JP2015184225A - Apparatus, method and program for calculating prescribed layer predictive water temperature - Google Patents

Abstract

PROBLEM TO BE SOLVED: To highly accurately predict a future water temperature.SOLUTION: A prediction program 11c which acquires weather forecasting data from a weather forecasting information provision server 51, acquires water temperature data obtained by observing a water temperature at a prescribed layer in a water depth direction, and performs calculation by the use of a weather-linked water temperature prediction model for calculating a predictive water temperature obtained by predicting a future water temperature in the prescribed layer on the basis of the weather forecasting data and the water temperature data. The prediction program 11c is configured such that, with a water upper side defined as an upstream and a deeper side in the water depth direction defined as a downstream, the predictive water temperature is calculated by substituting upstream prediction information, which is water temperature data on the upstream of the prescribed layer or the weather forecasting data, and the water temperature data on the prescribed layer for an operational expression. In addition to the weather-linked water temperature prediction model, an autoregressive water temperature prediction model is used in combination so that a highly accurate prediction model is employed.

Description

  The present invention relates to a predetermined layer predicted water temperature calculation device, a predetermined layer predicted water temperature calculation method, and a predetermined layer predicted water temperature calculation program that accurately predict a future water temperature.

  Conventionally, the applicant has made a fixed-point observation of the underwater environment such as the water temperature in order to observe the influence of the underwater environment of the ocean on living organisms. In operations such as marine aquaculture, work may be required depending on the water temperature and so on, and the observation data provided by observing this underwater environment is widely used by fishermen.

  With regard to such an underwater environment, if it is possible to know not only current observation values but also future water temperature values, fishermen can prepare for work early or carry out work early. It will be useful. On the other hand, if the prediction is inaccurate, the water temperature did not rise as much as predicted in spite of the fact that the work was performed earlier, which may lead to the work being unnecessary. For this reason, if prediction is performed, accurate prediction is desired.

  Here, when patent information was investigated as to whether there is a technique for predicting an underwater environment such as a water temperature, an observation data assimilation method and an observation data assimilation apparatus have been proposed (see Patent Document 1). However, this method merely estimates (predicts) the water temperature or the like at a location where no sensor is installed, using the observation values (water temperature or the like) at a plurality of locations in the sea, and predicts the future water temperature or the like. It was not a thing.

  In addition, a sea state information providing system, a sea state information providing method, and a program thereof have been proposed (see Patent Document 2). However, this system pre-registers the ship's drainage volume, aggregates the received ship's position information and sea state information, calculates the ship's danger level from the sea state information and the ship's drainage volume, and displays the danger information.・ This is a warning, not a prediction of the future water temperature.

  As described above, there is no technique in the known literature useful for predicting the future water temperature.

JP 2008-241433 A JP 2004-354069 A

  In view of the above problems, the present invention provides a predetermined layer predicted water temperature calculation device, a predetermined layer predicted water temperature calculation method, and a predetermined layer predicted water temperature calculation program for accurately predicting a future water temperature. The purpose is to improve.

  The present invention includes a predicted meteorological information acquisition unit that acquires predicted meteorological information related to future weather, a predetermined layer measured water temperature information acquisition unit that acquires predetermined layer measured water temperature information in which the water temperature is observed in a predetermined layer in the depth direction, and A weather-linked predetermined layer predicted water temperature calculating means for calculating a predetermined layer predicted water temperature information for predicting a future water temperature of the predetermined layer based on the predicted weather information and the predetermined layer actually measured water temperature information, and the weather-linked type Predetermined layer predicted water temperature calculation means, the upstream side is upstream and the deeper one in the depth direction is downstream, predetermined layer predicted water temperature information upstream from the predetermined layer or upstream predicted information that is the predicted weather information, and the predetermined layer A predetermined layer predicted water temperature calculation device, a method thereof, and a program thereof are configured to calculate the predetermined layer predicted water temperature information by substituting the predetermined layer actual water temperature information into an arithmetic expression.

  According to the present invention, it is possible to provide a predetermined layer predicted water temperature calculation device, a predetermined layer predicted water temperature calculation method, and a predetermined layer predicted water temperature calculation program that accurately predict a future water temperature.

The system block diagram of a sea state data integrated management system. Explanatory drawing of the data format of observation data and weather forecast data. The block diagram which shows the data structure for browsing. The screen block diagram of a water temperature prediction display screen. Flow chart of prediction processing by observation information integrated processing server

  An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing a system configuration of the sea state data comprehensive management system A.
The ocean state data comprehensive management system A is used by the ocean state data comprehensive processing system 1, the observation device management system 2, the simple observation device management system 3, the ubiquitous observation device management system 4, the weather forecast information providing system 5, and the user. It consists of user terminals 7 and 8.

  The sea state data comprehensive processing system 1 manages the observation data general database 12 that stores the observation data 13 for browsing, and uploads data such as the browsing data 17b to the WEB server 17 by the observation data browsing screen creation program 11a to browse the sea state data. An observation information general processing server 11 (predetermined layer predicted water temperature calculation device, computer) that also functions as a screen creation device, an operation terminal 14 that performs data processing, internal browsing processing, and the like for the observation information general processing server 11, and a browsing program 17a A WEB server 17 is provided that provides an external WEB page and an internal WEB page using the browsing data 17b and the graph image 17c, and allows the user to browse the observation data.

