JPS5860370A - Snowfall estimating device - Google Patents

Abstract

PURPOSE:To improve the estimation accuracy of snowfall by mutually compensating data obtained from a LANDSAT satellite and snowfall data obtained on the ground. CONSTITUTION:A Picture data 11 in a prescribed area obtained from a LANDSAT satellite is topographically compensated 12 and the compensated data are sent to an objective area extracting device 13 to extract an area requiring the measurement of snowfall. The extracted picture data are applied to a snowfall area discriminating device 14 to find out snowfall and no snowfall areas and the picture data in the snowfall area are sent to a photographic output device 16 through a magnetic tape unit 15 to output a photographed snowfall distribution figure 17. Snowfall in each mesh is applied to multivariate analysis 21 on the basis of height data 18 corresponding to each mesh in an area to be observed, specific attitude data 19 showing snowfall conditions and measured and snowfall variation is estimated 29 on the basis of a sample checking data 27 and aerial photographing data 28. Subsequently melted snow volume is estimated 31 from automatic observation data 30 and the outflow of snow is estimated 32 to estimate snowfall.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a snow amount estimating device that estimates the amount of snow in an observation target area from ground image data obtained from, for example, a Landsatellite satellite, and topographical and meteorological data.

As is well known, accurately knowing the amount of snow that has fallen in mountainous areas greatly contributes to the planned operation of hydroelectric power plants and related thermal power plants.
*It is extremely important for appropriate river management such as effective use of water resources such as irrigation water, nature protection, and flood prevention.

Conventionally, the amount of snowfall in mountainous areas has been measured using snowboarding, pay methods, aerial photogrammetry, and satellite information.

The Snow Survey 7 method measures snow depth manually at measurement points within a mountain river basin, and then aggregates the data to determine the amount of snow within the basin.

However, this is expensive because it requires manpower, and exploring snowy mountains is dangerous. Moreover, direct measurement is limited to about 5 meters of snow. In addition, because sufficient measurement data cannot be obtained, this data lacks surface representativeness, making it difficult to accurately estimate the amount of snowfall. Marks are placed on transmission line towers) are measured by aircraft during the snowy season, and the amount of snow in the basin is determined by aggregating the measurements. If the observation target area is small, the amount of snowfall can be determined with relatively high accuracy, but it is difficult to collect data simultaneously over a large area.

In addition, satellite information includes, for example, creating a snow distribution map from image data of an observation target area obtained from a Randnut satellite, and estimating the amount of snow from changes in this snow distribution. In other words, since image data is obtained from the Randnut satellite at intervals of several days, the current data is compared with the previous data to determine how much snow has melted using snow line information (the boundary between snow-covered areas and snow-free areas). ), and the amount of snow that melted (the amount of snow that had accumulated) was estimated from this and the amount of river flow actually observed.
is calculated backwards. However, this method only determines snow-covered areas based on image data obtained from Randnut satellites.
It is difficult to know the current amount of snow since it is estimated from snow melting conditions.

As described above, there are various methods for determining the amount of snowfall, but all of them are difficult to estimate the current amount of snowfall with high accuracy and are not practical.

This invention was made based on the above circumstances.11 Its purpose is to estimate the amount of snowfall with high accuracy using image data of the observation target area, and geographical and meteorological data.
The present invention aims to provide a snowfall estimation device that enables planned operation of hydropower and thermal power plants, as well as effective utilization of water resources and appropriate river management.

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

In FIG. 1, reference numeral 11 indicates multispectral image data of a predetermined area obtained by, for example, a Randnut satellite. As is well known, this image data is obtained by scanning the ground with a width of 185 km, and the area of one image is 185 km x 185 -, of which a range of approximately 57 m x 57 m is processed as one pixel. This one pixel consists of 8 bits of data, and the same area has a wavelength of 5 pads (0, 5
~0.6μm, 0.6~0.7μm, 0.7~0, 8μm
m, 0.8-1.1 μnl t 10.4-12.
6 μm), image data was obtained separately for each.

