows: ; Where Y represents the abiotic stress assessment index, x i W represents the membership value of the i-th core indicator. i This represents the weight value of the i-th core indicator. This calculation formula provides a precise and standardized quantitative tool for the early detection and evaluation of the abiotic stress tolerance of new forest tree germplasm, significantly enhancing the reliability, operability, and objectivity of the assessment model.

[0039] In some of the solutions described above in this application, the proposed method further includes: validating the abiotic stress assessment index algorithm using K-fold cross-validation, employing the coefficient of determination R² and root mean square error (RMSE) as evaluation indicators to quantify the algorithm's goodness of fit and prediction accuracy, and optimizing the weighting rules for core indicators and the threshold for dividing numerical intervals based on the validation results. By introducing a systematic validation and optimization mechanism, the potential shortcomings of the abiotic stress assessment index algorithm in terms of goodness of fit and prediction accuracy are addressed, thereby ensuring the reliability and adaptability of the assessment index.

[0040] In some of the solutions described above in this application, it is further proposed that the tolerance level in the aforementioned evaluation method can be divided into three levels: highly tolerant, intermediate, and sensitive. These clear classification labels provide breeders with intuitive judgment criteria, enabling them to quickly identify and distinguish forest tree germplasm with different stress resistance characteristics, thereby guiding the selection and utilization of breeding materials.

[0041] Preferably, the method provided in this embodiment further includes: using the variance contribution rate of principal component analysis, grey relational analysis algorithm, weight calculation, and biological stress assessment index calculation as an assessment model to conduct early detection and assessment of the abiotic stress tolerance of new forest tree germplasm. A 10-fold cross-validation method is used to validate the assessment model, with the coefficient of determination R² and root mean square error (RMSE) as core evaluation indicators to quantify the model's fit and prediction accuracy. The assessment index values ​​of the samples are correlated with the quantitative values ​​of tolerance levels identified by experts to construct a validation dataset. The validation dataset is randomly divided into 10 groups, with 9 groups used as the training set and 1 group as the test set, repeated 10 times. After each training iteration, R² and RMSE are calculated. The average of the 10 validation results is calculated, yielding an average R² = 0.85 and an average RMSE = 0.27, both meeting the accuracy requirements of R² ≥ 0.8 and RMSE ≤ 0.3, indicating that the fit and prediction accuracy meet the standards. If the verification results do not meet the standard, i.e. R² < 0.8 or RMSE > 0.3, the weight assignment rules of the core indicators and the threshold for dividing the tolerance value range are adjusted iteratively, the evaluation index value is recalculated and verified, until the accuracy of the evaluation model meets the standard.

[0042] Example 2 This embodiment provides an early detection and evaluation system for the tolerance of new forest tree germplasm to abiotic stresses, including a multi-dimensional index acquisition module, a principal component analysis module, a grey relational analysis module, a core index screening module, an abiotic stress assessment index construction module, and a tolerance level classification module.

[0043] A multi-dimensional indicator acquisition module is used to collect chlorophyll fluorescence, phenotypic, and physiological and biochemical indicators of forest trees under abiotic stress conditions, and to clean and standardize the collected raw data. This module uses a chlorophyll fluorometer for non-invasive in-situ measurements, collecting indicators such as electron transport rate, maximum photochemical efficiency of photosystem II, and actual photochemical efficiency; it measures plant height and leaf shape parameters using digital calipers, and determines chlorophyll content using a SPAD-502 chlorophyll meter; and it employs minimally invasive multi-point sampling technology to collect physiological and biochemical indicators, including malondialdehyde content, peroxidase activity, and catalase activity. Data cleaning uses the interquartile range method to remove outliers, and missing values ​​are imputed using the mean. Standardization uses the Z-score method to eliminate dimensional differences between different indicators, resulting in a multi-source standardized dataset.

[0044] The principal component analysis module performs principal component analysis on multi-source standardized datasets to screen out candidate indicators with high explanatory power for data variation.

[0045] The grey relational analysis module uses the comprehensive score of forest stress tolerance as the reference sequence and candidate indicators as the comparison sequence to perform grey relational analysis and calculate the correlation degree. This module uses the grey relational analysis algorithm to calculate the correlation degree between each candidate indicator and the reference sequence. The correlation degree ranges from [0,1], and the closer the correlation degree is to 1, the stronger the correlation between the indicator and the forest stress tolerance.

