A method for quantitatively analyzing drought tolerance of alfalfa based on root architecture parameters

Abstract

The application discloses a kind of alfalfa drought tolerance quantitative analysis methods based on root system configuration parameter, it is related to the cross technical field of agricultural science and biotechnology, the method includes: selecting different root type alfalfa germplasm, establish intelligent water gradient processing system to carry out drought stress, obtain root system in situ three-dimensional structure and high-resolution image, using improved U-Net deep learning model automatic segmentation extraction root skeleton and calculate basic morphological parameters, topological index, fractal dimension and specific root length;Synchronous endogenous hormone spectrum quantitative analysis, transcriptome sequencing and molecular marker screening;Build parameter time sequence dynamic curve, calculate comprehensive score by principal component analysis, combined with structural equation model analysis Abscisic acid on the core regulation path of root system configuration;Through field unmanned aerial vehicle remote sensing verification and establish indoor-field transfer function, finally divide drought resistance grade and provide application setting.

Description

Technical Field

[0001] This invention relates to the field of interdisciplinary technology of agricultural science and biotechnology, specifically to a quantitative analysis method for drought resistance of alfalfa based on root system configuration parameters. Background Technology

[0002] Alfalfa, as one of the world's most important high-quality legume forage crops, is widely cultivated in arid and semi-arid regions. Its yield and quality are significantly affected by drought stress. Breeding and screening drought-resistant varieties is an important way to ensure the sustainable development of the alfalfa industry, and accurate and efficient quantitative analysis methods for drought resistance are the key technical foundation for variety selection and germplasm resource identification.

[0003] Currently, the evaluation of alfalfa drought resistance mainly relies on field phenotypic observations and routine physiological index measurements, which have the following prominent problems and shortcomings: The evaluation indicators are singular and the phenotypic acquisition efficiency is low: traditional methods mostly focus on aboveground traits, and there is insufficient systematic quantification of the root system, the core organ of plants that senses and responds to drought. Existing root system measurements mostly use manual excavation, two-dimensional scanning and manual calibration, which have low throughput, high subjectivity and make it difficult to obtain detailed root system configuration features, especially lacking quantitative descriptions of topological and fractal properties such as root branching patterns and space-filling capacity.

[0004] Ignoring the temporal dynamic response of root system architecture to drought: Existing studies are mostly static comparisons at a single time point, which cannot reflect the adaptive remodeling process of root system during the continuous drought process, and lack dynamic evaluation dimensions of response rate, adaptation range and recovery potential among varieties.

[0005] The disconnect between multi-omics information and phenotype: Hormonal regulation and differential transcriptome expression have been proven to be closely related to drought resistance, but existing evaluation systems often analyze physiological and biochemical indicators separately from root morphology and structure, lacking structural equation modeling with key hormones as the core mediating variables, making it difficult to reveal causal pathways.

[0006] Insufficient coverage of variety types and poor universality of evaluation systems: Alfalfa has different root types, such as rhizomatous, taproot, and root tiller, with rich variations in root system architecture. However, most quantitative analysis methods only target single varieties or a few types, lacking variety materials that cover a wide range of root system architectures, thus limiting the applicability and representativeness of the evaluation results.

[0007] The disconnect between indoor evaluation and field productivity prediction: Many evaluation results under controlled indoor conditions are difficult to directly translate into yield performance under dryland conditions in the field. The lack of a correction mechanism for establishing transfer functions through remote sensing feedback leads to unstable performance of drought-resistant materials screened indoors when promoted in the field. Summary of the Invention

[0008] To address the aforementioned technical problems, this invention proposes a quantitative analysis method for drought resistance of alfalfa based on root system architecture parameters, comprising the following steps: Alfalfa germplasm materials with different root types were selected and seedlings were obtained after cultivation. A smart moisture gradient processing system was established using a frequency domain reflectance soil moisture sensor and an Internet of Things (IoT) communication module. At least four moisture treatment levels were set and water was automatically replenished. Sampling was performed every four weeks. Before sampling, a micro-computed tomography system was used to obtain the in-situ three-dimensional structure of the root system. After the root system was removed, it was divided into two parts. One part was frozen for subsequent analysis, and the other part was used for two-dimensional image scanning. A dual-light source high-resolution scanner was used to acquire grayscale images of the root system, and an improved U-Net deep learning model was used to automatically segment and extract the root system skeleton. Based on the root system skeleton, basic morphological parameters, topological index, fractal dimension and specific root length were calculated; frozen root samples were taken and endogenous hormone profiles were quantitatively analyzed by high performance liquid chromatography-tandem mass spectrometry and hormone ratios were calculated. Transcriptome sequencing was performed on leaves or root tips before and after drought treatment to screen differentially expressed genes and obtain a set of molecular markers. Dynamic response curves of each parameter were plotted according to the sampling time, and the response rate and recovery potential were calculated. Principal component analysis was used to calculate the comprehensive score, and a structural equation model with abscisic acid as the core mediating variable was constructed to analyze the effects between various parameters. Different grades of varieties were validated in arid and semi-arid areas, and remote sensing indicators were obtained by UAVs and regressed with the indoor comprehensive score to establish a transfer function. Based on the comprehensive score, drought resistance levels were classified and application settings were set.

