Frost predicting system

Abstract

To provide a frost predicting system which can stably prevent frost from causing damages.SOLUTION: The present invention includes: a correlational expression estimation step S3 for holding a previous correlation between one of an actually measured lowest temperature and a predicted lowest temperature in a reference site 3 for several days in the past and an actually measured lowest temperature in a target site 2 on each of the corresponding days and estimating the estimated lowest temperature in the target site 2 on a prediction day from the predicted lowest temperature in the reference site 3 on the prediction day and the past correlation; and a predictions step S4 for predicting frost on leaf surfaces of the prediction day from the estimated lowest temperature estimated by the correlational expression estimation step S3.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention relates to a frost prediction system for predicting frost. [Background technology]

[0002] In order to prevent damage to crops caused by frost, it is necessary to take measures before frost occurs, such as covering crops with nonwoven fabric or mulch. To do this, it is essential to predict frost in advance. For example, frost warnings issued by the Japan Meteorological Agency are known as such frost predictions, but they are issued according to meteorological information at the observation point. Therefore, frost prediction systems are known that can predict frost even in environments with different distances, altitudes, topographical characteristics, etc. from the observation point.

[0003] For example, the frost prediction system shown in Patent Document 1 is connected to a weather forecast server, and can substitute weather forecast data obtained from the weather forecast server into a frost determination formula and determine that frost will occur if the conditions are met. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2006-189403 A (Pages 5 and 6, Figure 2) Summary of the Invention [Problem to be solved by the invention]

[0005] In a frost prediction system such as that described in Patent Document 1, the system is also connected to a local meteorological observation device installed in the target area for which frost prediction is required, and the frost judgment formula can be updated based on local meteorological data acquired from the local meteorological observation device. As a result, the more the frost judgment formula is updated, the more the prediction accuracy can be improved. However, since a number of trials are required until the prediction accuracy of a complex frost judgment formula is improved, there is a risk of frost damage occurring, especially during periods when the number of trials is insufficient.

[0006] The present invention has been made in consideration of such problems, and has an object to provide a frost prediction system that can stably prevent frost damage. [Means for solving the problem]

[0007] In order to solve the above problems, the frost prediction system of the present invention comprises: a correlation equation estimation step of storing a past correlation between the actual minimum temperature or the predicted minimum temperature at a reference location for a plurality of past days and the actual minimum temperature at a target location for each corresponding day, and estimating an estimated minimum temperature at the target location on a prediction date from the predicted minimum temperature at the reference location on the prediction date and the past correlation; and a prediction step of predicting frost on the leaf surfaces on the predicted day from the estimated minimum temperature estimated by the correlation equation estimation step. According to this feature, it is possible to predict frost on the leaves of the target area on the predicted date by using the highly accurate estimated minimum temperature, so that frost damage can be reliably prevented.

[0008] The past correlation is characterized in that it is formed using a regression line of a lower limit sample group among a sample group consisting of samples of the actual minimum temperature or predicted minimum temperature at the reference site and the actual minimum temperature at the target site for each day. This feature can improve the accuracy of frost prediction.

[0009] The sample is characterized in that it is taken on a day when at least one of the conditions of weather, wind speed, and rainfall is satisfied. This feature can improve the accuracy of frost prediction.

[0010] The samples are characterized in that they are taken on days when all the conditions of weather, wind speed, and rainfall are satisfied. This feature can further improve the accuracy of frost prediction.

[0011] The system is characterized in that a first sensor that detects at least one of weather, wind speed, and rainfall and a second sensor that measures air temperature are placed in different locations, and the second sensor is placed in a depression within the target area. This feature makes it possible to predict frost that is more suited to the environment of the target area.

[0012] The frost prediction system further includes an elevation difference estimation step of estimating an estimated minimum temperature of the target location on the prediction date from an elevation difference between the reference location and the target location and a predicted minimum temperature of the reference location on the prediction date, The prediction step is characterized in that frost on the leaf surfaces on the predicted day is predicted from the estimated minimum temperature estimated by the altitude difference estimation step. This feature makes it possible to prevent frost damage more reliably.

[0013] The prediction step is characterized in that it predicts that frost will occur on the leaf surface if at least one of the estimated minimum air temperature estimated by the correlation equation estimation step and the leaf surface temperature calculated from the estimated minimum air temperature estimated by the altitude difference equation estimation step satisfies a condition. This feature makes it possible to increase the frequency with which frost is predicted to occur.

[0014] The frost prediction system is characterized by predicting frost for several days. This feature allows for a certain amount of time to deal with frost. [Brief description of the drawings]

[0015] [Figure 1] FIG. 1 is a schematic diagram showing a frost prediction system in a first embodiment. [Diagram 2] 13 is a flowchart of a frost determination process. [Diagram 3] 1 is a graph showing a lower limit collapse line. [Figure 4] This is a distribution diagram of leaf surface temperatures calculated from the actual minimum air temperatures measured on high ground and leaf surface temperatures derived from the MLR method. [Diagram 5] This is a distribution diagram of leaf surface temperatures calculated from the actual minimum air temperatures in a depression and leaf surface temperatures derived from the MLR method. [Figure 6] This is a distribution diagram of leaf surface temperatures calculated from the actual minimum air temperatures on high ground and leaf surface temperatures derived from the LEM method on high ground. [Figure 7] This is a distribution diagram of leaf surface temperatures calculated from the actual minimum air temperatures in the depression and leaf surface temperatures derived from the LEM method in the depression. [Figure 8] FIG. 13 is a diagram showing the text of an email displayed on a user terminal. [Figure 9] 13 is a flowchart of an update process of the correlation estimation equation of the frost prediction system in the second embodiment. [Figure 10] This is a distribution map of leaf surface temperatures calculated from the actual minimum air temperatures in the depression and leaf surface temperatures derived from the LEM method in the depression before the correlation estimation equation was updated by the update process. [Figure 11] This is a distribution map of leaf surface temperatures calculated from the actual minimum air temperatures in a depression and leaf surface temperatures derived from the LEM method in a depression where the correlation estimation equation has been updated by the update process. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0016] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A frost prediction system according to the present invention will be described below based on an embodiment.

[0017] The frost prediction system of the present invention predicts the occurrence of frost in a target location by using the predicted minimum temperature of a reference location. In predicting frost, an estimation step is performed to calculate an estimated minimum temperature in the target location from the predicted minimum temperature of the reference location, and a prediction step is performed to predict whether frost will occur on the leaves from the estimated minimum temperature.

