Estimation device and estimation method

Abstract

An estimation device 1 comprises: a model generation unit 10 that, in each time step, calculates an OBM value including the growth rate of marine microorganisms using an ocean physical value calculated from an ocean physical model and an ocean biogeochemistry model (OBM), and trains a learning model for estimating the growth rate of marine microorganisms in each time step from the relationship between the ocean physical value and the OBM value in each time step; and an estimation unit 20 that, in each time step, calculates an OBM value other than the growth rate of the marine microorganism using the ocean physical value calculated from the ocean physical model and the OBM, and uses the learning model to estimate the growth rate of the marine microorganism from the ocean physical value and the OBM value other than the growth rate of the marine microorganism.

Description

Estimation device and estimation method

[0001] The present disclosure relates to an estimation device and an estimation method.

[0002] Earth System Models (ESMs) are important tools for understanding the impact of human activities on Earth's material cycles, including the global carbon cycle, and on marine ecosystems and resources. ESMs include ocean biogeochemistry models (OBMs) as one of their components. OBMs are used to describe the biogeochemical properties of the ocean and how these properties change over time.

[0003] For example, one example of OBM is the Biogeochemistry Model (BLING: Biogeochemistry with Light, Iron, Nutrients and Gas) (Non-Patent Document 1). The model disclosed in Non-Patent Document 1 represents the growth and loss rates of marine microorganisms such as small and large plankton.

[0004] Dunne, John P., et al. "Simple global ocean Biogeochemistry with Light, Iron, Nutrients and Gas version 2 (BLINGv2): Model description and simulation characteristics in GFDL's CM4. 0." Journal of Advances in Modeling Earth Systems 12.10 (2020)

[0005] The growth and loss rates of marine microorganisms in Non-Patent Document 1 are estimated based on ocean physical models and changes in the OBM value. However, the evolutionary rate of marine microorganisms is fast. Therefore, it is preferable to estimate the growth rate of marine microorganisms more accurately, for example, by taking the progression of time into account.

[0006] The present disclosure has been made in consideration of the above circumstances, and an object of the present disclosure is to provide a technology that enables more accurate estimation of the growth rate due to evolution of marine microorganisms.

[0007] An estimation device of one aspect of the present disclosure includes a model generation unit that calculates OBM values, including the growth rate of marine microorganisms, using marine physical values ​​calculated from a marine physical model and an OBM (marine biogeochemical model) at each time step, and learns a learning model that estimates the growth rate of the marine microorganisms at each time step from the relationship between the marine physical values ​​and the OBM values ​​at each time step; and an estimation unit that calculates OBM values ​​other than the growth rate of the marine microorganisms at each time step using the marine physical values ​​and the OBM calculated from the marine physical model, and estimates the growth rate of the marine microorganisms from the marine physical values ​​and the OBM values ​​other than the growth rate of the marine microorganisms using the learning model.

[0008] In one aspect of the estimation method of the present disclosure, a computer calculates OBM values, including the growth rate of marine microorganisms, using marine physical values ​​calculated from an oceanographic model and an OBM (marine biogeochemical model) at each time step, learns a learning model that estimates the growth rate of the marine microorganisms at each time step from the relationship between the marine physical values ​​and the OBM values ​​at each time step, calculates OBM values ​​other than the growth rate of the marine microorganisms at each time step using the marine physical values ​​and OBM calculated from the oceanographic model, and estimates the growth rate of the marine microorganisms from the marine physical values ​​and the OBM values ​​other than the growth rate of the marine microorganisms using the learning model.

[0009] According to the present disclosure, a technique can be provided that enables more accurate estimation of the evolutionary growth rate of marine microorganisms.

[0010] Fig. 1 is a diagram illustrating functional blocks of an estimation device according to an embodiment of the present disclosure. Fig. 2 is a block diagram illustrating the configuration of a model generation unit. Fig. 3 is a block diagram illustrating the configuration of an estimation unit. Fig. 4 is a flowchart illustrating estimation processing by the estimation device. Fig. 5 is a diagram illustrating the hardware configuration of a computer used in the estimation device.

