An ecological evolution simulation method based on rainfall runoff assimilation

Abstract

The application discloses an ecological evolution simulation method based on rainfall runoff assimilation, and comprises the following steps: estimating environmental effects such as topographic erosion, water resource distribution change and light and heat resource change by using rainfall runoff assimilation results output by a rainfall runoff assimilation model; and calculating resource supply conditions and species distribution conditions of each layer of a food chain to simulate changes of a target environment ecological circle on the basis. The application simulates rainfall runoff processes by using the rainfall runoff assimilation model, adds short-term influences of rainfall runoff in ecological environment evolution simulation, improves granularity of the ecological evolution simulation method, and promotes ecological coupling rainfall runoff assimilation to be more accurate.

Description

TECHNICAL FIELD

[0001] The present application provides a rainfall runoff assimilation-based ecological evolution simulation method, relating to the technical field of deep learning, the technical field of rainfall runoff assimilation and the technical field of ecological environment evolution simulation. BACKGROUND

[0002] Traditional ecological environment evolution simulation technology mainly considers the influence of the atmosphere on the change of ecological environment from the perspective of climate, and has a low granularity in simulating ecological environment, which is suitable for simulating long-term ecological environment change. Rainfall runoff process has high nonlinearity and non-stationarity, and can change the light and heat resources, water resources and terrain of a region in a short period of time, and has a great influence on the change of ecological environment. Sometimes, this influence shapes a small climate, and makes the ecological environment also change. Therefore, the present application combines rainfall runoff assimilation model and ecological environment evolution, simulates rainfall runoff process, improves the granularity of the ecological simulation method, and promotes the ecological coupling rainfall runoff assimilation to be more accurate.

[0003] Traditional methods mainly rely on process-driven hydrological models, set some parameters with physical significance according to hydrological processes, and establish complex algorithms to link the parameters according to existing knowledge to build hydrological models. In application, the model parameters are adjusted manually according to actual observation data to fit the actual situation. However, due to the high complexity and many parameters of the model, the process-driven hydrological model has some deficiencies in practical application.

[0004] Data-driven methods, such as regression analysis and time series analysis, can predict rainfall runoff by capturing numerical trends, which can reveal the rules of historical data to a certain extent, but the prediction accuracy and stability are limited for complex hydro-meteorological systems. In recent years, with the progress of machine learning technology, traditional machine learning and emerging methods such as deep learning have been applied to rainfall runoff assimilation. These methods can better capture the nonlinear characteristics of hydro-meteorological systems, thereby improving the prediction accuracy and stability. However, for complex runoff prediction, the prediction accuracy of these technologies has declined seriously. The present application introduces an input transposition mechanism to make the model focus more on the sequence characteristics of the variable itself, fully utilize the modeling capability of Transformer for nonlinear runoff sequence, and improve the learning ability and prediction accuracy of the model for complex data patterns of runoff sequence. SUMMARY

[0005] The present application aims to provide a rainfall runoff assimilation-based ecological evolution simulation method, which simulates rainfall-runoff process by combining rainfall runoff assimilation model, improves the fine granularity of ecological simulation method, and fills the gap in the prior art.

[0006] Specifically, the present application provides a rainfall runoff assimilation-based ecological evolution simulation method, comprising the following steps:

[0007] Step 1: Input the basic geographic data of the simulated region; provide a K-layer food chain with plants as the first layer; for each layer of the food chain, there will be a corresponding Resource Supply Map (RSG) and Species Density Map (SDM);

[0008] Step 2: Input the full-node hydrological dataset of the simulated regional river basin with a time scale of days and a time span of not less than 1 year, and use a neural network model for rainfall-runoff assimilation training;

[0009] Step 3: Combine latitude and longitude, altitude, sea level extreme temperature and rainfall-runoff assimilation model to determine the climate characteristics of the target environment and estimate its light and heat resources;

[0010] Step 4: Combine the digital elevation model, topographic orientation map, and mineral map to add the topography, runoff, and watershed of the target environment and initialize the target environment;

[0011] Step 5: Based on the initial topography, runoff and watershed, and light and heat resources of the target environment, construct the first layer of the RSG. 0 And so on, constructing subsequent layers to obtain the initial RSG; for subsequent layers RSG 1 ~RSG K-1 Iterative use of RSG i-1 Building RSG i ;

[0012] Step 6: Continuously simulate and iterate on an annual scale, including terrain, SDM, and RSG; finally output the latest terrain, SDM, and RSG.

[0013] Furthermore, the basic geographic data mentioned in step 1 includes the digital elevation model of the target environment, topographic orientation map, mineral map, latitude and longitude, altitude, and extreme sea-level temperature.