[0018] In the present invention, two types of estimation steps are used as the estimation steps: an elevation difference estimation step that calculates the estimated minimum temperature in a target area (hereinafter referred to as the MLR (moist-adiabatic lapse rate) method) by using a moist adiabatic lapse rate method (hereinafter referred to as the MLR-estimated minimum temperature), and a correlation equation estimation step that calculates the estimated minimum temperature in a target area (hereinafter referred to as the LEM (linear estimate) method) by using a linear estimation method (hereinafter referred to as the LEM-estimated minimum temperature).

[0019] The MLR method is a method for calculating the MLR-type estimated minimum temperature at a target site based on the temperature difference between a reference site and a target site. The temperature difference between the reference site and the target site can be calculated by multiplying the elevation difference between the reference site and the target site by the moist adiabatic lapse rate (temperature difference = elevation difference x moist adiabatic lapse rate). The MLR-type estimated minimum temperature at the target site can be calculated by subtracting the temperature difference from the predicted minimum temperature at the reference site on the day for which the presence or absence of frost is to be predicted (hereinafter simply referred to as the "prediction day") (MLR-type estimated minimum temperature = predicted minimum temperature forecast for the reference site - temperature difference).

[0020] The LEM method is a method for calculating the LEM estimated minimum temperature at a target location based on the past correlation determined from the actual measured minimum temperature or predicted minimum temperature at a reference location for several past days and the corresponding actual measured minimum temperatures at the target location on each of the corresponding days.

[0021] By using the equation representing the regression line obtained from the sample group for past correlation, the LEM estimated minimum temperature at the target location can be calculated from the predicted minimum temperature forecasted for the reference location on the prediction date.

[0022] A sample group is made up of the actual minimum temperature measured at a reference site on a certain day in the past (e.g., January 1) and the actual minimum temperature measured at a target site on the same day (e.g., January 1), and is made up of samples collected for multiple days. The more samples that make up the sample group, the easier it is to obtain a regression line with a strong correlation between the predicted minimum temperature and the LEM-estimated minimum temperature.

[0023] The sample may be the predicted minimum temperature forecasted for a reference location on a certain date in the past (e.g., January 1) and the actual minimum temperature measured at a target location on the same date (e.g., January 1). With this configuration, it is possible to calculate the LEM-based estimated minimum temperature that takes into account the correlation between the predicted minimum temperature at the reference location and the actual minimum temperature at the target location.

[0024] The accuracy of frost prediction can be improved by determining the past correlation from an extracted sample group obtained by extracting samples of days that fall under conditions that are likely to cause condensation or frost from the sample group. Note that a high accuracy of frost prediction in the present invention means that the proportion of days on which frost occurred that were predicted not to occur is small.

[0025] Conditions that make condensation and frost likely to occur include weather, wind speed, and rainfall. Frost is likely to occur when the weather is clear, when the wind speed is gentle, and the closer the precipitation is to zero, the more likely frost is to occur. It is considered that the more these conditions are met, the more likely condensation and frost are to occur. By finding a past correlation from an extracted sample group created by extracting samples of days that satisfy at least one of these conditions, it is possible to calculate the LEM-based estimated minimum temperature when at least one of the conditions that make condensation and frost likely to occur is met, regardless of the weather, wind speed, and rainfall on the predicted day in the target area. In this way, it is possible to predict the occurrence of frost by leaning toward the dangerous side, that is, conditions that are likely to cause frost, regardless of the weather, wind speed, and rainfall on the predicted day in the target area.

[0026] To extract the sample group, we used a first sensor that detects weather, wind speed, and rainfall, and a second sensor that measures temperature. For example, if the target area has undulating terrain such as high ground and depressions, we can obtain samples that are closer to conditions where frost is more likely to occur by placing the first sensor on the high ground and the second sensor in the depression.

[0027] For more details, if illuminance is used as an index to judge the weather, the illuminance tends to be higher on higher ground than in lower areas. By judging the conditions based on the illuminance on higher ground, the accuracy of frost prediction can be improved.

[0028] In addition, wind speeds tend to be faster on higher ground than in lower areas. By judging the conditions based on the wind speed on higher ground, the accuracy of frost detection can be improved.

[0029] In addition, when precipitation is used as an index to determine rainfall, the amount of precipitation tends to be greater on high ground than in low areas. By determining the condition based on the amount of precipitation on high ground, the accuracy of frost determination can be improved.

[0030] In addition, cold air tends to accumulate in depressions more easily than on higher ground, and the air temperature drops from the ground up. The minimum temperature tends to be lower in depressions than on higher ground. Measuring the temperature in depressions to obtain the actual minimum temperature can improve the accuracy of frost determination.

[0031] The first sensor is not limited to a configuration capable of detecting all of weather, wind speed, and rainfall, and may be a sensor that detects at least one of weather, wind speed, and rainfall, or a plurality of sensors with different detection targets may be used in combination. Also, the first sensor is only required to have at least a function of detecting data for determining conditions.

[0032] Furthermore, the accuracy of frost forecasts can be improved by finding the regression line from the lower limit sample group extracted from the sample group. The lower limit sample group is a group of samples extracted from which the measured minimum temperature at the target site is lower than the measured minimum temperature at the reference site and where there is a large difference in temperature between the measured minimum temperatures at the reference site and the measured minimum temperature at the target site, i.e. samples on the danger side. Using the regression line found from the lower limit sample group, it is possible to calculate the LEM-based estimated minimum temperature closer to the danger side.

[0033] In the prediction step, frost on the leaves on the forecast date is predicted based on the MLR-based and LEM-based estimated minimum temperatures.

[0034] Based on the above, the frost prediction system of this embodiment can calculate the estimated minimum temperature at the target location on the predicted date (e.g., January 1st) and predict frost on the leaves at the target location on the predicted date (e.g., January 1st) if it can obtain the expected minimum temperature forecasted for a reference location prior to (e.g., December 30th) the predicted date (e.g., January 1st).

[0035] In addition, the highly accurate LEM estimated minimum temperature calculated by the LEM method can be used to predict frost on the leaves of the target area on the forecast date, making it possible to reliably prevent frost damage. EXAMPLES

[0036] The first embodiment will be described below.

