JP2008185489A - Snow melt runoff prediction system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve prediction accuracy of snow melt runoff. <P>SOLUTION: This snow melt runoff prediction system 100 receives measured meteorological data including rainfall data before the current state in a target basin and predicted meteorological data after the current state with a data collector 200. In a data arithmetic device 300, a snow melt calculation section 320 calculates the water amount that flows down on a snow cover layer from a snow surface and is supplied to the soil based on the received meteorological data, and an outflow calculation section 330 calculates the predicted outflow amount using an outflow model based on the calculated water amount supplied to the soil. The snow melt calculation section 320 calculates water amount supplied to the soil based on the snow melt amount and rainfall using a snow cover infiltration model where the snow melt water infiltrates into the snow cover layer and rainfall directly arrives at the soil without infiltration into the snow cover layer. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a snowmelt runoff prediction system.

  In the management and operation of multipurpose dams in cold snowy regions, it is necessary to set a high water storage level during the snowmelt season in preparation for water use demand such as irrigation. On the other hand, because there is concern about abnormal water discharge accompanied by heavy rain during the snowy and melting seasons due to climate change due to global warming, large-scale dam inflows with high water storage levels are not desirable for safety management . Under such conflicting conditions of water resource management and flood management, it is important to accurately predict the amount of runoff associated with rainfall and snowmelt.

  Today, many methods for calculating runoff during the snowmelt period have been proposed, but most of the runoff is simply by combining the estimation of snowmelt on the snow surface (snowmelt model) and river runoff (runoff model). The amount is calculated. However, the snowmelt generated on the snow surface does not reach the soil directly, but reaches the soil after flowing down the snow. Since it takes time for the water to flow down the snow, the time when the snowmelt reaches the soil is delayed from the time when the snowmelt occurs on the snow surface. Regarding this lag time, it was shown that the flow time in the snow on the hillside of snowmelt water is almost the same as the flow time in the ground on the hillside, indicating that it is an important element of the snowmelt runoff process. Yes.

Thus, conventionally, as a method for predicting snowmelt runoff, a snowmelt runoff model that incorporates snow layer infiltration from a practical viewpoint such as flood prediction has been proposed (Non-patent Document 1). This snowmelt runoff model is a snowmelt runoff model in which the storage effect of the snow layer is modeled by a simple storage function method (hereinafter referred to as “snow infiltration model”). In this snow infiltration model, a change in snow storage accompanying a change in snow depth is expressed by changing one parameter included in the storage function method according to the snow depth.
Makoto Nakatsugawa and two others, "A general-purpose snowmelt runoff analysis considering snow storage," Hydrology, February 2004, Vol. 48, pp.37-42 Toshihide Usiya, 2 others, “Quantification of water circulation in the Ishikari River basin”, Hokkaido Development Public Works Research Institute Monthly Report, No.628, September 2005, pp.18-34 Kondo, J and Yamazki, T, "A Prediction Model for Snowmelt, Snow Surface Temperature and freezing Depth Using a Heat Balance Method", Journal of Applied Meteorology, May 1990, vol.29, pp.375-384 Takeshi Yamazaki, “Modeling of Snow Melting Mechanism”, Lecture on Recent Advances in Modeling Technology, 1993, pp.91-106 Kondo Jun, "Meteorology of water environment-water balance and heat balance of the ground surface", Asakura Shoten, 1994 Takashi Ishii and 2 others, “Estimation of leaf area index LAI from satellite data”, Journal of Hydrology and Water Resources, 1999, Vol. 12, No. 3, pp.210-220 Hisashi Kuchizawa, 1 other person, “Estimation of snowfall and evapotranspiration on the basin scale in consideration of heat and water balance”, Hokkaido Engineering Research Institute Monthly Report, No. 588, pp.19-38 Supervised by the Ministry of Land, Infrastructure, Transport and Tourism, Hokkaido Development Bureau Construction Department River Management Division, edited and published by Hokkaido River Disaster Prevention Research Center, Research Institute, "Real-time flood prediction system theory", May 2004 Hitoshi Baba and two others, "Development of a two-stage storage function type runoff model combining evapotranspiration and infiltration", Journal of Hydrology and Water Resources, 2001, Vol. 14, No. 5, pp.364-375 Hoshiyoshi, 1 other, “Interrelation between rainwater flow method and storage function method”, Proceedings of Hydraulic Lecture, 1982, 26, pp.273-278 Toshihide Usaya and two others, “Examination of a snowmelt runoff model incorporating the storage effect of a snow layer”, River Technology Papers, 2006, Vol. 12, pp.329-334 Nakatsugawa Makoto and two others, “Long-term Runoff Calculation Considering Changes in Snow Cover”, Journal of Hydraulic Engineering, 2003, Vol. 47, pp.157-162 Nash, JE and Sutcliffe, JV, "River flow forecasting through conceptual models part Ι-A discussion of principles-", J. of Hydrology, 1970, Vol.10, pp.282-290 Jungo Yoshida, “Penetration of snowmelt into snow”, Low Temperature Science Physics, 1965, 23, pp.1-16

  However, the conventional snowmelt runoff prediction method (snowmelt runoff model) has the following problems.

  The snow accumulation infiltration model incorporated in this snowmelt runoff model models snowmelt water infiltration, and it has not been verified whether it can also express heavy rain infiltration.

  That is, the snow infiltration model is intended to take this time into account, taking note of the time taken for the water supplied to the snow surface to reach the soil as described above. As water supplied to the snow surface, snowmelt water and rainfall are considered. The conventional snow accumulation infiltration model does not distinguish between snowmelt water and rainfall, and passes the snow layer through these two types of water. However, subsequent verification revealed that the conventional snow infiltration model cannot express heavy rain infiltration. Therefore, in the conventional method, there is a certain limit in the prediction accuracy of snowmelt water discharge.

  This invention is made | formed in view of this point, and it aims at providing the snowmelt runoff prediction system which can improve the prediction accuracy of snowmelt runoff.

  The snowmelt runoff prediction system according to the present invention is a snowmelt runoff prediction system that predicts runoff accompanying snowmelt, and inputs meteorological data including actual meteorological data before the current state of the target basin including rainfall data and predicted meteorological data thereafter. A data input unit, a soil supply water amount calculation unit that calculates the amount of water supplied to the soil by flowing down the snow layer from the snow surface based on the input weather data, and a predetermined value based on the calculated soil supply water amount A runoff amount calculation unit that calculates a predicted runoff amount using the runoff model of the soil, and the soil supply water amount calculation unit is configured to infiltrate the snowmelt water into the snow layer and the rain without penetrating the snow layer into the soil. As a method of directly reaching the above, a configuration is adopted in which the amount of water supplied to the soil is calculated based on the given amount of snowmelt and the input rainfall.

  According to the present invention, it is possible to improve the prediction accuracy of snowmelt water discharge.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a block diagram showing a configuration of a snowmelt runoff prediction system according to an embodiment of the present invention. FIG. 2 is a block diagram showing a configuration of each part of the snowmelt runoff prediction system of FIG. FIG. 3 is a schematic diagram showing the types and flows of data used in the snowmelt runoff prediction system of FIG.

  A snowmelt runoff prediction system 100 shown in FIG. 1 roughly includes a data collection device 200, a data calculation device 300, and an output device 400.

  The data collection device 200 is a device that inputs and collects data necessary for snowmelt runoff prediction calculation, and includes an input data processing unit 210 that processes the input data. For example, dam actual condition data, terrain data in the dam basin, meteorological agency AMeDAS data, meteorological agency radar data, meteorological agency precipitation prediction mesh, GPV (Grid Point Value) data, and the like are input to the data collection device 200. The data collection device 200 is composed of, for example, a computer provided with external communication means.

  The actual dam data is measured data of meteorological data and hydrological data in the dam basin. The weather data includes items such as temperature, humidity, precipitation, wind speed, amount of solar radiation, sunshine duration, atmospheric pressure, and snow depth. The hydrology data includes, for example, dam inflow, storage level, and discharge flow rate as items. The actual dam data is transmitted from the dam management office and the telemeter. Live dam data is measured in days or hours.

