Directional efficient cultivation method for dendrocalamus latiflorus and leaf dual-purpose forest

Abstract

The invention provides a dendrocalamus latiflorus and leaf dual-purpose forest oriented efficient cultivation method, and relates to the technical field of economic forest cultivation, and the method comprises the steps: obtaining record data of a dendrocalamus latiflorus and leaf dual-purpose forest cultivation process in a historical period, including manual processing data, soil physical and chemical factors and growth state data; the manual processing data comprises hook tip data and / or fertilization data; the growth state data comprises bamboo shoot quality and bamboo leaf quantity; taking each variable in the recorded data as a node, and training a Bayesian network model; by setting cultivation scenes of different hook tip data and fertilization data, inputting the cultivation scenes into the trained Bayesian network, and outputting prediction data of bamboo shoot quality and bamboo leaf quantity; and according to the prediction data, determining an optimal cultivation scene simultaneously meeting the requirements of bamboo shoot quality and bamboo leaf quantity. The method solves the problems that in the prior art, the complex relation between coprocessing of hook tips and fertilization and the growth state of dendrocalamus latiflorus leaves is difficult to quantify, the optimal hook tip strength and the fertilization level cannot be determined, and the cultivation effect is poor.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of economic forest cultivation, in particular to a directional and efficient cultivation method of Dendrocalamus latiflorus shoot and leaf dual-purpose forest. BACKGROUND

[0002] As an important shoot and leaf dual-purpose economic bamboo species in southern China, Dendrocalamus latiflorus can supply high-quality food materials to the market, and its leaves are widely used in food packaging, medicine extraction, and feed processing, etc., and have both ecological value and high economic value. Promoting the efficient cultivation of Dendrocalamus latiflorus shoot and leaf dual-purpose forest has become one of the key paths for the industrial upgrading and income increase of forest farmers in southern bamboo areas.

[0003] Currently, in the cultivation process of Dendrocalamus latiflorus shoot and leaf dual-purpose forest, pruning and fertilization are the core human intervention measures for regulating forest growth and balancing shoot and leaf output. Pruning can adjust the bamboo crown structure, reduce nutrient consumption, and enhance the ventilation and light penetration of the forest stand, which can not only prevent bamboo from falling, but also promote the improvement of bamboo leaf photosynthetic efficiency and bamboo shoot nutrient accumulation. Fertilization can supplement soil nitrogen, phosphorus, potassium and trace elements, improve soil fertility, directly affect the yield and quality of bamboo shoots, and provide basic nutrients for the growth of bamboo leaves.

[0004] However, the existing technology still cannot achieve the goal of efficient cultivation. On the one hand, existing researches mostly focus on the influence of single pruning intensity or single fertilization level on single growth index of shoots or leaves, such as exploring the effect of a certain pruning height on bamboo leaf quantity or the effect of a certain fertilization amount on the yield of bamboo shoots, but ignoring the synergistic effect of pruning and fertilization, as well as the dynamic changes of soil physical and chemical factors such as soil organic matter content, available nutrient concentration, and pH value under the combined action of pruning and fertilization. On the other hand, the relationship between pruning intensity, fertilization level, soil physical and chemical factors, shoot quality, and bamboo leaf quantity presents high nonlinearity and uncertainty, and the existing technology lacks effective means to quantify the complex correlation, and cannot accurately analyze the regulation mechanism of the combined treatment of pruning and fertilization on the dual output of shoots and leaves. SUMMARY

[0005] The embodiments of the present application provide a directional and efficient cultivation method of Dendrocalamus latiflorus shoot and leaf dual-purpose forest, which aims to solve the problem that the existing technology cannot quantify the complex relationship between the combined treatment of pruning and fertilization and the growth state of Dendrocalamus latiflorus shoot and leaf dual-purpose forest, resulting in the inability to determine the optimal pruning intensity and fertilization level, and poor cultivation effect.