The observation information integrated processing server 11 and the operation terminal 14 are connected to a switching hub 10 that is a communication monitoring device via a LAN, and are connected to the Internet 9 that is a public communication network from the switching hub 10 via a router 20 that is a communication monitoring device. It is connected. The observation information integrated processing server 11 includes an observation data browsing screen creation program 11a for creating an observation data browsing screen that can display a large number of types of observation data at once, and buoys (buoy robot 23, simple buoy 33, ubiquitous buoy 43). ) Is received from each observation device information providing server (21, 31, 41) and recorded in the browsing data 13 to manage the “invalid value” and “valid value” of the observation data and forcibly A management program 11b that performs a process of “forced missing measurement” as data missing, a prediction program 11c (predetermined layer predicted water temperature calculation program) that predicts a future water temperature, and an arithmetic expression for making a prediction linked to the weather 11d for creating a weather-linked prediction calculation formula, and auto-regression for creating a calculation formula for autoregressive prediction based on past actual measurement values Prediction calculation with equation generation program 11e is installed in the storage device.
The WEB server 17 is provided in an external network such as a rental server and is directly connected to the Internet 9. Thereby, the WEB server 17 operates the browsing program 17a to output the browsing data 17b and the graph image 17c based on the browsing observation data 13 to the user terminals 7 and 8 having the browser program as a WEB page. Is acceptable.

  The observation apparatus management system 2 includes an observation apparatus information providing server 21 and a plurality of observation apparatuses (hereinafter referred to as “buoy robots”) 23. The observation device information providing server 21 is connected to the Internet 9 via the router 20. In the configuration shown in the figure, the observation device information providing server 21 is connected to the LAN of the sea state data comprehensive processing system 1. However, the present invention is not limited to this, and the observation device information providing server 21 is the sea state data comprehensive processing system 1. It may be configured to connect to another Internet by connecting to another LAN.

  The buirobo 23 is driven by the power of the power supply unit, and measures a water temperature measurement unit that measures water temperature, a salinity measurement unit that measures salinity, a dissolved oxygen measurement unit that measures dissolved oxygen, and a flow direction flow velocity measurement that measures flow direction and flow velocity. A plurality of (four in the illustrated example) underwater measuring units, an air temperature measuring unit that measures the air temperature, a wind direction measuring unit that measures the wind direction, and an air speed measuring unit that measures the wind speed It has a measurement unit, and the control unit temporarily stores various observation data (water temperature, salinity, dissolved oxygen, flow direction, flow velocity, temperature, wind direction, wind speed) measured by these underwater measurement units and air measurement units in the storage unit. The information is stored and transmitted to the observation device information providing server 21 via the Internet 9 by the communication unit. For example, the underwater measurement unit is configured such that the first underwater measurement unit measures the water depth of 1 m, the second underwater measurement unit measures the water depth of 15 m, the third underwater measurement unit measures the water depth of 30 m, and the fourth underwater measurement unit measures the bottom layer. However, the present invention is not limited to this, and an appropriate depth and number of layers can be set.

The simple observation device management system 3 includes a simple observation device information providing server 31 and a plurality of simple observation devices (hereinafter referred to as simple buoys) 33 that are a kind of observation device.
The simple buoy 33 has a plurality of (four in the illustrated example) underwater measurement units that are driven by the power of the power supply unit and measure the water temperature, and various observations measured by these underwater measurement units. The control unit temporarily stores data (water temperature) in the storage unit, and transmits the data to the simple observation device information providing server via the Internet 9 by the communication unit. For example, the underwater measurement unit may be configured such that the first underwater measurement unit measures a water depth of 1 m, the second underwater measurement unit measures a water depth of 10 m, and the third underwater measurement unit measures a water depth of 20 m. The water depth and the number of layers can be set appropriately.

The ubiquitous observation apparatus management system 4 includes a ubiquitous observation apparatus information providing server 41 and a plurality of ubiquitous observation apparatuses (hereinafter referred to as ubiquitous buoys) 43 that are a kind of observation apparatus.
The ubiquitous buoy 43 is driven by the power of the power supply unit, and includes four underwater measurement units having a water temperature measurement unit that measures the water temperature, and an air measurement unit having an air temperature measurement unit that measures the temperature. The control unit temporarily stores various observation data (water temperature) measured by the measurement unit in the storage unit, and transmits the data to the ubiquitous observation apparatus information providing server 41 via the Internet 9 by the communication unit. The underwater measurement unit includes, for example, a first underwater measurement unit having a depth of 1 m, a second underwater measurement unit having a depth of 10 m, a third underwater measurement unit having a depth of 15 m, a fourth underwater measurement unit having a depth of 20 m, and a fifth underwater. The measuring part has a depth of 28 m.

The weather forecast information providing system 5 includes a weather forecast information providing server 51 and a HUB 52.
The weather forecast information providing server 51 is a server that stores the weather forecast of the Japan Meteorological Agency, and stores tomorrow weather forecast data and weekly weather forecast data today in the weather forecast database. The tomorrow weather forecast data and the weekly weather forecast data today include information (predicted weather information) such as maximum temperature, minimum temperature, wind speed, and precipitation probability.

The HUB 52 is a device that push-distributes an update notification indicating that an update has been made when at least one of today's tomorrow weather forecast data and weekly weather forecast data stored in the weather forecast database of the weather forecast providing server 51 is updated. It is.
Therefore, when receiving the update notification from the HUB 52, the control device of the integrated processing server 11 immediately accesses the weather forecast information providing server 51 and downloads today's tomorrow weather forecast data and weekly weather forecast data to obtain the latest weather. Forecast data (various data such as today's tomorrow weather forecast data and weekly weather forecast data) can be obtained. The control device of the integrated processing server 11 at this time functions as a predicted weather information acquisition unit.

The user terminal 7 includes an input device such as a touch panel, a display device such as a liquid crystal display, a communication device using a mobile phone network, a storage device that stores data and programs, and a control device that controls these, and a WEB server. 17 can be accessed and the observation data can be browsed.
The user terminal 8 is composed of a computer, and can access the WEB server 17 and browse observation data.

  Note that the user terminals 7 and 8, the observation information general processing server 11, the operation terminal 14, the WEB server 17, the observation device information providing server 21, the simple observation device information providing server 31, and the ubiquitous observation device information providing server 41 are all provided. Consists of computers such as personal computers or server computers, input devices such as a keyboard and mouse or touch panel, communication processing for data communication, display devices such as CRTs and liquid crystal displays, storage devices such as hard disks, and CPUs and ROMs And a RAM, and the controller executes arithmetic operations and controls various hardware operations in accordance with a program stored in the storage device.