[0037] As shown in Figure 1, the frost prediction system 1 of this embodiment predicts whether frost will occur on a predicted date in a field 2 as a target area from the expected minimum temperature forecasted for a reference area 3 on the predicted date, and is mainly composed of a frost determination server 10, a gateway 20, an MS (Meteorological Station) sensor 21 as a first sensor, and a depression sensor 22 as a second sensor.

[0038] The frost determination server 10 is connected to a public communication network 4 consisting of the Internet and a telephone network, and can communicate through the public communication network 4 with a weather forecast server 30, a cloud server 40, and a user terminal 50 of a user who operates the field 2.

[0039] The frost determination server 10 acquires data such as wind direction, wind speed, precipitation, illuminance, temperature, and humidity in the field 2 from the cloud server 40, and acquires data such as the expected minimum temperature and weather from the weather forecast server 30. In addition, the frost determination server 10 executes a frost determination process, which will be described later.

[0040] The gateway 20, the MS sensor 21, and the depression sensor 22 are arranged in the farm field 2. The farm field 2 is located in a mountainous area, and has a depression 2L that slopes downward in the southwest direction from a ridge-like high ground 2H. The altitude of the high ground 2H is about 710 m.

[0041] The gateway 20 and the MS sensor 21 are placed on the hill 2H. The depression sensor 22 is placed in the depression 2L, and is located southwest of the MS sensor 21. This is in anticipation of cold air flowing into the depression 2L in winter.

[0042] The gateway 20 is wirelessly connected to the public communications network 4, the MS sensor 21, and the depression sensor 22, and is intended to transmit various data input from the MS sensor 21 and the depression sensor 22 to the cloud server 40 via the public communications network 4.

[0043] The MS sensor 21 observes six items, namely wind direction, wind speed, precipitation, illuminance, temperature, and humidity, every 10 minutes and transmits the various data to the gateway 20. The observation device that observes the above data is installed at a position about 1.7 to 2.0 m above the ground on hill 2H.

[0044] The depression sensor 22 observes five items, namely air temperature, humidity, volumetric water content in the ground, electrical conductivity in the ground, and soil temperature, every hour, and transmits the various data to the gateway 20. The observation device that observes the air temperature and humidity data is installed at a position about 1.7 to 2.0 m above the ground in the depression 2L. The observation devices that observe the volumetric water content, electrical conductivity, and soil temperature data are partially buried in the ground in the depression 2L.

[0045] The weather forecast server 30 acquires various weather data from an observation device 31 installed at the reference site 3 via a public communication network 4, and publishes the weather, the measured minimum temperature, the predicted minimum temperature, etc. at the reference site 3. The reference site 3 is located in a plain area, with an altitude of about 10 m. The reference site 3 is also located about 40 km away from the farm field 2 in a straight line.

[0046] The cloud server 40 stores various data transmitted from the gateway 20 .

[0047] Next, a description will be given of the frost determination process performed by the frost determination server 10. The frost determination server 10 starts the frost determination process at 6 p.m. every day. The start time of the frost determination process may be changed as appropriate, and multiple times may be set in one day.

[0048] As shown in Figure 2, the frost determination process comprises an acquisition step S1 in which the predicted minimum temperature at the reference location 3 is obtained from the weather forecast server 30, an elevation difference equation estimation step S2 in which the MLR-type estimated minimum temperature of the field 2 is calculated using the MLR method, a correlation equation estimation step S3 in which the LEM-type estimated minimum temperature of the field 2 is calculated using the LEM method, a prediction step S4 in which the leaf surface temperature is calculated from the estimated minimum temperatures calculated in steps S2 and S3 and whether or not frost will occur on the leaf surface is determined, and a notification step S5 in which the determination result of the prediction step S4 is transmitted to the user terminal 50.

[0049] In an acquisition step S1, the frost determination server 10 acquires the predicted minimum temperature (predicted minimum temperature) at the reference location 3 for five days from one day (next day) to five days after the day on which the frost determination process is performed as the prediction day.

[0050] In the elevation difference estimation step S2, the predicted minimum temperatures are individually substituted into the elevation difference estimation formula to calculate the MLR-based estimated minimum temperatures MLR1 to MLR5 for five days. The elevation difference between the high ground 2H and the reference site 3 is about 700 m (700 = 710-10). The moist adiabatic lapse rate changes depending on the temperature and air pressure, and is about 0.5 degrees / 100 m when the average temperature near the ground on a sunny day is 10 to 20 degrees Celsius, and is 0.6 degrees / 100 m when the ground temperature is around 0 degrees Celsius, which is a condition where frost falls, so 0.6 degrees / 100 m was adopted. From these, the temperature difference between the high ground 2H and the reference site 3 is 4.2 degrees (4.2 = 700 × 0.6 / 100), and the following formula 1 was obtained as the elevation difference estimation formula of this embodiment. The moist adiabatic lapse rate does not have to be 0.6 degrees / 100 m, and may be changed as appropriate. MLR Estimated Minimum Temperature = Forecast Minimum Temperature - 4.2 (Equation 1)

[0051] To confirm the prediction accuracy of the MLR-based minimum temperature, an experiment was conducted over a period of 115 days. In this experiment, the MLR-based minimum temperature was calculated every day from 1 day to 5 days later using Equation 1.

[0052] Next, the difference between the actual minimum temperature measured at high ground 2H and the MLR-estimated minimum temperature calculated on the corresponding day was calculated for 1, 2, 3, 4, and 5 days later, and the average temperature error was obtained as shown in Table 1 below.

[0053] Specifically, if the day on which the actual minimum temperature was measured at high ground 2H was January 5th, then the difference between the actual minimum temperature measured on January 5th and the MLR-type estimated minimum temperature MLR5 estimated on December 31st was calculated, the difference between the actual minimum temperature measured on January 5th and the MLR-type estimated minimum temperature MLR4 estimated on January 1st was calculated, the difference between the actual minimum temperature measured on January 5th and the MLR-type estimated minimum temperature MLR3 estimated on January 2nd was calculated, the difference between the actual minimum temperature measured on January 5th and the MLR-type estimated minimum temperature MLR2 estimated on January 3rd was calculated, and the difference between the actual minimum temperature measured on January 5th and the MLR-type estimated minimum temperature MLR1 estimated on January 4th was calculated.

