JP5876286B2 - Precipitation prediction method - Google Patents

Description

  The present invention relates to a precipitation prediction method.

  Weather prediction includes prediction of changes in atmospheric pressure, temperature, humidity, wind direction, wind speed, cloud cover, precipitation, and so on. Among them, precipitation prediction is one of the weather prediction information that can be predicted, such as precipitation region and precipitation intensity (Patent Document 1). Particularly in recent years, damage due to local heavy rains has also occurred, and not only prediction information for changes in precipitation regions, but also prediction information for changes in regions with high precipitation intensity is important.

  As shown in FIG. 3A, the conventional precipitation prediction method is obtained from the precipitation region 200 obtained from the precipitation distribution data and the data after a predetermined time has elapsed in the specific region 100 that is the observation target region. The movement vector M is obtained based on the precipitation region 201. Then, as shown in FIG. 3B, based on the movement vector M, a predicted precipitation region 202 after a predetermined time has elapsed is created. Here, the precipitation distribution data indicates precipitation intensity (precipitation) in the specific region 100.

JP-A-9-189773

  The precipitation region 200 includes a portion (strong precipitation region 200A) that shows a locally strong rainfall. In the conventional precipitation prediction method, the prediction precipitation region 202 is created with the precipitation region 200 and the strong precipitation region 200A as one unit. Thereby, it is predicted that the strong precipitation region 200 </ b> A moves based on the movement vector M of the precipitation region 200 after the predetermined time has elapsed. However, the movement vector M of the precipitation region 200 and the movement vector of the strong precipitation region 200A do not necessarily match. For this reason, the predicted position of the strong precipitation region 200A may be different from the actual position, and there is a problem that the accuracy of the predicted position of the strong precipitation region 200A is not sufficient.

  Therefore, in view of such circumstances, an object of the present invention is to provide a precipitation prediction method capable of performing the position prediction of a localized precipitation region with high accuracy, for example, showing a locally strong rainfall. And

  In order to achieve the above object, the precipitation prediction method of the present invention specifies a precipitation region position after a lapse of a predetermined time based on a precipitation region indicating the entire precipitation location and a movement vector of the precipitation region, and In the region, a local precipitation region position after a predetermined time has been identified based on a local precipitation region having a horizontal scale smaller than the precipitation region and a movement vector of the local precipitation region, and a predetermined time It combines the precipitation area after the passage and the local precipitation area.

  According to the above method, the prediction of the movement of the precipitation region and the prediction of the movement of the local precipitation region can be performed separately. Therefore, the position of the predicted precipitation region is the actual local region. It is consistent with the position of the precipitation area.

  In the above configuration, the precipitation area and the local precipitation area may be a precipitation distribution observed by a weather radar. Thereby, the movement of the precipitation region and the local precipitation region can be observed at short time intervals.

  In the above configuration, by performing wavelet transform on the precipitation distribution data, a local precipitation area can be extracted from the precipitation area in the precipitation distribution data.

  In the above configuration, the precipitation distribution data of the precipitation region and the local precipitation region observed by the weather radar, and a rain gauge distributed and arranged at predetermined intervals in the precipitation region and the local precipitation region Using the observed rainfall, the precipitation distribution data in the precipitation region and the local precipitation region can be corrected based on the rainfall observed by the rain gauge.

  The precipitation data from the rain gauge is closer to the actual precipitation than the precipitation data obtained from weather radar. For this reason, by correcting the data obtained from the weather radar based on the data acquired by the rain gauge, the precipitation data becomes closer to the actual precipitation.