  The terrain data in the dam basin is data relating to the terrain in the dam basin, and includes, for example, longitude and latitude, altitude, slope amount, and LAI (Leaf Area Index). The amount of inclination is, for example, the amount of inclination in the north-south / east-west direction. The LAI is a monthly value, and the longitude / latitude, altitude, and inclination amount are fixed values. In the present embodiment, published values are used as the terrain data in the dam basin. The terrain data in the dam basin is transmitted from an appropriate external information source via a network, for example. The fixed value may be stored in advance in a memory (not shown) in the data collection device 200.

  The Japan Meteorological Agency AMeDAS data is the AMeDAS data released by the Japan Meteorological Agency, and the Japan Meteorological Agency radar data is the radar data also released by the JMA. Both the JMA AMeDAS data and the JMA radar data consist of, for example, precipitation and temperature (5 km mesh) as items, both of which are actually measured data. JMA AMeDAS data and JMA radar data are distributed from the JMA at regular time intervals (on-time distribution).

  The Japan Meteorological Agency Precipitation Prediction Mesh is one of the weather prediction data published by the Japan Meteorological Agency, and it predicts the precipitation amount of 6 hours ahead in 5km mesh units. The Japan Meteorological Agency precipitation forecast mesh is delivered from the Japan Meteorological Agency at regular time intervals (on-time delivery).

  The GPV data is one piece of weather prediction data, and predicts weather data (for example, temperature, humidity, precipitation, cloud cover, and atmospheric pressure) 24 hours ahead in units of 20 km mesh. For example, GPV data is distributed from the Japan Meteorological Agency at regular time intervals (on-time distribution).

  The input data processing unit 210 performs calibration based on actual measurement values and 1 km mesh on the input predetermined data, and also performs 1 km mesh of live weather data. For example, 1 km meshing is performed by spatially interpolating data.

  In calibration with actual measurement values and 1 km mesh, the forecast data of the Meteorological Agency precipitation forecast mesh (6 hours ahead precipitation) and GPV data (24 hours ahead temperature, humidity, precipitation, cloud cover, atmospheric pressure) are used as the Meteorological Agency AMeDAS data. In addition, the data is corrected (calibrated) by actual measurement data (precipitation amount and temperature of 5 km mesh) of the Japan Meteorological Agency radar data, and various data are further expanded to 1 km mesh (1 km mesh). At this time, the input data processing unit 210 calculates the solar radiation amount and the sunshine time from the cloud amount, and meshes the solar radiation amount and the sunshine time with 1 km instead of the cloud amount. For example, in the case of precipitation, as an example of correction (calibration), as shown in FIG. 3, correction is performed by making use of rainfall observations at the southeast and southwest points. As a result of this processing (calibration using actual values and 1 km meshing), the input data processing unit 210 outputs 1 km mesh predicted weather data (temperature, humidity, precipitation, solar radiation, sunshine duration, atmospheric pressure). Hereinafter, this 1 km mesh predicted weather data is referred to as “predicted mesh data”.

  Further, in the production of 1 km mesh of live weather data, the weather data is developed into 1 km mesh (1 km mesh) using the dam live data (particularly weather data) and the terrain data in the dam basin. As a result of this processing (1 km mesh of live weather data), the input data processing unit 210 converts live weather data (temperature, humidity, precipitation, wind speed, solar radiation, sunshine duration, atmospheric pressure, snow depth) of 1 km mesh. Output. Hereinafter, this live weather data of 1 km mesh will be referred to as “live mesh data”.

  As described above, the data collection device 200 inputs (collects) various data, and after the input data is processed by the input data processing unit 210, is output to the next data calculation device 300 as predicted mesh data and actual mesh data. To do. However, the dam inflow data among the inputted dam actual condition data is directly output to the data arithmetic unit 300 without being processed by the input data processing unit 210. In addition, among the collected data, for example, depending on the type of data such as a part of terrain data, the data is appropriately output to the data calculation device 300 as necessary.

  In the present embodiment, the data collection device 200 is physically separated from the data arithmetic device 300, but the present invention is not limited to this. For example, the data collection device 200 and the data calculation device 300 can be configured by a single computer.

  The data operation device 300 is configured by a computer, and roughly includes a snow accumulation calculation unit 310, a snow melting calculation unit 320, an outflow calculation unit 330, and a storage unit 340. The data calculation device 300 has a built-in snowmelt runoff model, and uses this snowmelt runoff model to calculate a predicted dam inflow (predicted runoff) from predetermined weather data. The snowmelt runoff model is roughly divided into three types of submodels: a snowmelt model, a snow infiltration model, and a runoff model. The snow melting model and the snow penetration model are built in the snow melting calculation unit 320, and the outflow model is built in the outflow calculation unit 330.

  The storage unit 340 stores various constants necessary for calculation based on a snowmelt runoff model (snowmelt model, snow infiltration model, runoff model). The constant is a fixed value. In the present embodiment, values listed in the literature are used for physical constants. The storage unit 340 includes, for example, a bulk coefficient, a constant pressure specific heat of air, a factor representing the inclination of the leaf surface with respect to radiation, a melting latent heat of snow, a thermal conductivity of snow, a Stefan-Boltzman constant, and a calculation time. The interval is stored.

  In the calculation based on this snowmelt runoff model, first, the snowmelt model and the snow infiltration model are applied to a mesh of about 1 km square (1 km mesh), and the amount of soil supply water for each mesh is estimated from meteorological and topographic factors. Next, the soil supply water amount for each mesh is ramped (basin average) over the entire basin, and the obtained value (basin average soil supply water amount) is used as an input value to the runoff model. Finally, this value is input to the runoff model to calculate the inflow.

  The snow cover calculation unit 310 calculates data related to snow cover (for example, snow cover depth and snow cover density) based on the live mesh data. Therefore, the snow cover calculation unit 310 includes a snow cover equivalent water amount estimation model unit 312 and a snow cover model unit 314 as shown in FIG. The snow-equivalent water amount estimation model unit 312 has a built-in snow-equivalent water amount estimation model for calculating the snow-equivalent water amount (the value obtained by converting the snow cover into the rain amount). The snow model unit 314 calculates the snow depth and the snow density. Built-in snow model. The snow calculation unit 310 calculates the snow equivalent water amount by the snow equivalent water amount estimation model based on the actual mesh data (the value obtained by converting the meteorological data observed on the ground to 1 km mesh), and also calculates the snow depth by the snow model. And calculate snow density. The snow cover calculation unit 310 performs the above calculation processing in units of mesh (about 1 km square) and in units of days.

  The snow equivalent water amount estimation model and the snow accumulation model to be used are not particularly limited. As an example, in the present embodiment, models based on Kojima's viscous compression theory are used as the snow-equivalent water amount estimation model and the snow accumulation model (Non-Patent Document 2). The calculation formula is given by the following formulas (1) to (5), where time is t.

Here, ρ _s is snow density (kg / m ³ ), S _d is snow depth (m), S _sf is snow depth (m), S _d ^′ is snow depth before consolidation (m), ρ _s ^′ is Average density of all layers (kg / m ³ ), S _w is the amount of snow equivalent to snow (mm), m is the amount of snow melting (mm), e is the amount of evapotranspiration (mm), ρ _w is the density of water (= 1000 kg / m ³⁾ ), K is a constant, and η ₀ is a viscosity coefficient (kg · day / m ² ). For example, K is fixed at 0.021, and η ₀ is determined in the range of 10 to 16 according to the snow depth reproduction result for each region.

  The snow melting calculation unit 320 includes a snow melting model unit 322, a snow accumulation infiltration model unit 324, and a basin averaging unit 326. The snow melting model unit 322 incorporates a snow melting model, and the snow accumulation infiltration model unit 324 includes a snow accumulation infiltration model. The snow melting calculation unit 320 first inputs meteorological data and terrain data, and calculates the amount of soil supply water based on the built-in snow melting model and snow accumulation infiltration model. This calculation process is performed in mesh units (about 1 km square) and in predetermined time units. In other words, the snow melting calculation unit 320 first applies the snow melting model and the snow infiltration model to a 1 km mesh, and estimates the amount of soil supply water for each mesh from the weather and topographic factors. Next, the snow melting calculation unit 320 ramps the soil water supply amount for each mesh over the entire basin (basin average), and outputs the obtained value (basin average soil water supply amount).