[0006] In order to achieve the above-mentioned purpose, the present application provides a directional and efficient cultivation method of Dendrocalamus latiflorus shoot and leaf dual-purpose forest, comprising the following steps: obtaining recorded data of the cultivation process of Dendrocalamus latiflorus shoot and leaf dual-purpose forest in the historical period, including human intervention data, soil physical and chemical factor data, and growth state data; the human intervention data includes pruning data and / or fertilization data; the growth state data includes shoot quality and bamboo leaf quantity; Take each variable in the record data as a node, train a Bayesian network model, and the edges of the Bayesian network model include: an edge taking the human treatment data as a parent node and the soil physicochemical factor as a child node; an edge taking the human treatment data as a parent node and the growth state data as a child node; an edge taking the soil physicochemical factor as a parent node and the growth state data as a child node; By setting different hooking data and fertilization data cultivation scenarios, the trained Bayesian network is input, and prediction data of bamboo shoot quality and bamboo leaf amount are output; According to the prediction data, the best cultivation scenario that meets the requirements of bamboo shoot quality and bamboo leaf amount at the same time is determined.

[0007] Further, the hooking data includes hooking strength and hooking time; and the fertilization data includes fertilization type and fertilization amount.

[0008] Further, the soil physicochemical factor includes ammonia nitrogen, nitric nitrogen, nitrous nitrogen, pH, organic carbon, inorganic carbon and / or total phosphorus.

[0009] Further, before training the Bayesian network model, the nodes are also screened, including the following steps: Taking the human treatment data and the soil physicochemical factor as explanatory variables and the growth state data as response variables, a random forest model is constructed. The importance of the explanatory variables is sorted by the random forest model, and the last N variables in the sorting are removed.

[0010] Further, the cultivation scenario uses an orthogonal method to set the hooking data and the fertilization data.

[0011] Further, in the training process of the Bayesian network, an interaction term between the hooking data and the fertilization data is also introduced.

[0012] Further, the bamboo shoot quality is comprehensively represented by bamboo shoot weight, bamboo shoot length, bamboo shoot diameter and / or bamboo shoot color.

[0013] Further, the Bayesian network model of different altitudes and / or different latitudes is also constructed to obtain the best cultivation scenario in different regions.

[0014] The above technical solutions have the following technical effects: The application obtains record data of the cultivation process of the bamboo shoot and leaf dual-purpose forest in the historical period, including human treatment data, soil physical and chemical factors and growth state data; the human treatment data includes pruning data and / or fertilization data; the growth state data includes shoot quality and bamboo leaf amount; each variable in the record data is taken as a node to train a Bayesian network model; different cultivation scenarios of the pruning data and the fertilization data are set, and the trained Bayesian network is inputted, and prediction data of the shoot quality and the bamboo leaf amount are outputted; and the best cultivation scenario meeting the requirements of the shoot quality and the bamboo leaf amount is determined according to the prediction data. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 A flowchart of the directional and efficient cultivation method of the bamboo shoot and leaf dual-purpose forest according to an embodiment of the application is shown in Figure 2 A structure diagram of the Bayesian network model according to an embodiment of the application is shown in DETAILED DESCRIPTION

[0016] To further illustrate the embodiments, the application provides accompanying drawings. These drawings are part of the disclosure of the application, mainly used to illustrate the embodiments, and can be used to explain the operating principle of the embodiments in conjunction with the related description of the specification. Those of ordinary skill in the art should be able to understand other possible implementations and advantages of the application by referring to these contents. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0017] The application will be further described in conjunction with the accompanying drawings and specific embodiments.