Further, the predicted precipitation R1 (t) after t hours by the precipitation prediction method of the present invention as a kinematic method, the predicted precipitation R2 (t) after t hours by the physical method, and the physical method Based on the weight function C (t) (1 ≧ C (t) ≧ 0) after t hours of the predicted precipitation R1 (t) by the kinematic method with respect to the predicted precipitation R2 (t) by 1 is used to obtain a predicted precipitation amount R (t) after t hours, and as a weighting function C (t), at least a predicted precipitation amount after t hours by a kinematic method based on a topographic factor, a wind direction factor, and a wind speed factor. Different weighting functions C (t) are prepared in correspondence with different weightings between R1 (t) and the predicted precipitation amount R2 (t) after t hours by the physical method, and these prepared weighting functions C (t) can be selected and used.
(Equation 1)
R (t) = C (t) * R1 (t) + (1-C (t)) * R2 (t)

  Prediction of precipitation includes kinematic methods and physical methods. When the precipitation is predicted using the predicted precipitation data obtained by the kinematic method obtained by the present invention and the predicted precipitation data obtained by the physical method separately obtained, it is more accurate according to the actual situation. Precipitation can be predicted.

  Precipitation intensity after a lapse of a predetermined time in the precipitation area and / or the local precipitation area is predicted from the time change of the precipitation intensity in the precipitation area and / or the local precipitation area, or the precipitation area and / or the station From the temporal change of the movement vector of the local precipitation region, the precipitation region position and / or the local precipitation region position after a predetermined time of the precipitation region and / or the local precipitation region can be predicted. As a result, it is possible to accurately predict precipitation even when the precipitation changes or the movement speed and direction change with the movement of the precipitation region and the local precipitation region. Predict the location of the rain region.

  According to the present invention, since the prediction of the movement of the precipitation region and the prediction of the movement of the local strong precipitation region can be performed separately, the predicted position of the local strong precipitation region is the actual This is consistent with the location of the local heavy precipitation area, and the position of the local heavy precipitation area can be predicted with high accuracy.

It is a figure which shows the outline of the precipitation prediction method of this invention, (a) is a figure which shows precipitation distribution, (b1) is a figure which shows the low frequency component of precipitation distribution, (b2) is a figure which shows a high frequency component, (C1) is the figure which showed the prediction position of a low frequency component, (c2) is the figure which showed the prediction position in a high frequency component, (d) is a figure which shows the precipitation distribution which synthesize | combined each component after prediction. (A) is a figure explaining the outline of a wavelet transformation, (b) is a figure explaining the outline of a Fourier transformation. It is a figure which shows the outline of the conventional precipitation prediction method, (a) shows the precipitation area | region obtained from precipitation data, (b) is the figure which added the prediction precipitation area | region.

  Hereinafter, a precipitation prediction method according to a first embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a diagram showing an outline of the precipitation prediction method of the present invention. FIG. 1A shows a precipitation distribution observed by a weather radar (not shown) in the specific region 10. The specific area 10 is an area to be observed by the weather radar, and can be set as appropriate, for example, the whole of Japan or some local areas. Weather radars are installed in various parts of Japan and emit radio waves (microwaves) while rotating antennas to observe rain and snow that exist within a radius of tens to hundreds of kilometers. . With this weather radar, precipitation data in the specific area 10 can be obtained at intervals of about 1 to 5 minutes. In addition, although precipitation is a general term for rain, snow, hail, etc., in the following description, an example of rain will be described.

  In FIG. 1A, 20 (T1) indicates a precipitation region indicating the entire precipitation position at time T1, and 20 (T2) indicates a precipitation region at time T2 after a predetermined time has elapsed from T1. That is, the precipitation regions 20 (T1) and 20 (T2) indicate the entire region where it is raining at times T1 and T2.

  Further, 21a (T1) and 21b (T1) are local precipitation areas having a horizontal scale smaller than the precipitation area 20 in the precipitation area 20 at time T1, and 21a (T2) and 21b (T2) are T2 In the precipitation area 20, a local precipitation area having a horizontal scale smaller than the precipitation area 20. Here, the local precipitation region 21 indicates a region having a large amount of precipitation locally in the precipitation region 20, that is, a region having a high precipitation intensity. The horizontal scale is a scale indicating the magnitude of the meteorological phenomenon.

  As shown in FIG. 1 (a), the precipitation region 20 is moving with the movement of the rain clouds from time T1 to T2. As the precipitation area 20 moves, the local precipitation area 21 in the precipitation area 20 also moves.