[0009] Furthermore, the at least four water treatment levels are control, mild stress, moderate stress, and severe stress, respectively. The control is 70%–80% of the field maximum water holding capacity, the mild stress is 60%–65% of the field maximum water holding capacity, the moderate stress is 45%–50% of the field maximum water holding capacity, and the severe stress is 30%–35% of the field maximum water holding capacity.

[0010] Furthermore, the improved U-Net deep learning model introduces a hollow spatial pyramid pooling module and an efficient channel attention module between the encoder and decoder.

[0011] Furthermore, the topological index is used to quantify the root branching pattern type, the fractal dimension is calculated using the box-counting method and together with the fractal abundance characterizes the root space filling capacity, the specific root length is calculated by dividing the total root length by the root dry weight, and the specific root surface area is calculated by dividing the total root surface area by the root dry weight.

[0012] Furthermore, the types of hormones included in the quantitative analysis of the endogenous hormone profile include at least abscisic acid, indoleacetic acid, gibberellin, and zeatin, and the hormone ratios include the ratio of abscisic acid to indoleacetic acid, and the ratio of the sum of the three promoting hormones (indoleacetic acid, gibberellin, and zeatin) to abscisic acid.

[0013] Furthermore, the screening criteria for differentially expressed genes are that the absolute value of the fold change in expression level is greater than 2 and the false discovery rate is less than 0.05, and the molecular marker set contains at least 20 single nucleotide polymorphism sites that are significantly associated with drought resistance index or relative leaf water content.

[0014] Furthermore, the principal component analysis method extracts principal components with eigenvalues ​​greater than 1, and calculates the comprehensive score of each variety by weighting the variance contribution rate of each principal component; the structural equation model uses the maximum likelihood method to estimate parameters, and its goodness-of-fit index includes a chi-square degree of freedom ratio of less than 2, an approximate root mean square error of less than 0.08, and a goodness-of-fit index of greater than 0.9.

[0015] Furthermore, the UAV remote sensing indicators include canopy temperature, normalized vegetation index (NDI), and soil-adjusted vegetation index (SDI). The NDI is calculated by dividing the difference between near-infrared reflectance and red reflectance by the sum of the two. The transfer function is used to convert indoor evaluation results into expected field productivity.

[0016] Furthermore, after selecting germplasm materials and before conducting the full-process evaluation, a simplified initial screening step is also included: after treating the germplasm to be tested under severe stress conditions for two weeks, only three indicators are measured: root abscisic acid content, fractal dimension, and root-shoot ratio. After normalizing each indicator, the initial screening score is obtained by weighted summation according to the weights. Germplasm with high scores is selected according to the ranking of the initial screening scores to enter the subsequent full-process fine re-evaluation.

[0017] Furthermore, the application settings are as follows: for drought-resistant varieties with a comprehensive score higher than the threshold, they are directly promoted and planted in arid and semi-arid regions and used as parental materials for drought-resistant breeding; for varieties with a moderate comprehensive score but with specific superior traits, the molecular marker set is used to screen the genotypes of the offspring population; for sensitive varieties with a comprehensive score lower than the threshold, endogenous abscisic acid accumulation is induced by exogenous spraying of abscisic acid or by using a deficit irrigation strategy to promote the transformation of root architecture towards improving water absorption efficiency.

[0018] The present invention has the following outstanding beneficial effects: This invention integrates intelligent sensing irrigation, deep learning image analysis, multi-omics detection, and field remote sensing feedback technologies to achieve multidimensional and precise quantification and intelligent analysis of alfalfa root architecture. An improved U-Net model is used to automatically segment root images, achieving an intersection-over-union ratio (IoU) of over 87%, pixel accuracy exceeding 91%, and a determination coefficient higher than 0.95. In addition to conventional parameters, topological indices are introduced to distinguish branching patterns, and fractal dimension and fractal abundance are combined to characterize spatial filling ability, thus providing refined indicators for revealing the morphological mechanisms of drought resistance differences. Simultaneously, by setting four moisture gradients and extending the treatment period to three months with sampling every four weeks, time-series dynamic response curves of root parameters are constructed, allowing for the calculation of the variety's response rate and recovery potential, overcoming the limitations of static evaluation.