[0018] Embodiment one: Figure 1 A flowchart of the directional and efficient cultivation method of the bamboo shoot and leaf dual-purpose forest according to an embodiment of the application is shown in Figure 1 As shown in the figure, the method of this embodiment includes the following steps: The application obtains record data of the cultivation process of the bamboo shoot and leaf dual-purpose forest in the historical period, including human treatment data, soil physical and chemical factors and growth state data; the human treatment data includes pruning data and / or fertilization data; the growth state data includes shoot quality and bamboo leaf amount; In a specific implementation, the pruning data includes pruning intensity and pruning time; the fertilization data includes fertilization type and fertilization amount. The soil physical and chemical factors include ammonia nitrogen, nitric acid nitrogen, nitrous acid nitrogen, pH, organic carbon, inorganic carbon and / or total phosphorus. The shoot quality is comprehensively characterized by shoot weight, shoot length, shoot diameter and / or shoot color.

[0019] In the embodiment, the artificial treatment data specifically includes two types of data of hooking and fertilization, the hooking data includes hooking strength such as the proportion of hooking off bamboo branches, the reserved bamboo crown height and the like, directly affects the bamboo crown photosynthetic efficiency and nutrient allocation and hooking time such as key nodes after bamboo shoot harvesting, before bamboo leaf germination and the like, and is related to the timing matching degree of nutrient supply; the fertilization data is specifically the type of fertilization such as nitrogen fertilizer, phosphorus fertilizer, compound fertilizer and the like, determines the type of nutrient supplement and the fertilization amount such as the specific application amount of nitrogen, phosphorus and potassium per unit area, and affects the soil nutrient supply strength.

[0020] As the key intermediate carrier connecting the artificial treatment and the growth state, the soil physicochemical factors are selected to reflect the soil available nitrogen content, directly relate to the nitrogen absorption of Dendrocalamus latiflorus, the pH affects the soil nutrient availability and microbial activity, the organic carbon and inorganic carbon represent the soil carbon pool level, relate to the nutrient storage capacity, the total phosphorus reflects the soil phosphorus supply basis, and captures the core changes of the soil environment after the artificial treatment. In the growth state data, the shoot quality is not a single index, but is comprehensively represented by quantifiable indexes such as shoot weight (measuring single shoot yield), shoot length and diameter (reflecting shoot size), shoot color (reflecting shoot freshness and commodity nature) and the like, combined with bamboo leaf amount such as unit area bamboo leaf fresh weight or dry weight, and comprehensively reflects the cultivation effect of Dendrocalamus latiflorus shoot and leaf double output.

[0021] Each variable in the recorded data is taken as a node to train the Bayesian network model. Figure 2 The structural diagram of the Bayesian network model in the embodiment of the application is shown in FIG. 1. Figure 2 As shown in FIG. 1, the edges of the Bayesian network model include: the edge taking the artificial treatment data as the parent node and the soil physicochemical factor as the child node; the edge taking the artificial treatment data as the parent node and the growth state data as the child node; the edge taking the soil physicochemical factor as the parent node and the growth state data as the child node; In the model construction stage, in the embodiment, each key variable in the historical record data is set as an independent node of the Bayesian network model, and the model learns and quantifies the dependency between nodes through data-driven training, instead of simply stacking variables. The edges of the Bayesian network (i.e. the causal relationship between nodes) are clearly divided into three categories: one is the edge with human treatment data as the parent node and soil physicochemical factors as the child node, which reflects the direct regulation of human measures such as pruning and fertilization on the soil environment, for example, the change of fertilizer amount will directly change the content of soil ammonia nitrogen and total phosphorus, and the selection of pruning time will affect the accumulation of soil organic carbon; the second is the edge with human treatment data as the parent node and growth state data as the child node, which reflects the direct influence of human measures on the growth of shoots and leaves, for example, reasonable pruning intensity can directly optimize the ventilation and light transmission of the bamboo crown and improve the amount of bamboo leaves; the third is the edge with soil physicochemical factors as the parent node and growth state data as the child node, which reflects the intermediate bridge role of the soil environment, for example, the standard soil nitric acid nitrogen content can provide sufficient nitrogen for bamboo shoots, thereby improving the shoot weight and shoot diameter.

[0022] By setting different pruning data and fertilization data cultivation scenarios, the trained Bayesian network is inputted, and the prediction data of shoot quality and bamboo leaf amount is outputted; In a specific implementation, the cultivation scenarios are set by using an orthogonal method for pruning data and fertilization data.