This image data 11 is recorded on a magnetic chip 7"K and is supplied to a terrain correction device Geometric distortions are corrected by device I2 and UTM (Univ@rsal
The sequences are recombined so that they correspond one to one to the topographic map of Transform M*rcator (transverse Mercator) coordinates. That is, the image data 11 is first supplied to the ground control point extractor 12 (VC), which extracts ground control points necessary for geometrically correcting the image.
Po1nt:GCP) is selected. This GCP
is an object that can be easily identified in an image, and is an object on the ground whose exact position (latitude, altitude) can be determined on a topographic map. Usually in the center of large bridges, at road intersections,
Artificial and easy-to-identify points such as the runway of an airport, the center of a pond of appropriate size, the center of an island, etc. are selected as GCPs. The number of GCPs selected from one scene of image data (equivalent to one image area) varies depending on the transformation formula, but usually about 10 points are sufficient when using the affine transformation formula described later.

GCP is displayed on the gray field and from this displayed data.
is inputted using a light (eg.

In this image orientation section xxb, GCP and GC in the image data are
The relationship between P and the latitude and longitude of the ground is determined, and the conversion coefficient is determined. That is, peak latitude universally represents a position on the earth, but its coordinate system is developed on the surface of the spheroid that defines the shape of the earth. On the other hand, the position of a point on the image data is a pixel (Pietur@Elem@nt: P
It is defined on a plane consisting of the lx@l ) axis and the line axis. An ellipsoid surface and a plane are generally related by an elliptic integral of the second kind, and for K, a practical conformal projection formula between latitude and longitude and transverse Mercator coordinates already exists. Therefore, here, the relationship between latitude and longitude coordinates and land coordinates (image data) coordinates is defined in two stages as follows.

Conversion ■ Latitude and longitude coordinates; Transverse Mercator coordinate conversion ■ Transverse Mercator coordinates = Landsat coordinates Here, the conversion ■ is a conversion between plane coordinate systems, and the transverse Mercator coordinates are distances on the earth (north-south X, east-west Y) Regarding K, it is a rectangular coordinate system, but it is thought that the image data of the Randnacht satellite is approximately an oblique coordinate system.

For these coordinate systems, relational expressions can be created using affine transformation expressions. Transverse Mercator coordinate value (x = y
) When the coordinate values of t and t are (P, A), the relational expression between them is as follows.The six p conversion coefficients m1 (1=1 to
6) is determined by the least squares method using the Radsat coordinate values of three or more GCPs and the transverse Mercator coordinate values. This obtained conversion coefficient is the image sample 1951
J e K supplied.

This image numbering section 12a actually converts the image data into
Image data is rearranged according to equation (1) above so that it can be superimposed on a topographic map. There are various methods to rearrange this image, but here we will use the simplest nearest neighbor method (N@ar@st W@ightbour
) applies. This is K] The rearranged image data is replaced with the original image data closest to that position, and final topographical correction is performed.

The image data corrected in this way is supplied to the target area extraction device t3, and the observation target area with the required amount of snowfall is extracted. This extracted image data is supplied to the snow area determination device 14.

This snow area determination device 14+ first includes a threshold selection unit 14.
*In Ilc, a training area is set in the silent area and snowy area that are obvious from the extracted nine-stroke data, and the average, variance, histogram, etc. for each hand are calculated for the data in this area. Next, a band that best matches the snow reflection characteristics among these values is selected, and a threshold value is determined from the data of this band. That is, as shown in Fig. 2, the intermediate value between the average value of the histogram Sfr of the training area in the snow-free region and the average value of the histogram S□ of the training area in the snow-covered region in the optimal band is set to 50.
This value is set as the threshold (second
(shown as sth in the figure). This threshold value is supplied to the determination unit 14bK, and a determination is made as to whether the image data of the selected optimal band is a snow-free area or a snow-covered area. Data with a value lower than the threshold value is determined to be snow-free, and data with a value higher than the threshold value is determined to be snow-covered. This is done pixel by pixel, and image data of only the snow-covered area is extracted1. Normally, the density slice method is used to extract the image data of only snow-covered areas. When extracting a snow-covered area, it is common to use only the training area of the snow-covered area, but in this case, it is difficult to set a training area that represents an intermediate area, so there is a risk that only areas close to the snow-covered area are extracted. . On the other hand, in the case of the present invention, since a snow-free training area is also used, it is easy to set an intermediate value. In addition, the reflection of the snow area (
Since the vector is considered to be relatively stable, if we select an area where the land cover remains the same as the training area for the snow-free area even if the scene changes, the threshold for extraction will also be relatively stable. It becomes possible to extract a good snow-covered area.