  FIG. 2 shows observation data and weather forecast data acquired by the observation information integrated processing server 11 from the observation device information provision server 21, the simple observation device management system 3, the ubiquitous observation device management system 4, and the weather forecast information provision system 5. It is explanatory drawing explaining a format.

FIG. 2A is a configuration diagram showing a folder structure of the brobo file 85 (observation data) stored in the storage device of the observation device information providing server 21.
The folder in which the buirobo file 85 is stored is an observation item (water temperature: WT, salinity: SLT) below the data type specifying folder 81 that specifies the data type (hourly: HOUR, daily average: DAY, half-season: HSON). , Temperature: AT, dissolved oxygen: DOP, flow direction: ADCPD, flow velocity: ADCPS, wind direction: WD, wind speed: WS) are provided. Below the observation item folder 82 is stored a buirobo file 85 that records observation values in unit of one hour. In the buirobo file 85, future files are also prepared and stored in advance for an appropriate number of years in addition to the past. The buoy robot file 85 is set to a file name that can identify the buoy name, year, layer, and the like. This file name can be set as “HOUR_WT1_AO2013.CSV” in the format of “data for every hour” _ “observation item + layer number” _ “buoy name + year” .csv, for example. A time stamp 86 is set in the brobo file 85. When the data content of the brobo file 85 is updated, the control device of the observation device information providing server 21 updates the date / time (date, time) of the time stamp 86 to the update date / time. Note that the structure of the folder hierarchy and file name for storing the buirobo file 85 is not limited to this, and an appropriate structure may be adopted.

FIG. 2B is a configuration diagram showing data contents of the brobo file 85 stored in the storage device of the observation device information providing server 21.
As shown in the figure, the buirobo file 85 is composed of an ID, a state, and an observed value. The ID is given for each measurement, and it is possible to determine the observation value measured at what time on what month and by this ID. As for the state, symbols representing the measurement state are recorded by numerical values, such as whether the measurement was successful or the measurement value was abnormal. As the observation value, an observation value corresponding to the observation item of the brobo file 85 is recorded. For example, if the observation item is the brobo file 85 having the water temperature, the water temperature is recorded in degrees Celsius.

  FIG. 2C is an explanatory diagram of wave height buoy mail text data 91 (observation data) indicating observation data of a wave height buoy handled as a kind of simple buoy 33 in the present system. The wave height buoy mail text data 91 is composed of data in which observation data is arranged in a predetermined order with a predetermined number of characters, and has necessary data such as observation date and time and observation data. The simple observation device information providing server 31 functions as an e-mail transmission server (SMTP server), and transmits an e-mail having the wave height buoy mail body data 91 to a pre-registered e-mail address. Note that an e-mail receiving server (POP server) for managing the receiving e-mail is not shown.

  FIG. 2D is an explanatory diagram of simple buoy mail text data 92 (observation data) transmitted by the simple observation device information providing server 31 with observation data from the simple buoy 33 described in the text of the e-mail. The simple buoy mail text data 92 records data such as the date and time observed by the simple buoy 33, the water temperature of each layer, the internal pressure, the internal temperature, the solar battery, the battery, the measurement interval, and the next communication date and time.

  FIG. 2E is an explanatory diagram of a simple buoy attachment file 93 (observation data) attached to the simple buoy mail body data 92. As shown in the figure, the simple buoy attachment file 93 records data including items of date and time, water temperature of each layer, internal pressure, internal temperature, solar cell, and battery.

  FIG. 2F is an explanatory diagram showing a ubiquitous buoy download file 95 (observation data) of the ubiquitous buoy 43 stored in the storage device by the ubiquitous observation apparatus information providing server 41. The ubiquitous buoy download file 95 records the buoy ID, alias, owner, operation start date, operation end date, latitude, and longitude as common data, and the measurement data of each buirobo 23 as measurement data. Yes. The ubiquitous buoy download file 95 is stored in the storage device of the ubiquitous observation device information providing server 41. The observation information general processing server 11 acquires the ubiquitous buoy download file 95 from the ubiquitous observation apparatus information providing server 41. This download is acquired by an appropriate method such as HTTP download by a PHP program or FTP download.

  FIG. 2 (G) is an explanatory diagram showing a tomorrow weather forecast download file 97 tomorrow that has been downloaded from the weather forecast information providing system 5 as tomorrow weather forecast data as one of the weather forecast data.

  Today's tomorrow weather forecast download file 97 includes regional data indicating the area to be forecasted, time data indicating the forecast target time such as this morning, daytime, tonight, tomorrow, the day after tomorrow, weather data indicating the weather such as sunny, cloudy, rainy etc. , Maximum temperature data indicating daytime maximum temperature, minimum temperature data indicating arbitrage temperature, wind direction data indicating wind direction, wind speed data indicating wind strength, precipitation probability data indicating precipitation probability, wave height Wave height data shown, tsunami data showing a tsunami, etc. are recorded in XML format.

  These observation data and weather forecast data are acquired and recorded as needed by the management program 11b. The management program 11b, for example, acquires observation data from each observation device information providing server 21, 31, 41 after 5 to 10 minutes, preferably after 7 minutes, and receives an update notification from the HUB 52. Weather forecast data is acquired from the providing server 51. The acquisition of the weather forecast data will be described in detail. The management program 11b checks the feed every 30 seconds, and if there is a feed, acquires the weather forecast data in XML format as needed. Then, the management program 11b processes each acquired data into an appropriate form as shown in FIG. 3 to be described next, and stores it in the observation data general database 12 and the storage device of the WEB server 17.