[0054] In addition, for the average temperature error one day later, the absolute value of the difference between the measured minimum temperature and the MLR formula estimated minimum temperature MLR1 estimated on the previous day (one day before) was obtained for 115 days, such as the absolute value of the difference between the actual minimum temperature measured on January 1st and the MLR formula estimated minimum temperature MLR1 for January 1st estimated on December 31st, and the absolute value of the difference between the actual minimum temperature measured on January 2nd and the MLR formula estimated minimum temperature MLR1 for January 2nd estimated on January 1st, and the average value was calculated.

[0055] Similarly, for the average temperature error two days later, the absolute value of the difference between the measured actual minimum temperature and the MLR-type estimated minimum temperature MLR2 estimated two days earlier was obtained for 115 days, and the average was calculated. For the average temperature error three days later, the absolute value of the difference between the measured actual minimum temperature and the MLR-type estimated minimum temperature MLR3 estimated three days earlier was obtained for 115 days, and the average was calculated. For the average temperature error four days later, the absolute value of the difference between the measured actual minimum temperature and the MLR-type estimated minimum temperature MLR4 estimated four days earlier was obtained for 115 days, and the average was calculated. For the average temperature error five days later, the absolute value of the difference between the measured actual minimum temperature and the MLR-type estimated minimum temperature MLR5 estimated five days earlier was obtained for 115 days, and the average was calculated.

[0056] As with upland 2H, the average temperature error was calculated by counting the difference between the actual minimum temperature measured in depression 2L and the MLR-estimated minimum temperature calculated on the corresponding day, divided into 1, 2, 3, 4, and 5 days later. Note that because the elevation difference between upland 2H and depression 2L is small, Equation 1 was also used for depression 2L.

[0057] These results are summarized in Table 1.

[0058] [Table 1]

[0059] As shown in Table 1, the average temperature error of the MLR-based estimated minimum temperature was less than 3 degrees, demonstrating high prediction accuracy. Additionally, the closer the prediction date gets, the smaller the average temperature error tends to be, and the higher the prediction accuracy becomes. This is thought to be because the accuracy of the predicted minimum temperature increases as the prediction date approaches.

[0060] In addition, while the average error in the temperature for the MLR-based estimated minimum temperature increases slightly the more different the environment is, such as the distance from the reference site and the difference in topographical characteristics from the reference site, good results were generally obtained unless the differences in the environments were extreme, such as when the reference site was on land and the target site was on water. For this reason, the MLR-based estimated minimum temperature is considered to be highly versatile.

[0061] In the correlation equation estimation step S3, the LEM-based estimated minimum temperature of field 2 is calculated based on the correlation estimation equation obtained from the past correlation between the past measured minimum temperatures at reference site 3 and the past measured minimum temperatures in depression 2L of field 2.

[0062] In this embodiment, a sample group consisting of samples for 169 days was obtained to obtain the correlation estimation equation. Each sample has the x-axis value representing the actual lowest temperature measured at the reference site 3 on a certain day (e.g., January 1), and the y-axis value representing the actual lowest temperature measured at the field 2 by the depression sensor 22 on the same day (e.g., January 1).

[0063] These samples were filtered based on weather, wind speed, and rainfall conditions, and 49 days' worth of samples were obtained that met all conditions.

[0064] Regarding the detailed judgment of each condition, the judgment was made acceptable if the weather was clear from sunset to dawn, and more specifically, if the average illuminance was 2.4 kLux or more. This judgment was made with reference to the hourly weather information issued by the Japan Meteorological Agency and the illuminance detected by the MS sensor 21. Note that, although the criterion for judging whether the weather was clear was that the average illuminance was 2.4 kLux or more, this is not limited to this and may be changed as appropriate. The same applies to wind speed and precipitation, which will be described later.

[0065] The wind speed was judged as acceptable if it was calm throughout the night, more specifically, if it was 1 m / s or less. This judgment was based on the wind speed detected by the MS sensor 21.

[0066] Rainfall was judged as acceptable if there was no precipitation throughout the night, more specifically, if the amount of precipitation was 0.5 mm or less. This judgment was based on the amount of precipitation detected by the MS sensor 21.

[0067] Furthermore, in order to calculate the LEM-type estimated minimum temperature on the more dangerous side, samples were extracted from the extracted 49-day samples in which the measured minimum temperature in depression 2L in field 2 was lower than the measured minimum temperature in reference site 3, and the difference between the measured minimum temperature in reference site 3 and the measured minimum temperature in depression 2L in field 2 was large. More specifically, samples in which the measured minimum temperature in depression 2L was lower than the measured minimum temperature in reference site 3, and the difference was -4.2 degrees or less (measured minimum temperature in depression 2L - measured minimum temperature in reference site 3 ≦ -4.2) were extracted. The difference between the measured minimum temperature in depression 2L and the measured minimum temperature in reference site 3, -4.2 degrees or less, is a condition based on the temperature difference between high ground 2H and reference site 3 described above, and is intended to enable frost prediction on the more dangerous side than the MLR-type estimated minimum temperature. In addition, when multiple measured minimum temperatures in depression 2L in field 2 corresponded to one measured minimum temperature in reference site 3, the lowest measured minimum temperature among them was adopted. As a result, a lower limit sample group consisting of seven lower limit samples was obtained, as shown in Figure 3.

[0068] The LEM-estimated minimum temperature can be calculated by substituting the predicted minimum temperature on the forecast date as the x-axis value into the equation representing the lower limit collapse line (regression line) obtained from the lower limit sample group. In other words, the equation representing the lower limit collapse line is the correlation estimation equation of this embodiment, and is as shown in the following Equation 2. LEM formula estimated minimum temperature = 0.85 x predicted minimum temperature - 2.5 (Equation 2)

[0069] In correlation equation estimation step S3, the predicted minimum temperatures for five days are individually substituted into equation 2 to calculate the LEM-equation estimated minimum temperatures LEM1 to LEM5 for the five days.

[0070] In order to confirm the prediction accuracy of the minimum temperature estimated by the LEM method, an experiment was conducted over a period of 115 days. Explanations that overlap with those of the MLR method will be omitted or simplified. In this experiment, a correlation estimation equation for depression 2L (Equation 2) and a correlation estimation equation for high ground 2H (see Equation 3 below) were created in a similar manner. Equation 3 was created for the experiment, and is not used in the frost determination process. LEM estimated minimum temperature = 0.83 x predicted minimum temperature - 1.6 (Equation 3)

[0071] In this experiment, the LEM-estimated minimum temperature using Equation 2 and the LEM-estimated minimum temperature using Equation 3 were calculated every day from 1 day to 5 days later.