[0019] This invention utilizes high-performance liquid chromatography-tandem mass spectrometry (HPLC-MS / MS) to simultaneously quantify abscisic acid (ABA), indoleacetic acid (IAA), gibberellin, and zeatin. Combined with structural equation modeling, it reveals the core regulatory pathway of ABA on root branching and spatial configuration, clarifying the physiological targets for drought resistance improvement. Furthermore, the fusion of transcriptome sequencing and drought-resistant SNP markers enhances the value of breeding-assisted selection, while regression modeling based on UAV remote sensing and indoor comprehensive scores establishes a reliable indoor-field transfer function, enabling the evaluation results to effectively predict actual productivity. For large-scale early-generation breeding populations, a simplified initial screening method measuring only ABA content, fractal dimension, and root-to-shoot ratio achieves a correlation coefficient of 0.93, increasing throughput fivefold, reducing cost by 60%, and achieving 90% accuracy. The entire method covers three types of varieties: rhizomatous, taproot, and root-tillering. Principal component analysis (PCA) dimensionality reduction classifies varieties into five drought resistance levels, and provides differentiated application schemes for different levels of varieties, including direct promotion, molecular marker-assisted selection, and exogenous regulation, forming a complete technical system from refined evaluation to field validation, and from initial screening to re-evaluation. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the quantitative analysis method for drought resistance of alfalfa based on root system architecture parameters according to the present invention. Figure 2 This is a drought stress diagram of the present invention; Figure 3 This is a diagram illustrating the root system morphology of the present invention. Figure 4 It is a classification standard for soil moisture status (soil humidity). Detailed Implementation

[0021] The present invention will be described in detail below with reference to specific embodiments. The embodiments of the present invention are only used to explain the present invention and are not intended to limit the scope of protection of the present invention.

[0022] like Figure 1 As shown, a quantitative analysis method for drought tolerance of alfalfa based on root system architecture parameters includes the following steps: Step 1: Germplasm resource preparation and genetic diversity coverage.

[0023] Alfalfa varieties or germplasm materials with clear genetic backgrounds and gradient differences in drought tolerance were selected. At least three different root types were included: rhizomatous, taproot, and sucker-type varieties, to cover a wide range of root system morphological variations. Each germplasm material was sterilized and then germinated in a standardized seedling substrate until it reached the seedling stage.

[0024] Step 2: Establishment of an intelligent moisture gradient treatment system.

[0025] A frequency domain reflectance soil moisture sensor was used to monitor the volumetric water content of the cultivation substrate in real time. The sensor uploaded the data to a cloud control system via an IoT communication module. The system has a built-in adaptive calibration algorithm that automatically adjusts the measurement parameters according to different substrate types. At least four water treatment levels were set: control (70-80% of field capacity), mild stress (60-65%), moderate stress (45-50%), and severe stress (30-35%). Figure 2 As shown. The system automatically replenishes water to the target level based on sensor feedback, extending the treatment cycle to three months. Destructive sampling is performed every four weeks to obtain temporal dynamic response data of root architecture to drought stress, such as... Figure 3 As shown, Figure 3 (a) and (b) represent the interaction effect. Under 60%-65% field water holding capacity, the root length, root diameter, surface area, and volume of AH and ZD all reach their minimum values. Figure 3 (c) and (d) show the main effects; at 60%-65% field capacity, root length, root diameter, surface area, and volume all reach their minimum values. This indicates that under suitable moisture conditions, the root system has a small distribution range in the growth medium, eliminating the need for a large root system to obtain water; for example... Figure 4 The figure shows the grading standards for soil moisture status (soil humidity).

[0026] Preferably, the frequency domain reflective sensor automatically collects three readings before each water replenishment, takes the average value, and transmits the data to the gateway via Bluetooth or LoRa wireless module. The gateway then aggregates the data and uploads it to the cloud. The cloud control system issues a water replenishment command based on the preset target moisture content threshold, controls the solenoid valve to open and close, and realizes automatic irrigation.

[0027] Step 3: Obtaining the in-situ three-dimensional structure of the root system and collecting samples.

[0028] At each sampling point, a micro-computed tomography system was first used to perform a non-destructive scan of the cultivation container to obtain in-situ three-dimensional structural data of the roots. After scanning, the roots were completely removed from the substrate and gently washed with running water. The washed roots were divided into two portions: one portion was flash-frozen in liquid nitrogen and then stored in an ultra-low temperature freezer at -80°C for subsequent endogenous hormone profiling and transcriptome determination; the other portion was used for high-resolution two-dimensional image scanning.

[0029] Step 4: High-resolution root system image acquisition and intelligent segmentation.

[0030] The root system used for imaging was laid flat in a transparent tray, with a suitable layer of water added to prevent root overlap. A dual-light source high-resolution scanner was used to acquire grayscale images of the root system at a resolution of 600 dots per inch. The dual-light source system illuminates simultaneously from the top and sides, eliminating shadows and uneven reflections.