  Next, a method for predicting the movement of the precipitation region 20 and the local precipitation regions 21a and 21b in the specific region 10 by the precipitation prediction method of the present invention will be described. First, as shown in FIG. 1A, precipitation distribution data by a weather radar is acquired at time T1. Then, the local precipitation region 21 is extracted from the precipitation region 20 by performing wavelet transform on the data in T1, and the low frequency component indicating the entire precipitation region 20 and the high frequency component indicating the local precipitation region 21 are extracted. And decompose. Hereinafter, the low frequency component means the precipitation region 20, and the high frequency component means the local precipitation region 21.

Here, the wavelet transform is a technique for expressing a given waveform by enlarging / reducing or translating a basic wavelet function which is a small wave (wavelet). As shown in FIG. 2A, the complex function f can be expressed by a sum of a local component of a wavelet function ψ having a different scale s. An expression indicating the wavelet transform is shown in the following formula 2.

  On the other hand, the Fourier transform conventionally used is a method of expressing data having a complex waveform by a combination of simple waveforms as shown in FIG. The Fourier transform is premised on a periodic function, and if there is a discontinuous region in the function, it will not be converted correctly. Further, the high frequency component is detected as noise. For this reason, when a high frequency component exists locally in the low frequency component as in the present invention, it is preferable to perform wavelet transform because the low frequency component and the high frequency component can be appropriately decomposed.

  Similarly to the time T1, the precipitation distribution data at the time T2 can be decomposed into regions indicating the low-frequency component and the high-frequency component by performing wavelet transform.

  Thereby, in the precipitation distribution data of FIG. 1, a low frequency component indicating the entire precipitation region 20 as shown in FIG. 1 (b1) and a high frequency component indicating the local precipitation region 21 as shown in FIG. 1 (b2). And can be disassembled. That is, the local precipitation area 21 can be extracted from the precipitation area 20. Here, the regions 20 (T1) and 20 (T2) showing the low frequency components in FIG. 1 (b1) correspond to the precipitation regions 20 (T1) and 20 (T2) in FIG. The regions 21a (T1), 21b (T1), 21a (T2), and 21b (T2) showing the high frequency components of 1 (b2) are the local precipitation regions 21a (T1) and 21b (T1) in FIG. ), 21a (T2), 21b (T2).

  Subsequently, movement is predicted individually for each of the low-frequency component in FIG. 1 (b1) and the high-frequency component in FIG. 1 (b2). For the movement prediction of each component, a method similar to the conventional method can be used. That is, for the low frequency component, as shown in FIG. 1 (b1), the movement vector ML is obtained based on the movement of the precipitation regions 20 (T1) and 20 (T2) indicating the low frequency component. Based on this movement vector ML, the position of the low frequency component at time T3 after a predetermined time has elapsed is further specified. 20 (T3) in FIG. 1 (c1) indicates the predicted position of the low frequency component at time T3.

  Similarly, for the high frequency component, as shown in FIG. 1 (b2), based on the movement of the regions 21a (T1), 21a (T2), and 21b (T1), 21b (T2) indicating the high frequency components. Movement vectors MHa and MHb are obtained. Based on the movement vectors MHa and MHb, the position of the high frequency component at time T3 after a predetermined time has elapsed is specified. 21a (T3) and 21b (T3) in FIG. 1 (c2) indicate the predicted positions of the high frequency components at time T3.

  When there are two local precipitation areas (21a, 21b) that are high-frequency components as in the present embodiment, the local precipitation area 21a, the local precipitation area 21b, As another group, the movement vectors MHa and MHb are obtained in each region, but the two regions 21a and 21b may be combined into one group to obtain one movement vector using a high frequency component.

  Further, by combining the low frequency component shown in FIG. 1 (c1) and the high frequency component shown in FIG. 1 (c2), as shown in FIG. 1 (d), the position and station of the precipitation region 20 at time T3. A predicted precipitation distribution in which the positions of the local precipitation areas 21a and 21b are predicted can be obtained.