[0023] In the model application stage, by constructing multiple groups of pruning and fertilization collaborative treatment cultivation scenarios, the trained Bayesian network is inputted, and the prediction data of shoot quality and bamboo leaf amount under each group of scenarios can be directly outputted, realizing the quantitative prediction of the effect of different cultivation schemes. The cultivation scenarios are set by using an orthogonal method, without the need for comprehensive combination of all levels of pruning data and fertilization data, but by scientifically screening representative combinations through an orthogonal table. For example, if the pruning data includes 2 intensity levels and 2 time nodes, and the fertilization data includes 2 fertilizer types and 2 application levels, comprehensive combination requires 16 scenarios, while the orthogonal method only needs 4 groups to cover different levels of parameters, and can uniformly reflect the interaction between parameters. This setting method not only avoids the increase in operation cost caused by redundant scenarios, but also ensures that the input model scenarios can completely cover the core combination dimensions of “pruning intensity-pruning time-fertilizer type-fertilization amount”, effectively reducing the risk of missing the best scheme. After inputting the model through the orthogonal scenarios, the prediction data of shoot quality and bamboo leaf amount will have more reference value, and can clearly present the influence difference of different pruning and fertilization combinations on double output.

[0024] According to the prediction data, the best cultivation scenario that meets the requirements of shoot quality and bamboo leaf amount is determined.

[0025] In a specific implementation, before training the Bayesian network model, the nodes are also screened, including the following steps: A random forest model is constructed by taking human processed data and soil physical and chemical factors as explanatory variables and growth state data as a response variable. The importance of the explanatory variables is sorted by the random forest model, and the last N variables are removed.

[0026] In this embodiment, to improve the training efficiency and prediction accuracy of the Bayesian network model and avoid the interference of redundant variables, a node screening step is also added before training the model. The importance of variables is determined by the random forest model. The specific process is as follows: first, taking human processed data and soil physical and chemical factors as explanatory variables and growth state data as a response variable, a random forest model is constructed to learn the actual influence of the two types of explanatory variables on growth effect; then, the importance scores of each explanatory variable are calculated and output by the random forest model, and the scores are sorted from high to low, and finally the last N variables are removed, N is set according to the data size and model requirements, such as removing the last 20% of variables with low importance. For example, if the influence score of inorganic carbon in soil physical and chemical factors on bamboo shoot quality and bamboo leaf amount is very low, it will be included in the removal range, and only the core variables such as ammonia nitrogen, fertilizer amount, and hooking strength are retained as the nodes of the Bayesian network, which not only reduces the operation cost of model training, but also avoids the prediction deviation caused by irrelevant variables, and lays a foundation for accurate training of the Bayesian network model.

[0027] In a specific implementation, an interaction term between hooking data and fertilization data is introduced in the Bayesian network training process.

[0028] In this embodiment, an interaction term between hooking data and fertilization data is introduced in the Bayesian network training process, rather than simply inputting the two as independent variables into the model. The interaction term can quantify the additive effect of the combined action of the two. After introducing the interaction term, the Bayesian network can not only learn the independent influence of hooking and fertilization on soil physical and chemical factors and growth state, but also capture the dynamic correlation of the combination of the two, making the model more consistent with the logic of the interaction and co-regulation of hooking and fertilization in cultivation practice, thereby improving the accuracy of the prediction data of bamboo shoot quality and bamboo leaf amount when inputting different cultivation scenarios, and avoiding misjudgment of the best cultivation scenario due to ignoring the interaction effect.

[0029] In a specific implementation, Bayesian network models for different altitudes and / or different latitudes are also constructed to obtain the best cultivation scenario in different regions.