  The snow melting model unit 322 incorporates a predetermined snow melting model, and uses this snow melting model to calculate the amount of snow melting (the amount of water given to the snow surface) from predetermined input information (meteorological data and terrain data). Of the input information to the snow melting model unit 322, weather data is input from the data collection device 200 as predicted mesh data and live mesh data. Further, the terrain data is also input from the data collection device 200. The calculated amount of snow melting is output to the snow accumulation infiltration model unit 324.

  The snow melting model to be used is not particularly limited. As an example, in the present embodiment, a snow melting model based on the heat balance method (Non-patent Documents 3 and 4) proposed by Kondo and Yamazaki et al. Is used in consideration of the effects of vegetation. The basic equation is composed of two heat balance equations relating to the entire snow cover and the snow surface, and is given by the following equations (6) and (7), respectively.

Where c _s is the specific heat of snow (J / kg / K), ρ _s is the density of snow (kg / m ³ ), l _f is the latent heat of melting of snow (J / kg), and T ₀ is 0 (° C.) , T _s is the snow temperature (° C.), T _sn is the snow temperature (° C.) after time Δt, W ₀ is the maximum moisture content (= 0.1), Z is the freezing depth (m), and ε is the injection rate (= 0.97), λ _s is the thermal conductivity of snow cover (= 0.42 W / m / K), σ is the Stefan-Boltzmann constant (W / m ² / K ⁴ ), Δt is the time interval (= 3600 s), f _v is the transmittance of the vegetation layer, G is the energy received by the snow (W / m ² ), H is the sensible heat (W / m ² ), lE is the latent heat (W / m ² ), L downward arrows are downward longwave radiation _{^{(W / m 2), M}} 0 is snow melting heat ^(W / m 2), frost depth after _{Z n} is the time Δt (m), _{T v} is the temperature of the canopy layer (° C.). Note that the downward longwave radiation amount L down arrow, the method of Kondo (Non-Patent Document 5, pp.90-91) given by, Also, the temperature T _v of the plant cover layer is conveniently used temperature.

  The sensible heat H and the latent heat 1E are given by the following equation (8) by the bulk method (Non-patent Document 5, pp.242).

Here, _{c p} is the specific heat at constant pressure (J / kg / K) of the air, [rho is the density of air _{^{(kg / m 3), C}} H bulk coefficient for sensible heat (= 0.003), U is the wind speed (m / S), T is the temperature (° C.), l is the latent heat of water evaporation (J / kg), _CE is the bulk coefficient for latent heat (= 0.003), q _s (T _s ) is the saturation specific humidity, q Is the relative humidity, h is the relative humidity, Δ is the gradient with respect to the temperature of the saturated specific humidity, p is the atmospheric pressure (hPa), e is the water vapor pressure (hPa), and e _SAT is the saturated water vapor pressure (hPa).

Also, the planting transmittance f _v of the layers, Ishii et al. Using a leaf area index LAI that (Non-Patent Document 6) has been estimated, given by the following equation (9) (Non-Patent Document 5, Pp.231) .

  Here, F is a factor (= 0.5: isotropic) representing the inclination of the leaf surface with respect to radiation, and LAI is the leaf area index.

  Further, the energy G received by the above-mentioned snow cover is given by the following equation (10).

Here, S is the amount of solar radiation (W / m ² ), and α is an albedo. In addition, albedo is given by the empirical formula (nonpatent literature 7) using daily average temperature.

In the snow melting model of the present embodiment, the above-described equations (6) to (10) are simultaneously solved to simultaneously solve the snow melting heat M ₀ , snow temperature T _sn , and freezing depth Z _n . Then, using the value of the obtained snow melting heat M ₀ , the amount of snow melting (the amount of water given to the snow surface) q _rm (mm / h) is calculated by the following equation (11).

Here, q _rm is the amount of snow melting (mm / h), M ₀ is the heat of melting snow (W / m ² ), Δt is the time (= 3600 s), l _f is the latent heat of melting (melting heat) (J / kg) , Ρ _w is the density of water (= 1000 kg / m ³ ).

  The snow accumulation infiltration model unit 324 incorporates a predetermined snow accumulation infiltration model, and uses this snow accumulation infiltration model to calculate the amount of soil supply water for each mesh from predetermined input information (snow melting amount, rain amount, snow accumulation depth). Of the input information to the snow accumulation infiltration model unit 324, the snow melting amount is input from the snow melting model unit 322, and the snow accumulation depth is input from the snow accumulation calculation unit 310. The rainfall is input from the data collection device 200 as predicted mesh data and live mesh data. The calculated soil supply water amount is output to the basin averaging unit 326.

Basically, the amount of soil supply water is given by the following equation.
Soil supply water amount = rainfall + snow melt water−snow accumulation amount Snow accumulation amount (retention amount remaining in the snow) varies with snow depth.

  In this embodiment, a monovalent linear storage function method (Non-patent Document 1) proposed by Nakatsugawa et al. Is used as a snow infiltration model. This model is derived by assuming Darcy's law (saturation seepage) in the water flow in the snow, and the basic formula is given by the following formulas (12) and (13).

Here, s _s is the snow storage amount (mm), k _s is the snow storage coefficient, q _rm is the amount of water given to the snow surface (especially the amount of snow melting) (mm / h), and q _s is the amount of soil supplied by snow infiltration. (Mm / h), k ₀₁ and k ₀₂ are coefficients, and H _s is snow depth (cm).

In equation (12), the snow storage effect is expressed by the storage coefficient k _s . The storage coefficient k _s is considered to change depending on the snow quality (snow accumulation density, ice particle size, etc.), but is parameterized by the snow accumulation depth H _s from the viewpoint of practicality.

  FIG. 4 is a diagram showing a configuration of a snow accumulation infiltration model in the present embodiment. In addition, FIG. 5 is a figure which shows the structure of the conventional snow accumulation penetration model.

  Originally, there is no distinction between rainfall and snowmelt water flow. However, in the snow infiltration model shown in FIG. 4, rainfall and snowmelt water flow are handled separately, and the snow melting water is referred to as the above-described conventional method (hereinafter referred to as “conventional method”). As in the case of the model, infiltration is considered and rainfall is allowed to reach the soil directly without considering infiltration (that is, the delay time when the rain flows down the snow layer is not considered). In this way, the method of causing the rainfall to directly reach the soil without taking into account snow penetration (that is, outputting the rainfall directly to the runoff model) will be referred to herein as a “correction method” for convenience. In the modified method (corresponding to the snow infiltration model shown in FIG. 4), the snow melt passes through the snow layer, while the rain reaches the soil directly regardless of the amount. By treating in this way, the snowmelt water at the time of no rain penetrates the snow layer, while heavy rain reaches the soil directly.

  In the conventional method, the sum of the amount of rain and the amount of snow melt given to the snow surface is assumed to flow down the snow layer, and as shown in FIG.

Thus, in this embodiment, soil supply amount of water is output, and snowmelt q _s reaching the soil by snow penetration, the sum of the rainfall. That is,
Soil supply water volume (output)
= Snow melting amount q _s + rain amount reaching the soil due to snow infiltration q Here, the snow melting amount (soil supply water amount due to snow infiltration) q _rm reaching the soil due to snow infiltration is expressed by the above equations (12) and (13). Calculated.

The coefficients k ₀₁ and k ₀₂ in the above equation (12) are obtained from FIG. 6, for example.

FIG. 6 is a diagram illustrating the relationship between the storage coefficient k _s and the snow depth H _s . In FIG. 6, the storage coefficient k _s is a value obtained from a remarkable snow melting case observed in 1997. Further, the snow depth H _s is a value observed when the storage coefficient k _s is obtained. The solid line in the figure is a regression line obtained from these points by the least square method.