  In this way, since the precipitation distribution data is decomposed into a low-frequency component and a high-frequency component, and the position of each component can be predicted based on the movement vector of the region in each component, the predicted position of the high-frequency component is It becomes consistent with the position of the actual high frequency component, and the local strong precipitation position can be predicted with high accuracy. Thereby, the prediction about the movement of the local heavy rain region can be performed with high accuracy. Moreover, since precipitation data is decomposed into a low-frequency component and a high-frequency component by wavelet transformation, it can be accurately decomposed into a low-frequency component and a high-frequency component.

  Next, the precipitation prediction method of 2nd Embodiment of this invention is demonstrated. In the first embodiment, the precipitation distribution data is a value obtained from an observation value obtained by a weather radar. Further, the precipitation distribution data is calculated based on a value measured using a precipitation gauge such as a rain gauge. Data may be corrected. In other words, rain gauges are distributed at predetermined intervals in the precipitation area and the local precipitation area, and the precipitation is observed. Thereby, in this embodiment, the data by a weather radar and a precipitation meter can be obtained. Then, the precipitation distribution data of the weather radar is corrected based on the precipitation observed by the rain gauge. As the correction method, correction can be performed by the method described in Japanese Patent No. 3828316.

  The precipitation data from the rain gauge is closer to the actual precipitation than the precipitation data obtained from weather radar. For this reason, by correcting the data obtained from the weather radar based on the data obtained by the rain gauge, the precipitation data becomes closer to the actual precipitation, and more accurate precipitation measurement and prediction Can do.

  Next, the precipitation prediction method of the third embodiment of the present invention will be described. Prediction of precipitation includes kinematic methods and physical methods. The kinematic method is a method for predicting the movement, development and weakness of the rain region in the future from the past and present rain cloud conditions by weather radar etc., and the physical method is the time integration of the equation for predicting atmospheric phenomena. And predict the future atmospheric conditions. The precipitation prediction methods of the first and second embodiments correspond to kinematic methods. In this embodiment, the kinematic method and the physical method described above are combined.

As a method for predicting precipitation by a combination of a kinematic method and a physical method, a method described in Japanese Patent No. 3851641 can be employed. That is, the motion of the predicted precipitation R1 (t) after t hours by the kinematic method, the predicted precipitation R2 (t) after t time by the physical method, and the predicted precipitation R2 (t) by the physical method On the basis of the weight function C (t) after t hours of the predicted precipitation R1 (t) by the scientific method, the predicted precipitation R (t) after t hours is obtained from the following equation.
(Equation 1)
R (t) = C (t) * R1 (t) + (1-C (t)) * R2 (t)

  Actually, the prediction by the kinematic method may be closer to the actual measurement value, or the prediction by the physical method may be closer to the actual measurement value. Therefore, as the weighting function C (t), for example, the predicted precipitation R1 (t) after t time by the kinematic method and the predicted precipitation after t time by the physical method depending on the terrain factor, the wind direction factor and the wind speed factor. Different weighting functions C (t) are prepared in correspondence with different weightings from the quantity R2 (t), and these prepared weighting functions C (t) are selected and used. Thereby, the more accurate precipitation prediction according to the actual situation can be performed.

  In the first to third embodiments, the precipitation intensity after a lapse of a predetermined time may be predicted from the temporal change in precipitation (precipitation intensity) in the precipitation area 20 or the local precipitation area 21. For example, when the precipitation in this region gradually increases as the local precipitation region 21 moves, the precipitation in the local precipitation region 21 after a predetermined time has been predicted based on the change over time. be able to. Similarly for the precipitation region 20, the precipitation amount in the precipitation region 20 after the elapse of a predetermined time can be predicted from the temporal change in the precipitation amount. Further, not only when the precipitation amount increases but also when the precipitation amount decreases with the movement of the local precipitation region 21 and the precipitation region 20, the prediction can be similarly made.

  In this way, since the change in precipitation can be predicted, it is possible to accurately predict not only the change in position of the precipitation region 20 and the local precipitation region 21, but also the change in precipitation in that region. it can.

  Further, in the first to third embodiments, the precipitation region position and the local precipitation region position after a predetermined time are predicted from the time change of the movement vector of the precipitation region 20 and the local precipitation region 21. May be. For example, when the moving speed gradually increases or the direction changes as the local precipitation area 21 moves, the position of the local precipitation area 21 after a predetermined time has elapsed based on the change over time. Can be predicted. Similarly for the precipitation region 20, the position of the precipitation region after a predetermined time can be predicted from the time change of the movement vector. Moreover, it can predict similarly about the case where a moving speed reduces with the movement of the local precipitation area | region 21 or the precipitation area | region 20. FIG.

  In this way, the position of the precipitation region 20 and the local precipitation region 21 can be accurately predicted even when the moving speed and the moving direction change as the precipitation region 20 and the local precipitation region 21 move. .

  The embodiment of the present invention has been described above, but the present invention is not limited to the above-described embodiment, and it is needless to say that various modifications can be made without departing from the gist of the present invention. In the present embodiment, the prediction method in the case of rain as precipitation has been described. However, the present invention is not limited to this, and may be used for prediction of snow or the like. The entire region may be divided into a plurality of areas, and the movement vector may be obtained in each area. The precipitation distribution data may be obtained by other than weather radar. In the embodiment, the precipitation area 20 and the local precipitation area 21 are obtained simultaneously by the weather radar, but the precipitation area 20 and the local precipitation area 21 are obtained separately and independently. Also, the present invention can be applied.

10 Specific area 20 Precipitation area 21a, 21b Local precipitation area ML, MHa, MHb Movement vector

Claims (5)

Based on the precipitation area showing the whole precipitation position and the movement vector of the precipitation area, the position of the precipitation area after a predetermined time has passed,
In the precipitation area, based on the local precipitation area showing a horizontal scale smaller than the precipitation area, and the movement vector of the local precipitation area, the local precipitation area position after a predetermined time has passed,
A precipitation prediction method for synthesizing a precipitation region after a predetermined time and a local precipitation region ,
The precipitation region and the local precipitation region are precipitation distributions observed by a weather radar, and by performing wavelet transform on the precipitation distribution data, the precipitation region data is localized from the precipitation region. A precipitation prediction method characterized by extracting a precipitation region . Precipitation distribution data in the precipitation area and local precipitation area observed by the weather radar, and observation with a rain gauge distributed at predetermined intervals along the path along which the precipitation area and local precipitation area move The precipitation prediction method according to claim 1, wherein the precipitation distribution data in the precipitation region and the local precipitation region is corrected based on the rainfall observed by a rain gauge using the measured rainfall. Predicted precipitation R1 (t) after t hours by the precipitation prediction method according to claim 1 or 2 as a kinematic technique, predicted precipitation R2 (t) after t hours by a physical technique, Based on the weight function C (t) (1 ≧ C (t) ≧ 0) after t hours of the predicted precipitation R1 (t) by the kinematic method with respect to the predicted precipitation R2 (t) by the physical method, The predicted precipitation amount R (t) after t hours is obtained by the following equation 1, and the weight function C (t) is calculated as the weight function C (t) at least after t hours by the kinematic method according to the topographic factor, wind direction factor, and wind speed factor. Different weighting functions C (t) are prepared in correspondence with different weightings between the predicted precipitation R1 (t) and the predicted precipitation R2 (t) after t hours by the physical method. The weight function C (t) is selected and used. Precipitation prediction method.
(Equation 1)
R (t) = C (t) * R1 (t) + (1-C (t)) * R2 (t) The precipitation intensity after a predetermined time elapses in the precipitation area and / or the local precipitation area from the temporal change in the precipitation intensity of the precipitation area and / or the local precipitation area. precipitation prediction method according to any one of the three. Predicting the precipitation region position and / or the local precipitation region position after a predetermined time of the precipitation region and / or the local precipitation region from the time change of the movement vector of the precipitation region and / or the local precipitation region. The precipitation prediction method according to any one of claims 1 to 4 , wherein: