Method for evaluating environmental adaptability of reed germplasm resources

Abstract

The present application relates to the technical field of plant germplasm resource evaluation, and discloses a reed germplasm resource environmental adaptability evaluation method. The method comprises the following steps: obtaining environmental data and reed phenotype data of multiple ecological test points, and matching to form a germplasm environmental response record; dividing the record according to a key reed phenophase, and recombining the record into multiple phenophase response sub-records; for each sub-record, extracting extreme values and duration of environmental factors that exceed a preset tolerance range, and constructing an environmental stress image; connecting all the images of the same reed germplasm at all test points in the order of phenophases to form a cross-regional whole growth period environmental adaptability path diagram; inputting the path diagram into a pre-trained multi-dimensional adaptability discrimination network for analysis, and outputting an adaptability grade classification label to complete the evaluation. The present application can more finely quantify environmental stress at a specific growth stage, and systematically discriminate the overall adaptability of germplasm to a complex adversity sequence, thereby improving the accuracy and practicality of the evaluation.

Description

Technical Field

[0001] This invention relates to the field of plant germplasm resource evaluation technology, specifically a method for evaluating the environmental adaptability of Phragmites australis germplasm resources. Background Technology

[0002] Existing assessments of the environmental adaptability of Reed sedge germplasm typically rely on multi-year cultivation trials conducted at single or a few locations. These trials analyze the correlation between average climatic factors throughout the growing season and comprehensive phenotypic traits such as final yield and plant height. This approach treats the environment as a homogeneous or static background, and evaluation indicators are often cumulative or average values ​​over the growth period. Another common practice is to set up test sites in different ecological zones, independently evaluate the performance of germplasm at each site, and finally conduct statistical analysis or simple comparisons of data from multiple sites to determine its suitable planting area.

[0003] These conventional technical solutions have significant shortcomings. Using average environmental indicators throughout the entire growth period as the evaluation basis fails to identify and quantify the transient or extreme environmental stresses encountered by *Arundinaria repens* at different critical growth stages. This averaged approach masks the substantial impact of specific phenological stress events on plant development, leading to a disconnect between the evaluation results and the true stress resistance of the germplasm. Furthermore, methods based on single-point independent analysis or simple comparison of multi-point data can only reflect the static performance of germplasm under discrete environmental conditions, failing to characterize and analyze the complex, dynamic, and temporally correlated combinations of environmental pressures experienced by the same germplasm throughout its entire growth cycle across different geographical regions. Accurately quantifying phenologically specific environmental stresses and, based on this, achieving holistic pattern recognition of the cross-regional, full-growth-cycle stress response trajectory of germplasm is a key challenge for improving the accuracy of *Arundinaria repens* germplasm adaptability evaluation. Summary of the Invention

[0004] The purpose of this invention is to provide a method for evaluating the environmental adaptability of Phragmites australis germplasm resources, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, this invention provides a method for evaluating the environmental adaptability of Phragmites australis germplasm resources, the method comprising:

[0006] Periodic environmental sequence data and synchronous phenotypic development data of Reed hyacinth germplasm cultivation experimental areas set up at multiple independent ecological test sites were obtained. The periodic environmental sequence data and the phenotypic development data were matched and aligned according to the same observation period identifier to form a germplasm environmental response record with spatiotemporal consistency.

[0007] For each germplasm environment response record, based on the preset key phenological period division points of Phragmites australis, the germplasm environment response record is extracted and recombined to generate multiple phenological response sub-records corresponding to different phenological stages;

[0008] An exclusive environmental stress profile is constructed for each of the phenological response sub-records. The environmental stress profile is generated by extracting the extreme values ​​and durations of environmental factors that exceed the preset tolerance range of the corresponding phenological period of Phragmites australis from the phenological response sub-records.

[0009] By linking all environmental stress profiles of the same Arundo donax germplasm at all test sites, arranged in phenological order, an environmental adaptation pathway map of the Arundo donax germplasm across regions and throughout its entire growth period is formed.

[0010] The environmental adaptability path map is input into a pre-trained multidimensional adaptive discriminant network for feature parsing. The multidimensional adaptive discriminant network outputs classification labels that characterize the adaptability level of Arundo donax germplasm under different environmental stress modes. The environmental adaptability evaluation of Arundo donax germplasm resources is completed based on the classification labels.

[0011] Preferably, the acquisition of periodic environmental sequence data and synchronous phenotypic development data of *Arundinaria mirifica* germplasm cultivation experimental areas arranged at multiple independent ecological testing sites includes:

[0012] A uniform observation time granularity is set, and environmental parameters including temperature, humidity, light intensity, soil moisture content, and soil salinity are continuously collected at each independent ecological test point through a deployed sensor array. The data are packaged according to the observation time granularity and labeled with the observation period to form the periodic environmental sequence data.

[0013] Under the same observation time granularity, the plant height, stem diameter, number of leaves, and biomass increment of Arundo donax are recorded by image acquisition device or manual measurement, and the recorded data are bound with the same observation period identifier to form the Arundo donax phenotypic development data;

[0014] A central data pool is established to collect and store the periodic environmental sequence data with observation period identifiers from all independent ecological test sites, along with the Phragmites australis phenotypic development data.

[0015] Preferably, for each germplasm environmental response record, based on the preset key phenological period division points of *Arundinaria rapa*, the germplasm environmental response record is truncated and recombined to generate multiple phenological response sub-records corresponding to different phenological stages, including:

[0016] The complete growth period of the evaluated Phragmites australis germplasm is defined, and multiple key phenological period division points are preset within the complete growth period. The key phenological period division points include the seedling stage, tillering stage, jointing stage, and maturity stage.

[0017] Read a germplasm environment response record and, based on the time information contained in the record, locate the observation cycle identifier corresponding to each preset key phenological period division point in the germplasm environment response record;

[0018] Using the observation cycle markers corresponding to two adjacent key phenological period division points as the start and end boundaries, all environmental sequence data and Reed phenotypic development data within the corresponding time period are extracted from the germplasm environmental response record to form a phenological response sub-record;

[0019] Traverse all preset key phenological period division intervals, complete the extraction and recombination of a single germplasm environmental response record, and generate multiple phenological response sub-records arranged in chronological order for a single experiment of the same Reed germplasm at the same test point.

[0020] Preferably, constructing a unique environmental stress profile for each of the phenological response sub-records includes:

[0021] For a phenological response sub-record, the preset tolerance range of the Arundinaria germplasm to various environmental factors in the current phenological stage is retrieved from the Arundinaria germplasm characteristics database. The preset tolerance range includes an upper threshold and a lower threshold.

[0022] The environmental factor sequence data in the phenological response sub-record are analyzed one by one to identify all data points in the sequence that exceed the corresponding upper limit threshold or fall below the corresponding lower limit threshold, and the environmental factor value at each time the threshold is exceeded and the number of observation cycles for each exceeding event are recorded.

[0023] By accumulating the values ​​and durations of the same environmental factor in all out-of-event events, and combining the statistical distribution characteristics of the environmental factor throughout the entire time period of the phenological response sub-record, a sub-profile is generated to describe the degree of stress of the environmental factor on Reed truncata during this stage.

[0024] The sub-portraits of all analyzed environmental factors during the phenological stage are aggregated to form a comprehensive environmental pressure profile specific to this phenological response sub-record, which is stored in a structured data format.

[0025] Preferably, the environmental stress profiles of the same *Arundinaria mirifica* germplasm, arranged sequentially by phenological stage at all test sites, form an environmental adaptation pathway map of the *Arundinaria mirifica* germplasm across regions and throughout its entire growth period, including:

[0026] Using the germplasm identifier of Phragmites australis as an index, the database was used to retrieve the complete environmental stress profiles of the Phragmites australis germplasm produced by each independent ecological test site.

[0027] The independent ecological test sites are sorted according to their geographical or climatic characteristics, and within each test site, the environmental pressure profiles of multiple phenological stages are arranged in chronological order of the occurrence of phenological events.