[0046] The core indicator selection module determines the core indicators and their weights from the candidate indicators based on the variance contribution rate of principal component analysis (PCA) and the correlation degree of grey relational analysis. By combining the results of PCA variance contribution rate and grey relational analysis, the comprehensive score of each candidate indicator is calculated. Based on the comprehensive score ranking, the top three indicators with the highest comprehensive scores are selected as core indicators.

[0047] The abiotic stress assessment index construction module uses a membership function to convert the standardized values ​​of each core indicator into membership values, and then performs a weighted summation of the membership values ​​according to their weights to obtain the abiotic stress assessment index. This module calculates the membership values ​​of the core indicators based on an S-shaped membership function. For indicators positively correlated with tolerance, the membership value is directly calculated; a higher membership value indicates that the indicator better reflects the tolerance of the forest trees. Based on the membership values ​​and weights of each core indicator, the abiotic stress assessment index is constructed.

[0048] The tolerance level classification module, based on the abiotic stress assessment index, uses a clustering algorithm to classify the tolerance levels of forest tree germplasm. This module employs the K-means clustering algorithm to divide the tolerance capacity of forest trees into three levels, corresponding to three standardized tolerance capacity numerical ranges.

[0049] In a preferred embodiment, the variance contribution rate of principal component analysis, grey relational analysis algorithm, weight calculation, and biotic stress assessment index calculation are used as the evaluation model to conduct early detection and assessment of the abiotic stress tolerance of new forest tree germplasm. A 10-fold cross-validation method is employed to validate the evaluation model, using the coefficient of determination (R²) and root mean square error (RMSE) as core evaluation indicators to quantify the model's fit and prediction accuracy. The evaluation index values ​​of the samples are correlated with the quantitative values ​​of tolerance levels identified by experts to construct a validation dataset. This dataset is randomly divided into 10 groups, with 9 groups serving as the training set and 1 group as the test set, repeated 10 times. After each training iteration, R² and RMSE are calculated. The average of the 10 validation results is calculated, yielding an average R² = 0.85 and an average RMSE = 0.27, both meeting the accuracy requirements of R² ≥ 0.8 and RMSE ≤ 0.3, indicating that the fit and prediction accuracy meet the standards. If the verification results do not meet the standard, i.e. R² < 0.8 or RMSE > 0.3, the weight assignment rules of the core indicators and the threshold for dividing the tolerance value range are adjusted iteratively, the evaluation index value is recalculated and verified, until the accuracy of the evaluation model meets the standard.

[0050] In another preferred embodiment, 20 randomly selected forest materials were used as test materials. After standardized seedling cultivation, each progeny material was divided into experimental units with 5 cuttings, and 5 biological replicates were set up for each progeny material. During the high light period from June to August, a dual-scenario planting experiment was carried out simultaneously in a laboratory controlled environment and a natural environment. After 20 days of high light stress treatment, the test materials underwent systematic calculations and analyses, including standardized sample pretreatment, multi-dimensional index measurement, correlation analysis, principal component analysis dimensionality reduction, and grey relational analysis. At the same time, the survival rate and top-view leaf area of ​​the materials in the experimental field were measured and recorded to verify the accuracy of the assessment and the scientificity, applicability, and cross-scenario stability of the entire assessment system. Field verification results showed that the survival rate of the low high light tolerance material was 83%, while the survival rate of the high high light tolerance material reached 100%. The top-view leaf area of ​​the high high light tolerance material was 2.62 times that of the low high light tolerance material. The field measurement results were highly consistent with the assessment prediction results.

[0051] Example 3 This embodiment uses *Salix matsudana* as an example to illustrate the above technical solution in more detail: S1. Collect chlorophyll fluorescence index, phenotypic index and physiological and biochemical index of forest trees under abiotic stress conditions, and clean and standardize the collected raw data to obtain a multi-source standardized dataset.