[0072] Next, the difference between the actual minimum temperature measured at high ground 2H and the LEM-estimated minimum temperature calculated using Equation 3 assuming high ground 2H on the corresponding day was tabulated for 1, 2, 3, 4, and 5 days later to obtain the average temperature error. In addition, the difference between the actual minimum temperature measured at depression 2L and the LEM-estimated minimum temperature calculated using Equation 2 assuming depression 2L on the corresponding day was tabulated for 1, 2, 3, 4, and 5 days later to obtain the average temperature error.

[0073] These results are summarized in Table 2.

[0074] [Table 2]

[0075] As shown in Table 2, the average temperature error of the LEM-based estimated minimum temperature was less than 2 degrees, demonstrating high prediction accuracy. Additionally, the closer the prediction date, the smaller the average temperature error tends to be, and the higher the prediction accuracy. This is thought to be because the accuracy of the predicted minimum temperature increases as the prediction date approaches.

[0076] In addition, the average error in the minimum temperature estimated by the LEM method was smaller than that in the minimum temperature estimated by the MLR method. This suggests that the minimum temperature estimated by the LEM method has higher prediction accuracy than the minimum temperature estimated by the MLR method.

[0077] The LEM method estimated minimum temperature is based on past correlations, i.e., it statistically reflects differences in distance from a reference site, altitude, topographical characteristics, etc.; therefore, although it takes time to obtain sufficient samples, it can reflect the effects of distance from a reference site, topographical characteristics, etc. that cannot be reflected by the MLR method estimated minimum temperature.

[0078] In the prediction step S4, the leaf surface temperatures TsMLR1-TsMLR5, TsLEM1-TsLEM5 for each predicted day are calculated by individually substituting the estimated minimum temperatures MLR1-MLR5, LEM1-LEM5 for each predicted day into the leaf surface temperature calculation formula. The leaf surface temperature calculation formula uses a well-known nighttime heat balance formula, so a detailed explanation will be omitted, but the only variable is the estimated minimum temperature, and values ​​such as relative humidity, effective input radiation, evaporation efficiency, exchange rate, and black body index are set to fixed values ​​on the dangerous side.

[0079] Then, it is determined whether each of the leaf surface temperatures TsMLR1 to TsMLR5, TsLEM1 to TsLEM5 is equal to or less than zero. If at least one of the leaf surface temperatures derived from the RMF method and the leaf surface temperature derived from the LEM method for the same prediction date (for example, January 1st) is equal to or less than zero, it is determined that frost will occur on that prediction date, and if both of these are above zero, it is determined that frost will not occur on that prediction date.

[0080] To confirm the accuracy of frost prediction, an experiment was conducted over a period of 115 days. Explanations that overlap with the MLR method will be omitted or simplified. In this experiment, the leaf surface temperature derived from the MLR method was calculated every day from 1 day to 5 days later, and the leaf surface temperature derived from the LEM method was calculated from 1 day to 5 days later, and the leaf surface temperature was calculated from the actual minimum temperature obtained on that day.

[0081] In addition, the leaf surface temperature derived from the LEM method was calculated as the leaf surface temperature derived from Equation 2 and the leaf surface temperature derived from Equation 3. The leaf surface temperature calculated from the actual minimum air temperature was obtained by substituting the actual minimum air temperature measured at high ground 2H into the leaf surface temperature calculation formula, and by substituting the actual minimum air temperature measured at depression 2L into the leaf surface temperature calculation formula.

[0082] Next, the difference between the leaf surface temperature at high ground 2H calculated from the actual minimum air temperature and the leaf surface temperature derived from the MLR method calculated on the corresponding day was tabulated for 1, 2, 3, 4, and 5 days later to obtain the average leaf surface temperature error. Also, the difference between the leaf surface temperature at low ground 2L calculated from the actual minimum air temperature and the leaf surface temperature derived from the MLR method calculated on the corresponding day was tabulated for 1, 2, 3, 4, and 5 days later to obtain the average leaf surface temperature error.

[0083] Similarly, the difference between the leaf surface temperature of high ground 2H calculated from the actual minimum air temperature and the leaf surface temperature of high ground 2H derived from the LEM method calculated on the corresponding day was tabulated for 1, 2, 3, 4, and 5 days later to obtain the average leaf surface temperature error. Also, the difference between the leaf surface temperature of depression 2L calculated from the actual minimum air temperature and the leaf surface temperature of depression 2L derived from the LEM method calculated on the corresponding day was tabulated for 1, 2, 3, 4, and 5 days later to obtain the average leaf surface temperature error.

[0084] These results are summarized in Table 3.

[0085] [Table 3]

[0086] As shown in Table 3, the leaf surface temperature derived from the MLR method had a high prediction accuracy, with an average leaf surface temperature error of less than 3 degrees. In addition, the closer the prediction date, the smaller the average leaf surface temperature error tends to be, and the higher the prediction accuracy.

[0087] In addition, the leaf surface temperature derived from the LEM method had an average leaf surface temperature error of less than 2 degrees, indicating high prediction accuracy. In addition, the closer the prediction date gets, the smaller the average leaf surface temperature error tends to be, indicating that the prediction accuracy increases.

[0088] Figure 4 is a distribution diagram in which the leaf surface temperature derived from the MLR method is the x-axis value, and the leaf surface temperature at 2H high ground calculated from the actual minimum temperature on the same day is the y-axis value. As shown in Figure 4, no data was obtained in which the leaf surface temperature calculated from the actual minimum temperature was a negative value, even though the leaf surface temperature derived from the MLR method was a positive value. In other words, it was found that there were no days on which frost occurred, despite the prediction that no frost would occur.

[0089] Figure 5 is a distribution diagram in which the leaf surface temperature derived from the MLR method is the x-axis value, and the leaf surface temperature in depression 2L calculated from the actual minimum air temperature on the same day is the y-axis value. As shown in Figure 5, no data was obtained in which the leaf surface temperature calculated from the actual minimum air temperature was a negative value, even though the leaf surface temperature derived from the MLR method was a positive value. In other words, it was found that there were no days on which frost occurred, despite the prediction that no frost would occur.