[0028] The sorted environmental stress profiles from all test points are logically linked and integrated according to the order of test points and phenological stages to construct a multi-dimensional data graph structure, which is the environmental adaptability path graph. Each node represents an environmental stress profile of a phenological stage, and the connecting lines represent spatiotemporal or phenological sequence relationships.

[0029] Preferably, the multidimensional adaptive discriminant network is a pre-trained machine learning model, and its pre-training process includes:

[0030] We collected environmental adaptation pathway maps of various plant germplasms with known final adaptability that had been evaluated in the past as training samples, and labeled each training sample with its true and comprehensive environmental adaptability level label.

[0031] An initial network model with a multi-layer nonlinear transformation structure is constructed, and the multi-dimensional data graph structure of the environmental adaptation path graph is transformed into feature vectors, which are used as inputs to the initial network model.

[0032] The initial network model is iteratively trained using labeled training samples. By adjusting the internal parameters of the network, the difference between the predicted classification label output by the initial network model and the real environment adaptability level label of the training samples is minimized.

[0033] When the prediction accuracy reaches the preset standard, training stops, and the multidimensional adaptive discriminant network that can be used to evaluate new Arundinaria germplasm is obtained.

[0034] Preferably, the step of inputting the environmental adaptive path map into a pre-trained multidimensional adaptive discriminant network for feature parsing includes:

[0035] The environmental adaptability path graph is normalized to ensure that its dimensions are consistent with the input dimensions required by the multidimensional adaptive discriminant network.

[0036] The normalized environmental adaptability path map data is input into the input layer of the multidimensional adaptive discriminant network;

[0037] The multidimensional adaptive discriminant network extracts and abstracts features from the input data layer by layer through multiple hidden layers, and finally generates a probability distribution vector in the output layer. Each element in the probability distribution vector corresponds to a preset fitness level.

[0038] The fitness level corresponding to the element with the highest probability value in the probability distribution vector is selected as the classification label output by the multidimensional adaptive discriminant network.

[0039] Preferably, after completing the environmental adaptability evaluation of the Reed germplasm resources based on the classification labels, the method further includes steps of verifying the evaluation results and updating the database.

[0040] The germplasm identifiers of Phragmites australis generated in this evaluation, their corresponding environmental adaptability pathway maps, and the classification labels output by the multidimensional adaptability discrimination network are packaged together to generate a germplasm adaptability evaluation file.

[0041] The germplasm adaptability evaluation archive is compared with the historical evaluation records of the Reed germplasm in the germplasm resource database. If the classification label of this evaluation is consistent with the historical mainstream evaluation conclusion, it is directly archived; if there is a significant inconsistency, a manual review is triggered.

[0042] For evaluation files that trigger manual review, an expert system or human will conduct a secondary analysis based on the original environmental stress profile and phenological response sub-records, and update the final adaptability conclusions after review and confirmation to the germplasm resource database. At the same time, the complete file of this evaluation will be used as a new training sample for subsequent incremental training of the multidimensional adaptability discrimination network.

[0043] Preferably, the step of extracting all environmental sequence data and *Arundinaria lobata* phenotypic development data within the corresponding time period from the germplasm environmental response record, using the observation period markers corresponding to two adjacent key phenological period division points as the start and end boundaries, to form a phenological response sub-record includes:

[0044] Identify the phenological stage that needs to be addressed, wherein the phenological stage is defined by two adjacent critical phenological period dividing points;

[0045] In the timeline of the germplasm environment response record, locate the starting observation period marker corresponding to the starting key phenological period division point, and the ending observation period marker corresponding to the ending key phenological period division point.

[0046] Extract all data entries whose observation period identifier is greater than or equal to the start observation period identifier and less than or equal to the end observation period identifier from the data structure stored in the germplasm environment response record.

[0047] The extracted data entries were categorized and organized into two types: environmental sequence data and Reed phenotypic development data, ensuring that each data entry contained complete observation period identifiers, environmental parameter values, and Reed phenotypic measurements.

[0048] The categorized and organized data is packaged into a new independent data set, which is a phenological response sub-record corresponding to a specific phenological stage.

[0049] Preferably, the step of analyzing the environmental factor sequence data in the phenological response sub-record one by one, identifying all data points in the sequence that exceed the corresponding upper limit threshold or fall below the corresponding lower limit threshold, and recording the environmental factor value at each time the threshold is exceeded and the number of observation periods for each exceeding event includes:

[0050] Read all time-series data points of the first environmental factor sequentially from the current phenological response sub-record;

[0051] The value of each data point is compared with the preset upper limit threshold and preset lower limit threshold of the environmental factor in the current phenological stage;

[0052] When the value of a data point is detected to be greater than a preset upper threshold or less than a preset lower threshold, the data point is marked as the starting point of an abnormal event, and the specific value of the environmental factor when the abnormal event occurs is recorded.

[0053] Continue scanning the time series data points until a data point whose value recovers to the preset tolerance range is encountered, and mark it as the end point of the abnormal event;

[0054] Calculate the number of consecutive observation periods between the start point and the end point, and use this as the number of continuous observation periods for this out-of-bounds event;

[0055] The environmental factor values, anomaly start point location, anomaly end point location, and number of continuous observation periods for this event will be stored as a complete record.

[0056] Repeat the process until all time-series data points of the current environmental factor in the phenological response sub-record have been traversed and all out-of-series events have been identified;

[0057] Next, the time series data of the next environmental factor is read sequentially, and the above identification and recording steps are repeated until the analysis of all environmental factor sequence data in the phenological response sub-record is completed.

[0058] Compared with the prior art, the beneficial effects of the present invention are:

[0059] By segmenting observation records according to preset key phenological periods and extracting extreme values ​​and durations of environmental factors exceeding preset tolerance ranges for each phenological stage, an environmental stress profile is constructed, refining the granularity of environmental analysis from the entire growing season to specific physiological stages. This method can accurately capture discrete stress events that substantially affect the growth and development of Reed Bamboo, avoiding the problem of diluted or masked stress signals caused by using average environmental data. It transforms the assessment basis of environmental adaptability from a general climatic background to a clear and specific list of adverse disturbances, improving the assessment's targeting and sensitivity to abiotic stresses such as wind, cold, drought, and floods in actual production.

[0060] By concatenating all environmental stress profiles of the same *Arundinaria mirifica* germplasm across all test sites in phenological order, a cross-regional environmental adaptability pathmap throughout the entire growth period is formed. This sequenced pathmap is then input into a pre-trained multidimensional adaptability discriminant network for feature analysis. The evaluation object is transformed from the static performance of germplasm at a single point to the dynamic stress trajectory experienced by the germplasm in the spatiotemporal dimension. The network model can learn and recognize the deep patterns contained in this complex sequence. This achieves holistic and systematic discrimination of germplasm adaptability, distinguishing germplasm resources that perform similarly in simple multi-point comparisons but have different adaptation strategies to complex stress sequences, thereby uncovering superior materials with greater broad adaptability or specific stress resistance. Attached Figure Description

[0061] Figure 1 This is a schematic diagram illustrating the working principle of the method for evaluating the environmental adaptability of Phyllostachys amurensis germplasm resources according to the present invention.