[0052] S101. Select 10cm diameter, disease-free *Salix matsudana* cuttings of the 9901 × F1 generation (from the Yanjiang River) with uniform diameter. Each cutting retains 2-3 buds, with 3 cuttings per group. Each generation is replicated 3 times. The cuttings are placed in 100ml disposable plastic cups filled with clean water and cultured in an artificial climate chamber for 15 days. Culture conditions are: temperature 25±1℃, light intensity 70μmol·m⁻²·s⁻¹, photoperiod 14h / 10h (light / dark). Water is replenished daily to maintain a culture volume of 50ml of clean water. After the acclimatization period, the seedlings are transferred to a high-light stress treatment. The high-light stress conditions are set as follows: light intensity 160μmol·m⁻²·s⁻¹, with all other culture conditions remaining the same as the acclimatization period. Twenty days after the high-light stress treatment, chlorophyll fluorescence, phenotypic, and physiological and biochemical indicators of the *Salix matsudana* seedlings are collected simultaneously.

[0053] Chlorophyll fluorescence index was measured non-invasively in situ using a chlorophyll fluorometer (MINI-PAM) during the light-stable period from 10:00 to 16:00. Fully expanded leaves of *Salix matsudana* were selected. After 30 minutes of dark adaptation, 3-5 fully expanded leaves from the same location were selected from each plant. Measurements were taken at three sites on each leaf: the tip, middle, and base. The time interval between data collection points for each progeny group did not exceed 15 minutes, and the average value was taken as the data for that plant. The measured parameters included electron transport rate (ETR), maximum photochemical efficiency (Fv / Fm) of photosystem II, actual photochemical efficiency (Y(II)), photochemical quenching coefficient (qP), regulated energy dissipation quantum efficiency (Y(NPQ)), non-regulated energy dissipation quantum efficiency (Y(NO)), and non-photochemical quenching NPQ. This technique requires no leaf harvesting, causes zero damage to the plant, and allows continuous tracking of the dynamic changes of the same leaf, provided the instrument operating procedures are strictly followed.

[0054] Immediately after chlorophyll fluorescence measurement of individual plants, phenotypic indicators were collected using minimally invasive / non-destructive sampling methods, with sampling times ranging from T0+5 min to T0+10 min. In vivo morphological indicators were collected using non-destructive sampling methods, with sampling times ranging from T0+5 min to T0+10 min. Plant height was measured using digital vernier calipers with an accuracy of 0.1 cm, and the vertical distance from the bud to the terminal bud of the cutting was measured, with three measurements taken per plant and the arithmetic mean taken. Leaf shape parameters were collected using vernier calipers with an accuracy of 0.01 cm, measuring the maximum leaf length L and maximum leaf width W, and calculating the leaf width ratio, W / L. Three mature, fully expanded leaves were selected from each plant, prioritizing leaves from the middle of lateral branches, avoiding the main stem growth point, for minimally invasive in vitro sampling. The area of ​​the leaf notch was controlled to be within 5% of the single leaf area to minimize the impact on plant growth. Chlorophyll content was measured in vivo non-destructively using a SPAD-502 chlorophyll meter. Three fully expanded leaves were selected from each plant, and three sampling sites were selected along both sides of the midrib of each leaf, avoiding the veins. The measurements were repeated five times, and the average value was taken. Biomass allocation indicators were measured minimally invasively, with sampling time from T0+10 min to T0+15 min.

[0055] Stratified random sampling was used when chlorophyll fluorescence measurements were completed for each group. One offspring from each group was selected from three grades: good, average, and poor growth, for a total of three plants. The aboveground parts (stems) and underground parts (roots) were completely separated, blanched at 105℃ for 30 min, and then dried at 65℃ to constant weight. The plants were weighed using an electronic analytical balance with an accuracy of 0.0001 g, and the stem-to-root ratio (dry weight of aboveground parts / dry weight of underground parts) was calculated.

[0056] Physiological and biochemical indicators were collected using a minimally invasive multi-point sampling technique, with sampling times ranging from T0+10 min to T0+15 min. Three to five fully expanded leaves were selected from each plant. Using a 6mm diameter punch, leaf discs were collected from the middle of the leaf, avoiding the midrib area. One to two discs (approximately 20-30 mg) were collected from each leaf, with the sampling volume controlled to within 5% of the leaf area to ensure the plant did not die due to sampling. The samples collected from multiple points were mixed and divided into three biological replicates, frozen in liquid nitrogen, and stored at -80°C for subsequent measurements. Malondialdehyde (MDA) content, peroxidase (POD) activity, and catalase (CAT) activity were determined spectrophotometrically, following the procedures outlined in *Plant Physiology Experimental Guide* (4th edition). Three blank controls were included in the measurements to eliminate systematic errors and ensure data reliability.

[0057] S102. After 20 days of high light stress treatment, chlorophyll fluorescence index, phenotypic index and physiological and biochemical index were collected. Each group of offspring was used as a sample, and a total of 80 samples were collected. Each sample was measured 9 times, and the average value was taken as the raw data. Python version 3.9 and Pandas version 1.5.3 were used to clean the collected raw data and process it using the Z-score standardization method.

[0058] Using Python's Pandas library, outliers for each indicator were removed using the interquartile range (IQR). Data exceeding the range Q1-1.5IQR to Q3+1.5IQR were considered outliers. For missing values ​​(missing rate ≤ 5%), the average of all valid data for that indicator was used for imputation to ensure dataset integrity. The cleaned raw data was then standardized using Python's Pandas library to eliminate dimensional differences between indicators. The standardization formula is as follows: ; Where, x std This represents the standardized data, where x represents the original data. mean x represents the average of all the raw data for this indicator. std This represents the standard deviation of all the original data for the indicator; after standardization, a standardized dataset is obtained, which is used for subsequent indicator selection and weight calculation.

[0059] S2. Perform principal component analysis on the multi-source standardized dataset to screen out candidate indicators with high explanatory power for data variation.

[0060] Based on the multi-source standardized dataset obtained in step S1, using Python 3.9 and Scikit-learn 1.2.2, combined with a custom algorithm, we completed the selection of core indicators and weight assignment. The specific operations are as follows: Principal component analysis (PCA) was performed on the multi-source standardized dataset using the sklearn.decomposition package in Python. With five principal components, the variance contribution rate of each indicator was calculated, and indicators with a variance contribution rate ≥8% were selected as candidate indicators. A higher variance contribution rate indicates a stronger explanatory power for data variation. PCA analysis identified five candidate indicators: ETR, Fv / Fm, leaf width ratio, SPAD value, and POD activity, with variance contribution rates of 12.3%, 9.8%, 10.5%, 8.7%, and 8.2%, respectively, totaling 49.5%. These indicators effectively reflect the core variation information of the dataset. (See reference...) Figure 2 .

[0061] S3. Using the comprehensive score of forest stress tolerance as a reference sequence and the candidate indicators as a comparison sequence, perform grey relational analysis to calculate the correlation degree.

[0062] Using the light tolerance of *Salix matsudana* as the reference sequence (high light tolerance = 2, medium light tolerance = 1, low light tolerance = 0), and the previously selected five candidate indicators as the comparison sequence, a grey relational analysis algorithm was written in Python. The reference sequence is X0 = {x0(1), x0(2), ..., x0(n)}, and the comparison sequence is X... i = {x i (1), x i (2), ..., x i (n)}, calculate the correlation degree and correlation coefficient between each candidate index and the reference sequence.

[0063] The formula for calculating the correlation coefficient is: ; in, Let xi represent the correlation coefficient of the i-th comparison sequence at point k, and x0(k) represent the value of the reference sequence (parent sequence) at point k. i (k) represents the value of the i-th comparison sequence (subsequence) at point k. This represents the minimum difference between two levels (the minimum value among all differences). ρ represents the maximum difference between the two levels (the maximum value among all differences), and ρ represents the resolution coefficient, which is usually taken as ρ∈[0,1], and generally ρ=0.5.

[0064] correlation The calculation formula is: ; The correlation coefficient ranges from [0,1], with a correlation coefficient closer to 1 indicating a stronger correlation between the indicator and the high light tolerance of *Salix matsudana*. The calculated correlation coefficients for the five candidate indicators are: ETR=0.89, Fv / Fm=0.76, leaf width ratio=0.85, SPAD value=0.72, and POD activity=0.87. (See reference...) Figure 3 A.

[0065] S4. Based on the variance contribution rate of the principal component analysis and the correlation degree of the grey relational analysis, determine the core indicators and their weights from the candidate indicators.