[0090] In addition, in this experiment, the number of days when the leaf surface temperature derived from the MLR method was lower than the leaf surface temperature at high ground 2H calculated from the actual minimum temperature on the same day, that is, the number of days distributed above the solid line y = x in Figure 4, was 98 days out of 115 (approximately 85%) (see Figure 4). In addition, the number of days when the leaf surface temperature derived from the MLR method was lower than the leaf surface temperature at low ground 2L calculated from the actual minimum temperature on the same day was 82 days out of 115 (approximately 61%) (see Figure 5). This shows that the leaf surface temperatures derived from the MLR method tend to be lower than the leaf surface temperatures calculated from the actual minimum temperatures in both high ground 2H and low ground 2L.

[0091] As shown above, it was found that the leaf surface temperature derived from the MLR method can accurately predict frost, whether it is on the high ground 2H or in the depression 2L.

[0092] Figure 6 is a distribution diagram in which the leaf surface temperature at high ground 2H derived from the LEM method is taken as the x-axis value, and the leaf surface temperature at high ground 2H calculated from the actual minimum air temperature on the same day is taken as the y-axis value. As shown in Figure 6, there was a small amount of data in which the leaf surface temperature calculated from the actual minimum air temperature was negative, even though the leaf surface temperature derived from the LEM method was a positive value. This is thought to be because not enough samples were obtained.

[0093] Figure 7 is a distribution diagram in which the leaf surface temperature in depression 2L derived from the LEM method is taken as the x-axis value, and the leaf surface temperature in depression 2L calculated from the actual measured minimum air temperature on the same day is taken as the y-axis value. As shown in Figure 7, no data was obtained in which the leaf surface temperature calculated from the actual measured minimum air temperature was a negative value, even though the leaf surface temperature derived from the LEM method was a positive value. In other words, it was found that there were no days on which frost occurred, despite the prediction that no frost would occur.

[0094] Furthermore, in this experiment, there were 57 days out of 115 (approximately 50%) when the leaf surface temperature on high ground 2H derived from the LEM method was lower than the leaf surface temperature on high ground 2H calculated from the actual minimum temperature on the same day (see Figure 6). There were 50 days out of 115 (approximately 43%) when the leaf surface temperature on depression 2L derived from the LEM method was lower than the leaf surface temperature on depression 2L calculated from the actual minimum temperature on the same day (see Figure 7). This shows that the percentage of leaf surface temperatures derived from the LEM method that were lower and higher than the leaf surface temperatures calculated from the actual minimum temperature measured at the corresponding point tended to be almost the same.

[0095] As shown above, it was found that the leaf surface temperature derived from the LEM method can accurately predict frost, especially in depression 2L, where frost is thought to be more likely to occur.

[0096] 8, in notification step S5, an email summarizing the determination results in prediction step S4 from one day later to five days later is created and transmitted to user terminal 50. This allows the user to know the frost prediction for up to five days later. Therefore, the frost prediction system 1 of this embodiment can secure a period for the user to deal with frost.

[0097] In addition, the email will contain both the leaf surface temperatures derived from the MLR method and the leaf surface temperatures derived from the LEM method for each forecast date, allowing users to check the leaf surface temperatures derived from the LEM method and the MLR method and decide whether or not to take measures against frost.

[0098] As explained above, in this embodiment, both Equation 1, which is the elevation difference estimation equation, and Equation 2, which is the correlation estimation equation, are linear functions, and when the predicted minimum temperature falls below approximately 11.3°C, the minimum temperature estimated by the MLR equation is lower than the minimum temperature estimated by the LEM equation. This suggests that in this embodiment, the effects of differences in elevation, distance, topographical characteristics, etc. between the target field 2 and the reference field 3 were smaller than the effects of the elevation difference.

[0099] For example, if the elevation difference estimation formula is formula 1 and the correlation estimation formula is formula 4 below, when the predicted minimum temperature falls below 3 degrees, the minimum temperature estimated by the LEM formula will be lower than the minimum temperature estimated by the MLR formula. In such a case, it is considered that the influence of the differences in elevation, distance, topographical characteristics, etc. between the target site and reference site 3 was greater than the influence of the elevation difference. Estimated minimum temperature = 0.85 × predicted minimum temperature - 3.75 (Equation 4)

[0100] From the above, by using the LEM-based estimated minimum temperature and the MLR-based estimated minimum temperature together, the occurrence of frost that cannot be predicted from one estimated minimum temperature can be predicted from the other estimated minimum temperature, thereby improving the accuracy of frost forecasts.

[0101] In addition, the correlation estimation formula of this embodiment can be updated to a new correlation estimation formula obtained based on the increased sample group. By updating the correlation estimation formula, the average temperature error of the LEM-estimated minimum temperature is expected to decrease, improving accuracy. For this reason, by updating the correlation estimation formula, the leaf surface temperature derived from the LEM method is also expected to have a smaller average leaf surface temperature error, improving accuracy of frost prediction.

[0102] In addition, although the MLR-based estimated minimum temperature does not update Equation 1, as mentioned above, there were no days when frost occurred even though it was predicted that no frost would occur, whether on high ground 2H or in depression 2L, and the accuracy of the frost prediction is high.

[0103] Based on these findings, it is believed that in the short term, when not enough days have passed since the start of the frost determination process, the leaf surface temperature derived from the MLR method is more accurate in predicting frost and therefore more reliable.

[0104] In contrast, after a sufficient number of days have passed since the frost determination process was started, the sum of the probability that frost will occur when frost is predicted to occur and the probability that frost will not occur when frost is predicted to not occur, i.e., the prediction accuracy rate, will increase, and it is thought that the reliability will increase.

[0105] The prediction accuracy rate of the leaf surface temperature derived from the MLR method and the prediction accuracy rate of the leaf surface temperature derived from the LEM method may be included in the email. With this configuration, the prediction accuracy rate of the leaf surface temperature derived from the MLR method and the LEM method may be used as indicators for determining whether or not the user should take measures against frost.

[0106] Furthermore, since the leaf surface temperature derived from the LEM method is calculated as a leaf surface temperature on the danger side regardless of the weather, wind speed, and rainfall in the target location on the predicted day, for example, in notification step S5, information on the weather, wind speed, and rainfall in the target location on the predicted day may be obtained from the weather forecast server 30, and the reliability of the leaf surface temperature derived from the LEM method may be recorded. Specifically, if the predicted weather in the target location is sunny, the wind speed is 1 m / s or less, and there is no rainfall, it may be displayed as "high reliability", if any one of them does not apply, "medium reliability", if any two of them do not apply, "low-medium reliability", and if none of them apply, "low reliability", etc.