[0031] For the acquired root system images, an automatic segmentation and parameter extraction system based on an improved U-Net deep learning model was used. This improved U-Net network introduces a hollow spatial pyramid pooling module and an efficient channel attention module between the encoder and decoder to enhance the ability to capture features of fine root tips and intersecting branches. The segmented binarized image is then superimposed on the original image to obtain a clean root skeleton map. This segmentation system achieves an average intersection-union ratio (IUU) of over 87%, an average pixel accuracy of over 91%, and determination coefficients for root length and area measurements both exceed 0.95, demonstrating extremely high consistency with manually calibrated results.

[0032] Preferably, the training dataset of the improved U-Net model contains no less than one thousand root-labeled images of different varieties, different water treatments and different growth stages. The labeled content includes two types of pixels: root skeleton and background. Data augmentation techniques, including rotation, flipping, scaling and brightness adjustment, are used during training to improve the model's generalization ability.

[0033] Step 5: Calculation of multi-dimensional parameters of root system configuration.

[0034] Based on the segmented root system skeleton diagram, the following parameters are extracted and calculated using root system analysis software or a self-developed algorithm.

[0035] Basic morphological parameters include total root length, total root surface area, root volume, average root diameter, number of root tips, and number of root branches.

[0036] The topological index TI is calculated to quantify the type of root branching pattern. Its specific form is as follows: Where A is the total number of internal links, that is, the maximum number of nodes (or the total number of internal links in the longest channel) from the base of the root system to any end point; M is the total number of external links, that is, the total number of all end point root tips.

[0037] When TI=1, the root system has a typical "V" shaped branch; when TI=0.5, the root system has a typical "forked" branch; the actual root system is somewhere in between.

[0038] The root system space-filling capacity is calculated using the box-counting method, and its specific form is as follows: in, Let denoted as r, and let FD be the number of square grids intersecting the root system image. Let FD be the fractal dimension, i.e., the negative value of the slope of the regression line, and let lgK be the fractal abundance. Fractal dimension and fractal abundance together characterize the root system's spatial filling ability and distribution uniformity.

[0039] Specific root length and specific root surface area are calculated by dividing the total root length by the dry weight of the root system and the total root surface area by the dry weight of the root system, respectively.

[0040] Step six: Quantitative analysis of endogenous hormone profile.

[0041] The frozen root samples preserved in step three were used for simultaneous quantification of multiple hormones using a high-performance liquid chromatography-tandem mass spectrometry (HPLC-MS / MS) system. The hormones included at least abscisic acid (ABA), indoleacetic acid (IAA), gibberellin, and zeatin. Chromatographic separation was performed using a reversed-phase C18 column with a mobile phase consisting of methanol and ultrapure water containing trace amounts of formic acid, using gradient elution. Mass spectrometry employed an electrospray ionization source in multiple reaction monitoring (MRM) mode, with external standard method for quantification. The content and ratios of each hormone were calculated, including the ABA to IAA ratio and the ratio of the sum of the three promoting hormones (IAA, gibberellin, and zeatin) to ABA.

[0042] Preferably, the chromatographic conditions for high-performance liquid chromatography-tandem mass spectrometry (HPLC-MS / MS) are as follows: column size 2.1 mm x 50 mm, particle size 1.8 μm; mobile phase A is ultrapure water containing 0.1% formic acid, and mobile phase B is methanol containing 0.1% formic acid; gradient elution program is as follows: initial mobile phase B 5%, linearly increased to 95% in 0-2 minutes, maintained at 95% in 2-4 minutes, decreased to 5% in 4-4.1 minutes, and equilibrated in 4.1-6 minutes; flow rate 0.3 mL / min, column temperature 40 °C, injection volume 2 μL. The mass spectrometry conditions are: electrospray ionization source in negative ion mode, spray voltage -2800 volts, curtain gas pressure 15 psi, nebulizer gas pressure 65 psi, auxiliary gas pressure 70 psi, and nebulization temperature 400 °C. The parent ion of abscisic acid is 263, and the daughter ions are 219 and 153; the parent ion of indoleacetic acid is 176, and the daughter ion is 130; the parent ion of gibberellin is 345, and the daughter ion is 143; the parent ion of zeatin is 220, and the daughter ion is 136.

[0043] Step 7: Transcriptome data acquisition and molecular marker association.

[0044] Before drought treatment and after severe stress treatment, leaf or root tip samples were collected from each variety, and total RNA was extracted for transcriptome sequencing. A high-throughput sequencing system was used to obtain gene expression profiles for each variety under normal and drought conditions. The expression profile data were compared and annotated with known alfalfa multi-omics databases to screen differentially expressed genes in response to drought. Simultaneously, using published genome-wide association studies, single nucleotide polymorphisms (SNPs) significantly associated with drought-related traits such as drought resistance index, relative leaf water content, and yield were extracted and used as a molecular marker set.

[0045] Preferably, the screening criteria for differentially expressed genes in transcriptome sequencing are an absolute fold change in expression greater than 2 and a false discovery rate of less than 0.05. The obtained molecular marker set contains at least twenty single nucleotide polymorphism sites that are significantly associated with drought resistance index or relative leaf water content.

[0046] Step 8: Constructing the time-series dynamic response curve.

[0047] Dynamic curves were plotted chronologically based on root architecture parameters, hormone content, and biomass data measured at each sampling time point. The response rate, maximum response amplitude, and recovery potential of each variety during the stress process were calculated. Varieties were classified into three categories—rapid-response, gradual-adaptation, and slow-responding—based on the dynamic curve morphology, serving as supplementary indicators for evaluating drought tolerance.

[0048] Step 9: Establish a multi-dimensional comprehensive evaluation model.

[0049] Principal component analysis (PCA) was used to reduce the dimensionality of all the above parameters, and principal components with eigenvalues ​​greater than 1 were extracted. The variance contribution rate of each principal component was used as a weight to calculate the comprehensive score (D-value) for each variety. The parameters used for the comprehensive score included, but were not limited to, total root length, total surface area, fractal dimension, fractal abundance, topological index, specific root length, abscisic acid content, the ratio of promoting hormones to abscisic acid, root-shoot ratio, and the fold change in expression levels of key differentially expressed genes in the transcriptome. Based on the D-value, the germplasm population was divided into five levels: highly drought-resistant, relatively drought-resistant, moderately drought-resistant, weakly drought-resistant, and drought-sensitive.

[0050] Simultaneously, a structural equation model was constructed, using abscisic acid (ABA) content as the core mediating variable, to analyze its direct and indirect effects on root topological index, fractal dimension, root tip number, branch number, root biomass, and aboveground biomass, clarifying the hormone regulatory pathways underlying differences in drought tolerance among different varieties. The maximum likelihood method was used to estimate the parameters in the structural equation model, and goodness-of-fit indices included a chi-square degree-of-freedom ratio of less than 2, an approximate root mean square error of less than 0.08, and a goodness-of-fit index greater than 0.9.

[0051] Preferably, all parameters are standardized before principal component analysis, and the applicability of factor analysis of the data is verified by Kaiser-Meyer-Olkin test and Bartlett's test of sphericity. Principal component extraction can only be performed when the Kaiser-Meyer-Olkin value is greater than 0.6 and the Bartlett test is significant.

[0052] Step 10: Field verification and remote sensing feedback.

[0053] The drought-resistant varieties selected in step nine were subjected to field natural drought verification experiments in typical arid and semi-arid regions. Two treatments were set up in the field: normal irrigation and natural drought. During the key growth stages, portable airborne multispectral or hyperspectral UAVs were used to acquire remote sensing indicators such as canopy temperature, normalized difference vegetation index (NDVI), and soil-adjusted vegetation index (SDI). After harvest, aboveground biomass and root biomass were measured. Regression modeling was performed between the field remote sensing indicators and the indoor comprehensive score (D-value) to establish a transfer function from the indoor evaluation results to field productivity prediction, which was used to correct and optimize the weight parameters of the indoor evaluation system.

[0054] Preferably, the UAV remote sensing data acquisition time in step ten is selected between 10:00 AM and 2:00 PM on a clear, windless day, with a flight altitude of 50 meters. The multispectral camera includes five bands: blue, green, red, red-edge, and near-infrared, while the hyperspectral camera includes a continuous spectrum from 400 to 1000 nanometers. The normalized difference in vegetation index (NDDI) is calculated by dividing the difference between the near-infrared reflectance and the red reflectance by the sum of the two.

[0055] Step 11: Application based on the above evaluation results.

[0056] For drought-resistant varieties with a comprehensive score above the threshold, they should be directly planted in arid and semi-arid regions and used as parental materials for drought-resistant breeding. For varieties with a moderate comprehensive score but specific desirable traits, marker-assisted selection should be used to screen the drought-resistant single nucleotide polymorphisms obtained in step seven for genotyping of offspring populations, accelerating the breeding process of drought-resistant varieties. For sensitive varieties with a low comprehensive score, exogenous abscisic acid spraying or deficit irrigation strategies can be used to induce the accumulation of endogenous abscisic acid, promoting the transformation of root architecture towards forked branching and high fractal dimension, thereby improving water absorption efficiency.

[0057] Example 1 Multi-variety, multi-gradient comprehensive drought resistance evaluation Eight alfalfa varieties were selected, including two rhizomatous varieties, three taproot varieties, and three root sucker varieties. All varieties were sourced from the National Germplasm Bank and had clear pedigree information. Seeds were disinfected with 1% sodium hypochlorite for 30 minutes, rinsed four times with deionized water, and then sown in plastic pots containing a 5:1 mixture of potting soil and sand, with one seedling per pot. All pots were placed in an intelligent artificial climate chamber, with the temperature controlled at 25-30℃, a light intensity of 120 watts per square meter, and a photoperiod of 16 hours of light followed by 8 hours of darkness. After two months of seedling growth, when the root systems of each variety were uniformly developed, water treatment began.

[0058] A frequency domain reflectance type soil moisture sensor was inserted into each pot of substrate. The sensor connected to a gateway via Bluetooth, and the gateway uploaded the data to a cloud-based control system. The system was configured with four moisture treatment levels: 78%, 63%, 48%, and 33% of field maximum water holding capacity, with an error margin controlled within ±2%. The treatments lasted for twelve weeks, with three plants of each variety randomly selected from each treatment every four weeks for sampling, for a total of four samplings.

[0059] Before sampling, the pot was moved to a miniature CT scanning system for in-situ scanning of the root system to obtain three-dimensional voxel data. The root system was then completely removed, washed, and divided into two parts. Approximately 0.5 grams of root tip tissue was rapidly frozen in liquid nitrogen and stored at -80°C; the other part of the root system was laid flat on a transparent tray containing a 1 cm layer of water, and images were obtained using a ScanMaker i800 Plus scanner at 600 dpi.

[0060] Scanned images were input into a pre-trained improved U-Net model for automatic segmentation. The model's training set contained 1,200 manually labeled alfalfa root images, covering different varieties and water treatments. The segmentation results were manually verified, with an average intersection-over-union ratio of 87.5% and an average pixel accuracy of 92.1%. Total root length, total surface area, root volume, average diameter, number of root tips, and number of branches were extracted from the segmented skeleton image. Topological index, fractal dimension, and fractal abundance were calculated. The scanned roots were dried to constant weight in a 65℃ oven, weighed, and their specific root length and specific root surface area were calculated.

[0061] The contents of abscisic acid, indoleacetic acid, gibberellin, and zeatin in frozen root samples were determined by high-performance liquid chromatography-tandem mass spectrometry (HPLC-MS / MS). The chromatographic and mass spectrometric conditions were as described above. Results are expressed as picograms per gram of fresh weight. The ratios of each hormone were calculated.

[0062] Before and after treatment, leaf samples from each variety were collected for RNA extraction and transcriptome sequencing. After quality control, alignment, and quantification of the sequencing data, differentially expressed genes associated with drought response were screened. Simultaneously, known genome-wide association study (GWAS) molecular marker genotypes for these eight varieties were obtained from public databases, totaling 25 markers covering loci related to drought resistance index and relative leaf water content.

[0063] Plot dynamic curves for all the above parameters according to the sampling time, and calculate the average response rate and maximum change of fractal dimension for each variety from the start of treatment to week 12.

[0064] Principal component analysis was then performed. The Kaiser-Meyer-Olkin test result was 0.72, and the Bartlett's test for sphericity was significant, indicating suitability for factor analysis. Five principal components with eigenvalues ​​greater than 1 were extracted, representing a cumulative contribution rate of 82.1%. The loading matrices of each principal component showed that the first principal component was mainly determined by total root length, total surface area, fractal dimension, and fractal abundance; the second principal component was mainly determined by topological index and specific root length; the third principal component was determined by abscisic acid content and the ratio of promoting hormones to abscisic acid; the fourth principal component was determined by root-to-shoot ratio and root biomass; and the fifth principal component was determined by the expression levels of key differentially expressed genes. The comprehensive score (D-value) for each variety was calculated based on the contribution rate of each principal component. Based on the D-value ranking, the eight varieties were divided into two highly drought-resistant, three relatively highly drought-resistant, two moderately drought-resistant, and one weakly drought-resistant type. Rhizomatous varieties were all placed in the highly drought-resistant and relatively highly drought-resistant categories, taproot varieties were more dispersed, and root sucker varieties were moderately to above average.

[0065] The structural equation model used abscisic acid (ABA) as an endogenous mediator, topological index, fractal dimension, root tip number, and branching number as exogenous variables, and root biomass and aboveground biomass as the final response variables. The model fit results showed a chi-square degree of freedom ratio of 1.05, an approximate root mean square error of 0.049, and a goodness-of-fit index of 0.93, indicating a good model fit. Path coefficients showed that ABA had positive direct effects of 0.67 and 0.82 on root tip number and branching number, respectively; a positive effect of 0.65 on fractal dimension; a negative effect of -0.55 on topological index; a positive effect of 0.48 on root biomass; and a negative effect of -0.53 on aboveground biomass. All path coefficients were highly significant.