[0011] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. In the description of the drawings, the same parts are designated by the same reference numerals and the description thereof will be omitted.

[0012] (Estimation Device) The estimation device 1 according to the present disclosure more accurately estimates the evolutionary growth rate of marine microorganisms in a marine biogeochemical model (hereinafter referred to as OBM). The estimation device 1 estimates the growth rate or loss rate of marine microorganisms taking into account the passage of time. The estimation device 1 makes it possible to reflect the evolutionary speed of marine microorganisms in the growth rate or loss rate.

[0013] 1 , the estimation device 1 includes the functions of a model generation unit 10 and an estimation unit 20, an input data storage unit 30 that stores input data, learning model data 41, and estimation data 28. Each function is implemented in a CPU 901. Each piece of data is stored in a storage device such as a memory 902 or a storage 903.

[0014] The model generation unit 10 generates a learning model that estimates the growth rate of marine microorganisms at each time step by referring to each data in the input data storage unit 30. The model generation unit 10 outputs learning model data 41 that identifies the generated learning model.

[0015] The estimation unit 20 estimates the growth rate of marine microorganisms at each time step by referring to each data in the input data storage unit 30 and the learning model generated by the model generation unit 10. The estimation unit 20 associates each time step with the growth rate of marine microorganisms at that time step to generate estimated data 28.

[0016] The estimation device 1 according to the present disclosure may generate a learning model that estimates the growth rate for each type of marine microorganism. The growth rate of marine microorganisms is estimated for each time step and type.

[0017] In the present disclosure, the estimation unit 20 may estimate the growth rate of marine microorganisms at a lower resolution than the resolution of the learning model generated by the model generation unit 10. This reduces the processing load on the estimation unit 20 compared to when the growth rate of marine microorganisms is estimated at a higher resolution.

[0018] The estimation device 1 according to the present disclosure generates a learning model for estimating the growth rate of marine microorganisms and estimates the growth rate of marine microorganisms at each time step, but the present disclosure is not limited to this. The estimation device 1 may also estimate the loss rate of marine microorganisms. In the present disclosure, marine microorganisms are phytoplankton present in the ocean.

[0019] The ocean physical model is a model that calculates parameters that specify the physical state of the ocean, such as ocean current velocity vectors and salinity concentrations. The ocean physical values ​​are the values ​​of the parameters that specify the physical state of the ocean. The ocean physical model and the ocean physical values ​​may each include multiple elements.

[0020] The OBM is a model that calculates parameters that specify biogeochemical properties in the ocean, such as concentrations of nutrients or plankton, growth rates of phytoplankton, and parameters that represent iron deficiency. The OBM value is the value of a parameter that specifies biogeochemical properties in the ocean. The OBM and the OBM value may each include multiple elements.

[0021] Generally, the growth rate μ of phytoplankton is expressed by Equation (1): Equation (1) indicates that the growth rate of phytoplankton does not change over time.

[0022]

[0023] In contrast, the growth rate of phytoplankton estimated by the estimation device 1 according to the present disclosure is estimated for each time step, taking into account the evolution of phytoplankton. This allows for more accurate estimation of changes in the growth rate of phytoplankton. Furthermore, the estimation device 1 according to the present disclosure simulates the physical state and biogeochemical state of the ocean using existing models, allowing for flexible estimation.

[0024] (Model Generation Unit) The model generation unit 10 will be described with reference to Fig. 2. The time step in the model generation unit 10 is expressed by a variable n1. N is the maximum value of the time step calculated by the estimation device 1.

[0025] The model generation unit 10 references each piece of data in the input data storage unit 30 to generate learning model data 41. The model generation unit 10 has the functions of a calculation unit 11, an output processing unit 15, and a learning unit 16, as well as output data 18 and learning data 19.

[0026] The input data storage unit 30 includes ocean physics initial values ​​31, ocean physics model data 32, OBM initial values ​​33, and OBM data .

[0027] The oceanographic initial values ​​31 are oceanographic values ​​at time step n1=0.