[0107] Furthermore, in a configuration in which steps S2 and S3 are used in combination as in this embodiment, information on the weather, wind speed, and rainfall in the target location on the prediction date is obtained from the weather forecast server 30, and if one or more of these do not apply, the leaf surface temperature derived from the LEM method may not be included in the email, and the LEM-estimated minimum temperature or the leaf surface temperature derived from the LEM method may not be calculated. EXAMPLES

[0108] Second Embodiment Next, a frost prediction system according to a second embodiment will be described with reference to Fig. 9 to Fig. 11. Note that the same components as those shown in the first embodiment will be denoted by the same reference numerals and duplicated description will be omitted.

[0109] 9, the frost prediction system of this embodiment includes an update process for updating the correlation estimation formula for calculating the LEM-based estimated minimum temperature of a target area. This update process is executed by the frost determination server 10 in this embodiment, but is not limited to this and may be executed by another server, and the execution device may be changed as appropriate.

[0110] The frost determination server 10 starts the update process at 10:00. In this embodiment, the update process is performed once a day, but is not limited to this and may be performed multiple times a day, or may be performed once every multiple days (for example, every three days). The start time of the update process may be changed as appropriate, and may be performed at predetermined intervals.

[0111] The update process comprises an MS data acquisition step Sr1, an MS data comparison step Sr2, a reference site actual minimum temperature acquisition step Sr3, a target site actual minimum temperature acquisition step Sr4, a lower limit sample determination step Sr5, a storage step Sr6, a set date determination step Sr7, and an update step Sr8.

[0112] In the MS data acquisition step Sr1, the wind speed, precipitation, and illuminance for 2H of high ground from 8 PM of the previous day to 9 AM of the same day are acquired from the cloud server 40.

[0113] In the MS data comparison step Sr2, based on the wind speed, precipitation, and illuminance acquired in the MS data acquisition step Sr1, it is determined whether the average illuminance is 2.4 kLux or more, whether the wind speed is 1 m / s or less, and whether the precipitation is 0.5 mm or less.

[0114] If the average illuminance is 2.4 kLux or more, the wind speed is 1 m / s or less, and the precipitation is 0.5 mm or less (Sr2: Yes), the process proceeds to step Sr3 for acquiring the reference site's actually measured minimum temperature. On the other hand, if the average illuminance is less than 2.4 kLux, the wind speed is more than 1 m / s, or the precipitation is more than 0.5 mm (Sr2: No), the process proceeds to step Sr7 for determining the set date.

[0115] In the reference location measured minimum temperature acquisition step Sr3, the measured minimum temperature in the time period from 0:00 to 8:00 on the same day at the reference location 3 is acquired from the weather forecast server 30. Note that the time period from 0:00 to 8:00 set in this step Sr3 may be the same time period as in the target location measured minimum temperature acquisition step Sr4, or may be changed as appropriate.

[0116] In target location actual minimum temperature acquisition step Sr4, the actual minimum temperature in the depression 2L from 2:00 to 8:00 on the same day is acquired from the cloud server 40. Note that the time period from 2:00 to 8:00 set in this step Sr4 is an assumed time period before dawn, but may be changed as appropriate.

[0117] In the lower limit sample determination step Sr5, it is determined whether the temperature difference calculated by subtracting the measured minimum temperature at reference location 3 obtained in reference location measured minimum temperature acquisition step Sr3 from the measured minimum temperature at depression 2L obtained in target location measured minimum temperature acquisition step Sr4 is -4.2 degrees or less (measured minimum temperature at depression 2L - measured minimum temperature at reference location 3 ≦ -4.2).

[0118] If the temperature difference is -4.2 degrees or less (Sr5: Yes), the measured minimum temperature at reference site 3 acquired in this reference site measured minimum temperature acquisition step Sr3 and the measured minimum temperature at depression 2L acquired in target site measured minimum temperature acquisition step Sr4 are stored as new lower limit samples in a storage area (e.g., a hard disk drive) of frost determination server 10 (Sr6). The lower limit sample may be stored in cloud server 40, and the storage area may be changed as appropriate.

[0119] On the other hand, if the temperature difference is greater than -4.2 degrees (Sr5: No), the process proceeds to the setting date determination step Sr7 without storing a new lower limit sample.

[0120] In the set date determination step Sr7, it is determined whether or not the current day is a preset set date (for example, December 31 or April 30). The set date may be once a year (for example, only April 30), or may be every other month, and may be changed as appropriate, including the number of times and the date.

[0121] If the current day is the set date (Sr7: Yes), the process proceeds to update step Sr8. On the other hand, if the current day is not the set date (Sr7: No), the update process for the current day ends.

[0122] In the update step Sr8, a lower limit sample group consisting of all the lower limit samples stored in the storage area is acquired. A new correlation estimation formula is acquired from the new lower limit sample group, and this correlation estimation formula is updated as the correlation estimation formula to be used in the correlation equation estimation step S3. After that, the update process for the day is terminated.

[0123] A new experiment was conducted to confirm the advantages of having an update process. In this experiment, the LEM-estimated minimum temperature using the correlation estimation formula (Formula 2) before the update and the LEM-estimated minimum temperature using the correlation estimation formula (Formula 5) after the update were calculated from 1 day to 5 days later for 44 days.

[0124] When an experiment was carried out using Equation 2 before it was updated by the update process, the results shown in Figure 10 were obtained. Figure 10 is a distribution diagram in which the leaf surface temperature in depression 2L derived from the LEM method using Equation 2 is shown as the x-axis value, and the leaf surface temperature in depression 2L calculated from the actual minimum temperature on the same day is shown as the y-axis value. In Figure 10, 220 pieces of data (hereinafter simply referred to as "data") consisting of an x-axis value and a y-axis value are shown (44 days x 5 days (1 day later to 5 days later)). Note that in Figures 10 and 11, the data is shown as 1 day earlier to 5 days earlier, with the day on which the actual minimum temperature was measured as the base day.

[0125] In this experiment, the leaf surface temperature calculated from the minimum measured air temperature was negative, even though the leaf surface temperature derived from the LEM method was positive. In other words, there were days when frost occurred, even though it was predicted that frost would not occur.