As a result of snow melting observation, k ₀₁ = 0.16 h / cm, k ₀₂ = 8.24 h was reported from the correlation between the k _{s of the} remarkable snow melting case and the snow depth H _{s at} that time (FIG. 6). Even in the present embodiment, this value is used as the coefficients k ₀₁ and k _{02 in the} equation (12).

  The basin averaging unit 326 ramps the soil water supply amount for each mesh calculated by the snow infiltration model unit 324 over the entire basin (basin averaging). The obtained value (basin average soil supply water amount) is output to the runoff calculation unit 330.

  The outflow calculation unit 330 includes an outflow model unit 332 and an outflow model constant update unit 334. The runoff model unit 332 has a built-in runoff model, and calculates the predicted dam inflow from the soil supply water based on this runoff model. The outflow model constant updating unit 334 includes a Kalman filter, and updates the outflow model constant based on the Kalman filter theory so that the difference between the predicted value of the dam inflow amount and the actual measurement value is minimized. The Kalman filter is a method of correcting (correcting) the model constant of the outflow model in real time while correcting the predicted dam inflow amount using an actual measurement value (dam inflow amount) obtained every hour in an actual operational scene. The model constant is updated (modified) by applying the formula of Kalman filter theory. Since the method of applying the Kalman filter to the outflow model is described in detail in Non-Patent Document 8, for example, description thereof is omitted here.

  FIG. 7 is a diagram showing the configuration of the outflow model in the present embodiment. In the present embodiment, as shown in FIG. 7, a storage function method including a subsurface outflow component (a so-called two-stage tank type storage function model) (Non-Patent Document 9) is used as the outflow model. In this runoff model, the whole runoff process is separated into two components, surface / intermediate runoff and groundwater runoff, and both are expressed by separate storage function methods.

  The surface / intermediate outflow component (first stage tank) is composed of the equation of motion (Non-patent Document 10) that concentrates the kinematic wave equation and the continuous equation, and is given by the following equations (14) and (15). .

Here, s ₁ is the storage height (mm) of the first tank, q _s is the amount of soil supply water (mm / h), q ₁ is the outflow height (mm / h) of the surface / intermediate outflow component, and f _b is 1 Permeation supply amount from the tank to the second tank (mm / h), k ₁₁ and k ₁₂ are storage coefficients, p ₁ and p ₂ are storage indexes, A is a basin area (km ² ), and bar q _s is The average strength (mm / h) of the soil water supply amount, c ₁ , c ₂ , and c ₃ are model constants (unknown constants).

  On the other hand, the groundwater outflow component (second stage tank) is given by the following basic formulas (16) and (17).

Here, s ₂ is the storage height (mm) of the second tank, k ₂₁ and k ₂₂ are storage coefficients, q ₂ is the outflow height (mm / h) of the groundwater outflow component, and c ₄ is a model constant (unknown constant). It is.

Therefore, the predicted dam inflow calculated by the runoff model is obtained as the sum of the surface / intermediate runoff component (first tank) and the groundwater runoff component (second tank), that is, q ₁ + q ₂ . In short, the runoff model unit 332 receives the basin average soil supply water amount q _s from the snow melting calculation unit 320 and calculates the predicted dam inflow amount (q ₁ + q ₂ ) based on the above runoff model. More specifically, the dam inflow (m ³ / s) is expressed by the following unit conversion formula, that is,
Dam inflow (m ³ / s) = Outflow height (mm / h) x (basin area km ² ) / 3.6
Is used to convert the unit of outflow height (q ₁ + q ₂ ) (mm / h) obtained as a calculation result.

It is necessary to give _four model constants (c ₁ , c ₂ , c ₃ , c ₄ ) to calculate the outflow amount by this storage function method. This combination of model constant values is desirably one type in consideration of snowmelt runoff prediction. In general, since the model constant takes a different value for each water discharge, it is often difficult to determine a representative value of the model constant from a plurality of water discharge examples. Therefore, in the present embodiment, the model constant is determined by searching for the optimum value for one water discharge of a relatively large scale, instead of identifying the model constant from a plurality of water discharge cases. Note that Newton's method was used to search for the optimum value of the model constant (Non-Patent Document 11). The validity of this simple optimal identification method was confirmed by demonstrating the high reproducibility of a multi-year flood hydrograph with a set of model constants, as described below. Note that the Newton method here is used as a method of tuning the outflow model so that the outflow phenomenon can be well reproduced.

  The output device 400 outputs the data collected by the data collection device 200 (including mesh data obtained by the input data processing unit 210) and the data obtained by the data operation device 300 to the outside in a predetermined form. The output device 400 includes, for example, a display, a printer, a communication unit, and the like.

  When the output device 400 is a display, for example, as shown in FIG. 3, as a collection result of the data collection device 200, a telemeter actual value display screen, a dam actual value latest data / past data display screen, a long-term snow melting The trend study screen is displayed, the precipitation mesh prediction screen is displayed as the predicted mesh data, and the snow distribution screen, the latest data / past data display screen of the snow depth and snow density are displayed as the calculation result of the snow calculation unit 310. As the calculation result of the snow melting calculation unit 320, the snow melting prediction screen, the latest data / past data display screen of the soil supply water amount, the latest data / past data display screen of the predicted dam inflow amount as the calculation result of the runoff calculation unit 330, the snow melting dam The inflow screen is displayed.

  Next, the operation of the snowmelt runoff prediction system 100 having the above configuration will be described using the flowcharts shown in FIGS. FIG. 8 is a main flowchart showing the overall operation of the snowmelt water discharge prediction system 100. FIG. 9 is a flowchart showing the contents of the snow melting calculation process (S1400) of FIG. FIG. 10 is a flowchart showing the contents of the snow melting amount calculation process (S1410) of FIG. FIG. 11 is a flowchart showing the contents of the soil supply water amount calculation process (S1420) of FIG. FIG. 12 is a flowchart showing the contents of the outflow calculation process (S1500) of FIG. Note that the flowcharts shown in FIGS. 8 to 12 are stored as a control program in a ROM of a computer or the like and executed by the CPU.

  First, in step S1000, the data collection device 200 inputs and collects data necessary for snowmelt runoff prediction calculation. Specifically, for example, as described above, dam actual condition data, terrain data in the dam basin, JMA AMeDAS data, JMA radar data, JMA precipitation forecast mesh, and GPV data are input.

  In step S1100, the input data processing unit 210 in the data collection device 200 performs input data processing. Specifically, the predetermined data input in step S1000 is calibrated with actual measurement values and converted into a 1 km mesh, and the actual weather data is converted into a 1 km mesh. For example, as described above, in the calibration with the actual measurement value and the 1 km mesh, the prediction data of the Meteorological Agency precipitation prediction mesh and GPV data is corrected (calibrated) by the meteorological agency AMeDAS data and the actual measurement data of the JMA radar data, and further 1 km By developing the mesh (1 km mesh), predicted mesh data is obtained. In addition, in the production of 1km mesh of live weather data, live weather mesh data is obtained by expanding the weather data into 1km mesh (1km mesh) using dam live data (especially weather data) and topographic data in the dam basin. .

  In step S1200, the data calculation device 300 determines whether or not to perform snow cover calculation. As described above, since the process of the snow cover calculation unit 310 is performed on a daily basis, it is determined whether or not it is time to execute the snow cover calculation in step S1200. As a result of this determination, if snow calculation is to be executed (S1200: YES), the process proceeds to step S1300, and if snow calculation is not to be executed (S1200: NO), the process immediately proceeds to step S1400.

  In step S1300, the snow calculation unit 310 in the data arithmetic device 300 performs snow calculation. Specifically, based on the live mesh data obtained in step S1100, using the above equations (1) to (5), the snow equivalent water amount is calculated by the snow equivalent water amount estimation model, and the snow accumulation by the snow model is calculated. Calculate depth and snow density.

  In step S1400, the snow melting calculation unit 320 in the data arithmetic device 300 performs snow melting calculation processing. This snow melting calculation process is performed according to the procedure shown in the flowchart of FIG.