[0066] Based on the variance contribution rate from principal component analysis and the results of grey relational analysis, the comprehensive score of each candidate indicator was calculated: Comprehensive Score = Variance Contribution Rate × 40% + Relational Degree × 60%. According to the ranking of comprehensive scores, the top three indicators were selected as core indicators: ETR (0.86), leaf width ratio (0.88), and POD activity (0.87), covering chlorophyll fluorescence, phenotype, and physiological dimensions to ensure the representativeness and comprehensiveness of the core indicators. (See reference...) Figure 3 B.

[0067] The weight values ​​W of each core indicator are calculated using the normalization method. i Ensure that the sum of the weights of all core indicators is 1. The calculation formula is as follows: ; Among them, S i Let be the comprehensive score of the i-th core indicator, and n be the number of core indicators (n=3). After calculation, the weight values ​​of the three core indicators are: W1 (ETR) = 0.42, W2 (leaf width ratio) = 0.35, and W3 (POD activity) = 0.23, and W1+W2+W3=1, which meets the weight assignment requirements.

[0068] S5. Use the membership function to convert the standardized values ​​of each core indicator into membership values, and then perform a weighted summation of the membership values ​​according to the weights to obtain the abiotic stress assessment index.

[0069] Based on the sigmoid membership function, a function was written in Python to calculate the membership values ​​x of the three selected core indicators. i x i The S-type membership function formula is as follows: ∈[0,1] ; Where k represents the adjustment coefficient, k=2 in this embodiment, x represents the standardized value of the core indicator, and x0 represents the average value of the standardized value of the core indicator; for indicators positively correlated with high light tolerance, namely ETR, leaf width ratio, and POD activity, the membership value is directly calculated using the above formula. The higher the membership value, the better the indicator reflects the high light tolerance of *Salix matsudana*. (See [reference]) Figure 3 C.

[0070] Based on the membership values ​​x of each core indicator i and weight value W i The High Light Tolerance Index (HLTI) of Salix matsudana was constructed, and the calculation formula is as follows: ; Where, x i W represents the membership value of the i-th core indicator. iLet x be the weight value of the i-th core indicator; and let x be the weight value of the 80 samples. i and W i Substituting into the formula, the HLTI value for each sample is calculated, with the HLTI value ranging from 0.03 to 4.12.

[0071] S6. Based on the aforementioned abiotic stress assessment index, a clustering algorithm is used to classify the tolerance levels of forest tree germplasm.

[0072] Based on the distribution characteristics of the HLTI values ​​of 80 *Salix matsudana* samples obtained from calculation, the K-means clustering algorithm was executed using the KMeans module in the Scikit-learn library of Python, with n_clusters=3, to divide the high light tolerance of *Salix matsudana* into three levels, and three standardized high light tolerance value intervals were established accordingly. The specific division results are shown in Table 1.

[0073] (1) Strong light-resistant type HLTI≥4.328, corresponding to HLTI values ​​of 3.500~4.120, a total of 30 samples. This type of sample has strong photosynthetic system stability, significant stress physiological response, and good growth status; (2) Medium-high light tolerant type 0.513≤HLTI<3.566, corresponding to HLTI values ​​ranging from 0.513 to 3.560, a total of 15 samples. The photosynthetic system of this type of sample is basically stable, the physiological response to stress is moderate, and the growth status is normal. (3) The HLTI of the weak high light tolerance type is <0.413. The corresponding sample HLTI values ​​range from -1.897 to 0.413, with a total of 35 samples. The photosynthetic system of this type of sample is significantly damaged, the physiological response to stress is weak, and the growth status is poor.

[0074] S7. The 10-fold cross-validation method was adopted, and the HLTI evaluation model was validated using the cross_val_score module in the Scikit-learn library of Python. The coefficient of determination R² and root mean square error RMSE were used as the core evaluation indicators to quantify the model's fit and prediction accuracy.

[0075] The formula for calculating the coefficient of determination is: .

[0076] The formula for calculating the root mean square error is: .

[0077] The specific verification process is as follows: A validation dataset was constructed by correlating the HLTI values ​​of 80 samples with the expert-assessed gloss tolerance level. This dataset was randomly divided into 10 groups: 9 groups for training and 1 group for testing, repeated 10 times. After each training iteration, the model's R² and RMSE were calculated. The average of the 10 validation results was calculated, yielding an average R² of 0.85 and an average RMSE of 0.27, both meeting the accuracy requirements of R² ≥ 0.8 and RMSE ≤ 0.3. This indicates that the HLTI evaluation model's fit and prediction accuracy are satisfactory and require no further optimization. If the validation results do not meet the requirements (R² < 0.8 or RMSE > 0.3), the weighting rules for the core indicators are iteratively adjusted. This involves adjusting the weighting percentages of the PCA variance contribution rate and gray relational degree, as well as the threshold for dividing the gloss tolerance value range. The HLTI values ​​are then recalculated and validated until the evaluation model's accuracy meets the requirements. (See [reference needed]). Figure 4 .