[0028] The ocean physical model data 32 is data that specifies a model used to calculate ocean physical values ​​in a time step to be processed, where n1>0 and n1<N.

[0029] Examples of ocean physical models include MITgcm (Massachusetts Institute of Technology General Circulation Model), NEMO (Nucleus for European Modelling of the Ocean), MICOM (Miami Isopycnal Cordinate Ocean Model), MPIOM (Max Planck Institute Ocean Model), CSIRO (Commonwealth Scientific and Industrial Research Organisation), and MOM6 (Modular Ocean Model 6).

[0030] The OBM initial value 33 is the OBM value at time step=0.

[0031] The OBM data 34 is data specifying a model used to calculate the OBM value at a time step to be processed, where time step n1>0 and n1<N. In the present disclosure, the OBM data 34 includes data specifying, for example, the calculation formula shown in Equation (1) as a model for calculating the growth rate of phytoplankton.

[0032] Examples of OBMs include BLING, PlankTOM5, HAMOCC (Hamburg Model of the Ocean Carbon Cycle), PISCES, REcom2 (Regulated Ecosystem Model 2), and COBALT.

[0033] The calculation unit 11 calculates the OBM value at each time step through simulation. At each time step, the calculation unit 11 calculates the OBM value, including the growth rate of phytoplankton, using the ocean physical values ​​calculated from the ocean physical model specified by the ocean physical model data 32 and the OBM specified by the OBM data 34. Here, when the time step is 0, the calculation unit 11 refers to the ocean physical initial value 31 and the OBM initial value 33 instead of referring to the ocean physical model data 32 and the OBM data 34.

[0034] The calculation unit 11 includes an oceanographic calculation unit 12 , a coupling unit 13 , and an OBM calculation unit 14 .

[0035] The oceanographic calculation unit 12 calculates oceanographic physical values ​​at the time step to be calculated from the oceanographic physical model specified by the oceanographic model data 32. The combining unit 13 integrates the oceanographic physical values ​​calculated from each oceanographic physical model to specify the ocean state at that time step.

[0036] The OBM calculation unit 14 calculates an OBM value for a time step to be calculated using the ocean state identified by the combining unit 13 and the OBM identified by the OBM data 34. Here, the OBM value includes data on the growth rate of phytoplankton. For the time step to be calculated, the OBM calculation unit 14 calculates the growth rate of phytoplankton from the parameters of the ocean state for that time step, for example, using Equation (1).

[0037] The OBM calculation unit 14 inserts into the output data 18 a record that associates the oceanographic physical value calculated by the oceanographic physical calculation unit 12 with the OBM value calculated by the OBM calculation unit 14 for the time step to be calculated.

[0038] The calculation unit 11 repeats the processes of the oceanographic physical calculation unit 12, the coupling unit 13, and the OBM calculation unit 14 while incrementing the time step n1 from time step n1=0 to n1<n. The calculation unit 11 generates output data 18 that associates the oceanographic physical values ​​calculated by the oceanographic physical calculation unit 12 with the OBM values ​​calculated by the OBM calculation unit 14 for each time step.

[0039] The output processing unit 15 generates learning data 19 from the output data 18, which serves as teacher data for learning in the learning unit 16.

[0040] The learning unit 16 learns a learning model that estimates the growth rate of phytoplankton at each time step from the relationship between the oceanographic physical value and the OBM value at each time step. In the present disclosure, the learning unit 16 learns a learning model for estimating the output data shown in Equation (3) from the input data shown in Equation (2) by referring to the learning data 19. The learning unit 16 outputs learning model data 41 that identifies the learned learning model.

[0041] The learning unit 16 uses a general deep learning algorithm, such as a convolutional neural network or a recurrent neural network.

[0042] The estimation device 1 may estimate the growth rate for each type of marine microorganism. For example, the estimation device 1 classifies marine microorganisms into small phytoplankton and large phytoplankton and calculates the growth rate for each type. The OBM calculation unit 14 of the model generation unit 10 calculates an OBM value including the growth rate for each type of marine microorganism, and the learning unit 16 trains a learning model that estimates the growth rate for each type of marine microorganism.