  FIG. 3 shows setting data (101, 102) stored as browsing data 17b (see FIG. 1) in the observation data general database 12 of the observation information general processing server 11 and the storage device of the WEB server 17 after processing the acquired observation data. ), Priority order determination data 103, browsing data (104, 105, 106, 107, 108), and graph image data 109b. The WEB server 17 only needs to store the brobo browsing data 106 as data to be calculated and displayed each time. If the graph image 17c and HTML data are stored, other browsing data (104 , 105, 107, 108) need not be stored.

  FIG. 3A shows the structure of the weather-linked prediction calculation setting data 101. The weather-linked prediction calculation setting data 101 includes an X item, a Y item, a water temperature half-season, a temperature half-season, a start half-season, a half-end, a normal value coefficient, an upstream element coefficient, a (self-α) item coefficient, and Each section is composed of items.

The X item indicates an upstream element such as an air temperature (upstream actual measurement value or upstream prediction information). That is, if the predetermined layer which is the calculation target is the first layer, the temperature is indicated. If the predetermined layer which is the calculation target is the second layer or less, the water temperature of the one upstream layer (predetermined layer predicted water temperature information) ). The upstream element is one upstream in this embodiment, but is not limited to this, and various upstream elements that can be used when calculating each layer may be used. In this case, it is possible to freely select and calculate an upstream element optimum for prediction in the layer.
Y item shows which water temperature of which layer of which buoy is predicted.
The half-water temperature indicates how many half-months ago (Y item) value (measured value of the water temperature) is used. The water temperature designated here becomes the predetermined layer actually measured water temperature information in the calculation.
The mid-air temperature indicates how many half-months ago is used as the X item.

The first half indicates the beginning of the period in which this data is used.
The last half shows the period of the period when this data is used. The start half and end half indicate, for example, whether the data is used for the entire period of one year or only for a predetermined period such as spring, summer, autumn, winter. Therefore, it is possible to vary the data for each season.

The normal value coefficient indicates a coefficient for multiplying the normal value.
The upstream element coefficient indicates a coefficient for multiplying the upstream element. Here, the upstream is an element specified by the “X item” in the middle of the season specified by “temperature half-season”, and if it is a 1 m layer at the first water depth, the upstream is the temperature, and itself is the second If the water depth is 15 m, the upstream water temperature of the 1 m layer, which is the first water depth, and if it is 30 m water layer, which is the third water depth, the upstream is the water temperature of the 15 m layer, which is the second water depth. It refers to the upper layer (temperature if it is the top layer, otherwise water temperature one layer above itself).

The (self-α) item coefficient indicates a coefficient to be multiplied by the water temperature (measured value) of the buoy specified by the “Y item” in the half season specified by “water temperature half-season”.
An intercept shows the number added to the calculation result using the data.

  Using these data, when calculating with a weather-linked water temperature prediction model by the prediction program 11c, the calculation is performed using the following equation.

Y = “average value coefficient” x “average water temperature average for the forecasted half-season”
+ "Upstream element coefficient" x "Upstream actual or predicted value before specified half-year"
+ "(Self-α) item coefficient" x "Own water temperature before the specified half-year"
+ Intercept * Y: Indicates the predicted water temperature of the target buoy and layer specified in the Y item in the mid-season.
* Forecast half-year average value: Water temperature average value for the buoy forecast half specified by “Y item” * Upstream measured or predicted value before the designated half-time: Designated by “temperature half-season” The actual or predicted value of the buoy specified in the “X item” before half of the season (that is, the specified past half-season) (temperature or water temperature)
* The water temperature of the buoy specified in the “Y item” before the half-time specified in the “water temperature half-time” (that is, the specified half-past in the past) (water temperature)

  In this calculation formula, if “before half of the designated time” is “0”, the current half of the season is indicated. For example, if the calculation is performed on the first day of the current half-day, today's tomorrow weather forecast data today and tomorrow's expected maximum and minimum temperatures, and the forecast from 2 to 4 days after the weekly weather forecast data The highest temperature and the lowest expected temperature are acquired, and the expected average temperature in the middle of this period (that is, the current half) is calculated. In the case of half-season, such as calculating on the third day of the current half-season, use the past meteorological forecast data to obtain the expected maximum temperature and the expected minimum temperature for each applicable half-day, and the corresponding period. Calculate the expected average temperature in mid-season. In the case where measured values are obtained for the past, this measured value may be used, or past weather forecast data may be used as it is without using the measured value.

  The predicted average temperature obtained in this way can also be used as a predicted value upstream of 0 half an hour ago, and the predicted water temperature value of the 1 m layer, which is the first layer, can be calculated. When calculating the predicted water temperature of the 15m layer, which is the second layer, use the predicted value upstream of what half a month earlier by calculating the predicted water temperature of the first layer as described above. Even if it can be calculated.

  The normal value coefficient, the upstream element coefficient, the (self-α) item coefficient, and the intercept value are calculated by the weather-linked prediction calculation formula creation program 11d using the past water temperature and predicted temperature for each buoy. It is calculated in advance by multiple regression analysis.

  FIG. 3B shows the structure of the setting data 102 for autoregressive prediction calculation. The autoregressive prediction calculation setting data 102 includes items and coefficients. In the items, a1, a2,..., Which mean coefficients of the autoregressive water temperature prediction model, and wt1_ta, wt2_ta, wt3_ta,..., Which mean eigenvectors (degrees of fluctuation) of each buoy and layer are stored. The coefficient is set.

  Using these data, when the auto-regressive water temperature prediction model is calculated by the prediction program 11c, the calculation is performed using the following arithmetic expressions (Expression 1) to (Expression 4).