[0078] Twenty randomly selected *Salix matsudana* materials were used as test materials. After standardized seedling cultivation in the laboratory, each progeny material was divided into experimental units with five cuttings, and five biological replicates were set up for each progeny material. During the high light period from June to August, a dual-scene planting experiment was carried out simultaneously in a controlled laboratory environment and a natural environment experimental field in Qidong City, Jiangsu Province. After subjecting the test materials to 20 days of high light stress treatment, the systematic calculations and analyses, including sample standardization pretreatment, multi-dimensional index determination, correlation analysis, principal component analysis dimensionality reduction, and grey relational analysis, were strictly performed according to the method described in this invention. At the same time, the survival rate and top-view leaf area of ​​the materials in the Qidong experimental field were measured and recorded to verify the classification accuracy of the evaluation model of this invention, as well as the scientificity, applicability, and cross-scene stability of the entire evaluation system.

[0079] Field verification results showed that the survival rate of the low light-tolerant material was 83%, while the survival rate of the high light-tolerant material was 100%. The top-view leaf area of ​​the high light-tolerant material was 2.62 times that of the low light-tolerant material. The field measurement results were highly consistent with the prediction results of the evaluation model of this invention.

[0080] Figure 2 The variance contribution rate results of the principal component analysis provided in this embodiment are shown. PCA analysis was performed on the standardized dataset to calculate the variance contribution rate of the five candidate indicators. The red dashed line represents the screening threshold (8%), and all indicators meet the threshold requirement. ETR had the highest variance contribution rate (12.3%), followed by leaf width ratio (10.5%), Fv / Fm (9.8%), SPAD value (8.7%), and POD activity (8.2%), with a cumulative variance contribution rate of 49.5%.

[0081] Figure 3A represents the grey relational analysis results of the candidate indicators provided in this embodiment. Using the comprehensive score of *Salix matsudana*'s high light tolerance as a reference sequence, the grey relational degree of five candidate indicators (ETR, Fv / Fm, leaf width ratio, SPAD value, and POD activity) was calculated. The red dashed line represents the screening threshold (0.7), and all indicators meet the threshold requirement. ETR had the highest correlation (0.89), followed by POD activity (0.87), leaf width ratio (0.85), Fv / Fm (0.76), and SPAD value (0.72).

[0082] Figure 3 B represents the results of the comprehensive score calculation and screening of the core indicators provided in this embodiment. The comprehensive score of each indicator is calculated by combining the PCA variance contribution rate (weight 40%) and the grey relational degree (weight 60%). The top 3 core indicators with the highest comprehensive scores are marked with red pentagrams: ETR (highest comprehensive score), leaf width ratio, and POD activity. These three indicators represent the three dimensions of chlorophyll fluorescence, phenotype, and physiology, respectively, and will be used in the subsequent construction of the HLTI index.

[0083] Figure 3 C is a schematic diagram of the S-shaped membership function curve provided in this embodiment, and the formula is: μ(x) = 1 / (1+e^(-k(x-x0))); Where k is the adjustment coefficient (k=2 in this embodiment), and x0 is the average value of the standardized index (x0=0). When the standardized index value x=x0, the membership degree μ(x)=0.5. This function is used to convert the standardized core index value into a membership degree value in the interval [0,1], and then calculate the HLTI index.

[0084] Figure 4 This is a schematic diagram showing the regression fitting between the measured classification results and the model classification results.