[0030] In this embodiment, further considering the significant influence of geographical environment differences such as altitude and latitude on the growth of bamboo shoot and leaf dual-purpose forest, by constructing a Bayesian network model specific to different altitudes and / or different latitudes, the regional adaptation of the optimal cultivation scenario is realized. Due to the difference in altitude, the climate conditions such as temperature, light and precipitation will be directly changed, and the difference in latitude will affect the accumulated temperature and seasonal rhythm, which will lead to completely different effects of the same hooking and fertilization scheme in different regions. For example, a high-fertilization scheme in low-latitude areas may be wasted due to slow nutrient absorption, and the appropriate hooking strength in high-altitude areas may be insufficient in low-latitude areas.

[0031] Therefore, by training the model for different regions respectively and inputting the historical cultivation data corresponding to the region, the model can accurately learn the unique association logic under the regional environment. When the cultivation scenario is input later, the prediction data of bamboo shoot quality and bamboo leaf quantity adapted to the region can also be output, and the finally determined optimal cultivation scenario is more suitable for the local actual situation.

[0032] Although the present application is specifically shown and introduced in combination with the preferred embodiments, it should be understood by those skilled in the art that various changes can be made in form and details without departing from the spirit and scope of the present application as defined in the appended claims, and all such changes are within the protection scope of the present application.

Claims

1. A method for high-efficiency and directional cultivation of Dendrocalamus latiflorus and its leaves, characterized in that, Includes the following steps: Recorded data on the cultivation process of bamboo shoot and leaf dual-purpose forests during historical periods were obtained, including human intervention data, soil physicochemical factors, and growth status data; the human intervention data included shoot hooking data and / or fertilization data; the growth status data included shoot quality and bamboo leaf quantity. Using each variable in the recorded data as a node, a Bayesian network model is trained; the edges of the Bayesian network model include: The human-processed data is taken as the parent node, and the soil physicochemical factors are the edges of the child nodes; The human-processed data is taken as the parent node, and the growth state data is taken as the edges of the child nodes; The soil physicochemical factors are used as parent nodes, and the growth state data are used as edges of child nodes; By setting different cultivation scenarios with different hooking data and fertilization data, the trained Bayesian network is input into the network to output predicted data on bamboo shoot quality and bamboo leaf quantity. Based on the predicted data, the optimal cultivation scenario that simultaneously meets the requirements for bamboo shoot quality and bamboo leaf quantity is determined.

2. The method according to claim 1, wherein the bamboo shoot and leaf dual-purpose forest is Dendrocalamus latiflorus. The hooking data includes hooking intensity and hooking time; the fertilization data includes fertilizer type and fertilizer amount. ​ 3. The method according to claim 1, wherein the method is characterized by the steps of: The soil physicochemical factors include ammonia nitrogen, nitrate nitrogen, nitrite nitrogen, pH, organic carbon, inorganic carbon, and / or total phosphorus.

4. The method for directional and efficient cultivation of bamboo shoot and leaf dual-purpose forests according to claim 1, characterized in that, Before training the Bayesian network model, nodes are also screened, including the following steps: A random forest model was constructed using human-processed data and soil physicochemical factors as explanatory variables and growth status data as response variables. The importance of explanatory variables was ranked using a random forest model, and the last N variables in the ranking were removed.

5. The method for directional and efficient cultivation of bamboo shoot and leaf dual-purpose forests according to claim 1, characterized in that, The cultivation scenario was set using an orthogonal method for the hooking data and fertilization data.

6. The method for directional and efficient cultivation of bamboo shoot and leaf dual-purpose forests according to claim 1, characterized in that, The Bayesian network also incorporates interaction terms between hooking data and fertilization data during training.

7. The method for directional and efficient cultivation of bamboo shoot and leaf dual-purpose forests according to claim 1, characterized in that, The quality of the bamboo shoots is comprehensively characterized by their weight, length, diameter, and / or color.

8. The method for directional and efficient cultivation of bamboo shoot and leaf dual-purpose forests according to claim 1, characterized in that, Furthermore, Bayesian network models were constructed for regions with different altitudes and / or latitudes to obtain the optimal breeding scenarios for different regions.

Images