  First, in step S1410, the snow melting model unit 322 in the snow melting calculation unit 320 performs a snow melting amount calculation process. This snow melting amount calculation processing is performed according to the procedure shown in the flowchart of FIG.

  First, in step S1411, the meteorological data of the predicted mesh data and the live mesh data obtained in step S1100, the terrain data in the dam basin obtained in step S1000, and the predetermined constant stored in the storage unit 340 are obtained by the snow melting model. Import as necessary information for calculation.

  In step S1412, a variable is calculated. Specifically, the transmittance, albedo α, water vapor pressure, air density, and specific humidity are calculated using the information captured in step S1411. For example, the transmittance is calculated by the above equation (9). Moreover, albedo (alpha) is given by the empirical formula using daily average temperature. Further, the water vapor pressure, the air density, and the specific humidity are calculated by well-known formulas that are generally used.

  In step S1413, primary parameters are calculated. Specifically, the terms (primary parameters) in the above formulas (6) and (7) are calculated using the information captured in step S1411 and the variables calculated in step S1412. The primary parameters are the downward long wave radiation amount L downward arrow, the sensible heat H, the latent heat 1E, and the energy G received by the snow cover. As described above, the downward long wave radiation amount L downward arrow is given by Kondo's method (Non-Patent Document 5, pp. 90-91). Further, the sensible heat H and the latent heat 1E are given by the above equation (8). The energy G received by the snow cover is given by the above equation (10).

In step S1414, a secondary parameter is calculated. Specifically, using the information captured in step S1411 and the primary parameter calculated in step S1413, the above equation (6) and equation (7) are solved to calculate the secondary parameter. Secondary parameters are snow melting heat M ₀ , snow temperature T _sn , and freezing depth Z _n .

In step S1415, the amount of snow melt is calculated. Specifically, the amount of snow melting q _rm (mm / h) is calculated by the above equation (11) using the information captured in step S1411 and the snow melting heat M ₀ calculated in step S1414.

  In step S1416, the amount of snow melt calculated in step S1415 is temporarily stored in a memory (not shown), and the process returns to the flowchart of FIG.

  And in step S1420 of FIG. 9, the snow supply infiltration model part 324 in the snow melting calculation part 320 performs a soil supply water amount calculation process. Specifically, using the predetermined information (rainfall amount) input in step S1000, the snowmelt amount calculated in step S1410, the snow depth calculated in step S1300, and the predetermined constant stored in the storage unit 340, FIG. The amount of soil supply water is calculated by the snow infiltration model (correction method) of No. 4. This soil supply water amount calculation process is performed according to the procedure shown in the flowchart of FIG.

First, in step S1421, the information (specifically, rainfall) necessary for calculation by the snow infiltration model (correction method) in FIG. 2 is taken from the information input in step S1000, and the snow melting calculated in step S1410 is performed. The amount, the snow depth calculated in step S1300, and predetermined constants (specifically, coefficients k ₀₁ and k ₀₂ ) stored in the storage unit 340 are taken in.

In step S1422, a snow storage coefficient is determined. Specifically, the snow storage coefficient k _s is determined by the above equation (13) using the snow depth H _s and the coefficients k ₀₁ and k ₀₂ taken in step S1421.

In step S1423, the amount of snow melting that reaches the soil is calculated. Specifically, by using the snow melting amount q _rm calculated in step S1410 and the snow accumulation coefficient k _s determined in step S1422, the snow melting amount q _s reaching the soil by snow infiltration is calculated by the above equation (12). calculate.

In step S1424, the amount of soil supply water is calculated. Specifically, the rainfall captured in step S1421, by summing up the snow melting amount q _s reaching the soil calculated in step S 1423, to calculate the soil supply amount of water.

  In step S1425, the soil supply water amount calculated in step S1424 is temporarily stored in a memory (not shown), and the process returns to the flowchart of FIG.

  In step S1430 of FIG. 9, the basin averaging unit 326 in the snow melting calculation unit 320 performs the basin averaging process. Specifically, the soil supply water amount for each mesh calculated in step S1420 is ramped over the entire basin (basin averaging) to obtain the basin average soil supply water amount.

  In step S1440, the basin average soil water supply obtained in step S1430 is temporarily stored in a memory (not shown), and the process returns to the flowchart of FIG.

  Note that the snow melting calculation process in step S1400 is performed separately on the weather data of the predicted mesh data and the live mesh data obtained in step S1100. For example, in FIG. 3, the flow of processing for the weather data of the live mesh data (the latest observed weather data) is indicated by a solid line, and the flow of the processing for the weather data of the predicted mesh data (predicted weather data) is indicated by a bold line. Yes.

  In step S1500 of FIG. 8, the outflow calculation unit 330 in the data arithmetic device 300 performs outflow calculation processing. This outflow calculation process is performed according to the procedure shown in the flowchart of FIG.

  First, in step S1510, the two types of basin average soil supply water calculated in step S1400, that is, the basin average soil supply water based on the latest observed weather data and the basin average soil supply water based on the predicted weather data are captured. In addition, the dam inflow data is fetched from the dam actual condition data input in step S1000.

  In step S1520, the outflow model unit 332 in the outflow calculation unit 330 uses the basin average soil supply water amount based on the latest observed weather data among the two types of basin average soil supply water amount captured in step S1510. Then, the dam inflow reproducible value (reproduced dam inflow amount) is calculated by the outflow model (two-stage tank type storage function model).

  In step S1530, the outflow model constant update unit 334 in the outflow calculation unit 330 compares the measured dam inflow amount acquired in step S1510 with the reproduced dam inflow amount calculated in step S1520, and based on the Kalman filter theory. Update (correct) outflow model constants.

  In step S1540, the outflow model unit 332 in the outflow calculation unit 330 again uses the outflow model constant updated in step S1530, and the predicted meteorological data out of the two types of basin average soil supply water amount captured in step S1510. Based on the average basin soil water supply based on the basin, the predicted value of dam inflow (predicted dam inflow) is calculated by the runoff model (two-stage tank type storage function model).

  In step S1550, the predicted dam inflow amount calculated in step S1540 is temporarily stored in a memory (not shown), and the process returns to the flowchart of FIG.

  In step S1600 of FIG. 8, the output device 400 performs output processing. Specifically, the data collected in step S1000, the mesh data obtained in step S1100, the calculation result obtained in step S1300, the calculation result obtained in step S1400, and the calculation result obtained in step S1500 for each type of information. They are displayed on a display in a form set in advance according to the application, printed by a printer, or transmitted (distributed) to the outside by a communication means.

  As described above, according to the present embodiment, in the snow infiltration model, the snow melt passes through the snow layer, while the rain directly reaches the soil regardless of the amount. The error (difference between the measured value and the calculated value) is reduced, and the prediction accuracy of the amount of water supplied to the soil can be improved. Therefore, the accuracy of the snowmelt runoff model incorporating the snow penetration model can be improved, and the prediction accuracy of snowmelt runoff can be improved.

  In order to verify the above effect, the present inventors verified the reproducibility of the snowmelt runoff model considering the storage effect of the snow layer with the dam basin (Toyohirakyo Dam and Jozankei Dam) in the snowy cold region as the analysis target point. . In the verification, when analyzing the interrelationship between rainfall during heavy rain and snow inflow during the snow melting season and estimating runoff during the snow melting season, it is more reasonable to think that there is no storage effect of the snow layer for heavy rain It was confirmed that.

  FIG. 13 is a schematic diagram showing the configuration of a snowmelt runoff model used for verification.

  As shown in FIG. 13, the snowmelt runoff model used for the verification is composed of three types of submodels: snowmelt, snow infiltration and runoff. In the calculation based on the snowmelt runoff model, first, the snowmelt model and the snow infiltration model are applied to a mesh of about 1 km square, and the soil water supply amount for each mesh is estimated from meteorological / terrain factors. Next, the soil water supply amount for each mesh is ramped over the entire basin, and the obtained value is used as an input value to the runoff model. Finally, this value is input to the runoff model to calculate the runoff (dam inflow). In addition, since it is as above-mentioned about the snowmelt outflow model, the description is abbreviate | omitted.

Target Basin and Analysis Data FIG. 14 is a diagram showing an analysis target basin.

As shown in FIG. 14, the analysis target basin was the Hoheikyo dam basin (basin area: 134 km ² ) located in the southern part of Sapporo, Hokkaido, and the Jozankei dam basin (basin area: 104 km ² ) located about 10 km south. . The elevation distribution of the basin is 400 to 1300 m for both dams, and 50% of the total is in the range of 700 to 900 m.

  The snow cover period lasts for about 6 months from late November to mid-May, and the increase in water accompanying snow melting is seen from mid-April to late May. The amount of snowfall (value converted to rainfall) is 1000 mm for Toyohirakyo Dam and 1200 mm for Jozankei Dam, and snowfall accounts for about 50% of annual precipitation.

  FIG. 15 is a table showing meteorological observation items in the Toyohirakyo Dam and Jozankei Dam.

  In each dam, the dam inflow and meteorological elements (see FIG. 15) are observed at the dam management office (see FIG. 14). In this verification, the observation data is arranged every hour. The analysis period was from 1996 to 2000. Since meteorological elements other than precipitation are values observed at the dam management office, mesh values need to be created from data observed at the dam management office in order to input them into the snowmelt runoff model.

  Regarding the temperature, assuming a temperature decrease rate (0.65 ° C./100 m), the temperature of the dam management station was corrected with the altitude, and a mesh value was created. The elevation of each mesh used the Ishikari River basin landscape information (edited and published by Hokkaido River Disaster Prevention Research Center (1998)). As the amount of solar radiation, the value obtained by correcting the observation value of the dam management office by the inclination of the mesh was used. As for other meteorological elements, the observation values were treated as mesh values because the area of the dam basin was small. The snow depth distribution in the basin was estimated based on a report by Nakatsugawa et al. (Non-Patent Document 12).

Application to the Toyohira Gorge Basin Here, we estimated the amount of soil supply using a snowmelt model and a snow infiltration model, and investigated whether the snow infiltration model calculated an appropriate amount of soil supply even in heavy rain.

  FIG. 16 is a table showing rainfall cases (2000) during the snowmelt period in Toyohirakyo Dam. Moreover, FIG. 17 is a figure which shows the snow cover distribution (2000) on the rainy start date in the Toyohirakyo dam basin. In FIG. 17, the white area is a snow area and the gray area is a no snow area.

  As an example of rainfall, we picked up three rainy periods that occurred during the 2000 snowmelt season (Fig. 16). In all cases, the rainfall for three days exceeded 70 mm, and the scale of rainfall is the top three cases in the snowmelt season from 1996 to 2000, which was the subject of analysis.

  The distribution of snow cover in each case is that there is snow in the entire basin on April 21 (Fig. 17 (A)). On May 12, although snow melts at low altitudes, there is no snow. In a wide range, snow still remains (FIG. 17C). Note that FIG. 17B shows a snow cover state on the middle of May 2.

  18 to 20 show the amount of water supplied to the snow surface (rainfall amount and snowmelt amount), and the calculated soil supply water amount and dam inflow amount for these cases. 18 is a diagram showing a comparison of the amount of water supplied to the snow surface and soil in the case of April 21, 2000, and FIG. 19 is the amount of water supplied to the snow surface and soil in the case of May 2, 2000. FIG. 20 is a diagram showing a comparison of the amount of water supplied to the snow surface and soil in the case of May 12, 2000. In each figure, the left side shows the amount of rain and the amount of snow melt (the amount of water before passing through the snow infiltration layer) supplied to the snow surface (FIGS. 18A, 19A, and 20A), The right side shows the amount of water supplied to the soil (the amount of water passing through the snow infiltration layer: a value calculated by the snow infiltration model) (FIGS. 18B, 19B, and 20B). However, the amount of soil supply water shown in FIGS. 18 to 20 is a calculated value based on a conventional snow infiltration model (see FIG. 5).

  18 to 20, the following features are grasped.

  1) The fluctuation of dam inflow corresponds well to the rainfall supplied to the snow surface with a time delay. In particular, in the second flood (FIG. 19), a plurality of small peaks in the dam inflow hydrograph are sensitive to the hyetograph peaks.

  2) The response of the dam inflow to the soil supply water flowing down the snow layer is not good. Specifically, in the first discharge (FIG. 18), the amount of soil supply water shows the same fluctuation as the amount of dam inflow, and in the second discharge (FIG. 19), the multiple peaks seen in the dam inflow hydrograph indicate the soil supply. Not seen in the amount of water.

  Based on this consideration, the snow infiltration model acts as a kind of filter, and it is considered that the snow layer is over-smoothed when the snow layer flows down against heavy rain. That is, for heavy rain, the conventional snow penetration model does not output an appropriate value. However, as shown in FIG. 19, focusing on the very good correspondence between rainfall and dam inflow, it is considered that reproducibility is likely to be improved by inputting rainfall directly into the runoff model. Therefore, an attempt was made to reproduce the snowmelt hydrograph by a correction method (a method in which rainfall is directly input to the runoff model) employed by the soil water supply amount prediction apparatus 100 according to the present embodiment.

  Here, the differences between the above-described conventional method and the correction method are summarized as follows. In the conventional method, the sum of the amount of rain and the amount of snow melt applied to the snow surface is considered to flow down the snow layer, and the penetration of the snow layer is considered for both (FIG. 5). On the other hand, in the modified method, rainfall and snowmelt water flow are handled separately. For snowmelt water, infiltration is considered in the same way as in the conventional method, and for rainwater, it reaches the soil directly without considering infiltration (Fig. 4). . That is, in the modified method, the snowmelt water passes through the snow layer, while the rainfall reaches the soil directly regardless of the amount. As a result, the snowmelt water during no rain flows down the snow layer, while heavy rain reaches the soil directly.

  In the following, the superiority and inferiority of the conventional method and the modified method will be verified by the reproducibility of the outflow (hydrograph).

  Here, the amount of soil water supply was estimated using the conventional method and the modified method, and the dam inflow calculated from each value was compared. This evaluated the superiority or inferiority of the conventional method and the modified method. The model constant used for the reproduction calculation is a value optimally identified by the case from May 12 to 14, 2000 when the maximum water discharge was observed.

A reproduction result obtained by optimizing the maximum water discharge example is shown in FIG. FIG. 21 is a diagram showing a comparison of reproducibility between the conventional method and the modified method (Toyohirakyo Dam, 2000). In FIG. 21, the gray portion indicates the measured dam inflow, the solid line indicates the reproduction result by the correction method, and the circle indicates the reproduction result by the conventional method. As shown in FIG. 21, the same dam inflow hydrograph was obtained using either method. The model constant and the accuracy evaluation index at this time are shown in the optimization result column of FIG. FIG. 22 is a table showing comparison of model constants and reproduction accuracy (Toyohirakyo Dam). As shown in FIG. 22, although there is a difference between the model constants (c ₁ , c ₂ ) obtained by the correction method and the conventional method, there is almost no difference with respect to the value of the accuracy evaluation index.

  The accuracy evaluation index shown in FIG. 22 is calculated by the following formulas (18) to (20).

Here, q _oi is the observed flow rate (m ³ / s), q _ci is the calculated flow rate (m ³ / s), and N is the number of data. Nash-Sutcliffe index E _i indicates that the closer the value is to 1, the higher the degree of matching between the observed value and the calculated value. Generally, if the value exceeds about 0.8, it is judged that a good reproduction result has been obtained. (Non-patent Document 13).

  From the above results, there is no difference between superiority and inferiority between the correction method and the conventional method with respect to the reproduction accuracy of one large-scale water discharge. The reason is considered that the difference in rainfall was reflected in the model constant during the optimization process.

The model constants (c ₁ , c ₂ ) give the storage coefficients (k ₁₁ , k ₁₂ ) of the first-stage tank, and determine the response of the storage height s ₁ (see formulas (14) and (15)). Changes in these model constants change the storage height s ₁ and also cause the following changes to the calculated hydrograph.