[0126] In addition, in this experiment, there were 155 out of 220 data points (approximately 70%) in which the leaf surface temperature in depression 2L derived from the LEM method was lower than the leaf surface temperature in depression 2L calculated from the actual minimum air temperature on the same day.

[0127] Next, the experimental results using the correlation estimation formula updated by the update process will be described. First, the following formula 5 was obtained as the correlation estimation formula updated by the above-mentioned update process. LEM estimated minimum temperature = 1.02 x predicted minimum temperature - 5.7 (Equation 5)

[0128] When an experiment was conducted using Equation 5 updated by the update process, the results shown in Figure 11 were obtained. Figure 11 is a distribution diagram in which the leaf surface temperature in depression 2L derived from the LEM method using Equation 5 is the x-axis value, and the leaf surface temperature in depression 2L calculated from the actual minimum air temperature on the same day is the y-axis value. In Figure 11, 220 pieces of data are shown, just like in Figure 10.

[0129] In this result, although the leaf surface temperature derived from the LEM method was a positive value, no data was obtained in which the leaf surface temperature calculated from the actually measured minimum air temperature was a negative value. In other words, it was found that there were no days when frost actually occurred, despite the prediction that no frost would occur.

[0130] In addition, in this experiment, there were 214 out of 220 data points (approximately 97%) in which the leaf surface temperature in depression 2L derived from the LEM method was lower than the leaf surface temperature in depression 2L calculated from the actual minimum air temperature on the same day.

[0131] These experimental results confirmed the advantages of updating the correlation estimation equation.

[0132] Although the embodiments of the present invention have been described above with reference to the drawings, the specific configurations are not limited to these embodiments, and additions and modifications that do not deviate from the gist of the present invention are also included in the present invention.

[0133] For example, in the first and second embodiments, the correlation equation estimation step is configured to obtain the LEM-based estimated minimum temperature of the target area by substituting the predicted minimum temperature of the reference area into the lower limit collapse line, but the present invention is not limited to this, and the LEM-based estimated minimum temperature of the target area may be obtained by consulting the predicted minimum temperature of the reference area in a table that summarizes past correlations, and the configuration may be changed as appropriate. The same applies to the elevation difference estimation step.

[0134] In addition, in the above-mentioned Examples 1 and 2, the past correlation has been described as being represented by the equation of the lower limit collapse line, but this is not limited to this and may be a regression line of a sample group, or a regression line of an extracted sample group extracted according to at least one condition of weather, wind speed, and rainfall, or may be represented by a regression line other than the lower limit collapse line.

[0135] In addition, in the above-mentioned Examples 1 and 2, the target land has been described as having a terrain including high ground and depressions, but this is not limited to this and may be flat land or an alluvial fan, and the terrain characteristics may be changed as appropriate.

[0136] Further, in the first and second embodiments, the correlation equation estimating step and the altitude difference equation estimating step are included. However, the present invention is not limited to this, and only the correlation equation estimating step may be included.

[0137] In addition, in the above-mentioned Examples 1 and 2, a configuration has been described in which frost prediction is performed for a period of five days, but this is not limited to this, and it may be performed for only one day later, only two days later, or for three days, or this may be changed as appropriate.

[0138] In addition, in the above-mentioned Examples 1 and 2, a configuration was described in which the leaf surface temperature is calculated using a leaf surface temperature calculation formula to determine whether frost has formed on the leaf surface. However, this is not limited to this, and it may be determined that frost will occur if the estimated minimum air temperature is below a predetermined value, and the method of determining whether frost has formed on the leaf surface may be changed as appropriate.

[0139] In addition, in the above-mentioned first and second embodiments, the frost determination server has been described as being configured to be able to acquire weather data for the target area from a cloud server connected via a public communications network, but this is not limited to this, and the frost determination server may be directly connected to the gateway via a wired or wireless connection, the frost determination server may also function as a gateway and may also function as a weather forecast server, and the configuration of the frost determination server may be changed as appropriate. [Explanation of symbols]

[0140] 1. Frost prediction system 2. Field (target area) 2H High ground 2L depression 3 Reference location 4 Public communication network 10 Frost Judgment Server 21 MS sensor (first sensor) 22 Depression sensor (second sensor) 30 Weather forecast server 40 Cloud Server 50 User terminals S1 Acquisition step S2 Elevation difference estimation step S3 Correlation equation estimation step S4 Prediction step S5 Notification Step

Claims

1. a correlation equation estimation step of storing a past correlation between the actual minimum temperature or the predicted minimum temperature at a reference location for a plurality of past days and the actual minimum temperature at a target location for each corresponding day, and estimating an estimated minimum temperature at the target location on a prediction date from the predicted minimum temperature at the reference location on the prediction date and the past correlation; a prediction step of predicting frost on the leaf surfaces on the predicted day from the estimated minimum temperature estimated by the correlation equation estimation step.

2. The frost prediction system according to claim 1, characterized in that the past correlation is formed using a regression line of a lower limit sample group among a sample group consisting of samples of the actual minimum temperature or predicted minimum temperature at the reference site and the actual minimum temperature at the target site for each day.

3. 3. The frost prediction system according to claim 2, wherein the sample is taken from a day when at least one of the following conditions is satisfied: weather, wind speed, and rainfall.

4. 4. The frost prediction system according to claim 3, wherein the samples are taken on days when all the conditions of weather, wind speed, and rainfall are satisfied.

5. The frost prediction system described in claim 2, characterized in that a first sensor that detects at least one of weather, wind speed, and rainfall and a second sensor that measures air temperature are located in different locations, and the second sensor is located in a depression within the target area.

6. The frost prediction system further includes an elevation difference estimation step of estimating an estimated minimum temperature of the target location on the prediction date from an elevation difference between the reference location and the target location and a predicted minimum temperature of the reference location on the prediction date, 2. The frost prediction system according to claim 1, wherein the prediction step also predicts frost on the leaves on the predicted day based on the estimated minimum temperature estimated by the elevation difference estimation step.

7. 7. The frost prediction system according to claim 6, wherein the prediction step predicts that frost will occur on the leaf surface if at least one of the estimated minimum air temperature estimated by the correlation equation estimation step and the leaf surface temperature calculated from the estimated minimum air temperature estimated by the altitude difference equation estimation step satisfies a condition.

8. 8. The frost prediction system according to claim 1, wherein the frost prediction system predicts frost for several days.

Images