[0014] Furthermore, the specific steps of step 2 are as follows:

[0015] Step 2.1: Input the full-node hydrological dataset of the river basin in the simulated region with a time scale of days and a time span of not less than 1 year. The data of each node consists of rainfall and runoff.

[0016] Step 2.2: Construct a Transformer model with transposed input as a rainfall-runoff assimilation model; the rainfall-runoff assimilation model is constructed through input... and predict and Where P represents rainfall, Q represents runoff, i represents the date to be predicted, j represents the node to be predicted, and j-1 represents the upstream node of j; the rainfall-runoff assimilation model achieves rainfall-runoff assimilation through the method of reusing predicted values;

[0017] Step 2.3: Set the loss function and optimizer, select suitable training hyperparameters, and train the rainfall-runoff assimilation model.

[0018] Furthermore, in step 2.2, the Transformer model with transposed input combines the values ​​of the same variable at different time steps into a vector for embedding operation at the embedding layer. Specifically, for an input variable X with N time steps, each time step recording M features, and its shape being N×M, the calculation formula for transposed embedding is: in This represents a fully connected layer with an input dimension of N and an output dimension of D.

[0019] Furthermore, the hidden layers of the Transformer model with transposed input described in step 2.2 are implemented by stacking multi-head attention encoders of the Transformer model, and the forward propagation calculation steps of each encoder are as follows:

[0020] Step T1: For input X, calculate the output X of the multi-head attention module. attn :

[0021] X attn =MultiHead(Q,K,V)+X=Concat(head l , ..., head h W O +X,

[0022] in

[0023] Step T2, for the multi-head attention module output X attn Calculate the output X of the feedforward module FFN :X FFN =max(0, X) attn W1+b2)W2+b2+X attn X FFN This is the result of the encoder's forward propagation;

[0024] In the formula, Concat(*, **) represents tensor concatenation operation, softmax(*) represents the softmax function, and max(*, **) is the maximum value function.

[0025] Furthermore, the specific steps of step 3 are as follows:

[0026] Step 3.1: Obtain runoff observation data with a time span of one year from the rainfall-runoff assimilation model described in Step 2, accumulate the rainfall at each node on the same date, and statistically analyze annual rainfall characteristics such as total annual rainfall, number of rainy days per year, maximum monthly rainfall, and minimum monthly rainfall.

[0027] Step 3.2: Based on existing geographical knowledge, combined with the annual rainfall characteristics, latitude and longitude, altitude, and extreme sea-level temperatures, determine the climate characteristics of the target environment and estimate the total annual solar thermal resources (hereinafter referred to as solar thermal resources).

[0028] Furthermore, the specific steps of step 5 are as follows:

[0029] Step 5.1: Based on the topography, runoff and watershed, and light and heat resources of the initial target environment, construct the first layer of the RSG. 0 RSG 0 Represent each cell on the map as a node on the map;

[0030] Step 5.2, according to RSG i-1 Building RSG i Until RSG is completed 1 ~RSG K-1 The construction process involves executing steps 5.3.1 to 5.3.4 each time.

[0031] Step 5.3.1: Encode the resources of each species at the i-th layer of the food chain, their location on the terrain, and their accessibility;

[0032] Step 5.3.2: Connect multiple nodes into local regions, calculate the fitness score of species at level i of the food chain in each local region, and simulate the competition between species at level i based on the fitness score.

[0033] Step 5.3.3: Based on the simulation results, set the population unit size for each region to obtain the i-th layer SDM. i ;

[0034] Step 5.3.4, from RSG i-1 and SDM i Calculate the remaining graph of unused resources And create RSG i .

[0035] 6. Further, the specific steps of step 6 are as follows:

[0036] Step 6.1: Obtain one year's rainfall-runoff assimilation data from the rainfall-runoff assimilation model, perform hydrodynamic analysis on the runoff at each node during the flood season in combination with the topography, calculate possible river course changes, and update the topography;

[0037] Step 6.2: Update RSG based on rainfall at each node. 0 The water resource node in the middle completes RSG. 0 Update;

[0038] Step 6.3, according to RSG i-1 Update RSG i Until RSG is fully updated 1 ~RSG K-1 The steps are the same as in step 5.2.

[0039] Step 6.4: Output the final terrain, SDM, and RSG as results.

[0040] All parts not covered in this invention are the same as or can be implemented using existing technologies. Attached Figure Description

[0041] Figure 1 This is a flowchart of the method described in this invention.

[0042] Figure 2 This is the rainfall-runoff assimilation result of the rainfall-runoff assimilation model described in this invention.