[0085] Table 1: Classification results of high light resistance of 80 groups of *Salix matsudana* materials

Claims

1. A method for early detection and evaluation of the abiotic stress tolerance of new forest tree germplasm, characterized in that, The method includes: Chlorophyll fluorescence, phenotypic, and physiological and biochemical indicators of forest trees under abiotic stress conditions were collected. The collected raw data were cleaned and standardized to obtain a multi-source standardized dataset. Principal component analysis was performed on the multi-source standardized dataset to screen out candidate indicators with high explanatory power for data variation; Using the comprehensive score of forest stress tolerance as a reference sequence and the candidate indicators as a comparison sequence, a grey relational analysis was performed to calculate the correlation degree. Based on the variance contribution rate of the principal component analysis and the correlation degree of the grey relational analysis, the core indicators and their weights are determined from the candidate indicators. The standardized values ​​of each core indicator are converted into membership values ​​using a membership function, and the membership values ​​are weighted and summed according to the weights to obtain the abiotic stress assessment index. Based on the aforementioned abiotic stress assessment index, a clustering algorithm was used to classify the tolerance levels of forest tree germplasm.

2. The evaluation method according to claim 1, characterized in that, The chlorophyll fluorescence index is used to reflect the state of the photosynthetic system and includes at least the electron transport rate, maximum photochemical efficiency of photosystem II, actual photochemical efficiency, photochemical quenching coefficient, regulated energy dissipation quantum efficiency, unregulated energy dissipation quantum efficiency, and non-photochemical quenching.

3. The evaluation method according to claim 1, characterized in that, The phenotypic indicators are morphological parameters used to characterize the growth status of plants, including at least plant height, leaf width ratio, stem-to-root ratio, and SPAD.

4. The evaluation method according to claim 1, characterized in that, The physiological and biochemical indicators mentioned above are biochemical parameters used to reflect the plant's physiological response to stress, including at least malondialdehyde content, peroxidase content, and catalase content.

5. The evaluation method according to claim 1, characterized in that, The steps for determining the core indicators include: calculating the variance contribution rate of each indicator in the multi-source standardized dataset based on principal component analysis, selecting indicators whose variance contribution rate exceeds a preset threshold as candidate indicators, and using the candidate indicators as a comparison sequence. Using the comprehensive score of forest stress tolerance as a reference sequence, the correlation between the comparison sequence and the reference sequence is calculated based on the grey relational analysis algorithm; Based on the results of variance contribution rate and grey relational degree, the comprehensive score of each candidate indicator is calculated, and the indicators are sorted according to the comprehensive score to select the core indicators that meet the preset conditions.

6. The evaluation method according to claim 1, characterized in that, The membership function can be an sigmoid membership function or a linear membership function.

7. The evaluation method according to claim 1, characterized in that, The formula for calculating the abiotic stress assessment index is as follows: ； Where Y represents the abiotic stress assessment index, x i W represents the membership value of the i-th core indicator. i This represents the weight value of the i-th core indicator.

8. The evaluation method according to claim 7, characterized in that, The method further includes: validating the non-biological stress assessment index algorithm using the K-fold cross-validation method, using the coefficient of determination R² and root mean square error RMSE as evaluation indicators to quantify the algorithm's fit and prediction accuracy, and optimizing the core indicator weight assignment rules and numerical interval division thresholds based on the validation results.

9. The evaluation method according to claim 1, characterized in that, The tolerance levels can be divided into three levels: high tolerance, intermediate tolerance, and sensitive tolerance.

10. An early detection and evaluation system for the abiotic stress tolerance of new forest tree germplasm, characterized in that, include: The multi-dimensional indicator acquisition module is used to collect chlorophyll fluorescence indicators, phenotypic indicators and physiological and biochemical indicators of forest trees under abiotic stress conditions, and to clean and standardize the collected raw data to obtain a multi-source standardized dataset. The principal component analysis module is used to perform principal component analysis on the multi-source standardized dataset to screen out candidate indicators with high explanatory power for data variation. The grey relational analysis module is used to perform grey relational analysis and calculate the correlation degree using the comprehensive score of forest stress tolerance as a reference sequence and the candidate indicators as a comparison sequence. The core indicator screening module is used to determine the core indicators and their weights from the candidate indicators based on the variance contribution rate of the principal component analysis and the correlation between the grey relational analysis. The abiotic stress assessment index construction module is used to convert the standardized values ​​of each core indicator into membership values ​​using a membership function, and to perform a weighted summation of the membership values ​​according to the weights to obtain the abiotic stress assessment index. The tolerance level classification module is used to classify the tolerance level of forest tree germplasm based on the abiotic stress assessment index and using a clustering algorithm.

Images