(Formula 1)
Principal component score Z ₁ = “coefficient A” × “first buoy first layer average deviation”
+ "Coefficient B" x "1st buoy 2nd layer average deviation"
+ "Coefficient C" x "1st buoy 3rd layer average deviation"
+ "Coefficient D" x "1st buoy 4th layer average deviation"
+ "Coefficient E" x "second buoy first layer average deviation"
+ "Coefficient F" x "2nd buoy 2nd layer average deviation"
+…
(Example of Formula 1)
Z ₁ = 0.275 _TA1 + 0.263 _TA15 + 0.250 _TA30
+ 0.223 _TAb + _{0.317 AO1} + 0.309 _AO15
+ 0.309 _AO30 + 0.270 _AOb + 0.334 _TO1
+ 0.314 _TO15 + 0.303 _TO30 + 0.278 _TOb
(* Values are coefficients. TA1 is flat buoy 1m layer, TA15 is flat buoy 15m layer, TA30 is flat buoy 30m layer, TAb is flat buoy bottom layer. Similarly, AO1-b are Aomori buoy 1m layer-bottom layer, TO1-b. Is East Bay buoy 1m layer ~ bottom layer.)

(Formula 2)
Autoregressive score value Z _n for the forecasted half-season
= “Coefficient a1” × “principal component score one half before the target layer”
+ "Coefficient a2" x "principal component score of two and a half months before the target layer"
+ "Coefficient a3" x "principal component score before 3 half months of target layer"
+ "Coefficient a4" x "principal component score of the target layer 4 quarters before"
+ "Coefficient a5" x "principal component score before half of the target layer"
(Example of Formula 2)
Z _n = 1.09983 × Z _n−1 −0.24522 × Z _n−2
+ 0.08854 × Z _n−3 −0.08006 × Z _n−4
+ 0.07005 × Z _n-5
(* Numbers are coefficients. Zn _{-1 to} Zn _-5 are principal component scores before the designated half-year)

(Formula 3)
Prediction deviation = “Eigenvector of the target layer” x Autoregressive score score value for the prediction target half-season

(Formula 4)
Predicted water temperature = "Predicted deviation" + "Normal water temperature"

Hereinafter, the calculations according to the above (formula 1) to (formula 4) will be described.
First, the autoregressive water temperature prediction model is a model for obtaining a future predicted water temperature from the measured water temperature, that is, a model for predicting the future self using its own data. In this model, first, principal component analysis is performed in order to grasp the variation pattern.

  Principal component analysis is an analysis method in which the values of many variables are represented by one or a few comprehensive indicators with as little information loss as possible. In selecting the principal component, it is necessary to compare the eigenvalue and contribution ratio of each principal component.

As a guideline for selecting the principal component, (1) the eigenvalue is 1 or more, (2) the difference from the latest contribution rate is large, and (3) the cumulative contribution rate is 80% or more. Of the principal components of the autoregressive water temperature prediction model, only the first principal component (z ₁ ) corresponds to this, and this is adopted. Using this result, calculation is performed using (Equation 1) described above as an autoregressive model for water temperature prediction using the first principal component.

  Next, the order of the model is performed using the Akaike information criterion. AIC is a statistic indicating how well the model is applied, and the smaller the numerical value, the better. From the results of AIC, if it is appropriate to adopt principal component scores up to, for example, five and a half months before (for example) from the results of AIC, the water temperature prediction autoregressive model becomes as described above (Formula 2). In addition, although it was described here that the period before half a month was appropriate, there may be cases where other half a month is also appropriate. In that case, it is better to prepare the necessary pre-half-time coefficient.

As shown in (Equation 3), multiplied by eigenvectors in the autoregressive score Z _n, can be calculated by adding average year value as shown in (Equation 4) (average year temperature), the predicted value (predicted temperature) .

  In this manner, the autoregressive water temperature prediction model can perform prediction using data of all layers 1 to 5 months before (by a plurality of past periods). Note that the configuration is not limited to using all layers of all buoys, and a specific buoy may be predicted using data of all layers (or a predetermined number of layers) of the buoy. In this case, what is necessary is just to estimate about the specific buoy using the data until 3-8 months ago.

  FIG. 3C shows the structure of the priority determination data 103 (priority determination information). The priority determination data 103 has items Y, wt_HJ, AT_HJ, Method, and R ^ 2, and one calculation setting data (calculation setting information) is recorded in one record (one line). .

Y indicates which priority is given to predicting the water temperature of which layer of which buoy.
wt_HJ is actual measurement designation information indicating how many months before the water temperature data is used for calculation.

  AT_HJ is upstream designation information that indicates how many months before the upstream water temperature data (or temperature data) is used for the calculation.

  Method is calculation method designation information indicating which of the weather-linked water temperature prediction model (when the value is 2) or the autoregressive water temperature prediction model (when the value is 1) is used as the water temperature prediction model.

  R ^ 2 indicates the degree of coincidence when calculation is performed using the contents specified by wt_HJ, AT_HJ, and Method. This degree of coincidence is calculated by a calculation simulation using past data, and the degree of coincidence is higher as the value is closer to 1 (that is, the value is larger) which is the maximum value.

  The data in the priority determination data 103 is sorted in descending order of the degree of coincidence stored in R ^ 2, and is used by searching in order from the top. Specifically, the degree of coincidence is defined as data (record) in which “Y” matches a predetermined layer of a predetermined buoy to be calculated and each data specified by “wt_HJ” and “AT_HJ” is obtained. Using the data at the time of a search and hit in order from the highest data, the calculation is performed by the water temperature prediction model specified by the data.

  The priority setting data 103 stores operation setting data for all buoys and layers. In addition, the priority determination data 103 is different for each buoy and stratum when to use half of the weather forecast data (AT_HJ) and when to use half of its own observation data (wt_HJ). Accordingly, a plurality of types of calculation setting data, in which other values are also changed accordingly, are recorded. Therefore, when predicting the water temperature of a layer of a buoy, the one with the highest degree of coincidence that can be calculated from the currently obtained data is selected from the multiple types of calculation setting data that exist for that buoy and layer. Is done.

  This priority determination data 103 is sorted only by the degree of coincidence without being grouped according to the water temperature prediction model, which half-time data is used, etc., so the most accurate calculation method can be selected easily and appropriately. it can.

  FIG. 3D shows the structure of the brobo browsing data 104 that stores the water temperature, and FIG. 3E shows the structure of the brobo browsing data 105 that stores the daily average value obtained by calculation from the observation data. FIG. 3 (F) shows the structure of the semi-averaged brobo browsing data 106 obtained by calculation from the daily average. Note that “half-season” refers to half of a seasonal day, that is, five days, and in this embodiment, half-season is used as a predetermined time unit. The brobo browsing data 104, 105, 106 has the same structure as the brobo file 85 (see FIG. 2B), and is composed of an ID, a state, and an observed value. Note that the observation items other than the water temperature are stored in the same structure as the brobo browsing data 104, and thus the description thereof is omitted.

  FIG. 3G shows the data structure of the simple buoy browsing data 107. The simple buoy browsing data 107 is in the same format as the brobo browsing data 104, 105, 106 based on the wave height bumail text data 91, the simple buoy text data 92, and the simple buoy attachment file 93 described with reference to FIG. It is processed data, and consists of ID, state, and observed value (1m water temperature etc.). Note that the time of the mail text is not the hour because it is the mail transmission time, but since the observation is performed accurately at the hour based on the time obtained from the GPS, the data obtained from the email is the data measured at the hour Treat as. The hour on the hour means 0 minutes per hour.

  FIG. 3H shows the data structure of the ubiquitous buoy browsing data 108. The ubiquitous buoy browsing data 108 extracts necessary observation data (date, time, outside temperature, water temperature of each layer) from the ubiquitous buoy download file 95 (see FIG. 2 (F)), and buoy robot browsing data 104, 105. , 106 and processed in the same format as the ID, state, and observed value (1 m water temperature, etc.). The data of the ubiquitous buoy browsing data 108 stores a value corrected to the reference time value because the measurement time is different from the reference time (normal time).

  These simple buoy browsing data 107 and ubiquitous buoy browsing data 108 are also provided with observation item folders indicating observation items below the data type specific folder, similar to the buirobo browsing data 104, 105, 106. A file for recording data in units of one hour, which is a unit time, is stored in the lower part, and each file is set to a file name that can identify a buoy name, year, layer, and the like.

  FIG. 3I shows the data structure of the output replacement table 109a on the home page. This output replacement table 109a is text data in which variable names and values are arranged, and one variable name and value are arranged in one record. Note that the output replacement table 109a is data in which variable names and values are stored in a comma-separated form, but is shown in a table format for easy understanding in the drawing. This output replacement table 109a stores all information for appropriately replacing HTML models, such as predicted values for each layer of each buoy, measured values for each layer of each buoy, and weather. Thereby, the observation data browsing screen creation program 11a can easily display the actual measurement value and the predicted value on the screen.

  FIG. 3J is a configuration diagram of a prediction graph image 109b showing water temperature prediction. The vertical axis is the observation item value (water temperature), and the horizontal axis is the date and time. The predicted graph image 109b displays the measured value and the predicted value in one screen using a line color, a line type, or the like so that the measured value is displayed in black and the predicted value is displayed in red. When the predicted value is calculated, the predicted graph image 109b is created and stored in the storage device or the like of the WEB server 17, so that the predicted graph image 109b displaying the measured value up to now and the predicted value in the future is displayed. It can be immediately provided to users who have accessed the WEB site.

  FIG. 4 is a screen configuration diagram of the water temperature prediction display screen 110 that the browsing program 17 a of the WEB server 17 displays on the user terminals 7 and 8. This water temperature prediction display screen 110 includes a half-season predicted water temperature data display unit 111, a plurality of predicted water temperature graph display units 115 that display the predicted water temperature for each buoy and layer, and the degree of coincidence of prediction based on past data for each buoy and layer. And a prediction accuracy data display unit 119 indicating the hit rate.

  The half-season predicted water temperature data display unit 111 includes an actual measurement value display unit 112 that indicates an actual measurement value in the latest half-season, and a predicted value display unit 113 that indicates an estimated value up to six half-ahead. The measured value display unit 112 and the predicted value display unit 113 display the measured water temperature or the predicted water temperature for each buoy and layer.

  The predicted water temperature graph display unit 115 is a graph in which the vertical axis represents the water temperature and the horizontal axis represents half-season. The predicted water temperature graph display unit 115 includes a measured value graph 116 for the past six and a half and a predicted value graph 117 for the next six and a half years including the present. Displayed as a continuous line graph. The measured value graph 116 and the predicted value graph 117 are common elements such as, for example, a point (plot) indicating a value and a line thickness and shape that are the same but different in color so that they can be clearly distinguished. And display with different elements.

Further, the normal value graph 118 indicating the past normal value is displayed in a form different from the actual value graph 116 and the predicted value graph 117 (for example, the color and line type are changed).
The prediction accuracy data display unit 119 displays the degree of coincidence and the hit rate for each buoy and layer.

  FIG. 5 is a flowchart of an operation in which the control device of the observation information general processing server 11 executes the prediction process by the prediction program 11c.

The control device waits until an update signal is received from the HUB 52 (step S1: No). When the update signal is received (step S1: YES), the control device acquires tomorrow weather forecast data and weekly weather forecast data from the weather forecast providing server 51 (step S2).
The control device acquires observation data in a range necessary for prediction from the browsing observation data 13 (step S3). At this time, the control device of the observation information integrated processing server 11 functions as a predetermined layer actually measured water temperature information acquisition unit.

  In order to calculate the predicted value of the predetermined half of the predetermined layer of the predetermined buoy, the control device first refers to the priority order determination data 103 and the data for calculating this predicted value is obtained most. A calculation method having a high degree of coincidence is determined (step S4). At this time, the control device functions as a selection unit that selects a calculation method having a high degree of coincidence.