[0043] In the present disclosure, an example of training data 19 generated by the model generation unit 10 will be described. The training data 19 is data that associates input data and output data during training. Equation (2) is input data during training. Equation (3) is output data during training. Note that T represents a tensor.

[0044]

[0045]

[0046] In the present disclosure, the estimation unit 20 estimates the growth rate at a lower resolution than the model generation unit 10. Therefore, the values ​​obtained by the calculation unit 11 are calculated by a simulation with a higher resolution than the resolution of the estimation by the estimation unit 20, and therefore each parameter in equation (2) is given a subscript H. In the following description, the low-resolution values ​​calculated by the estimation unit 20 are given a subscript L.

[0047] Using the learning data shown in equations (2) and (3), the learning model generated by the model generation unit 10 can estimate the growth rate for each type of phytoplankton, and estimates the growth rates of large and small phytoplankton.

[0048] In this case, the ocean physical initial values ​​31 and the ocean physical model data 32 include data related to seawater temperature. The OBM initial values ​​33 and the OBM data 34 include at least data related to nutrient salt concentrations and data on plankton growth rates. Here, the nutrients are NO3, PO4, and Fe. The growth rate data is the growth rate of small plankton and the growth rate of large plankton. Equations (2) and (3) set growth rates for each size of plankton, but growth rates may be set for other types, such as attributes, rather than just size.

[0049] The model generation unit 10 according to the present disclosure simulates oceanographic conditions using a general model and calculates values ​​of a marine biogeochemical model for the simulated oceanographic conditions. The model generation unit 10 calculates each value at a higher resolution than the observed data. The learning model generated by the model generation unit 10 enables more accurate estimation of phytoplankton.

[0050] (Estimation Unit) The estimation unit 20 will be described with reference to Fig. 3. The time step in the estimation unit 20 is expressed by a variable n2. N is the maximum value of the time step calculated by the estimation device 1.

[0051] The estimation unit 20 generates estimation data 28 by referring to each data in the input data storage unit 30 and the learning model data 41 generated by the model generation unit 10. The estimation unit 20 has the functions of a calculation unit 21, an output processing unit 15, and a learning unit 16, and each data of output data 18 and learning data 19.

[0052] The input data storage unit 30 has oceanographic initial values ​​31, oceanographic model data 32, OBM initial values ​​33, and OBM data 34'. Each piece of data in the input data storage unit 30 is the same as each piece of data described with reference to FIG. 2. However, unlike the OBM data 34 shown in FIG. 2, the OBM data 34' does not include a model for estimating the growth rate of phytoplankton. When estimating the growth rate of phytoplankton, the estimation unit 20 refers to the learning model data 41 generated by the model generation unit 10.

[0053] The estimated data 28 is data including the growth rate of phytoplankton at each time step.

[0054] At each time step, the calculation unit 21 calculates OBM values ​​other than the phytoplankton growth rate using ocean physical values ​​calculated from the ocean physical model specified by the ocean physical model data 32 and the OBM specified by the OBM data 34. The calculation unit 21 estimates the phytoplankton growth rate from the ocean physical values ​​and the OBM values ​​other than the phytoplankton growth rate using the learning model specified by the learning model data 41. Here, when the time step = 0, the calculation unit 21 references the ocean physical initial value 31 and the OBM initial value 33 instead of the ocean physical model data 32, the OBM data 34', and the learning model data 41.

[0055] The calculation unit 21 includes an oceanographic calculation unit 22 , a coupling unit 23 , and an OBM calculation unit 24 .

[0056] The oceanographic physical calculation unit 22 and the coupling unit 23 are similar to the oceanographic physical calculation unit 12 and the coupling unit 13 described with reference to Fig. 2. The oceanographic physical calculation unit 22 and the coupling unit 23 identify the ocean state at the time step to be calculated.

[0057] The OBM calculation unit 24 calculates OBM values ​​other than the growth rate of phytoplankton for the time step to be calculated using the ocean state and the OBM. The OBM calculation unit 24 further estimates the growth rate of phytoplankton for the time step to be calculated from the ocean physical values ​​and the OBM values ​​other than the growth rate of phytoplankton using the learning model generated by the model generation unit 10.