[0062] Figure 2 A flowchart for data acquisition and aggregation;

[0063] Figure 3 A flowchart for generating phenological response sub-records;

[0064] Figure 4 A graph showing the predicted environmental adaptability levels of Phyllostachys amurensis germplasm resources;

[0065] Figure 5 A comparison of the stress intensity and duration of environmental factors affecting the cultivation of Phragmites australis. Detailed Implementation

[0066] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] Please see Figure 1This invention provides a method for evaluating the environmental adaptability of *Arundinaria mirifica* germplasm resources. The method includes: acquiring periodic environmental sequence data and synchronous *Arundinaria mirifica* phenotypic development data from multiple independent ecological testing sites in a *Arundinaria mirifica* germplasm cultivation experimental area; matching and aligning the periodic environmental sequence data and the *Arundinaria mirifica* phenotypic development data according to the same observation period identifier to form a germplasm environmental response record with spatiotemporal consistency; for each germplasm environmental response record, based on preset key phenological period division points for *Arundinaria mirifica*, extracting and recombining the record to generate multiple phenological response sub-records corresponding to different phenological stages; constructing a unique environmental stress profile for each phenological response sub-record, the environmental stress profile being generated by extracting the extreme values ​​and durations of environmental factors exceeding the preset tolerance range of the corresponding phenological period of *Arundinaria mirifica* from the phenological response sub-record; and concatenating all environmental stress profiles of the same *Arundinaria mirifica* germplasm at all testing sites, arranged in phenological stage order, to form a cross-regional, full-growth-period environmental adaptability path map of the *Arundinaria mirifica* germplasm. The environmental adaptability path map is input into a pre-trained multidimensional adaptive discriminant network for feature parsing. The multidimensional adaptive discriminant network outputs classification labels that characterize the adaptability level of Arundo donax germplasm under different environmental stress modes. The environmental adaptability evaluation of Arundo donax germplasm resources is completed based on the classification labels.

[0068] In one embodiment of the present invention, see [reference] Figure 2 A unified observation time granularity was established. At each independent ecological test site, environmental parameters, including temperature, humidity, light intensity, soil moisture content, and soil salinity, were continuously collected using a deployed sensor array. These parameters were packaged according to the observation time granularity and labeled with an observation period identifier to form the periodic environmental sequence data. At the same observation time granularity, the plant height, stem diameter, number of leaves, and biomass increment of *Arundinaria salsa* were recorded using image acquisition devices or manual measurement. The recorded data were then bound to the same observation period identifier to form the *Arundinaria salsa* phenotypic development data. A central data pool was established to collect and store the periodic environmental sequence data with observation period identifiers from all independent ecological test sites, along with the *Arundinaria salsa* phenotypic development data.

[0069] In practice, a unified observation time granularity is set, defined as 1 day. At each independent ecological test site, environmental parameters including temperature, humidity, light intensity, soil moisture content, and soil salinity are continuously collected through a deployed sensor array. The temperature sensor in the array records the daily average, maximum, and minimum temperatures; the humidity sensor records the relative humidity; the light intensity sensor records the daily cumulative photosynthetically active radiation; and the soil moisture content and soil salinity sensors record the volumetric water content and electrical conductivity of the root zone soil, respectively. All environmental parameter values ​​collected daily are packaged according to the observation time granularity, and each data packet is assigned a unique observation period identifier in the format "Test Site Number - Year Month Day," thus forming a periodic environmental sequence data.

[0070] In some embodiments, at the same observation time granularity, top-view and side-view images of *Arundinaria mirifica* plants are captured daily using an image acquisition device fixed in the experimental plot. The images are automatically analyzed by an image analysis algorithm, which extracts information on plant height, stem diameter, and number of leaves. For biomass increment, at each preset key phenological stage, the fresh weight of the aboveground parts is obtained and recorded through manual sampling and weighing. The plant height, stem diameter, and number of leaves data obtained from manual measurement and image analysis are all linked to the corresponding observation period identifier. For example, within the observation period identifier "S01-20230517", the plant height of *Arundinaria mirifica* germplasm A is recorded as 85 cm, the stem diameter as 8.2 mm, and the number of leaves as 12. These data are then associated with and stored with the observation period identifier "S01-20230517" to form *Arundinaria mirifica* phenotypic development data.

[0071] It is understandable that a central data pool is established, constructed using a relational database. Periodic environmental sequence data and *Arundinaria lobata* phenotypic development data, both bearing observation period identifiers, from all independent ecological testing sites, are aggregated into the central data pool via network transmission protocols. Within the central data pool, data tables are stored in association based on the observation period identifier and the *Arundinaria lobata* germplasm identifier. For a record with the observation period identifier "S02-20230601," the central data pool simultaneously stores the environmental parameter set from test site S02, as well as the plant height, stem diameter, and leaf number measurements of *Arundinaria lobata* germplasm B on the same day.

[0072] In practice, the data packaging process for periodic environmental sequence data follows defined integration rules. For some raw sensor data sampled at higher frequencies, aggregation calculations are required based on the granularity of the observation time to generate representative values ​​for each day. For example, for temperature data sampled every minute, when packaging to generate periodic environmental sequence data, the arithmetic mean of all sampling points within that day needs to be calculated as the daily average temperature, and the maximum and minimum values ​​within that day are selected as the daily maximum and minimum temperatures. This aggregation process can be expressed by the following formula:

[0073] ;

[0074] in: This represents the aggregated value of a specific environmental factor within the observation time granularity. This represents the total number of raw data points collected by the sensor within this observation time granularity. Indicates the first The values ​​of the original data points.

[0075] In some embodiments, the biomass increment data in the *Arundinaria pulmonata* phenotypic development data has different binding logic. Since destructive biomass measurements cannot be performed daily, the observation period identifier for biomass increment is not a single-day identifier, but rather aligned with the observation period identifier of key phenological stage division points. For example, the change in biomass is calculated between the observation period identifier at the start of the tillering stage and the observation period identifier at the end of the tillering stage, and bound to the observation period identifier marking the end of that phenological stage. The phenotypic data bound to the observation period identifier "S03-20230720" includes, in addition to the plant height and stem diameter data for that day, the accumulated biomass increment value from the previous phenological stage to "S03-20230720". Optionally, the data structure of the central data pool includes an environmental parameter table and a phenotypic parameter table. The primary key of the environmental parameter table is composed of the observation period identifier, test point number, and sensor number, and each record stores the specific value of an environmental parameter under the corresponding observation period identifier. The primary key of the phenotypic parameter table consists of an observation period identifier, a test point number, and a reed germplasm identifier. Each record stores all phenotypic measurements of the reed germplasm under the corresponding observation period identifier. It can be understood that a unified setting of the observation time granularity is a prerequisite for data matching and alignment. In practice, if the sensor array at a certain test point malfunctions, resulting in missing data for a particular day, the periodic environmental sequence data under the corresponding observation period identifier will be marked as null or an estimated value will be generated using an interpolation algorithm to maintain the continuity of the data sequence on the time axis.

[0076] In one embodiment of the present invention, see [reference] Figure 3To evaluate the *Arundinaria mirifica* germplasm, a complete growth period is defined, and multiple key phenological period division points are preset within this period. These key phenological period division points include emergence, tillering, jointing, and maturity. A germplasm environmental response record is read, and based on the time information contained in the record, the observation period identifier corresponding to each preset key phenological period division point is located. Using the observation period identifiers corresponding to two adjacent key phenological period division points as start and end boundaries, all environmental sequence data and *Arundinaria mirifica* phenotypic development data within the corresponding time period are extracted from the germplasm environmental response record to form a phenological response sub-record. The phenological stage that needs to be processed is determined, defined by two adjacent key phenological period division points. In the timeline of the germplasm environmental response record, the starting observation period identifier corresponding to the starting key phenological period division point and the ending observation period identifier corresponding to the ending key phenological period division point are located. From the data structure stored in the germplasm environmental response record, all data entries with observation period identifiers greater than or equal to the starting observation period identifier and less than or equal to the ending observation period identifier are extracted. The extracted data entries were categorized and organized into two types: environmental sequence data and Reed phenotypic development data, ensuring that each data entry included complete observation period identifiers, environmental parameter values, and Reed phenotypic measurements. The categorized data was then packaged into a new, independent dataset, namely, a phenological response sub-record corresponding to a specific phenological stage. By traversing all preset key phenological period division intervals, the environmental response records of each individual germplasm were extracted and recombined, generating multiple phenological response sub-records arranged chronologically for a single experiment of the same Reed germplasm at the same test site.