[0043] Figure 3 This refers to the RSG described in this invention.

[0044] Figure 4 This describes the relationship between RSG and SDM as described in this invention. Detailed Implementation

[0045] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0046] 1. Embedding mechanism for input transpose

[0047] For an input X with N time steps, each time step recording M features, and a shape of N×M, the time dimension and feature dimension are swapped through a matrix transpose operation to obtain X. T The input is of shape M×N and embedding is performed. The input transpose enables the model to apply self-attention in the time dimension, directly learning the temporal relevance of the input data.

[0048] 2. Self-attention mechanism

[0049] For each input sequence X, multiply by the weight matrix W. Q W K and W V Given the query matrix Q, key matrix K, and value matrix V, calculate QK. T Calculate the attention score and divide by To avoid numerical overflow, the scaled attention score is converted into weights using the softmax function. Multiplying the weights by V yields the self-attention output.

[0050] 3. RSG Composition

[0051] RSG is an atlas containing K images, corresponding to each layer of the food chain. Figure 3 Given a three-layered food chain RSG, with subscripts indicating resource availability for plants, herbivores, and carnivores respectively, from smallest to largest. Taking RSG1 as an example, the black lines delineate the antelope's main activity areas, black arrows represent the antelope's daily activity paths, and gray arrows indicate connectivity of the same resources. RSG1 shows that antelopes have four main activity areas, but due to terrain, they only move between two of them, forming three relatively enclosed activity areas; the largest activity area possesses the largest quantity and most diverse range of resources.

Claims

1. An ecological evolution simulation method based on rainfall-runoff assimilation, belonging to the category of computer graphics and artificial intelligence, characterized in that... Includes the following steps: Step 1: Input the basic geographic data of the simulated region; provide a K-layer food chain with plants as the first layer; for each layer of the food chain, there will be a corresponding Resource Supply Map (RSG) and Species Density Map (SDM); Step 2: Input a full-node hydrological dataset of the simulated regional river basin with a time scale of daily and a time span of no less than 1 year, and use a neural network model for rainfall-runoff assimilation training; Step 3: Combining latitude and longitude, altitude, extreme sea-level temperature, and rainfall-runoff assimilation model, determine the climate characteristics of the target environment and estimate its light and heat resources; Step 4: Combine the digital elevation model, topographic orientation map, and mineral map to add the topography, runoff, and watershed of the target environment and initialize the target environment; Step 5: Based on the initial topography, runoff and watershed, and light and heat resources of the target environment, construct the first layer of the RSG. 0 And so on, constructing subsequent layers to obtain the initial RSG; for subsequent layers RSG 1 ~RSG K-1 Iterative use of RSG i-1 Building RSG i ; Step 6: Continuously simulate and iterate on an annual scale, with iterations including terrain, SDM, and RSG; the output is the latest terrain, SDM, and RSG. Furthermore, the specific steps of step 6 are as follows: Step 6.1: Obtain one year's rainfall-runoff assimilation data from the rainfall-runoff assimilation model, perform hydrodynamic analysis on the runoff at each node during the flood season in combination with the topography, calculate possible river course changes, and update the topography; Step 6.2: Update RSG based on rainfall at each node. 0 The water resource node in the middle completes RSG. 0 Update; Step 6.3, according to RSG i-1 Update RSG i Until RSG is fully updated 1 ~RSG K-1 ; Step 6.4: Output the final terrain, SDM, and RSG as results.

2. The ecological evolution simulation method based on rainfall-runoff assimilation according to claim 1, characterized in that, The basic geographic data mentioned in step 1 includes the digital elevation model of the target environment, topographic orientation map, mineral map, latitude and longitude, altitude and extreme sea-level temperature.

3. The ecological evolution simulation method based on rainfall-runoff assimilation according to claim 1, characterized in that, Step 2 uses a Transformer model with transposed input as the neural network model for training rainfall-runoff assimilation. In the embedding layer, the values ​​of the same variable at different time steps are combined into a vector to perform the embedding operation. The calculation formula is: Where N is the number of time steps, M is the number of features, and X is the input variable. This represents a fully connected layer with an input dimension of N and an output dimension of D.

4. The ecological evolution simulation method based on rainfall-runoff assimilation according to claim 1, characterized in that, In step 5, the competition between species in each local region is simulated by calculating the fitness scores of species at the i-th level of the food chain in local regions, thereby generating the i-th level SDM of the SDM. i The fitness score is calculated as follows: Where s represents the species, R represents the set of necessary resources, and r represents a certain necessary resource. min r max Let be the minimum and maximum values ​​of resource r in a year, respectively, and let F(s, r) represent the demand function of species s for resource r.

Images