[0058] The OBM calculation unit 14 inserts into the output data 18 a record that associates the oceanographic physical value calculated by the oceanographic physical calculation unit 12 with the OBM value calculated by the OBM calculation unit 14 for the time step to be calculated.

[0059] The calculation unit 21 increments the time step from time step n1=0 to n1<n, and repeats the processes of the oceanographic calculation unit 22, the coupling unit 23, and the OBM calculation unit 24. The calculation unit 21 generates estimated data 28 that associates each time step with the growth rate of phytoplankton at that time step.

[0060] The output processing unit 25 outputs the estimated data 28 .

[0061] The estimation unit 20 may output the estimated data 28 at a lower resolution than the model generation unit 10. For example, the OBM calculation unit 24 calculates OBM values ​​other than the phytoplankton growth rate at each time step, and then calculates low-resolution OBM values ​​from the calculated OBM values ​​other than the phytoplankton growth rate. The OBM calculation unit 24 analyzes the calculated high-resolution BM values ​​other than the phytoplankton growth rate to calculate low-resolution OBM values ​​other than the phytoplankton growth rate. Any method may be used to calculate the low-resolution values. The OBM calculation unit 24 further estimates the plankton growth rate from the oceanographic physical values ​​and the low-resolution OBM values ​​other than the phytoplankton growth rate using a learning model.

[0062] Here, the low-resolution OBM value other than the growth rate of phytoplankton is expressed by, for example, equation (4).

[0063]

[0064] The OBM calculation unit 24 inputs the value shown in equation (4) into the learning model, and outputs the value shown in equation (5).

[0065]

[0066] This allows the estimation unit 20 to calculate the time-dependent growth rate of phytoplankton while reducing the processing load. The time-dependent growth rate of phytoplankton is used, for example, to calculate the primary production and carbon fixation amount of phytoplankton.

[0067] When the estimation device 1 estimates the growth rate for each species of marine microorganism, the model generation unit 10 trains a learning model that estimates the growth rate for each species of marine microorganism. In this case, the OBM calculation unit 24 uses the learning model that estimates the growth rate for each species of marine microorganism to estimate the growth rate for each species of marine microorganism from oceanographic physical values ​​and OBM values ​​other than the growth rate for each species of marine microorganism. The estimation device 1 can, for example, classify marine microorganisms into small phytoplankton and large phytoplankton and calculate the growth rate for each species.

[0068] (Estimation Method) The estimation method according to the present disclosure will be described with reference to Fig. 4. Of the processes shown in Fig. 4, steps S11 to S15 correspond to the processes of the model generation unit 10. Steps S21 to S27 correspond to the processes of the estimation unit 20.

[0069] In step S11, the estimation device 1 initializes a time step n1.

[0070] In step S12, the estimation device 1 calculates the oceanographic physical values ​​for time step n1. In step S13, the estimation device 1 calculates the oceanographic physical values ​​for time step n1.

[0071] In step S14, the estimation device 1 compares the time step n1 with the termination condition. If the time step n1 does not satisfy the termination condition, the process proceeds to step S16. In step S16, the estimation device 1 increments the time step n1 and repeats the processes of steps S12 and S13.

[0072] If the time step n1 satisfies the termination condition, the process proceeds to step S15. In step S15, the estimation device 1 generates a learning model for estimating the growth rate of phytoplankton.

[0073] In step S21, the estimation device 1 initializes the time step n2.

[0074] In step S22, the estimation device 1 calculates oceanographic physical values ​​for time step n2. In step S23, the estimation device 1 refers to the OBM to calculate oceanographic physical values ​​other than the growth rate of phytoplankton. In step S24, the estimation device 1 refers to the learning model generated in step S15 to estimate the growth rate of phytoplankton.

[0075] In step S25, the estimation device 1 compares the time step n2 with the termination condition. If the time step n2 does not satisfy the termination condition, the process proceeds to step S26. In step S26, the estimation device 1 increments the time step n2 and repeats the processes of steps S22 to S24.