  The control device calculates a predicted value based on the determined calculation method and the obtained data (step S5). At this time, a predicted value is calculated using the model determined by the priority determination data 103 among the autoregressive water temperature prediction model and the weather-linked water temperature prediction model and the obtained data. When the weather-linked water temperature prediction model is used, the control device functions as a weather-linked predetermined layer predicted water temperature calculation means, and when the autoregressive water temperature prediction model is used, the control device functions as an auto-regressive predetermined layer predicted water temperature calculation means. To do.

  The control device advances the prediction target next (step S7) until the calculation of the predicted value of the necessary range is completed (step S6: No), returns to step S4, and repeats the process. At this time, if there is a layer downstream from the current layer for which the predicted value is calculated, one downstream layer is targeted for prediction, and if the lowest layer (for example, the bottom) is reached, The most upstream half layer (for example, 1 m layer) ahead is set as the prediction target. Then, when prediction for all layers in the future half of the current buoy, which is the prediction limit (for example, six half years ahead), is completed, prediction values are calculated in order from the most recent highest layer for another buoy. .

  When the calculation of the predicted value of the necessary range is completed (step S6: Yes), the control device outputs the obtained buoy-by-layer stratified forecast data to be recorded in the browsing data 17b of the WEB server 17 ( Step S8). At this time, the control device also creates a predicted water temperature graph display unit 115 and stores it in the storage device of the WEB server 17.

With the above configuration and operation, the future water temperature can be accurately predicted.
Since the water temperature is predicted for each layer, it is possible to predict with high accuracy both the layer such as the 1 m layer that is easily affected by sunlight and the layer such as the bottom layer that is not easily affected by sunlight.

  In addition, since it is calculated using the upstream predicted value, the most upstream layer (for example, 1 m layer) is predicted using the predicted temperature, and the immediately downstream layer (for example, 15 m layer) is predicted using the prediction result. In addition, it is possible to predict up to the deep layer one after another.

  In addition, by using the autoregressive prediction formula creation program 11e, a highly accurate and stable prediction based on the past water temperature change can be performed for a relatively long period of time using data of all layers at least 1 to 5 months ago. For example, it can be carried out over a period of 4 to 6 months.

  Further, by using the autoregressive prediction calculation formula creation program 11e, it is possible to calculate only by the water temperature even if the weather prediction cannot be obtained. Further, the autoregressive prediction formula creation program 11e can faithfully reproduce the water temperature curve for the whole year.

  In addition, by using the weather-linked water temperature prediction model of the prediction program 11c, the latest water temperature (for example, two to three and a half years ahead, 4 to 5 half a year ahead) can be used with high accuracy in response to sudden weather fluctuations. Can be predicted.

  In addition, by using the weather-linked water temperature prediction model of the prediction program 11c, it is possible to cope with inexperienced water temperature, and prediction can be made with a small amount of data even if data of all layers before the first to the fifth half is not obtained. .

  In addition, the prediction program 11c that calculates with the weather-linked water temperature prediction model realizes more accurate prediction by using the past normal value in addition to the upstream predicted value and its own actual measured value. can do.

  In addition, the prediction program 11c that calculates with the weather-linked water temperature prediction model performs calculation by further substituting the water temperature average value (average water temperature information) of the predetermined layer that is the calculation target into the calculation formula, so that more stable and highly accurate Can be predicted.

  Moreover, since the prediction program 11c is a structure which selects the higher prediction accuracy according to a buoy layer among a weather-linked water temperature prediction model and an autoregressive water temperature prediction model, it takes high points of both models and provides high accuracy. Prediction can be realized.

  In particular, it is an autoregressive water temperature prediction model that can perform calculations with a high degree of coincidence of approximately 0.96, etc., and even for summer peak water temperatures where the degree of coincidence falls to 0.75 or 0.89, weather-linked water temperature prediction By using the model, the accuracy can be improved to a degree of coincidence such as 0.91 and 0.95.

  As a result, for scallop farming where large-scale mortality is a problem due to high water temperature during the summer, it is possible to prevent mortality by moving the depth of the culturing early with high-precision prediction. In other words, in the past, the scallops were moved from 23 ° C., where the growth of scallops stopped, to 23 ° C., where physical exhaustion began, and to 25 ° C., where there was a risk of death. This is because if the water is moved faster than that, the water temperature will not increase so much and it will be a wasteful work, and if it is moved deeper, the scallops will eat less and the growth will worsen. This is because it was optimal in terms of work efficiency. On the other hand, according to the present invention, it is possible to predict and indicate how much the water temperature will rise, so that it is possible to determine whether or not to move and to give a movement instruction at a point earlier than before. As a result, it is possible to prevent scallop mortality without causing unnecessary work.

  In addition, using the priority order determination data 103 arranged according to the priority order, prediction is performed using a calculation method having the highest degree of coincidence (high accuracy) among the calculation methods that can be calculated using the obtained data. Precise predictions can be realized over the entire period.

  In addition, the priority order determination data 103 includes a plurality of different weather forecast data and observation data to be used for the weather forecast data and observation data used for calculation for one buoy and layer. Various types of calculation setting data are recorded and can be selected in descending order of coincidence. For this reason, even if there is a missing measurement in the observation data, it is possible to make a prediction using the observed observation data and weather forecast data in mid-season. In addition, since the calculation setting data having the highest degree of coincidence is selected and calculated from the calculation setting data that can be calculated from the obtained observation data and weather forecast data, the best prediction can be performed using the obtained data.

  The water temperature prediction display screen 110 displays the observation data that has already been observed and the prediction data ahead of it as a continuous line graph, and has different elements by color-coding the observation data portion and the prediction data portion. Because it is displayed, the user can intuitively easily understand the actual water temperature transition and the predicted water temperature transition ahead.

The present invention is not limited only to the configuration of the above-described embodiment, and many embodiments can be obtained.
For example, the observation information general processing server 11 and the WEB server 17 as the observation data browsing screen creation device are constructed as separate servers. However, the present invention is not limited to this, and a single server computer has both functions. It can be configured as follows.