1) As the value of c ₁ decreases, the peak of the calculated hydrograph increases and, at the same time, the peak appears earlier.

2) c ₂ is the appearance of the peak of the computed hydrograph with decreasing values accelerated, resulting in the flooding ends earlier.

Considering based on the characteristics of such model constants, it is assumed that the difference between the optimum model constants obtained by the conventional method and the correction method is caused as follows. That is, paying attention to the amount of soil supply water and the peak value of the hydrograph, the value of the peak of soil supply water amount calculated by the correction method is larger than the value by the conventional method. For this reason, in order to adapt the peak of the hydrograph calculated by the correction method to the actual measurement value, a value smaller than c _{1 of} the conventional method is taken. Peak appearance of calculation hydrograph is delayed along with this, the optimization is performed so as to reduce the value of c ₂ in order to eliminate this delay. As a result, it is considered that c ₁ of the correction method is larger than the value of the conventional method, and c ₂ is a small value.

  Next, using this optimal model constant, the hydrograph of the entire snow melting period is reproduced, and the accuracy of the correction method and the conventional method is compared.

  FIG. 23 is a diagram showing a reproduction result (Toyohira Gorge Dam, 2000) by the correction method and the conventional method, and is a result of reproducing a dam inflow hydrograph for the entire snow melting period of 2000 using the optimum model constant. In FIG. 23, the gray portion indicates the measured dam inflow, the solid line indicates the reproduction result by the correction method, and the circle indicates the reproduction result by the conventional method.

According to FIG. 23, it can be seen that the calculation result by the correction method has improved relevance in the first and second water discharges while maintaining the reproducibility of the same level as the conventional method at the maximum water discharge. Further, when comparing the accuracy evaluation indexes shown in the reproduction result column of FIG. 22, in the correction method, the relative error (J _RE ) decreases from 24% of the conventional method to 19%, and the Nash-Sutcliffe index (E _i ) Increases from 0.87 to 0.95. As a result of the above analysis, it was confirmed that the reproducibility of the whole snow melting period was improved by inputting the rainfall directly into the runoff model.

  One of the reasons why the reproducibility of the correction method is better than that of the conventional method is considered that the flow method of heavy rain was changed in the correction method. It has been pointed out that the flow of water in snow is roughly divided into two types (Non-Patent Document 14). One is a film flow in which water forms a film and covers ice particles, and water flows down the surface. The other is a water channel flow in which a large amount of water fills the gaps in the snow and flows down in the form of columns. is there. There is a difference in the flow velocity between the two, and generally the flow velocity in the water channel is larger than that in the film flow. The snow infiltration model is a model for snowmelt water flow, and the conventional method can be said to be a method assuming film flow. On the other hand, it is considered that the correction method takes into account the large flow rate (water channel flow) different from the film flow by inputting rainfall directly into the runoff model. Although the correction method is simple, it accurately reflects the difference in the flow form, and this is considered to be the cause of improving the reproducibility.

  Next, using the model constants (correction method) obtained above, a hydrograph during the snowmelt period from 1996 to 1999 was reproduced. FIG. 24 is a diagram showing the reproduction result (Toyohirakyo Dam, 1996-1999). Fig. 24 (A) is the reproduction result of the dam inflow in 1996, Fig. 24 (B) is the reproduction result of the dam inflow in 1997, and Fig. 24 (C) is the dam inflow in 1998. FIG. 24D shows a reproduction result of the dam inflow amount in 1999. FIG. In FIG. 24, the gray portion represents the measured dam inflow, the solid line represents the calculated dam inflow, and the black portion represents the soil supply water amount.

As shown in FIG. 24, in any year, the short-term water discharge is accurately reproduced, and at the same time, the long-term trend of the entire snowmelt period is well reproduced. The maximum error from 1996 to 1999 was 38.06 m ³ / s and the RMSE was 7.91 m ³ / s.

  As shown in the analysis example of FIG. 24, it can be said that the good reproduction hydrograph was obtained over a plurality of years with different weather conditions, confirming the validity of the snowmelt runoff model of the present invention. In addition, it has been proved that model constants representing watersheds can be obtained even if model constants are identified in only one case with a large flood. In other words, it was found that if the optimal runoff model constant can be obtained with one flood at a relatively large scale during the snowmelt period, the hydrograph for the entire period of snowmelt can be reproduced with sufficient practical accuracy over multiple years.

  The reason why the dam inflow was able to be reproduced for a long time with one model constant is that the snow cover the soil. During the snow melting season, a small amount of snow melting water is supplied to the soil every day, while snow accumulation suppresses evaporation of moisture from the ground surface. It is considered that the hydrograph inflow into the dam could be reproduced well even if the model constants were fixed, as a result of the effects of snow cover to suppress changes in the wet and dry conditions of the soil.

Above the application to Jozankei Dam, we verified the effectiveness of the modified snow infiltration model and the simple method to determine the runoff model constant. Here, in order to confirm the versatility of the modified snowmelt runoff model, the analysis target was moved to the Jozankei Dam basin and the applicability of the model was examined.

  Here, first, the runoff model constant was optimized based on the case of April 21 to 24, 2000, where the flood volume was large, and the entire snow melting period of the same year was reproduced using the optimum value. FIG. 25 shows the result of this optimization. FIG. 26 shows the result of reproducing the 2000 hydrograph using the optimal model constant (correction method). In FIG. 25B, the gray portion represents the measured dam inflow, and the white circle represents the calculated dam inflow. Moreover, in FIG. 25, the soil supply water amount estimated by the correction method and the conventional method was compared and shown. The vertical axis of the amount of water supplied to the soil is positive in the correction method and positive in the conventional method. In FIG. 26, the gray portion represents the measured dam inflow, the solid line represents the calculated dam inflow, and the black portion represents the soil supply water amount. FIG. 27 is a table showing a comparison of model constants and reproduction accuracy at this time.

  As shown in FIG. 25, in the correction method, a model constant having a high degree of fitness is obtained, and as shown in FIGS. 26 and 27, the dam inflow hydrograph of the entire snow melting period can be reproduced well. On the other hand, in the conventional method, the solution vibrates and the optimum value is not obtained. The reason why the optimum value could not be obtained is thought to be that the rainfall waveform was excessively smoothed by the rain flowing down the filter of the snow layer (FIG. 25). Similar to the results of the application to the Toyohira Gorge basin described above, it was confirmed that the modified method is appropriate for handling rainfall.

  To summarize the above results, it can be said that it is more reasonable to think that if heavy rain occurs during the snow cover period, the water on the snow surface will reach the soil without any time delay and will flow into the river.

Finally, the result of reproducing the dam inflow hydrograph from 1996 to 1999 is shown in FIG. In Jozankei Dam, the calculated dam inflow well reproduced the long-term runoff situation of the whole snowmelt season, indicating the validity of the snowmelt runoff model in the present invention. The maximum error during this period is 38.25 m ³ / s and the RMSE is 5.96 m ³ / s. Like the analysis results at Toyohira Gorge Dam, It was verified that multi-year dam inflow hydrographs can be reproduced with sufficient practical accuracy.

  As mentioned above, in this verification, the reproducibility of the snowmelt runoff model considering snow storage was verified using the dam basin (Toyohirakyo Dam and Jozankei Dam) in the cold snowy region as the analysis target point. As a result, the following findings were obtained as a summary.

  1) According to past research results, it is pointed out that the flow form of snowmelt water in the snow layer is classified into film flow and water flow. As a result of this analysis, if concentrated heavy rain occurs during the snowmelt season, it is reasonable to consider it as a channel flow, and it is better to input the rainfall directly into the runoff model to improve the reproducibility of the dam inflow hydrograph. Confirmed to do.

  2) In order to easily and objectively search the runoff model constant, the model constant is identified in only one case with a large flooding scale using a mathematical optimization method (Newton method), and the dam inflow during the entire period of snow melting. Was reproduced. As a result, it was found that even with a model constant obtained in only one example, a sufficient reproduction accuracy was obtained in practical use.

  3) Based on the results of 1) and 2) above, a dam inflow hydrograph during the snowmelt period over several years was estimated. As a result, in the two catchment areas analyzed, the short-term water discharge and the long-term runoff situation of the whole snowmelt period were successfully reproduced.

  In this verification, it was proved that sufficient reproducibility could be obtained even if the runoff model constant of the modified snowmelt runoff model was fixed to one. This indicates that it is not necessary to change the model constant for each flood, which is a great advantage in actual flood forecasting.

  In this embodiment, the two-stage tank type storage function model is used as the outflow model, but the present invention is not limited to this. As an outflow model, in addition to a two-stage tank type storage function model, for example, a storage function method using effective rainfall (a so-called generalized storage function model) or a storage function method including a loss term (a so-called one-stage tank type storage function) Model) and the like are known (for example, Non-Patent Document 8). Therefore, for example, as shown in FIG. 29, these runoff models are used in combination, the runoff model having the smallest prediction error value is automatically selected as the optimum model, and the predicted dam inflow amount is calculated using the selected optimum model. You may make it calculate.

  The snowmelt runoff prediction system according to the present invention is useful as a snowmelt runoff prediction system that can improve the prediction accuracy of snowmelt runoff.

The block diagram which shows the structure of the snowmelt runoff prediction system which concerns on one embodiment of this invention. The block diagram which shows the structure of each part of the snowmelt runoff prediction system of FIG. Schematic diagram showing the type and flow of data used in the snowmelt runoff prediction system of FIG. The figure which shows the structure of the snow-cover infiltration model in this Embodiment Diagram showing the structure of a conventional snow penetration model The figure which shows the relationship between the storage coefficient k _s and snow depth H _s The figure which shows the structure of the outflow model in this Embodiment The main flowchart which shows the whole operation | movement of the snowmelt runoff prediction system of FIG. The flowchart which shows the content of the snow melting calculation process (S1400) of FIG. The flowchart which shows the content of the snowmelt amount calculation process (S1410) of FIG. The flowchart which shows the content of the soil water supply amount calculation process (S1420) of FIG. The flowchart which shows the content of the outflow calculation process (S1500) of FIG. Schematic showing the configuration of the snowmelt runoff model used for verification Diagram showing analysis target basin Table showing meteorological observation items at Toyohirakyo Dam and Jozankei Dam Table showing examples of rainfall during the snowmelt period (2000) at Toyohirakyo Dam Figure showing the snow distribution (2000) on the rainy start date in the Toyohira Gorge basin The figure which shows an example (April 21, 2000) of the comparison of the amount of water supplied to the snow surface and soil The figure which shows the other example (May 2, 2000) of the comparison of the amount of water supplied to the snow surface and the soil The figure which shows the further another example (May 12, 2000) of the comparison of the amount of water supplied to the snow surface and soil Figure showing the comparison (2000) of the reproducibility of the conventional method and the modified method at Toyohirakyo Dam Table showing comparison of model constants and reproduction accuracy at Toyohirakyo Dam The figure which shows the reproduction result (2000) by the correction method and the conventional method in Toyohirakyo Dam Figure showing the results (1996-1999) of the dam inflow during the multi-year snowmelt period at Toyohirakyo Dam The figure which shows the reproduction comparison (2000) by the optimal model constant in Jozankei Dam The figure which shows the reproduction result (2000) by the optimal model constant (correction method) in Jozankei Dam Table showing comparison of model constants and reproduction accuracy at Jozankei Dam Figure showing the reproduction results (1996-1999) of the dam inflow during the multi-year snowmelt period at Jozankei Dam The figure which shows the example of a change of this Embodiment

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Snow melting runoff prediction system 200 Data collection device 210 Input data processing part 300 Data operation part 310 Snow calculation part 312 Snow equivalent water amount estimation model 314 Snow model part 320 Snow melting calculation part 322 Snow melting model part 324 Snow accumulation infiltration model part 326 Watershed average part 330 Outflow Calculation Unit 332 Outflow Model Unit 334 Outflow Model Constant Update Unit 340 Storage Unit 400 Output Device

Claims (10)

A snowmelt runoff prediction system for predicting runoff due to snowmelt,
A data input unit for inputting meteorological data including actual meteorological data before the present state of the target basin including rainfall data and predicted meteorological data thereafter;
A soil supply water amount calculation unit that calculates the amount of water supplied to the soil by flowing down the snow layer from the snow surface based on the input weather data,
An outflow amount calculation unit that calculates a predicted outflow amount using a predetermined outflow model based on the calculated soil supply water amount,
The soil supply water amount calculation unit
The amount of water supplied to the soil is calculated based on the given amount of snow melting and the amount of rain input, assuming that the snow melting water penetrates the snow layer and the rain reaches the soil directly without penetrating the snow layer.
Snowmelt runoff prediction system. The soil supply water amount calculation unit
It has a snow infiltration model part including a snow infiltration model in which the storage effect of the snow layer is modeled by the storage function method,
Applying the snow infiltration model to snowmelt water, and to reach the soil directly without applying the snow infiltration model for rainfall, based on the amount of snowmelt given and the input rainfall, Calculate the amount of water supplied to the soil,
The snowmelt runoff prediction system according to claim 1. The soil supply water amount calculation unit
It has a snow infiltration model part including a snow infiltration model in which the storage effect of the snow layer is modeled by the storage function method,
For the given amount of snowmelt, the snowmelt infiltration model is applied to calculate the amount of snowmelt that penetrates the snow layer and reaches the soil, and for the input rainfall, the snow accumulation infiltration model is applied. Without calculating, the amount of water supplied to the soil is calculated by adding the calculated amount of snow melting to the soil and the amount of rain that has been input.
The snowmelt runoff prediction system according to claim 1. The soil supply water amount calculation unit
A snow melting model unit that calculates the amount of snow melting on the snow surface using a predetermined snow melting model based on the input weather data and the given terrain data;
The snow penetration model part is
Input the amount of snowmelt calculated by the snowmelt model unit,
The snowmelt runoff prediction system according to claim 2 or 3.   5. The snowmelt runoff prediction system according to claim 4, wherein the snowmelt model is based on a principle of a heat balance method and is composed of a heat balance equation that considers the influence of vegetation. The data input unit includes:
A means for processing the input weather data into a predetermined mesh unit,
Providing processed mesh unit weather data to the snowmelt model part,
The snow melting model part and the snow accumulation infiltration model part are:
Calculate the amount of snowmelt and the amount of soil supply water in a predetermined mesh unit,
The soil supply water amount calculation unit
A basin averaging unit that calculates the basin average soil supply water amount by averaging the calculated mesh unit soil supply water amount in the target basin, and
The outflow amount calculation unit
Based on the calculated basin average soil water supply, calculate the predicted runoff.
The snowmelt runoff prediction system according to claim 4 or 5.   The snowmelt runoff prediction system according to any one of claims 1 to 6, wherein the runoff model is a two-stage tank type storage function model including an underground runoff component. The outflow amount calculation unit
The runoff model includes a generalized storage function model using effective rainfall, a one-stage tank type storage function model including a loss term, and a two-stage tank type storage function model including an underground runoff component. These three storage functions Use the most appropriate model to calculate the predicted runoff amount.
The snowmelt runoff prediction system according to any one of claims 1 to 6. The outflow amount calculation unit
Applying a Kalman filter to the outflow model to calculate the predicted outflow amount,
The snowmelt runoff prediction system according to any one of claims 1 to 8. A snowmelt runoff prediction method for predicting runoff due to snowmelt,
A data input step for inputting meteorological data including actual meteorological data before the present state of the target basin including rainfall data and predicted meteorological data thereafter;
A soil supply water amount calculating step for calculating the amount of water supplied to the soil by flowing down the snow layer from the snow surface based on the input weather data;
A runoff amount calculating step for calculating a predicted runoff amount using a predetermined runoff model based on the calculated amount of lodge water supply,
The soil supply water amount calculating step includes:
The amount of water supplied to the soil is calculated based on the given amount of snowmelt and the amount of rain input, assuming that the snowmelt water penetrates the snow layer and the rain reaches the soil directly without penetrating the snow layer.
Snowmelt runoff prediction method.

Images