[0076] If the time step n2 satisfies the termination condition, the process proceeds to step S27. In step S27, the estimation device 1 outputs the growth rate of phytoplankton for each time step.

[0077] The estimation device 1 according to the present disclosure can output the growth rate of phytoplankton for each time step, taking into account the fast evolutionary rate of marine microorganisms, and can more accurately estimate the growth rate due to evolution of marine microorganisms.

[0078] The estimation device 1 of the present embodiment described above uses, for example, a general-purpose computer system including a CPU (Central Processing Unit, processor) 901, a memory 902, a storage 903 (HDD: Hard Disk Drive, SSD: Solid State Drive), a communication device 904, an input device 905, and an output device 906. In this computer system, the CPU 901 executes a program loaded on the memory 902, thereby realizing each function of the estimation device 1.

[0079] The estimation device 1 may be implemented by one computer or by multiple computers, or may be a virtual machine implemented on a computer.

[0080] The program of the estimation device 1 can be stored in a computer-readable recording medium such as a HDD, an SSD, a Universal Serial Bus (USB) memory, a Compact Disc (CD), or a Digital Versatile Disc (DVD), or can be distributed via a network. The computer-readable recording medium is, for example, a non-transitory recording medium.

[0081] The present disclosure is not limited to the above-described embodiments, and various modifications are possible within the scope of the present disclosure.

[0082] REFERENCE SIGNS LIST 1 Estimation device 10 Model generation unit 11, 21 Calculation unit 12, 22 Oceanographic physics calculation unit 13, 23 Coupling unit 14, 24 OBM calculation unit 15, 25 Output processing unit 18 Output data 19 Learning data 16 Learning unit 20 Estimation unit 28 Estimation data 30 Input data storage unit 31 Oceanographic physics initial value 32 Oceanographic physics model data 33 OBM initial value 34 OBM data 41 Learning model data 901 CPU 902 Memory 903 Storage 904 Communication device 905 Input device 906 Output device

Claims

1. An estimation device comprising: a model generation unit that calculates OBM values, including the growth rate of marine microorganisms, at each time step using marine physical values ​​calculated from an ocean physical model and an OBM (marine biogeochemical model), and learns a learning model that estimates the growth rate of the marine microorganisms at each time step from the relationship between the marine physical values ​​and the OBM values ​​at each time step; and an estimation unit that calculates OBM values ​​other than the growth rate of the marine microorganisms at each time step using the marine physical values ​​and OBM calculated from the ocean physical model, and estimates the growth rate of the marine microorganisms from the marine physical values ​​and the OBM values ​​other than the growth rate of the marine microorganisms using the learning model.

2. The estimation device described in claim 1, wherein the estimation unit calculates OBM values ​​other than the growth rate of the marine microorganisms at each time step, calculates low-resolution OBM values ​​from the calculated values, and estimates the growth rate of the marine microorganisms from the oceanographic physical values ​​and the low-resolution OBM values ​​using the learning model.

3. The estimation device described in claim 1, wherein the model generation unit calculates OBM values ​​including the growth rate of each species of the marine microorganism, and learns another learning model that estimates the growth rate of each species of the marine microorganism, and the estimation unit uses the other learning model to estimate the growth rate of each species of the marine microorganism from the oceanographic physical values ​​and OBM values ​​other than the growth rate of each species of the marine microorganism.

4. An estimation method in which a computer calculates OBM values, including the growth rate of marine microorganisms, at each time step using oceanographic values ​​calculated from an oceanographic model and an OBM (marine biogeochemical model), learns a learning model that estimates the growth rate of the marine microorganisms at each time step from the relationship between the oceanographic values ​​and the OBM values ​​at each time step, calculates OBM values ​​other than the growth rate of the marine microorganisms at each time step using the oceanographic values ​​and OBM calculated from the oceanographic model, and estimates the growth rate of the marine microorganisms from the oceanographic values ​​and the OBM values ​​other than the growth rate of the marine microorganisms using the learning model.

Images