  Also, it corresponds to the three types of buirobo 23, simple buoy 33, and ubiquitous buoy 43, and the data format also corresponds to the format downloaded by FTP and the format received by e-mail, but it corresponds to some of these, or It is possible to adopt an appropriate configuration such as handling more than this.

In addition, the observation data and the weather forecast data are calculated and output in half-season units. However, the present invention is not limited to this. It may be configured to output. In this case, the predicted value can be calculated in a determined period unit.
Moreover, although the predicted temperature obtained by averaging the average value of the predicted maximum temperature and the predicted minimum temperature in half-season is used as the predicted weather information, the present invention is not limited to this, and appropriate predicted weather information is used according to the obtained information. be able to.
Moreover, it is good also as a structure which takes in abnormal weather early warning information as weather forecast data, and improves a prediction precision.

  Moreover, although the upstream and downstream were made into the gravitational direction (vertical direction), not only this but a horizontal direction is good. In this case, for example, calculation may be performed at the same water depth with the water temperature on the inlet side being upstream and the water temperature on the back side being downstream. As a result, it is possible to cope with a marine environment in which the prediction in the horizontal direction is more accurate than the prediction in the vertical direction, and high-precision prediction can be realized in more types of environments.

  The present invention can be used in a system that accurately predicts the future water temperature in the ocean. However, the present invention is not limited to this, and various prediction systems, particularly systems in which multiple observation points exist in the depth direction. Can be used to predict

11 ... Integrated processing server 11c ... Prediction program

Claims (6)

Predicted weather information acquisition means for acquiring predicted weather information about future weather;
Predetermined layer actual water temperature information acquisition means for acquiring predetermined layer actual water temperature information in which the water temperature was observed in a predetermined layer in the water depth direction;
A weather-linked predetermined layer predicted water temperature calculating means for calculating a predetermined layer predicted water temperature information obtained by predicting a future water temperature of the predetermined layer based on the predicted weather information and the predetermined layer actually measured water temperature information;
The weather-linked predetermined layer predicted water temperature calculating means is:
Calculates the predetermined layer predicted water temperature information upstream of the predetermined layer or the upstream predicted information that is the predicted weather information, and the predetermined layer actually measured water temperature information of the predetermined layer, with the upstream side as the upstream and the deeper in the depth direction as the downstream. A predetermined layer predicted water temperature calculation device configured to calculate the predetermined layer predicted water temperature information by substituting into an equation. The predetermined layer actually measured water temperature information acquisition means is configured to acquire the predetermined layer actually measured water temperature information for each of a plurality of past periods,
Autoregressive predetermined layer predicted water temperature calculating means for calculating predetermined layer predicted water temperature information, which is a future predicted water temperature of the predetermined layer, based on the acquired predetermined layer actual water temperature information;
A selection unit that selects the one with higher prediction accuracy for each predetermined layer to be predicted for calculating the predicted water temperature among the weather-linked predetermined layer predicted water temperature calculating unit and the autoregressive predetermined layer predicted water temperature calculating unit. The predetermined layer predicted water temperature calculation device according to 1. Calculation method designation information for designating which of the weather-linked predetermined layer predicted water temperature calculating means and the autoregressive predetermined layer predicted water temperature calculating means is to be used for calculation;
Upstream designation information for designating which time of the upstream prediction information is used;
Calculation setting information composed of actual measurement designation information for designating which period of information is used as the predetermined layer actual water temperature information,
It has priority determination information that memorizes the level of prediction accuracy for each layer so that it can be distinguished,
The selection means includes
The predetermined layer predicted water temperature calculation device according to claim 2, wherein the predetermined setting water temperature calculation device is configured so as to employ a calculation setting that can be calculated based on the acquired information with reference to the priority determination information. The predetermined layer actually measured water temperature information is stored for each of a plurality of layers from upstream to downstream,
The process of calculating the predetermined layer predicted water temperature information by the calculation method selected by the selection unit is sequentially executed from the upstream layer to the downstream layer, and the time direction is sequentially executed from the latest to the future. The predetermined layer predicted water temperature calculation device according to claim 3, wherein the predicted value of a plurality of times is calculated for a plurality of layers. Predicted weather information acquisition means for acquiring predicted weather information about future weather;
Predetermined layer actual water temperature information acquisition means for acquiring predetermined layer actual water temperature information in which the water temperature was observed in a predetermined layer in the water depth direction;
Using a weather-linked predetermined layer predicted water temperature calculating means for calculating a predetermined layer predicted water temperature information for predicting a future water temperature of the predetermined layer based on the predicted weather information and the predetermined layer actually measured water temperature information;
By the weather interlocking type predetermined layer predicted water temperature calculating means,
Calculates the predetermined layer predicted water temperature information upstream of the predetermined layer or the upstream predicted information that is the predicted weather information, and the predetermined layer actually measured water temperature information of the predetermined layer, with the upstream side as the upstream and the deeper in the depth direction as the downstream. A predetermined layer predicted water temperature calculation method for calculating the predetermined layer predicted water temperature information by substituting into a formula. Computer
Predicted weather information acquisition means for acquiring predicted weather information about future weather;
Predetermined layer actual water temperature information acquisition means for acquiring predetermined layer actual water temperature information in which the water temperature was observed in a predetermined layer in the water depth direction;
Functioning as a weather-coupled predetermined layer predicted water temperature calculating means for calculating predetermined layer predicted water temperature information for predicting a future water temperature of the predetermined layer based on the predicted weather information and the predetermined layer actually measured water temperature information;
The weather-linked predetermined layer predicted water temperature calculating means is:
Calculates the predetermined layer predicted water temperature information upstream of the predetermined layer or the upstream predicted information that is the predicted weather information, and the predetermined layer actually measured water temperature information of the predetermined layer, with the upstream side as the upstream and the deeper in the depth direction as the downstream. A predetermined layer predicted water temperature calculation program for calculating the predetermined layer predicted water temperature information by substituting into a formula.