The image data of the snow covered area obtained as described above is supplied to, for example, a magnetic tape device JjiC and stored therein, and is also supplied to the photo output device 16. In this device 16, first, the data is converted into a photographic film in a film output section 16a, and this film is adjusted in scale, orientation, etc. in a correction section xtsb, and then converted into a photograph in a photographic processing section 16e. is output.

On the other hand, 18 is altitude data. This elevation data defines the observation target area (which matches the output data of the target area extraction device 13) as 1/4 (
It is divided into 278 m x 231 m) and represents the average altitude within these 174 meshes. This elevation data 18 is supplied to the ground data processing device 19, and contains data on topographic features related to snowfall in the observation target area, such as elevation, slope direction, distance from the main ridge, distance from the Sea of Japan, The amount of relief is determined for each mesh.

That is, the elevation is the elevation data 18 itself, and the slope direction is the measurement object mesh M, the mesh M0 that is in contact with the surrounding mesh M0, and the slope direction as shown in FIG.
The elevation differences of M, to M8 are determined, and the direction of the mesh showing the maximum value is determined as the inclination direction of the slope of M0, the mech to be measured. Further, the distance from the main ridge is input as the coordinate value of a mountain range that has a shadow W on snowfall, and the shortest distance between this coordinate value and each mesh is indicated by l. In addition, the distance from the Sea of Japan can be calculated by inputting the coordinate data of the coastline of the Sea of Japan and performing the same calculation as the distance from the main ridge. Furthermore, one undulation amount is determined by the difference between the maximum and minimum elevations of meshes within a radius of, for example, 20 degrees around the mech to be measured, and this is the mech to be measured.

It is considered to be the amount of undulation of the ridge. The characteristic data of the observation target area obtained in this manner are stored in the magnetic tape device 20, respectively.

It is supplied to the multivariate analysis device 21.

This multivariate analysis device 2) is supplied with snow depth data 22 (for example, observation data from the Japan Meteorological Agency) actually measured at a predetermined point d within the observation target area.

This snow depth data 22 is data of, for example, ten or more locations within the observation target area. The snow depth for each mesh is estimated from this snow depth data 22 and the determined data for each mesh. That is, the multivariate analysis device 21 performs, for example, multiple regression analysis to estimate the snow depth (surface information) for each mesh from the snow depth data (point information). The snow depth data and each snow depth data are shown in Tables 1 and 2 below.

K that best estimates the snow depth y (objective variable) on the measurement day
A linear combination of a set of variables 11 (explanatory variables) that may be related to h y is found, and y is estimated in step 5. That is, first, Y(= be + b Ixth + blxs * +-
・+ bpxIJ −・−(2) is calculated, and this estimated value −
The sum of squares of the difference between me and the actual measurement value y1 S = Σ (yl - 100) 2゛・・・・・・・・・・・・(3
Find l) @+1) 1...b, (partial regression coefficient) that minimizes ) 1 value. K that minimizes 8 is bopb S@
Partially differentiate with l...b, set it as O, and solve the folded simultaneous equations to find bO+bl...b. Now we have found a multiple regression equation that expresses the relationship between the objective variable and the explanatory variable, and we add the data xlj(
xll,..., x15:i=1,2...
・) By substituting K, the snow depth y1 for each mesh can be found.

The obtained snow depth data is supplied to the magnetic tape device 23 and stored therein, and is also supplied to the graphic processing device 24 to output a snow depth distribution map 25.