[0066] Example 2 Root architecture regulation experiment using exogenous abscisic acid The weakly drought-resistant variety with the lowest overall score in Example 1 was selected and divided into three treatment groups, with twelve pots in each group. The water treatment was uniformly set at 48% of the field maximum water holding capacity. The control group was sprayed with distilled water; the abscisic acid group was sprayed with an abscisic acid solution at a concentration of 10 μmol / L on the seventh, fourteenth, and twenty-first days after the start of drought treatment, 10 ml per pot; the inhibitor group was sprayed with fluroxypyrone at a concentration of 50 μmol / L, also sprayed at the three time points. After four weeks of treatment, root architecture parameters, hormone content, and biomass were measured according to the method in Example 1.

[0067] The results showed that compared with the control group, the abscisic acid group had a 40% higher root abscisic acid content, a 30% increase in the number of branches and a 35% increase in the number of root tips, a decrease in the topological index from 0.75 to 0.62, an increase in the fractal dimension from 1.21 to 1.36, a 22% increase in specific root length, a 28% increase in root biomass, and a 14% decrease in aboveground biomass. The inhibitor group showed the opposite trend, with the topological index increasing to 0.84, the fractal dimension decreasing to 1.14, and a decrease in root biomass. This example demonstrates that exogenously increasing abscisic acid levels can effectively induce root architecture remodeling towards branching and high complexity, verifying the regulatory pathway revealed by the structural equation model of this invention, and providing a direct technical means for the cultivation improvement of sensitive varieties.

[0068] Example 3 Field remote sensing verification and transfer function construction The remaining seedlings of the eight varieties evaluated in the greenhouse in Example 1 were transplanted to the Gansu Dingxi Dryland Agriculture Experimental Station the following spring. Normal irrigation and natural drought zones were set up in the field, with three replicate plots of each variety planted in each zone, each plot covering 20 square meters. During the flowering period, on a clear, windless day at 11:00 AM, a DJI M600 drone equipped with a RedEdge-MX multispectral camera was used at an altitude of 50 meters to acquire multispectral images of each plot. After radiometric calibration and geometric correction, the normalized difference vegetation index (NDI) and soil-adjusted vegetation index (SDI) of each plot were extracted. The aboveground biomass fresh weight and dry weight of each plot were simultaneously measured.

[0069] Linear regression analysis was performed between the normalized difference vegetation index (NDE) measured in the field and the comprehensive score (D) of each variety in Example 1. The coefficient of determination (R-squared) was 0.89, indicating a highly significant positive correlation. The coefficient of determination between the soil-adjusted vegetation index (SDE) and the D value was 0.85. Simultaneously, the coefficient of determination between the field aboveground biomass dry weight and the D value was 0.91. These results demonstrate that the indoor comprehensive score can effectively predict growth performance under drought conditions in the field. Based on the regression equation, a transfer function was established to directly convert the indoor evaluation results of subsequent new varieties into expected field productivity, significantly reducing the cost and time of large-scale field screening.

[0070] Example 4 Using the simplified initial screening method provided by this invention, a rapid drought tolerance assessment was conducted on a population containing two hundred alfalfa germplasm resources. All germplasm samples were treated for two weeks under severe stress (33% of field maximum water holding capacity), and only three indicators—root abscisic acid content, fractal dimension, and root-to-shoot ratio—were measured. After normalization, each indicator was weighted and summed using coefficients of 0.4, 0.4, and 0.2 to obtain the initial screening score. Based on the score ranking, the top twenty and bottom twenty samples were selected for a comprehensive, detailed re-evaluation using the complete method of this invention. The re-evaluation results showed a correlation coefficient of 0.93 between the initial screening score and the overall re-evaluation score (D value), indicating extremely high consistency. Eighteen of the top twenty samples in the initial screening were classified as strongly drought-resistant or relatively drought-resistant in the re-evaluation, while nineteen of the bottom twenty were classified as weakly drought-resistant or sensitive. This simplified method can increase screening throughput by more than five times and reduce costs by 60%, making it suitable for population reduction in early breeding generations.

[0071] In summary, the precise quantitative analysis method for drought resistance of alfalfa root architecture based on multi-omics and intelligent phenomics provided by this invention overcomes many bottlenecks of traditional evaluation by systematically integrating cutting-edge technologies. It provides reliable technical support for drought-resistant breeding and water management of alfalfa, possessing clear scientific basis and broad industrialization prospects. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A quantitative analysis method for drought tolerance of alfalfa based on root system architecture parameters, characterized in that, Includes the following steps: Alfalfa germplasm materials with different root types were selected and seedlings were obtained after cultivation. A smart moisture gradient processing system was established using a frequency domain reflectance soil moisture sensor and an Internet of Things (IoT) communication module. At least four moisture treatment levels were set and water was automatically replenished. Sampling was performed every four weeks. Before sampling, a micro-computed tomography system was used to obtain the in-situ three-dimensional structure of the root system. After the root system was removed, it was divided into two parts. One part was frozen for subsequent analysis, and the other part was used for two-dimensional image scanning. A dual-light source high-resolution scanner was used to acquire grayscale images of the root system, and an improved U-Net deep learning model was used to automatically segment and extract the root system skeleton. Based on the root system skeleton, basic morphological parameters, topological index, fractal dimension and specific root length were calculated; frozen root samples were taken and endogenous hormone profiles were quantitatively analyzed by high performance liquid chromatography-tandem mass spectrometry and hormone ratios were calculated. Transcriptome sequencing was performed on leaves or root tips before and after drought treatment to screen differentially expressed genes and obtain a set of molecular markers. Dynamic response curves of each parameter were plotted according to the sampling time, and the response rate and recovery potential were calculated. Principal component analysis was used to calculate the comprehensive score, and a structural equation model with abscisic acid as the core mediating variable was constructed to analyze the effects between various parameters. Different grades of varieties were field-tested in arid and semi-arid regions. Remote sensing indicators were obtained using UAVs and regressed with the indoor comprehensive score to establish a transfer function. Based on the comprehensive score, drought resistance levels were classified and application settings were implemented.

2. The method according to claim 1, characterized in that, The at least four water treatment levels are control, mild stress, moderate stress, and severe stress, respectively. The control is 70%–80% of field maximum water holding capacity, the mild stress is 60%–65% of field maximum water holding capacity, the moderate stress is 45%–50% of field maximum water holding capacity, and the severe stress is 30%–35% of field maximum water holding capacity.

3. The method according to claim 1, characterized in that, The improved U-Net deep learning model introduces a hollow spatial pyramid pooling module and an efficient channel attention module between the encoder and decoder.

4. The method according to claim 1, characterized in that, The topological index is used to quantify the root branching pattern type. The fractal dimension is calculated using the box-counting method and, together with the fractal abundance, characterizes the root space filling capacity. The specific root length is calculated by dividing the total root length by the root dry weight, and the specific root surface area is calculated by dividing the total root surface area by the root dry weight.

5. The method according to claim 1, characterized in that, The types of hormones included in the quantitative analysis of the endogenous hormone profile are at least abscisic acid, indoleacetic acid, gibberellin, and zeatin. The hormone ratios include the ratio of abscisic acid to indoleacetic acid, and the ratio of the sum of the three promoting hormones (indoleacetic acid, gibberellin, and zeatin) to abscisic acid.

6. The method according to claim 1, characterized in that, The screening criteria for differentially expressed genes are an absolute value of expression level change greater than 2 and a false discovery rate of less than 0.

05. The molecular marker set contains at least 20 single nucleotide polymorphism sites that are significantly associated with drought resistance index or relative leaf water content.

7. The method according to claim 1, characterized in that, The principal component analysis method extracts principal components with eigenvalues ​​greater than 1, and calculates the comprehensive score of each variety by weighting the variance contribution rate of each principal component; the structural equation model uses the maximum likelihood method to estimate parameters, and its goodness-of-fit index includes a chi-square degree of freedom ratio of less than 2, an approximate root mean square error of less than 0.08, and a goodness-of-fit index of greater than 0.

9.

8. The method according to claim 1, characterized in that, The UAV remote sensing indicators include canopy temperature, normalized vegetation index (NDI), and soil-adjusted vegetation index (SDI). The NDI is calculated by dividing the difference between near-infrared reflectance and red reflectance by the sum of the two. The transfer function is used to convert indoor evaluation results into expected field productivity.

9. The method according to claim 1, characterized in that, After selecting germplasm materials and before conducting a full-process evaluation, a simplified initial screening step is also included: after treating the germplasm to be tested under severe stress conditions for two weeks, only three indicators are measured: root abscisic acid content, fractal dimension, and root-shoot ratio. After normalizing each indicator, the initial screening score is obtained by weighted summation according to the weights. Germplasm with high scores is selected according to the initial screening scores to enter the subsequent full-process fine review.

10. The method according to claim 1, characterized in that, The specific application settings are as follows: for drought-resistant varieties with a comprehensive score higher than the threshold, they are directly promoted and planted in arid and semi-arid regions and used as parent materials for drought-resistant breeding; for varieties with a moderate comprehensive score but with specific superior traits, the molecular marker set is used to screen the genotypes of the offspring population; for sensitive varieties with a comprehensive score lower than the threshold, endogenous abscisic acid accumulation is induced by exogenous spraying of abscisic acid or by using a deficit irrigation strategy to promote the transformation of root architecture towards improving water absorption efficiency.

Images