[0077] In practice, a complete growth period is defined for the evaluated *Arundinaria mirifica* germplasm. A complete growth period refers to the total number of days from seed germination to technological maturity. For example, for the *Arundinaria mirifica* germplasm "Jianghe No. 1," its complete growth period is defined as 125 days. Within this complete growth period, several key phenological stages are pre-defined, including emergence, tillering, jointing, and maturity. These pre-defined key phenological stages are determined based on the general growth and development patterns of *Arundinaria mirifica* germplasm. The emergence stage is defined as the date when 50% of the seedlings emerge; the tillering stage is defined as the date when 50% of the plants begin to produce tillers; the jointing stage is defined as the date when 50% of the plants' main stem begins to elongate; and the maturity stage is defined as the date when 50% of the plants' lower stem leaves turn yellow and the stem walls harden.

[0078] In some embodiments, a germplasm environment response record is read. This record is stored in a central data pool and associated with the Phragmites australis germplasm identifier "Jianghe No. 1" and the test site number "N01". Based on the time information contained in the germplasm environment response record, the observation period identifier corresponding to each preset key phenological stage division point is located. For example, by querying the agricultural records and image determination records associated with "Jianghe No. 1" at the "N01" test site, it is determined that the seedling stage division point corresponds to the observation period identifier "N01-20230410", the tillering stage division point corresponds to the observation period identifier "N01-20230520", the jointing stage division point corresponds to the observation period identifier "N01-20230625", and the maturity stage division point corresponds to the observation period identifier "N01-20230812". In practice, the observation cycle identifiers corresponding to two adjacent key phenological period division points are used as the start and end boundaries. All environmental sequence data and *Arundinaria ruthes* phenotypic development data within the corresponding time period are extracted from the germplasm environmental response record to form a phenological response sub-record. The phenological stage that needs to be processed is determined. The phenological stage is defined by two adjacent key phenological period division points; for example, the early vegetative growth stage between seedling emergence and tillering needs to be processed. In the timeline of the germplasm environmental response record, the starting observation cycle identifier corresponding to the starting key phenological period division point is located (starting observation cycle identifier: "N01-20230410"), and the ending observation cycle identifier corresponding to the ending key phenological period division point is located (ending observation cycle identifier: "N01-20230520"). From the data structure stored in the germplasm environmental response record, all data entries with observation cycle identifiers greater than or equal to the starting observation cycle identifier "N01-20230410" and less than or equal to the ending observation cycle identifier "N01-20230520" are extracted. The extracted data entries include daily environmental parameter records and Reed phenotypic measurement records from April 10, 2023 to May 20, 2023.

[0079] It is understandable that the extracted data entries are categorized and organized into two types: environmental sequence data and Reed phenotypic development data. This categorization ensures that each data entry includes a complete observation period identifier, environmental parameter values, and Reed phenotypic measurements. For example, the data entry with the observation period identifier "N01-20230415" includes, after categorization, the average temperature of that day (21.5℃), maximum temperature (28℃), minimum temperature (16℃), average humidity (65%), cumulative light intensity (18.2 MJ / m²), soil moisture content (22.3%), soil salinity (0.8 dS / m²), and the plant height (15.2 cm), stem diameter (1.8 mm), and number of leaves (5) of the Reed germplasm "Jianghe No. 1" on that day. Optionally, all preset key phenological period intervals can be traversed to complete the extraction and recombination of individual germplasm environmental response records. For the germplasm environmental response record of the Phragmites australis germplasm “Jianghe No. 1” at test site “N01”, the preset key phenological period division intervals include: “emergence to tillering stage”, “tillering stage to jointing stage”, and “jointing stage to maturity stage”. The above-mentioned location, extraction, classification, and encapsulation operations are performed sequentially for each interval, thereby generating multiple phenological response sub-records arranged in chronological order for a single experiment of the same Phragmites australis germplasm at the same test site. The generation order is: “emergence-tillering stage” phenological response sub-record, “tillering-jointing stage” phenological response sub-record, and “jointing stage-maturity stage” phenological response sub-record.

[0080] In some embodiments, the data encapsulation of phenological response sub-records follows a uniform structured format. The structured format includes a header and body data. The header records the *Arundinaria praecox* germplasm identifier, test site number, start observation period identifier, end observation period identifier, and the name of the phenological stage it represents. The body data is an ordered list, where each item corresponds to an observation period identifier, and each item stores all environmental sequence data fields and *Arundinaria praecox* phenotypic development data fields corresponding to that observation period identifier. The duration of the phenological response sub-record can be calculated using the start and end observation period identifiers, using the following formula:

[0081] ;

[0082] in: Indicates the number of days that a phenological stage lasts. This indicates the date sequence number corresponding to the termination observation period identifier. This indicates the date sequence number corresponding to the starting observation period identifier. For example, if the starting observation period identifier "N01-20230410" corresponds to sequence number 100, and the ending observation period identifier "N01-20230520" corresponds to sequence number 140, then this represents the number of days in the phenological response sub-record for the "seedling stage - tillering stage". sky.

[0083] Understandably, the truncation and recombination process ensures precise alignment of environmental sequence data with *Arundinaria mirifica* phenotypic development data at the phenological scale. For example, in the phenological response sub-records of the "tillering-jointing stage," the daily cumulative light intensity data and the daily increase in *Arundinaria mirifica* plant height on the same day are perfectly matched temporally. This matching relationship allows for subsequent analysis of the correlation between environmental factor changes and *Arundinaria mirifica* growth responses within this phenological stage. For test sites with multiple replicate experimental plots within the complete growth period, the germplasm environmental response records of each experimental plot are independently subjected to the above truncation and recombination operations to generate their own set of phenological response sub-records for subsequent statistical analysis.

[0084] In one embodiment of the present invention, for a phenological response sub-record, the preset tolerance ranges of the *Arundinaria przewalskii* germplasm to various environmental factors at the current phenological stage are retrieved from the *Arundinaria przewalskii* germplasm characteristic database. The preset tolerance ranges include an upper threshold and a lower threshold. The sequence data of each environmental factor in the phenological response sub-record are analyzed one by one, identifying all data points in the sequence that exceed the corresponding upper threshold or fall below the corresponding lower threshold, and recording the environmental factor value at each exceedance and the number of observation periods for each exceedance event. All time-series data points of the first environmental factor are read sequentially from the current phenological response sub-record. The value of each data point is compared with the preset upper threshold and preset lower threshold of the environmental factor at the current phenological stage. When the value of a data point is detected to be greater than the preset upper threshold or less than the preset lower threshold, the data point is marked as the starting point of an abnormal event, and the specific value of the environmental factor at the time of the abnormal event is recorded. The time-series data points are scanned backward until a data point is encountered whose value recovers to the preset tolerance range, which is then marked as the ending point of the abnormal event. The number of consecutive observation periods from the start point to the end point is calculated as the duration of this over-event. The environmental factor values, abnormal start point location, abnormal end point location, and duration of observation periods for this over-event are stored as a complete record. This process is repeated until all time-series data points of the current environmental factor in the phenological response sub-record have been traversed and all over-events have been identified. Then, the time-series data of the next environmental factor is read sequentially, and the above identification and recording steps are repeated until the analysis of all environmental factor sequence data in the phenological response sub-record is completed. The values ​​and durations of the same environmental factor in all over-events are accumulated, and combined with the statistical distribution characteristics of the environmental factor throughout the entire time period of the phenological response sub-record, a sub-profile describing the degree of stress of the environmental factor on Reed twigs at this stage is generated. The sub-profiles of all analyzed environmental factors in the phenological stage are aggregated to form a comprehensive environmental pressure profile specific to this phenological response sub-record, which is stored in a structured data format.

[0085] In practical implementation, for a phenological response sub-record, such as the "jointing-maturity" phenological response sub-record of the Phragmites australis germplasm "Jianghe No. 1" at test point "N01", the preset tolerance ranges of the Phragmites australis germplasm "Jianghe No. 1" to various environmental factors during the current "jointing-maturity" phenological stage are retrieved from the Phragmites australis germplasm characteristic database. The preset tolerance range includes upper and lower thresholds. For example, the preset tolerance range for the Phragmites australis germplasm "Jianghe No. 1" to the maximum daily temperature during the jointing to maturity stage is an upper threshold of 35 degrees Celsius and a lower threshold of 10 degrees Celsius; the preset tolerance range for soil moisture content is an upper threshold of 90% of field capacity and a lower threshold of 50% of field capacity; and the preset tolerance range for soil salinity is an upper threshold of 4.0 dS / m.

[0086] In some embodiments, the environmental factor sequence data in the phenological response sub-record are analyzed one by one to identify all data points in the sequence that exceed the corresponding upper threshold or fall below the corresponding lower threshold. For example, the daily maximum temperature time series data in the "jointing-maturity" phenological response sub-record is analyzed. This time series data includes daily values ​​from the observation period identifier "N01-20230625" to "N01-20230812". The value of each data point is compared with the preset upper threshold of 35 degrees Celsius and the preset lower threshold of 10 degrees Celsius for the daily maximum temperature in the current phenological stage. When the value of a data point is detected to be greater than the preset upper threshold of 35 degrees Celsius or less than the preset lower threshold of 10 degrees Celsius, this data point is marked as the starting point of an abnormal event, and the specific value of the environmental factor at the time of the abnormal event is recorded. The time series data points are scanned backward until a data point is encountered whose value recovers to the preset tolerance range, which is marked as the ending point of the abnormal event. The number of consecutive observation periods contained from the starting point to the ending point is calculated as the number of continuous observation periods of this abnormal event. The environmental factor values, anomaly start and end points, and number of observation periods for each out-of-range event are stored as a complete record. This process is repeated until all time-series data points for the current environmental factor in the phenological response sub-record have been traversed and all out-of-range events have been identified. Then, the time-series data for the next environmental factor is read sequentially, and the above identification and recording steps are repeated until the analysis of all environmental factor sequence data in the phenological response sub-record is complete.

[0087] In practice, the values ​​and durations of the same environmental factor in all exceedance events are accumulated. Combined with the statistical distribution characteristics of the environmental factor throughout the entire phenological response sub-record period, a sub-profile describing the stress level of the environmental factor on *Arundinaria repens* during this stage is generated. The accumulation process includes counting the total number of exceedance events for the environmental factor, the total duration exceeding the preset tolerance range, and the average and maximum deviations of the environmental factor from the threshold in all exceedance events. The statistical distribution characteristics of the environmental factor throughout the entire phenological response sub-record period include the mean, standard deviation, maximum, and minimum values ​​of the environmental factor throughout the entire phenological stage. For example, for the daily maximum temperature factor, three exceedance events were identified: the first exceeded 35 degrees Celsius, reaching 36.2 degrees Celsius and lasting for 3 days; the second reached 37.5 degrees Celsius and lasted for 2 days; and the third reached 35.8 degrees Celsius and lasted for 1 day. The total cumulative exceedance duration was 6 days, the average exceedance was 1.2 degrees Celsius, and the maximum exceedance was 2.5 degrees Celsius. The average daily maximum temperature during the entire "jointing-maturity" period was 31.5 degrees Celsius, with a standard deviation of 3.2 degrees Celsius. Based on these statistics and records of events exceeding the maximum temperature, a temperature stress sub-profile is generated. This sub-profile is a structured data object containing the aforementioned statistical characteristic fields. To quantify the degree of stress, a formula for calculating a stress intensity index is introduced:

[0088] ;

[0089] in: This represents the stress intensity index relative to a preset upper limit threshold. This indicates the total number of events that exceed the preset upper limit threshold. Indicates the first This time the average value of environmental factors in the event exceeded the limit. This represents the preset upper limit threshold of environmental factors. Indicates the first The number of days the event exceeds the limit. For cases below the lower threshold, the formula can be modified similarly.

[0090] It can be understood that the sub-portfolios of all analyzed environmental factors during the phenological stage form a comprehensive environmental stress profile specific to this phenological response sub-record. The environmental stress profile is stored in a structured data format, such as JSON. An environmental stress profile for the "jointing-maturity" phenological response sub-record contains the following structure: {"Phenological Stage": "Jointing-Maturity", "Temperature Stress": {"Total Days Exceeding Standard": 6, "Average Exceedance": 1.2, "Maximum Exceedance": 2.5, "Stress Intensity Index S": 15.8, ...}, "Soil Salinity Stress": {"Total Days Exceeding Standard": 0, ...}, "Water Stress": {...}, ...}. This environmental stress profile summarizes the characteristics of the main abiotic stresses experienced by the *Phragmites australis* germplasm "Jianghe 1" at test site "N01" during a specific phenological stage.

[0091] Optionally, when generating sub-portfolios, the combination of statistical distribution features can be reflected in the calculation of relative stress intensity. In addition to calculating the absolute magnitude and duration of the exceedance, the characteristics of the exceedance event are compared with the background fluctuations of the environmental factor throughout the phenological stage. In some embodiments, the process of identifying exceedance events and recording their durations needs to handle boundary cases. For example, when scanning to the end of the time-series data, if an exceedance event has not yet ended, the last data point is recorded as a temporary end point for this exceedance event, and the number of continuous observation periods is calculated up to this temporary end point. Simultaneously, the event is marked as "not recovered" in the sub-portfolio to distinguish it from exceedance events that have ended within the phenological stage. It is understood that for cases where the preset tolerance range is a one-sided threshold, such as light intensity having only a lower threshold and soil salinity having only an upper threshold, the analysis process only requires one-sided comparisons. For factors with upper and lower thresholds, such as soil moisture content, exceedance events above the upper threshold and below the lower threshold need to be identified and recorded separately, and these events are distinguished into two different types of stress sub-portfolios when generating sub-portfolios. Ultimately, the stress profiles of all monitored environmental factors are aggregated to form a comprehensive environmental stress profile for this phenological stage.

[0092] In one embodiment of the present invention, using the *Arundinaria spp.* germplasm identifier as an index, all constructed environmental stress profiles generated by the *Arundinaria spp.* germplasm trials at each independent ecological test site are retrieved from the database. These profiles are sorted according to the geographical or climatic characteristics of the independent ecological test sites, and within each test site, the environmental stress profiles for multiple phenological stages are arranged in chronological order of phenological occurrence. The sorted environmental stress profiles from all test sites are logically linked and integrated according to the test site order and phenological stage order to construct a multidimensional data graph structure. This multidimensional data graph structure is the environmental adaptability path graph, where each node represents an environmental stress profile for a phenological stage, and connecting lines represent spatiotemporal or phenological sequence relationships. The multidimensional adaptability discriminant network is a pre-trained machine learning model. Its pre-training process includes: collecting environmental adaptability path graphs of various plant germplasms with known final adaptability that have been evaluated historically as training samples, and labeling each training sample with its true and comprehensive environmental adaptability level label. An initial network model with a multi-layer nonlinear transformation structure is constructed, transforming the multi-dimensional data graph structure of the environmental adaptability path map into feature vectors, which serve as the input to the initial network model. The initial network model is iteratively trained using labeled training samples. By adjusting the network's internal parameters, the difference between the predicted classification labels output by the initial network model and the true environmental adaptability level labels of the training samples is minimized. Training is stopped when the prediction accuracy reaches a preset standard, resulting in the multi-dimensional adaptability discriminant network that can be used to evaluate new Arundinaria mirifica germplasm.

[0093] In practice, using the Phragmites australis germplasm identifier as an index, all pre-constructed environmental stress profiles generated by the Phragmites australis germplasm "Jianghe No. 1" at each independent ecological test site were retrieved from the database. For example, the Phragmites australis germplasm "Jianghe No. 1" underwent cultivation trials at three independent ecological test sites "N01", "S02", and "W03". The trials at each test site generated environmental stress profiles for three phenological stages for the Phragmites australis germplasm "Jianghe No. 1", thus retrieving a total of nine environmental stress profiles.

[0094] In some embodiments, the test points are sorted according to their geographical or climatic characteristics. The sorting criteria can be single or combined indicators such as the latitude and longitude of the test points, average annual temperature, and annual precipitation. For example, if the test points are sorted from lowest to highest average annual temperature, and test point "N01" has an average annual temperature of 10.5℃, test point "W03" has an average annual temperature of 12.8℃, and test point "S02" has an average annual temperature of 18.2℃, then the sorted order of the test points would be "N01", "W03", and "S02". Within each test point, the environmental pressure profiles of multiple phenological stages are arranged in chronological order of phenological occurrence, specifically in the order of "seedling stage - tillering stage", "tillering stage - jointing stage", and "jointing stage - maturity stage".

[0095] In practice, the environmental stress profiles from all test points, after being sorted, are logically linked and integrated according to the order of the test points and the order of the phenological stages. Taking the Phragmites australis germplasm “Jianghe No. 1” as an example, its test point order is “N01”, “W03”, and “S02”, and the phenological stage order within each test point is three stages. After logical linking, a sequence containing nine nodes is formed. A multidimensional data graph structure is constructed, which is the environmental adaptability path graph. Each node represents an environmental stress profile of a phenological stage, and the node attributes store all the structured data of that profile. Connecting lines represent spatiotemporal or phenological sequence relationships. The connecting lines are directional, pointing from earlier phenological stages to later phenological stages, and from earlier-sorted test points to later-sorted test points, as shown in Table 1.

[0096] Table 1: Concatenated Construction Table of Environmental Stress Profiles

[0097]

[0098] It is understandable that a multidimensional adaptive discriminant network is a pre-trained machine learning model. The pre-training process of a multidimensional adaptive discriminant network includes: collecting environmental adaptation path maps of various plant germplasms with known final adaptability, which have been evaluated historically, as training samples. The known final adaptability is determined by field performance experts and categorized into "high adaptability," "medium adaptability," and "low adaptability" levels. For example, the database stores 1000 environmental adaptation path maps of different plant germplasms and their corresponding adaptability level labels. An initial network model with a multi-layer nonlinear transformation structure is constructed. This initial network model can be a graph neural network or an adapted long short-term memory network. The multidimensional data graph structure of the environmental adaptation path map is transformed into feature vectors, which serve as input to the initial network model. This transformation process requires mapping the structured data of each node into numerical vectors through an embedding layer and representing the connections between nodes in the form of an adjacency matrix.

[0099] Optionally, the initial network model can be iteratively trained using labeled training samples. In the specific implementation of iterative training, labeled training samples are used to train the initial network model. These training samples consist of environmental adaptation path maps of various plant germplasms that have been evaluated historically, and each training sample is labeled with a true comprehensive environmental adaptability level label. The initial network model is constructed as a machine learning model with a multi-layer nonlinear transformation structure, such as a graph neural network or a long short-term memory network. Its input layer receives feature vectors transformed from the multi-dimensional data graph structure of the environmental adaptability path map. During each iteration of training, an environmental adaptability path map is input into the initial network model. The model processes the input data through internal graph convolutional layers or similar propagation mechanisms, aggregating node features and neighbor information, and finally generating a predicted classification label at the output layer. The difference between the predicted classification label and the true environmental adaptability level label of the training samples is calculated. This difference is typically quantified using the cross-entropy loss function, and the network's internal parameters are adjusted using the backpropagation algorithm to minimize the difference. In each iteration, an environmental adaptation path graph is input into the initial network model. This model aggregates node features and neighbor information through its internal graph convolutional layers or similar propagation mechanisms, ultimately outputting a predicted classification label. By adjusting the network's internal parameters, the difference between the predicted classification label output by the initial network model and the true environmental adaptation level label of the training samples is minimized. This difference is typically calculated using the cross-entropy loss function and optimized using the backpropagation algorithm. When the prediction accuracy reaches a preset standard, such as a stable classification accuracy above 95% on a reserved validation set, training stops, resulting in a multidimensional adaptive discriminant network that can be used to evaluate new *Arundinaria mirifica* germplasm.

[0100] In some embodiments, the generation of node feature vectors during the construction of the initial network model can incorporate the encoding of geographic features of test points. For example, the feature vector of each node... The feature vector of its corresponding environmental stress profile and the standardized climate feature vector of its respective test point It is composed of multiple parts, and the formula is expressed as:

[0101] ;

[0102] in: This represents the node feature vectors input to the network. This represents the numerical feature vector extracted from the environmental stress profile. The climate feature vector representing the test point, with the symbol... This indicates a vector concatenation operation.

[0103] It is understandable that the collection of training samples needs to ensure diversity, which is reflected in plant germplasm types, test site environmental types, and the stress patterns experienced. For a newly input Reed arum germplasm environmental adaptation path map, the trained multidimensional adaptation discriminant network can analyze its path features and output adaptation level classification labels.

[0104] See Figure 4 This is a prediction chart of the environmental adaptability levels of Arundo donax germplasm resources. Its core purpose is to display the adaptability scores and classifications of different Arundo donax germplasms. The adaptability scores of different germplasms vary considerably, reflecting the differences in their environmental adaptability. This chart serves the screening and promotion of Arundo donax germplasm resources. Highly adaptable germplasms such as "Huaihe No. 3" should be prioritized for planting in multiple environmental regions; medium / low adaptable germplasms (such as Taihu No. 2) can be limited to specific ecological areas where they are adapted; and targeted breeding can be conducted based on the genetic characteristics of highly adaptable germplasms to improve the adaptability of other germplasms.

[0105] In one embodiment of the present invention, the environmental adaptability path map is normalized to ensure its dimensions match the input dimensions required by the multidimensional adaptive discriminant network. The normalized environmental adaptability path map data is then input into the input layer of the multidimensional adaptive discriminant network. The multidimensional adaptive discriminant network extracts and abstracts features from the input data layer by layer through multiple hidden layers, ultimately generating a probability distribution vector at the output layer. Each element in the probability distribution vector corresponds to a preset adaptability level. The adaptability level corresponding to the element with the highest probability value in the probability distribution vector is selected as the classification label output by the multidimensional adaptive discriminant network. The generated Phragmites australis germplasm identifier, its corresponding environmental adaptability path map, and the classification label output by the multidimensional adaptive discriminant network are packaged together to generate a germplasm adaptability evaluation file. The germplasm adaptability evaluation file is compared with historical evaluation records of Phragmites australis germplasm in the germplasm resource database. If the classification label of the current evaluation is consistent with the historical mainstream evaluation conclusions, it is directly archived; if significant inconsistencies are found, a manual review is triggered. For evaluation files that trigger manual review, an expert system or human will conduct a secondary analysis based on the original environmental stress profile and phenological response sub-records, and update the final adaptability conclusions after review and confirmation to the germplasm resource database. At the same time, the complete file of this evaluation will be used as a new training sample for subsequent incremental training of the multidimensional adaptability discrimination network.

[0106] In practical implementation, the environmental adaptability path graph undergoes data normalization. This normalization aims to unify features with different dimensions and value ranges to the same scale. For example, each node in the environmental adaptability path graph contains numerical fields such as temperature stress intensity index, total number of days exceeding water stress limits, and average exceedance of salinity stress limits. The numerical ranges of these fields vary significantly. Data normalization ensures that the dimensions of the environmental adaptability path graph data are consistent with the input dimensions required by the multidimensional adaptive discriminant network. During the training phase, the multidimensional adaptive discriminant network requires the input features to have a mean of 0 and a standard deviation of 1.

[0107] In some embodiments, the normalized environmental adaptation path graph data is input into the input layer of a multidimensional adaptive discriminant network. The multidimensional adaptive discriminant network extracts and abstracts features from the input data layer by layer through multiple hidden layers. These hidden layers can include convolutional layers, graph attention layers, or fully connected layers. For example, for a normalized environmental adaptation path graph representing the Reed germplasm "Jianghe 1", the input layer of the multidimensional adaptive discriminant network receives graph data consisting of feature vectors of all nodes and an adjacency matrix. The first hidden layer performs graph convolution operations to aggregate the feature information of each node and its neighboring nodes. Subsequent hidden layers further combine and abstract these local features to form a representation of the global path.

[0108] In practice, a probability distribution vector is generated at the output layer. Each element in the probability distribution vector corresponds to a preset fitness level, such as "high fitness," "medium fitness," and "low fitness." The probability distribution vector is calculated using the Softmax activation function of the output layer of the multidimensional adaptive discriminant network. The number of neurons in the output layer equals the number of preset fitness levels. The fitness level corresponding to the element with the highest probability value in the probability distribution vector is selected as the classification label output by the multidimensional adaptive discriminant network. For example, for an input, if the probability distribution vector output by the multidimensional adaptive discriminant network is [0.15, 0.80, 0.05], corresponding to "high fitness," "medium fitness," and "low fitness" respectively, then "medium fitness," corresponding to a probability value of 0.80, is selected as the classification label for this evaluation.

[0109] It is understandable that data normalization can employ the Z-score standardization method. For each numerical feature field in the environmental adaptation path graph, the mean and standard deviation of that field across all training samples are calculated. When processing a new environmental adaptation path graph, the feature value corresponding to each node in the path graph is transformed using the following formula:

[0110] ;

[0111] in: Represents the normalized eigenvalues. Represents the original feature values. This represents the mean of the feature in the entire training sample population. This represents the standard deviation of the feature in the training sample population. All feature fields are calculated in sequence to ensure that the normalized values ​​of each feature conform to a standard normal distribution, thereby ensuring the scale consistency of the input data.

[0112] Optionally, the Arundinaria germplasm identifier generated in this evaluation, its corresponding environmental adaptability path map, and the classification label output by the multidimensional adaptability discriminant network are packaged together to generate a germplasm adaptability evaluation file. The germplasm adaptability evaluation file is stored in a structured electronic file format, containing the Arundinaria germplasm identifier field, the characteristic hash value field of the environmental adaptability path map, the classification label field, and the evaluation timestamp field. The germplasm adaptability evaluation file is compared with the historical evaluation records of Arundinaria germplasm in the germplasm resource database. The historical evaluation records contain adaptability conclusions from early trials or expert assessments. If the classification label of this evaluation is consistent with the historical mainstream evaluation conclusion, for example, if the historical mainstream evaluation conclusion is "moderately adaptable," and the classification label output by the multidimensional adaptability discriminant network is also "moderately adaptable," then this germplasm adaptability evaluation file is directly archived to the corresponding record in the germplasm resource database.

[0113] In some embodiments, a manual review flag is triggered if significant inconsistency is found. The determination of significant inconsistency can be based on preset rules, such as when the current classification label differs from the mainstream conclusions of the three most recent historical evaluations, or when the current classification label is "low fitness" while the historical mainstream conclusion is "high fitness." In such cases, a significant inconsistency is identified, and a manual review flag is triggered. For evaluation files that trigger the manual review flag, a secondary analysis is performed by an expert system or a human based on the original environmental stress profile and phenological response sub-records. The expert system can perform reasoning based on a preset stress-response knowledge base, while the manual review involves breeding or cultivation experts reviewing the original data records and the intermediate features output by the multidimensional fitness discriminant network.

[0114] It is understandable that the final adaptability conclusions, after verification and confirmation, will be updated in the germplasm resource database. The update operation includes adding new adaptability evaluation entries to the records of *Arundinaria mirifica* germplasm. These entries will contain the final adaptability conclusions, the source of the conclusions, and an index of the associated germplasm adaptability evaluation archive. Simultaneously, the complete evaluation files will be used as new training samples for subsequent incremental training of the multidimensional adaptability discriminant network. The incremental training process involves using the newly added training samples to perform a limited number of iterative training rounds based on the already trained multidimensional adaptability discriminant network parameters, in order to fine-tune the network parameters and enable the multidimensional adaptability discriminant network to adapt to new data patterns or correct potential discrimination biases.

[0115] See Figure 5 This is a comparative chart of the stress intensity and duration of environmental factors affecting Arundo donax cultivation. Its core purpose is to analyze the stress impact characteristics of different environmental factors on Arundo donax. Humidity exhibits the highest stress intensity (approximately 1) and duration (approximately 26 days), making it the most influential factor on Arundo donax in the current environment. Most factors show a "low intensity, long duration" stress characteristic, while humidity possesses both high intensity and long duration effects. This chart serves to optimize the Arundo donax cultivation environment. For long-duration, low-intensity factors such as soil moisture content, periodic control strategies need to be developed to avoid long-term cumulative effects. The high stress characteristic of humidity can serve as a key reference indicator for selecting Arundo donax cultivation areas.

[0116] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, 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 process, method, article, or apparatus.

[0117] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for evaluating the environmental adaptability of Phyllostachys amurensis germplasm resources, characterized in that, include: Periodic environmental sequence data and synchronous phenotypic development data of Reed hyacinth germplasm cultivation experimental areas set up at multiple independent ecological test sites were obtained. The periodic environmental sequence data and the phenotypic development data were matched and aligned according to the same observation period identifier to form a germplasm environmental response record with spatiotemporal consistency. For each germplasm environment response record, based on the preset key phenological period division points of Phragmites australis, the germplasm environment response record is extracted and recombined to generate multiple phenological response sub-records corresponding to different phenological stages; An exclusive environmental stress profile is constructed for each of the phenological response sub-records. The environmental stress profile is generated by extracting the extreme values ​​and durations of environmental factors that exceed the preset tolerance range of the corresponding phenological period of Phragmites australis from the phenological response sub-records. By linking all environmental stress profiles of the same Arundo donax germplasm at all test sites, arranged in phenological order, an environmental adaptation pathway map of the Arundo donax germplasm across regions and throughout its entire growth period is formed. The environmental adaptability path map is input into a pre-trained multidimensional adaptive discriminant network for feature parsing. The multidimensional adaptive discriminant network outputs classification labels that characterize the adaptability level of Arundo donax germplasm under different environmental stress modes. The environmental adaptability evaluation of Arundo donax germplasm resources is completed based on the classification labels. The construction of a unique environmental stress profile for each of the phenological response sub-records includes: For a phenological response sub-record, the preset tolerance range of the Arundinaria germplasm to various environmental factors in the current phenological stage is retrieved from the Arundinaria germplasm characteristics database. The preset tolerance range includes an upper threshold and a lower threshold. The environmental factor sequence data in the phenological response sub-record are analyzed one by one to identify all data points in the sequence that exceed the corresponding upper limit threshold or fall below the corresponding lower limit threshold, and the environmental factor value at each time the threshold is exceeded and the number of observation cycles for each exceeding event are recorded. By accumulating the values ​​and durations of the same environmental factor in all out-of-event events, and combining the statistical distribution characteristics of the environmental factor throughout the entire time period of the phenological response sub-record, a sub-profile is generated to describe the degree of stress of the environmental factor on Reed truncata during this stage. The sub-portraits of all analyzed environmental factors in the phenological stage are aggregated to form a comprehensive environmental pressure profile specific to this phenological response sub-record, and the environmental pressure profile is stored in a structured data format. The environmental stress profiles of the same *Arundinaria mirifica* germplasm, arranged sequentially by phenological stage at all test sites, form an environmental adaptation pathway map of the *Arundinaria mirifica* germplasm across regions and throughout its entire growth period, including: Using the germplasm identifier of Phragmites australis as an index, the database was used to retrieve the complete environmental stress profiles of the Phragmites australis germplasm produced by each independent ecological test site. The independent ecological test sites are sorted according to their geographical or climatic characteristics, and within each test site, the environmental pressure profiles of multiple phenological stages are arranged in chronological order of the occurrence of phenological events. The sorted environmental stress profiles from all test points are logically linked and integrated according to the order of test points and phenological stages to construct a multi-dimensional data graph structure, which is the environmental adaptability path graph. Each node represents an environmental stress profile of a phenological stage, and the connecting lines represent spatiotemporal or phenological sequence relationships.

2. The method for evaluating the environmental adaptability of Phragmites australis germplasm resources according to claim 1, characterized in that, The acquisition of periodic environmental sequence data and synchronous phenotypic development data of *Arundinaria lobata* germplasm cultivation experimental areas arranged at multiple independent ecological testing sites includes: A uniform observation time granularity is set, and environmental parameters including temperature, humidity, light intensity, soil moisture content, and soil salinity are continuously collected at each independent ecological test point through a deployed sensor array. The data are packaged according to the observation time granularity and labeled with the observation period to form the periodic environmental sequence data. Under the same observation time granularity, the plant height, stem diameter, number of leaves, and biomass increment of Arundo donax are recorded by image acquisition device or manual measurement, and the recorded data are bound with the same observation period identifier to form the Arundo donax phenotypic development data; A central data pool is established to collect and store the periodic environmental sequence data with observation period identifiers from all independent ecological test sites, along with the Phragmites australis phenotypic development data.

3. The method for evaluating the environmental adaptability of Phragmites australis germplasm resources according to claim 2, characterized in that, For each germplasm environmental response record, based on the preset key phenological period division points of *Arundinaria zebrina*, the germplasm environmental response record is truncated and recombined to generate multiple phenological response sub-records corresponding to different phenological stages, including: The complete growth period of the evaluated Phragmites australis germplasm is defined, and multiple key phenological period division points are preset within the complete growth period. The key phenological period division points include the seedling stage, tillering stage, jointing stage, and maturity stage. Read a germplasm environment response record and, based on the time information contained in the record, locate the observation cycle identifier corresponding to each preset key phenological period division point in the germplasm environment response record; Using the observation cycle markers corresponding to two adjacent key phenological period division points as the start and end boundaries, all environmental sequence data and Reed phenotypic development data within the corresponding time period are extracted from the germplasm environmental response record to form a phenological response sub-record; Traverse all preset key phenological period division intervals, complete the extraction and recombination of a single germplasm environmental response record, and generate multiple phenological response sub-records arranged in chronological order for a single experiment of the same Reed germplasm at the same test point.

4. The method for evaluating the environmental adaptability of Phragmites australis germplasm resources according to claim 3, characterized in that, The multidimensional adaptive discriminant network is a pre-trained machine learning model, and its pre-training process includes: We collected environmental adaptation pathway maps of various plant germplasms with known final adaptability that had been evaluated in the past as training samples, and labeled each training sample with its true and comprehensive environmental adaptability level label. An initial network model with a multi-layer nonlinear transformation structure is constructed, and the multi-dimensional data graph structure of the environmental adaptation path graph is transformed into feature vectors, which are used as inputs to the initial network model. The initial network model is iteratively trained using labeled training samples. By adjusting the internal parameters of the network, the difference between the predicted classification label output by the initial network model and the real environment adaptability level label of the training samples is minimized. When the prediction accuracy reaches the preset standard, training stops, and the multidimensional adaptive discriminant network that can be used to evaluate new Arundinaria germplasm is obtained.

5. The method for evaluating the environmental adaptability of Phragmites australis germplasm resources according to claim 4, characterized in that, The step of inputting the environmental adaptive path map into a pre-trained multidimensional adaptive discriminant network for feature parsing includes: The environmental adaptability path graph is normalized to ensure that its dimensions are consistent with the input dimensions required by the multidimensional adaptive discriminant network. The normalized environmental adaptability path map data is input into the input layer of the multidimensional adaptive discriminant network; The multidimensional adaptive discriminant network extracts and abstracts features from the input data layer by layer through multiple hidden layers, and finally generates a probability distribution vector in the output layer. Each element in the probability distribution vector corresponds to a preset fitness level. The fitness level corresponding to the element with the highest probability value in the probability distribution vector is selected as the classification label output by the multidimensional adaptive discriminant network.

6. The method for evaluating the environmental adaptability of Phragmites australis germplasm resources according to claim 5, characterized in that, After evaluating the environmental adaptability of *Arundinaria mirifica* germplasm resources based on the aforementioned classification labels, the process also includes verifying the evaluation results and updating the database. The germplasm identifiers of Phragmites australis generated in this evaluation, their corresponding environmental adaptability pathway maps, and the classification labels output by the multidimensional adaptability discrimination network are packaged together to generate a germplasm adaptability evaluation file. The germplasm adaptability evaluation archive is compared with the historical evaluation records of the Reed germplasm in the germplasm resource database. If the classification label of this evaluation is consistent with the historical mainstream evaluation conclusion, it is directly archived; if there is a significant inconsistency, a manual review is triggered. For evaluation files that trigger manual review, an expert system or human will conduct a secondary analysis based on the original environmental stress profile and phenological response sub-records, and update the final adaptability conclusions after review and confirmation to the germplasm resource database. At the same time, the complete file of this evaluation will be used as a new training sample for subsequent incremental training of the multidimensional adaptability discrimination network.

7. The method for evaluating the environmental adaptability of Phragmites australis germplasm resources according to claim 3, characterized in that, The observation period markers corresponding to two adjacent key phenological period division points are used as start and end boundaries. All environmental sequence data and *Arundinaria lobata* phenotypic development data within the corresponding time period are extracted from the germplasm environmental response record to form a phenological response sub-record, including: Identify the phenological stage that needs to be addressed, wherein the phenological stage is defined by two adjacent critical phenological period dividing points; In the timeline of the germplasm environment response record, locate the starting observation period marker corresponding to the starting key phenological period division point, and the ending observation period marker corresponding to the ending key phenological period division point. Extract all data entries whose observation period identifier is greater than or equal to the start observation period identifier and less than or equal to the end observation period identifier from the data structure stored in the germplasm environment response record. The extracted data entries were categorized and organized into two types: environmental sequence data and Reed phenotypic development data, ensuring that each data entry contained complete observation period identifiers, environmental parameter values, and Reed phenotypic measurements. The categorized and organized data is packaged into a new independent data set, which is a phenological response sub-record corresponding to a specific phenological stage.

8. The method for evaluating the environmental adaptability of Phragmites australis germplasm resources according to claim 4, characterized in that, The step of analyzing the environmental factor sequence data in the phenological response sub-record one by one, identifying all data points in the sequence that exceed the corresponding upper limit threshold or fall below the corresponding lower limit threshold, and recording the environmental factor value at each time it exceeds the threshold and the number of observation periods for each exceeding event includes: Read all time-series data points of the first environmental factor sequentially from the current phenological response sub-record; The value of each data point is compared with the preset upper limit threshold and preset lower limit threshold of the environmental factor in the current phenological stage; When the value of a data point is detected to be greater than a preset upper threshold or less than a preset lower threshold, the data point is marked as the starting point of an abnormal event, and the specific value of the environmental factor when the abnormal event occurs is recorded. Continue scanning the time series data points until a data point whose value recovers to the preset tolerance range is encountered, and mark it as the end point of the abnormal event; Calculate the number of consecutive observation periods between the start point and the end point, and use this as the number of continuous observation periods for this out-of-bounds event; The environmental factor values, anomaly start point location, anomaly end point location, and number of continuous observation periods for this event will be stored as a complete record. Repeat the process until all time-series data points of the current environmental factor in the phenological response sub-record have been traversed and all out-of-series events have been identified; Next, the time series data of the next environmental factor is read sequentially, and the above identification and recording steps are repeated until the analysis of all environmental factor sequence data in the phenological response sub-record is completed.

Images