Furthermore, the snow depth data for each mesh obtained by the multivariate analysis device 21 and the image data of the snow region outputted from the snow region determining device 14 are supplied to the snow amount estimating device 26. This snowfall amount estimating device 26 estimates the total snowfall amount in the snowfall area by correlating the image data of the snowfall area with the snow depth data.

One pixel of the nine-stroke data obtained by the Landsa and To satellites is, as mentioned above, <57m x 57m, which means that the zone of transverse Mercator coordinates is divided into 640 equal parts in the X direction,
800 in the Y direction) obtained mesh (5
7,8mXSN7m)K are almost identical. Also, the snow depth data corresponds to 231 m x 278 m.
m, the relationship between this mesh and one pixel of Landsat satellite image data is shown in FIG. In other words, 1 image of snow depth data, 1 image of Shu, Landsa, and 2 satellites, and 2 images of Cao data.
This corresponds to 0 pixel. Since the snow depth and the size of each pixel in each mesh are known, the amount of snow in that mesh can be calculated from the number of pixels in the mesh and the snow depth. By determining the amount of snow in each corresponding mesh and summing them, it is possible to determine the total amount of snow in the snow area.

Furthermore, the image data of the Landsat satellites in each mesh is averaged, and the snow depth data is associated with this average value.

Roughly speaking, the snowfall amount estimating device zgVc is supplied with sample survey data 27 and aerial photography data 28. These data are observation data such as the position and linear density of trees in the observation target area, topographical conditions, and snow stains. The image data obtained by the LANDSA and TO satellites is obtained by receiving reflected light from the ground, as described above. For this reason, if trees are planted on the ground, their shadows will reduce the reflectance of the snow.
Additionally, if the surface of the snow is dirty or the terrain is such that the slope is oriented in a shadow from the sun, the reflectance will decrease and accurate data cannot be obtained.

Therefore, even though the amount of snowfall is actually small, it may not be accurately obtained as reflection data. Therefore, rounding is performed to improve the accuracy of snowfall estimation, and the snowfall amount for each mesh determined using the observation data is corrected. Further, this observation data is spread over the observation target area by, for example, the Thiessen division method, and the snow depth data obtained by the multivariate analysis device 21 is corrected.

The snowfall amount data of the observation target area obtained in this way is supplied to the snowfall change estimation device jt9. Since images of the same location are taken by Landsat satellites at intervals of about 18 days, the above processing yields snowfall amounts at intervals of 18 days.

In this snow cover change amount estimating device 29, the difference between the snow cover data having time intervals is determined for each mesh, and the change in the snow cover amount is estimated. This snow amount change data includes elevation data, slope data supplied from the ground data processing device 19, and snow depth data, temperature data, precipitation data, and river flow rate data outputted from the automatic observation device 30. It is also supplied to the snowmelt amount estimating device 31. The amount of snowmelt changes depending on the elevation difference, temperature, and precipitation, and the amount of runoff into rivers changes depending on the slope slope. The above-mentioned temperature data, precipitation data, and runoff amount data are automatically obtained by the 1° snowmelt amount estimation device No. 3, which is automatically performed by location observation devices installed at predetermined locations within the observation target area and rivers within the observation target area. First, the coefficients of the following equation (4) are determined by the method of least squares, using the data obtained from the above-mentioned observations and the amount of change in snowfall in the nine days corresponding to this Roget observation station.

Y=α* X+ +α* Xs +(Xs Xs+α4
X4+α,

Xl: Slope data yXs: Lightning depth data.

x4: Temperature data, X: Precipitation data.

α1 ~ α@: Coefficient, C: Real number Next 1 The coefficients α■ ~ α found here, and C (4
), and by substituting the elevation data and slope data of each mesh for By summing up the amount of change for each mesh, the amount of change in snow cover in the observation target area, ie, the amount of snow melt, is determined. This snowmelt amount data is supplied to the runoff amount estimating device 32.

The runoff amount estimating device 32 estimates the runoff amount by calculating the coefficient of the following equation (5) using the least squares method from the snowmelt amount data and the runoff data obtained by the automatic observation device 30.

I=β, M+C・・・・・・・・・・・・・(5
) Here, a case will be described in which the future amount of snowmelt and runoff is estimated using equations (4) and (5) above.

(4) In the second section, substitute the elevation data and slope data for Xl tx,
,, By substituting, for example, past monthly average snow depth data, temperature data, and rainfall data at a predetermined observation point within the observation target area, it is possible to calculate the future snowmelt amount for each metsushi from K. By doing so, it is possible to estimate the future amount of snowmelt in the observation target area. In other words,
For example, if the current time is March, the amount of snowmelt in April can be estimated by substituting April's monthly average snow depth data, temperature data, and rainfall data into (4)2.

Then, by substituting the obtained snowmelt amount data into equation (5), the amount of runoff in April can be estimated. Similarly, by substituting the average data for each month, the amount of snowmelt and runoff for that month can be estimated. However, this is only possible until the monthly average snow depth data reaches zero, and estimation cannot be made after this point.

According to the above configuration, it is difficult to use ground data to compensate for the drawback that snow depth cannot be determined directly from data obtained by conventional Landsatellite satellites, and snow cover in observation target areas that cannot be confirmed from ground data. By supplementing the distribution with data obtained by the Landsa and To satellites, it became possible to estimate the snowfall amount in the observation target area with higher accuracy than before. Furthermore, since the estimated snowfall amount is corrected according to the snow surface shape mVC actually observed on site, it is possible to further improve the estimation accuracy.

In addition, if data from Landsa and Tosa satellites and ground observation data are supplied, the current amount of snow can be estimated immediately.
Moreover, since it is possible to estimate the amount of snowmelt and runoff in the future using data from two LANDSA satellites spaced apart in time, data can be obtained more quickly than in the past.

Therefore, by enabling planned operation of current and future hydropower and thermal power plants, effective use of water resources and appropriate river management will be possible.

Numerical information and snow depth possessed by data obtained from Zara K, Landsa, and To satellites The relationship becomes clear. Therefore, it is possible to easily estimate snow depth, amount of snow, amount of runoff, etc. by correcting the data obtained from LANDSA and TO satellites with the data obtained from KAT observation.

In addition, almost 20% of the observation target area was
If image data of the entire area is obtained, it is possible to estimate the amount of snowfall using ground data, which is extremely advantageous compared to the conventional case, which requires image data of the entire area. It is possible to prevent defects in snowfall estimation as much as possible.

As detailed above, according to the present invention, the planned operation of hydropower and thermal power plants can be improved by estimating the amount of snow with high accuracy using image data of the observation target area and geographical and meteorological data. It is possible to provide an apparatus for estimating the amount of snow that enables efficient use of water resources and appropriate river management.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the snowfall amount estimating device according to the present invention, FIG. 2 is a diagram showing the operation of the snowfall area determination device, and FIG. 3(a). (b) each describes the operation of the ground data processing device.
FIG. 4 is a diagram shown for explaining the operation of the snow amount estimating device. 1 No. Image data, 12. Terrain correction device, 13
... One target area extraction device, 14... Snow area determination device, 111-- Elevation data, 19... Ground data processing device, 21... Multivariate analysis device, 22... Snow depth data, 26... Snow cover amount estimating device, 29... Snow cover change amount estimating device, 31... Snow melting amount estimating device, 32... Outflow amount estimating device.

Claims (1)

[Claims] Digital image data of the observation target area obtained from the sky is supplied, and the data of this area is added to the topographic map of the observation target area.
A means for performing geometric correction so as to correspond to one pair, and a predetermined threshold value corresponding to the value of the image data, and using this threshold value, it is determined whether there is snow or no snow for each pixel of the corrected data. means for determining the snowfall area; means for dividing the observation target area into predetermined meshes, supplying elevation data corresponding to each mesh; and determining terrain feature data representing snowfall conditions for each mesh; data and the snow depth actually measured at a predetermined point in the observation target area are supplied, and the snow depth for each mesh is estimated by multivariate analysis. The system is provided with a means for estimating the amount of camphor in the entire observation target area by correlating the snow area and snow depth, and a means for estimating the runoff flow rate based on the estimated snow amount. A snowfall amount estimation device characterized by: